2023-03-07 10:14:36,176 INFO [train2.py:879] (1/4) Training started 2023-03-07 10:14:36,177 INFO [train2.py:880] (1/4) {'frame_shift_ms': 10.0, 'allowed_excess_duration_ratio': 0.1, 'best_train_loss': inf, 'best_valid_loss': inf, 'best_train_epoch': -1, 'best_valid_epoch': -1, 'batch_idx_train': 0, 'log_interval': 50, 'reset_interval': 200, 'valid_interval': 3000, 'feature_dim': 80, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.23.3', 'k2-build-type': 'Debug', 'k2-with-cuda': True, 'k2-git-sha1': '3b81ac9686aee539d447bb2085b2cdfc131c7c91', 'k2-git-date': 'Thu Jan 26 20:40:25 2023', 'lhotse-version': '1.9.0.dev+git.97bf4b0.dirty', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'surt', 'icefall-git-sha1': 'e9931b7-dirty', 'icefall-git-date': 'Fri Mar 3 16:27:17 2023', 'icefall-path': '/exp/draj/mini_scale_2022/icefall', 'k2-path': '/exp/draj/mini_scale_2022/k2/k2/python/k2/__init__.py', 'lhotse-path': '/exp/draj/mini_scale_2022/lhotse/lhotse/__init__.py', 'hostname': 'r8n04', 'IP address': '10.1.8.4'}, 'beam_size': 10, 'reduction': 'sum', 'use_double_scores': True, 'world_size': 4, 'master_port': 12368, 'tensorboard': True, 'num_epochs': 30, 'start_epoch': 1, 'start_batch': 0, 'exp_dir': PosixPath('zipformer_ctc_att/exp/v0'), 'lang_dir': PosixPath('data/lang_bpe_500'), 'base_lr': 0.05, 'lr_batches': 5000, 'lr_epochs': 3.5, 'att_rate': 0.8, 'num_decoder_layers': 6, 'seed': 42, 'print_diagnostics': False, 'inf_check': False, 'save_every_n': 2000, 'keep_last_k': 10, 'average_period': 200, 'use_fp16': True, 'num_encoder_layers': '2,4,3,2,4', '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', 'full_libri': True, 'manifest_dir': PosixPath('data/manifests'), 'max_duration': 1000, 'bucketing_sampler': True, 'num_buckets': 30, 'concatenate_cuts': False, 'duration_factor': 1.0, 'gap': 1.0, 'on_the_fly_feats': False, 'shuffle': True, 'drop_last': True, 'return_cuts': True, 'num_workers': 2, 'enable_spec_aug': True, 'spec_aug_time_warp_factor': 80, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures'} 2023-03-07 10:14:36,468 INFO [lexicon.py:168] (1/4) Loading pre-compiled data/lang_bpe_500/Linv.pt 2023-03-07 10:14:37,047 INFO [train2.py:902] (1/4) About to create model 2023-03-07 10:14:37,543 INFO [zipformer.py:178] (1/4) At encoder stack 4, which has downsampling_factor=2, we will combine the outputs of layers 1 and 3, with downsampling_factors=2 and 8. 2023-03-07 10:14:37,604 INFO [train2.py:906] (1/4) Number of model parameters: 86083707 2023-03-07 10:14:42,054 INFO [train2.py:921] (1/4) Using DDP 2023-03-07 10:14:42,773 INFO [asr_datamodule.py:420] (1/4) About to get the shuffled train-clean-100, train-clean-360 and train-other-500 cuts 2023-03-07 10:14:42,878 INFO [asr_datamodule.py:224] (1/4) Enable MUSAN 2023-03-07 10:14:42,878 INFO [asr_datamodule.py:225] (1/4) About to get Musan cuts 2023-03-07 10:14:44,224 INFO [asr_datamodule.py:249] (1/4) Enable SpecAugment 2023-03-07 10:14:44,225 INFO [asr_datamodule.py:250] (1/4) Time warp factor: 80 2023-03-07 10:14:44,225 INFO [asr_datamodule.py:260] (1/4) Num frame mask: 10 2023-03-07 10:14:44,225 INFO [asr_datamodule.py:273] (1/4) About to create train dataset 2023-03-07 10:14:44,225 INFO [asr_datamodule.py:300] (1/4) Using DynamicBucketingSampler. 2023-03-07 10:14:46,722 INFO [asr_datamodule.py:316] (1/4) About to create train dataloader 2023-03-07 10:14:46,723 INFO [asr_datamodule.py:440] (1/4) About to get dev-clean cuts 2023-03-07 10:14:46,724 INFO [asr_datamodule.py:447] (1/4) About to get dev-other cuts 2023-03-07 10:14:46,724 INFO [asr_datamodule.py:347] (1/4) About to create dev dataset 2023-03-07 10:14:47,008 INFO [asr_datamodule.py:364] (1/4) About to create dev dataloader 2023-03-07 10:15:00,122 INFO [train2.py:809] (1/4) Epoch 1, batch 0, loss[ctc_loss=5.375, att_loss=1.381, loss=2.18, over 16861.00 frames. utt_duration=1378 frames, utt_pad_proportion=0.007875, over 49.00 utterances.], tot_loss[ctc_loss=5.375, att_loss=1.381, loss=2.18, over 16861.00 frames. utt_duration=1378 frames, utt_pad_proportion=0.007875, over 49.00 utterances.], batch size: 49, lr: 2.50e-02, grad_scale: 2.0 2023-03-07 10:15:00,122 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-07 10:15:12,517 INFO [train2.py:843] (1/4) Epoch 1, validation: ctc_loss=5.278, att_loss=1.419, loss=2.191, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-07 10:15:12,518 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 13325MB 2023-03-07 10:15:18,067 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5.0, num_to_drop=2, layers_to_drop={1, 3} 2023-03-07 10:15:43,086 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=23.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 10:16:22,274 INFO [train2.py:809] (1/4) Epoch 1, batch 50, loss[ctc_loss=1.119, att_loss=0.943, loss=0.9782, over 15651.00 frames. utt_duration=1693 frames, utt_pad_proportion=0.008624, over 37.00 utterances.], tot_loss[ctc_loss=2.298, att_loss=1.128, loss=1.362, over 735284.85 frames. utt_duration=1299 frames, utt_pad_proportion=0.0422, over 2267.63 utterances.], batch size: 37, lr: 2.75e-02, grad_scale: 2.0 2023-03-07 10:16:35,851 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=59.64 vs. limit=5.0 2023-03-07 10:17:06,717 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 10:17:21,431 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.6013, 5.5410, 5.5649, 5.9074, 5.5097, 5.4018, 5.2589, 5.9133], device='cuda:1'), covar=tensor([0.0320, 0.0046, 0.0262, 0.0157, 0.0172, 0.0151, 0.0046, 0.0100], device='cuda:1'), in_proj_covar=tensor([0.0009, 0.0008, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0008], device='cuda:1'), out_proj_covar=tensor([8.6161e-06, 8.5751e-06, 8.5283e-06, 8.7755e-06, 8.5823e-06, 8.5795e-06, 8.5126e-06, 8.6684e-06], device='cuda:1') 2023-03-07 10:17:31,214 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 4.328e+01 1.076e+02 2.257e+02 4.349e+02 2.620e+03, threshold=4.514e+02, percent-clipped=0.0 2023-03-07 10:17:31,259 INFO [train2.py:809] (1/4) Epoch 1, batch 100, loss[ctc_loss=1.168, att_loss=0.981, loss=1.018, over 16543.00 frames. utt_duration=1472 frames, utt_pad_proportion=0.005435, over 45.00 utterances.], tot_loss[ctc_loss=1.669, att_loss=1.046, loss=1.171, over 1295403.59 frames. utt_duration=1266 frames, utt_pad_proportion=0.05009, over 4098.03 utterances.], batch size: 45, lr: 3.00e-02, grad_scale: 2.0 2023-03-07 10:18:30,929 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.8044, 3.8595, 3.8510, 3.8354, 3.8585, 3.8246, 3.8213, 3.8574], device='cuda:1'), covar=tensor([0.0011, 0.0030, 0.0046, 0.0017, 0.0030, 0.0023, 0.0013, 0.0028], device='cuda:1'), in_proj_covar=tensor([0.0008, 0.0008, 0.0008, 0.0008, 0.0008, 0.0009, 0.0008, 0.0008], device='cuda:1'), out_proj_covar=tensor([8.1901e-06, 8.3179e-06, 8.2548e-06, 8.2292e-06, 8.4302e-06, 8.3979e-06, 8.4542e-06, 8.3302e-06], device='cuda:1') 2023-03-07 10:18:33,458 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=144.0, num_to_drop=2, layers_to_drop={0, 3} 2023-03-07 10:18:42,514 INFO [train2.py:809] (1/4) Epoch 1, batch 150, loss[ctc_loss=1.143, att_loss=0.9329, loss=0.9749, over 16544.00 frames. utt_duration=1472 frames, utt_pad_proportion=0.006242, over 45.00 utterances.], tot_loss[ctc_loss=1.469, att_loss=1.022, loss=1.112, over 1740164.52 frames. utt_duration=1230 frames, utt_pad_proportion=0.05695, over 5665.81 utterances.], batch size: 45, lr: 3.25e-02, grad_scale: 2.0 2023-03-07 10:19:05,474 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=2.92 vs. limit=2.0 2023-03-07 10:19:33,937 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=18.69 vs. limit=5.0 2023-03-07 10:19:48,620 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 4.064e+01 7.732e+01 1.287e+02 1.925e+02 9.711e+02, threshold=2.575e+02, percent-clipped=2.0 2023-03-07 10:19:48,667 INFO [train2.py:809] (1/4) Epoch 1, batch 200, loss[ctc_loss=1.112, att_loss=0.9106, loss=0.951, over 16112.00 frames. utt_duration=1536 frames, utt_pad_proportion=0.007187, over 42.00 utterances.], tot_loss[ctc_loss=1.365, att_loss=1.001, loss=1.074, over 2078279.75 frames. utt_duration=1250 frames, utt_pad_proportion=0.05294, over 6660.43 utterances.], batch size: 42, lr: 3.50e-02, grad_scale: 2.0 2023-03-07 10:19:52,677 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2731, 4.2420, 4.2644, 4.2590, 4.2177, 4.3307, 4.2743, 4.2156], device='cuda:1'), covar=tensor([0.0053, 0.0041, 0.0052, 0.0049, 0.0051, 0.0018, 0.0034, 0.0025], device='cuda:1'), in_proj_covar=tensor([0.0009, 0.0008, 0.0009, 0.0008, 0.0009, 0.0008, 0.0008, 0.0008], device='cuda:1'), out_proj_covar=tensor([8.7006e-06, 8.6402e-06, 8.8433e-06, 8.4604e-06, 8.9305e-06, 8.7630e-06, 8.7670e-06, 8.7414e-06], device='cuda:1') 2023-03-07 10:19:55,731 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.90 vs. limit=2.0 2023-03-07 10:20:27,680 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.7768, 5.7693, 5.7767, 5.7768, 5.7789, 5.6975, 5.7723, 5.7768], device='cuda:1'), covar=tensor([0.0026, 0.0045, 0.0013, 0.0049, 0.0025, 0.0221, 0.0030, 0.0018], device='cuda:1'), in_proj_covar=tensor([0.0008, 0.0008, 0.0008, 0.0008, 0.0008, 0.0008, 0.0008, 0.0008], device='cuda:1'), out_proj_covar=tensor([8.5481e-06, 8.3818e-06, 8.4212e-06, 8.4098e-06, 8.4488e-06, 8.4447e-06, 8.3735e-06, 8.5300e-06], device='cuda:1') 2023-03-07 10:20:54,632 INFO [train2.py:809] (1/4) Epoch 1, batch 250, loss[ctc_loss=1.23, att_loss=1.008, loss=1.053, over 17326.00 frames. utt_duration=1006 frames, utt_pad_proportion=0.04914, over 69.00 utterances.], tot_loss[ctc_loss=1.306, att_loss=0.9882, loss=1.052, over 2348563.40 frames. utt_duration=1247 frames, utt_pad_proportion=0.0525, over 7545.68 utterances.], batch size: 69, lr: 3.75e-02, grad_scale: 2.0 2023-03-07 10:21:54,688 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 10:21:59,451 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=300.0, num_to_drop=2, layers_to_drop={1, 2} 2023-03-07 10:22:00,458 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 4.252e+01 6.827e+01 1.043e+02 1.996e+02 1.331e+03, threshold=2.085e+02, percent-clipped=17.0 2023-03-07 10:22:00,502 INFO [train2.py:809] (1/4) Epoch 1, batch 300, loss[ctc_loss=1.291, att_loss=1.043, loss=1.092, over 17306.00 frames. utt_duration=1175 frames, utt_pad_proportion=0.02263, over 59.00 utterances.], tot_loss[ctc_loss=1.27, att_loss=0.9794, loss=1.038, over 2550100.79 frames. utt_duration=1222 frames, utt_pad_proportion=0.06179, over 8359.54 utterances.], batch size: 59, lr: 4.00e-02, grad_scale: 2.0 2023-03-07 10:22:47,743 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=4.77 vs. limit=2.0 2023-03-07 10:23:06,531 INFO [train2.py:809] (1/4) Epoch 1, batch 350, loss[ctc_loss=1.187, att_loss=0.96, loss=1.005, over 17296.00 frames. utt_duration=1259 frames, utt_pad_proportion=0.01245, over 55.00 utterances.], tot_loss[ctc_loss=1.245, att_loss=0.9685, loss=1.024, over 2724596.34 frames. utt_duration=1243 frames, utt_pad_proportion=0.05186, over 8779.48 utterances.], batch size: 55, lr: 4.25e-02, grad_scale: 2.0 2023-03-07 10:23:13,078 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2262, 3.2380, 3.4478, 5.3900, 5.2036, 4.1680, 4.3545, 3.7177], device='cuda:1'), covar=tensor([0.0155, 0.2426, 0.1849, 0.0072, 0.0120, 0.0313, 0.0300, 0.0196], device='cuda:1'), in_proj_covar=tensor([0.0008, 0.0008, 0.0009, 0.0008, 0.0008, 0.0008, 0.0008, 0.0008], device='cuda:1'), out_proj_covar=tensor([8.7608e-06, 8.6313e-06, 9.4336e-06, 8.6777e-06, 8.9284e-06, 8.5322e-06, 8.6620e-06, 8.5265e-06], device='cuda:1') 2023-03-07 10:23:14,298 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=357.0, num_to_drop=2, layers_to_drop={1, 2} 2023-03-07 10:23:54,067 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=387.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 10:24:12,774 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 4.983e+01 8.126e+01 1.923e+02 2.981e+02 6.249e+02, threshold=3.846e+02, percent-clipped=46.0 2023-03-07 10:24:12,820 INFO [train2.py:809] (1/4) Epoch 1, batch 400, loss[ctc_loss=1.175, att_loss=0.9397, loss=0.9867, over 16459.00 frames. utt_duration=1433 frames, utt_pad_proportion=0.00702, over 46.00 utterances.], tot_loss[ctc_loss=1.217, att_loss=0.9538, loss=1.006, over 2842715.58 frames. utt_duration=1255 frames, utt_pad_proportion=0.05169, over 9074.32 utterances.], batch size: 46, lr: 4.50e-02, grad_scale: 4.0 2023-03-07 10:25:02,005 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=439.0, num_to_drop=2, layers_to_drop={0, 1} 2023-03-07 10:25:13,989 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=448.0, num_to_drop=2, layers_to_drop={0, 1} 2023-03-07 10:25:17,465 INFO [train2.py:809] (1/4) Epoch 1, batch 450, loss[ctc_loss=1.058, att_loss=0.8438, loss=0.8866, over 15868.00 frames. utt_duration=1629 frames, utt_pad_proportion=0.0103, over 39.00 utterances.], tot_loss[ctc_loss=1.198, att_loss=0.9435, loss=0.9944, over 2942842.94 frames. utt_duration=1240 frames, utt_pad_proportion=0.05507, over 9505.84 utterances.], batch size: 39, lr: 4.75e-02, grad_scale: 4.0 2023-03-07 10:25:22,230 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=5.73 vs. limit=2.0 2023-03-07 10:26:22,551 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 5.753e+01 1.641e+02 2.817e+02 4.953e+02 1.195e+03, threshold=5.634e+02, percent-clipped=34.0 2023-03-07 10:26:22,598 INFO [train2.py:809] (1/4) Epoch 1, batch 500, loss[ctc_loss=1.109, att_loss=0.8864, loss=0.9308, over 16133.00 frames. utt_duration=1538 frames, utt_pad_proportion=0.005879, over 42.00 utterances.], tot_loss[ctc_loss=1.175, att_loss=0.9287, loss=0.9779, over 3017874.17 frames. utt_duration=1255 frames, utt_pad_proportion=0.05162, over 9633.24 utterances.], batch size: 42, lr: 4.99e-02, grad_scale: 4.0 2023-03-07 10:27:27,828 INFO [train2.py:809] (1/4) Epoch 1, batch 550, loss[ctc_loss=1.009, att_loss=0.8327, loss=0.868, over 16278.00 frames. utt_duration=1516 frames, utt_pad_proportion=0.007402, over 43.00 utterances.], tot_loss[ctc_loss=1.149, att_loss=0.9131, loss=0.9603, over 3075529.06 frames. utt_duration=1239 frames, utt_pad_proportion=0.05441, over 9940.12 utterances.], batch size: 43, lr: 4.98e-02, grad_scale: 4.0 2023-03-07 10:27:41,738 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=562.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 10:27:49,269 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=568.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 10:28:19,200 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=590.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 10:28:32,099 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=600.0, num_to_drop=2, layers_to_drop={0, 2} 2023-03-07 10:28:33,065 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 8.073e+01 2.172e+02 3.350e+02 5.891e+02 2.324e+03, threshold=6.699e+02, percent-clipped=26.0 2023-03-07 10:28:33,109 INFO [train2.py:809] (1/4) Epoch 1, batch 600, loss[ctc_loss=1.009, att_loss=0.8661, loss=0.8948, over 16773.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.005582, over 48.00 utterances.], tot_loss[ctc_loss=1.125, att_loss=0.9011, loss=0.9458, over 3123270.66 frames. utt_duration=1248 frames, utt_pad_proportion=0.05125, over 10018.53 utterances.], batch size: 48, lr: 4.98e-02, grad_scale: 4.0 2023-03-07 10:29:01,211 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=623.0, num_to_drop=2, layers_to_drop={0, 1} 2023-03-07 10:29:09,943 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=629.0, num_to_drop=2, layers_to_drop={1, 3} 2023-03-07 10:29:33,581 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=648.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 10:29:37,166 INFO [train2.py:809] (1/4) Epoch 1, batch 650, loss[ctc_loss=0.9096, att_loss=0.8047, loss=0.8257, over 15762.00 frames. utt_duration=1661 frames, utt_pad_proportion=0.008017, over 38.00 utterances.], tot_loss[ctc_loss=1.087, att_loss=0.8864, loss=0.9266, over 3153065.68 frames. utt_duration=1233 frames, utt_pad_proportion=0.0566, over 10237.79 utterances.], batch size: 38, lr: 4.98e-02, grad_scale: 4.0 2023-03-07 10:29:37,394 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=651.0, num_to_drop=2, layers_to_drop={1, 3} 2023-03-07 10:29:38,586 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=652.0, num_to_drop=2, layers_to_drop={0, 2} 2023-03-07 10:29:57,491 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.35 vs. limit=2.0 2023-03-07 10:30:42,277 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.243e+02 2.720e+02 3.813e+02 5.930e+02 1.074e+03, threshold=7.626e+02, percent-clipped=15.0 2023-03-07 10:30:42,322 INFO [train2.py:809] (1/4) Epoch 1, batch 700, loss[ctc_loss=0.8067, att_loss=0.7982, loss=0.7999, over 15784.00 frames. utt_duration=1663 frames, utt_pad_proportion=0.008039, over 38.00 utterances.], tot_loss[ctc_loss=1.047, att_loss=0.8766, loss=0.9106, over 3178980.87 frames. utt_duration=1242 frames, utt_pad_proportion=0.05549, over 10251.24 utterances.], batch size: 38, lr: 4.98e-02, grad_scale: 4.0 2023-03-07 10:31:32,105 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=739.0, num_to_drop=2, layers_to_drop={0, 2} 2023-03-07 10:31:37,686 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=743.0, num_to_drop=2, layers_to_drop={0, 3} 2023-03-07 10:31:47,680 INFO [train2.py:809] (1/4) Epoch 1, batch 750, loss[ctc_loss=0.7795, att_loss=0.7522, loss=0.7577, over 11491.00 frames. utt_duration=1840 frames, utt_pad_proportion=0.1897, over 25.00 utterances.], tot_loss[ctc_loss=0.9981, att_loss=0.8639, loss=0.8908, over 3191406.47 frames. utt_duration=1237 frames, utt_pad_proportion=0.05755, over 10330.42 utterances.], batch size: 25, lr: 4.97e-02, grad_scale: 4.0 2023-03-07 10:32:18,280 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.92 vs. limit=2.0 2023-03-07 10:32:34,573 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=787.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 10:32:52,870 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.221e+02 2.689e+02 3.178e+02 4.558e+02 8.298e+02, threshold=6.355e+02, percent-clipped=3.0 2023-03-07 10:32:52,913 INFO [train2.py:809] (1/4) Epoch 1, batch 800, loss[ctc_loss=0.6994, att_loss=0.7347, loss=0.7276, over 15644.00 frames. utt_duration=1693 frames, utt_pad_proportion=0.008719, over 37.00 utterances.], tot_loss[ctc_loss=0.9501, att_loss=0.8524, loss=0.872, over 3210223.62 frames. utt_duration=1233 frames, utt_pad_proportion=0.05663, over 10430.49 utterances.], batch size: 37, lr: 4.97e-02, grad_scale: 8.0 2023-03-07 10:33:01,893 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=2.22 vs. limit=2.0 2023-03-07 10:33:52,592 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=847.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 10:33:57,215 INFO [train2.py:809] (1/4) Epoch 1, batch 850, loss[ctc_loss=0.8669, att_loss=0.8289, loss=0.8365, over 17384.00 frames. utt_duration=1105 frames, utt_pad_proportion=0.03298, over 63.00 utterances.], tot_loss[ctc_loss=0.9089, att_loss=0.8409, loss=0.8545, over 3225920.12 frames. utt_duration=1241 frames, utt_pad_proportion=0.05532, over 10410.75 utterances.], batch size: 63, lr: 4.96e-02, grad_scale: 8.0 2023-03-07 10:34:06,571 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.20 vs. limit=2.0 2023-03-07 10:35:01,442 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.412e+02 3.625e+02 4.631e+02 5.738e+02 1.582e+03, threshold=9.262e+02, percent-clipped=18.0 2023-03-07 10:35:01,486 INFO [train2.py:809] (1/4) Epoch 1, batch 900, loss[ctc_loss=0.6888, att_loss=0.6815, loss=0.6829, over 15755.00 frames. utt_duration=1660 frames, utt_pad_proportion=0.01007, over 38.00 utterances.], tot_loss[ctc_loss=0.8739, att_loss=0.8225, loss=0.8327, over 3235165.42 frames. utt_duration=1219 frames, utt_pad_proportion=0.0617, over 10631.54 utterances.], batch size: 38, lr: 4.96e-02, grad_scale: 8.0 2023-03-07 10:35:10,226 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=908.0, num_to_drop=2, layers_to_drop={0, 2} 2023-03-07 10:35:23,188 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=918.0, num_to_drop=2, layers_to_drop={1, 3} 2023-03-07 10:35:31,238 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=924.0, num_to_drop=2, layers_to_drop={1, 3} 2023-03-07 10:35:58,371 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.4634, 2.5809, 2.4937, 2.5806, 2.6717, 2.7655, 2.3053, 2.5165], device='cuda:1'), covar=tensor([0.3128, 0.3155, 0.3544, 0.2587, 0.2793, 0.2593, 0.3693, 0.3321], device='cuda:1'), in_proj_covar=tensor([0.0048, 0.0042, 0.0045, 0.0044, 0.0044, 0.0045, 0.0051, 0.0049], device='cuda:1'), out_proj_covar=tensor([4.2872e-05, 3.4370e-05, 3.6589e-05, 3.7227e-05, 3.8148e-05, 3.6591e-05, 4.5280e-05, 4.2688e-05], device='cuda:1') 2023-03-07 10:35:59,550 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=946.0, num_to_drop=2, layers_to_drop={0, 2} 2023-03-07 10:36:05,678 INFO [train2.py:809] (1/4) Epoch 1, batch 950, loss[ctc_loss=0.707, att_loss=0.6805, loss=0.6858, over 16754.00 frames. utt_duration=1398 frames, utt_pad_proportion=0.007339, over 48.00 utterances.], tot_loss[ctc_loss=0.838, att_loss=0.7939, loss=0.8027, over 3238148.15 frames. utt_duration=1243 frames, utt_pad_proportion=0.05736, over 10435.33 utterances.], batch size: 48, lr: 4.96e-02, grad_scale: 8.0 2023-03-07 10:36:07,095 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=952.0, num_to_drop=2, layers_to_drop={0, 2} 2023-03-07 10:36:45,336 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9291, 5.8875, 5.5716, 6.0390, 5.7004, 5.7392, 5.8523, 5.6890], device='cuda:1'), covar=tensor([0.0391, 0.0336, 0.0555, 0.0252, 0.0611, 0.0693, 0.0511, 0.0468], device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0048, 0.0053, 0.0044, 0.0047, 0.0057, 0.0051, 0.0051], device='cuda:1'), out_proj_covar=tensor([3.8723e-05, 4.4197e-05, 5.0050e-05, 4.1873e-05, 4.5065e-05, 5.8281e-05, 4.7541e-05, 5.2354e-05], device='cuda:1') 2023-03-07 10:37:08,807 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=1000.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 10:37:09,961 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 3.067e+02 5.280e+02 6.177e+02 7.418e+02 1.446e+03, threshold=1.235e+03, percent-clipped=9.0 2023-03-07 10:37:10,006 INFO [train2.py:809] (1/4) Epoch 1, batch 1000, loss[ctc_loss=0.6384, att_loss=0.6058, loss=0.6123, over 16533.00 frames. utt_duration=1471 frames, utt_pad_proportion=0.006437, over 45.00 utterances.], tot_loss[ctc_loss=0.8054, att_loss=0.7601, loss=0.7691, over 3244364.09 frames. utt_duration=1248 frames, utt_pad_proportion=0.05662, over 10409.05 utterances.], batch size: 45, lr: 4.95e-02, grad_scale: 8.0 2023-03-07 10:38:02,140 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.95 vs. limit=2.0 2023-03-07 10:38:04,786 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1043.0, num_to_drop=2, layers_to_drop={0, 1} 2023-03-07 10:38:15,415 INFO [train2.py:809] (1/4) Epoch 1, batch 1050, loss[ctc_loss=0.5894, att_loss=0.5463, loss=0.5549, over 15515.00 frames. utt_duration=1725 frames, utt_pad_proportion=0.008035, over 36.00 utterances.], tot_loss[ctc_loss=0.7738, att_loss=0.7249, loss=0.7347, over 3260488.05 frames. utt_duration=1267 frames, utt_pad_proportion=0.04871, over 10305.72 utterances.], batch size: 36, lr: 4.95e-02, grad_scale: 8.0 2023-03-07 10:38:39,076 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.83 vs. limit=2.0 2023-03-07 10:39:07,962 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=1091.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 10:39:21,064 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 3.042e+02 5.471e+02 6.537e+02 8.714e+02 1.508e+03, threshold=1.307e+03, percent-clipped=3.0 2023-03-07 10:39:21,110 INFO [train2.py:809] (1/4) Epoch 1, batch 1100, loss[ctc_loss=0.6149, att_loss=0.5468, loss=0.5604, over 16780.00 frames. utt_duration=1400 frames, utt_pad_proportion=0.005741, over 48.00 utterances.], tot_loss[ctc_loss=0.7444, att_loss=0.6895, loss=0.7005, over 3262834.30 frames. utt_duration=1287 frames, utt_pad_proportion=0.04335, over 10148.93 utterances.], batch size: 48, lr: 4.94e-02, grad_scale: 8.0 2023-03-07 10:40:27,182 INFO [train2.py:809] (1/4) Epoch 1, batch 1150, loss[ctc_loss=0.6187, att_loss=0.5502, loss=0.5639, over 16974.00 frames. utt_duration=1359 frames, utt_pad_proportion=0.0071, over 50.00 utterances.], tot_loss[ctc_loss=0.7185, att_loss=0.6583, loss=0.6703, over 3271025.66 frames. utt_duration=1278 frames, utt_pad_proportion=0.04538, over 10246.89 utterances.], batch size: 50, lr: 4.94e-02, grad_scale: 8.0 2023-03-07 10:40:30,063 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1153.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 10:41:32,900 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 3.510e+02 5.540e+02 7.016e+02 8.798e+02 2.623e+03, threshold=1.403e+03, percent-clipped=4.0 2023-03-07 10:41:32,945 INFO [train2.py:809] (1/4) Epoch 1, batch 1200, loss[ctc_loss=0.6271, att_loss=0.5418, loss=0.5589, over 17357.00 frames. utt_duration=1178 frames, utt_pad_proportion=0.02068, over 59.00 utterances.], tot_loss[ctc_loss=0.6938, att_loss=0.6289, loss=0.6419, over 3274414.75 frames. utt_duration=1254 frames, utt_pad_proportion=0.05032, over 10457.25 utterances.], batch size: 59, lr: 4.93e-02, grad_scale: 8.0 2023-03-07 10:41:35,567 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1203.0, num_to_drop=2, layers_to_drop={0, 2} 2023-03-07 10:41:49,248 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1214.0, num_to_drop=2, layers_to_drop={0, 2} 2023-03-07 10:41:54,869 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1218.0, num_to_drop=2, layers_to_drop={0, 2} 2023-03-07 10:41:57,216 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=13.47 vs. limit=5.0 2023-03-07 10:42:03,496 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1224.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 10:42:05,986 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2902, 4.6322, 4.5972, 4.8404, 4.2484, 4.6283, 4.7552, 4.8477], device='cuda:1'), covar=tensor([0.0760, 0.0526, 0.0670, 0.0327, 0.0589, 0.0503, 0.0838, 0.0353], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0043, 0.0044, 0.0037, 0.0047, 0.0040, 0.0046, 0.0038], device='cuda:1'), out_proj_covar=tensor([4.3762e-05, 3.6530e-05, 3.9028e-05, 3.0373e-05, 4.0591e-05, 3.4159e-05, 4.0055e-05, 3.0589e-05], device='cuda:1') 2023-03-07 10:42:31,483 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1246.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 10:42:37,644 INFO [train2.py:809] (1/4) Epoch 1, batch 1250, loss[ctc_loss=0.5761, att_loss=0.5129, loss=0.5256, over 16390.00 frames. utt_duration=1492 frames, utt_pad_proportion=0.007546, over 44.00 utterances.], tot_loss[ctc_loss=0.6697, att_loss=0.6008, loss=0.6146, over 3271730.50 frames. utt_duration=1268 frames, utt_pad_proportion=0.04775, over 10329.32 utterances.], batch size: 44, lr: 4.92e-02, grad_scale: 8.0 2023-03-07 10:42:56,388 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=1266.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 10:43:05,041 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=1272.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 10:43:34,641 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=1294.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 10:43:43,368 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 3.330e+02 5.046e+02 6.291e+02 8.119e+02 1.875e+03, threshold=1.258e+03, percent-clipped=2.0 2023-03-07 10:43:43,411 INFO [train2.py:809] (1/4) Epoch 1, batch 1300, loss[ctc_loss=0.6202, att_loss=0.5024, loss=0.526, over 17305.00 frames. utt_duration=693.7 frames, utt_pad_proportion=0.1186, over 100.00 utterances.], tot_loss[ctc_loss=0.6515, att_loss=0.5787, loss=0.5932, over 3275057.16 frames. utt_duration=1281 frames, utt_pad_proportion=0.04487, over 10235.02 utterances.], batch size: 100, lr: 4.92e-02, grad_scale: 8.0 2023-03-07 10:44:24,583 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.75 vs. limit=2.0 2023-03-07 10:44:50,639 INFO [train2.py:809] (1/4) Epoch 1, batch 1350, loss[ctc_loss=0.5586, att_loss=0.485, loss=0.4998, over 16492.00 frames. utt_duration=1436 frames, utt_pad_proportion=0.005167, over 46.00 utterances.], tot_loss[ctc_loss=0.6337, att_loss=0.5588, loss=0.5738, over 3282289.55 frames. utt_duration=1291 frames, utt_pad_proportion=0.04036, over 10184.01 utterances.], batch size: 46, lr: 4.91e-02, grad_scale: 8.0 2023-03-07 10:45:59,937 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.741e+02 4.980e+02 5.949e+02 7.400e+02 1.342e+03, threshold=1.190e+03, percent-clipped=3.0 2023-03-07 10:45:59,980 INFO [train2.py:809] (1/4) Epoch 1, batch 1400, loss[ctc_loss=0.4526, att_loss=0.395, loss=0.4065, over 15895.00 frames. utt_duration=1632 frames, utt_pad_proportion=0.008505, over 39.00 utterances.], tot_loss[ctc_loss=0.6137, att_loss=0.5382, loss=0.5533, over 3272568.08 frames. utt_duration=1306 frames, utt_pad_proportion=0.03967, over 10036.74 utterances.], batch size: 39, lr: 4.91e-02, grad_scale: 8.0 2023-03-07 10:46:45,515 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1434.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 10:47:08,141 INFO [train2.py:809] (1/4) Epoch 1, batch 1450, loss[ctc_loss=0.6031, att_loss=0.4917, loss=0.514, over 16685.00 frames. utt_duration=1452 frames, utt_pad_proportion=0.005922, over 46.00 utterances.], tot_loss[ctc_loss=0.6005, att_loss=0.5239, loss=0.5392, over 3274584.99 frames. utt_duration=1288 frames, utt_pad_proportion=0.04347, over 10184.65 utterances.], batch size: 46, lr: 4.90e-02, grad_scale: 8.0 2023-03-07 10:48:08,691 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1495.0, num_to_drop=2, layers_to_drop={0, 3} 2023-03-07 10:48:16,252 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.693e+02 5.276e+02 6.895e+02 9.021e+02 2.982e+03, threshold=1.379e+03, percent-clipped=6.0 2023-03-07 10:48:16,297 INFO [train2.py:809] (1/4) Epoch 1, batch 1500, loss[ctc_loss=0.5554, att_loss=0.4629, loss=0.4814, over 16982.00 frames. utt_duration=687.7 frames, utt_pad_proportion=0.1372, over 99.00 utterances.], tot_loss[ctc_loss=0.5844, att_loss=0.5086, loss=0.5238, over 3282346.03 frames. utt_duration=1258 frames, utt_pad_proportion=0.0472, over 10452.17 utterances.], batch size: 99, lr: 4.89e-02, grad_scale: 8.0 2023-03-07 10:48:19,291 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1503.0, num_to_drop=1, layers_to_drop={2} 2023-03-07 10:48:25,239 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.98 vs. limit=2.0 2023-03-07 10:48:27,279 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1509.0, num_to_drop=2, layers_to_drop={1, 3} 2023-03-07 10:48:39,863 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=6.07 vs. limit=5.0 2023-03-07 10:49:08,622 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.88 vs. limit=2.0 2023-03-07 10:49:26,089 INFO [train2.py:809] (1/4) Epoch 1, batch 1550, loss[ctc_loss=0.434, att_loss=0.3927, loss=0.4009, over 15872.00 frames. utt_duration=1629 frames, utt_pad_proportion=0.01009, over 39.00 utterances.], tot_loss[ctc_loss=0.5721, att_loss=0.497, loss=0.512, over 3282946.74 frames. utt_duration=1250 frames, utt_pad_proportion=0.04803, over 10514.95 utterances.], batch size: 39, lr: 4.89e-02, grad_scale: 8.0 2023-03-07 10:49:26,171 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=1551.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 10:49:42,285 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2825, 4.6015, 4.3771, 4.5121, 4.6550, 4.3675, 4.6404, 4.8241], device='cuda:1'), covar=tensor([0.0502, 0.0325, 0.0460, 0.0443, 0.0353, 0.0526, 0.0311, 0.0348], device='cuda:1'), in_proj_covar=tensor([0.0034, 0.0029, 0.0025, 0.0026, 0.0029, 0.0033, 0.0027, 0.0029], device='cuda:1'), out_proj_covar=tensor([2.9424e-05, 2.2542e-05, 2.0752e-05, 2.0724e-05, 2.4609e-05, 2.8530e-05, 2.2205e-05, 2.4800e-05], device='cuda:1') 2023-03-07 10:49:46,969 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1566.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 10:50:35,402 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2915, 4.3412, 3.1917, 5.2853, 5.0816, 4.0000, 5.3805, 5.3142], device='cuda:1'), covar=tensor([0.2781, 0.1213, 0.0631, 0.0149, 0.0245, 0.1494, 0.0190, 0.0215], device='cuda:1'), in_proj_covar=tensor([0.0035, 0.0029, 0.0025, 0.0037, 0.0038, 0.0029, 0.0035, 0.0036], device='cuda:1'), out_proj_covar=tensor([2.7823e-05, 2.0493e-05, 1.5838e-05, 2.2749e-05, 2.4958e-05, 1.9713e-05, 2.2038e-05, 2.3427e-05], device='cuda:1') 2023-03-07 10:50:36,313 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 3.585e+02 5.591e+02 7.426e+02 9.323e+02 2.170e+03, threshold=1.485e+03, percent-clipped=5.0 2023-03-07 10:50:36,359 INFO [train2.py:809] (1/4) Epoch 1, batch 1600, loss[ctc_loss=0.4596, att_loss=0.4169, loss=0.4254, over 15872.00 frames. utt_duration=1630 frames, utt_pad_proportion=0.009985, over 39.00 utterances.], tot_loss[ctc_loss=0.5594, att_loss=0.4861, loss=0.5007, over 3282775.97 frames. utt_duration=1257 frames, utt_pad_proportion=0.04718, over 10457.31 utterances.], batch size: 39, lr: 4.88e-02, grad_scale: 8.0 2023-03-07 10:50:50,108 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1611.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 10:51:05,273 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 2023-03-07 10:51:13,095 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1627.0, num_to_drop=2, layers_to_drop={0, 2} 2023-03-07 10:51:44,857 INFO [train2.py:809] (1/4) Epoch 1, batch 1650, loss[ctc_loss=0.4918, att_loss=0.4454, loss=0.4547, over 15935.00 frames. utt_duration=1556 frames, utt_pad_proportion=0.007591, over 41.00 utterances.], tot_loss[ctc_loss=0.5454, att_loss=0.4753, loss=0.4893, over 3279913.57 frames. utt_duration=1266 frames, utt_pad_proportion=0.0468, over 10377.97 utterances.], batch size: 41, lr: 4.87e-02, grad_scale: 8.0 2023-03-07 10:52:15,858 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1672.0, num_to_drop=2, layers_to_drop={0, 2} 2023-03-07 10:52:55,824 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.825e+02 4.712e+02 6.880e+02 8.240e+02 2.356e+03, threshold=1.376e+03, percent-clipped=2.0 2023-03-07 10:52:55,869 INFO [train2.py:809] (1/4) Epoch 1, batch 1700, loss[ctc_loss=0.5871, att_loss=0.4936, loss=0.5123, over 13484.00 frames. utt_duration=373.6 frames, utt_pad_proportion=0.3502, over 145.00 utterances.], tot_loss[ctc_loss=0.5317, att_loss=0.4649, loss=0.4783, over 3274495.92 frames. utt_duration=1248 frames, utt_pad_proportion=0.05499, over 10505.53 utterances.], batch size: 145, lr: 4.86e-02, grad_scale: 8.0 2023-03-07 10:54:07,067 INFO [train2.py:809] (1/4) Epoch 1, batch 1750, loss[ctc_loss=0.4187, att_loss=0.3901, loss=0.3958, over 16178.00 frames. utt_duration=1580 frames, utt_pad_proportion=0.00632, over 41.00 utterances.], tot_loss[ctc_loss=0.5169, att_loss=0.4549, loss=0.4673, over 3277266.04 frames. utt_duration=1257 frames, utt_pad_proportion=0.05202, over 10442.93 utterances.], batch size: 41, lr: 4.86e-02, grad_scale: 8.0 2023-03-07 10:55:02,400 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1790.0, num_to_drop=1, layers_to_drop={3} 2023-03-07 10:55:02,482 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2470, 4.1143, 4.1432, 4.5152, 4.6096, 4.3385, 4.4753, 4.6901], device='cuda:1'), covar=tensor([0.0392, 0.0633, 0.0438, 0.0296, 0.0361, 0.0317, 0.0314, 0.0285], device='cuda:1'), in_proj_covar=tensor([0.0033, 0.0031, 0.0031, 0.0028, 0.0029, 0.0033, 0.0025, 0.0028], device='cuda:1'), out_proj_covar=tensor([2.6181e-05, 2.6419e-05, 2.5066e-05, 2.2568e-05, 2.3688e-05, 2.6674e-05, 1.8928e-05, 2.2568e-05], device='cuda:1') 2023-03-07 10:55:17,469 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 3.104e+02 4.727e+02 5.545e+02 7.358e+02 1.291e+03, threshold=1.109e+03, percent-clipped=0.0 2023-03-07 10:55:17,512 INFO [train2.py:809] (1/4) Epoch 1, batch 1800, loss[ctc_loss=0.3932, att_loss=0.3753, loss=0.3789, over 15634.00 frames. utt_duration=1692 frames, utt_pad_proportion=0.009384, over 37.00 utterances.], tot_loss[ctc_loss=0.5049, att_loss=0.4466, loss=0.4583, over 3266631.81 frames. utt_duration=1252 frames, utt_pad_proportion=0.05313, over 10445.05 utterances.], batch size: 37, lr: 4.85e-02, grad_scale: 8.0 2023-03-07 10:55:28,577 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1809.0, num_to_drop=1, layers_to_drop={2} 2023-03-07 10:56:09,991 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1838.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 10:56:27,798 INFO [train2.py:809] (1/4) Epoch 1, batch 1850, loss[ctc_loss=0.4813, att_loss=0.4403, loss=0.4485, over 16457.00 frames. utt_duration=1433 frames, utt_pad_proportion=0.007261, over 46.00 utterances.], tot_loss[ctc_loss=0.4975, att_loss=0.4425, loss=0.4535, over 3269417.47 frames. utt_duration=1233 frames, utt_pad_proportion=0.05805, over 10617.27 utterances.], batch size: 46, lr: 4.84e-02, grad_scale: 8.0 2023-03-07 10:56:36,582 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=1857.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 10:56:36,772 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1857.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 10:56:56,789 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1870.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 10:57:06,467 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9470, 4.1559, 3.5499, 4.0806, 4.4792, 4.1925, 3.9461, 3.5297], device='cuda:1'), covar=tensor([0.0323, 0.0225, 0.0537, 0.0300, 0.0172, 0.0338, 0.0335, 0.0475], device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0033, 0.0036, 0.0033, 0.0030, 0.0032, 0.0035, 0.0034], device='cuda:1'), out_proj_covar=tensor([2.2839e-05, 2.4501e-05, 2.8037e-05, 2.4994e-05, 2.2354e-05, 2.5412e-05, 2.7199e-05, 2.6012e-05], device='cuda:1') 2023-03-07 10:57:22,993 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4031, 4.5273, 3.3346, 5.0623, 4.1283, 4.6175, 5.2488, 5.0480], device='cuda:1'), covar=tensor([0.0956, 0.0535, 0.0424, 0.0150, 0.0775, 0.0414, 0.0117, 0.0148], device='cuda:1'), in_proj_covar=tensor([0.0034, 0.0033, 0.0030, 0.0044, 0.0049, 0.0030, 0.0041, 0.0039], device='cuda:1'), out_proj_covar=tensor([2.7196e-05, 2.2867e-05, 1.7399e-05, 2.6151e-05, 3.3267e-05, 1.9055e-05, 2.4208e-05, 2.2499e-05], device='cuda:1') 2023-03-07 10:57:37,052 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1899.0, num_to_drop=2, layers_to_drop={1, 3} 2023-03-07 10:57:39,638 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.994e+02 5.025e+02 6.383e+02 8.643e+02 1.619e+03, threshold=1.277e+03, percent-clipped=10.0 2023-03-07 10:57:39,683 INFO [train2.py:809] (1/4) Epoch 1, batch 1900, loss[ctc_loss=0.5032, att_loss=0.4544, loss=0.4642, over 16860.00 frames. utt_duration=682.6 frames, utt_pad_proportion=0.1414, over 99.00 utterances.], tot_loss[ctc_loss=0.4878, att_loss=0.4369, loss=0.4471, over 3275425.56 frames. utt_duration=1241 frames, utt_pad_proportion=0.05466, over 10571.71 utterances.], batch size: 99, lr: 4.83e-02, grad_scale: 8.0 2023-03-07 10:58:04,500 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-03-07 10:58:05,195 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1918.0, num_to_drop=2, layers_to_drop={0, 1} 2023-03-07 10:58:10,228 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1922.0, num_to_drop=1, layers_to_drop={3} 2023-03-07 10:58:10,790 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=5.22 vs. limit=5.0 2023-03-07 10:58:22,736 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1931.0, num_to_drop=2, layers_to_drop={0, 1} 2023-03-07 10:58:50,738 INFO [train2.py:809] (1/4) Epoch 1, batch 1950, loss[ctc_loss=0.3952, att_loss=0.3813, loss=0.3841, over 15887.00 frames. utt_duration=1631 frames, utt_pad_proportion=0.009191, over 39.00 utterances.], tot_loss[ctc_loss=0.4782, att_loss=0.4313, loss=0.4407, over 3269345.39 frames. utt_duration=1214 frames, utt_pad_proportion=0.0628, over 10782.08 utterances.], batch size: 39, lr: 4.83e-02, grad_scale: 8.0 2023-03-07 10:59:14,932 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1967.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 10:59:22,116 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2912, 4.9454, 5.3248, 5.1853, 4.6728, 4.6578, 5.2690, 5.0387], device='cuda:1'), covar=tensor([0.0394, 0.0351, 0.0196, 0.0185, 0.0284, 0.0267, 0.0321, 0.0231], device='cuda:1'), in_proj_covar=tensor([0.0054, 0.0042, 0.0043, 0.0037, 0.0047, 0.0042, 0.0046, 0.0040], device='cuda:1'), out_proj_covar=tensor([4.7917e-05, 3.5533e-05, 3.5953e-05, 2.7427e-05, 3.8172e-05, 3.2206e-05, 3.7671e-05, 2.8930e-05], device='cuda:1') 2023-03-07 10:59:30,481 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.5070, 2.7084, 4.0483, 3.9394, 3.6081, 3.3717, 3.2293, 3.3255], device='cuda:1'), covar=tensor([0.0552, 0.0907, 0.0234, 0.0408, 0.0366, 0.1019, 0.0583, 0.0971], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0049, 0.0062, 0.0084, 0.0078, 0.0100, 0.0046, 0.0085], device='cuda:1'), out_proj_covar=tensor([4.2175e-05, 4.2734e-05, 4.4948e-05, 5.9753e-05, 5.4885e-05, 8.2328e-05, 4.1084e-05, 6.4870e-05], device='cuda:1') 2023-03-07 11:00:06,966 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 3.041e+02 5.008e+02 6.125e+02 7.778e+02 1.503e+03, threshold=1.225e+03, percent-clipped=3.0 2023-03-07 11:00:07,010 INFO [train2.py:809] (1/4) Epoch 1, batch 2000, loss[ctc_loss=0.3862, att_loss=0.3698, loss=0.3731, over 16164.00 frames. utt_duration=1578 frames, utt_pad_proportion=0.00744, over 41.00 utterances.], tot_loss[ctc_loss=0.4683, att_loss=0.4256, loss=0.4341, over 3270438.81 frames. utt_duration=1224 frames, utt_pad_proportion=0.05923, over 10701.69 utterances.], batch size: 41, lr: 4.82e-02, grad_scale: 16.0 2023-03-07 11:00:19,392 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.6094, 2.3454, 3.8489, 3.7682, 3.6161, 3.4062, 3.0972, 3.4124], device='cuda:1'), covar=tensor([0.0360, 0.1210, 0.0268, 0.0318, 0.0349, 0.0897, 0.0710, 0.0770], device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0051, 0.0063, 0.0084, 0.0078, 0.0102, 0.0049, 0.0085], device='cuda:1'), out_proj_covar=tensor([4.2949e-05, 4.3573e-05, 4.5239e-05, 5.9180e-05, 5.4367e-05, 8.3846e-05, 4.1769e-05, 6.5421e-05], device='cuda:1') 2023-03-07 11:01:04,632 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.8298, 3.9579, 3.8890, 3.9038, 3.9315, 3.6092, 3.9164, 3.9467], device='cuda:1'), covar=tensor([0.0127, 0.0102, 0.0105, 0.0115, 0.0102, 0.0181, 0.0123, 0.0102], device='cuda:1'), in_proj_covar=tensor([0.0016, 0.0015, 0.0014, 0.0014, 0.0015, 0.0018, 0.0016, 0.0016], device='cuda:1'), out_proj_covar=tensor([1.3108e-05, 1.0658e-05, 1.0378e-05, 1.0196e-05, 1.1503e-05, 1.4388e-05, 1.1754e-05, 1.2706e-05], device='cuda:1') 2023-03-07 11:01:23,777 INFO [train2.py:809] (1/4) Epoch 1, batch 2050, loss[ctc_loss=0.366, att_loss=0.3709, loss=0.3699, over 16266.00 frames. utt_duration=1515 frames, utt_pad_proportion=0.008102, over 43.00 utterances.], tot_loss[ctc_loss=0.4531, att_loss=0.4177, loss=0.4248, over 3259453.94 frames. utt_duration=1193 frames, utt_pad_proportion=0.06987, over 10939.47 utterances.], batch size: 43, lr: 4.81e-02, grad_scale: 16.0 2023-03-07 11:01:57,112 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3746, 4.7712, 4.5750, 4.7982, 4.6083, 3.8314, 4.6155, 5.0459], device='cuda:1'), covar=tensor([0.0146, 0.0076, 0.0128, 0.0099, 0.0097, 0.0275, 0.0126, 0.0052], device='cuda:1'), in_proj_covar=tensor([0.0016, 0.0015, 0.0013, 0.0013, 0.0015, 0.0018, 0.0015, 0.0015], device='cuda:1'), out_proj_covar=tensor([1.2475e-05, 1.0221e-05, 9.8014e-06, 9.7226e-06, 1.0759e-05, 1.4171e-05, 1.1031e-05, 1.2003e-05], device='cuda:1') 2023-03-07 11:02:23,672 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2090.0, num_to_drop=2, layers_to_drop={0, 1} 2023-03-07 11:02:39,888 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.629e+02 3.921e+02 4.697e+02 6.442e+02 1.106e+03, threshold=9.394e+02, percent-clipped=0.0 2023-03-07 11:02:39,933 INFO [train2.py:809] (1/4) Epoch 1, batch 2100, loss[ctc_loss=0.446, att_loss=0.4362, loss=0.4381, over 17057.00 frames. utt_duration=1314 frames, utt_pad_proportion=0.008534, over 52.00 utterances.], tot_loss[ctc_loss=0.4392, att_loss=0.4111, loss=0.4167, over 3260839.71 frames. utt_duration=1204 frames, utt_pad_proportion=0.06904, over 10844.54 utterances.], batch size: 52, lr: 4.80e-02, grad_scale: 16.0 2023-03-07 11:02:47,546 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5233, 4.7259, 4.7851, 4.8518, 4.6956, 3.8098, 4.8599, 5.0923], device='cuda:1'), covar=tensor([0.0130, 0.0089, 0.0125, 0.0092, 0.0092, 0.0284, 0.0089, 0.0074], device='cuda:1'), in_proj_covar=tensor([0.0016, 0.0015, 0.0014, 0.0013, 0.0015, 0.0018, 0.0015, 0.0015], device='cuda:1'), out_proj_covar=tensor([1.2517e-05, 1.0595e-05, 1.0099e-05, 9.7774e-06, 1.0924e-05, 1.4510e-05, 1.0998e-05, 1.1953e-05], device='cuda:1') 2023-03-07 11:03:35,217 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=2138.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 11:03:55,160 INFO [train2.py:809] (1/4) Epoch 1, batch 2150, loss[ctc_loss=0.3603, att_loss=0.3451, loss=0.3482, over 15769.00 frames. utt_duration=1662 frames, utt_pad_proportion=0.008573, over 38.00 utterances.], tot_loss[ctc_loss=0.4273, att_loss=0.4058, loss=0.4101, over 3269764.95 frames. utt_duration=1224 frames, utt_pad_proportion=0.06259, over 10700.56 utterances.], batch size: 38, lr: 4.79e-02, grad_scale: 16.0 2023-03-07 11:04:15,377 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.2456, 5.5910, 5.6235, 6.0511, 5.7416, 5.9905, 5.6363, 5.9460], device='cuda:1'), covar=tensor([0.0191, 0.0258, 0.0341, 0.0196, 0.0323, 0.0320, 0.0227, 0.0335], device='cuda:1'), in_proj_covar=tensor([0.0086, 0.0102, 0.0088, 0.0096, 0.0105, 0.0103, 0.0095, 0.0110], device='cuda:1'), out_proj_covar=tensor([7.9033e-05, 8.8757e-05, 8.1616e-05, 9.0988e-05, 9.5666e-05, 1.0542e-04, 9.5821e-05, 1.0778e-04], device='cuda:1') 2023-03-07 11:04:41,902 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-03-07 11:05:00,215 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2194.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 11:05:10,567 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.636e+02 4.045e+02 5.653e+02 7.128e+02 1.127e+03, threshold=1.131e+03, percent-clipped=5.0 2023-03-07 11:05:10,613 INFO [train2.py:809] (1/4) Epoch 1, batch 2200, loss[ctc_loss=0.3297, att_loss=0.3513, loss=0.347, over 16376.00 frames. utt_duration=1490 frames, utt_pad_proportion=0.008453, over 44.00 utterances.], tot_loss[ctc_loss=0.4163, att_loss=0.4007, loss=0.4038, over 3261474.29 frames. utt_duration=1201 frames, utt_pad_proportion=0.06863, over 10874.77 utterances.], batch size: 44, lr: 4.78e-02, grad_scale: 16.0 2023-03-07 11:05:30,182 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2213.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 11:05:32,321 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-03-07 11:05:43,456 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2222.0, num_to_drop=2, layers_to_drop={0, 2} 2023-03-07 11:05:49,161 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2226.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 11:06:26,399 INFO [train2.py:809] (1/4) Epoch 1, batch 2250, loss[ctc_loss=0.3585, att_loss=0.3611, loss=0.3606, over 15366.00 frames. utt_duration=1758 frames, utt_pad_proportion=0.01138, over 35.00 utterances.], tot_loss[ctc_loss=0.4047, att_loss=0.395, loss=0.397, over 3255359.28 frames. utt_duration=1199 frames, utt_pad_proportion=0.07, over 10873.22 utterances.], batch size: 35, lr: 4.77e-02, grad_scale: 16.0 2023-03-07 11:06:43,033 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.93 vs. limit=2.0 2023-03-07 11:06:52,774 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2267.0, num_to_drop=1, layers_to_drop={2} 2023-03-07 11:06:57,129 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=2270.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 11:07:43,372 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.224e+02 3.835e+02 4.889e+02 6.269e+02 9.585e+02, threshold=9.779e+02, percent-clipped=0.0 2023-03-07 11:07:43,419 INFO [train2.py:809] (1/4) Epoch 1, batch 2300, loss[ctc_loss=0.3182, att_loss=0.346, loss=0.3405, over 16694.00 frames. utt_duration=1453 frames, utt_pad_proportion=0.006245, over 46.00 utterances.], tot_loss[ctc_loss=0.3965, att_loss=0.392, loss=0.3929, over 3257734.35 frames. utt_duration=1190 frames, utt_pad_proportion=0.07134, over 10964.88 utterances.], batch size: 46, lr: 4.77e-02, grad_scale: 16.0 2023-03-07 11:08:06,526 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=2315.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 11:08:43,533 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2340.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 11:08:59,804 INFO [train2.py:809] (1/4) Epoch 1, batch 2350, loss[ctc_loss=0.3689, att_loss=0.3756, loss=0.3742, over 17498.00 frames. utt_duration=1016 frames, utt_pad_proportion=0.0381, over 69.00 utterances.], tot_loss[ctc_loss=0.3884, att_loss=0.3892, loss=0.3891, over 3259380.17 frames. utt_duration=1209 frames, utt_pad_proportion=0.06733, over 10800.88 utterances.], batch size: 69, lr: 4.76e-02, grad_scale: 16.0 2023-03-07 11:10:15,554 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.741e+02 4.246e+02 5.457e+02 7.092e+02 1.687e+03, threshold=1.091e+03, percent-clipped=6.0 2023-03-07 11:10:15,599 INFO [train2.py:809] (1/4) Epoch 1, batch 2400, loss[ctc_loss=0.3609, att_loss=0.3868, loss=0.3816, over 16766.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.006644, over 48.00 utterances.], tot_loss[ctc_loss=0.3796, att_loss=0.385, loss=0.3839, over 3260860.53 frames. utt_duration=1222 frames, utt_pad_proportion=0.06296, over 10686.82 utterances.], batch size: 48, lr: 4.75e-02, grad_scale: 16.0 2023-03-07 11:10:15,991 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2401.0, num_to_drop=2, layers_to_drop={2, 3} 2023-03-07 11:11:30,936 INFO [train2.py:809] (1/4) Epoch 1, batch 2450, loss[ctc_loss=0.4539, att_loss=0.4262, loss=0.4317, over 14169.00 frames. utt_duration=392.2 frames, utt_pad_proportion=0.3191, over 145.00 utterances.], tot_loss[ctc_loss=0.3737, att_loss=0.3828, loss=0.381, over 3262398.57 frames. utt_duration=1207 frames, utt_pad_proportion=0.0662, over 10821.84 utterances.], batch size: 145, lr: 4.74e-02, grad_scale: 16.0 2023-03-07 11:12:36,929 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2494.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 11:12:47,385 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.845e+02 4.360e+02 5.388e+02 6.966e+02 1.787e+03, threshold=1.078e+03, percent-clipped=1.0 2023-03-07 11:12:47,430 INFO [train2.py:809] (1/4) Epoch 1, batch 2500, loss[ctc_loss=0.365, att_loss=0.3625, loss=0.363, over 15875.00 frames. utt_duration=1630 frames, utt_pad_proportion=0.009954, over 39.00 utterances.], tot_loss[ctc_loss=0.3676, att_loss=0.3805, loss=0.3779, over 3258135.41 frames. utt_duration=1212 frames, utt_pad_proportion=0.06663, over 10766.70 utterances.], batch size: 39, lr: 4.73e-02, grad_scale: 16.0 2023-03-07 11:12:59,065 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.7137, 4.0274, 4.2593, 3.6123, 3.9949, 4.2329, 4.0998, 3.6449], device='cuda:1'), covar=tensor([0.0193, 0.0128, 0.0091, 0.0292, 0.0241, 0.0115, 0.0141, 0.0350], device='cuda:1'), in_proj_covar=tensor([0.0020, 0.0020, 0.0019, 0.0020, 0.0018, 0.0018, 0.0018, 0.0024], device='cuda:1'), out_proj_covar=tensor([1.4144e-05, 1.4012e-05, 1.4085e-05, 1.5116e-05, 1.3387e-05, 1.3248e-05, 1.2672e-05, 1.7824e-05], device='cuda:1') 2023-03-07 11:13:06,121 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2513.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 11:13:25,384 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2526.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 11:13:28,880 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.94 vs. limit=2.0 2023-03-07 11:13:35,045 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.99 vs. limit=2.0 2023-03-07 11:13:49,476 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=2542.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 11:14:02,693 INFO [train2.py:809] (1/4) Epoch 1, batch 2550, loss[ctc_loss=0.2687, att_loss=0.3254, loss=0.3141, over 16401.00 frames. utt_duration=1492 frames, utt_pad_proportion=0.007676, over 44.00 utterances.], tot_loss[ctc_loss=0.3605, att_loss=0.3769, loss=0.3736, over 3264588.24 frames. utt_duration=1244 frames, utt_pad_proportion=0.05746, over 10511.02 utterances.], batch size: 44, lr: 4.72e-02, grad_scale: 16.0 2023-03-07 11:14:19,970 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=2561.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 11:14:39,272 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=2574.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 11:14:47,037 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5917, 4.8614, 4.9464, 4.7754, 3.5817, 5.1173, 5.3334, 5.2336], device='cuda:1'), covar=tensor([0.0771, 0.0585, 0.0139, 0.0264, 0.2283, 0.0199, 0.0086, 0.0090], device='cuda:1'), in_proj_covar=tensor([0.0043, 0.0047, 0.0045, 0.0063, 0.0107, 0.0039, 0.0055, 0.0053], device='cuda:1'), out_proj_covar=tensor([3.3714e-05, 3.1143e-05, 2.5729e-05, 3.8454e-05, 7.3636e-05, 2.4957e-05, 2.9494e-05, 2.6490e-05], device='cuda:1') 2023-03-07 11:15:19,992 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.773e+02 4.230e+02 5.272e+02 6.534e+02 1.107e+03, threshold=1.054e+03, percent-clipped=1.0 2023-03-07 11:15:20,037 INFO [train2.py:809] (1/4) Epoch 1, batch 2600, loss[ctc_loss=0.2935, att_loss=0.3297, loss=0.3225, over 12777.00 frames. utt_duration=1827 frames, utt_pad_proportion=0.1175, over 28.00 utterances.], tot_loss[ctc_loss=0.355, att_loss=0.3748, loss=0.3708, over 3264265.56 frames. utt_duration=1256 frames, utt_pad_proportion=0.05558, over 10406.10 utterances.], batch size: 28, lr: 4.71e-02, grad_scale: 16.0 2023-03-07 11:15:32,142 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.09 vs. limit=2.0 2023-03-07 11:16:37,077 INFO [train2.py:809] (1/4) Epoch 1, batch 2650, loss[ctc_loss=0.3188, att_loss=0.339, loss=0.335, over 16393.00 frames. utt_duration=1492 frames, utt_pad_proportion=0.007364, over 44.00 utterances.], tot_loss[ctc_loss=0.3473, att_loss=0.3707, loss=0.366, over 3265340.96 frames. utt_duration=1264 frames, utt_pad_proportion=0.05332, over 10347.11 utterances.], batch size: 44, lr: 4.70e-02, grad_scale: 16.0 2023-03-07 11:17:29,905 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2685.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 11:17:40,085 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.8122, 4.5988, 4.9196, 4.7998, 4.4186, 4.5993, 4.9692, 4.9198], device='cuda:1'), covar=tensor([0.0394, 0.0265, 0.0135, 0.0145, 0.0304, 0.0121, 0.0270, 0.0149], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0052, 0.0057, 0.0048, 0.0063, 0.0052, 0.0061, 0.0048], device='cuda:1'), out_proj_covar=tensor([7.3914e-05, 5.1025e-05, 5.2578e-05, 4.4413e-05, 6.3416e-05, 5.3045e-05, 6.1246e-05, 4.0816e-05], device='cuda:1') 2023-03-07 11:17:45,200 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.91 vs. limit=2.0 2023-03-07 11:17:45,985 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2696.0, num_to_drop=1, layers_to_drop={2} 2023-03-07 11:17:53,968 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.491e+02 4.149e+02 5.114e+02 6.615e+02 1.308e+03, threshold=1.023e+03, percent-clipped=1.0 2023-03-07 11:17:54,013 INFO [train2.py:809] (1/4) Epoch 1, batch 2700, loss[ctc_loss=0.3802, att_loss=0.3868, loss=0.3855, over 16973.00 frames. utt_duration=1359 frames, utt_pad_proportion=0.0071, over 50.00 utterances.], tot_loss[ctc_loss=0.3447, att_loss=0.3712, loss=0.3659, over 3273042.68 frames. utt_duration=1233 frames, utt_pad_proportion=0.05832, over 10631.66 utterances.], batch size: 50, lr: 4.69e-02, grad_scale: 16.0 2023-03-07 11:19:03,821 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2746.0, num_to_drop=2, layers_to_drop={0, 3} 2023-03-07 11:19:10,755 INFO [train2.py:809] (1/4) Epoch 1, batch 2750, loss[ctc_loss=0.2864, att_loss=0.3494, loss=0.3368, over 16469.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.007181, over 46.00 utterances.], tot_loss[ctc_loss=0.3403, att_loss=0.3694, loss=0.3636, over 3273111.31 frames. utt_duration=1250 frames, utt_pad_proportion=0.05405, over 10484.68 utterances.], batch size: 46, lr: 4.68e-02, grad_scale: 16.0 2023-03-07 11:20:27,838 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.038e+02 4.978e+02 5.756e+02 7.064e+02 1.353e+03, threshold=1.151e+03, percent-clipped=3.0 2023-03-07 11:20:27,887 INFO [train2.py:809] (1/4) Epoch 1, batch 2800, loss[ctc_loss=0.332, att_loss=0.3864, loss=0.3755, over 17263.00 frames. utt_duration=1172 frames, utt_pad_proportion=0.02584, over 59.00 utterances.], tot_loss[ctc_loss=0.3383, att_loss=0.3689, loss=0.3627, over 3276927.07 frames. utt_duration=1261 frames, utt_pad_proportion=0.05109, over 10411.21 utterances.], batch size: 59, lr: 4.67e-02, grad_scale: 16.0 2023-03-07 11:21:04,333 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2824.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 11:21:15,347 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 2023-03-07 11:21:44,977 INFO [train2.py:809] (1/4) Epoch 1, batch 2850, loss[ctc_loss=0.3994, att_loss=0.4074, loss=0.4058, over 17315.00 frames. utt_duration=1005 frames, utt_pad_proportion=0.05079, over 69.00 utterances.], tot_loss[ctc_loss=0.3368, att_loss=0.369, loss=0.3626, over 3275342.87 frames. utt_duration=1232 frames, utt_pad_proportion=0.05821, over 10645.52 utterances.], batch size: 69, lr: 4.66e-02, grad_scale: 16.0 2023-03-07 11:22:36,850 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2885.0, num_to_drop=2, layers_to_drop={1, 3} 2023-03-07 11:22:43,969 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-03-07 11:23:01,456 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.575e+02 4.414e+02 5.747e+02 7.042e+02 1.597e+03, threshold=1.149e+03, percent-clipped=5.0 2023-03-07 11:23:01,500 INFO [train2.py:809] (1/4) Epoch 1, batch 2900, loss[ctc_loss=0.33, att_loss=0.381, loss=0.3708, over 17333.00 frames. utt_duration=1177 frames, utt_pad_proportion=0.02196, over 59.00 utterances.], tot_loss[ctc_loss=0.3357, att_loss=0.3688, loss=0.3622, over 3269439.86 frames. utt_duration=1212 frames, utt_pad_proportion=0.06493, over 10804.12 utterances.], batch size: 59, lr: 4.65e-02, grad_scale: 16.0 2023-03-07 11:23:57,979 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2937.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 11:24:03,087 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-03-07 11:24:18,940 INFO [train2.py:809] (1/4) Epoch 1, batch 2950, loss[ctc_loss=0.3032, att_loss=0.3626, loss=0.3507, over 16963.00 frames. utt_duration=1359 frames, utt_pad_proportion=0.007582, over 50.00 utterances.], tot_loss[ctc_loss=0.3292, att_loss=0.3658, loss=0.3584, over 3268184.78 frames. utt_duration=1238 frames, utt_pad_proportion=0.05883, over 10574.14 utterances.], batch size: 50, lr: 4.64e-02, grad_scale: 16.0 2023-03-07 11:25:29,032 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2996.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 11:25:32,180 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2998.0, num_to_drop=2, layers_to_drop={0, 1} 2023-03-07 11:25:36,472 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.432e+02 4.332e+02 5.294e+02 6.689e+02 1.698e+03, threshold=1.059e+03, percent-clipped=3.0 2023-03-07 11:25:36,519 INFO [train2.py:809] (1/4) Epoch 1, batch 3000, loss[ctc_loss=0.3156, att_loss=0.362, loss=0.3527, over 17395.00 frames. utt_duration=882 frames, utt_pad_proportion=0.07641, over 79.00 utterances.], tot_loss[ctc_loss=0.3273, att_loss=0.365, loss=0.3575, over 3273995.97 frames. utt_duration=1231 frames, utt_pad_proportion=0.05856, over 10648.07 utterances.], batch size: 79, lr: 4.63e-02, grad_scale: 16.0 2023-03-07 11:25:36,519 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-07 11:25:51,186 INFO [train2.py:843] (1/4) Epoch 1, validation: ctc_loss=0.2388, att_loss=0.3154, loss=0.3001, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-07 11:25:51,187 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 14610MB 2023-03-07 11:26:08,408 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1824, 4.4026, 4.3376, 4.6582, 4.8320, 4.5567, 4.5397, 4.7663], device='cuda:1'), covar=tensor([0.0183, 0.0235, 0.0129, 0.0105, 0.0098, 0.0120, 0.0118, 0.0076], device='cuda:1'), in_proj_covar=tensor([0.0034, 0.0032, 0.0033, 0.0028, 0.0029, 0.0030, 0.0029, 0.0028], device='cuda:1'), out_proj_covar=tensor([3.6145e-05, 3.3296e-05, 3.4639e-05, 2.7391e-05, 2.6952e-05, 3.1120e-05, 2.6994e-05, 2.6159e-05], device='cuda:1') 2023-03-07 11:26:18,649 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3019.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 11:26:37,636 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.6939, 5.3329, 5.3710, 5.5495, 5.1480, 5.6507, 5.3179, 5.6971], device='cuda:1'), covar=tensor([0.0400, 0.0307, 0.0363, 0.0357, 0.0748, 0.0568, 0.0337, 0.0512], device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0134, 0.0111, 0.0125, 0.0171, 0.0130, 0.0115, 0.0140], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-07 11:26:51,061 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3041.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 11:26:55,339 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=3044.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 11:26:55,620 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2642, 5.6699, 5.0189, 4.5722, 4.6273, 4.6175, 5.0015, 4.7041], device='cuda:1'), covar=tensor([0.0221, 0.0051, 0.0218, 0.0440, 0.0402, 0.0414, 0.0280, 0.0445], device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0073, 0.0066, 0.0091, 0.0094, 0.0079, 0.0050, 0.0046], device='cuda:1'), out_proj_covar=tensor([2.3960e-05, 2.9113e-05, 2.6660e-05, 5.7122e-05, 5.2286e-05, 4.5640e-05, 3.0579e-05, 2.8170e-05], device='cuda:1') 2023-03-07 11:27:07,503 INFO [train2.py:809] (1/4) Epoch 1, batch 3050, loss[ctc_loss=0.2806, att_loss=0.338, loss=0.3265, over 16132.00 frames. utt_duration=1538 frames, utt_pad_proportion=0.005298, over 42.00 utterances.], tot_loss[ctc_loss=0.3229, att_loss=0.3625, loss=0.3546, over 3259436.32 frames. utt_duration=1229 frames, utt_pad_proportion=0.06418, over 10620.77 utterances.], batch size: 42, lr: 4.62e-02, grad_scale: 16.0 2023-03-07 11:27:28,995 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.20 vs. limit=2.0 2023-03-07 11:27:39,746 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.75 vs. limit=5.0 2023-03-07 11:27:52,688 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3080.0, num_to_drop=2, layers_to_drop={0, 3} 2023-03-07 11:28:24,852 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.428e+02 4.335e+02 5.357e+02 6.400e+02 1.084e+03, threshold=1.071e+03, percent-clipped=0.0 2023-03-07 11:28:24,900 INFO [train2.py:809] (1/4) Epoch 1, batch 3100, loss[ctc_loss=0.3538, att_loss=0.369, loss=0.366, over 15651.00 frames. utt_duration=1694 frames, utt_pad_proportion=0.00845, over 37.00 utterances.], tot_loss[ctc_loss=0.3178, att_loss=0.3595, loss=0.3512, over 3255153.69 frames. utt_duration=1263 frames, utt_pad_proportion=0.05798, over 10322.16 utterances.], batch size: 37, lr: 4.61e-02, grad_scale: 16.0 2023-03-07 11:28:42,247 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.8024, 5.4724, 5.4685, 5.6433, 5.1877, 5.7863, 5.3413, 5.6648], device='cuda:1'), covar=tensor([0.0359, 0.0235, 0.0274, 0.0315, 0.0806, 0.0391, 0.0357, 0.0456], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0140, 0.0114, 0.0129, 0.0184, 0.0133, 0.0122, 0.0146], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-07 11:29:25,844 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3562, 5.2088, 4.9825, 4.4483, 4.3936, 4.5187, 4.9452, 4.6306], device='cuda:1'), covar=tensor([0.0185, 0.0142, 0.0146, 0.0499, 0.0666, 0.0407, 0.0205, 0.0330], device='cuda:1'), in_proj_covar=tensor([0.0054, 0.0076, 0.0067, 0.0093, 0.0102, 0.0081, 0.0053, 0.0048], device='cuda:1'), out_proj_covar=tensor([2.5194e-05, 3.0318e-05, 2.7510e-05, 5.8191e-05, 5.7415e-05, 4.7517e-05, 3.1736e-05, 2.9470e-05], device='cuda:1') 2023-03-07 11:29:41,932 INFO [train2.py:809] (1/4) Epoch 1, batch 3150, loss[ctc_loss=0.306, att_loss=0.365, loss=0.3532, over 16629.00 frames. utt_duration=1417 frames, utt_pad_proportion=0.00511, over 47.00 utterances.], tot_loss[ctc_loss=0.3154, att_loss=0.3594, loss=0.3506, over 3259918.45 frames. utt_duration=1264 frames, utt_pad_proportion=0.05688, over 10328.12 utterances.], batch size: 47, lr: 4.60e-02, grad_scale: 16.0 2023-03-07 11:30:18,646 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.4850, 2.2730, 2.8642, 3.4587, 3.2114, 2.8545, 2.8737, 3.0869], device='cuda:1'), covar=tensor([0.0147, 0.0518, 0.0262, 0.0134, 0.0263, 0.0649, 0.0465, 0.0247], device='cuda:1'), in_proj_covar=tensor([0.0049, 0.0037, 0.0038, 0.0046, 0.0041, 0.0054, 0.0057, 0.0051], device='cuda:1'), out_proj_covar=tensor([3.3075e-05, 3.0973e-05, 3.1764e-05, 2.9333e-05, 3.2080e-05, 5.8951e-05, 4.9363e-05, 3.5839e-05], device='cuda:1') 2023-03-07 11:30:22,051 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=2.04 vs. limit=2.0 2023-03-07 11:30:25,827 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3180.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 11:30:58,388 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.970e+02 4.582e+02 5.665e+02 7.020e+02 1.937e+03, threshold=1.133e+03, percent-clipped=6.0 2023-03-07 11:30:58,431 INFO [train2.py:809] (1/4) Epoch 1, batch 3200, loss[ctc_loss=0.3003, att_loss=0.3622, loss=0.3498, over 16867.00 frames. utt_duration=1378 frames, utt_pad_proportion=0.007728, over 49.00 utterances.], tot_loss[ctc_loss=0.3144, att_loss=0.3592, loss=0.3502, over 3267280.37 frames. utt_duration=1268 frames, utt_pad_proportion=0.05332, over 10316.17 utterances.], batch size: 49, lr: 4.59e-02, grad_scale: 16.0 2023-03-07 11:31:23,963 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4063, 4.4580, 4.4196, 4.7590, 4.7017, 4.4982, 4.6451, 4.9383], device='cuda:1'), covar=tensor([0.0162, 0.0279, 0.0139, 0.0103, 0.0134, 0.0156, 0.0171, 0.0094], device='cuda:1'), in_proj_covar=tensor([0.0033, 0.0032, 0.0033, 0.0028, 0.0029, 0.0030, 0.0029, 0.0028], device='cuda:1'), out_proj_covar=tensor([3.7821e-05, 3.4234e-05, 3.6197e-05, 2.7630e-05, 2.8247e-05, 3.3427e-05, 2.8685e-05, 2.7986e-05], device='cuda:1') 2023-03-07 11:32:13,829 INFO [train2.py:809] (1/4) Epoch 1, batch 3250, loss[ctc_loss=0.335, att_loss=0.38, loss=0.371, over 17287.00 frames. utt_duration=1174 frames, utt_pad_proportion=0.02444, over 59.00 utterances.], tot_loss[ctc_loss=0.3129, att_loss=0.3591, loss=0.3498, over 3272429.82 frames. utt_duration=1282 frames, utt_pad_proportion=0.0491, over 10222.94 utterances.], batch size: 59, lr: 4.58e-02, grad_scale: 16.0 2023-03-07 11:32:24,145 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9284, 4.6958, 5.0469, 4.7068, 4.3564, 4.5239, 4.8547, 4.8112], device='cuda:1'), covar=tensor([0.0331, 0.0140, 0.0127, 0.0217, 0.0254, 0.0126, 0.0275, 0.0141], device='cuda:1'), in_proj_covar=tensor([0.0076, 0.0058, 0.0065, 0.0053, 0.0071, 0.0060, 0.0070, 0.0053], device='cuda:1'), out_proj_covar=tensor([9.2183e-05, 6.3162e-05, 6.6901e-05, 5.5743e-05, 7.9667e-05, 7.2667e-05, 8.1466e-05, 5.1015e-05], device='cuda:1') 2023-03-07 11:33:09,815 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3288.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 11:33:17,871 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3293.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 11:33:29,893 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.939e+02 4.361e+02 5.805e+02 6.999e+02 1.268e+03, threshold=1.161e+03, percent-clipped=3.0 2023-03-07 11:33:29,938 INFO [train2.py:809] (1/4) Epoch 1, batch 3300, loss[ctc_loss=0.2263, att_loss=0.3012, loss=0.2862, over 15905.00 frames. utt_duration=1633 frames, utt_pad_proportion=0.007976, over 39.00 utterances.], tot_loss[ctc_loss=0.3087, att_loss=0.3564, loss=0.3468, over 3270271.22 frames. utt_duration=1286 frames, utt_pad_proportion=0.04966, over 10184.58 utterances.], batch size: 39, lr: 4.57e-02, grad_scale: 16.0 2023-03-07 11:33:39,797 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.96 vs. limit=5.0 2023-03-07 11:34:19,610 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2627, 2.3167, 2.6366, 3.4569, 3.0469, 2.8016, 2.6176, 3.3980], device='cuda:1'), covar=tensor([0.0249, 0.0485, 0.0377, 0.0173, 0.0276, 0.0659, 0.0639, 0.0194], device='cuda:1'), in_proj_covar=tensor([0.0048, 0.0035, 0.0036, 0.0044, 0.0038, 0.0052, 0.0055, 0.0046], device='cuda:1'), out_proj_covar=tensor([3.2786e-05, 2.9905e-05, 3.1758e-05, 2.8950e-05, 3.0106e-05, 5.6022e-05, 4.7153e-05, 3.2461e-05], device='cuda:1') 2023-03-07 11:34:30,980 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3341.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 11:34:43,779 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3349.0, num_to_drop=2, layers_to_drop={2, 3} 2023-03-07 11:34:46,405 INFO [train2.py:809] (1/4) Epoch 1, batch 3350, loss[ctc_loss=0.2497, att_loss=0.3195, loss=0.3055, over 16177.00 frames. utt_duration=1580 frames, utt_pad_proportion=0.006519, over 41.00 utterances.], tot_loss[ctc_loss=0.3047, att_loss=0.3542, loss=0.3443, over 3271515.81 frames. utt_duration=1296 frames, utt_pad_proportion=0.04521, over 10110.03 utterances.], batch size: 41, lr: 4.56e-02, grad_scale: 16.0 2023-03-07 11:35:23,026 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3375.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 11:35:44,663 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=3389.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 11:36:04,118 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.598e+02 4.455e+02 5.465e+02 6.622e+02 1.273e+03, threshold=1.093e+03, percent-clipped=2.0 2023-03-07 11:36:04,165 INFO [train2.py:809] (1/4) Epoch 1, batch 3400, loss[ctc_loss=0.437, att_loss=0.4329, loss=0.4337, over 14188.00 frames. utt_duration=392.9 frames, utt_pad_proportion=0.3167, over 145.00 utterances.], tot_loss[ctc_loss=0.3034, att_loss=0.3542, loss=0.344, over 3261372.70 frames. utt_duration=1249 frames, utt_pad_proportion=0.05914, over 10453.62 utterances.], batch size: 145, lr: 4.55e-02, grad_scale: 16.0 2023-03-07 11:37:16,655 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.81 vs. limit=5.0 2023-03-07 11:37:19,750 INFO [train2.py:809] (1/4) Epoch 1, batch 3450, loss[ctc_loss=0.2828, att_loss=0.3412, loss=0.3295, over 16531.00 frames. utt_duration=1471 frames, utt_pad_proportion=0.006617, over 45.00 utterances.], tot_loss[ctc_loss=0.3033, att_loss=0.3545, loss=0.3442, over 3266586.14 frames. utt_duration=1258 frames, utt_pad_proportion=0.05549, over 10400.23 utterances.], batch size: 45, lr: 4.54e-02, grad_scale: 16.0 2023-03-07 11:37:41,371 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.89 vs. limit=2.0 2023-03-07 11:37:54,412 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.70 vs. limit=5.0 2023-03-07 11:38:03,901 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3480.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 11:38:35,896 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.427e+02 4.628e+02 6.013e+02 7.728e+02 1.949e+03, threshold=1.203e+03, percent-clipped=6.0 2023-03-07 11:38:35,940 INFO [train2.py:809] (1/4) Epoch 1, batch 3500, loss[ctc_loss=0.2983, att_loss=0.3424, loss=0.3336, over 15878.00 frames. utt_duration=1630 frames, utt_pad_proportion=0.008641, over 39.00 utterances.], tot_loss[ctc_loss=0.3016, att_loss=0.353, loss=0.3427, over 3263408.56 frames. utt_duration=1268 frames, utt_pad_proportion=0.05296, over 10306.92 utterances.], batch size: 39, lr: 4.53e-02, grad_scale: 16.0 2023-03-07 11:39:11,450 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1629, 4.6680, 4.6929, 4.7457, 2.9158, 4.7747, 5.0607, 5.2016], device='cuda:1'), covar=tensor([0.0817, 0.0351, 0.0323, 0.0336, 0.3491, 0.0249, 0.0150, 0.0082], device='cuda:1'), in_proj_covar=tensor([0.0058, 0.0066, 0.0081, 0.0118, 0.0187, 0.0065, 0.0096, 0.0091], device='cuda:1'), out_proj_covar=tensor([4.5706e-05, 4.4742e-05, 4.6712e-05, 7.2394e-05, 1.2672e-04, 3.8913e-05, 5.2572e-05, 4.5545e-05], device='cuda:1') 2023-03-07 11:39:13,361 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.09 vs. limit=2.0 2023-03-07 11:39:15,024 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-03-07 11:39:16,981 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=3528.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 11:39:19,098 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.03 vs. limit=2.0 2023-03-07 11:39:26,282 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.1225, 1.9758, 2.7999, 3.3185, 3.3977, 3.1215, 3.0122, 1.9582], device='cuda:1'), covar=tensor([0.0662, 0.1007, 0.0380, 0.0294, 0.0388, 0.0468, 0.0603, 0.1957], device='cuda:1'), in_proj_covar=tensor([0.0072, 0.0065, 0.0065, 0.0091, 0.0087, 0.0082, 0.0060, 0.0107], device='cuda:1'), out_proj_covar=tensor([5.4345e-05, 4.8117e-05, 4.4792e-05, 5.8858e-05, 5.2456e-05, 7.5577e-05, 4.4015e-05, 8.9074e-05], device='cuda:1') 2023-03-07 11:39:45,692 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3546.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 11:39:52,694 INFO [train2.py:809] (1/4) Epoch 1, batch 3550, loss[ctc_loss=0.3076, att_loss=0.3683, loss=0.3561, over 17033.00 frames. utt_duration=1337 frames, utt_pad_proportion=0.007797, over 51.00 utterances.], tot_loss[ctc_loss=0.3057, att_loss=0.3556, loss=0.3456, over 3273757.64 frames. utt_duration=1267 frames, utt_pad_proportion=0.05068, over 10351.47 utterances.], batch size: 51, lr: 4.51e-02, grad_scale: 16.0 2023-03-07 11:40:58,964 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3593.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 11:41:10,753 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.449e+02 6.503e+02 7.868e+02 1.011e+03 2.300e+03, threshold=1.574e+03, percent-clipped=15.0 2023-03-07 11:41:10,798 INFO [train2.py:809] (1/4) Epoch 1, batch 3600, loss[ctc_loss=0.2675, att_loss=0.3195, loss=0.3091, over 16538.00 frames. utt_duration=1471 frames, utt_pad_proportion=0.005781, over 45.00 utterances.], tot_loss[ctc_loss=0.3021, att_loss=0.3533, loss=0.3431, over 3270390.26 frames. utt_duration=1281 frames, utt_pad_proportion=0.0475, over 10223.03 utterances.], batch size: 45, lr: 4.50e-02, grad_scale: 16.0 2023-03-07 11:41:11,029 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.8647, 5.3453, 5.3421, 5.6972, 5.1831, 5.8040, 5.2679, 5.6584], device='cuda:1'), covar=tensor([0.0432, 0.0316, 0.0371, 0.0275, 0.1088, 0.0500, 0.0335, 0.0533], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0155, 0.0127, 0.0145, 0.0218, 0.0158, 0.0135, 0.0159], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-07 11:41:20,381 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3607.0, num_to_drop=2, layers_to_drop={2, 3} 2023-03-07 11:42:12,270 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=3641.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 11:42:17,634 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3644.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 11:42:27,538 INFO [train2.py:809] (1/4) Epoch 1, batch 3650, loss[ctc_loss=0.2825, att_loss=0.3314, loss=0.3217, over 15906.00 frames. utt_duration=1632 frames, utt_pad_proportion=0.008241, over 39.00 utterances.], tot_loss[ctc_loss=0.302, att_loss=0.3539, loss=0.3436, over 3274779.73 frames. utt_duration=1263 frames, utt_pad_proportion=0.05156, over 10382.77 utterances.], batch size: 39, lr: 4.49e-02, grad_scale: 16.0 2023-03-07 11:43:05,251 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3675.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 11:43:06,907 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3676.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 11:43:47,094 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 3.430e+02 5.749e+02 7.107e+02 9.009e+02 1.799e+03, threshold=1.421e+03, percent-clipped=2.0 2023-03-07 11:43:47,140 INFO [train2.py:809] (1/4) Epoch 1, batch 3700, loss[ctc_loss=0.2977, att_loss=0.3423, loss=0.3334, over 15938.00 frames. utt_duration=1557 frames, utt_pad_proportion=0.00728, over 41.00 utterances.], tot_loss[ctc_loss=0.3003, att_loss=0.3532, loss=0.3426, over 3281062.75 frames. utt_duration=1263 frames, utt_pad_proportion=0.04844, over 10400.44 utterances.], batch size: 41, lr: 4.48e-02, grad_scale: 16.0 2023-03-07 11:44:21,453 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=3723.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 11:44:34,167 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5485, 4.8566, 4.6452, 5.0416, 4.9423, 4.8355, 4.7719, 2.7735], device='cuda:1'), covar=tensor([0.0951, 0.0585, 0.0946, 0.0326, 0.0868, 0.0541, 0.0875, 0.8239], device='cuda:1'), in_proj_covar=tensor([0.0113, 0.0075, 0.0053, 0.0083, 0.0088, 0.0089, 0.0062, 0.0174], device='cuda:1'), out_proj_covar=tensor([7.0499e-05, 3.4812e-05, 3.8082e-05, 3.6574e-05, 5.0366e-05, 4.6936e-05, 4.1199e-05, 1.0645e-04], device='cuda:1') 2023-03-07 11:44:45,201 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3737.0, num_to_drop=2, layers_to_drop={2, 3} 2023-03-07 11:45:00,913 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.88 vs. limit=2.0 2023-03-07 11:45:06,157 INFO [train2.py:809] (1/4) Epoch 1, batch 3750, loss[ctc_loss=0.2591, att_loss=0.3218, loss=0.3093, over 16018.00 frames. utt_duration=1603 frames, utt_pad_proportion=0.006645, over 40.00 utterances.], tot_loss[ctc_loss=0.2963, att_loss=0.3509, loss=0.34, over 3261801.62 frames. utt_duration=1266 frames, utt_pad_proportion=0.05383, over 10318.59 utterances.], batch size: 40, lr: 4.47e-02, grad_scale: 16.0 2023-03-07 11:46:23,998 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 3.021e+02 4.717e+02 6.819e+02 8.989e+02 1.986e+03, threshold=1.364e+03, percent-clipped=6.0 2023-03-07 11:46:24,043 INFO [train2.py:809] (1/4) Epoch 1, batch 3800, loss[ctc_loss=0.2962, att_loss=0.3337, loss=0.3262, over 11107.00 frames. utt_duration=1852 frames, utt_pad_proportion=0.1877, over 24.00 utterances.], tot_loss[ctc_loss=0.294, att_loss=0.3496, loss=0.3385, over 3259172.67 frames. utt_duration=1287 frames, utt_pad_proportion=0.04962, over 10139.62 utterances.], batch size: 24, lr: 4.46e-02, grad_scale: 16.0 2023-03-07 11:47:08,750 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=2.64 vs. limit=2.0 2023-03-07 11:47:43,587 INFO [train2.py:809] (1/4) Epoch 1, batch 3850, loss[ctc_loss=0.2529, att_loss=0.3398, loss=0.3224, over 16126.00 frames. utt_duration=1537 frames, utt_pad_proportion=0.006495, over 42.00 utterances.], tot_loss[ctc_loss=0.2905, att_loss=0.348, loss=0.3365, over 3263661.64 frames. utt_duration=1301 frames, utt_pad_proportion=0.04393, over 10048.11 utterances.], batch size: 42, lr: 4.45e-02, grad_scale: 16.0 2023-03-07 11:47:59,793 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=2.35 vs. limit=2.0 2023-03-07 11:49:01,029 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 3.357e+02 5.503e+02 6.521e+02 8.081e+02 1.434e+03, threshold=1.304e+03, percent-clipped=2.0 2023-03-07 11:49:01,075 INFO [train2.py:809] (1/4) Epoch 1, batch 3900, loss[ctc_loss=0.3479, att_loss=0.3918, loss=0.383, over 17183.00 frames. utt_duration=871.7 frames, utt_pad_proportion=0.08719, over 79.00 utterances.], tot_loss[ctc_loss=0.2919, att_loss=0.3494, loss=0.3379, over 3272769.24 frames. utt_duration=1273 frames, utt_pad_proportion=0.04765, over 10296.05 utterances.], batch size: 79, lr: 4.44e-02, grad_scale: 16.0 2023-03-07 11:49:02,686 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3902.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 11:49:39,869 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3926.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 11:50:08,118 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3944.0, num_to_drop=1, layers_to_drop={2} 2023-03-07 11:50:18,520 INFO [train2.py:809] (1/4) Epoch 1, batch 3950, loss[ctc_loss=0.2715, att_loss=0.3333, loss=0.3209, over 14545.00 frames. utt_duration=1820 frames, utt_pad_proportion=0.03303, over 32.00 utterances.], tot_loss[ctc_loss=0.2927, att_loss=0.3503, loss=0.3388, over 3265029.29 frames. utt_duration=1215 frames, utt_pad_proportion=0.06488, over 10760.94 utterances.], batch size: 32, lr: 4.43e-02, grad_scale: 16.0 2023-03-07 11:51:37,722 INFO [train2.py:809] (1/4) Epoch 2, batch 0, loss[ctc_loss=0.2516, att_loss=0.3228, loss=0.3085, over 16413.00 frames. utt_duration=1494 frames, utt_pad_proportion=0.006891, over 44.00 utterances.], tot_loss[ctc_loss=0.2516, att_loss=0.3228, loss=0.3085, over 16413.00 frames. utt_duration=1494 frames, utt_pad_proportion=0.006891, over 44.00 utterances.], batch size: 44, lr: 4.34e-02, grad_scale: 8.0 2023-03-07 11:51:37,723 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-07 11:51:49,515 INFO [train2.py:843] (1/4) Epoch 2, validation: ctc_loss=0.1604, att_loss=0.2954, loss=0.2684, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-07 11:51:49,516 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 15912MB 2023-03-07 11:51:52,993 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3987.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 11:52:01,256 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=3992.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 11:52:20,460 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.599e+02 5.286e+02 6.545e+02 8.035e+02 1.268e+03, threshold=1.309e+03, percent-clipped=0.0 2023-03-07 11:53:10,013 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4032.0, num_to_drop=1, layers_to_drop={2} 2023-03-07 11:53:14,643 INFO [train2.py:809] (1/4) Epoch 2, batch 50, loss[ctc_loss=0.2976, att_loss=0.3345, loss=0.3271, over 15650.00 frames. utt_duration=1693 frames, utt_pad_proportion=0.008545, over 37.00 utterances.], tot_loss[ctc_loss=0.2858, att_loss=0.3485, loss=0.336, over 728611.15 frames. utt_duration=1217 frames, utt_pad_proportion=0.07906, over 2397.32 utterances.], batch size: 37, lr: 4.33e-02, grad_scale: 8.0 2023-03-07 11:53:23,066 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.23 vs. limit=5.0 2023-03-07 11:54:03,210 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9030, 4.4828, 5.0853, 3.9375, 3.9658, 4.0101, 4.6901, 4.5496], device='cuda:1'), covar=tensor([0.1403, 0.1167, 0.0390, 0.2414, 0.3144, 0.3319, 0.0670, 0.1160], device='cuda:1'), in_proj_covar=tensor([0.0054, 0.0094, 0.0065, 0.0096, 0.0121, 0.0078, 0.0052, 0.0048], device='cuda:1'), out_proj_covar=tensor([2.4746e-05, 4.1386e-05, 2.7301e-05, 5.6009e-05, 7.0617e-05, 4.8732e-05, 2.7612e-05, 2.6279e-05], device='cuda:1') 2023-03-07 11:54:37,352 INFO [train2.py:809] (1/4) Epoch 2, batch 100, loss[ctc_loss=0.306, att_loss=0.3668, loss=0.3547, over 17319.00 frames. utt_duration=1101 frames, utt_pad_proportion=0.03321, over 63.00 utterances.], tot_loss[ctc_loss=0.2843, att_loss=0.3472, loss=0.3346, over 1303392.27 frames. utt_duration=1276 frames, utt_pad_proportion=0.04792, over 4089.74 utterances.], batch size: 63, lr: 4.31e-02, grad_scale: 8.0 2023-03-07 11:55:04,165 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.3369, 2.6853, 3.0112, 3.1936, 2.4975, 2.8533, 2.5129, 3.4994], device='cuda:1'), covar=tensor([0.0229, 0.0351, 0.0576, 0.0299, 0.0427, 0.0574, 0.0636, 0.0116], device='cuda:1'), in_proj_covar=tensor([0.0031, 0.0024, 0.0041, 0.0035, 0.0027, 0.0037, 0.0040, 0.0024], device='cuda:1'), out_proj_covar=tensor([2.8703e-05, 3.0018e-05, 4.6737e-05, 3.2612e-05, 2.8364e-05, 3.8330e-05, 3.8441e-05, 2.5428e-05], device='cuda:1') 2023-03-07 11:55:05,392 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.894e+02 5.347e+02 6.194e+02 8.279e+02 2.356e+03, threshold=1.239e+03, percent-clipped=4.0 2023-03-07 11:55:36,705 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.04 vs. limit=2.0 2023-03-07 11:56:00,900 INFO [train2.py:809] (1/4) Epoch 2, batch 150, loss[ctc_loss=0.2767, att_loss=0.3465, loss=0.3326, over 16966.00 frames. utt_duration=1359 frames, utt_pad_proportion=0.007524, over 50.00 utterances.], tot_loss[ctc_loss=0.284, att_loss=0.3469, loss=0.3344, over 1748511.03 frames. utt_duration=1257 frames, utt_pad_proportion=0.04781, over 5570.78 utterances.], batch size: 50, lr: 4.30e-02, grad_scale: 8.0 2023-03-07 11:56:11,609 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=4141.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 11:56:42,902 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2937, 4.4390, 4.2438, 4.5038, 4.6153, 4.3495, 4.5350, 4.4214], device='cuda:1'), covar=tensor([0.0117, 0.0222, 0.0138, 0.0148, 0.0085, 0.0135, 0.0149, 0.0131], device='cuda:1'), in_proj_covar=tensor([0.0033, 0.0036, 0.0036, 0.0030, 0.0027, 0.0031, 0.0033, 0.0033], device='cuda:1'), out_proj_covar=tensor([4.2660e-05, 4.6623e-05, 4.8277e-05, 3.8681e-05, 3.1706e-05, 4.2696e-05, 4.1209e-05, 4.0219e-05], device='cuda:1') 2023-03-07 11:57:12,330 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=5.60 vs. limit=5.0 2023-03-07 11:57:24,055 INFO [train2.py:809] (1/4) Epoch 2, batch 200, loss[ctc_loss=0.2679, att_loss=0.3378, loss=0.3238, over 16609.00 frames. utt_duration=1415 frames, utt_pad_proportion=0.005682, over 47.00 utterances.], tot_loss[ctc_loss=0.2826, att_loss=0.3452, loss=0.3327, over 2084461.11 frames. utt_duration=1231 frames, utt_pad_proportion=0.05621, over 6779.59 utterances.], batch size: 47, lr: 4.29e-02, grad_scale: 8.0 2023-03-07 11:57:52,230 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 3.168e+02 5.150e+02 6.618e+02 8.460e+02 1.655e+03, threshold=1.324e+03, percent-clipped=5.0 2023-03-07 11:57:52,603 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4202.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 11:57:52,685 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=4202.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 11:57:58,303 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-03-07 11:58:13,062 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.87 vs. limit=2.0 2023-03-07 11:58:46,862 INFO [train2.py:809] (1/4) Epoch 2, batch 250, loss[ctc_loss=0.3335, att_loss=0.3822, loss=0.3725, over 17062.00 frames. utt_duration=1289 frames, utt_pad_proportion=0.009267, over 53.00 utterances.], tot_loss[ctc_loss=0.2809, att_loss=0.3435, loss=0.331, over 2346996.53 frames. utt_duration=1214 frames, utt_pad_proportion=0.06313, over 7745.69 utterances.], batch size: 53, lr: 4.28e-02, grad_scale: 8.0 2023-03-07 11:59:11,786 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=4250.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 11:59:27,816 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0264, 2.8707, 3.0258, 3.6150, 4.1300, 3.7548, 2.7777, 1.9702], device='cuda:1'), covar=tensor([0.0219, 0.0525, 0.0544, 0.0547, 0.0216, 0.0214, 0.1127, 0.1559], device='cuda:1'), in_proj_covar=tensor([0.0056, 0.0051, 0.0062, 0.0046, 0.0040, 0.0052, 0.0083, 0.0079], device='cuda:1'), out_proj_covar=tensor([4.3020e-05, 4.7360e-05, 5.0563e-05, 4.5291e-05, 3.7221e-05, 3.9562e-05, 7.7786e-05, 6.5922e-05], device='cuda:1') 2023-03-07 12:00:03,999 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4282.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:00:08,483 INFO [train2.py:809] (1/4) Epoch 2, batch 300, loss[ctc_loss=0.2641, att_loss=0.3527, loss=0.335, over 17036.00 frames. utt_duration=1288 frames, utt_pad_proportion=0.01035, over 53.00 utterances.], tot_loss[ctc_loss=0.2802, att_loss=0.3436, loss=0.3309, over 2547700.81 frames. utt_duration=1230 frames, utt_pad_proportion=0.06148, over 8298.69 utterances.], batch size: 53, lr: 4.27e-02, grad_scale: 8.0 2023-03-07 12:00:14,414 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=4288.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:00:36,611 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.675e+02 5.336e+02 6.631e+02 8.677e+02 1.956e+03, threshold=1.326e+03, percent-clipped=4.0 2023-03-07 12:00:43,036 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5530, 5.0747, 4.8647, 4.7159, 5.2316, 5.1351, 5.0065, 4.9582], device='cuda:1'), covar=tensor([0.0636, 0.0244, 0.0286, 0.0458, 0.0266, 0.0221, 0.0240, 0.0314], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0117, 0.0086, 0.0088, 0.0121, 0.0138, 0.0101, 0.0117], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-07 12:01:04,754 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5084, 4.9327, 5.2015, 5.1643, 4.7328, 4.7352, 5.1371, 5.1905], device='cuda:1'), covar=tensor([0.0205, 0.0180, 0.0093, 0.0113, 0.0177, 0.0091, 0.0252, 0.0080], device='cuda:1'), in_proj_covar=tensor([0.0088, 0.0071, 0.0079, 0.0062, 0.0083, 0.0072, 0.0080, 0.0063], device='cuda:1'), out_proj_covar=tensor([1.2821e-04, 9.8500e-05, 9.6220e-05, 8.3629e-05, 1.1422e-04, 1.1217e-04, 1.1416e-04, 7.6547e-05], device='cuda:1') 2023-03-07 12:01:05,564 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-03-07 12:01:25,977 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4332.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 12:01:30,248 INFO [train2.py:809] (1/4) Epoch 2, batch 350, loss[ctc_loss=0.3065, att_loss=0.3623, loss=0.3511, over 17477.00 frames. utt_duration=886.5 frames, utt_pad_proportion=0.07172, over 79.00 utterances.], tot_loss[ctc_loss=0.2798, att_loss=0.3435, loss=0.3308, over 2712779.12 frames. utt_duration=1231 frames, utt_pad_proportion=0.05832, over 8828.38 utterances.], batch size: 79, lr: 4.26e-02, grad_scale: 8.0 2023-03-07 12:01:37,204 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0707, 5.4482, 5.4435, 5.7855, 5.2080, 5.9605, 5.2272, 5.9356], device='cuda:1'), covar=tensor([0.0410, 0.0392, 0.0359, 0.0356, 0.1452, 0.0445, 0.0340, 0.0431], device='cuda:1'), in_proj_covar=tensor([0.0201, 0.0172, 0.0148, 0.0171, 0.0254, 0.0172, 0.0137, 0.0185], 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') 2023-03-07 12:01:52,903 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=4349.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:02:04,404 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.91 vs. limit=2.0 2023-03-07 12:02:07,512 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.97 vs. limit=2.0 2023-03-07 12:02:42,478 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=4380.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 12:02:48,449 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.1502, 6.1842, 5.7543, 6.2377, 5.9999, 5.8079, 5.6742, 5.7878], device='cuda:1'), covar=tensor([0.0737, 0.0728, 0.0512, 0.0434, 0.0538, 0.0788, 0.2301, 0.1582], device='cuda:1'), in_proj_covar=tensor([0.0190, 0.0238, 0.0197, 0.0177, 0.0159, 0.0242, 0.0264, 0.0225], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-07 12:02:49,850 INFO [train2.py:809] (1/4) Epoch 2, batch 400, loss[ctc_loss=0.3041, att_loss=0.3757, loss=0.3614, over 14080.00 frames. utt_duration=1818 frames, utt_pad_proportion=0.04896, over 31.00 utterances.], tot_loss[ctc_loss=0.2799, att_loss=0.3439, loss=0.3311, over 2831408.98 frames. utt_duration=1236 frames, utt_pad_proportion=0.05853, over 9173.32 utterances.], batch size: 31, lr: 4.25e-02, grad_scale: 8.0 2023-03-07 12:02:56,237 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.05 vs. limit=2.0 2023-03-07 12:03:17,595 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.419e+02 5.648e+02 6.669e+02 8.746e+02 3.020e+03, threshold=1.334e+03, percent-clipped=4.0 2023-03-07 12:03:36,987 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3447, 5.1787, 5.1484, 5.1709, 4.5003, 4.8873, 5.1511, 5.0691], device='cuda:1'), covar=tensor([0.0297, 0.0136, 0.0159, 0.0145, 0.0402, 0.0109, 0.0302, 0.0135], device='cuda:1'), in_proj_covar=tensor([0.0092, 0.0074, 0.0083, 0.0065, 0.0088, 0.0075, 0.0083, 0.0066], device='cuda:1'), out_proj_covar=tensor([1.3629e-04, 1.0482e-04, 1.0324e-04, 8.9037e-05, 1.2222e-04, 1.1887e-04, 1.1995e-04, 8.3284e-05], device='cuda:1') 2023-03-07 12:04:10,869 INFO [train2.py:809] (1/4) Epoch 2, batch 450, loss[ctc_loss=0.3151, att_loss=0.3798, loss=0.3668, over 17295.00 frames. utt_duration=1259 frames, utt_pad_proportion=0.01242, over 55.00 utterances.], tot_loss[ctc_loss=0.2804, att_loss=0.3456, loss=0.3325, over 2941034.22 frames. utt_duration=1239 frames, utt_pad_proportion=0.05394, over 9505.06 utterances.], batch size: 55, lr: 4.24e-02, grad_scale: 8.0 2023-03-07 12:04:34,193 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1712, 4.2705, 4.1136, 4.5573, 4.5680, 4.2212, 4.2829, 4.3905], device='cuda:1'), covar=tensor([0.0120, 0.0312, 0.0157, 0.0124, 0.0109, 0.0136, 0.0188, 0.0165], device='cuda:1'), in_proj_covar=tensor([0.0033, 0.0037, 0.0036, 0.0029, 0.0027, 0.0030, 0.0035, 0.0035], device='cuda:1'), out_proj_covar=tensor([4.4732e-05, 4.9155e-05, 5.2124e-05, 3.9448e-05, 3.4989e-05, 4.3758e-05, 4.5093e-05, 4.3981e-05], device='cuda:1') 2023-03-07 12:05:18,982 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([1.6661, 1.8179, 2.2716, 1.9847, 2.3067, 1.8591, 1.7223, 2.9266], device='cuda:1'), covar=tensor([0.0624, 0.0873, 0.0508, 0.0787, 0.0459, 0.0892, 0.0874, 0.0276], device='cuda:1'), in_proj_covar=tensor([0.0059, 0.0053, 0.0044, 0.0065, 0.0050, 0.0063, 0.0057, 0.0061], device='cuda:1'), out_proj_covar=tensor([4.3857e-05, 4.4652e-05, 4.0496e-05, 4.5608e-05, 4.1679e-05, 6.8158e-05, 5.7048e-05, 4.2103e-05], device='cuda:1') 2023-03-07 12:05:30,521 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5295, 4.2159, 4.5940, 4.6634, 4.6284, 3.6983, 4.4530, 3.6682], device='cuda:1'), covar=tensor([0.0103, 0.0094, 0.0154, 0.0076, 0.0066, 0.0147, 0.0110, 0.0265], device='cuda:1'), in_proj_covar=tensor([0.0020, 0.0017, 0.0018, 0.0022, 0.0019, 0.0020, 0.0022, 0.0032], device='cuda:1'), out_proj_covar=tensor([2.6684e-05, 2.7196e-05, 3.0533e-05, 2.7819e-05, 2.5321e-05, 3.1502e-05, 2.8431e-05, 4.3043e-05], device='cuda:1') 2023-03-07 12:05:33,258 INFO [train2.py:809] (1/4) Epoch 2, batch 500, loss[ctc_loss=0.1906, att_loss=0.2734, loss=0.2568, over 15489.00 frames. utt_duration=1722 frames, utt_pad_proportion=0.009488, over 36.00 utterances.], tot_loss[ctc_loss=0.2773, att_loss=0.3434, loss=0.3302, over 3012780.44 frames. utt_duration=1222 frames, utt_pad_proportion=0.06037, over 9874.73 utterances.], batch size: 36, lr: 4.23e-02, grad_scale: 8.0 2023-03-07 12:05:52,573 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4497.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:06:00,028 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 3.132e+02 4.968e+02 5.977e+02 8.490e+02 1.955e+03, threshold=1.195e+03, percent-clipped=9.0 2023-03-07 12:06:17,759 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0780, 5.3544, 5.6277, 5.8546, 5.1743, 5.8848, 5.3943, 5.9348], device='cuda:1'), covar=tensor([0.0391, 0.0369, 0.0343, 0.0288, 0.1710, 0.0486, 0.0380, 0.0501], device='cuda:1'), in_proj_covar=tensor([0.0209, 0.0180, 0.0150, 0.0177, 0.0267, 0.0176, 0.0145, 0.0192], 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') 2023-03-07 12:06:53,454 INFO [train2.py:809] (1/4) Epoch 2, batch 550, loss[ctc_loss=0.2603, att_loss=0.3444, loss=0.3276, over 17269.00 frames. utt_duration=1172 frames, utt_pad_proportion=0.02537, over 59.00 utterances.], tot_loss[ctc_loss=0.2745, att_loss=0.3413, loss=0.3279, over 3064962.75 frames. utt_duration=1224 frames, utt_pad_proportion=0.06241, over 10027.99 utterances.], batch size: 59, lr: 4.22e-02, grad_scale: 8.0 2023-03-07 12:07:22,035 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.7808, 4.3840, 4.1244, 3.9552, 4.5322, 4.3074, 4.2139, 4.1336], device='cuda:1'), covar=tensor([0.0858, 0.0328, 0.0315, 0.0567, 0.0306, 0.0319, 0.0293, 0.0327], device='cuda:1'), in_proj_covar=tensor([0.0153, 0.0120, 0.0089, 0.0093, 0.0129, 0.0143, 0.0104, 0.0119], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-07 12:07:41,706 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1963, 4.6457, 4.9406, 4.9915, 1.8411, 3.1377, 5.0754, 3.8543], device='cuda:1'), covar=tensor([0.2218, 0.0974, 0.0643, 0.1013, 3.5122, 0.5660, 0.0603, 0.8493], device='cuda:1'), in_proj_covar=tensor([0.0115, 0.0067, 0.0094, 0.0104, 0.0275, 0.0152, 0.0102, 0.0097], device='cuda:1'), out_proj_covar=tensor([5.6048e-05, 3.0970e-05, 3.5297e-05, 4.0112e-05, 1.5075e-04, 7.0540e-05, 3.8169e-05, 5.5659e-05], device='cuda:1') 2023-03-07 12:08:08,823 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4582.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:08:13,848 INFO [train2.py:809] (1/4) Epoch 2, batch 600, loss[ctc_loss=0.2324, att_loss=0.3317, loss=0.3118, over 17126.00 frames. utt_duration=1225 frames, utt_pad_proportion=0.01455, over 56.00 utterances.], tot_loss[ctc_loss=0.2744, att_loss=0.3408, loss=0.3275, over 3113895.18 frames. utt_duration=1249 frames, utt_pad_proportion=0.05457, over 9981.97 utterances.], batch size: 56, lr: 4.21e-02, grad_scale: 8.0 2023-03-07 12:08:17,656 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3588, 4.2414, 4.4994, 4.4241, 2.2414, 4.4391, 3.6839, 4.6964], device='cuda:1'), covar=tensor([0.0262, 0.0193, 0.0334, 0.0269, 0.4031, 0.0169, 0.0750, 0.0140], device='cuda:1'), in_proj_covar=tensor([0.0065, 0.0070, 0.0106, 0.0127, 0.0212, 0.0066, 0.0122, 0.0102], device='cuda:1'), out_proj_covar=tensor([5.0746e-05, 5.1896e-05, 7.0569e-05, 8.4347e-05, 1.3908e-04, 4.8139e-05, 7.8693e-05, 6.0927e-05], device='cuda:1') 2023-03-07 12:08:40,958 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.676e+02 4.778e+02 6.448e+02 8.266e+02 2.224e+03, threshold=1.290e+03, percent-clipped=6.0 2023-03-07 12:09:26,948 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=4630.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:09:35,422 INFO [train2.py:809] (1/4) Epoch 2, batch 650, loss[ctc_loss=0.265, att_loss=0.3453, loss=0.3292, over 16874.00 frames. utt_duration=1379 frames, utt_pad_proportion=0.007199, over 49.00 utterances.], tot_loss[ctc_loss=0.2735, att_loss=0.3404, loss=0.327, over 3151286.44 frames. utt_duration=1234 frames, utt_pad_proportion=0.05707, over 10226.39 utterances.], batch size: 49, lr: 4.20e-02, grad_scale: 8.0 2023-03-07 12:09:50,191 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4644.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:10:56,701 INFO [train2.py:809] (1/4) Epoch 2, batch 700, loss[ctc_loss=0.2477, att_loss=0.3283, loss=0.3122, over 16535.00 frames. utt_duration=1471 frames, utt_pad_proportion=0.005931, over 45.00 utterances.], tot_loss[ctc_loss=0.2718, att_loss=0.3393, loss=0.3258, over 3171156.86 frames. utt_duration=1228 frames, utt_pad_proportion=0.06195, over 10338.42 utterances.], batch size: 45, lr: 4.19e-02, grad_scale: 8.0 2023-03-07 12:11:23,313 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.556e+02 4.876e+02 6.372e+02 7.940e+02 1.497e+03, threshold=1.274e+03, percent-clipped=2.0 2023-03-07 12:12:18,133 INFO [train2.py:809] (1/4) Epoch 2, batch 750, loss[ctc_loss=0.2571, att_loss=0.3365, loss=0.3206, over 16613.00 frames. utt_duration=1415 frames, utt_pad_proportion=0.005992, over 47.00 utterances.], tot_loss[ctc_loss=0.2713, att_loss=0.3393, loss=0.3257, over 3201235.25 frames. utt_duration=1206 frames, utt_pad_proportion=0.06414, over 10629.70 utterances.], batch size: 47, lr: 4.18e-02, grad_scale: 8.0 2023-03-07 12:12:53,599 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1743, 3.8727, 3.2695, 1.6826, 3.7290, 4.4105, 4.0156, 2.6825], device='cuda:1'), covar=tensor([0.0258, 0.0248, 0.0650, 0.1856, 0.0367, 0.0079, 0.0323, 0.1192], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0062, 0.0067, 0.0094, 0.0076, 0.0045, 0.0050, 0.0090], device='cuda:1'), out_proj_covar=tensor([5.6064e-05, 5.1272e-05, 6.9002e-05, 8.2807e-05, 6.5758e-05, 3.7456e-05, 5.2758e-05, 8.0074e-05], device='cuda:1') 2023-03-07 12:12:56,058 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.75 vs. limit=2.0 2023-03-07 12:13:38,644 INFO [train2.py:809] (1/4) Epoch 2, batch 800, loss[ctc_loss=0.3147, att_loss=0.3646, loss=0.3547, over 17024.00 frames. utt_duration=689.3 frames, utt_pad_proportion=0.1341, over 99.00 utterances.], tot_loss[ctc_loss=0.2695, att_loss=0.338, loss=0.3243, over 3217793.14 frames. utt_duration=1230 frames, utt_pad_proportion=0.05709, over 10479.10 utterances.], batch size: 99, lr: 4.17e-02, grad_scale: 8.0 2023-03-07 12:13:57,567 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4797.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:14:05,359 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-03-07 12:14:05,889 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 3.406e+02 5.666e+02 6.834e+02 8.456e+02 1.435e+03, threshold=1.367e+03, percent-clipped=1.0 2023-03-07 12:14:50,935 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.66 vs. limit=2.0 2023-03-07 12:14:58,936 INFO [train2.py:809] (1/4) Epoch 2, batch 850, loss[ctc_loss=0.2988, att_loss=0.3626, loss=0.3498, over 17418.00 frames. utt_duration=1011 frames, utt_pad_proportion=0.04711, over 69.00 utterances.], tot_loss[ctc_loss=0.2687, att_loss=0.3374, loss=0.3237, over 3232339.71 frames. utt_duration=1249 frames, utt_pad_proportion=0.05237, over 10363.98 utterances.], batch size: 69, lr: 4.16e-02, grad_scale: 8.0 2023-03-07 12:15:04,187 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.0506, 1.8545, 1.6740, 1.9397, 2.0868, 1.5179, 1.5781, 2.1909], device='cuda:1'), covar=tensor([0.0562, 0.0851, 0.0816, 0.0599, 0.0538, 0.1164, 0.0928, 0.0329], device='cuda:1'), in_proj_covar=tensor([0.0065, 0.0056, 0.0051, 0.0065, 0.0054, 0.0066, 0.0061, 0.0066], device='cuda:1'), out_proj_covar=tensor([4.9647e-05, 4.7340e-05, 4.7563e-05, 4.6941e-05, 4.3941e-05, 7.2467e-05, 6.2465e-05, 4.4898e-05], device='cuda:1') 2023-03-07 12:15:14,890 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=4845.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:15:48,093 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4591, 4.9796, 4.4063, 4.9567, 5.0289, 4.5467, 4.6583, 4.9991], device='cuda:1'), covar=tensor([0.0114, 0.0211, 0.0148, 0.0151, 0.0095, 0.0150, 0.0168, 0.0139], device='cuda:1'), in_proj_covar=tensor([0.0036, 0.0039, 0.0039, 0.0032, 0.0028, 0.0032, 0.0039, 0.0038], device='cuda:1'), out_proj_covar=tensor([5.1064e-05, 5.5711e-05, 6.0923e-05, 4.5520e-05, 3.7418e-05, 4.8351e-05, 5.3392e-05, 5.2992e-05], device='cuda:1') 2023-03-07 12:16:19,355 INFO [train2.py:809] (1/4) Epoch 2, batch 900, loss[ctc_loss=0.2476, att_loss=0.3147, loss=0.3013, over 15957.00 frames. utt_duration=1558 frames, utt_pad_proportion=0.00684, over 41.00 utterances.], tot_loss[ctc_loss=0.267, att_loss=0.337, loss=0.323, over 3237532.48 frames. utt_duration=1259 frames, utt_pad_proportion=0.05043, over 10300.85 utterances.], batch size: 41, lr: 4.15e-02, grad_scale: 8.0 2023-03-07 12:16:46,358 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.888e+02 4.698e+02 5.859e+02 7.531e+02 1.383e+03, threshold=1.172e+03, percent-clipped=2.0 2023-03-07 12:17:40,696 INFO [train2.py:809] (1/4) Epoch 2, batch 950, loss[ctc_loss=0.2887, att_loss=0.3565, loss=0.3429, over 17289.00 frames. utt_duration=1174 frames, utt_pad_proportion=0.02428, over 59.00 utterances.], tot_loss[ctc_loss=0.2682, att_loss=0.338, loss=0.324, over 3242251.36 frames. utt_duration=1233 frames, utt_pad_proportion=0.05876, over 10531.63 utterances.], batch size: 59, lr: 4.14e-02, grad_scale: 8.0 2023-03-07 12:17:55,072 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4944.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:18:55,770 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=9.33 vs. limit=5.0 2023-03-07 12:19:01,003 INFO [train2.py:809] (1/4) Epoch 2, batch 1000, loss[ctc_loss=0.2576, att_loss=0.3472, loss=0.3293, over 16968.00 frames. utt_duration=1359 frames, utt_pad_proportion=0.007509, over 50.00 utterances.], tot_loss[ctc_loss=0.2667, att_loss=0.3375, loss=0.3233, over 3257167.16 frames. utt_duration=1233 frames, utt_pad_proportion=0.05594, over 10577.89 utterances.], batch size: 50, lr: 4.13e-02, grad_scale: 8.0 2023-03-07 12:19:12,003 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=4992.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:19:27,690 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.851e+02 4.989e+02 6.201e+02 7.841e+02 1.460e+03, threshold=1.240e+03, percent-clipped=6.0 2023-03-07 12:20:21,710 INFO [train2.py:809] (1/4) Epoch 2, batch 1050, loss[ctc_loss=0.2585, att_loss=0.3351, loss=0.3198, over 16893.00 frames. utt_duration=1381 frames, utt_pad_proportion=0.006039, over 49.00 utterances.], tot_loss[ctc_loss=0.266, att_loss=0.3372, loss=0.323, over 3245870.06 frames. utt_duration=1217 frames, utt_pad_proportion=0.06377, over 10684.37 utterances.], batch size: 49, lr: 4.12e-02, grad_scale: 8.0 2023-03-07 12:21:42,521 INFO [train2.py:809] (1/4) Epoch 2, batch 1100, loss[ctc_loss=0.3174, att_loss=0.3707, loss=0.3601, over 17330.00 frames. utt_duration=878.8 frames, utt_pad_proportion=0.0807, over 79.00 utterances.], tot_loss[ctc_loss=0.2658, att_loss=0.3368, loss=0.3226, over 3245543.59 frames. utt_duration=1206 frames, utt_pad_proportion=0.06677, over 10775.78 utterances.], batch size: 79, lr: 4.11e-02, grad_scale: 8.0 2023-03-07 12:22:09,101 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.39 vs. limit=5.0 2023-03-07 12:22:09,664 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.609e+02 5.009e+02 6.335e+02 7.458e+02 1.646e+03, threshold=1.267e+03, percent-clipped=3.0 2023-03-07 12:22:45,436 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=2.23 vs. limit=2.0 2023-03-07 12:23:03,369 INFO [train2.py:809] (1/4) Epoch 2, batch 1150, loss[ctc_loss=0.268, att_loss=0.3542, loss=0.3369, over 16461.00 frames. utt_duration=1433 frames, utt_pad_proportion=0.00684, over 46.00 utterances.], tot_loss[ctc_loss=0.2653, att_loss=0.3364, loss=0.3222, over 3256687.00 frames. utt_duration=1214 frames, utt_pad_proportion=0.06284, over 10743.47 utterances.], batch size: 46, lr: 4.10e-02, grad_scale: 8.0 2023-03-07 12:24:24,940 INFO [train2.py:809] (1/4) Epoch 2, batch 1200, loss[ctc_loss=0.2858, att_loss=0.3633, loss=0.3478, over 16767.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.00651, over 48.00 utterances.], tot_loss[ctc_loss=0.266, att_loss=0.3369, loss=0.3227, over 3256161.13 frames. utt_duration=1211 frames, utt_pad_proportion=0.06364, over 10764.43 utterances.], batch size: 48, lr: 4.08e-02, grad_scale: 8.0 2023-03-07 12:24:53,173 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 3.294e+02 5.285e+02 6.236e+02 7.599e+02 1.317e+03, threshold=1.247e+03, percent-clipped=1.0 2023-03-07 12:25:45,338 INFO [train2.py:809] (1/4) Epoch 2, batch 1250, loss[ctc_loss=0.2498, att_loss=0.3246, loss=0.3096, over 15955.00 frames. utt_duration=1558 frames, utt_pad_proportion=0.007073, over 41.00 utterances.], tot_loss[ctc_loss=0.2641, att_loss=0.3357, loss=0.3214, over 3262768.16 frames. utt_duration=1222 frames, utt_pad_proportion=0.06108, over 10693.27 utterances.], batch size: 41, lr: 4.07e-02, grad_scale: 8.0 2023-03-07 12:26:09,688 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1427, 2.7023, 3.7497, 4.0061, 4.4703, 4.5449, 2.6883, 2.2242], device='cuda:1'), covar=tensor([0.0252, 0.1026, 0.0515, 0.0621, 0.0195, 0.0123, 0.1554, 0.1895], device='cuda:1'), in_proj_covar=tensor([0.0071, 0.0087, 0.0087, 0.0061, 0.0050, 0.0060, 0.0106, 0.0102], device='cuda:1'), out_proj_covar=tensor([6.0320e-05, 7.8596e-05, 7.6081e-05, 6.5245e-05, 5.0624e-05, 4.9220e-05, 9.8586e-05, 8.9588e-05], device='cuda:1') 2023-03-07 12:26:14,155 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.95 vs. limit=2.0 2023-03-07 12:26:16,993 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.98 vs. limit=2.0 2023-03-07 12:26:42,134 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5847, 3.9117, 4.5343, 4.1380, 2.0751, 4.4356, 3.6558, 4.6295], device='cuda:1'), covar=tensor([0.0212, 0.0220, 0.0260, 0.0475, 0.3886, 0.0278, 0.0782, 0.0170], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0070, 0.0113, 0.0127, 0.0213, 0.0069, 0.0126, 0.0107], device='cuda:1'), out_proj_covar=tensor([5.5655e-05, 5.4897e-05, 8.0065e-05, 8.7293e-05, 1.4243e-04, 5.2473e-05, 8.4921e-05, 6.8906e-05], device='cuda:1') 2023-03-07 12:27:05,711 INFO [train2.py:809] (1/4) Epoch 2, batch 1300, loss[ctc_loss=0.2454, att_loss=0.3111, loss=0.298, over 15373.00 frames. utt_duration=1758 frames, utt_pad_proportion=0.01114, over 35.00 utterances.], tot_loss[ctc_loss=0.2629, att_loss=0.3346, loss=0.3203, over 3255184.69 frames. utt_duration=1250 frames, utt_pad_proportion=0.05628, over 10430.20 utterances.], batch size: 35, lr: 4.06e-02, grad_scale: 8.0 2023-03-07 12:27:15,687 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2623, 2.6798, 3.5829, 3.7279, 2.8846, 3.0854, 2.1772, 3.7610], device='cuda:1'), covar=tensor([0.0318, 0.0433, 0.0360, 0.0264, 0.0293, 0.0594, 0.0974, 0.0134], device='cuda:1'), in_proj_covar=tensor([0.0043, 0.0032, 0.0059, 0.0051, 0.0038, 0.0067, 0.0061, 0.0033], device='cuda:1'), out_proj_covar=tensor([4.6203e-05, 4.4849e-05, 7.8680e-05, 5.5507e-05, 4.7889e-05, 8.1168e-05, 6.6630e-05, 4.0045e-05], device='cuda:1') 2023-03-07 12:27:20,452 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2878, 2.0858, 3.7294, 4.1775, 4.3045, 4.5782, 2.9403, 2.0577], device='cuda:1'), covar=tensor([0.0172, 0.1192, 0.0465, 0.0258, 0.0163, 0.0097, 0.1143, 0.1606], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0088, 0.0086, 0.0059, 0.0049, 0.0061, 0.0103, 0.0101], device='cuda:1'), out_proj_covar=tensor([5.9344e-05, 7.9704e-05, 7.5560e-05, 6.3603e-05, 5.0369e-05, 4.9707e-05, 9.6673e-05, 8.9033e-05], device='cuda:1') 2023-03-07 12:27:33,502 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 3.402e+02 4.574e+02 5.700e+02 7.129e+02 3.450e+03, threshold=1.140e+03, percent-clipped=2.0 2023-03-07 12:28:26,301 INFO [train2.py:809] (1/4) Epoch 2, batch 1350, loss[ctc_loss=0.2351, att_loss=0.3207, loss=0.3036, over 15864.00 frames. utt_duration=1628 frames, utt_pad_proportion=0.01064, over 39.00 utterances.], tot_loss[ctc_loss=0.2612, att_loss=0.3334, loss=0.3189, over 3257992.11 frames. utt_duration=1271 frames, utt_pad_proportion=0.05247, over 10265.18 utterances.], batch size: 39, lr: 4.05e-02, grad_scale: 8.0 2023-03-07 12:28:28,762 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=2.16 vs. limit=2.0 2023-03-07 12:28:49,441 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4911, 5.0033, 4.6253, 4.9116, 5.0526, 4.8514, 4.6100, 4.8379], device='cuda:1'), covar=tensor([0.0107, 0.0167, 0.0108, 0.0096, 0.0090, 0.0095, 0.0195, 0.0114], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0040, 0.0041, 0.0032, 0.0028, 0.0033, 0.0044, 0.0040], device='cuda:1'), out_proj_covar=tensor([5.8629e-05, 6.2268e-05, 6.9677e-05, 5.0008e-05, 4.1963e-05, 5.3634e-05, 6.5643e-05, 6.1113e-05], device='cuda:1') 2023-03-07 12:29:47,503 INFO [train2.py:809] (1/4) Epoch 2, batch 1400, loss[ctc_loss=0.2132, att_loss=0.2863, loss=0.2717, over 15494.00 frames. utt_duration=1723 frames, utt_pad_proportion=0.009328, over 36.00 utterances.], tot_loss[ctc_loss=0.2592, att_loss=0.3329, loss=0.3181, over 3269417.15 frames. utt_duration=1268 frames, utt_pad_proportion=0.05033, over 10328.52 utterances.], batch size: 36, lr: 4.04e-02, grad_scale: 8.0 2023-03-07 12:30:15,419 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.881e+02 4.556e+02 5.757e+02 7.116e+02 2.241e+03, threshold=1.151e+03, percent-clipped=2.0 2023-03-07 12:31:08,630 INFO [train2.py:809] (1/4) Epoch 2, batch 1450, loss[ctc_loss=0.2795, att_loss=0.3609, loss=0.3446, over 17036.00 frames. utt_duration=1312 frames, utt_pad_proportion=0.00902, over 52.00 utterances.], tot_loss[ctc_loss=0.2576, att_loss=0.3327, loss=0.3177, over 3280353.32 frames. utt_duration=1271 frames, utt_pad_proportion=0.04712, over 10332.20 utterances.], batch size: 52, lr: 4.03e-02, grad_scale: 8.0 2023-03-07 12:32:29,332 INFO [train2.py:809] (1/4) Epoch 2, batch 1500, loss[ctc_loss=0.3034, att_loss=0.3643, loss=0.3521, over 17470.00 frames. utt_duration=1014 frames, utt_pad_proportion=0.04237, over 69.00 utterances.], tot_loss[ctc_loss=0.2567, att_loss=0.3324, loss=0.3172, over 3277747.00 frames. utt_duration=1263 frames, utt_pad_proportion=0.05014, over 10392.67 utterances.], batch size: 69, lr: 4.02e-02, grad_scale: 8.0 2023-03-07 12:32:34,936 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.04 vs. limit=2.0 2023-03-07 12:32:57,037 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.303e+02 4.697e+02 5.525e+02 6.949e+02 3.924e+03, threshold=1.105e+03, percent-clipped=4.0 2023-03-07 12:33:49,181 INFO [train2.py:809] (1/4) Epoch 2, batch 1550, loss[ctc_loss=0.2667, att_loss=0.3489, loss=0.3324, over 16617.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.005812, over 47.00 utterances.], tot_loss[ctc_loss=0.2565, att_loss=0.3325, loss=0.3173, over 3282633.02 frames. utt_duration=1250 frames, utt_pad_proportion=0.05143, over 10519.79 utterances.], batch size: 47, lr: 4.01e-02, grad_scale: 8.0 2023-03-07 12:34:53,040 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4583, 5.0936, 5.2284, 5.2836, 4.6269, 5.0188, 5.3056, 5.2593], device='cuda:1'), covar=tensor([0.0280, 0.0182, 0.0150, 0.0128, 0.0293, 0.0099, 0.0297, 0.0102], device='cuda:1'), in_proj_covar=tensor([0.0104, 0.0088, 0.0095, 0.0072, 0.0101, 0.0085, 0.0098, 0.0076], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-03-07 12:35:10,351 INFO [train2.py:809] (1/4) Epoch 2, batch 1600, loss[ctc_loss=0.2973, att_loss=0.3508, loss=0.3401, over 14046.00 frames. utt_duration=386.4 frames, utt_pad_proportion=0.3268, over 146.00 utterances.], tot_loss[ctc_loss=0.2536, att_loss=0.3303, loss=0.3149, over 3278754.81 frames. utt_duration=1241 frames, utt_pad_proportion=0.05439, over 10579.90 utterances.], batch size: 146, lr: 4.00e-02, grad_scale: 8.0 2023-03-07 12:35:19,279 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3107, 4.4581, 5.2609, 4.2356, 3.6964, 4.1133, 4.8413, 4.6151], device='cuda:1'), covar=tensor([0.0488, 0.1198, 0.0282, 0.1859, 0.3486, 0.1811, 0.0528, 0.0528], device='cuda:1'), in_proj_covar=tensor([0.0076, 0.0162, 0.0092, 0.0188, 0.0211, 0.0119, 0.0074, 0.0079], device='cuda:1'), out_proj_covar=tensor([4.8046e-05, 9.3572e-05, 5.2279e-05, 1.2427e-04, 1.3276e-04, 8.3330e-05, 4.7734e-05, 5.3086e-05], device='cuda:1') 2023-03-07 12:35:38,630 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 3.176e+02 4.466e+02 5.617e+02 6.515e+02 1.895e+03, threshold=1.123e+03, percent-clipped=5.0 2023-03-07 12:36:30,216 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7250, 4.5572, 4.5962, 4.7446, 5.0328, 4.6522, 4.2423, 2.5545], device='cuda:1'), covar=tensor([0.0157, 0.0184, 0.0150, 0.0083, 0.0440, 0.0124, 0.0351, 0.2682], device='cuda:1'), in_proj_covar=tensor([0.0132, 0.0090, 0.0082, 0.0092, 0.0138, 0.0109, 0.0075, 0.0208], device='cuda:1'), out_proj_covar=tensor([9.1329e-05, 5.5799e-05, 5.6705e-05, 5.4797e-05, 1.0702e-04, 6.9103e-05, 5.3351e-05, 1.4087e-04], device='cuda:1') 2023-03-07 12:36:31,275 INFO [train2.py:809] (1/4) Epoch 2, batch 1650, loss[ctc_loss=0.2606, att_loss=0.3404, loss=0.3244, over 17353.00 frames. utt_duration=1178 frames, utt_pad_proportion=0.02108, over 59.00 utterances.], tot_loss[ctc_loss=0.2536, att_loss=0.3307, loss=0.3152, over 3282001.67 frames. utt_duration=1261 frames, utt_pad_proportion=0.0484, over 10425.29 utterances.], batch size: 59, lr: 3.99e-02, grad_scale: 8.0 2023-03-07 12:36:41,255 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3578, 5.0450, 5.2746, 5.2025, 2.0460, 4.0400, 5.3826, 4.1175], device='cuda:1'), covar=tensor([0.1394, 0.0506, 0.0348, 0.0673, 1.9205, 0.2353, 0.0339, 0.6096], device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0097, 0.0125, 0.0133, 0.0346, 0.0214, 0.0133, 0.0148], device='cuda:1'), out_proj_covar=tensor([8.2782e-05, 4.4487e-05, 5.0829e-05, 5.7525e-05, 1.7574e-04, 1.0175e-04, 5.4968e-05, 8.3779e-05], device='cuda:1') 2023-03-07 12:36:55,825 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5650.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 12:37:51,243 INFO [train2.py:809] (1/4) Epoch 2, batch 1700, loss[ctc_loss=0.2371, att_loss=0.3309, loss=0.3121, over 16682.00 frames. utt_duration=1452 frames, utt_pad_proportion=0.006011, over 46.00 utterances.], tot_loss[ctc_loss=0.2543, att_loss=0.3309, loss=0.3155, over 3278247.68 frames. utt_duration=1238 frames, utt_pad_proportion=0.05564, over 10608.27 utterances.], batch size: 46, lr: 3.98e-02, grad_scale: 8.0 2023-03-07 12:37:54,969 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9440, 5.1769, 5.5443, 5.7270, 5.0444, 5.8927, 5.3183, 5.7498], device='cuda:1'), covar=tensor([0.0391, 0.0459, 0.0340, 0.0355, 0.1806, 0.0448, 0.0378, 0.0521], device='cuda:1'), in_proj_covar=tensor([0.0254, 0.0201, 0.0174, 0.0198, 0.0320, 0.0193, 0.0165, 0.0221], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-03-07 12:37:55,162 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.8139, 1.0117, 2.9324, 3.5704, 2.8893, 3.3282, 2.6909, 1.5597], device='cuda:1'), covar=tensor([0.0629, 0.1538, 0.0405, 0.0304, 0.0421, 0.0285, 0.0742, 0.1850], device='cuda:1'), in_proj_covar=tensor([0.0078, 0.0066, 0.0059, 0.0082, 0.0085, 0.0061, 0.0069, 0.0100], device='cuda:1'), out_proj_covar=tensor([5.1954e-05, 5.1666e-05, 4.6406e-05, 4.9132e-05, 5.1238e-05, 4.3848e-05, 4.8962e-05, 7.8969e-05], device='cuda:1') 2023-03-07 12:38:18,539 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.296e+02 4.646e+02 5.757e+02 7.379e+02 2.321e+03, threshold=1.151e+03, percent-clipped=8.0 2023-03-07 12:38:34,143 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5711.0, num_to_drop=1, layers_to_drop={2} 2023-03-07 12:39:10,814 INFO [train2.py:809] (1/4) Epoch 2, batch 1750, loss[ctc_loss=0.2284, att_loss=0.2973, loss=0.2835, over 15790.00 frames. utt_duration=1663 frames, utt_pad_proportion=0.006367, over 38.00 utterances.], tot_loss[ctc_loss=0.2541, att_loss=0.3307, loss=0.3154, over 3284162.42 frames. utt_duration=1253 frames, utt_pad_proportion=0.04976, over 10494.82 utterances.], batch size: 38, lr: 3.97e-02, grad_scale: 8.0 2023-03-07 12:39:44,217 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5755.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:40:14,638 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.7949, 5.1055, 5.3112, 5.7044, 4.7861, 5.6999, 5.1162, 5.6672], device='cuda:1'), covar=tensor([0.0467, 0.0477, 0.0394, 0.0429, 0.2028, 0.0574, 0.0375, 0.0577], device='cuda:1'), in_proj_covar=tensor([0.0265, 0.0206, 0.0177, 0.0202, 0.0325, 0.0199, 0.0170, 0.0230], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-03-07 12:40:31,799 INFO [train2.py:809] (1/4) Epoch 2, batch 1800, loss[ctc_loss=0.2645, att_loss=0.3353, loss=0.3212, over 16898.00 frames. utt_duration=1381 frames, utt_pad_proportion=0.00668, over 49.00 utterances.], tot_loss[ctc_loss=0.2522, att_loss=0.3298, loss=0.3143, over 3280920.31 frames. utt_duration=1242 frames, utt_pad_proportion=0.05428, over 10580.76 utterances.], batch size: 49, lr: 3.96e-02, grad_scale: 8.0 2023-03-07 12:40:59,739 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.935e+02 4.735e+02 5.722e+02 7.362e+02 1.615e+03, threshold=1.144e+03, percent-clipped=5.0 2023-03-07 12:41:23,386 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5816.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:41:44,239 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.63 vs. limit=5.0 2023-03-07 12:41:53,266 INFO [train2.py:809] (1/4) Epoch 2, batch 1850, loss[ctc_loss=0.2624, att_loss=0.3444, loss=0.328, over 17401.00 frames. utt_duration=1106 frames, utt_pad_proportion=0.03297, over 63.00 utterances.], tot_loss[ctc_loss=0.2496, att_loss=0.328, loss=0.3123, over 3285405.33 frames. utt_duration=1261 frames, utt_pad_proportion=0.0483, over 10431.17 utterances.], batch size: 63, lr: 3.95e-02, grad_scale: 8.0 2023-03-07 12:42:30,745 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5858.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:43:13,668 INFO [train2.py:809] (1/4) Epoch 2, batch 1900, loss[ctc_loss=0.1796, att_loss=0.2816, loss=0.2612, over 14548.00 frames. utt_duration=1820 frames, utt_pad_proportion=0.03443, over 32.00 utterances.], tot_loss[ctc_loss=0.248, att_loss=0.327, loss=0.3112, over 3280274.56 frames. utt_duration=1267 frames, utt_pad_proportion=0.0473, over 10367.73 utterances.], batch size: 32, lr: 3.95e-02, grad_scale: 8.0 2023-03-07 12:43:28,922 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2294, 4.4290, 4.0802, 4.4226, 4.4200, 4.3298, 4.2130, 4.2543], device='cuda:1'), covar=tensor([0.0131, 0.0210, 0.0145, 0.0124, 0.0116, 0.0108, 0.0254, 0.0190], device='cuda:1'), in_proj_covar=tensor([0.0036, 0.0041, 0.0040, 0.0032, 0.0027, 0.0031, 0.0045, 0.0039], device='cuda:1'), out_proj_covar=tensor([6.1652e-05, 6.8264e-05, 7.3765e-05, 5.5277e-05, 4.3249e-05, 5.5405e-05, 7.3933e-05, 6.5690e-05], device='cuda:1') 2023-03-07 12:43:41,191 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.885e+02 5.042e+02 5.993e+02 7.245e+02 1.405e+03, threshold=1.199e+03, percent-clipped=1.0 2023-03-07 12:44:09,803 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5919.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:44:35,116 INFO [train2.py:809] (1/4) Epoch 2, batch 1950, loss[ctc_loss=0.2768, att_loss=0.3545, loss=0.3389, over 17330.00 frames. utt_duration=1177 frames, utt_pad_proportion=0.02199, over 59.00 utterances.], tot_loss[ctc_loss=0.2476, att_loss=0.3269, loss=0.311, over 3281336.71 frames. utt_duration=1257 frames, utt_pad_proportion=0.04927, over 10451.11 utterances.], batch size: 59, lr: 3.94e-02, grad_scale: 8.0 2023-03-07 12:45:21,416 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4853, 4.0693, 5.0308, 3.8059, 3.7559, 4.0403, 4.7168, 4.5355], device='cuda:1'), covar=tensor([0.0278, 0.1164, 0.0299, 0.2202, 0.3044, 0.1570, 0.0363, 0.0462], device='cuda:1'), in_proj_covar=tensor([0.0079, 0.0167, 0.0099, 0.0197, 0.0230, 0.0124, 0.0076, 0.0083], device='cuda:1'), out_proj_covar=tensor([5.4108e-05, 1.0049e-04, 5.8892e-05, 1.3405e-04, 1.4707e-04, 9.1321e-05, 5.4008e-05, 5.7962e-05], device='cuda:1') 2023-03-07 12:45:55,278 INFO [train2.py:809] (1/4) Epoch 2, batch 2000, loss[ctc_loss=0.2142, att_loss=0.3204, loss=0.2992, over 17064.00 frames. utt_duration=1289 frames, utt_pad_proportion=0.008963, over 53.00 utterances.], tot_loss[ctc_loss=0.2467, att_loss=0.3262, loss=0.3103, over 3276978.16 frames. utt_duration=1286 frames, utt_pad_proportion=0.04336, over 10202.33 utterances.], batch size: 53, lr: 3.93e-02, grad_scale: 16.0 2023-03-07 12:46:04,112 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5990.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:46:26,499 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 3.443e+02 4.806e+02 6.010e+02 7.878e+02 2.247e+03, threshold=1.202e+03, percent-clipped=5.0 2023-03-07 12:46:33,504 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6006.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 12:47:19,314 INFO [train2.py:809] (1/4) Epoch 2, batch 2050, loss[ctc_loss=0.182, att_loss=0.2586, loss=0.2433, over 15366.00 frames. utt_duration=1757 frames, utt_pad_proportion=0.01159, over 35.00 utterances.], tot_loss[ctc_loss=0.2493, att_loss=0.3281, loss=0.3124, over 3276890.29 frames. utt_duration=1264 frames, utt_pad_proportion=0.04916, over 10381.04 utterances.], batch size: 35, lr: 3.92e-02, grad_scale: 8.0 2023-03-07 12:47:33,231 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6043.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:47:41,226 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.97 vs. limit=2.0 2023-03-07 12:47:45,331 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6051.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:47:58,300 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9992, 5.2498, 5.5690, 5.8727, 5.0263, 5.9154, 5.1948, 5.9200], device='cuda:1'), covar=tensor([0.0484, 0.0399, 0.0346, 0.0343, 0.2326, 0.0510, 0.0368, 0.0456], device='cuda:1'), in_proj_covar=tensor([0.0267, 0.0212, 0.0183, 0.0208, 0.0332, 0.0198, 0.0165, 0.0231], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-03-07 12:48:40,079 INFO [train2.py:809] (1/4) Epoch 2, batch 2100, loss[ctc_loss=0.2556, att_loss=0.3384, loss=0.3218, over 16477.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.00682, over 46.00 utterances.], tot_loss[ctc_loss=0.2487, att_loss=0.328, loss=0.3121, over 3281088.02 frames. utt_duration=1271 frames, utt_pad_proportion=0.04556, over 10342.09 utterances.], batch size: 46, lr: 3.91e-02, grad_scale: 8.0 2023-03-07 12:49:08,983 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 3.112e+02 4.379e+02 5.307e+02 7.391e+02 2.758e+03, threshold=1.061e+03, percent-clipped=4.0 2023-03-07 12:49:10,922 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6104.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:49:23,171 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6111.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:50:01,356 INFO [train2.py:809] (1/4) Epoch 2, batch 2150, loss[ctc_loss=0.2527, att_loss=0.3451, loss=0.3266, over 17061.00 frames. utt_duration=1289 frames, utt_pad_proportion=0.00836, over 53.00 utterances.], tot_loss[ctc_loss=0.2484, att_loss=0.3278, loss=0.3119, over 3279560.05 frames. utt_duration=1249 frames, utt_pad_proportion=0.05204, over 10512.28 utterances.], batch size: 53, lr: 3.90e-02, grad_scale: 8.0 2023-03-07 12:50:01,752 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6135.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:50:11,648 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([1.4160, 1.3343, 2.0850, 1.0168, 1.8058, 2.2369, 1.9417, 2.3924], device='cuda:1'), covar=tensor([0.0919, 0.0871, 0.0527, 0.0916, 0.0479, 0.0555, 0.0684, 0.0452], device='cuda:1'), in_proj_covar=tensor([0.0077, 0.0062, 0.0060, 0.0068, 0.0059, 0.0070, 0.0064, 0.0086], device='cuda:1'), out_proj_covar=tensor([4.9772e-05, 4.8049e-05, 4.8809e-05, 5.4589e-05, 3.7963e-05, 6.3787e-05, 5.4315e-05, 4.5464e-05], device='cuda:1') 2023-03-07 12:50:57,746 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6170.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:51:21,405 INFO [train2.py:809] (1/4) Epoch 2, batch 2200, loss[ctc_loss=0.2383, att_loss=0.3182, loss=0.3022, over 16384.00 frames. utt_duration=1491 frames, utt_pad_proportion=0.007863, over 44.00 utterances.], tot_loss[ctc_loss=0.251, att_loss=0.3297, loss=0.3139, over 3276516.69 frames. utt_duration=1235 frames, utt_pad_proportion=0.05627, over 10628.20 utterances.], batch size: 44, lr: 3.89e-02, grad_scale: 8.0 2023-03-07 12:51:39,652 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6196.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:51:50,201 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 3.181e+02 4.728e+02 5.597e+02 7.112e+02 1.691e+03, threshold=1.119e+03, percent-clipped=4.0 2023-03-07 12:51:50,636 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6203.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:52:04,331 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6211.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:52:08,959 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6214.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:52:19,934 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.4654, 1.7034, 2.3635, 2.7433, 2.4851, 2.0598, 2.4583, 1.2610], device='cuda:1'), covar=tensor([0.0459, 0.1055, 0.0385, 0.0497, 0.0476, 0.0538, 0.0625, 0.2098], device='cuda:1'), in_proj_covar=tensor([0.0078, 0.0060, 0.0051, 0.0087, 0.0084, 0.0069, 0.0071, 0.0099], device='cuda:1'), out_proj_covar=tensor([4.9658e-05, 4.5536e-05, 4.1276e-05, 4.8986e-05, 5.1649e-05, 4.6638e-05, 4.6994e-05, 7.4926e-05], device='cuda:1') 2023-03-07 12:52:22,277 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.80 vs. limit=2.0 2023-03-07 12:52:35,466 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6231.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:52:42,164 INFO [train2.py:809] (1/4) Epoch 2, batch 2250, loss[ctc_loss=0.2381, att_loss=0.3396, loss=0.3193, over 16775.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.006067, over 48.00 utterances.], tot_loss[ctc_loss=0.2489, att_loss=0.3288, loss=0.3128, over 3278611.41 frames. utt_duration=1243 frames, utt_pad_proportion=0.05457, over 10565.00 utterances.], batch size: 48, lr: 3.88e-02, grad_scale: 8.0 2023-03-07 12:53:22,487 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.87 vs. limit=2.0 2023-03-07 12:53:29,789 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6264.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:53:32,829 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6266.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:53:42,108 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6272.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:54:02,557 INFO [train2.py:809] (1/4) Epoch 2, batch 2300, loss[ctc_loss=0.2766, att_loss=0.3444, loss=0.3308, over 16670.00 frames. utt_duration=675 frames, utt_pad_proportion=0.1477, over 99.00 utterances.], tot_loss[ctc_loss=0.2491, att_loss=0.3284, loss=0.3125, over 3273535.96 frames. utt_duration=1231 frames, utt_pad_proportion=0.05832, over 10648.20 utterances.], batch size: 99, lr: 3.87e-02, grad_scale: 8.0 2023-03-07 12:54:30,724 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.781e+02 4.301e+02 5.622e+02 7.152e+02 1.270e+03, threshold=1.124e+03, percent-clipped=3.0 2023-03-07 12:54:32,580 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9278, 4.6530, 4.6825, 4.5553, 4.7486, 3.6147, 4.5158, 2.4915], device='cuda:1'), covar=tensor([0.0108, 0.0094, 0.0276, 0.0133, 0.0103, 0.0218, 0.0131, 0.0908], device='cuda:1'), in_proj_covar=tensor([0.0027, 0.0025, 0.0026, 0.0033, 0.0027, 0.0029, 0.0032, 0.0058], device='cuda:1'), out_proj_covar=tensor([4.9765e-05, 5.3043e-05, 6.1627e-05, 5.9114e-05, 5.1158e-05, 6.6309e-05, 5.7479e-05, 1.0246e-04], device='cuda:1') 2023-03-07 12:54:36,242 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6306.0, num_to_drop=1, layers_to_drop={2} 2023-03-07 12:54:55,604 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5782, 4.2067, 4.9347, 4.6199, 2.1836, 4.7241, 3.7754, 5.2126], device='cuda:1'), covar=tensor([0.0168, 0.0305, 0.0374, 0.0281, 0.4065, 0.0148, 0.0673, 0.0107], device='cuda:1'), in_proj_covar=tensor([0.0069, 0.0069, 0.0134, 0.0119, 0.0206, 0.0074, 0.0132, 0.0113], device='cuda:1'), out_proj_covar=tensor([5.9095e-05, 5.8796e-05, 1.0100e-04, 8.6174e-05, 1.4145e-04, 6.0206e-05, 9.4870e-05, 7.9561e-05], device='cuda:1') 2023-03-07 12:55:09,185 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6327.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:55:21,934 INFO [train2.py:809] (1/4) Epoch 2, batch 2350, loss[ctc_loss=0.2947, att_loss=0.3582, loss=0.3455, over 17316.00 frames. utt_duration=1175 frames, utt_pad_proportion=0.02382, over 59.00 utterances.], tot_loss[ctc_loss=0.2474, att_loss=0.3271, loss=0.3112, over 3268676.33 frames. utt_duration=1256 frames, utt_pad_proportion=0.05257, over 10425.89 utterances.], batch size: 59, lr: 3.86e-02, grad_scale: 8.0 2023-03-07 12:55:39,730 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6346.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:55:52,499 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6354.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 12:56:41,290 INFO [train2.py:809] (1/4) Epoch 2, batch 2400, loss[ctc_loss=0.2123, att_loss=0.2989, loss=0.2816, over 16428.00 frames. utt_duration=1495 frames, utt_pad_proportion=0.006105, over 44.00 utterances.], tot_loss[ctc_loss=0.2483, att_loss=0.3277, loss=0.3118, over 3274711.24 frames. utt_duration=1236 frames, utt_pad_proportion=0.0563, over 10612.19 utterances.], batch size: 44, lr: 3.85e-02, grad_scale: 8.0 2023-03-07 12:57:03,928 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6399.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:57:10,898 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 3.496e+02 4.794e+02 5.680e+02 6.939e+02 1.517e+03, threshold=1.136e+03, percent-clipped=5.0 2023-03-07 12:57:24,376 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6411.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:57:59,327 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.96 vs. limit=5.0 2023-03-07 12:58:03,303 INFO [train2.py:809] (1/4) Epoch 2, batch 2450, loss[ctc_loss=0.2091, att_loss=0.2946, loss=0.2775, over 15892.00 frames. utt_duration=1631 frames, utt_pad_proportion=0.008833, over 39.00 utterances.], tot_loss[ctc_loss=0.2445, att_loss=0.3253, loss=0.3092, over 3268516.03 frames. utt_duration=1266 frames, utt_pad_proportion=0.05026, over 10339.18 utterances.], batch size: 39, lr: 3.84e-02, grad_scale: 8.0 2023-03-07 12:58:42,225 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6459.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:58:47,195 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6462.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:58:54,680 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6467.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:59:23,404 INFO [train2.py:809] (1/4) Epoch 2, batch 2500, loss[ctc_loss=0.2375, att_loss=0.3226, loss=0.3056, over 16386.00 frames. utt_duration=1491 frames, utt_pad_proportion=0.008553, over 44.00 utterances.], tot_loss[ctc_loss=0.2442, att_loss=0.325, loss=0.3089, over 3270333.54 frames. utt_duration=1265 frames, utt_pad_proportion=0.0493, over 10352.21 utterances.], batch size: 44, lr: 3.83e-02, grad_scale: 8.0 2023-03-07 12:59:33,173 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6491.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 12:59:51,987 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.651e+02 4.550e+02 5.536e+02 6.688e+02 2.770e+03, threshold=1.107e+03, percent-clipped=3.0 2023-03-07 12:59:59,507 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.17 vs. limit=5.0 2023-03-07 13:00:09,863 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6514.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:00:24,400 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6523.0, num_to_drop=1, layers_to_drop={2} 2023-03-07 13:00:29,496 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6526.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:00:32,881 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6528.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:00:44,103 INFO [train2.py:809] (1/4) Epoch 2, batch 2550, loss[ctc_loss=0.1979, att_loss=0.2856, loss=0.2681, over 15870.00 frames. utt_duration=1629 frames, utt_pad_proportion=0.01023, over 39.00 utterances.], tot_loss[ctc_loss=0.2435, att_loss=0.3242, loss=0.3081, over 3268009.17 frames. utt_duration=1258 frames, utt_pad_proportion=0.05312, over 10407.28 utterances.], batch size: 39, lr: 3.82e-02, grad_scale: 8.0 2023-03-07 13:01:23,802 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6559.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:01:28,550 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6562.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:01:36,567 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6567.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:02:05,600 INFO [train2.py:809] (1/4) Epoch 2, batch 2600, loss[ctc_loss=0.2077, att_loss=0.3099, loss=0.2895, over 16531.00 frames. utt_duration=1471 frames, utt_pad_proportion=0.006647, over 45.00 utterances.], tot_loss[ctc_loss=0.2417, att_loss=0.3234, loss=0.3071, over 3264086.09 frames. utt_duration=1257 frames, utt_pad_proportion=0.05459, over 10402.80 utterances.], batch size: 45, lr: 3.81e-02, grad_scale: 8.0 2023-03-07 13:02:32,556 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.5080, 5.6972, 5.2110, 5.8140, 5.2567, 5.2548, 5.0084, 5.0996], device='cuda:1'), covar=tensor([0.1307, 0.0752, 0.0867, 0.0517, 0.0631, 0.1122, 0.2320, 0.1664], device='cuda:1'), in_proj_covar=tensor([0.0218, 0.0261, 0.0225, 0.0189, 0.0167, 0.0265, 0.0282, 0.0266], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-07 13:02:35,428 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.190e+02 4.654e+02 5.537e+02 6.841e+02 1.492e+03, threshold=1.107e+03, percent-clipped=3.0 2023-03-07 13:03:05,299 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6622.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:03:10,854 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2652, 4.8992, 5.0659, 4.8673, 2.0362, 3.1265, 5.1926, 3.9430], device='cuda:1'), covar=tensor([0.1159, 0.0333, 0.0204, 0.0603, 1.5694, 0.3044, 0.0234, 0.3625], device='cuda:1'), in_proj_covar=tensor([0.0204, 0.0111, 0.0138, 0.0158, 0.0384, 0.0243, 0.0141, 0.0179], device='cuda:1'), out_proj_covar=tensor([1.0504e-04, 5.6707e-05, 6.4950e-05, 7.4103e-05, 1.8706e-04, 1.1813e-04, 6.4942e-05, 9.9459e-05], device='cuda:1') 2023-03-07 13:03:16,843 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6629.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:03:16,979 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1295, 4.8945, 5.0667, 4.8684, 1.8050, 3.2645, 5.0547, 3.9146], device='cuda:1'), covar=tensor([0.1200, 0.0255, 0.0149, 0.0536, 1.5686, 0.2365, 0.0219, 0.3573], device='cuda:1'), in_proj_covar=tensor([0.0206, 0.0112, 0.0139, 0.0159, 0.0387, 0.0245, 0.0142, 0.0180], device='cuda:1'), out_proj_covar=tensor([1.0608e-04, 5.6973e-05, 6.5276e-05, 7.4765e-05, 1.8830e-04, 1.1880e-04, 6.5316e-05, 1.0014e-04], device='cuda:1') 2023-03-07 13:03:26,463 INFO [train2.py:809] (1/4) Epoch 2, batch 2650, loss[ctc_loss=0.2012, att_loss=0.2966, loss=0.2775, over 16278.00 frames. utt_duration=1515 frames, utt_pad_proportion=0.006919, over 43.00 utterances.], tot_loss[ctc_loss=0.2402, att_loss=0.3227, loss=0.3062, over 3266638.97 frames. utt_duration=1256 frames, utt_pad_proportion=0.05413, over 10416.13 utterances.], batch size: 43, lr: 3.80e-02, grad_scale: 8.0 2023-03-07 13:03:44,183 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4005, 4.9411, 4.9715, 5.0434, 4.4193, 4.7187, 5.2441, 4.9730], device='cuda:1'), covar=tensor([0.0281, 0.0216, 0.0165, 0.0126, 0.0313, 0.0136, 0.0231, 0.0136], device='cuda:1'), in_proj_covar=tensor([0.0108, 0.0092, 0.0105, 0.0072, 0.0106, 0.0085, 0.0100, 0.0082], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-03-07 13:03:44,221 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6646.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:04:30,946 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.91 vs. limit=2.0 2023-03-07 13:04:45,712 INFO [train2.py:809] (1/4) Epoch 2, batch 2700, loss[ctc_loss=0.2455, att_loss=0.3345, loss=0.3167, over 17398.00 frames. utt_duration=1106 frames, utt_pad_proportion=0.03305, over 63.00 utterances.], tot_loss[ctc_loss=0.2406, att_loss=0.3232, loss=0.3067, over 3277011.12 frames. utt_duration=1268 frames, utt_pad_proportion=0.0479, over 10348.71 utterances.], batch size: 63, lr: 3.79e-02, grad_scale: 8.0 2023-03-07 13:04:54,057 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6690.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:05:00,042 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6694.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:05:09,210 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6699.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:05:15,304 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.955e+02 4.204e+02 5.612e+02 7.162e+02 1.326e+03, threshold=1.122e+03, percent-clipped=2.0 2023-03-07 13:06:06,250 INFO [train2.py:809] (1/4) Epoch 2, batch 2750, loss[ctc_loss=0.3226, att_loss=0.3739, loss=0.3636, over 16465.00 frames. utt_duration=1433 frames, utt_pad_proportion=0.006719, over 46.00 utterances.], tot_loss[ctc_loss=0.2404, att_loss=0.3236, loss=0.307, over 3280221.59 frames. utt_duration=1261 frames, utt_pad_proportion=0.0494, over 10420.05 utterances.], batch size: 46, lr: 3.79e-02, grad_scale: 8.0 2023-03-07 13:06:26,269 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6747.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:06:27,116 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.91 vs. limit=2.0 2023-03-07 13:07:25,950 INFO [train2.py:809] (1/4) Epoch 2, batch 2800, loss[ctc_loss=0.2702, att_loss=0.3571, loss=0.3397, over 17292.00 frames. utt_duration=1174 frames, utt_pad_proportion=0.02352, over 59.00 utterances.], tot_loss[ctc_loss=0.2418, att_loss=0.3248, loss=0.3082, over 3278952.77 frames. utt_duration=1242 frames, utt_pad_proportion=0.05572, over 10575.86 utterances.], batch size: 59, lr: 3.78e-02, grad_scale: 8.0 2023-03-07 13:07:36,248 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6791.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:07:55,853 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.567e+02 4.477e+02 5.364e+02 6.587e+02 1.505e+03, threshold=1.073e+03, percent-clipped=3.0 2023-03-07 13:08:20,312 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6818.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 13:08:28,514 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6823.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:08:33,303 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6826.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:08:46,714 INFO [train2.py:809] (1/4) Epoch 2, batch 2850, loss[ctc_loss=0.1798, att_loss=0.2759, loss=0.2567, over 15878.00 frames. utt_duration=1630 frames, utt_pad_proportion=0.009611, over 39.00 utterances.], tot_loss[ctc_loss=0.2419, att_loss=0.325, loss=0.3084, over 3276610.42 frames. utt_duration=1228 frames, utt_pad_proportion=0.05964, over 10683.72 utterances.], batch size: 39, lr: 3.77e-02, grad_scale: 8.0 2023-03-07 13:08:53,175 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6839.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:09:25,247 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6859.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:09:35,803 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-03-07 13:09:38,408 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6867.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:09:49,559 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6874.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:09:56,310 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6878.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 13:10:06,297 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-03-07 13:10:06,817 INFO [train2.py:809] (1/4) Epoch 2, batch 2900, loss[ctc_loss=0.2079, att_loss=0.281, loss=0.2664, over 15779.00 frames. utt_duration=1663 frames, utt_pad_proportion=0.007976, over 38.00 utterances.], tot_loss[ctc_loss=0.2406, att_loss=0.3239, loss=0.3072, over 3267167.35 frames. utt_duration=1221 frames, utt_pad_proportion=0.06288, over 10719.00 utterances.], batch size: 38, lr: 3.76e-02, grad_scale: 8.0 2023-03-07 13:10:35,234 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4107, 5.0762, 4.7502, 4.5406, 5.0841, 4.9405, 4.8084, 4.6570], device='cuda:1'), covar=tensor([0.0841, 0.0249, 0.0210, 0.0519, 0.0248, 0.0293, 0.0240, 0.0283], device='cuda:1'), in_proj_covar=tensor([0.0208, 0.0153, 0.0098, 0.0113, 0.0156, 0.0171, 0.0130, 0.0143], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-03-07 13:10:36,473 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.536e+02 4.377e+02 5.489e+02 7.218e+02 1.392e+03, threshold=1.098e+03, percent-clipped=5.0 2023-03-07 13:10:42,899 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6907.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:10:43,209 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([1.0210, 1.2183, 1.5932, 1.3235, 1.6446, 1.6786, 2.2104, 1.9085], device='cuda:1'), covar=tensor([0.1017, 0.1132, 0.0685, 0.0996, 0.0609, 0.0945, 0.0541, 0.0671], device='cuda:1'), in_proj_covar=tensor([0.0074, 0.0073, 0.0067, 0.0070, 0.0068, 0.0080, 0.0068, 0.0089], device='cuda:1'), out_proj_covar=tensor([4.8810e-05, 5.4577e-05, 5.0930e-05, 5.6643e-05, 4.0979e-05, 6.2227e-05, 5.0909e-05, 4.7207e-05], device='cuda:1') 2023-03-07 13:10:46,204 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6909.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:10:55,830 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6915.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:11:07,561 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6922.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:11:27,226 INFO [train2.py:809] (1/4) Epoch 2, batch 2950, loss[ctc_loss=0.2022, att_loss=0.2913, loss=0.2735, over 16288.00 frames. utt_duration=1517 frames, utt_pad_proportion=0.006686, over 43.00 utterances.], tot_loss[ctc_loss=0.2397, att_loss=0.3233, loss=0.3066, over 3270673.48 frames. utt_duration=1233 frames, utt_pad_proportion=0.0585, over 10625.14 utterances.], batch size: 43, lr: 3.75e-02, grad_scale: 8.0 2023-03-07 13:11:35,170 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6939.0, num_to_drop=1, layers_to_drop={3} 2023-03-07 13:12:04,543 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4867, 2.4998, 3.4813, 4.3560, 4.5858, 4.4250, 2.8185, 2.5328], device='cuda:1'), covar=tensor([0.0183, 0.2017, 0.0808, 0.0290, 0.0110, 0.0174, 0.1646, 0.1776], device='cuda:1'), in_proj_covar=tensor([0.0091, 0.0136, 0.0122, 0.0081, 0.0064, 0.0074, 0.0134, 0.0131], device='cuda:1'), out_proj_covar=tensor([8.1112e-05, 1.2292e-04, 1.1507e-04, 9.1424e-05, 6.8545e-05, 6.5797e-05, 1.2790e-04, 1.1898e-04], device='cuda:1') 2023-03-07 13:12:14,534 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8179, 5.3883, 4.8089, 5.0864, 5.4570, 5.3462, 5.2279, 5.1708], device='cuda:1'), covar=tensor([0.0702, 0.0230, 0.0231, 0.0341, 0.0216, 0.0194, 0.0162, 0.0184], device='cuda:1'), in_proj_covar=tensor([0.0207, 0.0154, 0.0097, 0.0113, 0.0154, 0.0170, 0.0130, 0.0144], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-03-07 13:12:24,483 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=6970.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:12:24,766 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6970.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:12:47,622 INFO [train2.py:809] (1/4) Epoch 2, batch 3000, loss[ctc_loss=0.243, att_loss=0.333, loss=0.315, over 16478.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.005352, over 46.00 utterances.], tot_loss[ctc_loss=0.2398, att_loss=0.3237, loss=0.3069, over 3278547.47 frames. utt_duration=1235 frames, utt_pad_proportion=0.0555, over 10632.46 utterances.], batch size: 46, lr: 3.74e-02, grad_scale: 8.0 2023-03-07 13:12:47,623 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-07 13:13:01,226 INFO [train2.py:843] (1/4) Epoch 2, validation: ctc_loss=0.1245, att_loss=0.2759, loss=0.2456, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-07 13:13:01,227 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-07 13:13:01,447 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6985.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:13:30,571 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 3.232e+02 4.393e+02 5.286e+02 7.029e+02 1.971e+03, threshold=1.057e+03, percent-clipped=9.0 2023-03-07 13:13:40,235 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([1.0751, 1.3575, 1.4956, 1.1462, 1.9778, 1.4516, 1.5229, 1.5201], device='cuda:1'), covar=tensor([0.0972, 0.1081, 0.0665, 0.2395, 0.0522, 0.0905, 0.0678, 0.0698], device='cuda:1'), in_proj_covar=tensor([0.0074, 0.0072, 0.0066, 0.0069, 0.0068, 0.0079, 0.0068, 0.0088], device='cuda:1'), out_proj_covar=tensor([4.8471e-05, 5.4886e-05, 5.0998e-05, 5.5970e-05, 4.1339e-05, 6.1518e-05, 5.0679e-05, 4.7517e-05], device='cuda:1') 2023-03-07 13:14:12,244 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4480, 5.1305, 5.1270, 5.0631, 1.9671, 3.1967, 5.1409, 3.8811], device='cuda:1'), covar=tensor([0.0774, 0.0220, 0.0202, 0.0488, 1.3383, 0.2546, 0.0217, 0.3476], device='cuda:1'), in_proj_covar=tensor([0.0210, 0.0116, 0.0139, 0.0156, 0.0385, 0.0251, 0.0137, 0.0191], device='cuda:1'), out_proj_covar=tensor([1.1002e-04, 6.0971e-05, 6.6645e-05, 7.5574e-05, 1.8645e-04, 1.2312e-04, 6.6506e-05, 1.0588e-04], device='cuda:1') 2023-03-07 13:14:20,775 INFO [train2.py:809] (1/4) Epoch 2, batch 3050, loss[ctc_loss=0.2757, att_loss=0.3357, loss=0.3237, over 16764.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.006658, over 48.00 utterances.], tot_loss[ctc_loss=0.2403, att_loss=0.324, loss=0.3073, over 3278539.63 frames. utt_duration=1226 frames, utt_pad_proportion=0.05847, over 10709.95 utterances.], batch size: 48, lr: 3.73e-02, grad_scale: 8.0 2023-03-07 13:14:42,947 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7048.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 13:15:27,282 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.4149, 3.5851, 3.6271, 3.7115, 2.1416, 3.7549, 2.5524, 3.4773], device='cuda:1'), covar=tensor([0.0267, 0.0289, 0.0585, 0.0270, 0.3772, 0.0220, 0.1233, 0.0456], device='cuda:1'), in_proj_covar=tensor([0.0079, 0.0074, 0.0146, 0.0114, 0.0213, 0.0080, 0.0147, 0.0123], device='cuda:1'), out_proj_covar=tensor([6.6953e-05, 6.4091e-05, 1.1380e-04, 8.5119e-05, 1.4954e-04, 6.6619e-05, 1.0968e-04, 9.1646e-05], device='cuda:1') 2023-03-07 13:15:41,477 INFO [train2.py:809] (1/4) Epoch 2, batch 3100, loss[ctc_loss=0.2324, att_loss=0.3108, loss=0.2951, over 15961.00 frames. utt_duration=1559 frames, utt_pad_proportion=0.006622, over 41.00 utterances.], tot_loss[ctc_loss=0.2388, att_loss=0.3234, loss=0.3065, over 3279271.60 frames. utt_duration=1237 frames, utt_pad_proportion=0.0566, over 10619.04 utterances.], batch size: 41, lr: 3.72e-02, grad_scale: 8.0 2023-03-07 13:16:10,553 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.443e+02 4.796e+02 5.430e+02 6.688e+02 1.495e+03, threshold=1.086e+03, percent-clipped=6.0 2023-03-07 13:16:20,424 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7109.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 13:16:35,030 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7118.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 13:16:42,341 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7123.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:17:01,201 INFO [train2.py:809] (1/4) Epoch 2, batch 3150, loss[ctc_loss=0.2302, att_loss=0.3357, loss=0.3146, over 17061.00 frames. utt_duration=1314 frames, utt_pad_proportion=0.008418, over 52.00 utterances.], tot_loss[ctc_loss=0.2385, att_loss=0.323, loss=0.3061, over 3285624.43 frames. utt_duration=1239 frames, utt_pad_proportion=0.05385, over 10620.39 utterances.], batch size: 52, lr: 3.71e-02, grad_scale: 8.0 2023-03-07 13:17:46,728 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7163.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:17:51,644 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=7166.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:17:59,089 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=7171.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:18:22,234 INFO [train2.py:809] (1/4) Epoch 2, batch 3200, loss[ctc_loss=0.207, att_loss=0.3123, loss=0.2912, over 16274.00 frames. utt_duration=1515 frames, utt_pad_proportion=0.00698, over 43.00 utterances.], tot_loss[ctc_loss=0.2394, att_loss=0.3241, loss=0.3072, over 3292480.24 frames. utt_duration=1218 frames, utt_pad_proportion=0.05637, over 10828.44 utterances.], batch size: 43, lr: 3.71e-02, grad_scale: 8.0 2023-03-07 13:18:50,783 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.936e+02 4.512e+02 5.517e+02 7.088e+02 1.562e+03, threshold=1.103e+03, percent-clipped=2.0 2023-03-07 13:19:25,300 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7224.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 13:19:41,625 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7234.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 13:19:42,949 INFO [train2.py:809] (1/4) Epoch 2, batch 3250, loss[ctc_loss=0.2375, att_loss=0.3426, loss=0.3216, over 16870.00 frames. utt_duration=1379 frames, utt_pad_proportion=0.007523, over 49.00 utterances.], tot_loss[ctc_loss=0.2402, att_loss=0.3248, loss=0.3079, over 3295672.92 frames. utt_duration=1226 frames, utt_pad_proportion=0.05466, over 10768.29 utterances.], batch size: 49, lr: 3.70e-02, grad_scale: 8.0 2023-03-07 13:20:29,952 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7265.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:20:45,991 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7275.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:20:48,922 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4421, 5.0447, 4.7422, 4.5619, 5.0829, 4.9958, 4.7543, 4.7353], device='cuda:1'), covar=tensor([0.0959, 0.0367, 0.0251, 0.0588, 0.0336, 0.0269, 0.0252, 0.0281], device='cuda:1'), in_proj_covar=tensor([0.0220, 0.0159, 0.0101, 0.0121, 0.0162, 0.0178, 0.0139, 0.0151], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-03-07 13:21:02,607 INFO [train2.py:809] (1/4) Epoch 2, batch 3300, loss[ctc_loss=0.2601, att_loss=0.3291, loss=0.3153, over 17085.00 frames. utt_duration=691.6 frames, utt_pad_proportion=0.1322, over 99.00 utterances.], tot_loss[ctc_loss=0.2395, att_loss=0.3241, loss=0.3071, over 3291199.95 frames. utt_duration=1235 frames, utt_pad_proportion=0.05261, over 10668.49 utterances.], batch size: 99, lr: 3.69e-02, grad_scale: 8.0 2023-03-07 13:21:02,953 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7285.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:21:30,733 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.392e+02 4.714e+02 6.021e+02 7.352e+02 2.285e+03, threshold=1.204e+03, percent-clipped=4.0 2023-03-07 13:21:46,009 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5079, 4.2965, 4.8220, 4.7747, 5.0127, 4.3027, 4.2558, 2.3598], device='cuda:1'), covar=tensor([0.0277, 0.0487, 0.0176, 0.0156, 0.0559, 0.0268, 0.0437, 0.3946], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0099, 0.0094, 0.0098, 0.0170, 0.0117, 0.0082, 0.0228], device='cuda:1'), out_proj_covar=tensor([8.9293e-05, 6.9241e-05, 7.1025e-05, 7.0226e-05, 1.4427e-04, 8.2435e-05, 6.6064e-05, 1.6723e-04], device='cuda:1') 2023-03-07 13:21:52,220 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3199, 4.8586, 4.9923, 4.7160, 1.9293, 2.8826, 4.9802, 3.7535], device='cuda:1'), covar=tensor([0.0714, 0.0242, 0.0136, 0.0416, 1.1890, 0.2827, 0.0163, 0.3143], device='cuda:1'), in_proj_covar=tensor([0.0208, 0.0121, 0.0141, 0.0163, 0.0388, 0.0260, 0.0132, 0.0198], device='cuda:1'), out_proj_covar=tensor([1.1028e-04, 6.3778e-05, 6.8457e-05, 7.8606e-05, 1.8743e-04, 1.2930e-04, 6.5031e-05, 1.1021e-04], device='cuda:1') 2023-03-07 13:22:20,350 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=7333.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:22:23,592 INFO [train2.py:809] (1/4) Epoch 2, batch 3350, loss[ctc_loss=0.1705, att_loss=0.2653, loss=0.2463, over 15766.00 frames. utt_duration=1661 frames, utt_pad_proportion=0.008981, over 38.00 utterances.], tot_loss[ctc_loss=0.2369, att_loss=0.3229, loss=0.3057, over 3294658.27 frames. utt_duration=1251 frames, utt_pad_proportion=0.04856, over 10546.43 utterances.], batch size: 38, lr: 3.68e-02, grad_scale: 8.0 2023-03-07 13:22:25,546 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7336.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:22:32,525 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-03-07 13:22:35,373 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.72 vs. limit=5.0 2023-03-07 13:23:04,801 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5227, 5.1400, 4.9404, 4.7889, 5.1887, 5.0513, 4.8788, 4.8120], device='cuda:1'), covar=tensor([0.0799, 0.0321, 0.0211, 0.0428, 0.0235, 0.0225, 0.0255, 0.0262], device='cuda:1'), in_proj_covar=tensor([0.0221, 0.0161, 0.0102, 0.0120, 0.0162, 0.0178, 0.0140, 0.0150], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-03-07 13:23:43,568 INFO [train2.py:809] (1/4) Epoch 2, batch 3400, loss[ctc_loss=0.284, att_loss=0.3452, loss=0.333, over 17385.00 frames. utt_duration=881.9 frames, utt_pad_proportion=0.07368, over 79.00 utterances.], tot_loss[ctc_loss=0.2362, att_loss=0.3224, loss=0.3051, over 3279733.20 frames. utt_duration=1244 frames, utt_pad_proportion=0.05225, over 10562.81 utterances.], batch size: 79, lr: 3.67e-02, grad_scale: 8.0 2023-03-07 13:24:12,391 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.987e+02 4.124e+02 5.654e+02 7.446e+02 2.114e+03, threshold=1.131e+03, percent-clipped=4.0 2023-03-07 13:24:14,047 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7404.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 13:24:18,713 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7407.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:25:03,870 INFO [train2.py:809] (1/4) Epoch 2, batch 3450, loss[ctc_loss=0.2549, att_loss=0.3357, loss=0.3195, over 17304.00 frames. utt_duration=1260 frames, utt_pad_proportion=0.01191, over 55.00 utterances.], tot_loss[ctc_loss=0.2351, att_loss=0.322, loss=0.3046, over 3277293.70 frames. utt_duration=1267 frames, utt_pad_proportion=0.04664, over 10359.02 utterances.], batch size: 55, lr: 3.66e-02, grad_scale: 8.0 2023-03-07 13:25:42,297 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.08 vs. limit=5.0 2023-03-07 13:25:56,125 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7468.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:26:22,847 INFO [train2.py:809] (1/4) Epoch 2, batch 3500, loss[ctc_loss=0.201, att_loss=0.3136, loss=0.2911, over 16870.00 frames. utt_duration=1379 frames, utt_pad_proportion=0.008119, over 49.00 utterances.], tot_loss[ctc_loss=0.2352, att_loss=0.322, loss=0.3046, over 3277573.89 frames. utt_duration=1267 frames, utt_pad_proportion=0.04584, over 10357.62 utterances.], batch size: 49, lr: 3.65e-02, grad_scale: 8.0 2023-03-07 13:26:50,319 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.396e+02 4.119e+02 4.861e+02 5.923e+02 1.212e+03, threshold=9.723e+02, percent-clipped=1.0 2023-03-07 13:27:16,200 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7519.0, num_to_drop=1, layers_to_drop={3} 2023-03-07 13:27:39,887 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7534.0, num_to_drop=1, layers_to_drop={2} 2023-03-07 13:27:41,023 INFO [train2.py:809] (1/4) Epoch 2, batch 3550, loss[ctc_loss=0.2795, att_loss=0.3551, loss=0.34, over 17138.00 frames. utt_duration=1226 frames, utt_pad_proportion=0.01373, over 56.00 utterances.], tot_loss[ctc_loss=0.2367, att_loss=0.3222, loss=0.3051, over 3276199.67 frames. utt_duration=1249 frames, utt_pad_proportion=0.05043, over 10507.08 utterances.], batch size: 56, lr: 3.65e-02, grad_scale: 8.0 2023-03-07 13:28:27,995 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7565.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:28:55,954 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=7582.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 13:29:00,321 INFO [train2.py:809] (1/4) Epoch 2, batch 3600, loss[ctc_loss=0.2367, att_loss=0.3073, loss=0.2932, over 15627.00 frames. utt_duration=1691 frames, utt_pad_proportion=0.009373, over 37.00 utterances.], tot_loss[ctc_loss=0.2363, att_loss=0.3215, loss=0.3045, over 3278019.85 frames. utt_duration=1256 frames, utt_pad_proportion=0.04853, over 10453.30 utterances.], batch size: 37, lr: 3.64e-02, grad_scale: 8.0 2023-03-07 13:29:28,533 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.614e+02 5.143e+02 6.522e+02 9.491e+02 2.921e+03, threshold=1.304e+03, percent-clipped=20.0 2023-03-07 13:29:44,792 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=7613.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:30:03,094 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7624.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:30:14,834 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7631.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:30:20,988 INFO [train2.py:809] (1/4) Epoch 2, batch 3650, loss[ctc_loss=0.2286, att_loss=0.3338, loss=0.3127, over 17042.00 frames. utt_duration=1313 frames, utt_pad_proportion=0.008526, over 52.00 utterances.], tot_loss[ctc_loss=0.2336, att_loss=0.32, loss=0.3027, over 3274303.38 frames. utt_duration=1260 frames, utt_pad_proportion=0.0486, over 10409.96 utterances.], batch size: 52, lr: 3.63e-02, grad_scale: 8.0 2023-03-07 13:31:41,712 INFO [train2.py:809] (1/4) Epoch 2, batch 3700, loss[ctc_loss=0.2017, att_loss=0.2951, loss=0.2764, over 16385.00 frames. utt_duration=1491 frames, utt_pad_proportion=0.007697, over 44.00 utterances.], tot_loss[ctc_loss=0.2329, att_loss=0.32, loss=0.3026, over 3273059.30 frames. utt_duration=1264 frames, utt_pad_proportion=0.0484, over 10370.03 utterances.], batch size: 44, lr: 3.62e-02, grad_scale: 8.0 2023-03-07 13:31:42,153 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7685.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:32:10,452 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.948e+02 4.093e+02 5.003e+02 6.020e+02 1.181e+03, threshold=1.001e+03, percent-clipped=0.0 2023-03-07 13:32:12,241 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7704.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 13:32:47,110 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0753, 4.1030, 4.2284, 4.3713, 4.6236, 4.0568, 3.9086, 2.0546], device='cuda:1'), covar=tensor([0.0364, 0.0503, 0.0332, 0.0172, 0.0975, 0.0360, 0.0532, 0.4210], device='cuda:1'), in_proj_covar=tensor([0.0123, 0.0101, 0.0096, 0.0102, 0.0172, 0.0119, 0.0084, 0.0229], device='cuda:1'), out_proj_covar=tensor([9.2971e-05, 7.1783e-05, 7.4819e-05, 7.5784e-05, 1.4942e-04, 8.6240e-05, 6.8748e-05, 1.7198e-04], device='cuda:1') 2023-03-07 13:33:01,948 INFO [train2.py:809] (1/4) Epoch 2, batch 3750, loss[ctc_loss=0.1982, att_loss=0.3019, loss=0.2811, over 16176.00 frames. utt_duration=1580 frames, utt_pad_proportion=0.00696, over 41.00 utterances.], tot_loss[ctc_loss=0.2318, att_loss=0.3194, loss=0.3018, over 3273807.88 frames. utt_duration=1261 frames, utt_pad_proportion=0.05034, over 10394.97 utterances.], batch size: 41, lr: 3.61e-02, grad_scale: 8.0 2023-03-07 13:33:14,947 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7743.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:33:29,482 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=7752.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 13:33:46,950 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7763.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:34:21,949 INFO [train2.py:809] (1/4) Epoch 2, batch 3800, loss[ctc_loss=0.2042, att_loss=0.2911, loss=0.2737, over 15774.00 frames. utt_duration=1662 frames, utt_pad_proportion=0.008526, over 38.00 utterances.], tot_loss[ctc_loss=0.2326, att_loss=0.3201, loss=0.3026, over 3263981.17 frames. utt_duration=1245 frames, utt_pad_proportion=0.05615, over 10496.59 utterances.], batch size: 38, lr: 3.60e-02, grad_scale: 8.0 2023-03-07 13:34:49,623 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.1718, 3.1826, 3.2716, 3.2104, 3.4476, 3.1143, 1.8786, 3.8149], device='cuda:1'), covar=tensor([0.1118, 0.0279, 0.0913, 0.0541, 0.0349, 0.0929, 0.1077, 0.0093], device='cuda:1'), in_proj_covar=tensor([0.0093, 0.0057, 0.0101, 0.0084, 0.0066, 0.0103, 0.0090, 0.0048], device='cuda:1'), out_proj_covar=tensor([1.2609e-04, 9.3702e-05, 1.4833e-04, 1.1392e-04, 1.0803e-04, 1.5079e-04, 1.1896e-04, 7.4090e-05], device='cuda:1') 2023-03-07 13:34:49,633 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7802.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:34:50,773 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.304e+02 4.201e+02 5.009e+02 6.863e+02 2.448e+03, threshold=1.002e+03, percent-clipped=6.0 2023-03-07 13:34:52,742 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7804.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:35:14,354 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=2.00 vs. limit=2.0 2023-03-07 13:35:16,882 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7819.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 13:35:42,003 INFO [train2.py:809] (1/4) Epoch 2, batch 3850, loss[ctc_loss=0.27, att_loss=0.358, loss=0.3404, over 16967.00 frames. utt_duration=1359 frames, utt_pad_proportion=0.007305, over 50.00 utterances.], tot_loss[ctc_loss=0.2315, att_loss=0.3185, loss=0.3011, over 3244384.45 frames. utt_duration=1262 frames, utt_pad_proportion=0.05663, over 10293.93 utterances.], batch size: 50, lr: 3.60e-02, grad_scale: 8.0 2023-03-07 13:35:45,192 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8985, 5.1899, 4.8870, 4.9381, 5.3435, 5.1113, 4.6640, 5.1497], device='cuda:1'), covar=tensor([0.0074, 0.0110, 0.0094, 0.0074, 0.0045, 0.0062, 0.0255, 0.0124], device='cuda:1'), in_proj_covar=tensor([0.0038, 0.0040, 0.0041, 0.0030, 0.0028, 0.0036, 0.0051, 0.0045], device='cuda:1'), out_proj_covar=tensor([8.2388e-05, 8.5588e-05, 1.0089e-04, 6.7827e-05, 5.8292e-05, 8.1258e-05, 1.0741e-04, 9.8824e-05], device='cuda:1') 2023-03-07 13:36:01,517 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 2023-03-07 13:36:25,219 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7863.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:36:26,659 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.4877, 3.3706, 3.3462, 3.0845, 3.5375, 3.1534, 1.8529, 3.7180], device='cuda:1'), covar=tensor([0.0908, 0.0185, 0.0590, 0.0595, 0.0282, 0.0775, 0.1162, 0.0115], device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0059, 0.0104, 0.0089, 0.0068, 0.0107, 0.0093, 0.0050], device='cuda:1'), out_proj_covar=tensor([1.3179e-04, 9.7544e-05, 1.5389e-04, 1.2006e-04, 1.1177e-04, 1.5753e-04, 1.2352e-04, 7.8064e-05], device='cuda:1') 2023-03-07 13:36:31,179 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=7867.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:36:59,758 INFO [train2.py:809] (1/4) Epoch 2, batch 3900, loss[ctc_loss=0.2309, att_loss=0.3336, loss=0.3131, over 16873.00 frames. utt_duration=1379 frames, utt_pad_proportion=0.007214, over 49.00 utterances.], tot_loss[ctc_loss=0.2313, att_loss=0.3188, loss=0.3013, over 3253926.63 frames. utt_duration=1269 frames, utt_pad_proportion=0.05293, over 10267.54 utterances.], batch size: 49, lr: 3.59e-02, grad_scale: 8.0 2023-03-07 13:37:27,831 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.554e+02 4.818e+02 5.738e+02 6.928e+02 1.942e+03, threshold=1.148e+03, percent-clipped=11.0 2023-03-07 13:37:47,433 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.04 vs. limit=2.0 2023-03-07 13:38:11,525 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7931.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:38:17,768 INFO [train2.py:809] (1/4) Epoch 2, batch 3950, loss[ctc_loss=0.2075, att_loss=0.3028, loss=0.2838, over 16472.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.007061, over 46.00 utterances.], tot_loss[ctc_loss=0.2306, att_loss=0.3175, loss=0.3001, over 3244814.17 frames. utt_duration=1273 frames, utt_pad_proportion=0.05425, over 10205.15 utterances.], batch size: 46, lr: 3.58e-02, grad_scale: 8.0 2023-03-07 13:39:31,708 INFO [train2.py:809] (1/4) Epoch 3, batch 0, loss[ctc_loss=0.1883, att_loss=0.2737, loss=0.2566, over 15518.00 frames. utt_duration=1726 frames, utt_pad_proportion=0.007715, over 36.00 utterances.], tot_loss[ctc_loss=0.1883, att_loss=0.2737, loss=0.2566, over 15518.00 frames. utt_duration=1726 frames, utt_pad_proportion=0.007715, over 36.00 utterances.], batch size: 36, lr: 3.40e-02, grad_scale: 8.0 2023-03-07 13:39:31,708 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-07 13:39:44,018 INFO [train2.py:843] (1/4) Epoch 3, validation: ctc_loss=0.122, att_loss=0.2748, loss=0.2442, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-07 13:39:44,020 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-07 13:40:00,311 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=7979.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:40:01,859 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7980.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:40:42,077 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=2.06 vs. limit=2.0 2023-03-07 13:40:42,655 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.716e+02 4.111e+02 5.021e+02 6.373e+02 1.388e+03, threshold=1.004e+03, percent-clipped=3.0 2023-03-07 13:41:08,149 INFO [train2.py:809] (1/4) Epoch 3, batch 50, loss[ctc_loss=0.186, att_loss=0.2891, loss=0.2685, over 15626.00 frames. utt_duration=1691 frames, utt_pad_proportion=0.009858, over 37.00 utterances.], tot_loss[ctc_loss=0.2256, att_loss=0.3153, loss=0.2973, over 737853.27 frames. utt_duration=1147 frames, utt_pad_proportion=0.07963, over 2575.65 utterances.], batch size: 37, lr: 3.39e-02, grad_scale: 16.0 2023-03-07 13:42:17,949 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8063.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:42:23,009 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.8556, 6.0400, 5.6340, 6.0504, 5.6086, 5.6242, 5.5120, 5.5148], device='cuda:1'), covar=tensor([0.0962, 0.0693, 0.0587, 0.0481, 0.0534, 0.0974, 0.1798, 0.1641], device='cuda:1'), in_proj_covar=tensor([0.0237, 0.0288, 0.0235, 0.0215, 0.0193, 0.0292, 0.0306, 0.0285], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-07 13:42:27,411 INFO [train2.py:809] (1/4) Epoch 3, batch 100, loss[ctc_loss=0.2199, att_loss=0.3152, loss=0.2962, over 16954.00 frames. utt_duration=1358 frames, utt_pad_proportion=0.008137, over 50.00 utterances.], tot_loss[ctc_loss=0.2262, att_loss=0.3161, loss=0.2982, over 1296277.83 frames. utt_duration=1182 frames, utt_pad_proportion=0.07376, over 4391.29 utterances.], batch size: 50, lr: 3.38e-02, grad_scale: 16.0 2023-03-07 13:43:15,547 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=8099.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:43:19,268 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.11 vs. limit=2.0 2023-03-07 13:43:21,308 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.756e+02 4.455e+02 5.328e+02 6.355e+02 1.061e+03, threshold=1.066e+03, percent-clipped=2.0 2023-03-07 13:43:33,804 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=8111.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:43:47,467 INFO [train2.py:809] (1/4) Epoch 3, batch 150, loss[ctc_loss=0.2007, att_loss=0.3062, loss=0.2851, over 16401.00 frames. utt_duration=1493 frames, utt_pad_proportion=0.007631, over 44.00 utterances.], tot_loss[ctc_loss=0.2238, att_loss=0.3145, loss=0.2964, over 1727177.60 frames. utt_duration=1235 frames, utt_pad_proportion=0.06282, over 5599.15 utterances.], batch size: 44, lr: 3.37e-02, grad_scale: 16.0 2023-03-07 13:44:22,277 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-03-07 13:44:50,030 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=8158.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:44:56,422 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1229, 2.5315, 3.4640, 2.0273, 3.3585, 4.4736, 4.1040, 3.1230], device='cuda:1'), covar=tensor([0.0464, 0.1474, 0.0868, 0.1956, 0.1061, 0.0162, 0.0463, 0.1395], device='cuda:1'), in_proj_covar=tensor([0.0123, 0.0135, 0.0115, 0.0138, 0.0144, 0.0079, 0.0093, 0.0145], device='cuda:1'), out_proj_covar=tensor([1.2483e-04, 1.3637e-04, 1.3805e-04, 1.4068e-04, 1.5160e-04, 9.2797e-05, 1.1334e-04, 1.4686e-04], device='cuda:1') 2023-03-07 13:45:07,257 INFO [train2.py:809] (1/4) Epoch 3, batch 200, loss[ctc_loss=0.1845, att_loss=0.2747, loss=0.2567, over 15809.00 frames. utt_duration=1665 frames, utt_pad_proportion=0.006296, over 38.00 utterances.], tot_loss[ctc_loss=0.2193, att_loss=0.3117, loss=0.2932, over 2072713.31 frames. utt_duration=1284 frames, utt_pad_proportion=0.0484, over 6463.89 utterances.], batch size: 38, lr: 3.37e-02, grad_scale: 16.0 2023-03-07 13:46:02,074 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.226e+02 3.842e+02 5.004e+02 6.502e+02 2.003e+03, threshold=1.001e+03, percent-clipped=2.0 2023-03-07 13:46:28,360 INFO [train2.py:809] (1/4) Epoch 3, batch 250, loss[ctc_loss=0.2252, att_loss=0.3017, loss=0.2864, over 16117.00 frames. utt_duration=1536 frames, utt_pad_proportion=0.006787, over 42.00 utterances.], tot_loss[ctc_loss=0.2205, att_loss=0.3121, loss=0.2938, over 2339745.83 frames. utt_duration=1247 frames, utt_pad_proportion=0.05427, over 7516.07 utterances.], batch size: 42, lr: 3.36e-02, grad_scale: 16.0 2023-03-07 13:47:02,829 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-03-07 13:47:40,028 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8265.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:47:46,794 INFO [train2.py:809] (1/4) Epoch 3, batch 300, loss[ctc_loss=0.2521, att_loss=0.3258, loss=0.3111, over 16549.00 frames. utt_duration=1473 frames, utt_pad_proportion=0.005717, over 45.00 utterances.], tot_loss[ctc_loss=0.2205, att_loss=0.3124, loss=0.294, over 2546274.46 frames. utt_duration=1262 frames, utt_pad_proportion=0.0502, over 8078.06 utterances.], batch size: 45, lr: 3.35e-02, grad_scale: 16.0 2023-03-07 13:48:03,755 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8280.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:48:41,924 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.817e+02 4.123e+02 5.343e+02 7.099e+02 1.515e+03, threshold=1.069e+03, percent-clipped=5.0 2023-03-07 13:49:05,817 INFO [train2.py:809] (1/4) Epoch 3, batch 350, loss[ctc_loss=0.2144, att_loss=0.3155, loss=0.2953, over 17354.00 frames. utt_duration=880.3 frames, utt_pad_proportion=0.07631, over 79.00 utterances.], tot_loss[ctc_loss=0.2203, att_loss=0.3124, loss=0.294, over 2711710.57 frames. utt_duration=1245 frames, utt_pad_proportion=0.0528, over 8726.24 utterances.], batch size: 79, lr: 3.34e-02, grad_scale: 8.0 2023-03-07 13:49:16,790 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8326.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:49:19,541 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=8328.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:50:22,881 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.71 vs. limit=2.0 2023-03-07 13:50:25,025 INFO [train2.py:809] (1/4) Epoch 3, batch 400, loss[ctc_loss=0.2195, att_loss=0.3141, loss=0.2952, over 17369.00 frames. utt_duration=1104 frames, utt_pad_proportion=0.03476, over 63.00 utterances.], tot_loss[ctc_loss=0.2209, att_loss=0.3129, loss=0.2945, over 2835457.06 frames. utt_duration=1237 frames, utt_pad_proportion=0.05521, over 9183.61 utterances.], batch size: 63, lr: 3.34e-02, grad_scale: 8.0 2023-03-07 13:50:38,384 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-03-07 13:51:12,640 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8399.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:51:20,127 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.563e+02 4.039e+02 5.098e+02 6.528e+02 1.596e+03, threshold=1.020e+03, percent-clipped=2.0 2023-03-07 13:51:44,902 INFO [train2.py:809] (1/4) Epoch 3, batch 450, loss[ctc_loss=0.2344, att_loss=0.3184, loss=0.3016, over 16633.00 frames. utt_duration=1417 frames, utt_pad_proportion=0.00499, over 47.00 utterances.], tot_loss[ctc_loss=0.2197, att_loss=0.3123, loss=0.2938, over 2930945.43 frames. utt_duration=1244 frames, utt_pad_proportion=0.05301, over 9433.71 utterances.], batch size: 47, lr: 3.33e-02, grad_scale: 8.0 2023-03-07 13:52:30,021 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=8447.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:52:47,963 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8458.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:52:48,499 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.82 vs. limit=2.0 2023-03-07 13:52:58,029 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.1266, 1.3201, 2.0173, 2.1739, 1.0302, 2.8980, 1.2235, 1.9924], device='cuda:1'), covar=tensor([0.0464, 0.1177, 0.0823, 0.1267, 0.1910, 0.0307, 0.0825, 0.2127], device='cuda:1'), in_proj_covar=tensor([0.0066, 0.0058, 0.0055, 0.0078, 0.0066, 0.0053, 0.0070, 0.0087], device='cuda:1'), out_proj_covar=tensor([3.8156e-05, 3.9282e-05, 3.8285e-05, 4.3791e-05, 3.6521e-05, 3.2918e-05, 4.1402e-05, 5.5192e-05], device='cuda:1') 2023-03-07 13:53:05,350 INFO [train2.py:809] (1/4) Epoch 3, batch 500, loss[ctc_loss=0.2012, att_loss=0.2967, loss=0.2776, over 16549.00 frames. utt_duration=1473 frames, utt_pad_proportion=0.005537, over 45.00 utterances.], tot_loss[ctc_loss=0.2204, att_loss=0.3124, loss=0.294, over 3006558.47 frames. utt_duration=1247 frames, utt_pad_proportion=0.05188, over 9658.01 utterances.], batch size: 45, lr: 3.32e-02, grad_scale: 8.0 2023-03-07 13:53:35,785 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3320, 4.4007, 4.3288, 4.5855, 4.6724, 4.2134, 3.9005, 1.9501], device='cuda:1'), covar=tensor([0.0293, 0.0398, 0.0298, 0.0142, 0.0964, 0.0340, 0.0479, 0.3970], device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0104, 0.0097, 0.0106, 0.0188, 0.0125, 0.0090, 0.0236], device='cuda:1'), out_proj_covar=tensor([9.5239e-05, 7.6401e-05, 7.8489e-05, 8.1606e-05, 1.6369e-04, 9.2883e-05, 7.6577e-05, 1.8145e-04], device='cuda:1') 2023-03-07 13:54:00,442 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.491e+02 4.509e+02 5.776e+02 7.013e+02 1.879e+03, threshold=1.155e+03, percent-clipped=5.0 2023-03-07 13:54:04,276 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=8506.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:54:24,652 INFO [train2.py:809] (1/4) Epoch 3, batch 550, loss[ctc_loss=0.1955, att_loss=0.3023, loss=0.2809, over 17330.00 frames. utt_duration=1177 frames, utt_pad_proportion=0.02189, over 59.00 utterances.], tot_loss[ctc_loss=0.2204, att_loss=0.3127, loss=0.2942, over 3066618.12 frames. utt_duration=1221 frames, utt_pad_proportion=0.05829, over 10056.40 utterances.], batch size: 59, lr: 3.31e-02, grad_scale: 8.0 2023-03-07 13:54:38,185 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([1.6771, 1.0978, 1.7055, 2.5949, 0.8502, 2.9309, 1.9297, 1.7231], device='cuda:1'), covar=tensor([0.0848, 0.1742, 0.1340, 0.0716, 0.2329, 0.0305, 0.1359, 0.1866], device='cuda:1'), in_proj_covar=tensor([0.0067, 0.0061, 0.0058, 0.0081, 0.0070, 0.0055, 0.0071, 0.0091], device='cuda:1'), out_proj_covar=tensor([3.9274e-05, 4.0669e-05, 3.9983e-05, 4.5470e-05, 3.9889e-05, 3.3976e-05, 4.2309e-05, 5.7499e-05], device='cuda:1') 2023-03-07 13:55:32,620 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-03-07 13:55:45,140 INFO [train2.py:809] (1/4) Epoch 3, batch 600, loss[ctc_loss=0.1984, att_loss=0.3033, loss=0.2823, over 16263.00 frames. utt_duration=1514 frames, utt_pad_proportion=0.007015, over 43.00 utterances.], tot_loss[ctc_loss=0.2201, att_loss=0.3127, loss=0.2942, over 3113893.73 frames. utt_duration=1232 frames, utt_pad_proportion=0.05615, over 10123.94 utterances.], batch size: 43, lr: 3.31e-02, grad_scale: 8.0 2023-03-07 13:56:42,024 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.080e+02 4.122e+02 4.877e+02 6.146e+02 9.874e+02, threshold=9.754e+02, percent-clipped=0.0 2023-03-07 13:57:05,729 INFO [train2.py:809] (1/4) Epoch 3, batch 650, loss[ctc_loss=0.2483, att_loss=0.3187, loss=0.3046, over 16328.00 frames. utt_duration=1453 frames, utt_pad_proportion=0.006232, over 45.00 utterances.], tot_loss[ctc_loss=0.2201, att_loss=0.3128, loss=0.2942, over 3150223.93 frames. utt_duration=1224 frames, utt_pad_proportion=0.0569, over 10306.46 utterances.], batch size: 45, lr: 3.30e-02, grad_scale: 8.0 2023-03-07 13:57:09,600 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=8621.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 13:57:23,790 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9946, 3.0888, 5.1170, 3.5143, 3.5439, 4.4882, 4.9263, 4.7083], device='cuda:1'), covar=tensor([0.0105, 0.1240, 0.0166, 0.1796, 0.2383, 0.0515, 0.0148, 0.0285], device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0196, 0.0110, 0.0239, 0.0288, 0.0140, 0.0089, 0.0103], device='cuda:1'), out_proj_covar=tensor([7.6160e-05, 1.3968e-04, 8.0143e-05, 1.8338e-04, 2.0426e-04, 1.1355e-04, 7.0657e-05, 8.2011e-05], device='cuda:1') 2023-03-07 13:57:33,261 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.5758, 4.4886, 4.5999, 4.4638, 1.9661, 3.5982, 2.1357, 3.8326], device='cuda:1'), covar=tensor([0.0329, 0.0197, 0.0602, 0.0302, 0.4162, 0.0374, 0.1744, 0.0661], device='cuda:1'), in_proj_covar=tensor([0.0088, 0.0077, 0.0173, 0.0113, 0.0220, 0.0088, 0.0162, 0.0147], device='cuda:1'), out_proj_covar=tensor([7.7125e-05, 6.8797e-05, 1.3875e-04, 8.8763e-05, 1.6274e-04, 7.4799e-05, 1.2678e-04, 1.1604e-04], device='cuda:1') 2023-03-07 13:57:50,181 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-03-07 13:58:25,325 INFO [train2.py:809] (1/4) Epoch 3, batch 700, loss[ctc_loss=0.2507, att_loss=0.3391, loss=0.3214, over 17020.00 frames. utt_duration=1286 frames, utt_pad_proportion=0.01162, over 53.00 utterances.], tot_loss[ctc_loss=0.2233, att_loss=0.3149, loss=0.2966, over 3177587.51 frames. utt_duration=1218 frames, utt_pad_proportion=0.05962, over 10451.89 utterances.], batch size: 53, lr: 3.29e-02, grad_scale: 8.0 2023-03-07 13:58:46,382 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0176, 2.5196, 5.0887, 3.7594, 3.4990, 4.3591, 4.7182, 4.6866], device='cuda:1'), covar=tensor([0.0077, 0.1348, 0.0145, 0.1295, 0.2203, 0.0422, 0.0133, 0.0202], device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0193, 0.0110, 0.0240, 0.0289, 0.0141, 0.0089, 0.0104], device='cuda:1'), out_proj_covar=tensor([7.6648e-05, 1.3785e-04, 8.0203e-05, 1.8430e-04, 2.0482e-04, 1.1507e-04, 7.1225e-05, 8.3210e-05], device='cuda:1') 2023-03-07 13:58:48,397 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.70 vs. limit=2.0 2023-03-07 13:58:52,675 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.7990, 2.2608, 3.0437, 4.3213, 4.4508, 4.4432, 2.5927, 1.3285], device='cuda:1'), covar=tensor([0.0424, 0.2283, 0.1226, 0.0396, 0.0151, 0.0131, 0.1840, 0.3134], device='cuda:1'), in_proj_covar=tensor([0.0102, 0.0151, 0.0146, 0.0095, 0.0073, 0.0089, 0.0154, 0.0144], device='cuda:1'), out_proj_covar=tensor([9.6456e-05, 1.3973e-04, 1.3937e-04, 1.0658e-04, 7.6038e-05, 8.3063e-05, 1.4999e-04, 1.3390e-04], device='cuda:1') 2023-03-07 13:59:10,666 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-03-07 13:59:21,220 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.427e+02 4.229e+02 5.521e+02 7.710e+02 1.687e+03, threshold=1.104e+03, percent-clipped=9.0 2023-03-07 13:59:27,904 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8916, 2.7584, 5.0358, 3.7209, 3.4252, 4.4003, 4.6040, 4.5806], device='cuda:1'), covar=tensor([0.0087, 0.1482, 0.0137, 0.1410, 0.2371, 0.0464, 0.0198, 0.0314], device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0196, 0.0113, 0.0243, 0.0293, 0.0144, 0.0091, 0.0107], device='cuda:1'), out_proj_covar=tensor([7.7735e-05, 1.4015e-04, 8.2286e-05, 1.8704e-04, 2.0822e-04, 1.1700e-04, 7.2703e-05, 8.5311e-05], device='cuda:1') 2023-03-07 13:59:44,324 INFO [train2.py:809] (1/4) Epoch 3, batch 750, loss[ctc_loss=0.2006, att_loss=0.2963, loss=0.2772, over 16111.00 frames. utt_duration=1536 frames, utt_pad_proportion=0.007249, over 42.00 utterances.], tot_loss[ctc_loss=0.2208, att_loss=0.3133, loss=0.2948, over 3205356.45 frames. utt_duration=1255 frames, utt_pad_proportion=0.05021, over 10232.01 utterances.], batch size: 42, lr: 3.29e-02, grad_scale: 8.0 2023-03-07 14:00:51,039 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8760.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:01:04,408 INFO [train2.py:809] (1/4) Epoch 3, batch 800, loss[ctc_loss=0.21, att_loss=0.2932, loss=0.2766, over 16410.00 frames. utt_duration=1493 frames, utt_pad_proportion=0.007027, over 44.00 utterances.], tot_loss[ctc_loss=0.2211, att_loss=0.3136, loss=0.2951, over 3221115.84 frames. utt_duration=1216 frames, utt_pad_proportion=0.06155, over 10612.30 utterances.], batch size: 44, lr: 3.28e-02, grad_scale: 8.0 2023-03-07 14:01:35,700 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5330, 2.7296, 4.5331, 3.7300, 3.4094, 4.2457, 4.4368, 4.3733], device='cuda:1'), covar=tensor([0.0105, 0.1165, 0.0124, 0.1230, 0.2117, 0.0424, 0.0189, 0.0264], device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0194, 0.0113, 0.0238, 0.0285, 0.0140, 0.0087, 0.0105], device='cuda:1'), out_proj_covar=tensor([7.6803e-05, 1.3872e-04, 8.2422e-05, 1.8405e-04, 2.0340e-04, 1.1388e-04, 7.0140e-05, 8.4120e-05], device='cuda:1') 2023-03-07 14:02:01,735 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.534e+02 3.936e+02 4.764e+02 5.876e+02 1.316e+03, threshold=9.528e+02, percent-clipped=1.0 2023-03-07 14:02:25,016 INFO [train2.py:809] (1/4) Epoch 3, batch 850, loss[ctc_loss=0.1775, att_loss=0.2798, loss=0.2593, over 14545.00 frames. utt_duration=1819 frames, utt_pad_proportion=0.03835, over 32.00 utterances.], tot_loss[ctc_loss=0.2189, att_loss=0.3113, loss=0.2929, over 3229439.05 frames. utt_duration=1233 frames, utt_pad_proportion=0.0579, over 10490.06 utterances.], batch size: 32, lr: 3.27e-02, grad_scale: 8.0 2023-03-07 14:02:27,496 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7570, 5.3376, 5.0993, 5.0076, 5.3723, 5.2107, 5.0112, 5.0182], device='cuda:1'), covar=tensor([0.1023, 0.0289, 0.0189, 0.0386, 0.0309, 0.0255, 0.0269, 0.0235], device='cuda:1'), in_proj_covar=tensor([0.0263, 0.0169, 0.0113, 0.0135, 0.0183, 0.0197, 0.0156, 0.0168], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-03-07 14:02:29,357 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8821.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:02:48,036 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2224, 4.6743, 4.9706, 4.3949, 2.3253, 3.8154, 2.8124, 3.8649], device='cuda:1'), covar=tensor([0.0192, 0.0174, 0.0526, 0.0312, 0.3760, 0.0282, 0.1257, 0.0736], device='cuda:1'), in_proj_covar=tensor([0.0088, 0.0076, 0.0176, 0.0110, 0.0219, 0.0086, 0.0161, 0.0152], device='cuda:1'), out_proj_covar=tensor([7.8020e-05, 6.7862e-05, 1.4100e-04, 8.5983e-05, 1.6286e-04, 7.3751e-05, 1.2708e-04, 1.2024e-04], device='cuda:1') 2023-03-07 14:03:21,465 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([1.5117, 1.3735, 1.7716, 1.2942, 1.8581, 3.2664, 1.8649, 1.1803], device='cuda:1'), covar=tensor([0.0971, 0.1446, 0.0837, 0.1848, 0.0872, 0.0177, 0.1538, 0.2760], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0056, 0.0053, 0.0080, 0.0060, 0.0052, 0.0067, 0.0089], device='cuda:1'), out_proj_covar=tensor([3.8802e-05, 3.7945e-05, 3.7533e-05, 4.5470e-05, 3.6237e-05, 3.0824e-05, 3.9818e-05, 5.5890e-05], device='cuda:1') 2023-03-07 14:03:43,847 INFO [train2.py:809] (1/4) Epoch 3, batch 900, loss[ctc_loss=0.2499, att_loss=0.3227, loss=0.3081, over 16809.00 frames. utt_duration=680.6 frames, utt_pad_proportion=0.1449, over 99.00 utterances.], tot_loss[ctc_loss=0.22, att_loss=0.3117, loss=0.2934, over 3233172.93 frames. utt_duration=1229 frames, utt_pad_proportion=0.06002, over 10532.59 utterances.], batch size: 99, lr: 3.26e-02, grad_scale: 8.0 2023-03-07 14:03:46,433 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7223, 4.5432, 4.5323, 3.4750, 4.5431, 3.8937, 4.0892, 2.1773], device='cuda:1'), covar=tensor([0.0186, 0.0128, 0.0313, 0.0472, 0.0112, 0.0225, 0.0256, 0.1679], device='cuda:1'), in_proj_covar=tensor([0.0036, 0.0036, 0.0034, 0.0053, 0.0036, 0.0042, 0.0049, 0.0084], device='cuda:1'), out_proj_covar=tensor([8.6495e-05, 1.0280e-04, 1.0207e-04, 1.2638e-04, 9.1782e-05, 1.2442e-04, 1.1888e-04, 1.8985e-04], device='cuda:1') 2023-03-07 14:04:39,902 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.583e+02 4.192e+02 5.104e+02 6.098e+02 1.636e+03, threshold=1.021e+03, percent-clipped=5.0 2023-03-07 14:05:03,067 INFO [train2.py:809] (1/4) Epoch 3, batch 950, loss[ctc_loss=0.2014, att_loss=0.2882, loss=0.2709, over 15895.00 frames. utt_duration=1632 frames, utt_pad_proportion=0.008615, over 39.00 utterances.], tot_loss[ctc_loss=0.2195, att_loss=0.3121, loss=0.2936, over 3250109.51 frames. utt_duration=1244 frames, utt_pad_proportion=0.05338, over 10466.95 utterances.], batch size: 39, lr: 3.26e-02, grad_scale: 8.0 2023-03-07 14:05:07,424 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8921.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:05:16,835 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8927.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:06:16,271 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 2023-03-07 14:06:23,052 INFO [train2.py:809] (1/4) Epoch 3, batch 1000, loss[ctc_loss=0.2199, att_loss=0.3128, loss=0.2943, over 17297.00 frames. utt_duration=1004 frames, utt_pad_proportion=0.05258, over 69.00 utterances.], tot_loss[ctc_loss=0.22, att_loss=0.313, loss=0.2944, over 3261729.68 frames. utt_duration=1226 frames, utt_pad_proportion=0.05645, over 10653.28 utterances.], batch size: 69, lr: 3.25e-02, grad_scale: 8.0 2023-03-07 14:06:23,866 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=8969.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:06:28,759 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5774, 4.7864, 4.5278, 4.9459, 4.9549, 4.4428, 4.4075, 4.7485], device='cuda:1'), covar=tensor([0.0085, 0.0142, 0.0099, 0.0073, 0.0065, 0.0099, 0.0252, 0.0123], device='cuda:1'), in_proj_covar=tensor([0.0040, 0.0040, 0.0042, 0.0030, 0.0029, 0.0036, 0.0054, 0.0046], device='cuda:1'), out_proj_covar=tensor([9.2646e-05, 9.5394e-05, 1.1411e-04, 7.5647e-05, 6.7805e-05, 9.2046e-05, 1.2682e-04, 1.1254e-04], device='cuda:1') 2023-03-07 14:06:48,513 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.91 vs. limit=5.0 2023-03-07 14:06:50,833 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8986.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:06:54,007 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8988.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:07:19,494 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.658e+02 4.100e+02 4.728e+02 5.740e+02 1.456e+03, threshold=9.457e+02, percent-clipped=2.0 2023-03-07 14:07:34,363 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9013.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:07:43,216 INFO [train2.py:809] (1/4) Epoch 3, batch 1050, loss[ctc_loss=0.1795, att_loss=0.2779, loss=0.2582, over 16121.00 frames. utt_duration=1537 frames, utt_pad_proportion=0.006114, over 42.00 utterances.], tot_loss[ctc_loss=0.2184, att_loss=0.312, loss=0.2933, over 3271097.82 frames. utt_duration=1235 frames, utt_pad_proportion=0.05355, over 10611.39 utterances.], batch size: 42, lr: 3.24e-02, grad_scale: 8.0 2023-03-07 14:08:27,904 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9047.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:08:38,073 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.50 vs. limit=5.0 2023-03-07 14:09:02,457 INFO [train2.py:809] (1/4) Epoch 3, batch 1100, loss[ctc_loss=0.1919, att_loss=0.3072, loss=0.2841, over 16964.00 frames. utt_duration=1358 frames, utt_pad_proportion=0.007684, over 50.00 utterances.], tot_loss[ctc_loss=0.2177, att_loss=0.312, loss=0.2931, over 3279695.13 frames. utt_duration=1242 frames, utt_pad_proportion=0.05016, over 10573.08 utterances.], batch size: 50, lr: 3.24e-02, grad_scale: 8.0 2023-03-07 14:09:11,163 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9074.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:09:48,692 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([1.7765, 3.6560, 2.9006, 2.8928, 3.7270, 3.1872, 1.8335, 4.4802], device='cuda:1'), covar=tensor([0.1545, 0.0297, 0.1307, 0.0856, 0.0453, 0.0844, 0.1371, 0.0079], device='cuda:1'), in_proj_covar=tensor([0.0109, 0.0074, 0.0121, 0.0100, 0.0082, 0.0118, 0.0105, 0.0057], device='cuda:1'), out_proj_covar=tensor([1.5446e-04, 1.2478e-04, 1.8535e-04, 1.4351e-04, 1.3858e-04, 1.7989e-04, 1.4572e-04, 9.1136e-05], device='cuda:1') 2023-03-07 14:09:58,863 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.446e+02 3.857e+02 5.076e+02 6.311e+02 2.003e+03, threshold=1.015e+03, percent-clipped=7.0 2023-03-07 14:10:18,403 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9116.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:10:22,916 INFO [train2.py:809] (1/4) Epoch 3, batch 1150, loss[ctc_loss=0.1654, att_loss=0.2933, loss=0.2677, over 16472.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.007, over 46.00 utterances.], tot_loss[ctc_loss=0.2177, att_loss=0.3122, loss=0.2933, over 3287197.67 frames. utt_duration=1249 frames, utt_pad_proportion=0.04621, over 10536.97 utterances.], batch size: 46, lr: 3.23e-02, grad_scale: 8.0 2023-03-07 14:10:58,433 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3595, 4.4988, 4.2337, 4.6573, 4.8762, 4.3500, 4.0498, 1.9374], device='cuda:1'), covar=tensor([0.0275, 0.0454, 0.0458, 0.0213, 0.0855, 0.0293, 0.0482, 0.4246], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0108, 0.0100, 0.0103, 0.0190, 0.0125, 0.0092, 0.0239], device='cuda:1'), out_proj_covar=tensor([9.6747e-05, 8.2096e-05, 8.1700e-05, 8.4596e-05, 1.6888e-04, 9.7415e-05, 7.9997e-05, 1.8836e-04], device='cuda:1') 2023-03-07 14:11:42,364 INFO [train2.py:809] (1/4) Epoch 3, batch 1200, loss[ctc_loss=0.2675, att_loss=0.323, loss=0.3119, over 16882.00 frames. utt_duration=1380 frames, utt_pad_proportion=0.006685, over 49.00 utterances.], tot_loss[ctc_loss=0.2179, att_loss=0.3113, loss=0.2926, over 3278237.77 frames. utt_duration=1247 frames, utt_pad_proportion=0.05073, over 10527.15 utterances.], batch size: 49, lr: 3.22e-02, grad_scale: 8.0 2023-03-07 14:12:41,118 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.934e+02 4.184e+02 5.248e+02 6.761e+02 1.156e+03, threshold=1.050e+03, percent-clipped=4.0 2023-03-07 14:13:06,018 INFO [train2.py:809] (1/4) Epoch 3, batch 1250, loss[ctc_loss=0.1525, att_loss=0.2642, loss=0.2419, over 15996.00 frames. utt_duration=1601 frames, utt_pad_proportion=0.008024, over 40.00 utterances.], tot_loss[ctc_loss=0.2193, att_loss=0.3122, loss=0.2936, over 3283551.88 frames. utt_duration=1223 frames, utt_pad_proportion=0.05642, over 10751.43 utterances.], batch size: 40, lr: 3.22e-02, grad_scale: 8.0 2023-03-07 14:13:26,003 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.3295, 4.3295, 4.5277, 4.2296, 1.6300, 4.6309, 2.1928, 3.3440], device='cuda:1'), covar=tensor([0.0472, 0.0157, 0.0508, 0.0364, 0.4285, 0.0133, 0.1644, 0.0777], device='cuda:1'), in_proj_covar=tensor([0.0093, 0.0080, 0.0176, 0.0109, 0.0218, 0.0088, 0.0167, 0.0155], device='cuda:1'), out_proj_covar=tensor([8.1864e-05, 7.2004e-05, 1.4256e-04, 8.4758e-05, 1.6390e-04, 7.5031e-05, 1.3187e-04, 1.2325e-04], device='cuda:1') 2023-03-07 14:14:30,174 INFO [train2.py:809] (1/4) Epoch 3, batch 1300, loss[ctc_loss=0.3126, att_loss=0.3595, loss=0.3501, over 13914.00 frames. utt_duration=385.6 frames, utt_pad_proportion=0.3295, over 145.00 utterances.], tot_loss[ctc_loss=0.2193, att_loss=0.312, loss=0.2935, over 3279895.19 frames. utt_duration=1199 frames, utt_pad_proportion=0.0639, over 10960.07 utterances.], batch size: 145, lr: 3.21e-02, grad_scale: 8.0 2023-03-07 14:14:52,975 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9283.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:15:29,122 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.817e+02 3.943e+02 4.874e+02 6.147e+02 1.582e+03, threshold=9.748e+02, percent-clipped=2.0 2023-03-07 14:15:54,117 INFO [train2.py:809] (1/4) Epoch 3, batch 1350, loss[ctc_loss=0.1852, att_loss=0.2854, loss=0.2654, over 16177.00 frames. utt_duration=1580 frames, utt_pad_proportion=0.006335, over 41.00 utterances.], tot_loss[ctc_loss=0.2179, att_loss=0.3109, loss=0.2923, over 3275649.56 frames. utt_duration=1205 frames, utt_pad_proportion=0.0637, over 10891.19 utterances.], batch size: 41, lr: 3.20e-02, grad_scale: 8.0 2023-03-07 14:15:58,379 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.15 vs. limit=2.0 2023-03-07 14:16:32,146 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9522, 6.0666, 5.4291, 6.0401, 5.6772, 5.4971, 5.5133, 5.4869], device='cuda:1'), covar=tensor([0.0932, 0.0687, 0.0686, 0.0522, 0.0551, 0.0993, 0.1851, 0.1930], device='cuda:1'), in_proj_covar=tensor([0.0260, 0.0302, 0.0248, 0.0224, 0.0209, 0.0302, 0.0317, 0.0308], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-07 14:16:32,213 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9342.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:16:52,839 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7754, 4.5249, 4.5516, 3.5764, 4.5553, 4.0056, 4.3003, 2.2844], device='cuda:1'), covar=tensor([0.0089, 0.0097, 0.0171, 0.0355, 0.0130, 0.0172, 0.0176, 0.1459], device='cuda:1'), in_proj_covar=tensor([0.0037, 0.0039, 0.0035, 0.0057, 0.0039, 0.0044, 0.0050, 0.0089], device='cuda:1'), out_proj_covar=tensor([9.1877e-05, 1.1246e-04, 1.0904e-04, 1.3998e-04, 1.0531e-04, 1.3523e-04, 1.2469e-04, 2.0680e-04], device='cuda:1') 2023-03-07 14:17:13,536 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=5.95 vs. limit=5.0 2023-03-07 14:17:17,853 INFO [train2.py:809] (1/4) Epoch 3, batch 1400, loss[ctc_loss=0.2187, att_loss=0.3216, loss=0.301, over 16873.00 frames. utt_duration=1379 frames, utt_pad_proportion=0.007082, over 49.00 utterances.], tot_loss[ctc_loss=0.2166, att_loss=0.3098, loss=0.2912, over 3275076.74 frames. utt_duration=1216 frames, utt_pad_proportion=0.06055, over 10783.66 utterances.], batch size: 49, lr: 3.20e-02, grad_scale: 8.0 2023-03-07 14:17:18,078 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9369.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:17:41,159 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3118, 2.8799, 4.8708, 3.7508, 3.2032, 4.2468, 4.5870, 4.4314], device='cuda:1'), covar=tensor([0.0209, 0.1359, 0.0269, 0.1512, 0.2583, 0.0582, 0.0256, 0.0503], device='cuda:1'), in_proj_covar=tensor([0.0103, 0.0202, 0.0112, 0.0251, 0.0296, 0.0147, 0.0089, 0.0112], device='cuda:1'), out_proj_covar=tensor([8.2181e-05, 1.4662e-04, 8.3464e-05, 1.9642e-04, 2.1348e-04, 1.1908e-04, 7.3193e-05, 9.0857e-05], device='cuda:1') 2023-03-07 14:18:16,803 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.777e+02 3.704e+02 4.754e+02 6.457e+02 1.261e+03, threshold=9.507e+02, percent-clipped=3.0 2023-03-07 14:18:33,986 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9414.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:18:37,104 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9416.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:18:41,730 INFO [train2.py:809] (1/4) Epoch 3, batch 1450, loss[ctc_loss=0.1964, att_loss=0.3024, loss=0.2812, over 16949.00 frames. utt_duration=1357 frames, utt_pad_proportion=0.008488, over 50.00 utterances.], tot_loss[ctc_loss=0.2182, att_loss=0.3114, loss=0.2927, over 3270997.14 frames. utt_duration=1162 frames, utt_pad_proportion=0.07681, over 11277.89 utterances.], batch size: 50, lr: 3.19e-02, grad_scale: 8.0 2023-03-07 14:18:47,835 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-03-07 14:18:58,599 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4521, 5.0508, 4.8017, 4.9146, 5.1684, 5.0572, 4.9288, 4.5998], device='cuda:1'), covar=tensor([0.1330, 0.0438, 0.0264, 0.0422, 0.0247, 0.0310, 0.0264, 0.0349], device='cuda:1'), in_proj_covar=tensor([0.0270, 0.0173, 0.0112, 0.0137, 0.0181, 0.0196, 0.0159, 0.0169], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-03-07 14:19:58,095 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=9464.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:20:06,077 INFO [train2.py:809] (1/4) Epoch 3, batch 1500, loss[ctc_loss=0.2309, att_loss=0.3161, loss=0.299, over 17119.00 frames. utt_duration=1224 frames, utt_pad_proportion=0.01438, over 56.00 utterances.], tot_loss[ctc_loss=0.216, att_loss=0.3103, loss=0.2915, over 3266353.49 frames. utt_duration=1181 frames, utt_pad_proportion=0.07291, over 11081.24 utterances.], batch size: 56, lr: 3.18e-02, grad_scale: 8.0 2023-03-07 14:20:15,975 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9475.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:21:03,451 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.770e+02 4.105e+02 5.004e+02 6.164e+02 1.572e+03, threshold=1.001e+03, percent-clipped=6.0 2023-03-07 14:21:27,813 INFO [train2.py:809] (1/4) Epoch 3, batch 1550, loss[ctc_loss=0.2195, att_loss=0.3203, loss=0.3001, over 16872.00 frames. utt_duration=1379 frames, utt_pad_proportion=0.007346, over 49.00 utterances.], tot_loss[ctc_loss=0.2158, att_loss=0.3106, loss=0.2917, over 3271136.74 frames. utt_duration=1178 frames, utt_pad_proportion=0.07243, over 11119.01 utterances.], batch size: 49, lr: 3.18e-02, grad_scale: 8.0 2023-03-07 14:22:50,221 INFO [train2.py:809] (1/4) Epoch 3, batch 1600, loss[ctc_loss=0.2671, att_loss=0.3513, loss=0.3345, over 16946.00 frames. utt_duration=1357 frames, utt_pad_proportion=0.008722, over 50.00 utterances.], tot_loss[ctc_loss=0.2149, att_loss=0.3102, loss=0.2911, over 3281328.44 frames. utt_duration=1208 frames, utt_pad_proportion=0.06274, over 10881.96 utterances.], batch size: 50, lr: 3.17e-02, grad_scale: 8.0 2023-03-07 14:23:04,504 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-03-07 14:23:12,997 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9583.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:23:14,650 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9584.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:23:49,434 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.439e+02 4.014e+02 4.915e+02 6.528e+02 1.408e+03, threshold=9.829e+02, percent-clipped=8.0 2023-03-07 14:24:03,224 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=2.00 vs. limit=2.0 2023-03-07 14:24:09,259 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5769, 5.1843, 5.2540, 5.1870, 2.0473, 2.9925, 5.2832, 3.9395], device='cuda:1'), covar=tensor([0.0414, 0.0153, 0.0096, 0.0215, 0.8377, 0.2165, 0.0149, 0.1988], device='cuda:1'), in_proj_covar=tensor([0.0228, 0.0132, 0.0148, 0.0175, 0.0398, 0.0284, 0.0149, 0.0232], device='cuda:1'), out_proj_covar=tensor([1.2572e-04, 7.0230e-05, 7.7259e-05, 8.6575e-05, 1.9705e-04, 1.4267e-04, 7.6208e-05, 1.3205e-04], device='cuda:1') 2023-03-07 14:24:13,627 INFO [train2.py:809] (1/4) Epoch 3, batch 1650, loss[ctc_loss=0.2097, att_loss=0.3117, loss=0.2913, over 17381.00 frames. utt_duration=1180 frames, utt_pad_proportion=0.01915, over 59.00 utterances.], tot_loss[ctc_loss=0.2131, att_loss=0.309, loss=0.2898, over 3279992.44 frames. utt_duration=1228 frames, utt_pad_proportion=0.05823, over 10692.84 utterances.], batch size: 59, lr: 3.16e-02, grad_scale: 8.0 2023-03-07 14:24:33,198 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=9631.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:24:44,403 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.67 vs. limit=2.0 2023-03-07 14:24:52,648 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9642.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:24:57,530 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9645.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:25:17,993 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9658.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:25:35,001 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=2.10 vs. limit=2.0 2023-03-07 14:25:35,212 INFO [train2.py:809] (1/4) Epoch 3, batch 1700, loss[ctc_loss=0.2858, att_loss=0.345, loss=0.3331, over 13990.00 frames. utt_duration=384.9 frames, utt_pad_proportion=0.3307, over 146.00 utterances.], tot_loss[ctc_loss=0.2114, att_loss=0.3074, loss=0.2882, over 3275693.58 frames. utt_duration=1247 frames, utt_pad_proportion=0.05495, over 10517.92 utterances.], batch size: 146, lr: 3.16e-02, grad_scale: 8.0 2023-03-07 14:25:35,560 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9669.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:25:49,873 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6456, 5.0023, 4.6496, 4.9136, 5.1066, 4.9775, 4.5809, 4.9425], device='cuda:1'), covar=tensor([0.0101, 0.0194, 0.0119, 0.0103, 0.0116, 0.0089, 0.0290, 0.0170], device='cuda:1'), in_proj_covar=tensor([0.0044, 0.0044, 0.0044, 0.0032, 0.0032, 0.0037, 0.0058, 0.0051], device='cuda:1'), out_proj_covar=tensor([1.1065e-04, 1.0897e-04, 1.2741e-04, 8.7224e-05, 7.8429e-05, 1.0154e-04, 1.4708e-04, 1.3239e-04], device='cuda:1') 2023-03-07 14:26:08,858 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=9690.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:26:33,373 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.300e+02 3.813e+02 4.618e+02 5.851e+02 1.256e+03, threshold=9.236e+02, percent-clipped=5.0 2023-03-07 14:26:40,584 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.44 vs. limit=5.0 2023-03-07 14:26:54,272 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=9717.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:26:57,505 INFO [train2.py:809] (1/4) Epoch 3, batch 1750, loss[ctc_loss=0.1882, att_loss=0.2986, loss=0.2765, over 16773.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.006096, over 48.00 utterances.], tot_loss[ctc_loss=0.2118, att_loss=0.3072, loss=0.2882, over 3266440.41 frames. utt_duration=1224 frames, utt_pad_proportion=0.06338, over 10689.00 utterances.], batch size: 48, lr: 3.15e-02, grad_scale: 8.0 2023-03-07 14:26:57,971 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9719.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:27:37,592 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9743.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:28:19,378 INFO [train2.py:809] (1/4) Epoch 3, batch 1800, loss[ctc_loss=0.1706, att_loss=0.2869, loss=0.2636, over 16417.00 frames. utt_duration=1494 frames, utt_pad_proportion=0.006573, over 44.00 utterances.], tot_loss[ctc_loss=0.2118, att_loss=0.3079, loss=0.2887, over 3277816.03 frames. utt_duration=1229 frames, utt_pad_proportion=0.05874, over 10682.51 utterances.], batch size: 44, lr: 3.14e-02, grad_scale: 8.0 2023-03-07 14:28:21,133 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9770.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:29:17,431 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.661e+02 4.029e+02 5.153e+02 6.488e+02 1.407e+03, threshold=1.031e+03, percent-clipped=9.0 2023-03-07 14:29:17,894 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9804.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:29:41,466 INFO [train2.py:809] (1/4) Epoch 3, batch 1850, loss[ctc_loss=0.1755, att_loss=0.2825, loss=0.2611, over 16185.00 frames. utt_duration=1581 frames, utt_pad_proportion=0.005783, over 41.00 utterances.], tot_loss[ctc_loss=0.2115, att_loss=0.3087, loss=0.2893, over 3275623.52 frames. utt_duration=1233 frames, utt_pad_proportion=0.0582, over 10639.96 utterances.], batch size: 41, lr: 3.14e-02, grad_scale: 8.0 2023-03-07 14:30:17,843 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7841, 5.1786, 4.8412, 4.8397, 5.2633, 5.1879, 4.9423, 4.7347], device='cuda:1'), covar=tensor([0.1112, 0.0326, 0.0261, 0.0429, 0.0256, 0.0227, 0.0246, 0.0285], device='cuda:1'), in_proj_covar=tensor([0.0279, 0.0180, 0.0119, 0.0140, 0.0189, 0.0203, 0.0164, 0.0175], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-07 14:30:38,483 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.8232, 4.4242, 4.3682, 4.1397, 2.2566, 2.4789, 4.4350, 3.5544], device='cuda:1'), covar=tensor([0.0715, 0.0166, 0.0182, 0.0412, 0.7390, 0.2489, 0.0179, 0.2412], device='cuda:1'), in_proj_covar=tensor([0.0234, 0.0133, 0.0156, 0.0173, 0.0404, 0.0287, 0.0152, 0.0238], device='cuda:1'), out_proj_covar=tensor([1.2841e-04, 6.9624e-05, 8.1379e-05, 8.5907e-05, 1.9867e-04, 1.4428e-04, 7.6913e-05, 1.3444e-04], device='cuda:1') 2023-03-07 14:31:02,973 INFO [train2.py:809] (1/4) Epoch 3, batch 1900, loss[ctc_loss=0.1942, att_loss=0.3078, loss=0.285, over 16548.00 frames. utt_duration=1473 frames, utt_pad_proportion=0.005732, over 45.00 utterances.], tot_loss[ctc_loss=0.212, att_loss=0.3089, loss=0.2895, over 3273924.88 frames. utt_duration=1226 frames, utt_pad_proportion=0.05961, over 10695.71 utterances.], batch size: 45, lr: 3.13e-02, grad_scale: 8.0 2023-03-07 14:31:42,850 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1171, 4.8543, 4.7983, 3.8076, 4.9302, 4.1815, 4.5844, 2.7213], device='cuda:1'), covar=tensor([0.0089, 0.0100, 0.0188, 0.0403, 0.0100, 0.0149, 0.0163, 0.1198], device='cuda:1'), in_proj_covar=tensor([0.0039, 0.0041, 0.0036, 0.0062, 0.0039, 0.0046, 0.0053, 0.0090], device='cuda:1'), out_proj_covar=tensor([9.8400e-05, 1.2045e-04, 1.1446e-04, 1.5825e-04, 1.0496e-04, 1.4727e-04, 1.3993e-04, 2.1698e-04], device='cuda:1') 2023-03-07 14:32:00,232 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.657e+02 4.209e+02 5.043e+02 6.488e+02 1.113e+03, threshold=1.009e+03, percent-clipped=3.0 2023-03-07 14:32:23,705 INFO [train2.py:809] (1/4) Epoch 3, batch 1950, loss[ctc_loss=0.261, att_loss=0.345, loss=0.3282, over 17056.00 frames. utt_duration=1313 frames, utt_pad_proportion=0.008003, over 52.00 utterances.], tot_loss[ctc_loss=0.2093, att_loss=0.3071, loss=0.2876, over 3271574.69 frames. utt_duration=1237 frames, utt_pad_proportion=0.05846, over 10589.05 utterances.], batch size: 52, lr: 3.13e-02, grad_scale: 8.0 2023-03-07 14:32:58,027 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9940.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:33:25,961 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.1404, 1.8906, 1.8072, 1.7022, 1.0616, 2.9823, 1.2176, 1.2307], device='cuda:1'), covar=tensor([0.0551, 0.0882, 0.0754, 0.1721, 0.2042, 0.0340, 0.2034, 0.3374], device='cuda:1'), in_proj_covar=tensor([0.0060, 0.0050, 0.0050, 0.0068, 0.0058, 0.0052, 0.0061, 0.0082], device='cuda:1'), out_proj_covar=tensor([3.6361e-05, 3.3121e-05, 3.3180e-05, 4.1868e-05, 3.8581e-05, 2.8785e-05, 3.8394e-05, 5.3392e-05], device='cuda:1') 2023-03-07 14:33:44,093 INFO [train2.py:809] (1/4) Epoch 3, batch 2000, loss[ctc_loss=0.1964, att_loss=0.2937, loss=0.2743, over 16115.00 frames. utt_duration=1537 frames, utt_pad_proportion=0.006772, over 42.00 utterances.], tot_loss[ctc_loss=0.2072, att_loss=0.3058, loss=0.2861, over 3277639.56 frames. utt_duration=1255 frames, utt_pad_proportion=0.0519, over 10462.18 utterances.], batch size: 42, lr: 3.12e-02, grad_scale: 8.0 2023-03-07 14:34:19,238 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2679, 0.5627, 1.5375, 2.1724, 0.6917, 1.9565, 1.3921, 2.0210], device='cuda:1'), covar=tensor([0.0129, 0.1400, 0.1244, 0.0541, 0.1071, 0.0574, 0.0746, 0.0461], device='cuda:1'), in_proj_covar=tensor([0.0066, 0.0070, 0.0075, 0.0069, 0.0070, 0.0068, 0.0073, 0.0076], device='cuda:1'), out_proj_covar=tensor([3.4359e-05, 4.6976e-05, 4.7213e-05, 3.9309e-05, 4.1111e-05, 3.9396e-05, 4.2572e-05, 4.0803e-05], device='cuda:1') 2023-03-07 14:34:35,659 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-03-07 14:34:46,866 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.957e+02 4.346e+02 5.150e+02 6.489e+02 1.727e+03, threshold=1.030e+03, percent-clipped=5.0 2023-03-07 14:35:02,996 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10014.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:35:09,423 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1475, 2.7516, 3.3308, 2.5796, 3.3149, 4.4615, 4.1448, 3.1575], device='cuda:1'), covar=tensor([0.0357, 0.1133, 0.0749, 0.1340, 0.0910, 0.0177, 0.0388, 0.1225], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0157, 0.0133, 0.0153, 0.0169, 0.0098, 0.0113, 0.0159], 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') 2023-03-07 14:35:10,655 INFO [train2.py:809] (1/4) Epoch 3, batch 2050, loss[ctc_loss=0.212, att_loss=0.3237, loss=0.3013, over 16757.00 frames. utt_duration=1398 frames, utt_pad_proportion=0.00694, over 48.00 utterances.], tot_loss[ctc_loss=0.2075, att_loss=0.306, loss=0.2863, over 3275066.17 frames. utt_duration=1251 frames, utt_pad_proportion=0.05497, over 10483.38 utterances.], batch size: 48, lr: 3.11e-02, grad_scale: 8.0 2023-03-07 14:35:33,500 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1772, 4.6020, 4.4915, 4.7194, 1.8506, 4.5825, 2.3521, 2.7964], device='cuda:1'), covar=tensor([0.0235, 0.0164, 0.0583, 0.0200, 0.4134, 0.0159, 0.1604, 0.1199], device='cuda:1'), in_proj_covar=tensor([0.0089, 0.0080, 0.0183, 0.0103, 0.0215, 0.0085, 0.0171, 0.0157], device='cuda:1'), out_proj_covar=tensor([7.9083e-05, 7.4276e-05, 1.5010e-04, 8.2066e-05, 1.6459e-04, 7.5902e-05, 1.3604e-04, 1.2649e-04], device='cuda:1') 2023-03-07 14:35:43,996 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10039.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:36:21,340 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10062.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 14:36:26,891 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-03-07 14:36:27,589 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3390, 4.7623, 4.2973, 4.8413, 4.9181, 4.3607, 4.1782, 4.6418], device='cuda:1'), covar=tensor([0.0101, 0.0153, 0.0126, 0.0089, 0.0095, 0.0136, 0.0316, 0.0151], device='cuda:1'), in_proj_covar=tensor([0.0043, 0.0042, 0.0043, 0.0031, 0.0030, 0.0037, 0.0057, 0.0050], device='cuda:1'), out_proj_covar=tensor([1.1020e-04, 1.0646e-04, 1.2714e-04, 8.5562e-05, 7.7190e-05, 1.0460e-04, 1.4836e-04, 1.3361e-04], device='cuda:1') 2023-03-07 14:36:32,072 INFO [train2.py:809] (1/4) Epoch 3, batch 2100, loss[ctc_loss=0.1464, att_loss=0.2664, loss=0.2424, over 12373.00 frames. utt_duration=1835 frames, utt_pad_proportion=0.131, over 27.00 utterances.], tot_loss[ctc_loss=0.2074, att_loss=0.3047, loss=0.2852, over 3258452.77 frames. utt_duration=1246 frames, utt_pad_proportion=0.06044, over 10476.80 utterances.], batch size: 27, lr: 3.11e-02, grad_scale: 8.0 2023-03-07 14:36:33,999 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10070.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:36:49,458 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3653, 4.4719, 4.3416, 4.5810, 1.7938, 4.5208, 2.1186, 2.2558], device='cuda:1'), covar=tensor([0.0150, 0.0139, 0.0633, 0.0258, 0.3762, 0.0169, 0.1705, 0.1373], device='cuda:1'), in_proj_covar=tensor([0.0090, 0.0081, 0.0185, 0.0105, 0.0217, 0.0087, 0.0172, 0.0159], device='cuda:1'), out_proj_covar=tensor([7.9854e-05, 7.4540e-05, 1.5171e-04, 8.2693e-05, 1.6647e-04, 7.7362e-05, 1.3639e-04, 1.2835e-04], device='cuda:1') 2023-03-07 14:37:21,244 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10099.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:37:23,012 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10100.0, num_to_drop=1, layers_to_drop={2} 2023-03-07 14:37:28,711 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.690e+02 4.150e+02 4.955e+02 6.007e+02 1.097e+03, threshold=9.911e+02, percent-clipped=1.0 2023-03-07 14:37:50,924 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=10118.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:37:52,389 INFO [train2.py:809] (1/4) Epoch 3, batch 2150, loss[ctc_loss=0.2016, att_loss=0.3175, loss=0.2943, over 16780.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.006067, over 48.00 utterances.], tot_loss[ctc_loss=0.2072, att_loss=0.3043, loss=0.2849, over 3257722.70 frames. utt_duration=1260 frames, utt_pad_proportion=0.05536, over 10355.58 utterances.], batch size: 48, lr: 3.10e-02, grad_scale: 8.0 2023-03-07 14:37:58,687 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10123.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 14:39:12,830 INFO [train2.py:809] (1/4) Epoch 3, batch 2200, loss[ctc_loss=0.1742, att_loss=0.2887, loss=0.2658, over 15371.00 frames. utt_duration=1758 frames, utt_pad_proportion=0.0103, over 35.00 utterances.], tot_loss[ctc_loss=0.2071, att_loss=0.3047, loss=0.2852, over 3264789.88 frames. utt_duration=1261 frames, utt_pad_proportion=0.05323, over 10372.32 utterances.], batch size: 35, lr: 3.09e-02, grad_scale: 8.0 2023-03-07 14:39:35,153 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-03-07 14:39:38,337 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-03-07 14:40:07,300 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.4017, 3.8455, 3.3047, 3.5241, 4.0522, 3.6839, 2.6951, 4.3978], device='cuda:1'), covar=tensor([0.1370, 0.0357, 0.1048, 0.0637, 0.0332, 0.0837, 0.0986, 0.0177], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0081, 0.0135, 0.0109, 0.0091, 0.0131, 0.0113, 0.0069], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:1') 2023-03-07 14:40:09,925 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.374e+02 4.147e+02 5.296e+02 6.795e+02 2.410e+03, threshold=1.059e+03, percent-clipped=6.0 2023-03-07 14:40:33,285 INFO [train2.py:809] (1/4) Epoch 3, batch 2250, loss[ctc_loss=0.1904, att_loss=0.2804, loss=0.2624, over 15381.00 frames. utt_duration=1759 frames, utt_pad_proportion=0.01057, over 35.00 utterances.], tot_loss[ctc_loss=0.207, att_loss=0.3047, loss=0.2851, over 3267335.60 frames. utt_duration=1248 frames, utt_pad_proportion=0.05519, over 10480.76 utterances.], batch size: 35, lr: 3.09e-02, grad_scale: 8.0 2023-03-07 14:41:07,811 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10240.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:41:24,204 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([1.7232, 2.9248, 2.2759, 1.6355, 0.7330, 2.7755, 1.8712, 1.5455], device='cuda:1'), covar=tensor([0.0476, 0.0353, 0.0721, 0.2356, 0.1758, 0.0319, 0.0967, 0.2443], device='cuda:1'), in_proj_covar=tensor([0.0063, 0.0054, 0.0056, 0.0073, 0.0058, 0.0058, 0.0065, 0.0088], device='cuda:1'), out_proj_covar=tensor([3.7570e-05, 3.4742e-05, 3.7048e-05, 4.4682e-05, 3.9182e-05, 3.1820e-05, 4.0034e-05, 5.7269e-05], device='cuda:1') 2023-03-07 14:41:52,778 INFO [train2.py:809] (1/4) Epoch 3, batch 2300, loss[ctc_loss=0.2034, att_loss=0.3104, loss=0.289, over 16414.00 frames. utt_duration=1494 frames, utt_pad_proportion=0.006649, over 44.00 utterances.], tot_loss[ctc_loss=0.2059, att_loss=0.3041, loss=0.2844, over 3269087.15 frames. utt_duration=1251 frames, utt_pad_proportion=0.05517, over 10468.05 utterances.], batch size: 44, lr: 3.08e-02, grad_scale: 8.0 2023-03-07 14:42:10,675 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.8645, 4.9891, 5.3121, 5.3974, 4.9694, 5.6572, 5.0056, 5.8013], device='cuda:1'), covar=tensor([0.0434, 0.0608, 0.0491, 0.0444, 0.1889, 0.0644, 0.0528, 0.0465], device='cuda:1'), in_proj_covar=tensor([0.0367, 0.0256, 0.0245, 0.0298, 0.0438, 0.0242, 0.0220, 0.0279], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0004, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-03-07 14:42:23,458 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=10288.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:42:34,030 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3584, 4.8343, 4.6140, 4.4749, 4.9267, 4.7031, 4.5620, 4.4188], device='cuda:1'), covar=tensor([0.1089, 0.0390, 0.0238, 0.0599, 0.0269, 0.0284, 0.0265, 0.0312], device='cuda:1'), in_proj_covar=tensor([0.0294, 0.0187, 0.0126, 0.0152, 0.0199, 0.0217, 0.0169, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-07 14:42:49,026 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.993e+02 3.999e+02 4.937e+02 6.652e+02 1.498e+03, threshold=9.874e+02, percent-clipped=4.0 2023-03-07 14:43:04,901 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10314.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:43:12,504 INFO [train2.py:809] (1/4) Epoch 3, batch 2350, loss[ctc_loss=0.2841, att_loss=0.3549, loss=0.3407, over 13561.00 frames. utt_duration=373.1 frames, utt_pad_proportion=0.3512, over 146.00 utterances.], tot_loss[ctc_loss=0.2065, att_loss=0.3047, loss=0.2851, over 3266611.87 frames. utt_duration=1240 frames, utt_pad_proportion=0.05903, over 10552.92 utterances.], batch size: 146, lr: 3.08e-02, grad_scale: 16.0 2023-03-07 14:44:21,300 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=10362.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:44:32,342 INFO [train2.py:809] (1/4) Epoch 3, batch 2400, loss[ctc_loss=0.3156, att_loss=0.37, loss=0.3591, over 14232.00 frames. utt_duration=391.5 frames, utt_pad_proportion=0.318, over 146.00 utterances.], tot_loss[ctc_loss=0.2069, att_loss=0.3052, loss=0.2855, over 3262239.23 frames. utt_duration=1223 frames, utt_pad_proportion=0.0644, over 10684.54 utterances.], batch size: 146, lr: 3.07e-02, grad_scale: 16.0 2023-03-07 14:44:50,718 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2464, 2.0966, 2.8110, 4.2734, 4.3204, 4.6228, 2.9511, 2.0760], device='cuda:1'), covar=tensor([0.0202, 0.2875, 0.1714, 0.0485, 0.0131, 0.0076, 0.1628, 0.2693], device='cuda:1'), in_proj_covar=tensor([0.0113, 0.0175, 0.0166, 0.0116, 0.0089, 0.0094, 0.0173, 0.0161], device='cuda:1'), out_proj_covar=tensor([1.1125e-04, 1.6679e-04, 1.6352e-04, 1.3089e-04, 9.3001e-05, 9.2469e-05, 1.7087e-04, 1.5424e-04], device='cuda:1') 2023-03-07 14:44:56,799 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7112, 5.1169, 4.9199, 4.8407, 5.4195, 5.1594, 4.9860, 4.8972], device='cuda:1'), covar=tensor([0.1233, 0.0513, 0.0285, 0.0684, 0.0225, 0.0243, 0.0260, 0.0263], device='cuda:1'), in_proj_covar=tensor([0.0290, 0.0182, 0.0123, 0.0149, 0.0195, 0.0213, 0.0167, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-07 14:45:14,386 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10395.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 14:45:20,856 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10399.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:45:28,362 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.513e+02 3.987e+02 4.935e+02 6.057e+02 1.238e+03, threshold=9.870e+02, percent-clipped=1.0 2023-03-07 14:45:36,577 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10409.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:45:50,462 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10418.0, num_to_drop=1, layers_to_drop={2} 2023-03-07 14:45:51,371 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.45 vs. limit=5.0 2023-03-07 14:45:51,949 INFO [train2.py:809] (1/4) Epoch 3, batch 2450, loss[ctc_loss=0.2451, att_loss=0.3355, loss=0.3174, over 17042.00 frames. utt_duration=1313 frames, utt_pad_proportion=0.008541, over 52.00 utterances.], tot_loss[ctc_loss=0.2063, att_loss=0.3049, loss=0.2851, over 3256726.57 frames. utt_duration=1225 frames, utt_pad_proportion=0.06417, over 10645.91 utterances.], batch size: 52, lr: 3.06e-02, grad_scale: 16.0 2023-03-07 14:46:38,267 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=10447.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:46:49,300 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10454.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:47:07,997 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1250, 4.1953, 4.3029, 4.2041, 1.9928, 4.4036, 2.1817, 2.9414], device='cuda:1'), covar=tensor([0.0220, 0.0149, 0.0567, 0.0323, 0.3769, 0.0147, 0.1719, 0.1086], device='cuda:1'), in_proj_covar=tensor([0.0091, 0.0086, 0.0193, 0.0105, 0.0220, 0.0087, 0.0182, 0.0169], device='cuda:1'), out_proj_covar=tensor([8.1569e-05, 7.8358e-05, 1.5946e-04, 8.4395e-05, 1.7114e-04, 7.8100e-05, 1.4497e-04, 1.3656e-04], device='cuda:1') 2023-03-07 14:47:12,264 INFO [train2.py:809] (1/4) Epoch 3, batch 2500, loss[ctc_loss=0.178, att_loss=0.2649, loss=0.2475, over 15626.00 frames. utt_duration=1691 frames, utt_pad_proportion=0.01, over 37.00 utterances.], tot_loss[ctc_loss=0.205, att_loss=0.3041, loss=0.2843, over 3261437.84 frames. utt_duration=1240 frames, utt_pad_proportion=0.05774, over 10530.68 utterances.], batch size: 37, lr: 3.06e-02, grad_scale: 16.0 2023-03-07 14:47:14,182 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10470.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:48:09,700 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.331e+02 4.171e+02 5.202e+02 6.320e+02 1.110e+03, threshold=1.040e+03, percent-clipped=6.0 2023-03-07 14:48:16,583 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5591, 5.2547, 5.1130, 5.3548, 4.7402, 4.9411, 5.4772, 5.2884], device='cuda:1'), covar=tensor([0.0360, 0.0202, 0.0284, 0.0092, 0.0304, 0.0152, 0.0200, 0.0106], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0120, 0.0143, 0.0089, 0.0133, 0.0101, 0.0115, 0.0105], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:1') 2023-03-07 14:48:27,706 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10515.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 14:48:33,427 INFO [train2.py:809] (1/4) Epoch 3, batch 2550, loss[ctc_loss=0.2353, att_loss=0.3221, loss=0.3048, over 16313.00 frames. utt_duration=1452 frames, utt_pad_proportion=0.007022, over 45.00 utterances.], tot_loss[ctc_loss=0.2048, att_loss=0.3041, loss=0.2842, over 3271855.19 frames. utt_duration=1235 frames, utt_pad_proportion=0.05587, over 10606.00 utterances.], batch size: 45, lr: 3.05e-02, grad_scale: 16.0 2023-03-07 14:49:54,161 INFO [train2.py:809] (1/4) Epoch 3, batch 2600, loss[ctc_loss=0.2158, att_loss=0.3228, loss=0.3014, over 17033.00 frames. utt_duration=1338 frames, utt_pad_proportion=0.007016, over 51.00 utterances.], tot_loss[ctc_loss=0.204, att_loss=0.3043, loss=0.2842, over 3277798.58 frames. utt_duration=1229 frames, utt_pad_proportion=0.05605, over 10677.71 utterances.], batch size: 51, lr: 3.05e-02, grad_scale: 16.0 2023-03-07 14:50:50,041 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.348e+02 3.926e+02 4.748e+02 6.116e+02 1.074e+03, threshold=9.496e+02, percent-clipped=2.0 2023-03-07 14:50:50,233 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.4666, 5.6710, 4.9350, 5.7626, 5.3168, 5.0435, 5.0251, 5.0661], device='cuda:1'), covar=tensor([0.1290, 0.0909, 0.0790, 0.0571, 0.0677, 0.1267, 0.2311, 0.2204], device='cuda:1'), in_proj_covar=tensor([0.0269, 0.0311, 0.0248, 0.0244, 0.0222, 0.0312, 0.0332, 0.0326], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-07 14:51:00,789 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-03-07 14:51:04,300 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5742, 4.9791, 4.7668, 5.0726, 4.4957, 4.8067, 5.2859, 5.0440], device='cuda:1'), covar=tensor([0.0296, 0.0244, 0.0391, 0.0133, 0.0314, 0.0166, 0.0212, 0.0132], device='cuda:1'), in_proj_covar=tensor([0.0132, 0.0117, 0.0141, 0.0088, 0.0132, 0.0097, 0.0113, 0.0103], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:1') 2023-03-07 14:51:13,370 INFO [train2.py:809] (1/4) Epoch 3, batch 2650, loss[ctc_loss=0.2059, att_loss=0.321, loss=0.298, over 16876.00 frames. utt_duration=1379 frames, utt_pad_proportion=0.006322, over 49.00 utterances.], tot_loss[ctc_loss=0.2034, att_loss=0.3039, loss=0.2838, over 3282362.81 frames. utt_duration=1250 frames, utt_pad_proportion=0.05027, over 10516.41 utterances.], batch size: 49, lr: 3.04e-02, grad_scale: 16.0 2023-03-07 14:51:56,596 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.64 vs. limit=2.0 2023-03-07 14:51:57,122 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6104, 4.6011, 4.6114, 3.0062, 4.7413, 3.8695, 4.2302, 2.3072], device='cuda:1'), covar=tensor([0.0083, 0.0063, 0.0130, 0.0573, 0.0052, 0.0158, 0.0161, 0.1315], device='cuda:1'), in_proj_covar=tensor([0.0041, 0.0041, 0.0037, 0.0066, 0.0039, 0.0049, 0.0056, 0.0090], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-07 14:52:32,126 INFO [train2.py:809] (1/4) Epoch 3, batch 2700, loss[ctc_loss=0.1959, att_loss=0.2917, loss=0.2725, over 15995.00 frames. utt_duration=1601 frames, utt_pad_proportion=0.008576, over 40.00 utterances.], tot_loss[ctc_loss=0.2035, att_loss=0.3042, loss=0.2841, over 3291631.13 frames. utt_duration=1266 frames, utt_pad_proportion=0.04433, over 10410.73 utterances.], batch size: 40, lr: 3.03e-02, grad_scale: 16.0 2023-03-07 14:53:14,762 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10695.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:53:28,798 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.793e+02 4.160e+02 5.458e+02 6.683e+02 1.246e+03, threshold=1.092e+03, percent-clipped=5.0 2023-03-07 14:53:51,022 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10718.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 14:53:52,190 INFO [train2.py:809] (1/4) Epoch 3, batch 2750, loss[ctc_loss=0.2137, att_loss=0.303, loss=0.2852, over 16537.00 frames. utt_duration=1472 frames, utt_pad_proportion=0.005721, over 45.00 utterances.], tot_loss[ctc_loss=0.2051, att_loss=0.3052, loss=0.2852, over 3291894.12 frames. utt_duration=1247 frames, utt_pad_proportion=0.04929, over 10572.62 utterances.], batch size: 45, lr: 3.03e-02, grad_scale: 16.0 2023-03-07 14:54:32,777 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=10743.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:54:34,619 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10744.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:55:06,772 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10765.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:55:08,213 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=10766.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 14:55:12,969 INFO [train2.py:809] (1/4) Epoch 3, batch 2800, loss[ctc_loss=0.1997, att_loss=0.3007, loss=0.2805, over 16640.00 frames. utt_duration=1418 frames, utt_pad_proportion=0.004453, over 47.00 utterances.], tot_loss[ctc_loss=0.2036, att_loss=0.304, loss=0.284, over 3285017.70 frames. utt_duration=1251 frames, utt_pad_proportion=0.05069, over 10516.43 utterances.], batch size: 47, lr: 3.02e-02, grad_scale: 16.0 2023-03-07 14:55:21,931 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10774.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:56:10,552 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.287e+02 4.175e+02 5.006e+02 5.883e+02 1.707e+03, threshold=1.001e+03, percent-clipped=3.0 2023-03-07 14:56:12,580 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10805.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:56:20,297 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10810.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 14:56:33,911 INFO [train2.py:809] (1/4) Epoch 3, batch 2850, loss[ctc_loss=0.1648, att_loss=0.2874, loss=0.2629, over 16892.00 frames. utt_duration=1380 frames, utt_pad_proportion=0.006959, over 49.00 utterances.], tot_loss[ctc_loss=0.2023, att_loss=0.303, loss=0.2828, over 3272570.22 frames. utt_duration=1276 frames, utt_pad_proportion=0.04821, over 10274.49 utterances.], batch size: 49, lr: 3.02e-02, grad_scale: 16.0 2023-03-07 14:57:00,144 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10835.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 14:57:34,300 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.4448, 5.0180, 5.2455, 3.2798, 5.1853, 4.2977, 4.6267, 2.6199], device='cuda:1'), covar=tensor([0.0095, 0.0104, 0.0139, 0.0643, 0.0084, 0.0174, 0.0192, 0.1476], device='cuda:1'), in_proj_covar=tensor([0.0041, 0.0043, 0.0038, 0.0068, 0.0041, 0.0051, 0.0057, 0.0093], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-07 14:57:54,307 INFO [train2.py:809] (1/4) Epoch 3, batch 2900, loss[ctc_loss=0.2267, att_loss=0.3224, loss=0.3033, over 17324.00 frames. utt_duration=1176 frames, utt_pad_proportion=0.02183, over 59.00 utterances.], tot_loss[ctc_loss=0.2026, att_loss=0.3027, loss=0.2826, over 3274944.56 frames. utt_duration=1278 frames, utt_pad_proportion=0.0471, over 10262.04 utterances.], batch size: 59, lr: 3.01e-02, grad_scale: 4.0 2023-03-07 14:58:21,002 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10885.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 14:58:54,576 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.371e+02 4.021e+02 4.974e+02 6.236e+02 1.292e+03, threshold=9.948e+02, percent-clipped=6.0 2023-03-07 14:59:14,723 INFO [train2.py:809] (1/4) Epoch 3, batch 2950, loss[ctc_loss=0.2609, att_loss=0.3365, loss=0.3213, over 17284.00 frames. utt_duration=1259 frames, utt_pad_proportion=0.01206, over 55.00 utterances.], tot_loss[ctc_loss=0.2017, att_loss=0.3024, loss=0.2823, over 3279787.94 frames. utt_duration=1294 frames, utt_pad_proportion=0.04184, over 10153.88 utterances.], batch size: 55, lr: 3.01e-02, grad_scale: 4.0 2023-03-07 14:59:59,152 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10946.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 15:00:25,839 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.69 vs. limit=2.0 2023-03-07 15:00:34,062 INFO [train2.py:809] (1/4) Epoch 3, batch 3000, loss[ctc_loss=0.1707, att_loss=0.2695, loss=0.2497, over 14614.00 frames. utt_duration=1828 frames, utt_pad_proportion=0.0393, over 32.00 utterances.], tot_loss[ctc_loss=0.2032, att_loss=0.3036, loss=0.2835, over 3279046.89 frames. utt_duration=1255 frames, utt_pad_proportion=0.05136, over 10464.33 utterances.], batch size: 32, lr: 3.00e-02, grad_scale: 4.0 2023-03-07 15:00:34,063 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-07 15:00:46,937 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.8731, 4.4378, 4.3316, 4.0321, 4.4263, 4.3194, 4.1344, 4.0383], device='cuda:1'), covar=tensor([0.1263, 0.0419, 0.0251, 0.0608, 0.0378, 0.0424, 0.0304, 0.0373], device='cuda:1'), in_proj_covar=tensor([0.0303, 0.0189, 0.0132, 0.0156, 0.0206, 0.0222, 0.0172, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-07 15:00:47,745 INFO [train2.py:843] (1/4) Epoch 3, validation: ctc_loss=0.1004, att_loss=0.2657, loss=0.2327, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-07 15:00:47,745 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-07 15:01:17,156 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.5513, 1.9620, 2.5699, 4.0341, 3.9354, 3.8709, 3.1199, 1.8475], device='cuda:1'), covar=tensor([0.0329, 0.2484, 0.1447, 0.0431, 0.0165, 0.0148, 0.1115, 0.2394], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0172, 0.0163, 0.0115, 0.0093, 0.0096, 0.0168, 0.0158], device='cuda:1'), out_proj_covar=tensor([1.1799e-04, 1.6493e-04, 1.6154e-04, 1.2893e-04, 9.7473e-05, 9.4973e-05, 1.6718e-04, 1.5275e-04], device='cuda:1') 2023-03-07 15:01:25,827 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.69 vs. limit=2.0 2023-03-07 15:01:48,350 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.705e+02 4.292e+02 5.307e+02 6.516e+02 1.303e+03, threshold=1.061e+03, percent-clipped=4.0 2023-03-07 15:02:09,270 INFO [train2.py:809] (1/4) Epoch 3, batch 3050, loss[ctc_loss=0.1855, att_loss=0.2853, loss=0.2654, over 15951.00 frames. utt_duration=1558 frames, utt_pad_proportion=0.006595, over 41.00 utterances.], tot_loss[ctc_loss=0.203, att_loss=0.3039, loss=0.2837, over 3288105.90 frames. utt_duration=1257 frames, utt_pad_proportion=0.04861, over 10475.75 utterances.], batch size: 41, lr: 2.99e-02, grad_scale: 4.0 2023-03-07 15:03:14,003 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0581, 4.7660, 4.7980, 4.4616, 2.0515, 4.6637, 3.0030, 2.6506], device='cuda:1'), covar=tensor([0.0257, 0.0119, 0.0536, 0.0259, 0.3367, 0.0141, 0.1249, 0.1127], device='cuda:1'), in_proj_covar=tensor([0.0092, 0.0086, 0.0194, 0.0105, 0.0217, 0.0092, 0.0178, 0.0168], device='cuda:1'), out_proj_covar=tensor([8.3272e-05, 7.8783e-05, 1.6059e-04, 8.6066e-05, 1.7031e-04, 8.0337e-05, 1.4363e-04, 1.3736e-04], device='cuda:1') 2023-03-07 15:03:23,383 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11065.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:03:29,049 INFO [train2.py:809] (1/4) Epoch 3, batch 3100, loss[ctc_loss=0.1784, att_loss=0.2843, loss=0.2631, over 16013.00 frames. utt_duration=1603 frames, utt_pad_proportion=0.00743, over 40.00 utterances.], tot_loss[ctc_loss=0.2022, att_loss=0.3037, loss=0.2834, over 3285416.88 frames. utt_duration=1249 frames, utt_pad_proportion=0.05127, over 10532.37 utterances.], batch size: 40, lr: 2.99e-02, grad_scale: 4.0 2023-03-07 15:03:55,822 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1828, 5.1216, 4.9360, 3.4603, 4.9332, 4.3457, 4.2972, 2.9919], device='cuda:1'), covar=tensor([0.0088, 0.0064, 0.0270, 0.0549, 0.0081, 0.0121, 0.0197, 0.1110], device='cuda:1'), in_proj_covar=tensor([0.0041, 0.0042, 0.0038, 0.0068, 0.0041, 0.0051, 0.0057, 0.0090], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-07 15:04:20,595 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11100.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:04:29,506 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.231e+02 3.861e+02 4.979e+02 6.138e+02 1.576e+03, threshold=9.958e+02, percent-clipped=5.0 2023-03-07 15:04:36,010 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11110.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 15:04:40,466 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=11113.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:04:50,174 INFO [train2.py:809] (1/4) Epoch 3, batch 3150, loss[ctc_loss=0.243, att_loss=0.3203, loss=0.3049, over 16766.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.005819, over 48.00 utterances.], tot_loss[ctc_loss=0.2017, att_loss=0.3028, loss=0.2826, over 3275884.23 frames. utt_duration=1231 frames, utt_pad_proportion=0.05782, over 10657.97 utterances.], batch size: 48, lr: 2.98e-02, grad_scale: 4.0 2023-03-07 15:05:08,073 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11130.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:05:08,733 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=5.48 vs. limit=5.0 2023-03-07 15:05:17,667 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0778, 4.8866, 4.8674, 3.6551, 4.9717, 4.2045, 4.4503, 2.5927], device='cuda:1'), covar=tensor([0.0140, 0.0081, 0.0270, 0.0570, 0.0092, 0.0170, 0.0211, 0.1539], device='cuda:1'), in_proj_covar=tensor([0.0043, 0.0044, 0.0039, 0.0071, 0.0043, 0.0053, 0.0059, 0.0095], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-07 15:05:53,256 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=11158.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:06:11,079 INFO [train2.py:809] (1/4) Epoch 3, batch 3200, loss[ctc_loss=0.2043, att_loss=0.3143, loss=0.2923, over 17241.00 frames. utt_duration=1096 frames, utt_pad_proportion=0.04204, over 63.00 utterances.], tot_loss[ctc_loss=0.1989, att_loss=0.3013, loss=0.2809, over 3274603.88 frames. utt_duration=1244 frames, utt_pad_proportion=0.05452, over 10545.35 utterances.], batch size: 63, lr: 2.98e-02, grad_scale: 8.0 2023-03-07 15:06:19,937 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11174.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:07:10,783 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.395e+02 3.760e+02 4.677e+02 5.867e+02 1.602e+03, threshold=9.354e+02, percent-clipped=2.0 2023-03-07 15:07:31,884 INFO [train2.py:809] (1/4) Epoch 3, batch 3250, loss[ctc_loss=0.212, att_loss=0.3253, loss=0.3026, over 17309.00 frames. utt_duration=1260 frames, utt_pad_proportion=0.0109, over 55.00 utterances.], tot_loss[ctc_loss=0.199, att_loss=0.3014, loss=0.2809, over 3279008.18 frames. utt_duration=1252 frames, utt_pad_proportion=0.05165, over 10491.57 utterances.], batch size: 55, lr: 2.97e-02, grad_scale: 8.0 2023-03-07 15:07:33,827 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.2704, 3.7369, 3.1607, 3.2420, 3.8181, 3.4973, 2.4172, 4.3138], device='cuda:1'), covar=tensor([0.1363, 0.0344, 0.1088, 0.0645, 0.0418, 0.0637, 0.1055, 0.0175], device='cuda:1'), in_proj_covar=tensor([0.0125, 0.0094, 0.0141, 0.0113, 0.0102, 0.0134, 0.0119, 0.0076], 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') 2023-03-07 15:07:58,403 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11235.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:08:07,922 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11241.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 15:08:52,867 INFO [train2.py:809] (1/4) Epoch 3, batch 3300, loss[ctc_loss=0.2286, att_loss=0.3346, loss=0.3134, over 17296.00 frames. utt_duration=1174 frames, utt_pad_proportion=0.02301, over 59.00 utterances.], tot_loss[ctc_loss=0.1993, att_loss=0.302, loss=0.2815, over 3275681.11 frames. utt_duration=1213 frames, utt_pad_proportion=0.06339, over 10816.07 utterances.], batch size: 59, lr: 2.97e-02, grad_scale: 8.0 2023-03-07 15:09:10,907 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4860, 5.0885, 4.7576, 5.0763, 4.4132, 4.9849, 5.4107, 5.1447], device='cuda:1'), covar=tensor([0.0422, 0.0258, 0.0490, 0.0212, 0.0425, 0.0149, 0.0201, 0.0133], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0123, 0.0147, 0.0094, 0.0138, 0.0101, 0.0118, 0.0107], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-07 15:09:52,444 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.248e+02 3.797e+02 4.904e+02 6.059e+02 1.982e+03, threshold=9.808e+02, percent-clipped=6.0 2023-03-07 15:10:12,980 INFO [train2.py:809] (1/4) Epoch 3, batch 3350, loss[ctc_loss=0.1659, att_loss=0.2751, loss=0.2533, over 15873.00 frames. utt_duration=1629 frames, utt_pad_proportion=0.00943, over 39.00 utterances.], tot_loss[ctc_loss=0.1995, att_loss=0.3018, loss=0.2813, over 3262168.98 frames. utt_duration=1212 frames, utt_pad_proportion=0.06686, over 10781.07 utterances.], batch size: 39, lr: 2.96e-02, grad_scale: 8.0 2023-03-07 15:11:34,097 INFO [train2.py:809] (1/4) Epoch 3, batch 3400, loss[ctc_loss=0.1603, att_loss=0.262, loss=0.2416, over 15365.00 frames. utt_duration=1758 frames, utt_pad_proportion=0.01147, over 35.00 utterances.], tot_loss[ctc_loss=0.2005, att_loss=0.3022, loss=0.2819, over 3262666.88 frames. utt_duration=1197 frames, utt_pad_proportion=0.07079, over 10918.22 utterances.], batch size: 35, lr: 2.96e-02, grad_scale: 8.0 2023-03-07 15:12:12,857 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-03-07 15:12:25,243 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11400.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:12:34,187 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.436e+02 3.676e+02 4.769e+02 6.355e+02 1.428e+03, threshold=9.539e+02, percent-clipped=8.0 2023-03-07 15:12:55,343 INFO [train2.py:809] (1/4) Epoch 3, batch 3450, loss[ctc_loss=0.177, att_loss=0.2636, loss=0.2463, over 15383.00 frames. utt_duration=1759 frames, utt_pad_proportion=0.01056, over 35.00 utterances.], tot_loss[ctc_loss=0.1984, att_loss=0.3008, loss=0.2803, over 3262127.99 frames. utt_duration=1212 frames, utt_pad_proportion=0.0664, over 10782.59 utterances.], batch size: 35, lr: 2.95e-02, grad_scale: 8.0 2023-03-07 15:13:12,841 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11430.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:13:41,672 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=11448.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:14:04,165 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.8243, 5.1531, 5.4004, 5.5501, 5.0458, 5.7368, 5.1342, 5.8131], device='cuda:1'), covar=tensor([0.0535, 0.0498, 0.0398, 0.0512, 0.1986, 0.0684, 0.0548, 0.0478], device='cuda:1'), in_proj_covar=tensor([0.0381, 0.0268, 0.0260, 0.0314, 0.0462, 0.0251, 0.0227, 0.0283], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0004, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-03-07 15:14:15,456 INFO [train2.py:809] (1/4) Epoch 3, batch 3500, loss[ctc_loss=0.1499, att_loss=0.2788, loss=0.253, over 17051.00 frames. utt_duration=1313 frames, utt_pad_proportion=0.008294, over 52.00 utterances.], tot_loss[ctc_loss=0.1975, att_loss=0.3, loss=0.2795, over 3264650.45 frames. utt_duration=1222 frames, utt_pad_proportion=0.06365, over 10702.76 utterances.], batch size: 52, lr: 2.95e-02, grad_scale: 8.0 2023-03-07 15:14:29,619 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=11478.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:14:38,271 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=2.12 vs. limit=2.0 2023-03-07 15:15:14,772 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.985e+02 3.561e+02 4.473e+02 5.619e+02 1.545e+03, threshold=8.947e+02, percent-clipped=3.0 2023-03-07 15:15:23,098 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4245, 4.2451, 4.2435, 4.5027, 4.6472, 4.4305, 3.9508, 1.9961], device='cuda:1'), covar=tensor([0.0233, 0.0359, 0.0267, 0.0102, 0.1079, 0.0205, 0.0379, 0.3586], device='cuda:1'), in_proj_covar=tensor([0.0125, 0.0111, 0.0106, 0.0107, 0.0216, 0.0130, 0.0102, 0.0238], device='cuda:1'), out_proj_covar=tensor([1.1105e-04, 9.3279e-05, 8.8543e-05, 9.5624e-05, 1.9772e-04, 1.1126e-04, 9.2019e-05, 2.0286e-04], device='cuda:1') 2023-03-07 15:15:36,473 INFO [train2.py:809] (1/4) Epoch 3, batch 3550, loss[ctc_loss=0.2003, att_loss=0.3171, loss=0.2937, over 17301.00 frames. utt_duration=1260 frames, utt_pad_proportion=0.01184, over 55.00 utterances.], tot_loss[ctc_loss=0.1974, att_loss=0.3005, loss=0.2799, over 3270981.82 frames. utt_duration=1217 frames, utt_pad_proportion=0.06258, over 10762.99 utterances.], batch size: 55, lr: 2.94e-02, grad_scale: 8.0 2023-03-07 15:15:53,519 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11530.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:15:56,824 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11532.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:16:12,286 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11541.0, num_to_drop=1, layers_to_drop={2} 2023-03-07 15:16:56,925 INFO [train2.py:809] (1/4) Epoch 3, batch 3600, loss[ctc_loss=0.2125, att_loss=0.3051, loss=0.2866, over 16122.00 frames. utt_duration=1537 frames, utt_pad_proportion=0.006403, over 42.00 utterances.], tot_loss[ctc_loss=0.1976, att_loss=0.3004, loss=0.2798, over 3272037.64 frames. utt_duration=1233 frames, utt_pad_proportion=0.05878, over 10629.33 utterances.], batch size: 42, lr: 2.93e-02, grad_scale: 8.0 2023-03-07 15:17:29,076 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=11589.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 15:17:36,077 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11593.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:17:56,223 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.467e+02 4.035e+02 5.369e+02 6.421e+02 1.595e+03, threshold=1.074e+03, percent-clipped=5.0 2023-03-07 15:18:07,163 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-03-07 15:18:18,110 INFO [train2.py:809] (1/4) Epoch 3, batch 3650, loss[ctc_loss=0.164, att_loss=0.273, loss=0.2512, over 15966.00 frames. utt_duration=1559 frames, utt_pad_proportion=0.005584, over 41.00 utterances.], tot_loss[ctc_loss=0.1972, att_loss=0.3001, loss=0.2795, over 3271454.96 frames. utt_duration=1229 frames, utt_pad_proportion=0.06137, over 10658.16 utterances.], batch size: 41, lr: 2.93e-02, grad_scale: 8.0 2023-03-07 15:18:43,936 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11635.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:19:38,112 INFO [train2.py:809] (1/4) Epoch 3, batch 3700, loss[ctc_loss=0.1756, att_loss=0.2916, loss=0.2684, over 16483.00 frames. utt_duration=1435 frames, utt_pad_proportion=0.005604, over 46.00 utterances.], tot_loss[ctc_loss=0.1967, att_loss=0.2993, loss=0.2788, over 3269590.04 frames. utt_duration=1253 frames, utt_pad_proportion=0.05424, over 10447.11 utterances.], batch size: 46, lr: 2.92e-02, grad_scale: 8.0 2023-03-07 15:20:21,391 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=2.23 vs. limit=2.0 2023-03-07 15:20:22,426 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11696.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:20:26,158 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=2.05 vs. limit=2.0 2023-03-07 15:20:27,159 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11699.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:20:37,359 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.856e+02 3.737e+02 4.615e+02 5.715e+02 1.108e+03, threshold=9.229e+02, percent-clipped=1.0 2023-03-07 15:20:58,431 INFO [train2.py:809] (1/4) Epoch 3, batch 3750, loss[ctc_loss=0.1945, att_loss=0.3051, loss=0.283, over 17309.00 frames. utt_duration=1261 frames, utt_pad_proportion=0.01049, over 55.00 utterances.], tot_loss[ctc_loss=0.1982, att_loss=0.3001, loss=0.2797, over 3274491.26 frames. utt_duration=1221 frames, utt_pad_proportion=0.06052, over 10737.35 utterances.], batch size: 55, lr: 2.92e-02, grad_scale: 8.0 2023-03-07 15:21:22,945 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.4194, 1.4547, 1.2850, 1.1448, 1.3726, 1.7835, 1.9010, 2.0262], device='cuda:1'), covar=tensor([0.0558, 0.1336, 0.1394, 0.1397, 0.1145, 0.0972, 0.0733, 0.0702], device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0088, 0.0092, 0.0080, 0.0075, 0.0080, 0.0094, 0.0100], device='cuda:1'), out_proj_covar=tensor([3.9356e-05, 4.9953e-05, 4.8198e-05, 4.5874e-05, 4.4466e-05, 4.4720e-05, 4.2059e-05, 4.2836e-05], device='cuda:1') 2023-03-07 15:21:50,579 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-03-07 15:22:04,571 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11760.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:22:18,746 INFO [train2.py:809] (1/4) Epoch 3, batch 3800, loss[ctc_loss=0.1716, att_loss=0.2922, loss=0.2681, over 16624.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.005469, over 47.00 utterances.], tot_loss[ctc_loss=0.1983, att_loss=0.3007, loss=0.2803, over 3284208.16 frames. utt_duration=1223 frames, utt_pad_proportion=0.05843, over 10755.03 utterances.], batch size: 47, lr: 2.91e-02, grad_scale: 8.0 2023-03-07 15:22:31,769 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-03-07 15:23:18,125 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.859e+02 4.239e+02 5.484e+02 6.646e+02 1.603e+03, threshold=1.097e+03, percent-clipped=8.0 2023-03-07 15:23:33,650 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11816.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:23:37,909 INFO [train2.py:809] (1/4) Epoch 3, batch 3850, loss[ctc_loss=0.2102, att_loss=0.2923, loss=0.2759, over 14533.00 frames. utt_duration=1818 frames, utt_pad_proportion=0.04039, over 32.00 utterances.], tot_loss[ctc_loss=0.1964, att_loss=0.2996, loss=0.2789, over 3275720.21 frames. utt_duration=1235 frames, utt_pad_proportion=0.05787, over 10622.76 utterances.], batch size: 32, lr: 2.91e-02, grad_scale: 8.0 2023-03-07 15:23:55,923 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11830.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:24:07,834 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9660, 6.1039, 5.4689, 6.0423, 5.8393, 5.4728, 5.4909, 5.2656], device='cuda:1'), covar=tensor([0.0956, 0.0694, 0.0665, 0.0636, 0.0509, 0.1164, 0.2289, 0.2291], device='cuda:1'), in_proj_covar=tensor([0.0281, 0.0313, 0.0260, 0.0253, 0.0235, 0.0325, 0.0350, 0.0342], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-07 15:24:20,883 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11846.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:24:40,898 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3168, 4.4770, 4.4865, 4.6529, 4.9115, 4.2629, 4.2592, 1.8939], device='cuda:1'), covar=tensor([0.0302, 0.0373, 0.0263, 0.0199, 0.0987, 0.0289, 0.0489, 0.3937], device='cuda:1'), in_proj_covar=tensor([0.0129, 0.0119, 0.0110, 0.0114, 0.0229, 0.0135, 0.0105, 0.0255], device='cuda:1'), out_proj_covar=tensor([1.1546e-04, 1.0012e-04, 9.5235e-05, 1.0266e-04, 2.1004e-04, 1.1571e-04, 9.6102e-05, 2.1797e-04], device='cuda:1') 2023-03-07 15:24:49,390 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-03-07 15:24:55,938 INFO [train2.py:809] (1/4) Epoch 3, batch 3900, loss[ctc_loss=0.1971, att_loss=0.3024, loss=0.2814, over 17344.00 frames. utt_duration=1103 frames, utt_pad_proportion=0.03519, over 63.00 utterances.], tot_loss[ctc_loss=0.1943, att_loss=0.2987, loss=0.2779, over 3282481.67 frames. utt_duration=1258 frames, utt_pad_proportion=0.05007, over 10448.64 utterances.], batch size: 63, lr: 2.90e-02, grad_scale: 8.0 2023-03-07 15:25:02,238 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8076, 5.2678, 4.9466, 5.1822, 4.6712, 5.0729, 5.5356, 5.3198], device='cuda:1'), covar=tensor([0.0247, 0.0219, 0.0372, 0.0144, 0.0272, 0.0119, 0.0163, 0.0119], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0129, 0.0157, 0.0098, 0.0144, 0.0107, 0.0128, 0.0117], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-07 15:25:08,654 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11877.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:25:09,893 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=11878.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:25:16,449 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3288, 4.9590, 4.9494, 4.1212, 1.9493, 2.7868, 4.9353, 3.5680], device='cuda:1'), covar=tensor([0.0475, 0.0125, 0.0115, 0.0644, 0.7669, 0.2436, 0.0179, 0.2141], device='cuda:1'), in_proj_covar=tensor([0.0239, 0.0143, 0.0159, 0.0183, 0.0391, 0.0298, 0.0149, 0.0259], device='cuda:1'), out_proj_covar=tensor([1.3574e-04, 7.3672e-05, 8.5524e-05, 8.9261e-05, 1.9608e-04, 1.4860e-04, 7.8059e-05, 1.4467e-04], device='cuda:1') 2023-03-07 15:25:25,505 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11888.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:25:53,044 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.007e+02 3.699e+02 4.427e+02 5.534e+02 1.749e+03, threshold=8.855e+02, percent-clipped=3.0 2023-03-07 15:25:54,873 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11907.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:26:07,717 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.17 vs. limit=5.0 2023-03-07 15:26:13,116 INFO [train2.py:809] (1/4) Epoch 3, batch 3950, loss[ctc_loss=0.1631, att_loss=0.2705, loss=0.249, over 15959.00 frames. utt_duration=1558 frames, utt_pad_proportion=0.006222, over 41.00 utterances.], tot_loss[ctc_loss=0.1924, att_loss=0.2977, loss=0.2766, over 3277647.16 frames. utt_duration=1277 frames, utt_pad_proportion=0.04676, over 10282.55 utterances.], batch size: 41, lr: 2.90e-02, grad_scale: 8.0 2023-03-07 15:26:14,868 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11920.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:26:53,968 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.73 vs. limit=2.0 2023-03-07 15:26:57,731 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11948.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:27:27,007 INFO [train2.py:809] (1/4) Epoch 4, batch 0, loss[ctc_loss=0.1859, att_loss=0.2967, loss=0.2746, over 16421.00 frames. utt_duration=1494 frames, utt_pad_proportion=0.006543, over 44.00 utterances.], tot_loss[ctc_loss=0.1859, att_loss=0.2967, loss=0.2746, over 16421.00 frames. utt_duration=1494 frames, utt_pad_proportion=0.006543, over 44.00 utterances.], batch size: 44, lr: 2.71e-02, grad_scale: 8.0 2023-03-07 15:27:27,007 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-07 15:27:39,233 INFO [train2.py:843] (1/4) Epoch 4, validation: ctc_loss=0.101, att_loss=0.2624, loss=0.2301, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-07 15:27:39,235 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-07 15:28:24,375 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11981.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:28:28,839 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11984.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:28:39,310 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11991.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:28:51,136 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-03-07 15:29:03,781 INFO [train2.py:809] (1/4) Epoch 4, batch 50, loss[ctc_loss=0.1895, att_loss=0.3028, loss=0.2801, over 17067.00 frames. utt_duration=1314 frames, utt_pad_proportion=0.00807, over 52.00 utterances.], tot_loss[ctc_loss=0.194, att_loss=0.2965, loss=0.276, over 739420.02 frames. utt_duration=1380 frames, utt_pad_proportion=0.02094, over 2145.38 utterances.], batch size: 52, lr: 2.70e-02, grad_scale: 8.0 2023-03-07 15:29:08,346 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.811e+02 4.015e+02 4.845e+02 5.994e+02 1.313e+03, threshold=9.689e+02, percent-clipped=6.0 2023-03-07 15:29:08,859 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5169, 2.2056, 4.9012, 3.7615, 3.1277, 4.2550, 4.3982, 4.5621], device='cuda:1'), covar=tensor([0.0183, 0.1938, 0.0238, 0.1431, 0.2705, 0.0435, 0.0195, 0.0327], device='cuda:1'), in_proj_covar=tensor([0.0121, 0.0225, 0.0119, 0.0278, 0.0329, 0.0170, 0.0109, 0.0118], device='cuda:1'), out_proj_covar=tensor([9.9457e-05, 1.7065e-04, 9.6869e-05, 2.2003e-04, 2.4630e-04, 1.3749e-04, 9.2512e-05, 1.0005e-04], device='cuda:1') 2023-03-07 15:29:13,812 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12009.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:29:19,187 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1577, 2.5375, 3.6703, 2.0861, 3.1588, 4.4689, 4.1853, 3.0026], device='cuda:1'), covar=tensor([0.0440, 0.1693, 0.0820, 0.1774, 0.1141, 0.0315, 0.0420, 0.1398], device='cuda:1'), in_proj_covar=tensor([0.0163, 0.0173, 0.0154, 0.0165, 0.0176, 0.0123, 0.0118, 0.0170], 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') 2023-03-07 15:30:10,865 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12045.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:30:24,239 INFO [train2.py:809] (1/4) Epoch 4, batch 100, loss[ctc_loss=0.2004, att_loss=0.315, loss=0.2921, over 17038.00 frames. utt_duration=1288 frames, utt_pad_proportion=0.01012, over 53.00 utterances.], tot_loss[ctc_loss=0.1947, att_loss=0.2987, loss=0.2779, over 1295069.01 frames. utt_duration=1212 frames, utt_pad_proportion=0.06883, over 4281.20 utterances.], batch size: 53, lr: 2.70e-02, grad_scale: 8.0 2023-03-07 15:30:27,570 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12055.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:30:36,361 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8220, 5.3413, 5.0986, 5.1198, 5.3735, 5.2363, 5.0810, 4.9228], device='cuda:1'), covar=tensor([0.1262, 0.0327, 0.0187, 0.0438, 0.0281, 0.0220, 0.0251, 0.0241], device='cuda:1'), in_proj_covar=tensor([0.0320, 0.0192, 0.0130, 0.0164, 0.0207, 0.0225, 0.0181, 0.0191], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-07 15:30:57,091 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1262, 4.9736, 5.0376, 3.4243, 5.0386, 4.2927, 4.5011, 2.6758], device='cuda:1'), covar=tensor([0.0110, 0.0070, 0.0153, 0.0625, 0.0076, 0.0147, 0.0193, 0.1359], device='cuda:1'), in_proj_covar=tensor([0.0043, 0.0046, 0.0039, 0.0077, 0.0046, 0.0055, 0.0062, 0.0095], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-03-07 15:31:01,986 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12076.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:31:21,242 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.66 vs. limit=5.0 2023-03-07 15:31:43,892 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-03-07 15:31:44,680 INFO [train2.py:809] (1/4) Epoch 4, batch 150, loss[ctc_loss=0.1391, att_loss=0.2679, loss=0.2422, over 16136.00 frames. utt_duration=1538 frames, utt_pad_proportion=0.005571, over 42.00 utterances.], tot_loss[ctc_loss=0.191, att_loss=0.2944, loss=0.2737, over 1721554.46 frames. utt_duration=1268 frames, utt_pad_proportion=0.05845, over 5437.69 utterances.], batch size: 42, lr: 2.69e-02, grad_scale: 8.0 2023-03-07 15:31:49,489 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.353e+02 4.093e+02 5.170e+02 6.831e+02 1.205e+03, threshold=1.034e+03, percent-clipped=5.0 2023-03-07 15:32:17,336 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([1.6953, 1.5723, 2.8644, 1.5866, 1.9912, 2.5480, 1.5880, 1.6087], device='cuda:1'), covar=tensor([0.1695, 0.1817, 0.0571, 0.3852, 0.1205, 0.1264, 0.1169, 0.4043], device='cuda:1'), in_proj_covar=tensor([0.0065, 0.0059, 0.0053, 0.0068, 0.0049, 0.0062, 0.0056, 0.0082], device='cuda:1'), out_proj_covar=tensor([3.7841e-05, 3.5505e-05, 3.2725e-05, 4.4015e-05, 3.4431e-05, 3.7475e-05, 3.6470e-05, 5.6698e-05], device='cuda:1') 2023-03-07 15:32:34,824 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-03-07 15:32:38,634 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12137.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:33:02,328 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.8120, 5.0299, 5.3376, 5.4736, 5.0482, 5.7872, 4.9826, 5.8758], device='cuda:1'), covar=tensor([0.0457, 0.0551, 0.0401, 0.0507, 0.1884, 0.0510, 0.0494, 0.0455], device='cuda:1'), in_proj_covar=tensor([0.0394, 0.0272, 0.0264, 0.0315, 0.0473, 0.0263, 0.0229, 0.0295], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0004, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-03-07 15:33:04,153 INFO [train2.py:809] (1/4) Epoch 4, batch 200, loss[ctc_loss=0.2028, att_loss=0.3122, loss=0.2904, over 17005.00 frames. utt_duration=1335 frames, utt_pad_proportion=0.008588, over 51.00 utterances.], tot_loss[ctc_loss=0.1876, att_loss=0.2934, loss=0.2723, over 2069481.08 frames. utt_duration=1298 frames, utt_pad_proportion=0.04625, over 6382.58 utterances.], batch size: 51, lr: 2.69e-02, grad_scale: 8.0 2023-03-07 15:33:20,546 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5419, 4.9329, 5.1719, 4.9103, 2.2380, 2.4824, 5.2698, 3.8522], device='cuda:1'), covar=tensor([0.0407, 0.0217, 0.0153, 0.0306, 0.7735, 0.3141, 0.0208, 0.1789], device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0145, 0.0162, 0.0189, 0.0398, 0.0305, 0.0153, 0.0263], device='cuda:1'), out_proj_covar=tensor([1.3958e-04, 7.4987e-05, 8.7062e-05, 9.1965e-05, 1.9997e-04, 1.5288e-04, 7.9749e-05, 1.4745e-04], device='cuda:1') 2023-03-07 15:33:34,666 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12172.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:33:59,108 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12188.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:34:21,567 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12202.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:34:23,025 INFO [train2.py:809] (1/4) Epoch 4, batch 250, loss[ctc_loss=0.2421, att_loss=0.3267, loss=0.3098, over 14536.00 frames. utt_duration=402.5 frames, utt_pad_proportion=0.3, over 145.00 utterances.], tot_loss[ctc_loss=0.1886, att_loss=0.2938, loss=0.2727, over 2326963.43 frames. utt_duration=1260 frames, utt_pad_proportion=0.0575, over 7395.34 utterances.], batch size: 145, lr: 2.68e-02, grad_scale: 8.0 2023-03-07 15:34:28,452 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.253e+02 3.783e+02 4.490e+02 5.232e+02 8.659e+02, threshold=8.979e+02, percent-clipped=0.0 2023-03-07 15:34:49,334 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6610, 4.1724, 4.3972, 4.4780, 4.1115, 4.5998, 4.2933, 4.6250], device='cuda:1'), covar=tensor([0.0646, 0.0612, 0.0459, 0.0571, 0.1741, 0.0633, 0.1352, 0.0570], device='cuda:1'), in_proj_covar=tensor([0.0400, 0.0276, 0.0269, 0.0322, 0.0473, 0.0272, 0.0233, 0.0299], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0004, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-03-07 15:35:15,626 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=12236.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:35:42,724 INFO [train2.py:809] (1/4) Epoch 4, batch 300, loss[ctc_loss=0.1965, att_loss=0.3067, loss=0.2847, over 16888.00 frames. utt_duration=1380 frames, utt_pad_proportion=0.007136, over 49.00 utterances.], tot_loss[ctc_loss=0.1905, att_loss=0.2953, loss=0.2743, over 2530773.31 frames. utt_duration=1210 frames, utt_pad_proportion=0.06942, over 8377.19 utterances.], batch size: 49, lr: 2.68e-02, grad_scale: 8.0 2023-03-07 15:36:19,377 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12276.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:36:42,515 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12291.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:37:01,515 INFO [train2.py:809] (1/4) Epoch 4, batch 350, loss[ctc_loss=0.1894, att_loss=0.288, loss=0.2683, over 16402.00 frames. utt_duration=1492 frames, utt_pad_proportion=0.007661, over 44.00 utterances.], tot_loss[ctc_loss=0.1896, att_loss=0.2948, loss=0.2738, over 2701533.38 frames. utt_duration=1235 frames, utt_pad_proportion=0.05947, over 8757.17 utterances.], batch size: 44, lr: 2.67e-02, grad_scale: 8.0 2023-03-07 15:37:03,269 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12304.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:37:06,088 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.120e+02 4.027e+02 4.741e+02 6.835e+02 1.897e+03, threshold=9.483e+02, percent-clipped=6.0 2023-03-07 15:37:58,216 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=12339.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:37:59,864 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12340.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:38:20,462 INFO [train2.py:809] (1/4) Epoch 4, batch 400, loss[ctc_loss=0.1915, att_loss=0.2983, loss=0.2769, over 16975.00 frames. utt_duration=1359 frames, utt_pad_proportion=0.007056, over 50.00 utterances.], tot_loss[ctc_loss=0.1887, att_loss=0.295, loss=0.2738, over 2823620.34 frames. utt_duration=1242 frames, utt_pad_proportion=0.05816, over 9101.60 utterances.], batch size: 50, lr: 2.67e-02, grad_scale: 8.0 2023-03-07 15:38:23,737 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12355.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:39:39,639 INFO [train2.py:809] (1/4) Epoch 4, batch 450, loss[ctc_loss=0.2166, att_loss=0.2879, loss=0.2736, over 15790.00 frames. utt_duration=1664 frames, utt_pad_proportion=0.007348, over 38.00 utterances.], tot_loss[ctc_loss=0.1883, att_loss=0.2949, loss=0.2735, over 2923308.28 frames. utt_duration=1261 frames, utt_pad_proportion=0.0527, over 9284.39 utterances.], batch size: 38, lr: 2.66e-02, grad_scale: 8.0 2023-03-07 15:39:39,754 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=12403.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:39:44,236 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.374e+02 3.731e+02 4.738e+02 5.685e+02 9.673e+02, threshold=9.477e+02, percent-clipped=1.0 2023-03-07 15:40:18,227 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.8206, 2.3084, 2.9802, 4.3630, 4.3401, 4.5031, 2.8061, 1.8515], device='cuda:1'), covar=tensor([0.0390, 0.2644, 0.1515, 0.0746, 0.0279, 0.0177, 0.1762, 0.2778], device='cuda:1'), in_proj_covar=tensor([0.0128, 0.0184, 0.0179, 0.0124, 0.0107, 0.0101, 0.0175, 0.0169], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-07 15:40:25,797 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12432.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:40:47,558 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12446.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:40:50,946 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0098, 5.1557, 5.4888, 5.6717, 5.2195, 5.9249, 5.0186, 5.9880], device='cuda:1'), covar=tensor([0.0446, 0.0575, 0.0467, 0.0519, 0.1886, 0.0627, 0.0416, 0.0403], device='cuda:1'), in_proj_covar=tensor([0.0405, 0.0284, 0.0274, 0.0337, 0.0481, 0.0273, 0.0239, 0.0296], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0004, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-03-07 15:40:58,418 INFO [train2.py:809] (1/4) Epoch 4, batch 500, loss[ctc_loss=0.1788, att_loss=0.2758, loss=0.2564, over 16120.00 frames. utt_duration=1536 frames, utt_pad_proportion=0.006864, over 42.00 utterances.], tot_loss[ctc_loss=0.1883, att_loss=0.2948, loss=0.2735, over 3004934.98 frames. utt_duration=1266 frames, utt_pad_proportion=0.05145, over 9505.92 utterances.], batch size: 42, lr: 2.66e-02, grad_scale: 8.0 2023-03-07 15:41:28,701 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12472.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:42:15,278 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12502.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:42:16,490 INFO [train2.py:809] (1/4) Epoch 4, batch 550, loss[ctc_loss=0.2029, att_loss=0.3019, loss=0.2821, over 17005.00 frames. utt_duration=695.8 frames, utt_pad_proportion=0.1281, over 98.00 utterances.], tot_loss[ctc_loss=0.1874, att_loss=0.2945, loss=0.2731, over 3067670.54 frames. utt_duration=1278 frames, utt_pad_proportion=0.04806, over 9612.98 utterances.], batch size: 98, lr: 2.65e-02, grad_scale: 8.0 2023-03-07 15:42:21,769 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.239e+02 3.520e+02 4.552e+02 5.513e+02 1.151e+03, threshold=9.104e+02, percent-clipped=4.0 2023-03-07 15:42:23,575 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12507.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:42:43,557 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=12520.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:43:30,618 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-03-07 15:43:31,176 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=12550.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:43:35,781 INFO [train2.py:809] (1/4) Epoch 4, batch 600, loss[ctc_loss=0.1894, att_loss=0.3073, loss=0.2838, over 17414.00 frames. utt_duration=1011 frames, utt_pad_proportion=0.04451, over 69.00 utterances.], tot_loss[ctc_loss=0.1876, att_loss=0.2946, loss=0.2732, over 3121474.92 frames. utt_duration=1279 frames, utt_pad_proportion=0.04579, over 9776.08 utterances.], batch size: 69, lr: 2.65e-02, grad_scale: 8.0 2023-03-07 15:43:41,511 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2049, 4.4425, 4.7926, 4.7181, 1.9338, 4.6312, 2.9061, 2.7739], device='cuda:1'), covar=tensor([0.0214, 0.0220, 0.0648, 0.0237, 0.3583, 0.0193, 0.1393, 0.1387], device='cuda:1'), in_proj_covar=tensor([0.0093, 0.0092, 0.0213, 0.0111, 0.0223, 0.0099, 0.0196, 0.0187], device='cuda:1'), out_proj_covar=tensor([8.8394e-05, 8.4576e-05, 1.7971e-04, 9.5717e-05, 1.8109e-04, 8.7285e-05, 1.6354e-04, 1.5525e-04], device='cuda:1') 2023-03-07 15:44:13,022 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12576.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:44:55,223 INFO [train2.py:809] (1/4) Epoch 4, batch 650, loss[ctc_loss=0.2004, att_loss=0.2994, loss=0.2796, over 16464.00 frames. utt_duration=1433 frames, utt_pad_proportion=0.006809, over 46.00 utterances.], tot_loss[ctc_loss=0.1861, att_loss=0.2937, loss=0.2722, over 3149392.46 frames. utt_duration=1266 frames, utt_pad_proportion=0.05162, over 9963.07 utterances.], batch size: 46, lr: 2.65e-02, grad_scale: 8.0 2023-03-07 15:44:57,666 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12604.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:45:00,416 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.959e+02 3.647e+02 4.227e+02 5.409e+02 1.351e+03, threshold=8.454e+02, percent-clipped=4.0 2023-03-07 15:45:29,071 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=12624.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:45:52,237 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.25 vs. limit=5.0 2023-03-07 15:45:53,179 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12640.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:46:13,057 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=12652.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:46:14,655 INFO [train2.py:809] (1/4) Epoch 4, batch 700, loss[ctc_loss=0.191, att_loss=0.3071, loss=0.2839, over 17101.00 frames. utt_duration=1223 frames, utt_pad_proportion=0.01613, over 56.00 utterances.], tot_loss[ctc_loss=0.1875, att_loss=0.2942, loss=0.2728, over 3169697.12 frames. utt_duration=1245 frames, utt_pad_proportion=0.05932, over 10196.38 utterances.], batch size: 56, lr: 2.64e-02, grad_scale: 8.0 2023-03-07 15:46:27,653 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.73 vs. limit=2.0 2023-03-07 15:46:34,411 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.2146, 2.1357, 2.0723, 1.0302, 0.9365, 1.4678, 1.8229, 1.8133], device='cuda:1'), covar=tensor([0.0697, 0.1541, 0.1396, 0.1669, 0.1538, 0.1978, 0.0929, 0.1163], device='cuda:1'), in_proj_covar=tensor([0.0095, 0.0091, 0.0088, 0.0073, 0.0075, 0.0080, 0.0089, 0.0092], device='cuda:1'), out_proj_covar=tensor([3.9715e-05, 4.8954e-05, 4.6402e-05, 4.0436e-05, 4.0752e-05, 4.2863e-05, 4.1409e-05, 4.2663e-05], device='cuda:1') 2023-03-07 15:47:09,150 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=12688.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:47:34,089 INFO [train2.py:809] (1/4) Epoch 4, batch 750, loss[ctc_loss=0.1793, att_loss=0.2911, loss=0.2688, over 16322.00 frames. utt_duration=1453 frames, utt_pad_proportion=0.00649, over 45.00 utterances.], tot_loss[ctc_loss=0.1893, att_loss=0.2956, loss=0.2744, over 3190157.52 frames. utt_duration=1200 frames, utt_pad_proportion=0.07091, over 10643.84 utterances.], batch size: 45, lr: 2.64e-02, grad_scale: 8.0 2023-03-07 15:47:38,633 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.969e+02 3.712e+02 4.325e+02 6.002e+02 1.306e+03, threshold=8.650e+02, percent-clipped=6.0 2023-03-07 15:48:19,062 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12732.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:48:52,754 INFO [train2.py:809] (1/4) Epoch 4, batch 800, loss[ctc_loss=0.1805, att_loss=0.2939, loss=0.2712, over 16551.00 frames. utt_duration=1473 frames, utt_pad_proportion=0.005627, over 45.00 utterances.], tot_loss[ctc_loss=0.1885, att_loss=0.2948, loss=0.2736, over 3203176.23 frames. utt_duration=1205 frames, utt_pad_proportion=0.06966, over 10649.78 utterances.], batch size: 45, lr: 2.63e-02, grad_scale: 8.0 2023-03-07 15:49:02,426 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([1.9577, 1.5876, 1.7410, 1.5776, 1.2348, 1.2615, 1.1416, 1.9649], device='cuda:1'), covar=tensor([0.0640, 0.1302, 0.1179, 0.0780, 0.0806, 0.1291, 0.1199, 0.0746], device='cuda:1'), in_proj_covar=tensor([0.0095, 0.0091, 0.0091, 0.0074, 0.0076, 0.0079, 0.0092, 0.0091], device='cuda:1'), out_proj_covar=tensor([3.9733e-05, 4.9322e-05, 4.7377e-05, 4.1047e-05, 4.0988e-05, 4.2939e-05, 4.2318e-05, 4.2840e-05], device='cuda:1') 2023-03-07 15:49:08,762 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12763.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:49:14,927 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.74 vs. limit=2.0 2023-03-07 15:49:35,441 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=12780.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:50:11,367 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12802.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:50:12,734 INFO [train2.py:809] (1/4) Epoch 4, batch 850, loss[ctc_loss=0.1336, att_loss=0.2514, loss=0.2279, over 15761.00 frames. utt_duration=1661 frames, utt_pad_proportion=0.009107, over 38.00 utterances.], tot_loss[ctc_loss=0.1891, att_loss=0.2951, loss=0.2739, over 3229150.45 frames. utt_duration=1237 frames, utt_pad_proportion=0.05877, over 10455.46 utterances.], batch size: 38, lr: 2.63e-02, grad_scale: 8.0 2023-03-07 15:50:17,381 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.255e+02 3.893e+02 4.938e+02 6.078e+02 1.799e+03, threshold=9.877e+02, percent-clipped=7.0 2023-03-07 15:50:46,123 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12824.0, num_to_drop=1, layers_to_drop={3} 2023-03-07 15:51:32,706 INFO [train2.py:809] (1/4) Epoch 4, batch 900, loss[ctc_loss=0.1859, att_loss=0.2967, loss=0.2745, over 16320.00 frames. utt_duration=1452 frames, utt_pad_proportion=0.006779, over 45.00 utterances.], tot_loss[ctc_loss=0.1873, att_loss=0.2937, loss=0.2725, over 3228384.06 frames. utt_duration=1252 frames, utt_pad_proportion=0.05859, over 10328.60 utterances.], batch size: 45, lr: 2.62e-02, grad_scale: 16.0 2023-03-07 15:52:10,070 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3326, 2.7922, 3.4989, 2.0500, 3.2180, 4.3319, 4.2277, 3.1618], device='cuda:1'), covar=tensor([0.0435, 0.1604, 0.1005, 0.1897, 0.1167, 0.0478, 0.0459, 0.1508], device='cuda:1'), in_proj_covar=tensor([0.0175, 0.0184, 0.0169, 0.0172, 0.0185, 0.0140, 0.0131, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-07 15:52:53,759 INFO [train2.py:809] (1/4) Epoch 4, batch 950, loss[ctc_loss=0.1943, att_loss=0.3049, loss=0.2828, over 16885.00 frames. utt_duration=1380 frames, utt_pad_proportion=0.007429, over 49.00 utterances.], tot_loss[ctc_loss=0.1864, att_loss=0.2939, loss=0.2724, over 3243469.85 frames. utt_duration=1263 frames, utt_pad_proportion=0.05186, over 10283.60 utterances.], batch size: 49, lr: 2.62e-02, grad_scale: 16.0 2023-03-07 15:52:58,561 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.008e+02 3.388e+02 4.239e+02 5.070e+02 1.281e+03, threshold=8.479e+02, percent-clipped=6.0 2023-03-07 15:54:15,089 INFO [train2.py:809] (1/4) Epoch 4, batch 1000, loss[ctc_loss=0.1451, att_loss=0.2523, loss=0.2308, over 15473.00 frames. utt_duration=1721 frames, utt_pad_proportion=0.007641, over 36.00 utterances.], tot_loss[ctc_loss=0.1878, att_loss=0.2946, loss=0.2733, over 3242700.47 frames. utt_duration=1241 frames, utt_pad_proportion=0.05927, over 10460.63 utterances.], batch size: 36, lr: 2.61e-02, grad_scale: 8.0 2023-03-07 15:55:36,070 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1908, 2.6192, 3.4826, 1.9846, 3.3199, 4.3768, 4.2171, 2.8782], device='cuda:1'), covar=tensor([0.0341, 0.1453, 0.0833, 0.1636, 0.0962, 0.0358, 0.0283, 0.1462], device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0184, 0.0167, 0.0170, 0.0184, 0.0140, 0.0131, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-07 15:55:37,208 INFO [train2.py:809] (1/4) Epoch 4, batch 1050, loss[ctc_loss=0.2664, att_loss=0.3375, loss=0.3233, over 17403.00 frames. utt_duration=1106 frames, utt_pad_proportion=0.03301, over 63.00 utterances.], tot_loss[ctc_loss=0.1866, att_loss=0.294, loss=0.2725, over 3245744.92 frames. utt_duration=1246 frames, utt_pad_proportion=0.05905, over 10432.25 utterances.], batch size: 63, lr: 2.61e-02, grad_scale: 8.0 2023-03-07 15:55:43,241 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.497e+02 3.995e+02 4.964e+02 6.656e+02 1.651e+03, threshold=9.928e+02, percent-clipped=9.0 2023-03-07 15:56:58,339 INFO [train2.py:809] (1/4) Epoch 4, batch 1100, loss[ctc_loss=0.1498, att_loss=0.2755, loss=0.2504, over 16332.00 frames. utt_duration=1453 frames, utt_pad_proportion=0.006034, over 45.00 utterances.], tot_loss[ctc_loss=0.185, att_loss=0.2933, loss=0.2717, over 3251324.08 frames. utt_duration=1268 frames, utt_pad_proportion=0.05234, over 10270.54 utterances.], batch size: 45, lr: 2.61e-02, grad_scale: 8.0 2023-03-07 15:57:24,785 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6664, 3.1407, 3.6174, 2.5145, 3.5837, 4.4604, 4.2922, 3.3959], device='cuda:1'), covar=tensor([0.0408, 0.1552, 0.1064, 0.1786, 0.1067, 0.0551, 0.0706, 0.1408], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0184, 0.0169, 0.0170, 0.0185, 0.0142, 0.0133, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-07 15:57:31,237 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2902, 4.5554, 4.9593, 4.2481, 1.7978, 2.2857, 5.1100, 3.6515], device='cuda:1'), covar=tensor([0.0618, 0.0303, 0.0161, 0.0590, 0.8116, 0.2914, 0.0153, 0.2419], device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0145, 0.0162, 0.0181, 0.0382, 0.0308, 0.0149, 0.0278], device='cuda:1'), out_proj_covar=tensor([1.3908e-04, 7.3933e-05, 8.6065e-05, 8.7615e-05, 1.9470e-04, 1.5339e-04, 7.6304e-05, 1.5175e-04], device='cuda:1') 2023-03-07 15:58:18,888 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=13102.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:58:20,171 INFO [train2.py:809] (1/4) Epoch 4, batch 1150, loss[ctc_loss=0.1858, att_loss=0.2982, loss=0.2757, over 17391.00 frames. utt_duration=1222 frames, utt_pad_proportion=0.0145, over 57.00 utterances.], tot_loss[ctc_loss=0.1844, att_loss=0.293, loss=0.2713, over 3263077.35 frames. utt_duration=1281 frames, utt_pad_proportion=0.04653, over 10203.38 utterances.], batch size: 57, lr: 2.60e-02, grad_scale: 8.0 2023-03-07 15:58:26,384 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.150e+02 3.646e+02 4.698e+02 5.581e+02 1.233e+03, threshold=9.396e+02, percent-clipped=4.0 2023-03-07 15:58:46,188 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=13119.0, num_to_drop=1, layers_to_drop={3} 2023-03-07 15:59:14,264 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9896, 5.0184, 4.9056, 3.3987, 4.8291, 4.2657, 4.0252, 2.8445], device='cuda:1'), covar=tensor([0.0106, 0.0063, 0.0239, 0.0608, 0.0103, 0.0153, 0.0323, 0.1304], device='cuda:1'), in_proj_covar=tensor([0.0047, 0.0050, 0.0045, 0.0082, 0.0050, 0.0060, 0.0070, 0.0100], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-03-07 15:59:35,985 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=13150.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 15:59:40,595 INFO [train2.py:809] (1/4) Epoch 4, batch 1200, loss[ctc_loss=0.2049, att_loss=0.3134, loss=0.2917, over 17023.00 frames. utt_duration=1286 frames, utt_pad_proportion=0.01061, over 53.00 utterances.], tot_loss[ctc_loss=0.184, att_loss=0.2927, loss=0.2709, over 3270160.88 frames. utt_duration=1285 frames, utt_pad_proportion=0.04305, over 10189.84 utterances.], batch size: 53, lr: 2.60e-02, grad_scale: 8.0 2023-03-07 15:59:53,724 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3614, 2.4910, 4.9066, 3.7125, 3.1725, 4.2456, 4.4753, 4.4730], device='cuda:1'), covar=tensor([0.0167, 0.1833, 0.0145, 0.1281, 0.2433, 0.0310, 0.0271, 0.0345], device='cuda:1'), in_proj_covar=tensor([0.0124, 0.0233, 0.0126, 0.0290, 0.0337, 0.0177, 0.0113, 0.0128], device='cuda:1'), out_proj_covar=tensor([1.0304e-04, 1.7852e-04, 1.0370e-04, 2.2955e-04, 2.5576e-04, 1.4587e-04, 9.5524e-05, 1.0848e-04], device='cuda:1') 2023-03-07 16:00:33,313 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3325, 2.5623, 5.0027, 3.6696, 3.1441, 4.4894, 4.5496, 4.6285], device='cuda:1'), covar=tensor([0.0203, 0.1739, 0.0228, 0.1381, 0.2484, 0.0317, 0.0247, 0.0269], device='cuda:1'), in_proj_covar=tensor([0.0124, 0.0231, 0.0124, 0.0286, 0.0333, 0.0177, 0.0111, 0.0126], device='cuda:1'), out_proj_covar=tensor([1.0305e-04, 1.7757e-04, 1.0226e-04, 2.2627e-04, 2.5271e-04, 1.4625e-04, 9.4772e-05, 1.0739e-04], device='cuda:1') 2023-03-07 16:01:01,002 INFO [train2.py:809] (1/4) Epoch 4, batch 1250, loss[ctc_loss=0.1709, att_loss=0.2687, loss=0.2491, over 15491.00 frames. utt_duration=1722 frames, utt_pad_proportion=0.008966, over 36.00 utterances.], tot_loss[ctc_loss=0.1845, att_loss=0.2928, loss=0.2711, over 3271218.96 frames. utt_duration=1280 frames, utt_pad_proportion=0.04506, over 10234.07 utterances.], batch size: 36, lr: 2.59e-02, grad_scale: 8.0 2023-03-07 16:01:07,179 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.553e+02 3.747e+02 4.896e+02 5.785e+02 2.345e+03, threshold=9.791e+02, percent-clipped=4.0 2023-03-07 16:02:21,491 INFO [train2.py:809] (1/4) Epoch 4, batch 1300, loss[ctc_loss=0.1824, att_loss=0.2969, loss=0.274, over 17102.00 frames. utt_duration=1223 frames, utt_pad_proportion=0.01531, over 56.00 utterances.], tot_loss[ctc_loss=0.1837, att_loss=0.2929, loss=0.271, over 3275969.27 frames. utt_duration=1269 frames, utt_pad_proportion=0.04698, over 10341.67 utterances.], batch size: 56, lr: 2.59e-02, grad_scale: 8.0 2023-03-07 16:02:27,341 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8943, 2.8545, 5.0872, 4.0404, 3.1660, 4.5299, 4.6377, 4.5413], device='cuda:1'), covar=tensor([0.0101, 0.1592, 0.0148, 0.1126, 0.2265, 0.0278, 0.0167, 0.0227], device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0225, 0.0121, 0.0281, 0.0327, 0.0175, 0.0109, 0.0123], device='cuda:1'), out_proj_covar=tensor([9.9246e-05, 1.7291e-04, 9.9810e-05, 2.2306e-04, 2.4855e-04, 1.4360e-04, 9.2397e-05, 1.0440e-04], device='cuda:1') 2023-03-07 16:03:12,917 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.70 vs. limit=2.0 2023-03-07 16:03:41,844 INFO [train2.py:809] (1/4) Epoch 4, batch 1350, loss[ctc_loss=0.2048, att_loss=0.3022, loss=0.2827, over 17016.00 frames. utt_duration=1336 frames, utt_pad_proportion=0.007977, over 51.00 utterances.], tot_loss[ctc_loss=0.1836, att_loss=0.293, loss=0.2711, over 3270652.63 frames. utt_duration=1247 frames, utt_pad_proportion=0.05369, over 10504.01 utterances.], batch size: 51, lr: 2.58e-02, grad_scale: 8.0 2023-03-07 16:03:48,889 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.865e+02 3.582e+02 4.373e+02 5.653e+02 1.802e+03, threshold=8.745e+02, percent-clipped=3.0 2023-03-07 16:03:55,784 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.34 vs. limit=5.0 2023-03-07 16:05:02,321 INFO [train2.py:809] (1/4) Epoch 4, batch 1400, loss[ctc_loss=0.1823, att_loss=0.3028, loss=0.2787, over 17382.00 frames. utt_duration=881.5 frames, utt_pad_proportion=0.0731, over 79.00 utterances.], tot_loss[ctc_loss=0.183, att_loss=0.2927, loss=0.2707, over 3264225.36 frames. utt_duration=1249 frames, utt_pad_proportion=0.05564, over 10468.11 utterances.], batch size: 79, lr: 2.58e-02, grad_scale: 8.0 2023-03-07 16:05:15,404 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0333, 4.3177, 4.6733, 4.7659, 2.1454, 4.5002, 2.1290, 2.3252], device='cuda:1'), covar=tensor([0.0213, 0.0185, 0.0614, 0.0159, 0.3002, 0.0173, 0.1884, 0.1641], device='cuda:1'), in_proj_covar=tensor([0.0094, 0.0092, 0.0217, 0.0108, 0.0220, 0.0096, 0.0197, 0.0187], device='cuda:1'), out_proj_covar=tensor([8.7591e-05, 8.5254e-05, 1.8315e-04, 9.1864e-05, 1.8163e-04, 8.6205e-05, 1.6452e-04, 1.5651e-04], device='cuda:1') 2023-03-07 16:06:22,103 INFO [train2.py:809] (1/4) Epoch 4, batch 1450, loss[ctc_loss=0.2691, att_loss=0.3283, loss=0.3164, over 13885.00 frames. utt_duration=381.8 frames, utt_pad_proportion=0.3325, over 146.00 utterances.], tot_loss[ctc_loss=0.1847, att_loss=0.2928, loss=0.2712, over 3262128.74 frames. utt_duration=1241 frames, utt_pad_proportion=0.0593, over 10529.20 utterances.], batch size: 146, lr: 2.58e-02, grad_scale: 8.0 2023-03-07 16:06:28,371 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.701e+02 3.999e+02 4.753e+02 5.960e+02 1.214e+03, threshold=9.507e+02, percent-clipped=6.0 2023-03-07 16:06:47,402 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=13419.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 16:07:42,518 INFO [train2.py:809] (1/4) Epoch 4, batch 1500, loss[ctc_loss=0.1936, att_loss=0.2914, loss=0.2718, over 16272.00 frames. utt_duration=1515 frames, utt_pad_proportion=0.006995, over 43.00 utterances.], tot_loss[ctc_loss=0.1844, att_loss=0.2925, loss=0.2709, over 3257050.34 frames. utt_duration=1229 frames, utt_pad_proportion=0.0646, over 10610.60 utterances.], batch size: 43, lr: 2.57e-02, grad_scale: 8.0 2023-03-07 16:08:04,759 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=13467.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 16:09:03,097 INFO [train2.py:809] (1/4) Epoch 4, batch 1550, loss[ctc_loss=0.2076, att_loss=0.3193, loss=0.297, over 17333.00 frames. utt_duration=1177 frames, utt_pad_proportion=0.02259, over 59.00 utterances.], tot_loss[ctc_loss=0.1836, att_loss=0.292, loss=0.2703, over 3261385.80 frames. utt_duration=1235 frames, utt_pad_proportion=0.0613, over 10572.50 utterances.], batch size: 59, lr: 2.57e-02, grad_scale: 8.0 2023-03-07 16:09:09,259 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.195e+02 3.814e+02 4.608e+02 5.940e+02 1.228e+03, threshold=9.215e+02, percent-clipped=5.0 2023-03-07 16:10:23,310 INFO [train2.py:809] (1/4) Epoch 4, batch 1600, loss[ctc_loss=0.2169, att_loss=0.3028, loss=0.2856, over 16775.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.006141, over 48.00 utterances.], tot_loss[ctc_loss=0.1842, att_loss=0.2928, loss=0.2711, over 3269306.13 frames. utt_duration=1227 frames, utt_pad_proportion=0.06092, over 10670.80 utterances.], batch size: 48, lr: 2.56e-02, grad_scale: 8.0 2023-03-07 16:11:11,096 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7035, 3.3833, 3.8138, 2.8632, 3.7574, 4.6772, 4.6661, 3.3314], device='cuda:1'), covar=tensor([0.0326, 0.1172, 0.0891, 0.1329, 0.0925, 0.0356, 0.0270, 0.1423], device='cuda:1'), in_proj_covar=tensor([0.0178, 0.0186, 0.0171, 0.0168, 0.0185, 0.0145, 0.0132, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-07 16:11:33,280 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.7408, 5.8521, 5.2608, 5.8213, 5.4688, 5.2637, 5.3151, 5.1158], device='cuda:1'), covar=tensor([0.0861, 0.0716, 0.0657, 0.0559, 0.0597, 0.1172, 0.1617, 0.1719], device='cuda:1'), in_proj_covar=tensor([0.0302, 0.0336, 0.0274, 0.0261, 0.0254, 0.0345, 0.0366, 0.0355], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-07 16:11:42,752 INFO [train2.py:809] (1/4) Epoch 4, batch 1650, loss[ctc_loss=0.1784, att_loss=0.299, loss=0.2749, over 16957.00 frames. utt_duration=1358 frames, utt_pad_proportion=0.007208, over 50.00 utterances.], tot_loss[ctc_loss=0.184, att_loss=0.2934, loss=0.2715, over 3275751.62 frames. utt_duration=1228 frames, utt_pad_proportion=0.05884, over 10679.14 utterances.], batch size: 50, lr: 2.56e-02, grad_scale: 8.0 2023-03-07 16:11:48,975 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.288e+02 3.919e+02 5.311e+02 6.584e+02 2.912e+03, threshold=1.062e+03, percent-clipped=12.0 2023-03-07 16:13:01,506 INFO [train2.py:809] (1/4) Epoch 4, batch 1700, loss[ctc_loss=0.1617, att_loss=0.2796, loss=0.256, over 15774.00 frames. utt_duration=1662 frames, utt_pad_proportion=0.008495, over 38.00 utterances.], tot_loss[ctc_loss=0.1841, att_loss=0.2936, loss=0.2717, over 3276855.30 frames. utt_duration=1236 frames, utt_pad_proportion=0.0574, over 10617.58 utterances.], batch size: 38, lr: 2.55e-02, grad_scale: 8.0 2023-03-07 16:13:44,590 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.70 vs. limit=2.0 2023-03-07 16:14:20,719 INFO [train2.py:809] (1/4) Epoch 4, batch 1750, loss[ctc_loss=0.1682, att_loss=0.2755, loss=0.254, over 16186.00 frames. utt_duration=1581 frames, utt_pad_proportion=0.005798, over 41.00 utterances.], tot_loss[ctc_loss=0.1832, att_loss=0.293, loss=0.2711, over 3278439.51 frames. utt_duration=1242 frames, utt_pad_proportion=0.05475, over 10573.53 utterances.], batch size: 41, lr: 2.55e-02, grad_scale: 8.0 2023-03-07 16:14:27,018 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.465e+02 3.709e+02 4.336e+02 5.273e+02 1.516e+03, threshold=8.672e+02, percent-clipped=2.0 2023-03-07 16:15:15,149 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=13737.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 16:15:39,466 INFO [train2.py:809] (1/4) Epoch 4, batch 1800, loss[ctc_loss=0.1759, att_loss=0.308, loss=0.2816, over 16875.00 frames. utt_duration=1379 frames, utt_pad_proportion=0.007302, over 49.00 utterances.], tot_loss[ctc_loss=0.1831, att_loss=0.2925, loss=0.2707, over 3272213.59 frames. utt_duration=1251 frames, utt_pad_proportion=0.0553, over 10475.86 utterances.], batch size: 49, lr: 2.55e-02, grad_scale: 8.0 2023-03-07 16:16:14,106 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-03-07 16:16:44,025 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([1.8961, 2.0135, 2.5029, 1.8529, 2.8227, 2.9986, 1.9899, 1.3680], device='cuda:1'), covar=tensor([0.1501, 0.1369, 0.0750, 0.2989, 0.1375, 0.2107, 0.1145, 0.5660], device='cuda:1'), in_proj_covar=tensor([0.0062, 0.0057, 0.0059, 0.0072, 0.0055, 0.0067, 0.0061, 0.0090], device='cuda:1'), out_proj_covar=tensor([3.7973e-05, 3.5116e-05, 3.5607e-05, 4.7313e-05, 3.6964e-05, 4.5133e-05, 3.9638e-05, 6.4302e-05], device='cuda:1') 2023-03-07 16:16:48,529 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.7021, 4.0182, 3.8496, 3.9438, 4.1147, 4.0455, 3.8828, 3.8167], device='cuda:1'), covar=tensor([0.1038, 0.0452, 0.0319, 0.0425, 0.0275, 0.0303, 0.0265, 0.0323], device='cuda:1'), in_proj_covar=tensor([0.0333, 0.0200, 0.0138, 0.0178, 0.0222, 0.0241, 0.0187, 0.0210], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-07 16:16:51,918 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=13798.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 16:16:59,426 INFO [train2.py:809] (1/4) Epoch 4, batch 1850, loss[ctc_loss=0.2081, att_loss=0.3005, loss=0.282, over 16025.00 frames. utt_duration=1604 frames, utt_pad_proportion=0.006335, over 40.00 utterances.], tot_loss[ctc_loss=0.183, att_loss=0.2925, loss=0.2706, over 3268885.40 frames. utt_duration=1245 frames, utt_pad_proportion=0.05722, over 10518.10 utterances.], batch size: 40, lr: 2.54e-02, grad_scale: 8.0 2023-03-07 16:17:05,503 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.641e+02 3.804e+02 4.653e+02 6.004e+02 1.107e+03, threshold=9.305e+02, percent-clipped=7.0 2023-03-07 16:18:19,882 INFO [train2.py:809] (1/4) Epoch 4, batch 1900, loss[ctc_loss=0.2208, att_loss=0.3217, loss=0.3015, over 17298.00 frames. utt_duration=1259 frames, utt_pad_proportion=0.01226, over 55.00 utterances.], tot_loss[ctc_loss=0.182, att_loss=0.292, loss=0.27, over 3275734.97 frames. utt_duration=1255 frames, utt_pad_proportion=0.05318, over 10452.12 utterances.], batch size: 55, lr: 2.54e-02, grad_scale: 8.0 2023-03-07 16:19:39,439 INFO [train2.py:809] (1/4) Epoch 4, batch 1950, loss[ctc_loss=0.1729, att_loss=0.2995, loss=0.2742, over 17074.00 frames. utt_duration=1315 frames, utt_pad_proportion=0.007547, over 52.00 utterances.], tot_loss[ctc_loss=0.1826, att_loss=0.292, loss=0.2701, over 3270062.69 frames. utt_duration=1245 frames, utt_pad_proportion=0.05733, over 10516.09 utterances.], batch size: 52, lr: 2.53e-02, grad_scale: 8.0 2023-03-07 16:19:41,943 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-03-07 16:19:45,526 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.392e+02 3.636e+02 4.382e+02 5.668e+02 1.213e+03, threshold=8.764e+02, percent-clipped=2.0 2023-03-07 16:20:10,238 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9263, 2.6390, 5.0460, 3.8221, 3.2976, 4.4296, 4.4615, 4.6009], device='cuda:1'), covar=tensor([0.0118, 0.1644, 0.0205, 0.1257, 0.2327, 0.0297, 0.0202, 0.0293], device='cuda:1'), in_proj_covar=tensor([0.0121, 0.0235, 0.0120, 0.0288, 0.0333, 0.0179, 0.0109, 0.0129], device='cuda:1'), out_proj_covar=tensor([1.0073e-04, 1.7997e-04, 9.8238e-05, 2.2782e-04, 2.5602e-04, 1.4708e-04, 9.2549e-05, 1.0922e-04], device='cuda:1') 2023-03-07 16:21:00,024 INFO [train2.py:809] (1/4) Epoch 4, batch 2000, loss[ctc_loss=0.1596, att_loss=0.2818, loss=0.2574, over 16466.00 frames. utt_duration=1433 frames, utt_pad_proportion=0.006704, over 46.00 utterances.], tot_loss[ctc_loss=0.1823, att_loss=0.2916, loss=0.2697, over 3272790.54 frames. utt_duration=1250 frames, utt_pad_proportion=0.05477, over 10485.69 utterances.], batch size: 46, lr: 2.53e-02, grad_scale: 8.0 2023-03-07 16:21:30,787 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7514, 4.8520, 4.7446, 3.5281, 4.6311, 4.1416, 4.2386, 2.7077], device='cuda:1'), covar=tensor([0.0104, 0.0069, 0.0143, 0.0596, 0.0080, 0.0163, 0.0229, 0.1266], device='cuda:1'), in_proj_covar=tensor([0.0049, 0.0053, 0.0044, 0.0086, 0.0050, 0.0063, 0.0072, 0.0099], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-03-07 16:22:23,362 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.57 vs. limit=5.0 2023-03-07 16:22:23,890 INFO [train2.py:809] (1/4) Epoch 4, batch 2050, loss[ctc_loss=0.1768, att_loss=0.2739, loss=0.2545, over 16182.00 frames. utt_duration=1580 frames, utt_pad_proportion=0.006791, over 41.00 utterances.], tot_loss[ctc_loss=0.1831, att_loss=0.2923, loss=0.2705, over 3273371.08 frames. utt_duration=1225 frames, utt_pad_proportion=0.05976, over 10699.59 utterances.], batch size: 41, lr: 2.53e-02, grad_scale: 8.0 2023-03-07 16:22:30,194 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.271e+02 3.689e+02 4.518e+02 6.015e+02 1.676e+03, threshold=9.035e+02, percent-clipped=8.0 2023-03-07 16:23:43,382 INFO [train2.py:809] (1/4) Epoch 4, batch 2100, loss[ctc_loss=0.1919, att_loss=0.3075, loss=0.2844, over 17301.00 frames. utt_duration=1174 frames, utt_pad_proportion=0.0237, over 59.00 utterances.], tot_loss[ctc_loss=0.1818, att_loss=0.2916, loss=0.2697, over 3277301.71 frames. utt_duration=1254 frames, utt_pad_proportion=0.05194, over 10463.95 utterances.], batch size: 59, lr: 2.52e-02, grad_scale: 8.0 2023-03-07 16:24:14,191 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14072.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 16:24:44,971 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.7810, 2.1020, 3.2899, 4.3797, 4.0356, 4.4683, 2.5894, 1.8212], device='cuda:1'), covar=tensor([0.0483, 0.2688, 0.1150, 0.0398, 0.0333, 0.0144, 0.1912, 0.2796], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0185, 0.0180, 0.0134, 0.0114, 0.0105, 0.0185, 0.0169], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-07 16:24:47,822 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14093.0, num_to_drop=1, layers_to_drop={3} 2023-03-07 16:25:03,074 INFO [train2.py:809] (1/4) Epoch 4, batch 2150, loss[ctc_loss=0.1659, att_loss=0.2838, loss=0.2602, over 16346.00 frames. utt_duration=1455 frames, utt_pad_proportion=0.005062, over 45.00 utterances.], tot_loss[ctc_loss=0.1791, att_loss=0.29, loss=0.2678, over 3275902.72 frames. utt_duration=1283 frames, utt_pad_proportion=0.04523, over 10228.69 utterances.], batch size: 45, lr: 2.52e-02, grad_scale: 8.0 2023-03-07 16:25:03,416 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14103.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 16:25:09,250 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.138e+02 3.723e+02 4.436e+02 6.346e+02 3.042e+03, threshold=8.872e+02, percent-clipped=8.0 2023-03-07 16:25:30,146 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14120.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 16:25:51,832 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14133.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 16:25:55,605 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-03-07 16:26:23,571 INFO [train2.py:809] (1/4) Epoch 4, batch 2200, loss[ctc_loss=0.1701, att_loss=0.2957, loss=0.2706, over 16883.00 frames. utt_duration=1380 frames, utt_pad_proportion=0.007503, over 49.00 utterances.], tot_loss[ctc_loss=0.1801, att_loss=0.2907, loss=0.2686, over 3277604.90 frames. utt_duration=1265 frames, utt_pad_proportion=0.04914, over 10379.02 utterances.], batch size: 49, lr: 2.51e-02, grad_scale: 8.0 2023-03-07 16:26:41,854 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14164.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 16:27:09,497 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14181.0, num_to_drop=1, layers_to_drop={2} 2023-03-07 16:27:32,315 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4061, 4.2541, 4.4184, 4.6962, 1.6612, 4.6683, 2.2842, 1.8518], device='cuda:1'), covar=tensor([0.0153, 0.0238, 0.0787, 0.0241, 0.4057, 0.0201, 0.2058, 0.2074], device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0092, 0.0227, 0.0111, 0.0229, 0.0100, 0.0208, 0.0194], device='cuda:1'), out_proj_covar=tensor([9.1941e-05, 8.8193e-05, 1.9353e-04, 9.6101e-05, 1.9124e-04, 9.2107e-05, 1.7463e-04, 1.6344e-04], device='cuda:1') 2023-03-07 16:27:44,783 INFO [train2.py:809] (1/4) Epoch 4, batch 2250, loss[ctc_loss=0.1867, att_loss=0.3221, loss=0.295, over 17318.00 frames. utt_duration=1261 frames, utt_pad_proportion=0.01105, over 55.00 utterances.], tot_loss[ctc_loss=0.1813, att_loss=0.292, loss=0.2699, over 3284443.47 frames. utt_duration=1256 frames, utt_pad_proportion=0.04895, over 10475.41 utterances.], batch size: 55, lr: 2.51e-02, grad_scale: 8.0 2023-03-07 16:27:50,861 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.200e+02 3.354e+02 4.208e+02 5.126e+02 1.091e+03, threshold=8.417e+02, percent-clipped=2.0 2023-03-07 16:28:33,706 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-03-07 16:29:04,196 INFO [train2.py:809] (1/4) Epoch 4, batch 2300, loss[ctc_loss=0.1725, att_loss=0.2891, loss=0.2658, over 16536.00 frames. utt_duration=1472 frames, utt_pad_proportion=0.006242, over 45.00 utterances.], tot_loss[ctc_loss=0.1807, att_loss=0.2914, loss=0.2692, over 3278778.84 frames. utt_duration=1257 frames, utt_pad_proportion=0.05097, over 10446.84 utterances.], batch size: 45, lr: 2.51e-02, grad_scale: 8.0 2023-03-07 16:29:49,927 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.3235, 3.7296, 2.9581, 3.2816, 3.7921, 3.6700, 2.5399, 4.4710], device='cuda:1'), covar=tensor([0.1271, 0.0475, 0.1399, 0.0737, 0.0477, 0.0561, 0.1088, 0.0254], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0112, 0.0166, 0.0132, 0.0132, 0.0158, 0.0139, 0.0109], 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') 2023-03-07 16:30:10,037 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-03-07 16:30:13,694 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.6658, 3.8883, 3.8191, 3.8828, 4.0636, 4.0427, 3.8771, 3.8679], device='cuda:1'), covar=tensor([0.1346, 0.0678, 0.0299, 0.0536, 0.0349, 0.0360, 0.0283, 0.0314], device='cuda:1'), in_proj_covar=tensor([0.0354, 0.0210, 0.0143, 0.0187, 0.0229, 0.0247, 0.0191, 0.0211], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0003, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-07 16:30:24,650 INFO [train2.py:809] (1/4) Epoch 4, batch 2350, loss[ctc_loss=0.1767, att_loss=0.3086, loss=0.2822, over 17337.00 frames. utt_duration=1177 frames, utt_pad_proportion=0.02257, over 59.00 utterances.], tot_loss[ctc_loss=0.1803, att_loss=0.2913, loss=0.2691, over 3280683.02 frames. utt_duration=1235 frames, utt_pad_proportion=0.05659, over 10637.07 utterances.], batch size: 59, lr: 2.50e-02, grad_scale: 8.0 2023-03-07 16:30:31,148 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.697e+02 3.860e+02 4.546e+02 5.797e+02 1.491e+03, threshold=9.092e+02, percent-clipped=6.0 2023-03-07 16:31:01,905 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2700, 4.7112, 4.5403, 4.7498, 4.7197, 4.5405, 4.0704, 4.5429], device='cuda:1'), covar=tensor([0.0096, 0.0094, 0.0077, 0.0087, 0.0078, 0.0077, 0.0326, 0.0182], device='cuda:1'), in_proj_covar=tensor([0.0048, 0.0047, 0.0048, 0.0036, 0.0035, 0.0043, 0.0065, 0.0060], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-07 16:31:44,319 INFO [train2.py:809] (1/4) Epoch 4, batch 2400, loss[ctc_loss=0.1265, att_loss=0.2538, loss=0.2284, over 16107.00 frames. utt_duration=1535 frames, utt_pad_proportion=0.006884, over 42.00 utterances.], tot_loss[ctc_loss=0.1814, att_loss=0.292, loss=0.2699, over 3281250.76 frames. utt_duration=1215 frames, utt_pad_proportion=0.06165, over 10812.25 utterances.], batch size: 42, lr: 2.50e-02, grad_scale: 8.0 2023-03-07 16:32:48,044 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14393.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 16:33:03,729 INFO [train2.py:809] (1/4) Epoch 4, batch 2450, loss[ctc_loss=0.1891, att_loss=0.3041, loss=0.2811, over 16751.00 frames. utt_duration=1397 frames, utt_pad_proportion=0.007487, over 48.00 utterances.], tot_loss[ctc_loss=0.1788, att_loss=0.2897, loss=0.2675, over 3271800.37 frames. utt_duration=1258 frames, utt_pad_proportion=0.05328, over 10411.97 utterances.], batch size: 48, lr: 2.49e-02, grad_scale: 8.0 2023-03-07 16:33:09,772 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.088e+02 3.936e+02 4.992e+02 6.745e+02 1.784e+03, threshold=9.985e+02, percent-clipped=4.0 2023-03-07 16:33:20,188 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14413.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 16:33:42,755 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7582, 5.1474, 4.9896, 4.7356, 5.2817, 5.2351, 5.0179, 4.7698], device='cuda:1'), covar=tensor([0.1097, 0.0417, 0.0213, 0.0671, 0.0309, 0.0240, 0.0213, 0.0263], device='cuda:1'), in_proj_covar=tensor([0.0355, 0.0207, 0.0142, 0.0188, 0.0233, 0.0251, 0.0193, 0.0213], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0003, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-07 16:33:44,300 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14428.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 16:33:44,350 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8157, 5.2501, 4.8785, 5.3010, 4.7762, 5.0456, 5.4335, 5.2556], device='cuda:1'), covar=tensor([0.0299, 0.0232, 0.0506, 0.0111, 0.0336, 0.0124, 0.0198, 0.0127], device='cuda:1'), in_proj_covar=tensor([0.0168, 0.0149, 0.0187, 0.0117, 0.0167, 0.0118, 0.0146, 0.0134], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-03-07 16:34:04,968 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=14441.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 16:34:17,548 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.8967, 2.5788, 3.9167, 3.5544, 3.1223, 3.6746, 3.4548, 3.9079], device='cuda:1'), covar=tensor([0.0177, 0.1276, 0.0181, 0.0935, 0.1689, 0.0377, 0.0207, 0.0247], device='cuda:1'), in_proj_covar=tensor([0.0126, 0.0230, 0.0122, 0.0285, 0.0324, 0.0176, 0.0107, 0.0129], device='cuda:1'), out_proj_covar=tensor([1.0651e-04, 1.7764e-04, 1.0058e-04, 2.2723e-04, 2.5127e-04, 1.4444e-04, 9.1209e-05, 1.0820e-04], device='cuda:1') 2023-03-07 16:34:23,281 INFO [train2.py:809] (1/4) Epoch 4, batch 2500, loss[ctc_loss=0.1349, att_loss=0.2548, loss=0.2309, over 15775.00 frames. utt_duration=1662 frames, utt_pad_proportion=0.008992, over 38.00 utterances.], tot_loss[ctc_loss=0.1786, att_loss=0.2894, loss=0.2672, over 3268322.20 frames. utt_duration=1250 frames, utt_pad_proportion=0.05654, over 10474.05 utterances.], batch size: 38, lr: 2.49e-02, grad_scale: 8.0 2023-03-07 16:34:33,331 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14459.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 16:34:57,608 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14474.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 16:35:00,379 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14476.0, num_to_drop=1, layers_to_drop={2} 2023-03-07 16:35:14,939 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2708, 4.4695, 4.7507, 4.9635, 2.3410, 4.6462, 2.7703, 1.9374], device='cuda:1'), covar=tensor([0.0216, 0.0184, 0.0634, 0.0246, 0.3249, 0.0223, 0.1852, 0.2166], device='cuda:1'), in_proj_covar=tensor([0.0101, 0.0095, 0.0229, 0.0114, 0.0231, 0.0102, 0.0212, 0.0195], device='cuda:1'), out_proj_covar=tensor([9.4854e-05, 9.0263e-05, 1.9579e-04, 1.0097e-04, 1.9324e-04, 9.4983e-05, 1.7820e-04, 1.6421e-04], device='cuda:1') 2023-03-07 16:35:17,961 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14487.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 16:35:42,396 INFO [train2.py:809] (1/4) Epoch 4, batch 2550, loss[ctc_loss=0.2116, att_loss=0.3138, loss=0.2934, over 17050.00 frames. utt_duration=1288 frames, utt_pad_proportion=0.009144, over 53.00 utterances.], tot_loss[ctc_loss=0.1792, att_loss=0.2899, loss=0.2677, over 3261237.40 frames. utt_duration=1238 frames, utt_pad_proportion=0.05977, over 10550.61 utterances.], batch size: 53, lr: 2.49e-02, grad_scale: 8.0 2023-03-07 16:35:48,914 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.665e+02 3.493e+02 4.541e+02 5.557e+02 1.503e+03, threshold=9.083e+02, percent-clipped=2.0 2023-03-07 16:36:54,440 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14548.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 16:37:02,527 INFO [train2.py:809] (1/4) Epoch 4, batch 2600, loss[ctc_loss=0.1544, att_loss=0.256, loss=0.2357, over 14502.00 frames. utt_duration=1815 frames, utt_pad_proportion=0.03217, over 32.00 utterances.], tot_loss[ctc_loss=0.1804, att_loss=0.2909, loss=0.2688, over 3258099.41 frames. utt_duration=1211 frames, utt_pad_proportion=0.0687, over 10778.14 utterances.], batch size: 32, lr: 2.48e-02, grad_scale: 8.0 2023-03-07 16:37:26,437 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([1.5341, 3.0241, 2.5950, 2.7829, 2.8539, 2.8887, 1.8911, 1.2872], device='cuda:1'), covar=tensor([0.1171, 0.0617, 0.0934, 0.1236, 0.0980, 0.1915, 0.1473, 0.8652], device='cuda:1'), in_proj_covar=tensor([0.0065, 0.0057, 0.0060, 0.0070, 0.0057, 0.0071, 0.0060, 0.0094], device='cuda:1'), out_proj_covar=tensor([4.1056e-05, 3.5790e-05, 3.7061e-05, 4.7702e-05, 3.7717e-05, 4.9507e-05, 4.0811e-05, 6.7753e-05], device='cuda:1') 2023-03-07 16:37:27,119 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.75 vs. limit=2.0 2023-03-07 16:38:21,813 INFO [train2.py:809] (1/4) Epoch 4, batch 2650, loss[ctc_loss=0.1654, att_loss=0.29, loss=0.2651, over 17385.00 frames. utt_duration=1105 frames, utt_pad_proportion=0.0339, over 63.00 utterances.], tot_loss[ctc_loss=0.1781, att_loss=0.2891, loss=0.2669, over 3258808.48 frames. utt_duration=1249 frames, utt_pad_proportion=0.05946, over 10448.91 utterances.], batch size: 63, lr: 2.48e-02, grad_scale: 8.0 2023-03-07 16:38:28,424 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.304e+02 3.793e+02 4.783e+02 6.009e+02 1.165e+03, threshold=9.567e+02, percent-clipped=2.0 2023-03-07 16:38:36,544 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7830, 4.0731, 3.3437, 3.6516, 4.1105, 3.6723, 2.6784, 4.7144], device='cuda:1'), covar=tensor([0.1138, 0.0376, 0.1056, 0.0596, 0.0515, 0.0755, 0.1075, 0.0299], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0115, 0.0170, 0.0135, 0.0137, 0.0163, 0.0145, 0.0114], 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') 2023-03-07 16:39:41,961 INFO [train2.py:809] (1/4) Epoch 4, batch 2700, loss[ctc_loss=0.1771, att_loss=0.2989, loss=0.2746, over 17442.00 frames. utt_duration=1013 frames, utt_pad_proportion=0.04453, over 69.00 utterances.], tot_loss[ctc_loss=0.1765, att_loss=0.2879, loss=0.2656, over 3252300.85 frames. utt_duration=1277 frames, utt_pad_proportion=0.05428, over 10197.89 utterances.], batch size: 69, lr: 2.48e-02, grad_scale: 8.0 2023-03-07 16:40:04,086 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-03-07 16:40:37,833 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7787, 4.6313, 4.4548, 4.6468, 4.9118, 4.7154, 4.4765, 2.0751], device='cuda:1'), covar=tensor([0.0166, 0.0275, 0.0301, 0.0164, 0.1045, 0.0174, 0.0303, 0.3199], device='cuda:1'), in_proj_covar=tensor([0.0129, 0.0113, 0.0110, 0.0113, 0.0251, 0.0128, 0.0103, 0.0245], device='cuda:1'), out_proj_covar=tensor([1.1882e-04, 1.0097e-04, 9.9504e-05, 1.0780e-04, 2.3227e-04, 1.1851e-04, 9.9846e-05, 2.2214e-04], device='cuda:1') 2023-03-07 16:41:01,175 INFO [train2.py:809] (1/4) Epoch 4, batch 2750, loss[ctc_loss=0.1524, att_loss=0.283, loss=0.2568, over 16332.00 frames. utt_duration=1454 frames, utt_pad_proportion=0.005791, over 45.00 utterances.], tot_loss[ctc_loss=0.1775, att_loss=0.289, loss=0.2667, over 3267632.92 frames. utt_duration=1289 frames, utt_pad_proportion=0.04686, over 10149.55 utterances.], batch size: 45, lr: 2.47e-02, grad_scale: 8.0 2023-03-07 16:41:07,359 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.333e+02 3.898e+02 4.978e+02 5.848e+02 1.042e+03, threshold=9.955e+02, percent-clipped=2.0 2023-03-07 16:41:42,274 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14728.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 16:42:02,308 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14740.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 16:42:23,121 INFO [train2.py:809] (1/4) Epoch 4, batch 2800, loss[ctc_loss=0.1301, att_loss=0.2456, loss=0.2225, over 15357.00 frames. utt_duration=1757 frames, utt_pad_proportion=0.01207, over 35.00 utterances.], tot_loss[ctc_loss=0.1763, att_loss=0.2882, loss=0.2659, over 3266497.69 frames. utt_duration=1314 frames, utt_pad_proportion=0.04082, over 9952.35 utterances.], batch size: 35, lr: 2.47e-02, grad_scale: 8.0 2023-03-07 16:42:32,809 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14759.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 16:42:48,941 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14769.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 16:43:00,440 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=14776.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 16:43:00,641 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14776.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 16:43:28,063 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8475, 5.2368, 5.0139, 5.1243, 5.2379, 5.2076, 5.0758, 4.7430], device='cuda:1'), covar=tensor([0.0938, 0.0507, 0.0269, 0.0431, 0.0276, 0.0245, 0.0220, 0.0294], device='cuda:1'), in_proj_covar=tensor([0.0346, 0.0204, 0.0143, 0.0183, 0.0231, 0.0252, 0.0191, 0.0213], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-07 16:43:41,693 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14801.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 16:43:44,566 INFO [train2.py:809] (1/4) Epoch 4, batch 2850, loss[ctc_loss=0.1831, att_loss=0.2971, loss=0.2743, over 16468.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.006342, over 46.00 utterances.], tot_loss[ctc_loss=0.176, att_loss=0.2878, loss=0.2654, over 3261722.72 frames. utt_duration=1303 frames, utt_pad_proportion=0.04412, over 10022.16 utterances.], batch size: 46, lr: 2.46e-02, grad_scale: 8.0 2023-03-07 16:43:50,719 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.067e+02 3.258e+02 4.280e+02 5.701e+02 1.148e+03, threshold=8.559e+02, percent-clipped=5.0 2023-03-07 16:43:50,891 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=14807.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 16:44:18,605 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=14824.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 16:44:22,501 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-03-07 16:44:48,691 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14843.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 16:45:04,281 INFO [train2.py:809] (1/4) Epoch 4, batch 2900, loss[ctc_loss=0.1874, att_loss=0.2985, loss=0.2763, over 16977.00 frames. utt_duration=687.5 frames, utt_pad_proportion=0.133, over 99.00 utterances.], tot_loss[ctc_loss=0.1776, att_loss=0.2891, loss=0.2668, over 3265708.78 frames. utt_duration=1283 frames, utt_pad_proportion=0.04734, over 10191.57 utterances.], batch size: 99, lr: 2.46e-02, grad_scale: 8.0 2023-03-07 16:45:49,736 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5037, 3.2349, 3.8615, 2.5999, 3.6516, 4.5732, 4.4615, 3.2199], device='cuda:1'), covar=tensor([0.0384, 0.1337, 0.0678, 0.1477, 0.0805, 0.0335, 0.0571, 0.1318], device='cuda:1'), in_proj_covar=tensor([0.0182, 0.0187, 0.0179, 0.0172, 0.0192, 0.0162, 0.0143, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-07 16:45:50,353 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-03-07 16:46:25,743 INFO [train2.py:809] (1/4) Epoch 4, batch 2950, loss[ctc_loss=0.1668, att_loss=0.2943, loss=0.2688, over 17387.00 frames. utt_duration=1106 frames, utt_pad_proportion=0.03275, over 63.00 utterances.], tot_loss[ctc_loss=0.1767, att_loss=0.2887, loss=0.2663, over 3269057.79 frames. utt_duration=1275 frames, utt_pad_proportion=0.04878, over 10266.33 utterances.], batch size: 63, lr: 2.46e-02, grad_scale: 8.0 2023-03-07 16:46:32,080 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.413e+02 3.902e+02 4.882e+02 6.670e+02 1.203e+03, threshold=9.763e+02, percent-clipped=9.0 2023-03-07 16:47:01,441 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.81 vs. limit=2.0 2023-03-07 16:47:24,142 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6775, 4.9430, 4.7816, 4.7262, 5.1374, 5.0654, 4.7503, 4.6038], device='cuda:1'), covar=tensor([0.0992, 0.0524, 0.0287, 0.0632, 0.0253, 0.0236, 0.0257, 0.0313], device='cuda:1'), in_proj_covar=tensor([0.0344, 0.0205, 0.0142, 0.0187, 0.0231, 0.0253, 0.0193, 0.0215], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-07 16:47:46,084 INFO [train2.py:809] (1/4) Epoch 4, batch 3000, loss[ctc_loss=0.1806, att_loss=0.2991, loss=0.2754, over 17055.00 frames. utt_duration=1313 frames, utt_pad_proportion=0.008853, over 52.00 utterances.], tot_loss[ctc_loss=0.1768, att_loss=0.289, loss=0.2666, over 3271807.31 frames. utt_duration=1279 frames, utt_pad_proportion=0.04868, over 10246.96 utterances.], batch size: 52, lr: 2.45e-02, grad_scale: 16.0 2023-03-07 16:47:46,084 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-07 16:48:00,421 INFO [train2.py:843] (1/4) Epoch 4, validation: ctc_loss=0.08782, att_loss=0.2564, loss=0.2227, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-07 16:48:00,422 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-07 16:48:20,267 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-03-07 16:48:36,258 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-03-07 16:49:20,423 INFO [train2.py:809] (1/4) Epoch 4, batch 3050, loss[ctc_loss=0.1546, att_loss=0.2657, loss=0.2435, over 15747.00 frames. utt_duration=1659 frames, utt_pad_proportion=0.01008, over 38.00 utterances.], tot_loss[ctc_loss=0.1776, att_loss=0.2901, loss=0.2676, over 3275853.42 frames. utt_duration=1282 frames, utt_pad_proportion=0.0451, over 10236.65 utterances.], batch size: 38, lr: 2.45e-02, grad_scale: 16.0 2023-03-07 16:49:25,337 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.8274, 2.5380, 3.1180, 4.2920, 4.1936, 4.5504, 2.8926, 1.8658], device='cuda:1'), covar=tensor([0.0516, 0.2439, 0.1319, 0.0525, 0.0317, 0.0117, 0.1787, 0.2750], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0185, 0.0179, 0.0139, 0.0123, 0.0105, 0.0182, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-07 16:49:26,541 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.237e+02 3.805e+02 4.579e+02 6.197e+02 1.095e+03, threshold=9.158e+02, percent-clipped=2.0 2023-03-07 16:50:40,669 INFO [train2.py:809] (1/4) Epoch 4, batch 3100, loss[ctc_loss=0.1336, att_loss=0.2545, loss=0.2303, over 15478.00 frames. utt_duration=1721 frames, utt_pad_proportion=0.009291, over 36.00 utterances.], tot_loss[ctc_loss=0.1774, att_loss=0.2902, loss=0.2677, over 3281318.88 frames. utt_duration=1256 frames, utt_pad_proportion=0.04948, over 10458.43 utterances.], batch size: 36, lr: 2.45e-02, grad_scale: 16.0 2023-03-07 16:51:06,256 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15069.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 16:51:50,412 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=15096.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 16:52:01,855 INFO [train2.py:809] (1/4) Epoch 4, batch 3150, loss[ctc_loss=0.1804, att_loss=0.3044, loss=0.2796, over 17055.00 frames. utt_duration=1313 frames, utt_pad_proportion=0.008694, over 52.00 utterances.], tot_loss[ctc_loss=0.1761, att_loss=0.2898, loss=0.2671, over 3285820.97 frames. utt_duration=1263 frames, utt_pad_proportion=0.04818, over 10415.55 utterances.], batch size: 52, lr: 2.44e-02, grad_scale: 16.0 2023-03-07 16:52:07,991 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.047e+02 3.829e+02 4.318e+02 5.478e+02 1.412e+03, threshold=8.637e+02, percent-clipped=4.0 2023-03-07 16:52:23,549 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=15117.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 16:53:05,041 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15143.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 16:53:21,397 INFO [train2.py:809] (1/4) Epoch 4, batch 3200, loss[ctc_loss=0.2127, att_loss=0.321, loss=0.2993, over 17075.00 frames. utt_duration=1221 frames, utt_pad_proportion=0.01603, over 56.00 utterances.], tot_loss[ctc_loss=0.1757, att_loss=0.2896, loss=0.2668, over 3289642.76 frames. utt_duration=1251 frames, utt_pad_proportion=0.05002, over 10531.25 utterances.], batch size: 56, lr: 2.44e-02, grad_scale: 16.0 2023-03-07 16:54:07,841 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4846, 4.6110, 4.7389, 4.5854, 1.9935, 4.7650, 2.4029, 2.0637], device='cuda:1'), covar=tensor([0.0197, 0.0175, 0.0626, 0.0304, 0.3270, 0.0176, 0.1972, 0.1728], device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0092, 0.0223, 0.0109, 0.0220, 0.0092, 0.0206, 0.0187], device='cuda:1'), out_proj_covar=tensor([9.2491e-05, 8.8295e-05, 1.9221e-04, 9.6261e-05, 1.8774e-04, 8.7456e-05, 1.7421e-04, 1.5954e-04], device='cuda:1') 2023-03-07 16:54:17,723 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=2.06 vs. limit=2.0 2023-03-07 16:54:21,486 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=15191.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 16:54:28,152 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=15195.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 16:54:42,167 INFO [train2.py:809] (1/4) Epoch 4, batch 3250, loss[ctc_loss=0.149, att_loss=0.2649, loss=0.2417, over 16027.00 frames. utt_duration=1604 frames, utt_pad_proportion=0.00678, over 40.00 utterances.], tot_loss[ctc_loss=0.175, att_loss=0.2888, loss=0.266, over 3285584.42 frames. utt_duration=1266 frames, utt_pad_proportion=0.04604, over 10390.76 utterances.], batch size: 40, lr: 2.44e-02, grad_scale: 16.0 2023-03-07 16:54:44,149 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.0400, 1.9209, 2.0039, 0.9120, 2.8640, 1.2572, 2.2806, 2.6512], device='cuda:1'), covar=tensor([0.0316, 0.1505, 0.1571, 0.2529, 0.0496, 0.1639, 0.0911, 0.0610], device='cuda:1'), in_proj_covar=tensor([0.0093, 0.0090, 0.0094, 0.0083, 0.0077, 0.0080, 0.0090, 0.0087], device='cuda:1'), out_proj_covar=tensor([3.9411e-05, 4.6576e-05, 4.8484e-05, 4.5233e-05, 3.9539e-05, 4.3572e-05, 4.3702e-05, 4.3916e-05], device='cuda:1') 2023-03-07 16:54:48,521 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.387e+02 3.575e+02 4.506e+02 5.643e+02 1.593e+03, threshold=9.012e+02, percent-clipped=6.0 2023-03-07 16:55:01,549 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-03-07 16:55:27,279 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.78 vs. limit=2.0 2023-03-07 16:56:01,883 INFO [train2.py:809] (1/4) Epoch 4, batch 3300, loss[ctc_loss=0.1851, att_loss=0.3055, loss=0.2814, over 17370.00 frames. utt_duration=1105 frames, utt_pad_proportion=0.03365, over 63.00 utterances.], tot_loss[ctc_loss=0.175, att_loss=0.2884, loss=0.2657, over 3277145.26 frames. utt_duration=1250 frames, utt_pad_proportion=0.05331, over 10497.79 utterances.], batch size: 63, lr: 2.43e-02, grad_scale: 16.0 2023-03-07 16:56:06,931 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=15256.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 16:56:51,728 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3252, 2.3625, 3.5757, 2.2937, 3.3125, 4.4408, 4.3829, 2.7572], device='cuda:1'), covar=tensor([0.0376, 0.1775, 0.0884, 0.1634, 0.1017, 0.0461, 0.0370, 0.1566], device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0190, 0.0182, 0.0178, 0.0192, 0.0165, 0.0144, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-07 16:57:22,462 INFO [train2.py:809] (1/4) Epoch 4, batch 3350, loss[ctc_loss=0.1631, att_loss=0.2931, loss=0.2671, over 17035.00 frames. utt_duration=1338 frames, utt_pad_proportion=0.00674, over 51.00 utterances.], tot_loss[ctc_loss=0.1756, att_loss=0.2888, loss=0.2662, over 3277400.83 frames. utt_duration=1243 frames, utt_pad_proportion=0.05612, over 10561.57 utterances.], batch size: 51, lr: 2.43e-02, grad_scale: 16.0 2023-03-07 16:57:28,538 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.491e+02 3.835e+02 4.773e+02 5.805e+02 1.964e+03, threshold=9.546e+02, percent-clipped=8.0 2023-03-07 16:58:35,109 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6924, 5.1646, 4.9956, 4.8299, 5.1604, 5.0366, 4.7915, 4.6314], device='cuda:1'), covar=tensor([0.0955, 0.0322, 0.0171, 0.0452, 0.0227, 0.0234, 0.0241, 0.0288], device='cuda:1'), in_proj_covar=tensor([0.0348, 0.0205, 0.0146, 0.0183, 0.0231, 0.0251, 0.0192, 0.0216], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-07 16:58:42,387 INFO [train2.py:809] (1/4) Epoch 4, batch 3400, loss[ctc_loss=0.1533, att_loss=0.2684, loss=0.2453, over 16280.00 frames. utt_duration=1516 frames, utt_pad_proportion=0.006523, over 43.00 utterances.], tot_loss[ctc_loss=0.1763, att_loss=0.2899, loss=0.2671, over 3281696.69 frames. utt_duration=1242 frames, utt_pad_proportion=0.05373, over 10582.21 utterances.], batch size: 43, lr: 2.42e-02, grad_scale: 16.0 2023-03-07 16:59:20,615 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.1861, 5.3334, 5.7026, 5.7244, 5.3612, 6.1036, 5.2329, 6.1685], device='cuda:1'), covar=tensor([0.0459, 0.0577, 0.0457, 0.0604, 0.1867, 0.0577, 0.0429, 0.0470], device='cuda:1'), in_proj_covar=tensor([0.0469, 0.0304, 0.0306, 0.0367, 0.0527, 0.0306, 0.0255, 0.0321], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-03-07 16:59:50,428 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15396.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 17:00:02,645 INFO [train2.py:809] (1/4) Epoch 4, batch 3450, loss[ctc_loss=0.1775, att_loss=0.2971, loss=0.2732, over 17008.00 frames. utt_duration=688.5 frames, utt_pad_proportion=0.135, over 99.00 utterances.], tot_loss[ctc_loss=0.174, att_loss=0.2881, loss=0.2653, over 3277234.52 frames. utt_duration=1267 frames, utt_pad_proportion=0.04836, over 10361.16 utterances.], batch size: 99, lr: 2.42e-02, grad_scale: 8.0 2023-03-07 17:00:10,264 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.330e+02 3.236e+02 3.892e+02 5.378e+02 1.574e+03, threshold=7.784e+02, percent-clipped=3.0 2023-03-07 17:01:07,805 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=15444.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 17:01:22,737 INFO [train2.py:809] (1/4) Epoch 4, batch 3500, loss[ctc_loss=0.1855, att_loss=0.3009, loss=0.2778, over 16775.00 frames. utt_duration=679.5 frames, utt_pad_proportion=0.1464, over 99.00 utterances.], tot_loss[ctc_loss=0.1755, att_loss=0.2892, loss=0.2664, over 3275440.89 frames. utt_duration=1217 frames, utt_pad_proportion=0.06157, over 10776.17 utterances.], batch size: 99, lr: 2.42e-02, grad_scale: 8.0 2023-03-07 17:02:43,135 INFO [train2.py:809] (1/4) Epoch 4, batch 3550, loss[ctc_loss=0.1858, att_loss=0.3043, loss=0.2806, over 17011.00 frames. utt_duration=1336 frames, utt_pad_proportion=0.008195, over 51.00 utterances.], tot_loss[ctc_loss=0.1743, att_loss=0.2878, loss=0.2651, over 3264785.53 frames. utt_duration=1233 frames, utt_pad_proportion=0.06087, over 10606.25 utterances.], batch size: 51, lr: 2.41e-02, grad_scale: 8.0 2023-03-07 17:02:50,733 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.279e+02 3.812e+02 4.413e+02 5.065e+02 1.151e+03, threshold=8.826e+02, percent-clipped=3.0 2023-03-07 17:03:02,609 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.41 vs. limit=5.0 2023-03-07 17:03:03,862 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.61 vs. limit=5.0 2023-03-07 17:03:05,479 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.17 vs. limit=5.0 2023-03-07 17:04:00,618 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=15551.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 17:04:03,542 INFO [train2.py:809] (1/4) Epoch 4, batch 3600, loss[ctc_loss=0.1796, att_loss=0.2916, loss=0.2692, over 17295.00 frames. utt_duration=1174 frames, utt_pad_proportion=0.02408, over 59.00 utterances.], tot_loss[ctc_loss=0.1739, att_loss=0.288, loss=0.2651, over 3276269.77 frames. utt_duration=1253 frames, utt_pad_proportion=0.05255, over 10474.62 utterances.], batch size: 59, lr: 2.41e-02, grad_scale: 8.0 2023-03-07 17:05:04,331 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-03-07 17:05:24,927 INFO [train2.py:809] (1/4) Epoch 4, batch 3650, loss[ctc_loss=0.1367, att_loss=0.2647, loss=0.2391, over 16160.00 frames. utt_duration=1578 frames, utt_pad_proportion=0.007317, over 41.00 utterances.], tot_loss[ctc_loss=0.173, att_loss=0.2878, loss=0.2648, over 3281141.93 frames. utt_duration=1256 frames, utt_pad_proportion=0.05038, over 10463.21 utterances.], batch size: 41, lr: 2.41e-02, grad_scale: 8.0 2023-03-07 17:05:32,892 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.227e+02 3.822e+02 4.406e+02 5.174e+02 1.169e+03, threshold=8.811e+02, percent-clipped=2.0 2023-03-07 17:05:33,249 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1342, 4.5626, 4.2714, 4.7918, 4.4556, 4.4146, 3.8852, 4.5431], device='cuda:1'), covar=tensor([0.0118, 0.0116, 0.0110, 0.0058, 0.0098, 0.0087, 0.0371, 0.0181], device='cuda:1'), in_proj_covar=tensor([0.0049, 0.0050, 0.0052, 0.0036, 0.0037, 0.0046, 0.0068, 0.0063], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-07 17:06:44,838 INFO [train2.py:809] (1/4) Epoch 4, batch 3700, loss[ctc_loss=0.1509, att_loss=0.2688, loss=0.2452, over 16203.00 frames. utt_duration=1582 frames, utt_pad_proportion=0.005473, over 41.00 utterances.], tot_loss[ctc_loss=0.1738, att_loss=0.2879, loss=0.265, over 3274259.33 frames. utt_duration=1241 frames, utt_pad_proportion=0.05628, over 10565.09 utterances.], batch size: 41, lr: 2.40e-02, grad_scale: 8.0 2023-03-07 17:07:05,248 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7777, 2.6570, 4.9999, 4.0621, 3.4050, 4.6909, 4.6984, 4.8905], device='cuda:1'), covar=tensor([0.0149, 0.1617, 0.0253, 0.1136, 0.2052, 0.0213, 0.0155, 0.0182], device='cuda:1'), in_proj_covar=tensor([0.0127, 0.0236, 0.0121, 0.0294, 0.0323, 0.0181, 0.0107, 0.0130], device='cuda:1'), out_proj_covar=tensor([1.0929e-04, 1.8353e-04, 1.0334e-04, 2.3310e-04, 2.5416e-04, 1.5061e-04, 9.1205e-05, 1.1145e-04], device='cuda:1') 2023-03-07 17:08:05,616 INFO [train2.py:809] (1/4) Epoch 4, batch 3750, loss[ctc_loss=0.1631, att_loss=0.2914, loss=0.2657, over 16773.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.006067, over 48.00 utterances.], tot_loss[ctc_loss=0.1735, att_loss=0.2877, loss=0.2649, over 3262295.70 frames. utt_duration=1218 frames, utt_pad_proportion=0.06379, over 10727.22 utterances.], batch size: 48, lr: 2.40e-02, grad_scale: 8.0 2023-03-07 17:08:13,130 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.429e+02 3.564e+02 4.340e+02 5.643e+02 1.443e+03, threshold=8.680e+02, percent-clipped=5.0 2023-03-07 17:08:53,345 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9939, 2.2140, 2.7629, 4.4478, 4.3588, 4.4857, 2.6868, 1.6426], device='cuda:1'), covar=tensor([0.0382, 0.2795, 0.1599, 0.0379, 0.0296, 0.0151, 0.2000, 0.3142], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0188, 0.0182, 0.0134, 0.0125, 0.0107, 0.0184, 0.0169], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-07 17:09:25,256 INFO [train2.py:809] (1/4) Epoch 4, batch 3800, loss[ctc_loss=0.1735, att_loss=0.2986, loss=0.2736, over 17396.00 frames. utt_duration=1222 frames, utt_pad_proportion=0.0143, over 57.00 utterances.], tot_loss[ctc_loss=0.1737, att_loss=0.288, loss=0.2652, over 3266837.18 frames. utt_duration=1237 frames, utt_pad_proportion=0.05888, over 10579.06 utterances.], batch size: 57, lr: 2.40e-02, grad_scale: 8.0 2023-03-07 17:09:41,335 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7401, 4.7064, 4.7006, 4.9770, 4.9778, 4.5811, 4.5306, 2.0645], device='cuda:1'), covar=tensor([0.0229, 0.0231, 0.0165, 0.0073, 0.0744, 0.0236, 0.0271, 0.3162], device='cuda:1'), in_proj_covar=tensor([0.0130, 0.0117, 0.0113, 0.0119, 0.0267, 0.0130, 0.0110, 0.0248], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-07 17:10:46,094 INFO [train2.py:809] (1/4) Epoch 4, batch 3850, loss[ctc_loss=0.2199, att_loss=0.3268, loss=0.3055, over 17044.00 frames. utt_duration=1288 frames, utt_pad_proportion=0.009303, over 53.00 utterances.], tot_loss[ctc_loss=0.1733, att_loss=0.2876, loss=0.2647, over 3263300.05 frames. utt_duration=1231 frames, utt_pad_proportion=0.06159, over 10616.91 utterances.], batch size: 53, lr: 2.39e-02, grad_scale: 8.0 2023-03-07 17:10:53,737 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.413e+02 3.550e+02 4.549e+02 5.409e+02 1.209e+03, threshold=9.099e+02, percent-clipped=5.0 2023-03-07 17:11:57,628 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-03-07 17:12:00,193 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15851.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 17:12:02,975 INFO [train2.py:809] (1/4) Epoch 4, batch 3900, loss[ctc_loss=0.158, att_loss=0.2839, loss=0.2587, over 16307.00 frames. utt_duration=1451 frames, utt_pad_proportion=0.007661, over 45.00 utterances.], tot_loss[ctc_loss=0.1741, att_loss=0.2878, loss=0.2651, over 3261903.83 frames. utt_duration=1219 frames, utt_pad_proportion=0.06511, over 10718.10 utterances.], batch size: 45, lr: 2.39e-02, grad_scale: 8.0 2023-03-07 17:12:32,252 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9062, 5.2908, 5.1948, 5.1749, 5.4262, 5.2063, 5.1198, 4.8455], device='cuda:1'), covar=tensor([0.1084, 0.0431, 0.0209, 0.0577, 0.0226, 0.0283, 0.0208, 0.0301], device='cuda:1'), in_proj_covar=tensor([0.0348, 0.0206, 0.0146, 0.0189, 0.0228, 0.0253, 0.0195, 0.0219], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-07 17:12:48,168 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.2905, 1.3224, 1.8382, 2.2609, 2.2025, 1.4916, 1.8015, 1.7184], device='cuda:1'), covar=tensor([0.0518, 0.1502, 0.1142, 0.0617, 0.0272, 0.0903, 0.1106, 0.0862], device='cuda:1'), in_proj_covar=tensor([0.0086, 0.0088, 0.0088, 0.0075, 0.0071, 0.0075, 0.0088, 0.0081], device='cuda:1'), out_proj_covar=tensor([3.8168e-05, 4.5793e-05, 4.6804e-05, 4.2733e-05, 3.5657e-05, 3.9710e-05, 4.3989e-05, 4.1389e-05], device='cuda:1') 2023-03-07 17:12:49,842 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=15883.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 17:13:13,704 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=15899.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 17:13:19,703 INFO [train2.py:809] (1/4) Epoch 4, batch 3950, loss[ctc_loss=0.167, att_loss=0.2821, loss=0.2591, over 15950.00 frames. utt_duration=1558 frames, utt_pad_proportion=0.007089, over 41.00 utterances.], tot_loss[ctc_loss=0.1749, att_loss=0.2882, loss=0.2655, over 3260511.56 frames. utt_duration=1203 frames, utt_pad_proportion=0.06987, over 10855.06 utterances.], batch size: 41, lr: 2.39e-02, grad_scale: 8.0 2023-03-07 17:13:26,641 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5314, 4.3710, 4.4094, 3.0915, 4.2644, 4.0250, 4.0037, 2.5870], device='cuda:1'), covar=tensor([0.0067, 0.0094, 0.0188, 0.0736, 0.0088, 0.0170, 0.0238, 0.1342], device='cuda:1'), in_proj_covar=tensor([0.0049, 0.0057, 0.0046, 0.0092, 0.0053, 0.0066, 0.0075, 0.0099], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-03-07 17:13:27,824 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.270e+02 3.629e+02 4.356e+02 5.723e+02 1.612e+03, threshold=8.711e+02, percent-clipped=3.0 2023-03-07 17:14:39,238 INFO [train2.py:809] (1/4) Epoch 5, batch 0, loss[ctc_loss=0.195, att_loss=0.3135, loss=0.2898, over 17046.00 frames. utt_duration=1313 frames, utt_pad_proportion=0.008352, over 52.00 utterances.], tot_loss[ctc_loss=0.195, att_loss=0.3135, loss=0.2898, over 17046.00 frames. utt_duration=1313 frames, utt_pad_proportion=0.008352, over 52.00 utterances.], batch size: 52, lr: 2.22e-02, grad_scale: 8.0 2023-03-07 17:14:39,239 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-07 17:14:51,958 INFO [train2.py:843] (1/4) Epoch 5, validation: ctc_loss=0.08303, att_loss=0.2543, loss=0.22, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-07 17:14:51,959 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-07 17:15:06,132 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=15944.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 17:16:13,042 INFO [train2.py:809] (1/4) Epoch 5, batch 50, loss[ctc_loss=0.1684, att_loss=0.298, loss=0.2721, over 16593.00 frames. utt_duration=678.8 frames, utt_pad_proportion=0.1494, over 98.00 utterances.], tot_loss[ctc_loss=0.1667, att_loss=0.2847, loss=0.2611, over 740449.25 frames. utt_duration=1195 frames, utt_pad_proportion=0.0628, over 2481.75 utterances.], batch size: 98, lr: 2.22e-02, grad_scale: 8.0 2023-03-07 17:16:41,408 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16001.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 17:16:51,838 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.119e+02 3.417e+02 4.421e+02 5.634e+02 1.181e+03, threshold=8.842e+02, percent-clipped=4.0 2023-03-07 17:17:36,497 INFO [train2.py:809] (1/4) Epoch 5, batch 100, loss[ctc_loss=0.1809, att_loss=0.2924, loss=0.2701, over 16879.00 frames. utt_duration=1379 frames, utt_pad_proportion=0.00692, over 49.00 utterances.], tot_loss[ctc_loss=0.1668, att_loss=0.2831, loss=0.2598, over 1298332.85 frames. utt_duration=1226 frames, utt_pad_proportion=0.05916, over 4239.58 utterances.], batch size: 49, lr: 2.21e-02, grad_scale: 8.0 2023-03-07 17:18:18,127 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16062.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 17:18:55,094 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16085.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 17:18:56,279 INFO [train2.py:809] (1/4) Epoch 5, batch 150, loss[ctc_loss=0.1209, att_loss=0.2445, loss=0.2198, over 15501.00 frames. utt_duration=1724 frames, utt_pad_proportion=0.008626, over 36.00 utterances.], tot_loss[ctc_loss=0.1658, att_loss=0.2815, loss=0.2583, over 1724192.28 frames. utt_duration=1209 frames, utt_pad_proportion=0.06909, over 5711.89 utterances.], batch size: 36, lr: 2.21e-02, grad_scale: 8.0 2023-03-07 17:19:31,235 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.071e+02 3.158e+02 4.095e+02 5.546e+02 1.058e+03, threshold=8.189e+02, percent-clipped=5.0 2023-03-07 17:19:48,712 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.6105, 2.4383, 2.9444, 4.2329, 4.2284, 4.1468, 2.6341, 1.9012], device='cuda:1'), covar=tensor([0.0542, 0.2393, 0.1479, 0.0482, 0.0314, 0.0238, 0.1715, 0.2798], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0187, 0.0183, 0.0136, 0.0127, 0.0111, 0.0182, 0.0171], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-07 17:20:16,134 INFO [train2.py:809] (1/4) Epoch 5, batch 200, loss[ctc_loss=0.1833, att_loss=0.3015, loss=0.2779, over 17324.00 frames. utt_duration=1176 frames, utt_pad_proportion=0.02159, over 59.00 utterances.], tot_loss[ctc_loss=0.1653, att_loss=0.2806, loss=0.2575, over 2058728.83 frames. utt_duration=1234 frames, utt_pad_proportion=0.06577, over 6683.48 utterances.], batch size: 59, lr: 2.21e-02, grad_scale: 8.0 2023-03-07 17:20:32,635 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16146.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 17:20:37,372 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16149.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 17:20:37,397 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2471, 0.8575, 1.8161, 2.4875, 1.8151, 1.4021, 2.0082, 1.8056], device='cuda:1'), covar=tensor([0.0221, 0.2406, 0.1904, 0.0550, 0.0364, 0.1424, 0.1121, 0.0859], device='cuda:1'), in_proj_covar=tensor([0.0090, 0.0093, 0.0094, 0.0075, 0.0078, 0.0083, 0.0094, 0.0085], device='cuda:1'), out_proj_covar=tensor([3.9294e-05, 4.9632e-05, 5.0420e-05, 4.2860e-05, 3.8898e-05, 4.4123e-05, 4.7022e-05, 4.3449e-05], device='cuda:1') 2023-03-07 17:21:27,939 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.1237, 1.0388, 1.6369, 2.0684, 1.6245, 1.6345, 1.1654, 1.6429], device='cuda:1'), covar=tensor([0.0351, 0.2239, 0.1280, 0.0850, 0.0457, 0.1002, 0.2296, 0.0986], device='cuda:1'), in_proj_covar=tensor([0.0090, 0.0094, 0.0095, 0.0076, 0.0079, 0.0083, 0.0096, 0.0086], device='cuda:1'), out_proj_covar=tensor([3.9499e-05, 4.9869e-05, 5.0989e-05, 4.3373e-05, 3.9354e-05, 4.4220e-05, 4.7933e-05, 4.3762e-05], device='cuda:1') 2023-03-07 17:21:34,248 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3583, 3.0047, 3.7378, 2.5418, 3.4645, 4.4872, 4.3149, 3.2188], device='cuda:1'), covar=tensor([0.0427, 0.1452, 0.0782, 0.1573, 0.0959, 0.0421, 0.0501, 0.1377], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0193, 0.0184, 0.0182, 0.0201, 0.0177, 0.0148, 0.0192], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-07 17:21:35,467 INFO [train2.py:809] (1/4) Epoch 5, batch 250, loss[ctc_loss=0.1609, att_loss=0.2819, loss=0.2577, over 16612.00 frames. utt_duration=1415 frames, utt_pad_proportion=0.006246, over 47.00 utterances.], tot_loss[ctc_loss=0.1624, att_loss=0.2795, loss=0.2561, over 2330082.29 frames. utt_duration=1277 frames, utt_pad_proportion=0.05309, over 7309.92 utterances.], batch size: 47, lr: 2.20e-02, grad_scale: 8.0 2023-03-07 17:22:09,329 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6426, 3.1599, 4.9108, 4.1225, 3.2166, 4.6186, 4.7663, 4.8549], device='cuda:1'), covar=tensor([0.0236, 0.1576, 0.0301, 0.1163, 0.2334, 0.0250, 0.0182, 0.0244], device='cuda:1'), in_proj_covar=tensor([0.0131, 0.0241, 0.0122, 0.0302, 0.0326, 0.0183, 0.0111, 0.0133], device='cuda:1'), out_proj_covar=tensor([1.1192e-04, 1.8876e-04, 1.0467e-04, 2.4003e-04, 2.5749e-04, 1.5367e-04, 9.4959e-05, 1.1350e-04], device='cuda:1') 2023-03-07 17:22:12,003 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.315e+02 3.409e+02 3.964e+02 4.827e+02 7.562e+02, threshold=7.928e+02, percent-clipped=0.0 2023-03-07 17:22:13,032 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-03-07 17:22:15,465 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16210.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 17:22:57,400 INFO [train2.py:809] (1/4) Epoch 5, batch 300, loss[ctc_loss=0.1459, att_loss=0.2664, loss=0.2423, over 16001.00 frames. utt_duration=1601 frames, utt_pad_proportion=0.007176, over 40.00 utterances.], tot_loss[ctc_loss=0.1626, att_loss=0.2807, loss=0.2571, over 2540011.98 frames. utt_duration=1283 frames, utt_pad_proportion=0.04927, over 7929.75 utterances.], batch size: 40, lr: 2.20e-02, grad_scale: 8.0 2023-03-07 17:23:02,213 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16239.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 17:23:08,971 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-03-07 17:24:18,573 INFO [train2.py:809] (1/4) Epoch 5, batch 350, loss[ctc_loss=0.1593, att_loss=0.2782, loss=0.2544, over 15969.00 frames. utt_duration=1559 frames, utt_pad_proportion=0.006187, over 41.00 utterances.], tot_loss[ctc_loss=0.163, att_loss=0.2815, loss=0.2578, over 2703177.47 frames. utt_duration=1271 frames, utt_pad_proportion=0.05117, over 8515.37 utterances.], batch size: 41, lr: 2.20e-02, grad_scale: 8.0 2023-03-07 17:24:56,571 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.150e+02 3.423e+02 4.365e+02 5.343e+02 1.409e+03, threshold=8.730e+02, percent-clipped=9.0 2023-03-07 17:25:33,737 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.4768, 1.3565, 1.9973, 2.7512, 2.3197, 1.4756, 1.0702, 1.7762], device='cuda:1'), covar=tensor([0.0828, 0.1956, 0.1403, 0.0453, 0.0390, 0.1254, 0.2130, 0.1004], device='cuda:1'), in_proj_covar=tensor([0.0090, 0.0092, 0.0092, 0.0073, 0.0076, 0.0083, 0.0096, 0.0085], device='cuda:1'), out_proj_covar=tensor([3.9245e-05, 4.9033e-05, 4.9210e-05, 4.1434e-05, 3.7539e-05, 4.3926e-05, 4.8063e-05, 4.3109e-05], device='cuda:1') 2023-03-07 17:25:39,504 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.98 vs. limit=5.0 2023-03-07 17:25:41,304 INFO [train2.py:809] (1/4) Epoch 5, batch 400, loss[ctc_loss=0.2315, att_loss=0.3219, loss=0.3038, over 14480.00 frames. utt_duration=398.3 frames, utt_pad_proportion=0.3049, over 146.00 utterances.], tot_loss[ctc_loss=0.1647, att_loss=0.2822, loss=0.2587, over 2826341.06 frames. utt_duration=1250 frames, utt_pad_proportion=0.05724, over 9056.58 utterances.], batch size: 146, lr: 2.20e-02, grad_scale: 8.0 2023-03-07 17:25:57,973 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8299, 5.1528, 4.7979, 5.2869, 4.7108, 5.0497, 5.3859, 5.1798], device='cuda:1'), covar=tensor([0.0294, 0.0231, 0.0512, 0.0115, 0.0299, 0.0128, 0.0165, 0.0125], device='cuda:1'), in_proj_covar=tensor([0.0190, 0.0160, 0.0207, 0.0134, 0.0178, 0.0130, 0.0155, 0.0150], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-03-07 17:26:15,244 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16357.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 17:26:37,579 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16371.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 17:27:02,331 INFO [train2.py:809] (1/4) Epoch 5, batch 450, loss[ctc_loss=0.1696, att_loss=0.29, loss=0.266, over 16473.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.006433, over 46.00 utterances.], tot_loss[ctc_loss=0.1657, att_loss=0.2831, loss=0.2597, over 2931388.97 frames. utt_duration=1246 frames, utt_pad_proportion=0.05649, over 9418.46 utterances.], batch size: 46, lr: 2.19e-02, grad_scale: 8.0 2023-03-07 17:27:40,287 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.521e+02 3.844e+02 4.538e+02 5.770e+02 9.247e+02, threshold=9.077e+02, percent-clipped=1.0 2023-03-07 17:27:43,677 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7677, 5.2033, 4.7029, 5.3797, 4.6903, 5.1591, 5.5028, 5.2344], device='cuda:1'), covar=tensor([0.0346, 0.0235, 0.0669, 0.0140, 0.0382, 0.0116, 0.0175, 0.0153], device='cuda:1'), in_proj_covar=tensor([0.0187, 0.0159, 0.0208, 0.0134, 0.0177, 0.0128, 0.0153, 0.0149], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0005, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:1') 2023-03-07 17:28:17,552 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16432.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 17:28:24,054 INFO [train2.py:809] (1/4) Epoch 5, batch 500, loss[ctc_loss=0.1518, att_loss=0.2674, loss=0.2443, over 16112.00 frames. utt_duration=1536 frames, utt_pad_proportion=0.006622, over 42.00 utterances.], tot_loss[ctc_loss=0.1654, att_loss=0.2827, loss=0.2592, over 3009574.94 frames. utt_duration=1264 frames, utt_pad_proportion=0.05022, over 9538.65 utterances.], batch size: 42, lr: 2.19e-02, grad_scale: 8.0 2023-03-07 17:28:32,526 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16441.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 17:29:45,241 INFO [train2.py:809] (1/4) Epoch 5, batch 550, loss[ctc_loss=0.1631, att_loss=0.2804, loss=0.2569, over 16286.00 frames. utt_duration=1516 frames, utt_pad_proportion=0.00693, over 43.00 utterances.], tot_loss[ctc_loss=0.1656, att_loss=0.2823, loss=0.2589, over 3070820.07 frames. utt_duration=1271 frames, utt_pad_proportion=0.04613, over 9673.16 utterances.], batch size: 43, lr: 2.19e-02, grad_scale: 8.0 2023-03-07 17:29:49,838 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-03-07 17:29:52,598 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.12 vs. limit=5.0 2023-03-07 17:30:16,785 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16505.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 17:30:22,784 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.254e+02 3.300e+02 4.154e+02 5.468e+02 1.131e+03, threshold=8.309e+02, percent-clipped=7.0 2023-03-07 17:30:47,443 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.1022, 1.0156, 1.4104, 2.3683, 2.1258, 1.7127, 1.5447, 2.3173], device='cuda:1'), covar=tensor([0.0950, 0.1918, 0.1675, 0.0732, 0.0481, 0.1243, 0.1167, 0.0575], device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0095, 0.0100, 0.0080, 0.0082, 0.0086, 0.0099, 0.0086], device='cuda:1'), out_proj_covar=tensor([4.1542e-05, 5.1195e-05, 5.1409e-05, 4.3396e-05, 3.8807e-05, 4.6205e-05, 4.8713e-05, 4.4190e-05], device='cuda:1') 2023-03-07 17:31:06,240 INFO [train2.py:809] (1/4) Epoch 5, batch 600, loss[ctc_loss=0.2146, att_loss=0.3157, loss=0.2955, over 17302.00 frames. utt_duration=1100 frames, utt_pad_proportion=0.03849, over 63.00 utterances.], tot_loss[ctc_loss=0.1663, att_loss=0.2823, loss=0.2591, over 3109338.71 frames. utt_duration=1253 frames, utt_pad_proportion=0.05288, over 9934.62 utterances.], batch size: 63, lr: 2.18e-02, grad_scale: 8.0 2023-03-07 17:31:12,937 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16539.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 17:32:22,130 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.07 vs. limit=2.0 2023-03-07 17:32:27,509 INFO [train2.py:809] (1/4) Epoch 5, batch 650, loss[ctc_loss=0.1795, att_loss=0.297, loss=0.2735, over 17149.00 frames. utt_duration=1226 frames, utt_pad_proportion=0.01349, over 56.00 utterances.], tot_loss[ctc_loss=0.165, att_loss=0.2816, loss=0.2583, over 3144939.27 frames. utt_duration=1264 frames, utt_pad_proportion=0.05077, over 9967.61 utterances.], batch size: 56, lr: 2.18e-02, grad_scale: 8.0 2023-03-07 17:32:29,905 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=16587.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 17:33:07,230 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.993e+02 3.372e+02 4.098e+02 5.133e+02 8.733e+02, threshold=8.196e+02, percent-clipped=2.0 2023-03-07 17:33:14,079 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.7622, 1.9466, 2.8649, 4.3627, 4.3630, 4.2633, 2.8809, 1.9760], device='cuda:1'), covar=tensor([0.0543, 0.3196, 0.1509, 0.0528, 0.0329, 0.0203, 0.1540, 0.2884], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0184, 0.0182, 0.0137, 0.0123, 0.0110, 0.0177, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-07 17:33:51,405 INFO [train2.py:809] (1/4) Epoch 5, batch 700, loss[ctc_loss=0.1782, att_loss=0.2941, loss=0.2709, over 17369.00 frames. utt_duration=1179 frames, utt_pad_proportion=0.02012, over 59.00 utterances.], tot_loss[ctc_loss=0.1642, att_loss=0.2811, loss=0.2577, over 3167964.35 frames. utt_duration=1243 frames, utt_pad_proportion=0.0574, over 10208.13 utterances.], batch size: 59, lr: 2.18e-02, grad_scale: 8.0 2023-03-07 17:34:26,840 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16657.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 17:35:14,450 INFO [train2.py:809] (1/4) Epoch 5, batch 750, loss[ctc_loss=0.1436, att_loss=0.2557, loss=0.2333, over 15888.00 frames. utt_duration=1631 frames, utt_pad_proportion=0.008402, over 39.00 utterances.], tot_loss[ctc_loss=0.1628, att_loss=0.2801, loss=0.2566, over 3195217.21 frames. utt_duration=1262 frames, utt_pad_proportion=0.05067, over 10135.49 utterances.], batch size: 39, lr: 2.17e-02, grad_scale: 8.0 2023-03-07 17:35:45,790 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=16705.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 17:35:51,950 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.143e+02 3.355e+02 3.924e+02 5.306e+02 2.112e+03, threshold=7.847e+02, percent-clipped=6.0 2023-03-07 17:36:21,537 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16727.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 17:36:36,905 INFO [train2.py:809] (1/4) Epoch 5, batch 800, loss[ctc_loss=0.175, att_loss=0.2886, loss=0.2659, over 16275.00 frames. utt_duration=1515 frames, utt_pad_proportion=0.007163, over 43.00 utterances.], tot_loss[ctc_loss=0.163, att_loss=0.2806, loss=0.2571, over 3221623.77 frames. utt_duration=1274 frames, utt_pad_proportion=0.04551, over 10130.31 utterances.], batch size: 43, lr: 2.17e-02, grad_scale: 8.0 2023-03-07 17:36:45,913 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16741.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 17:36:50,733 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4953, 2.6839, 3.4336, 2.6417, 3.5370, 4.5830, 4.2478, 3.2713], device='cuda:1'), covar=tensor([0.0415, 0.1911, 0.1372, 0.1782, 0.1150, 0.0464, 0.0569, 0.1437], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0195, 0.0192, 0.0187, 0.0202, 0.0183, 0.0155, 0.0193], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-07 17:36:55,832 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.95 vs. limit=2.0 2023-03-07 17:37:17,367 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-03-07 17:37:34,726 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-03-07 17:37:59,759 INFO [train2.py:809] (1/4) Epoch 5, batch 850, loss[ctc_loss=0.1318, att_loss=0.2712, loss=0.2433, over 16465.00 frames. utt_duration=1433 frames, utt_pad_proportion=0.007573, over 46.00 utterances.], tot_loss[ctc_loss=0.1636, att_loss=0.2811, loss=0.2576, over 3238199.71 frames. utt_duration=1263 frames, utt_pad_proportion=0.04665, over 10269.40 utterances.], batch size: 46, lr: 2.17e-02, grad_scale: 8.0 2023-03-07 17:38:04,491 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=16789.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 17:38:27,992 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.73 vs. limit=2.0 2023-03-07 17:38:30,515 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16805.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 17:38:36,313 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.012e+02 3.471e+02 4.444e+02 5.932e+02 1.311e+03, threshold=8.889e+02, percent-clipped=13.0 2023-03-07 17:39:10,198 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.39 vs. limit=5.0 2023-03-07 17:39:22,342 INFO [train2.py:809] (1/4) Epoch 5, batch 900, loss[ctc_loss=0.1844, att_loss=0.2939, loss=0.272, over 16409.00 frames. utt_duration=1493 frames, utt_pad_proportion=0.006411, over 44.00 utterances.], tot_loss[ctc_loss=0.1635, att_loss=0.2809, loss=0.2574, over 3247553.54 frames. utt_duration=1272 frames, utt_pad_proportion=0.04533, over 10227.39 utterances.], batch size: 44, lr: 2.16e-02, grad_scale: 8.0 2023-03-07 17:39:48,824 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=16853.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 17:40:13,258 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16868.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 17:40:42,548 INFO [train2.py:809] (1/4) Epoch 5, batch 950, loss[ctc_loss=0.2018, att_loss=0.311, loss=0.2891, over 17074.00 frames. utt_duration=691.3 frames, utt_pad_proportion=0.1337, over 99.00 utterances.], tot_loss[ctc_loss=0.1655, att_loss=0.2819, loss=0.2586, over 3252797.19 frames. utt_duration=1240 frames, utt_pad_proportion=0.05326, over 10507.28 utterances.], batch size: 99, lr: 2.16e-02, grad_scale: 8.0 2023-03-07 17:41:14,649 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16906.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 17:41:18,948 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.139e+02 3.605e+02 4.438e+02 5.826e+02 1.823e+03, threshold=8.877e+02, percent-clipped=6.0 2023-03-07 17:41:28,846 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9140, 5.3155, 5.1195, 5.0952, 5.3698, 5.3090, 5.0553, 4.8845], device='cuda:1'), covar=tensor([0.1002, 0.0359, 0.0237, 0.0454, 0.0237, 0.0227, 0.0231, 0.0253], device='cuda:1'), in_proj_covar=tensor([0.0372, 0.0227, 0.0157, 0.0199, 0.0242, 0.0264, 0.0206, 0.0237], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-07 17:41:52,643 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16929.0, num_to_drop=1, layers_to_drop={2} 2023-03-07 17:42:04,882 INFO [train2.py:809] (1/4) Epoch 5, batch 1000, loss[ctc_loss=0.153, att_loss=0.2898, loss=0.2625, over 17032.00 frames. utt_duration=1311 frames, utt_pad_proportion=0.009543, over 52.00 utterances.], tot_loss[ctc_loss=0.1643, att_loss=0.2816, loss=0.2581, over 3262426.35 frames. utt_duration=1267 frames, utt_pad_proportion=0.04598, over 10314.93 utterances.], batch size: 52, lr: 2.16e-02, grad_scale: 8.0 2023-03-07 17:42:54,663 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16967.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 17:43:14,031 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.75 vs. limit=2.0 2023-03-07 17:43:25,862 INFO [train2.py:809] (1/4) Epoch 5, batch 1050, loss[ctc_loss=0.1923, att_loss=0.3124, loss=0.2883, over 17366.00 frames. utt_duration=1179 frames, utt_pad_proportion=0.02095, over 59.00 utterances.], tot_loss[ctc_loss=0.1656, att_loss=0.2824, loss=0.2591, over 3264538.05 frames. utt_duration=1267 frames, utt_pad_proportion=0.04612, over 10317.00 utterances.], batch size: 59, lr: 2.16e-02, grad_scale: 8.0 2023-03-07 17:44:02,517 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.326e+02 3.274e+02 4.192e+02 5.406e+02 1.067e+03, threshold=8.384e+02, percent-clipped=5.0 2023-03-07 17:44:32,615 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17027.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 17:44:47,438 INFO [train2.py:809] (1/4) Epoch 5, batch 1100, loss[ctc_loss=0.1489, att_loss=0.2851, loss=0.2579, over 17044.00 frames. utt_duration=1288 frames, utt_pad_proportion=0.01028, over 53.00 utterances.], tot_loss[ctc_loss=0.1664, att_loss=0.283, loss=0.2597, over 3265929.06 frames. utt_duration=1253 frames, utt_pad_proportion=0.05108, over 10438.34 utterances.], batch size: 53, lr: 2.15e-02, grad_scale: 8.0 2023-03-07 17:44:59,897 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-03-07 17:45:29,746 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.67 vs. limit=2.0 2023-03-07 17:45:43,540 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.08 vs. limit=5.0 2023-03-07 17:45:48,458 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-03-07 17:45:48,878 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9924, 4.4868, 4.2733, 4.5953, 4.6158, 4.2340, 3.7003, 4.3809], device='cuda:1'), covar=tensor([0.0106, 0.0144, 0.0125, 0.0090, 0.0100, 0.0133, 0.0428, 0.0252], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0051, 0.0055, 0.0038, 0.0038, 0.0048, 0.0069, 0.0067], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-07 17:45:50,936 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=17075.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 17:46:09,754 INFO [train2.py:809] (1/4) Epoch 5, batch 1150, loss[ctc_loss=0.1513, att_loss=0.269, loss=0.2454, over 15899.00 frames. utt_duration=1632 frames, utt_pad_proportion=0.007502, over 39.00 utterances.], tot_loss[ctc_loss=0.1665, att_loss=0.283, loss=0.2597, over 3268783.12 frames. utt_duration=1251 frames, utt_pad_proportion=0.05072, over 10467.93 utterances.], batch size: 39, lr: 2.15e-02, grad_scale: 8.0 2023-03-07 17:46:40,067 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.82 vs. limit=2.0 2023-03-07 17:46:45,377 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.659e+02 3.410e+02 4.345e+02 5.323e+02 1.056e+03, threshold=8.690e+02, percent-clipped=4.0 2023-03-07 17:47:31,220 INFO [train2.py:809] (1/4) Epoch 5, batch 1200, loss[ctc_loss=0.1348, att_loss=0.256, loss=0.2317, over 14565.00 frames. utt_duration=1822 frames, utt_pad_proportion=0.04055, over 32.00 utterances.], tot_loss[ctc_loss=0.1665, att_loss=0.283, loss=0.2597, over 3271052.18 frames. utt_duration=1244 frames, utt_pad_proportion=0.05361, over 10526.70 utterances.], batch size: 32, lr: 2.15e-02, grad_scale: 8.0 2023-03-07 17:47:55,975 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-03-07 17:48:20,375 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 2023-03-07 17:48:52,438 INFO [train2.py:809] (1/4) Epoch 5, batch 1250, loss[ctc_loss=0.183, att_loss=0.2668, loss=0.2501, over 15380.00 frames. utt_duration=1759 frames, utt_pad_proportion=0.01059, over 35.00 utterances.], tot_loss[ctc_loss=0.168, att_loss=0.2835, loss=0.2604, over 3253435.53 frames. utt_duration=1200 frames, utt_pad_proportion=0.07066, over 10861.78 utterances.], batch size: 35, lr: 2.14e-02, grad_scale: 8.0 2023-03-07 17:48:57,266 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17189.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 17:49:29,491 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.117e+02 3.900e+02 4.861e+02 6.296e+02 1.076e+03, threshold=9.721e+02, percent-clipped=3.0 2023-03-07 17:49:55,851 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17224.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 17:50:14,606 INFO [train2.py:809] (1/4) Epoch 5, batch 1300, loss[ctc_loss=0.1759, att_loss=0.269, loss=0.2504, over 15509.00 frames. utt_duration=1725 frames, utt_pad_proportion=0.008258, over 36.00 utterances.], tot_loss[ctc_loss=0.1667, att_loss=0.2825, loss=0.2593, over 3261362.78 frames. utt_duration=1214 frames, utt_pad_proportion=0.06503, over 10756.85 utterances.], batch size: 36, lr: 2.14e-02, grad_scale: 8.0 2023-03-07 17:50:36,811 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17250.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 17:50:56,131 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17262.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 17:51:15,802 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8752, 4.4642, 4.5456, 4.6667, 5.1657, 4.9258, 4.4891, 2.0579], device='cuda:1'), covar=tensor([0.0215, 0.0543, 0.0296, 0.0259, 0.0735, 0.0202, 0.0400, 0.3205], device='cuda:1'), in_proj_covar=tensor([0.0130, 0.0123, 0.0115, 0.0126, 0.0276, 0.0132, 0.0113, 0.0248], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-07 17:51:27,913 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2489, 4.9646, 4.5395, 4.7978, 4.7584, 4.5615, 4.0655, 4.6288], device='cuda:1'), covar=tensor([0.0103, 0.0108, 0.0097, 0.0079, 0.0109, 0.0085, 0.0360, 0.0200], device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0052, 0.0056, 0.0039, 0.0039, 0.0049, 0.0071, 0.0068], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-07 17:51:35,792 INFO [train2.py:809] (1/4) Epoch 5, batch 1350, loss[ctc_loss=0.2934, att_loss=0.3558, loss=0.3433, over 14079.00 frames. utt_duration=384.7 frames, utt_pad_proportion=0.3262, over 147.00 utterances.], tot_loss[ctc_loss=0.1668, att_loss=0.2832, loss=0.2599, over 3267931.28 frames. utt_duration=1200 frames, utt_pad_proportion=0.06719, over 10908.78 utterances.], batch size: 147, lr: 2.14e-02, grad_scale: 8.0 2023-03-07 17:51:36,260 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6607, 4.3071, 4.3642, 4.5778, 4.9279, 4.7035, 4.3197, 1.8441], device='cuda:1'), covar=tensor([0.0235, 0.0365, 0.0275, 0.0099, 0.0998, 0.0222, 0.0336, 0.3431], device='cuda:1'), in_proj_covar=tensor([0.0131, 0.0124, 0.0115, 0.0127, 0.0279, 0.0133, 0.0114, 0.0250], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-07 17:52:12,509 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.675e+02 3.130e+02 4.012e+02 5.410e+02 1.134e+03, threshold=8.024e+02, percent-clipped=3.0 2023-03-07 17:52:57,027 INFO [train2.py:809] (1/4) Epoch 5, batch 1400, loss[ctc_loss=0.162, att_loss=0.2839, loss=0.2595, over 17330.00 frames. utt_duration=1176 frames, utt_pad_proportion=0.02215, over 59.00 utterances.], tot_loss[ctc_loss=0.1649, att_loss=0.2829, loss=0.2593, over 3276600.33 frames. utt_duration=1224 frames, utt_pad_proportion=0.05967, over 10716.88 utterances.], batch size: 59, lr: 2.14e-02, grad_scale: 8.0 2023-03-07 17:53:38,721 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17362.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 17:53:47,348 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8246, 4.6396, 4.5792, 4.8497, 4.9942, 4.9172, 4.5917, 2.1190], device='cuda:1'), covar=tensor([0.0210, 0.0251, 0.0237, 0.0108, 0.1474, 0.0208, 0.0295, 0.3470], device='cuda:1'), in_proj_covar=tensor([0.0133, 0.0122, 0.0115, 0.0126, 0.0281, 0.0132, 0.0116, 0.0247], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-07 17:54:13,816 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17383.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 17:54:18,019 INFO [train2.py:809] (1/4) Epoch 5, batch 1450, loss[ctc_loss=0.1662, att_loss=0.2938, loss=0.2683, over 16874.00 frames. utt_duration=1379 frames, utt_pad_proportion=0.008046, over 49.00 utterances.], tot_loss[ctc_loss=0.1635, att_loss=0.2819, loss=0.2582, over 3275508.39 frames. utt_duration=1234 frames, utt_pad_proportion=0.05727, over 10631.07 utterances.], batch size: 49, lr: 2.13e-02, grad_scale: 8.0 2023-03-07 17:54:55,330 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.883e+02 3.556e+02 4.184e+02 5.156e+02 1.348e+03, threshold=8.369e+02, percent-clipped=7.0 2023-03-07 17:55:19,591 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17423.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 17:55:39,519 INFO [train2.py:809] (1/4) Epoch 5, batch 1500, loss[ctc_loss=0.1729, att_loss=0.2719, loss=0.2521, over 16004.00 frames. utt_duration=1602 frames, utt_pad_proportion=0.007667, over 40.00 utterances.], tot_loss[ctc_loss=0.164, att_loss=0.2828, loss=0.259, over 3279900.19 frames. utt_duration=1243 frames, utt_pad_proportion=0.05482, over 10567.57 utterances.], batch size: 40, lr: 2.13e-02, grad_scale: 8.0 2023-03-07 17:55:52,459 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17444.0, num_to_drop=1, layers_to_drop={2} 2023-03-07 17:56:59,734 INFO [train2.py:809] (1/4) Epoch 5, batch 1550, loss[ctc_loss=0.1619, att_loss=0.2804, loss=0.2567, over 16287.00 frames. utt_duration=1516 frames, utt_pad_proportion=0.00693, over 43.00 utterances.], tot_loss[ctc_loss=0.1643, att_loss=0.2823, loss=0.2587, over 3280124.11 frames. utt_duration=1247 frames, utt_pad_proportion=0.05406, over 10531.52 utterances.], batch size: 43, lr: 2.13e-02, grad_scale: 8.0 2023-03-07 17:57:36,008 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.215e+02 3.523e+02 4.297e+02 5.219e+02 1.408e+03, threshold=8.593e+02, percent-clipped=2.0 2023-03-07 17:58:01,416 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17524.0, num_to_drop=1, layers_to_drop={2} 2023-03-07 17:58:20,673 INFO [train2.py:809] (1/4) Epoch 5, batch 1600, loss[ctc_loss=0.2356, att_loss=0.3254, loss=0.3074, over 13994.00 frames. utt_duration=384.7 frames, utt_pad_proportion=0.3298, over 146.00 utterances.], tot_loss[ctc_loss=0.164, att_loss=0.282, loss=0.2584, over 3273987.32 frames. utt_duration=1250 frames, utt_pad_proportion=0.05507, over 10488.49 utterances.], batch size: 146, lr: 2.12e-02, grad_scale: 8.0 2023-03-07 17:58:34,840 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17545.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 17:58:49,544 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([1.7416, 2.7783, 3.4334, 2.4877, 2.9593, 3.1203, 2.4986, 1.3310], device='cuda:1'), covar=tensor([0.1817, 0.1557, 0.1034, 0.4270, 0.1232, 0.2758, 0.1386, 1.1509], device='cuda:1'), in_proj_covar=tensor([0.0066, 0.0066, 0.0067, 0.0086, 0.0063, 0.0084, 0.0063, 0.0109], device='cuda:1'), out_proj_covar=tensor([4.7823e-05, 4.4000e-05, 4.4207e-05, 6.1033e-05, 4.4267e-05, 6.2055e-05, 4.6011e-05, 7.9841e-05], device='cuda:1') 2023-03-07 17:59:03,245 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17562.0, num_to_drop=1, layers_to_drop={2} 2023-03-07 17:59:20,716 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=17572.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 17:59:41,751 INFO [train2.py:809] (1/4) Epoch 5, batch 1650, loss[ctc_loss=0.1929, att_loss=0.2966, loss=0.2758, over 16394.00 frames. utt_duration=1492 frames, utt_pad_proportion=0.007964, over 44.00 utterances.], tot_loss[ctc_loss=0.1645, att_loss=0.2825, loss=0.2589, over 3272687.91 frames. utt_duration=1254 frames, utt_pad_proportion=0.05123, over 10449.52 utterances.], batch size: 44, lr: 2.12e-02, grad_scale: 8.0 2023-03-07 17:59:53,517 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.68 vs. limit=2.0 2023-03-07 18:00:19,757 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.071e+02 3.498e+02 4.244e+02 5.016e+02 1.623e+03, threshold=8.488e+02, percent-clipped=3.0 2023-03-07 18:00:22,199 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=17610.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:01:04,843 INFO [train2.py:809] (1/4) Epoch 5, batch 1700, loss[ctc_loss=0.1229, att_loss=0.2737, loss=0.2436, over 16620.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.004844, over 47.00 utterances.], tot_loss[ctc_loss=0.1645, att_loss=0.2822, loss=0.2586, over 3275719.98 frames. utt_duration=1235 frames, utt_pad_proportion=0.05518, over 10624.81 utterances.], batch size: 47, lr: 2.12e-02, grad_scale: 8.0 2023-03-07 18:02:26,019 INFO [train2.py:809] (1/4) Epoch 5, batch 1750, loss[ctc_loss=0.1135, att_loss=0.2413, loss=0.2157, over 15950.00 frames. utt_duration=1558 frames, utt_pad_proportion=0.005977, over 41.00 utterances.], tot_loss[ctc_loss=0.1627, att_loss=0.2815, loss=0.2577, over 3278310.40 frames. utt_duration=1252 frames, utt_pad_proportion=0.05119, over 10482.29 utterances.], batch size: 41, lr: 2.12e-02, grad_scale: 8.0 2023-03-07 18:02:40,573 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7650, 2.1112, 4.9888, 4.1749, 3.1069, 4.3006, 4.6978, 4.7442], device='cuda:1'), covar=tensor([0.0121, 0.1917, 0.0123, 0.0840, 0.2179, 0.0230, 0.0108, 0.0183], device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0245, 0.0123, 0.0299, 0.0320, 0.0183, 0.0109, 0.0134], device='cuda:1'), out_proj_covar=tensor([1.1416e-04, 1.9415e-04, 1.0472e-04, 2.3862e-04, 2.5620e-04, 1.5526e-04, 9.4035e-05, 1.1755e-04], device='cuda:1') 2023-03-07 18:02:47,901 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17700.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:03:02,978 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.276e+02 3.443e+02 4.314e+02 5.475e+02 1.397e+03, threshold=8.628e+02, percent-clipped=3.0 2023-03-07 18:03:17,844 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17718.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:03:37,935 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17730.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:03:46,845 INFO [train2.py:809] (1/4) Epoch 5, batch 1800, loss[ctc_loss=0.1438, att_loss=0.2568, loss=0.2342, over 15943.00 frames. utt_duration=1557 frames, utt_pad_proportion=0.007773, over 41.00 utterances.], tot_loss[ctc_loss=0.1635, att_loss=0.2818, loss=0.2581, over 3270299.05 frames. utt_duration=1223 frames, utt_pad_proportion=0.06218, over 10713.36 utterances.], batch size: 41, lr: 2.11e-02, grad_scale: 8.0 2023-03-07 18:03:51,597 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17739.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 18:04:27,116 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17761.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:04:48,842 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.5772, 3.9255, 3.7352, 3.9405, 3.9068, 3.7517, 3.3080, 3.8583], device='cuda:1'), covar=tensor([0.0097, 0.0115, 0.0101, 0.0067, 0.0070, 0.0089, 0.0340, 0.0185], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0052, 0.0055, 0.0037, 0.0037, 0.0047, 0.0068, 0.0065], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-07 18:05:07,325 INFO [train2.py:809] (1/4) Epoch 5, batch 1850, loss[ctc_loss=0.1584, att_loss=0.2746, loss=0.2514, over 16465.00 frames. utt_duration=1433 frames, utt_pad_proportion=0.006629, over 46.00 utterances.], tot_loss[ctc_loss=0.1628, att_loss=0.2813, loss=0.2576, over 3265271.95 frames. utt_duration=1236 frames, utt_pad_proportion=0.05913, over 10580.36 utterances.], batch size: 46, lr: 2.11e-02, grad_scale: 8.0 2023-03-07 18:05:15,577 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17791.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:05:45,054 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.153e+02 3.254e+02 3.960e+02 5.162e+02 1.019e+03, threshold=7.921e+02, percent-clipped=3.0 2023-03-07 18:06:20,541 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6902, 5.1142, 4.3836, 5.1889, 4.5075, 4.8382, 5.1943, 5.0469], device='cuda:1'), covar=tensor([0.0296, 0.0245, 0.0820, 0.0138, 0.0413, 0.0181, 0.0254, 0.0163], device='cuda:1'), in_proj_covar=tensor([0.0200, 0.0168, 0.0216, 0.0138, 0.0186, 0.0137, 0.0161, 0.0155], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-07 18:06:28,243 INFO [train2.py:809] (1/4) Epoch 5, batch 1900, loss[ctc_loss=0.1196, att_loss=0.2576, loss=0.23, over 16262.00 frames. utt_duration=1514 frames, utt_pad_proportion=0.008392, over 43.00 utterances.], tot_loss[ctc_loss=0.1624, att_loss=0.2812, loss=0.2575, over 3275629.47 frames. utt_duration=1249 frames, utt_pad_proportion=0.05264, over 10501.43 utterances.], batch size: 43, lr: 2.11e-02, grad_scale: 8.0 2023-03-07 18:06:42,521 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17845.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:07:08,725 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.96 vs. limit=2.0 2023-03-07 18:07:41,113 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=2.50 vs. limit=2.0 2023-03-07 18:07:49,565 INFO [train2.py:809] (1/4) Epoch 5, batch 1950, loss[ctc_loss=0.1467, att_loss=0.2565, loss=0.2346, over 14511.00 frames. utt_duration=1815 frames, utt_pad_proportion=0.03128, over 32.00 utterances.], tot_loss[ctc_loss=0.1621, att_loss=0.2814, loss=0.2576, over 3271804.62 frames. utt_duration=1247 frames, utt_pad_proportion=0.05396, over 10509.83 utterances.], batch size: 32, lr: 2.11e-02, grad_scale: 8.0 2023-03-07 18:08:00,639 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=17893.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:08:27,505 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.984e+02 3.619e+02 4.276e+02 5.445e+02 1.197e+03, threshold=8.551e+02, percent-clipped=4.0 2023-03-07 18:09:10,810 INFO [train2.py:809] (1/4) Epoch 5, batch 2000, loss[ctc_loss=0.1443, att_loss=0.2739, loss=0.2479, over 16530.00 frames. utt_duration=1471 frames, utt_pad_proportion=0.006351, over 45.00 utterances.], tot_loss[ctc_loss=0.1612, att_loss=0.2803, loss=0.2565, over 3267023.12 frames. utt_duration=1244 frames, utt_pad_proportion=0.05588, over 10519.47 utterances.], batch size: 45, lr: 2.10e-02, grad_scale: 8.0 2023-03-07 18:09:49,812 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-03-07 18:09:50,818 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17960.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:09:58,501 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.6464, 2.5934, 3.5781, 4.3887, 4.2650, 4.3089, 2.8721, 1.8646], device='cuda:1'), covar=tensor([0.0462, 0.2248, 0.1026, 0.0532, 0.0425, 0.0247, 0.1744, 0.2917], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0191, 0.0193, 0.0148, 0.0136, 0.0118, 0.0190, 0.0179], 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') 2023-03-07 18:10:19,266 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-03-07 18:10:32,015 INFO [train2.py:809] (1/4) Epoch 5, batch 2050, loss[ctc_loss=0.1399, att_loss=0.2852, loss=0.2562, over 16957.00 frames. utt_duration=1358 frames, utt_pad_proportion=0.008137, over 50.00 utterances.], tot_loss[ctc_loss=0.1606, att_loss=0.2802, loss=0.2563, over 3271302.83 frames. utt_duration=1251 frames, utt_pad_proportion=0.05329, over 10470.90 utterances.], batch size: 50, lr: 2.10e-02, grad_scale: 8.0 2023-03-07 18:11:14,244 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.268e+02 3.678e+02 4.390e+02 5.847e+02 1.003e+03, threshold=8.779e+02, percent-clipped=6.0 2023-03-07 18:11:16,197 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.7126, 2.2328, 2.9725, 4.4670, 4.1470, 4.3487, 2.9970, 2.1100], device='cuda:1'), covar=tensor([0.0382, 0.2561, 0.1649, 0.0584, 0.0535, 0.0145, 0.1511, 0.2599], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0194, 0.0195, 0.0150, 0.0138, 0.0118, 0.0191, 0.0181], 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') 2023-03-07 18:11:29,695 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18018.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:11:34,412 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18021.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:11:39,766 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-03-07 18:11:57,537 INFO [train2.py:809] (1/4) Epoch 5, batch 2100, loss[ctc_loss=0.1665, att_loss=0.277, loss=0.2549, over 16402.00 frames. utt_duration=1493 frames, utt_pad_proportion=0.006729, over 44.00 utterances.], tot_loss[ctc_loss=0.1615, att_loss=0.281, loss=0.2571, over 3277555.97 frames. utt_duration=1249 frames, utt_pad_proportion=0.05299, over 10509.30 utterances.], batch size: 44, lr: 2.10e-02, grad_scale: 8.0 2023-03-07 18:12:02,543 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18039.0, num_to_drop=1, layers_to_drop={2} 2023-03-07 18:12:30,768 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18056.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:12:36,976 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.1805, 3.8734, 3.0816, 3.4523, 3.9838, 3.7068, 2.4861, 4.4944], device='cuda:1'), covar=tensor([0.1551, 0.0388, 0.1458, 0.0679, 0.0507, 0.0642, 0.1132, 0.0332], device='cuda:1'), in_proj_covar=tensor([0.0153, 0.0129, 0.0179, 0.0143, 0.0156, 0.0169, 0.0151, 0.0145], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-03-07 18:12:46,837 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=18066.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:13:17,729 INFO [train2.py:809] (1/4) Epoch 5, batch 2150, loss[ctc_loss=0.1352, att_loss=0.2491, loss=0.2263, over 15383.00 frames. utt_duration=1760 frames, utt_pad_proportion=0.01033, over 35.00 utterances.], tot_loss[ctc_loss=0.1616, att_loss=0.2807, loss=0.2569, over 3274098.84 frames. utt_duration=1245 frames, utt_pad_proportion=0.05635, over 10533.98 utterances.], batch size: 35, lr: 2.09e-02, grad_scale: 8.0 2023-03-07 18:13:17,953 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18086.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:13:19,451 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=18087.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 18:13:54,648 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.249e+02 3.704e+02 4.287e+02 5.437e+02 1.158e+03, threshold=8.574e+02, percent-clipped=4.0 2023-03-07 18:14:38,378 INFO [train2.py:809] (1/4) Epoch 5, batch 2200, loss[ctc_loss=0.1764, att_loss=0.2956, loss=0.2718, over 16391.00 frames. utt_duration=1492 frames, utt_pad_proportion=0.0075, over 44.00 utterances.], tot_loss[ctc_loss=0.1617, att_loss=0.2805, loss=0.2567, over 3269621.98 frames. utt_duration=1218 frames, utt_pad_proportion=0.06289, over 10750.96 utterances.], batch size: 44, lr: 2.09e-02, grad_scale: 8.0 2023-03-07 18:15:02,250 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([1.7395, 2.9115, 3.2127, 2.1822, 2.7755, 3.0110, 2.8841, 1.1794], device='cuda:1'), covar=tensor([0.1096, 0.0961, 0.1157, 0.3470, 0.1160, 0.2288, 0.0679, 0.7989], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0070, 0.0070, 0.0092, 0.0070, 0.0086, 0.0064, 0.0112], device='cuda:1'), out_proj_covar=tensor([4.8973e-05, 4.6216e-05, 4.7288e-05, 6.5454e-05, 4.8282e-05, 6.5305e-05, 4.6658e-05, 8.3263e-05], device='cuda:1') 2023-03-07 18:15:12,184 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6897, 5.1858, 4.8502, 5.0424, 5.1658, 4.7286, 3.8022, 4.9978], device='cuda:1'), covar=tensor([0.0097, 0.0090, 0.0081, 0.0066, 0.0054, 0.0090, 0.0427, 0.0145], device='cuda:1'), in_proj_covar=tensor([0.0052, 0.0053, 0.0055, 0.0038, 0.0037, 0.0047, 0.0071, 0.0066], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-07 18:15:19,120 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0850, 5.0791, 5.1394, 3.2333, 4.8547, 4.5568, 4.4252, 2.4280], device='cuda:1'), covar=tensor([0.0200, 0.0078, 0.0122, 0.0889, 0.0086, 0.0170, 0.0284, 0.1597], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0059, 0.0048, 0.0093, 0.0057, 0.0067, 0.0076, 0.0097], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-07 18:15:41,256 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18174.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:15:42,910 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18175.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:16:00,106 INFO [train2.py:809] (1/4) Epoch 5, batch 2250, loss[ctc_loss=0.155, att_loss=0.2951, loss=0.2671, over 16965.00 frames. utt_duration=1359 frames, utt_pad_proportion=0.007626, over 50.00 utterances.], tot_loss[ctc_loss=0.1613, att_loss=0.2807, loss=0.2568, over 3270766.40 frames. utt_duration=1242 frames, utt_pad_proportion=0.05726, over 10546.81 utterances.], batch size: 50, lr: 2.09e-02, grad_scale: 8.0 2023-03-07 18:16:38,767 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.354e+02 3.428e+02 4.207e+02 5.393e+02 1.191e+03, threshold=8.413e+02, percent-clipped=8.0 2023-03-07 18:17:04,358 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5063, 2.7333, 3.6235, 2.6184, 3.5411, 4.6524, 4.4014, 3.3420], device='cuda:1'), covar=tensor([0.0393, 0.1711, 0.1057, 0.1500, 0.0992, 0.0390, 0.0458, 0.1250], device='cuda:1'), in_proj_covar=tensor([0.0200, 0.0205, 0.0207, 0.0191, 0.0210, 0.0197, 0.0162, 0.0197], 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') 2023-03-07 18:17:20,585 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18235.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:17:21,715 INFO [train2.py:809] (1/4) Epoch 5, batch 2300, loss[ctc_loss=0.1819, att_loss=0.2997, loss=0.2761, over 17279.00 frames. utt_duration=1258 frames, utt_pad_proportion=0.01075, over 55.00 utterances.], tot_loss[ctc_loss=0.1603, att_loss=0.2802, loss=0.2562, over 3259664.88 frames. utt_duration=1240 frames, utt_pad_proportion=0.05889, over 10526.60 utterances.], batch size: 55, lr: 2.09e-02, grad_scale: 8.0 2023-03-07 18:17:22,106 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18236.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:17:24,502 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-03-07 18:17:41,120 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-03-07 18:17:56,050 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.7551, 2.4838, 3.1888, 4.5057, 4.2409, 4.3088, 2.8243, 1.9452], device='cuda:1'), covar=tensor([0.0390, 0.2225, 0.1216, 0.0410, 0.0365, 0.0197, 0.1773, 0.2384], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0192, 0.0189, 0.0147, 0.0134, 0.0116, 0.0183, 0.0172], 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') 2023-03-07 18:17:56,706 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=2.01 vs. limit=2.0 2023-03-07 18:18:04,087 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.73 vs. limit=5.0 2023-03-07 18:18:13,419 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.71 vs. limit=5.0 2023-03-07 18:18:43,285 INFO [train2.py:809] (1/4) Epoch 5, batch 2350, loss[ctc_loss=0.1674, att_loss=0.2924, loss=0.2674, over 16863.00 frames. utt_duration=1378 frames, utt_pad_proportion=0.007581, over 49.00 utterances.], tot_loss[ctc_loss=0.1607, att_loss=0.2808, loss=0.2568, over 3273149.09 frames. utt_duration=1242 frames, utt_pad_proportion=0.05364, over 10558.23 utterances.], batch size: 49, lr: 2.08e-02, grad_scale: 8.0 2023-03-07 18:18:48,296 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0208, 5.2063, 5.1774, 2.9933, 5.0227, 4.5823, 4.3181, 2.0637], device='cuda:1'), covar=tensor([0.0233, 0.0091, 0.0125, 0.1202, 0.0093, 0.0153, 0.0356, 0.2356], device='cuda:1'), in_proj_covar=tensor([0.0051, 0.0059, 0.0048, 0.0094, 0.0057, 0.0068, 0.0077, 0.0099], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-07 18:19:07,979 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7997, 5.1138, 5.1732, 5.0602, 5.2298, 5.0787, 4.8181, 4.7263], device='cuda:1'), covar=tensor([0.0935, 0.0460, 0.0202, 0.0384, 0.0228, 0.0275, 0.0248, 0.0324], device='cuda:1'), in_proj_covar=tensor([0.0388, 0.0232, 0.0163, 0.0205, 0.0253, 0.0277, 0.0213, 0.0242], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-07 18:19:21,029 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.258e+02 3.468e+02 4.143e+02 4.951e+02 1.211e+03, threshold=8.286e+02, percent-clipped=3.0 2023-03-07 18:19:32,544 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18316.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:19:35,867 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.8859, 2.4562, 3.4077, 4.5863, 4.2234, 4.4255, 2.8839, 1.9078], device='cuda:1'), covar=tensor([0.0405, 0.2461, 0.1180, 0.0342, 0.0360, 0.0157, 0.1627, 0.2772], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0190, 0.0188, 0.0146, 0.0133, 0.0114, 0.0181, 0.0171], 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') 2023-03-07 18:19:56,942 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.2834, 1.7442, 1.5384, 1.7549, 2.0904, 2.1005, 1.5770, 2.6860], device='cuda:1'), covar=tensor([0.0853, 0.1616, 0.1600, 0.0711, 0.0965, 0.0860, 0.1446, 0.0538], device='cuda:1'), in_proj_covar=tensor([0.0093, 0.0101, 0.0102, 0.0086, 0.0084, 0.0084, 0.0099, 0.0087], device='cuda:1'), out_proj_covar=tensor([4.1096e-05, 5.3998e-05, 5.3849e-05, 4.3816e-05, 4.0099e-05, 4.5662e-05, 5.1067e-05, 4.7765e-05], device='cuda:1') 2023-03-07 18:20:04,253 INFO [train2.py:809] (1/4) Epoch 5, batch 2400, loss[ctc_loss=0.1488, att_loss=0.2508, loss=0.2304, over 15368.00 frames. utt_duration=1758 frames, utt_pad_proportion=0.01084, over 35.00 utterances.], tot_loss[ctc_loss=0.1603, att_loss=0.2808, loss=0.2567, over 3277801.01 frames. utt_duration=1253 frames, utt_pad_proportion=0.05076, over 10472.16 utterances.], batch size: 35, lr: 2.08e-02, grad_scale: 16.0 2023-03-07 18:20:10,584 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-03-07 18:20:38,590 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18356.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:21:26,053 INFO [train2.py:809] (1/4) Epoch 5, batch 2450, loss[ctc_loss=0.1418, att_loss=0.2844, loss=0.2559, over 16766.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.00657, over 48.00 utterances.], tot_loss[ctc_loss=0.1607, att_loss=0.2808, loss=0.2568, over 3279663.60 frames. utt_duration=1245 frames, utt_pad_proportion=0.05258, over 10553.36 utterances.], batch size: 48, lr: 2.08e-02, grad_scale: 16.0 2023-03-07 18:21:26,377 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18386.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:21:39,314 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7312, 2.6002, 5.0227, 3.8522, 2.9825, 4.5129, 4.6766, 4.6939], device='cuda:1'), covar=tensor([0.0151, 0.1887, 0.0164, 0.1215, 0.2388, 0.0246, 0.0176, 0.0272], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0251, 0.0123, 0.0303, 0.0325, 0.0190, 0.0109, 0.0139], device='cuda:1'), out_proj_covar=tensor([1.1587e-04, 1.9962e-04, 1.0432e-04, 2.4312e-04, 2.6182e-04, 1.6051e-04, 9.5755e-05, 1.2149e-04], device='cuda:1') 2023-03-07 18:21:56,523 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=18404.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:22:04,373 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.068e+02 3.376e+02 4.245e+02 5.566e+02 1.055e+03, threshold=8.490e+02, percent-clipped=7.0 2023-03-07 18:22:44,600 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=18434.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:22:47,529 INFO [train2.py:809] (1/4) Epoch 5, batch 2500, loss[ctc_loss=0.1623, att_loss=0.2958, loss=0.2691, over 17004.00 frames. utt_duration=1335 frames, utt_pad_proportion=0.008749, over 51.00 utterances.], tot_loss[ctc_loss=0.16, att_loss=0.28, loss=0.256, over 3277832.79 frames. utt_duration=1261 frames, utt_pad_proportion=0.05002, over 10411.81 utterances.], batch size: 51, lr: 2.08e-02, grad_scale: 16.0 2023-03-07 18:24:09,075 INFO [train2.py:809] (1/4) Epoch 5, batch 2550, loss[ctc_loss=0.1236, att_loss=0.254, loss=0.2279, over 14498.00 frames. utt_duration=1814 frames, utt_pad_proportion=0.04469, over 32.00 utterances.], tot_loss[ctc_loss=0.1591, att_loss=0.2791, loss=0.2551, over 3270373.59 frames. utt_duration=1265 frames, utt_pad_proportion=0.0509, over 10351.90 utterances.], batch size: 32, lr: 2.07e-02, grad_scale: 16.0 2023-03-07 18:24:15,581 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18490.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:24:44,666 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18508.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 18:24:45,765 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.865e+02 3.434e+02 4.162e+02 5.436e+02 1.075e+03, threshold=8.324e+02, percent-clipped=2.0 2023-03-07 18:25:18,654 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9969, 2.6001, 3.3909, 4.4767, 4.2847, 4.4361, 2.7222, 1.9662], device='cuda:1'), covar=tensor([0.0440, 0.2550, 0.1245, 0.0653, 0.0472, 0.0178, 0.1973, 0.2989], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0195, 0.0190, 0.0148, 0.0135, 0.0117, 0.0185, 0.0175], 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') 2023-03-07 18:25:20,134 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18530.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:25:21,709 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18531.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:25:29,425 INFO [train2.py:809] (1/4) Epoch 5, batch 2600, loss[ctc_loss=0.1711, att_loss=0.2715, loss=0.2514, over 15794.00 frames. utt_duration=1664 frames, utt_pad_proportion=0.007223, over 38.00 utterances.], tot_loss[ctc_loss=0.1605, att_loss=0.2797, loss=0.2559, over 3264538.55 frames. utt_duration=1218 frames, utt_pad_proportion=0.06387, over 10731.36 utterances.], batch size: 38, lr: 2.07e-02, grad_scale: 16.0 2023-03-07 18:25:54,368 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18551.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:26:09,126 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.92 vs. limit=5.0 2023-03-07 18:26:22,966 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18569.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 18:26:50,081 INFO [train2.py:809] (1/4) Epoch 5, batch 2650, loss[ctc_loss=0.2407, att_loss=0.3243, loss=0.3076, over 14334.00 frames. utt_duration=397 frames, utt_pad_proportion=0.3108, over 145.00 utterances.], tot_loss[ctc_loss=0.1617, att_loss=0.2803, loss=0.2566, over 3262007.02 frames. utt_duration=1212 frames, utt_pad_proportion=0.06648, over 10776.78 utterances.], batch size: 145, lr: 2.07e-02, grad_scale: 16.0 2023-03-07 18:27:27,353 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.996e+02 3.747e+02 4.549e+02 5.816e+02 1.798e+03, threshold=9.097e+02, percent-clipped=9.0 2023-03-07 18:27:38,424 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18616.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:28:10,689 INFO [train2.py:809] (1/4) Epoch 5, batch 2700, loss[ctc_loss=0.1439, att_loss=0.2629, loss=0.2391, over 16004.00 frames. utt_duration=1602 frames, utt_pad_proportion=0.007497, over 40.00 utterances.], tot_loss[ctc_loss=0.1602, att_loss=0.2789, loss=0.2552, over 3259805.22 frames. utt_duration=1233 frames, utt_pad_proportion=0.0618, over 10587.36 utterances.], batch size: 40, lr: 2.07e-02, grad_scale: 16.0 2023-03-07 18:28:19,870 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=2.51 vs. limit=2.0 2023-03-07 18:28:24,278 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3892, 2.6972, 4.9725, 3.8962, 3.1997, 4.4639, 4.5916, 4.7120], device='cuda:1'), covar=tensor([0.0210, 0.1893, 0.0123, 0.1283, 0.2151, 0.0240, 0.0149, 0.0249], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0244, 0.0120, 0.0300, 0.0316, 0.0190, 0.0109, 0.0139], device='cuda:1'), out_proj_covar=tensor([1.1936e-04, 1.9461e-04, 1.0236e-04, 2.4058e-04, 2.5584e-04, 1.6001e-04, 9.6465e-05, 1.2180e-04], device='cuda:1') 2023-03-07 18:28:56,339 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=18664.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:29:13,352 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18674.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:29:31,904 INFO [train2.py:809] (1/4) Epoch 5, batch 2750, loss[ctc_loss=0.1545, att_loss=0.2799, loss=0.2548, over 16890.00 frames. utt_duration=1380 frames, utt_pad_proportion=0.007047, over 49.00 utterances.], tot_loss[ctc_loss=0.1595, att_loss=0.2788, loss=0.255, over 3267778.16 frames. utt_duration=1245 frames, utt_pad_proportion=0.05713, over 10513.60 utterances.], batch size: 49, lr: 2.06e-02, grad_scale: 16.0 2023-03-07 18:30:03,987 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.45 vs. limit=5.0 2023-03-07 18:30:10,719 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.386e+02 3.574e+02 4.434e+02 5.832e+02 1.352e+03, threshold=8.867e+02, percent-clipped=7.0 2023-03-07 18:30:14,873 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.35 vs. limit=5.0 2023-03-07 18:30:51,271 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18735.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:30:52,488 INFO [train2.py:809] (1/4) Epoch 5, batch 2800, loss[ctc_loss=0.1774, att_loss=0.2891, loss=0.2668, over 17315.00 frames. utt_duration=1175 frames, utt_pad_proportion=0.02287, over 59.00 utterances.], tot_loss[ctc_loss=0.1596, att_loss=0.279, loss=0.2551, over 3264845.67 frames. utt_duration=1249 frames, utt_pad_proportion=0.05684, over 10464.91 utterances.], batch size: 59, lr: 2.06e-02, grad_scale: 8.0 2023-03-07 18:30:55,057 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-03-07 18:31:10,691 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=2.11 vs. limit=2.0 2023-03-07 18:32:14,578 INFO [train2.py:809] (1/4) Epoch 5, batch 2850, loss[ctc_loss=0.1162, att_loss=0.2694, loss=0.2388, over 15937.00 frames. utt_duration=1557 frames, utt_pad_proportion=0.007233, over 41.00 utterances.], tot_loss[ctc_loss=0.1583, att_loss=0.279, loss=0.2549, over 3275801.65 frames. utt_duration=1254 frames, utt_pad_proportion=0.05176, over 10464.23 utterances.], batch size: 41, lr: 2.06e-02, grad_scale: 8.0 2023-03-07 18:32:48,971 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1501, 4.8395, 4.8510, 2.8224, 2.0893, 2.6479, 4.5176, 3.7333], device='cuda:1'), covar=tensor([0.0467, 0.0148, 0.0167, 0.2338, 0.5465, 0.2504, 0.0222, 0.1712], device='cuda:1'), in_proj_covar=tensor([0.0273, 0.0172, 0.0202, 0.0184, 0.0361, 0.0327, 0.0181, 0.0317], device='cuda:1'), out_proj_covar=tensor([1.4158e-04, 7.6481e-05, 9.1422e-05, 8.5911e-05, 1.7866e-04, 1.5106e-04, 8.0115e-05, 1.5520e-04], device='cuda:1') 2023-03-07 18:32:54,813 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.041e+02 3.458e+02 4.600e+02 5.611e+02 1.235e+03, threshold=9.200e+02, percent-clipped=3.0 2023-03-07 18:33:27,982 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18830.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:33:29,477 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18831.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:33:37,330 INFO [train2.py:809] (1/4) Epoch 5, batch 2900, loss[ctc_loss=0.2456, att_loss=0.3244, loss=0.3087, over 14023.00 frames. utt_duration=385.6 frames, utt_pad_proportion=0.3293, over 146.00 utterances.], tot_loss[ctc_loss=0.1585, att_loss=0.2795, loss=0.2553, over 3279633.27 frames. utt_duration=1229 frames, utt_pad_proportion=0.05716, over 10684.27 utterances.], batch size: 146, lr: 2.06e-02, grad_scale: 8.0 2023-03-07 18:33:53,704 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18846.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:34:03,010 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.8361, 2.4548, 3.3543, 2.6117, 3.0631, 3.9251, 3.9031, 3.0241], device='cuda:1'), covar=tensor([0.0374, 0.1805, 0.0887, 0.1256, 0.1065, 0.0509, 0.0374, 0.1197], device='cuda:1'), in_proj_covar=tensor([0.0203, 0.0209, 0.0206, 0.0190, 0.0208, 0.0200, 0.0167, 0.0199], 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') 2023-03-07 18:34:21,714 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18864.0, num_to_drop=1, layers_to_drop={3} 2023-03-07 18:34:44,647 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=18878.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:34:46,043 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=18879.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:34:56,886 INFO [train2.py:809] (1/4) Epoch 5, batch 2950, loss[ctc_loss=0.1518, att_loss=0.2804, loss=0.2547, over 16683.00 frames. utt_duration=1452 frames, utt_pad_proportion=0.006572, over 46.00 utterances.], tot_loss[ctc_loss=0.1582, att_loss=0.2797, loss=0.2554, over 3285462.83 frames. utt_duration=1237 frames, utt_pad_proportion=0.05237, over 10640.78 utterances.], batch size: 46, lr: 2.05e-02, grad_scale: 8.0 2023-03-07 18:35:34,921 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.226e+02 3.285e+02 4.050e+02 4.764e+02 8.688e+02, threshold=8.101e+02, percent-clipped=0.0 2023-03-07 18:36:07,986 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18930.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:36:17,499 INFO [train2.py:809] (1/4) Epoch 5, batch 3000, loss[ctc_loss=0.1234, att_loss=0.2695, loss=0.2403, over 16971.00 frames. utt_duration=1359 frames, utt_pad_proportion=0.00729, over 50.00 utterances.], tot_loss[ctc_loss=0.1594, att_loss=0.2799, loss=0.2558, over 3278521.64 frames. utt_duration=1223 frames, utt_pad_proportion=0.05872, over 10733.03 utterances.], batch size: 50, lr: 2.05e-02, grad_scale: 8.0 2023-03-07 18:36:17,500 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-07 18:36:30,407 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.4596, 2.1764, 2.9985, 2.1906, 2.7583, 3.5476, 3.3974, 2.4936], device='cuda:1'), covar=tensor([0.0503, 0.2070, 0.1288, 0.1533, 0.1330, 0.0812, 0.0555, 0.1755], device='cuda:1'), in_proj_covar=tensor([0.0199, 0.0203, 0.0201, 0.0188, 0.0205, 0.0201, 0.0166, 0.0198], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-03-07 18:36:31,698 INFO [train2.py:843] (1/4) Epoch 5, validation: ctc_loss=0.07531, att_loss=0.2513, loss=0.2161, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-07 18:36:31,699 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-07 18:37:02,340 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6374, 5.0315, 4.8163, 4.9713, 5.1225, 5.0701, 4.7669, 4.6218], device='cuda:1'), covar=tensor([0.1249, 0.0521, 0.0320, 0.0467, 0.0302, 0.0277, 0.0298, 0.0373], device='cuda:1'), in_proj_covar=tensor([0.0389, 0.0236, 0.0165, 0.0207, 0.0257, 0.0284, 0.0217, 0.0247], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-07 18:37:41,594 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-03-07 18:37:52,825 INFO [train2.py:809] (1/4) Epoch 5, batch 3050, loss[ctc_loss=0.1523, att_loss=0.276, loss=0.2512, over 16761.00 frames. utt_duration=1398 frames, utt_pad_proportion=0.006747, over 48.00 utterances.], tot_loss[ctc_loss=0.1592, att_loss=0.2797, loss=0.2556, over 3272113.44 frames. utt_duration=1235 frames, utt_pad_proportion=0.05787, over 10613.85 utterances.], batch size: 48, lr: 2.05e-02, grad_scale: 8.0 2023-03-07 18:38:01,036 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18991.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:38:31,922 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.565e+02 3.659e+02 4.471e+02 5.591e+02 1.197e+03, threshold=8.941e+02, percent-clipped=8.0 2023-03-07 18:39:05,002 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19030.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:39:13,977 INFO [train2.py:809] (1/4) Epoch 5, batch 3100, loss[ctc_loss=0.2256, att_loss=0.2988, loss=0.2842, over 16542.00 frames. utt_duration=1472 frames, utt_pad_proportion=0.006167, over 45.00 utterances.], tot_loss[ctc_loss=0.1594, att_loss=0.2799, loss=0.2558, over 3278105.04 frames. utt_duration=1219 frames, utt_pad_proportion=0.05959, over 10768.98 utterances.], batch size: 45, lr: 2.05e-02, grad_scale: 8.0 2023-03-07 18:40:00,862 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6452, 2.2566, 5.0507, 4.0837, 2.9767, 4.4039, 4.5663, 4.6367], device='cuda:1'), covar=tensor([0.0173, 0.2036, 0.0143, 0.0949, 0.2333, 0.0233, 0.0165, 0.0230], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0245, 0.0122, 0.0302, 0.0317, 0.0190, 0.0110, 0.0138], device='cuda:1'), out_proj_covar=tensor([1.1853e-04, 1.9550e-04, 1.0414e-04, 2.4228e-04, 2.5788e-04, 1.6059e-04, 9.8240e-05, 1.2202e-04], device='cuda:1') 2023-03-07 18:40:36,263 INFO [train2.py:809] (1/4) Epoch 5, batch 3150, loss[ctc_loss=0.1776, att_loss=0.295, loss=0.2715, over 17590.00 frames. utt_duration=892.1 frames, utt_pad_proportion=0.06582, over 79.00 utterances.], tot_loss[ctc_loss=0.16, att_loss=0.2801, loss=0.2561, over 3274034.29 frames. utt_duration=1217 frames, utt_pad_proportion=0.06036, over 10769.94 utterances.], batch size: 79, lr: 2.04e-02, grad_scale: 8.0 2023-03-07 18:41:14,954 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.126e+02 3.661e+02 4.363e+02 5.729e+02 1.353e+03, threshold=8.726e+02, percent-clipped=4.0 2023-03-07 18:41:46,853 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19129.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 18:41:57,271 INFO [train2.py:809] (1/4) Epoch 5, batch 3200, loss[ctc_loss=0.2333, att_loss=0.3221, loss=0.3043, over 13624.00 frames. utt_duration=377.3 frames, utt_pad_proportion=0.345, over 145.00 utterances.], tot_loss[ctc_loss=0.1592, att_loss=0.2799, loss=0.2558, over 3274929.51 frames. utt_duration=1229 frames, utt_pad_proportion=0.05838, over 10674.82 utterances.], batch size: 145, lr: 2.04e-02, grad_scale: 8.0 2023-03-07 18:42:14,020 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19146.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:42:43,196 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19164.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 18:42:54,134 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.5341, 2.4283, 3.0280, 4.3002, 3.9870, 4.2023, 2.4720, 1.6969], device='cuda:1'), covar=tensor([0.0495, 0.2315, 0.1264, 0.0502, 0.0424, 0.0210, 0.2008, 0.2861], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0195, 0.0189, 0.0150, 0.0140, 0.0121, 0.0192, 0.0177], 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') 2023-03-07 18:43:03,338 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9448, 5.3320, 4.7821, 5.4312, 4.7842, 5.0622, 5.5472, 5.3231], device='cuda:1'), covar=tensor([0.0350, 0.0210, 0.0743, 0.0128, 0.0368, 0.0141, 0.0142, 0.0126], device='cuda:1'), in_proj_covar=tensor([0.0209, 0.0173, 0.0227, 0.0146, 0.0191, 0.0144, 0.0165, 0.0156], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-07 18:43:19,470 INFO [train2.py:809] (1/4) Epoch 5, batch 3250, loss[ctc_loss=0.148, att_loss=0.2837, loss=0.2566, over 17116.00 frames. utt_duration=1224 frames, utt_pad_proportion=0.01521, over 56.00 utterances.], tot_loss[ctc_loss=0.1588, att_loss=0.2801, loss=0.2558, over 3278268.41 frames. utt_duration=1229 frames, utt_pad_proportion=0.05748, over 10683.36 utterances.], batch size: 56, lr: 2.04e-02, grad_scale: 8.0 2023-03-07 18:43:26,710 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19190.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 18:43:32,677 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=19194.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:43:58,426 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.144e+02 3.305e+02 4.293e+02 5.572e+02 8.816e+02, threshold=8.587e+02, percent-clipped=1.0 2023-03-07 18:43:58,818 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([1.7610, 3.0018, 3.3733, 2.5643, 3.0889, 3.1764, 3.0667, 2.0741], device='cuda:1'), covar=tensor([0.1489, 0.0639, 0.1487, 0.3857, 0.1775, 0.1710, 0.0616, 0.6624], device='cuda:1'), in_proj_covar=tensor([0.0067, 0.0066, 0.0068, 0.0089, 0.0064, 0.0083, 0.0061, 0.0106], device='cuda:1'), out_proj_covar=tensor([4.9852e-05, 4.5183e-05, 4.8593e-05, 6.4777e-05, 4.6354e-05, 6.4194e-05, 4.4760e-05, 7.9903e-05], device='cuda:1') 2023-03-07 18:44:01,730 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=19212.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 18:44:40,684 INFO [train2.py:809] (1/4) Epoch 5, batch 3300, loss[ctc_loss=0.1227, att_loss=0.2606, loss=0.233, over 13585.00 frames. utt_duration=1813 frames, utt_pad_proportion=0.07587, over 30.00 utterances.], tot_loss[ctc_loss=0.1587, att_loss=0.2797, loss=0.2555, over 3278281.19 frames. utt_duration=1224 frames, utt_pad_proportion=0.05801, over 10725.58 utterances.], batch size: 30, lr: 2.04e-02, grad_scale: 8.0 2023-03-07 18:46:01,521 INFO [train2.py:809] (1/4) Epoch 5, batch 3350, loss[ctc_loss=0.1364, att_loss=0.2795, loss=0.2509, over 17033.00 frames. utt_duration=1287 frames, utt_pad_proportion=0.01075, over 53.00 utterances.], tot_loss[ctc_loss=0.1575, att_loss=0.2789, loss=0.2546, over 3268689.18 frames. utt_duration=1226 frames, utt_pad_proportion=0.06132, over 10678.19 utterances.], batch size: 53, lr: 2.03e-02, grad_scale: 8.0 2023-03-07 18:46:01,701 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19286.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:46:37,318 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.03 vs. limit=5.0 2023-03-07 18:46:39,662 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.279e+02 3.326e+02 4.027e+02 5.244e+02 1.066e+03, threshold=8.053e+02, percent-clipped=2.0 2023-03-07 18:47:13,598 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19330.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:47:18,296 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.4716, 2.1645, 3.1769, 4.3415, 4.1644, 4.0444, 2.7095, 1.6309], device='cuda:1'), covar=tensor([0.0564, 0.2783, 0.1274, 0.0527, 0.0470, 0.0298, 0.1924, 0.3109], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0197, 0.0188, 0.0149, 0.0140, 0.0121, 0.0186, 0.0175], 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') 2023-03-07 18:47:22,650 INFO [train2.py:809] (1/4) Epoch 5, batch 3400, loss[ctc_loss=0.1258, att_loss=0.2406, loss=0.2177, over 15482.00 frames. utt_duration=1722 frames, utt_pad_proportion=0.009919, over 36.00 utterances.], tot_loss[ctc_loss=0.1565, att_loss=0.2774, loss=0.2532, over 3258428.37 frames. utt_duration=1247 frames, utt_pad_proportion=0.0594, over 10464.55 utterances.], batch size: 36, lr: 2.03e-02, grad_scale: 8.0 2023-03-07 18:48:07,737 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-03-07 18:48:30,579 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=19378.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:48:42,955 INFO [train2.py:809] (1/4) Epoch 5, batch 3450, loss[ctc_loss=0.1623, att_loss=0.2659, loss=0.2452, over 15496.00 frames. utt_duration=1724 frames, utt_pad_proportion=0.008897, over 36.00 utterances.], tot_loss[ctc_loss=0.1578, att_loss=0.2786, loss=0.2544, over 3264036.15 frames. utt_duration=1236 frames, utt_pad_proportion=0.06026, over 10572.13 utterances.], batch size: 36, lr: 2.03e-02, grad_scale: 8.0 2023-03-07 18:48:56,567 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19394.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:49:21,715 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.889e+02 3.490e+02 4.079e+02 5.311e+02 1.453e+03, threshold=8.158e+02, percent-clipped=3.0 2023-03-07 18:50:03,982 INFO [train2.py:809] (1/4) Epoch 5, batch 3500, loss[ctc_loss=0.1858, att_loss=0.3005, loss=0.2776, over 17159.00 frames. utt_duration=694.9 frames, utt_pad_proportion=0.1281, over 99.00 utterances.], tot_loss[ctc_loss=0.1569, att_loss=0.2785, loss=0.2542, over 3269905.00 frames. utt_duration=1240 frames, utt_pad_proportion=0.057, over 10560.36 utterances.], batch size: 99, lr: 2.03e-02, grad_scale: 8.0 2023-03-07 18:50:34,960 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19455.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:50:39,484 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.6996, 5.9397, 5.2272, 5.8152, 5.4778, 5.2330, 5.2779, 5.1409], device='cuda:1'), covar=tensor([0.1173, 0.0885, 0.0870, 0.0783, 0.0654, 0.1347, 0.2377, 0.2402], device='cuda:1'), in_proj_covar=tensor([0.0335, 0.0375, 0.0295, 0.0305, 0.0277, 0.0355, 0.0403, 0.0379], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-07 18:51:12,359 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-03-07 18:51:24,358 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19485.0, num_to_drop=1, layers_to_drop={2} 2023-03-07 18:51:25,710 INFO [train2.py:809] (1/4) Epoch 5, batch 3550, loss[ctc_loss=0.1463, att_loss=0.256, loss=0.2341, over 16189.00 frames. utt_duration=1581 frames, utt_pad_proportion=0.005127, over 41.00 utterances.], tot_loss[ctc_loss=0.1574, att_loss=0.2789, loss=0.2546, over 3275295.72 frames. utt_duration=1238 frames, utt_pad_proportion=0.05658, over 10597.92 utterances.], batch size: 41, lr: 2.02e-02, grad_scale: 8.0 2023-03-07 18:52:02,969 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.286e+02 3.371e+02 4.237e+02 4.987e+02 1.005e+03, threshold=8.474e+02, percent-clipped=3.0 2023-03-07 18:52:46,422 INFO [train2.py:809] (1/4) Epoch 5, batch 3600, loss[ctc_loss=0.1442, att_loss=0.2781, loss=0.2513, over 16631.00 frames. utt_duration=1417 frames, utt_pad_proportion=0.005229, over 47.00 utterances.], tot_loss[ctc_loss=0.1567, att_loss=0.2778, loss=0.2536, over 3265767.87 frames. utt_duration=1254 frames, utt_pad_proportion=0.05596, over 10428.83 utterances.], batch size: 47, lr: 2.02e-02, grad_scale: 8.0 2023-03-07 18:53:27,823 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.97 vs. limit=2.0 2023-03-07 18:54:07,426 INFO [train2.py:809] (1/4) Epoch 5, batch 3650, loss[ctc_loss=0.1443, att_loss=0.2762, loss=0.2498, over 17028.00 frames. utt_duration=1311 frames, utt_pad_proportion=0.01035, over 52.00 utterances.], tot_loss[ctc_loss=0.1574, att_loss=0.278, loss=0.2539, over 3255976.12 frames. utt_duration=1262 frames, utt_pad_proportion=0.05559, over 10330.92 utterances.], batch size: 52, lr: 2.02e-02, grad_scale: 8.0 2023-03-07 18:54:07,762 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19586.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:54:21,548 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8153, 5.1580, 5.0529, 4.9161, 5.2567, 5.1766, 4.9445, 4.6722], device='cuda:1'), covar=tensor([0.0911, 0.0390, 0.0206, 0.0652, 0.0241, 0.0277, 0.0218, 0.0317], device='cuda:1'), in_proj_covar=tensor([0.0394, 0.0231, 0.0170, 0.0210, 0.0259, 0.0289, 0.0215, 0.0248], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-07 18:54:44,974 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.264e+02 3.543e+02 4.665e+02 6.112e+02 1.723e+03, threshold=9.329e+02, percent-clipped=10.0 2023-03-07 18:54:59,217 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19618.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:55:25,672 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=19634.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:55:28,660 INFO [train2.py:809] (1/4) Epoch 5, batch 3700, loss[ctc_loss=0.1294, att_loss=0.2732, loss=0.2445, over 16543.00 frames. utt_duration=1472 frames, utt_pad_proportion=0.006047, over 45.00 utterances.], tot_loss[ctc_loss=0.1568, att_loss=0.2787, loss=0.2543, over 3262100.45 frames. utt_duration=1245 frames, utt_pad_proportion=0.05926, over 10495.92 utterances.], batch size: 45, lr: 2.02e-02, grad_scale: 8.0 2023-03-07 18:56:39,213 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19679.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:56:49,597 INFO [train2.py:809] (1/4) Epoch 5, batch 3750, loss[ctc_loss=0.1837, att_loss=0.3003, loss=0.277, over 17051.00 frames. utt_duration=1289 frames, utt_pad_proportion=0.009557, over 53.00 utterances.], tot_loss[ctc_loss=0.156, att_loss=0.2778, loss=0.2534, over 3261791.86 frames. utt_duration=1250 frames, utt_pad_proportion=0.05846, over 10446.82 utterances.], batch size: 53, lr: 2.01e-02, grad_scale: 8.0 2023-03-07 18:57:27,464 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.299e+02 3.451e+02 3.922e+02 5.029e+02 1.290e+03, threshold=7.843e+02, percent-clipped=3.0 2023-03-07 18:58:11,650 INFO [train2.py:809] (1/4) Epoch 5, batch 3800, loss[ctc_loss=0.1432, att_loss=0.279, loss=0.2519, over 16889.00 frames. utt_duration=1380 frames, utt_pad_proportion=0.006318, over 49.00 utterances.], tot_loss[ctc_loss=0.1543, att_loss=0.2766, loss=0.2522, over 3263540.21 frames. utt_duration=1261 frames, utt_pad_proportion=0.05487, over 10361.49 utterances.], batch size: 49, lr: 2.01e-02, grad_scale: 8.0 2023-03-07 18:58:33,867 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19750.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 18:59:08,557 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.04 vs. limit=2.0 2023-03-07 18:59:30,506 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19785.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 18:59:31,759 INFO [train2.py:809] (1/4) Epoch 5, batch 3850, loss[ctc_loss=0.1445, att_loss=0.2782, loss=0.2514, over 17304.00 frames. utt_duration=1260 frames, utt_pad_proportion=0.01096, over 55.00 utterances.], tot_loss[ctc_loss=0.1534, att_loss=0.276, loss=0.2515, over 3267888.17 frames. utt_duration=1276 frames, utt_pad_proportion=0.05076, over 10256.21 utterances.], batch size: 55, lr: 2.01e-02, grad_scale: 8.0 2023-03-07 19:00:06,473 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5198, 2.7440, 3.6626, 2.7107, 3.5080, 4.5565, 4.3977, 3.2135], device='cuda:1'), covar=tensor([0.0324, 0.1748, 0.0907, 0.1405, 0.1012, 0.0461, 0.0419, 0.1364], device='cuda:1'), in_proj_covar=tensor([0.0202, 0.0209, 0.0203, 0.0191, 0.0211, 0.0208, 0.0171, 0.0204], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-03-07 19:00:09,266 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.273e+02 3.504e+02 4.097e+02 4.952e+02 1.319e+03, threshold=8.195e+02, percent-clipped=3.0 2023-03-07 19:00:13,269 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.14 vs. limit=5.0 2023-03-07 19:00:44,703 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=19833.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 19:00:48,872 INFO [train2.py:809] (1/4) Epoch 5, batch 3900, loss[ctc_loss=0.1001, att_loss=0.2325, loss=0.206, over 15345.00 frames. utt_duration=1755 frames, utt_pad_proportion=0.01077, over 35.00 utterances.], tot_loss[ctc_loss=0.1536, att_loss=0.275, loss=0.2507, over 3253562.31 frames. utt_duration=1278 frames, utt_pad_proportion=0.0535, over 10196.28 utterances.], batch size: 35, lr: 2.01e-02, grad_scale: 8.0 2023-03-07 19:02:05,456 INFO [train2.py:809] (1/4) Epoch 5, batch 3950, loss[ctc_loss=0.2076, att_loss=0.3071, loss=0.2872, over 17103.00 frames. utt_duration=685.5 frames, utt_pad_proportion=0.1334, over 100.00 utterances.], tot_loss[ctc_loss=0.1537, att_loss=0.2756, loss=0.2512, over 3259362.30 frames. utt_duration=1285 frames, utt_pad_proportion=0.04975, over 10158.73 utterances.], batch size: 100, lr: 2.00e-02, grad_scale: 8.0 2023-03-07 19:02:19,640 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.1103, 6.2380, 5.6291, 6.0810, 5.9589, 5.7344, 5.8561, 5.6505], device='cuda:1'), covar=tensor([0.1109, 0.0888, 0.0599, 0.0673, 0.0582, 0.1300, 0.1747, 0.2075], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0383, 0.0299, 0.0308, 0.0275, 0.0369, 0.0405, 0.0379], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-07 19:02:43,148 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.585e+02 3.417e+02 4.181e+02 5.217e+02 1.635e+03, threshold=8.363e+02, percent-clipped=3.0 2023-03-07 19:03:26,314 INFO [train2.py:809] (1/4) Epoch 6, batch 0, loss[ctc_loss=0.1719, att_loss=0.2949, loss=0.2703, over 17052.00 frames. utt_duration=1288 frames, utt_pad_proportion=0.00992, over 53.00 utterances.], tot_loss[ctc_loss=0.1719, att_loss=0.2949, loss=0.2703, over 17052.00 frames. utt_duration=1288 frames, utt_pad_proportion=0.00992, over 53.00 utterances.], batch size: 53, lr: 1.87e-02, grad_scale: 8.0 2023-03-07 19:03:26,315 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-07 19:03:38,954 INFO [train2.py:843] (1/4) Epoch 6, validation: ctc_loss=0.07525, att_loss=0.2502, loss=0.2152, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-07 19:03:38,955 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-07 19:04:08,921 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7193, 5.0518, 4.5519, 5.0854, 4.5665, 4.8345, 5.2561, 5.0051], device='cuda:1'), covar=tensor([0.0382, 0.0276, 0.0646, 0.0189, 0.0404, 0.0194, 0.0194, 0.0138], device='cuda:1'), in_proj_covar=tensor([0.0217, 0.0179, 0.0231, 0.0151, 0.0196, 0.0146, 0.0170, 0.0163], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-07 19:04:26,016 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-03-07 19:04:58,946 INFO [train2.py:809] (1/4) Epoch 6, batch 50, loss[ctc_loss=0.1278, att_loss=0.2444, loss=0.2211, over 15612.00 frames. utt_duration=1689 frames, utt_pad_proportion=0.01032, over 37.00 utterances.], tot_loss[ctc_loss=0.1492, att_loss=0.2748, loss=0.2497, over 739033.08 frames. utt_duration=1270 frames, utt_pad_proportion=0.04727, over 2330.88 utterances.], batch size: 37, lr: 1.87e-02, grad_scale: 8.0 2023-03-07 19:05:05,355 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19974.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:05:56,609 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.5729, 1.4712, 1.7756, 1.3001, 2.9620, 1.8937, 1.2976, 1.5154], device='cuda:1'), covar=tensor([0.0454, 0.2672, 0.2426, 0.2027, 0.0505, 0.1425, 0.2754, 0.2051], device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0103, 0.0099, 0.0093, 0.0088, 0.0087, 0.0106, 0.0087], device='cuda:1'), out_proj_covar=tensor([4.2727e-05, 5.5617e-05, 5.4141e-05, 4.5832e-05, 4.1338e-05, 4.9346e-05, 5.4951e-05, 4.8506e-05], device='cuda:1') 2023-03-07 19:05:59,254 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-03-07 19:06:05,489 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.789e+02 3.240e+02 3.986e+02 5.094e+02 1.358e+03, threshold=7.972e+02, percent-clipped=5.0 2023-03-07 19:06:22,312 INFO [train2.py:809] (1/4) Epoch 6, batch 100, loss[ctc_loss=0.1432, att_loss=0.2651, loss=0.2407, over 16418.00 frames. utt_duration=1494 frames, utt_pad_proportion=0.006604, over 44.00 utterances.], tot_loss[ctc_loss=0.1484, att_loss=0.2722, loss=0.2475, over 1290866.28 frames. utt_duration=1364 frames, utt_pad_proportion=0.03216, over 3788.88 utterances.], batch size: 44, lr: 1.86e-02, grad_scale: 8.0 2023-03-07 19:06:22,650 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20020.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:07:09,766 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20050.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:07:41,940 INFO [train2.py:809] (1/4) Epoch 6, batch 150, loss[ctc_loss=0.1812, att_loss=0.2973, loss=0.2741, over 16831.00 frames. utt_duration=688.5 frames, utt_pad_proportion=0.1373, over 98.00 utterances.], tot_loss[ctc_loss=0.1497, att_loss=0.2744, loss=0.2495, over 1735841.94 frames. utt_duration=1304 frames, utt_pad_proportion=0.04226, over 5331.19 utterances.], batch size: 98, lr: 1.86e-02, grad_scale: 8.0 2023-03-07 19:08:00,424 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20081.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:08:12,687 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20089.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:08:27,241 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=20098.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:08:33,258 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.64 vs. limit=2.0 2023-03-07 19:08:45,954 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.802e+02 3.260e+02 3.980e+02 4.904e+02 1.174e+03, threshold=7.960e+02, percent-clipped=3.0 2023-03-07 19:08:49,533 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20112.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:08:53,219 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([1.3656, 2.7328, 3.0690, 2.0965, 2.8842, 2.5083, 2.7501, 1.7916], device='cuda:1'), covar=tensor([0.1410, 0.0804, 0.1940, 0.5663, 0.1420, 0.2692, 0.1084, 0.8091], device='cuda:1'), in_proj_covar=tensor([0.0067, 0.0071, 0.0072, 0.0100, 0.0068, 0.0091, 0.0066, 0.0113], device='cuda:1'), out_proj_covar=tensor([5.1072e-05, 4.8351e-05, 5.2799e-05, 7.2793e-05, 4.9410e-05, 7.0267e-05, 4.8407e-05, 8.6294e-05], device='cuda:1') 2023-03-07 19:09:02,827 INFO [train2.py:809] (1/4) Epoch 6, batch 200, loss[ctc_loss=0.1339, att_loss=0.2752, loss=0.247, over 16763.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.005908, over 48.00 utterances.], tot_loss[ctc_loss=0.1505, att_loss=0.2745, loss=0.2497, over 2076137.17 frames. utt_duration=1258 frames, utt_pad_proportion=0.05112, over 6606.61 utterances.], batch size: 48, lr: 1.86e-02, grad_scale: 8.0 2023-03-07 19:09:51,022 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20150.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:09:56,463 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-03-07 19:10:23,258 INFO [train2.py:809] (1/4) Epoch 6, batch 250, loss[ctc_loss=0.1902, att_loss=0.3034, loss=0.2808, over 17058.00 frames. utt_duration=1289 frames, utt_pad_proportion=0.009253, over 53.00 utterances.], tot_loss[ctc_loss=0.1506, att_loss=0.2755, loss=0.2505, over 2355401.45 frames. utt_duration=1273 frames, utt_pad_proportion=0.04228, over 7407.81 utterances.], batch size: 53, lr: 1.86e-02, grad_scale: 8.0 2023-03-07 19:10:28,300 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20173.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:10:32,995 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7698, 4.7902, 4.8590, 2.6746, 4.6281, 4.2683, 4.0905, 2.2329], device='cuda:1'), covar=tensor([0.0164, 0.0091, 0.0106, 0.1091, 0.0086, 0.0212, 0.0277, 0.1531], device='cuda:1'), in_proj_covar=tensor([0.0055, 0.0064, 0.0053, 0.0099, 0.0059, 0.0071, 0.0080, 0.0102], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-07 19:10:35,229 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5512, 2.8638, 3.5833, 2.5491, 3.5616, 4.6349, 4.3631, 3.1556], device='cuda:1'), covar=tensor([0.0390, 0.2005, 0.1232, 0.1818, 0.1140, 0.0458, 0.0585, 0.1484], device='cuda:1'), in_proj_covar=tensor([0.0205, 0.0210, 0.0207, 0.0192, 0.0216, 0.0212, 0.0171, 0.0202], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-03-07 19:11:06,506 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.0748, 4.1717, 3.5616, 3.8134, 4.0360, 3.9085, 2.5804, 4.7435], device='cuda:1'), covar=tensor([0.1024, 0.0402, 0.1004, 0.0579, 0.0601, 0.0626, 0.1108, 0.0366], device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0137, 0.0184, 0.0149, 0.0169, 0.0176, 0.0157, 0.0161], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-03-07 19:11:15,414 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-03-07 19:11:17,511 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20204.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:11:27,957 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.870e+02 2.978e+02 3.813e+02 4.681e+02 1.334e+03, threshold=7.626e+02, percent-clipped=7.0 2023-03-07 19:11:43,059 INFO [train2.py:809] (1/4) Epoch 6, batch 300, loss[ctc_loss=0.1076, att_loss=0.2391, loss=0.2128, over 15998.00 frames. utt_duration=1601 frames, utt_pad_proportion=0.006068, over 40.00 utterances.], tot_loss[ctc_loss=0.1504, att_loss=0.2746, loss=0.2497, over 2554426.41 frames. utt_duration=1262 frames, utt_pad_proportion=0.04768, over 8104.26 utterances.], batch size: 40, lr: 1.86e-02, grad_scale: 4.0 2023-03-07 19:12:24,434 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.6873, 4.0828, 3.2226, 3.6980, 3.8795, 3.7109, 2.6842, 4.5144], device='cuda:1'), covar=tensor([0.1152, 0.0284, 0.1271, 0.0584, 0.0503, 0.0613, 0.0950, 0.0290], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0136, 0.0182, 0.0147, 0.0166, 0.0173, 0.0153, 0.0158], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-03-07 19:12:51,521 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5494, 4.7567, 4.7735, 4.9778, 2.0651, 4.8356, 2.7071, 2.4018], device='cuda:1'), covar=tensor([0.0160, 0.0143, 0.0661, 0.0143, 0.2777, 0.0169, 0.1719, 0.1811], device='cuda:1'), in_proj_covar=tensor([0.0106, 0.0100, 0.0255, 0.0112, 0.0230, 0.0107, 0.0228, 0.0209], device='cuda:1'), out_proj_covar=tensor([1.0554e-04, 9.9460e-05, 2.2474e-04, 1.0176e-04, 2.0566e-04, 1.0002e-04, 2.0024e-04, 1.8436e-04], device='cuda:1') 2023-03-07 19:12:54,591 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20265.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:13:01,866 INFO [train2.py:809] (1/4) Epoch 6, batch 350, loss[ctc_loss=0.1674, att_loss=0.2879, loss=0.2638, over 16900.00 frames. utt_duration=1381 frames, utt_pad_proportion=0.005701, over 49.00 utterances.], tot_loss[ctc_loss=0.151, att_loss=0.2743, loss=0.2497, over 2709831.00 frames. utt_duration=1250 frames, utt_pad_proportion=0.05263, over 8681.19 utterances.], batch size: 49, lr: 1.85e-02, grad_scale: 4.0 2023-03-07 19:13:08,361 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20274.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:13:14,435 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.21 vs. limit=5.0 2023-03-07 19:14:07,910 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.145e+02 3.150e+02 3.613e+02 4.726e+02 1.759e+03, threshold=7.227e+02, percent-clipped=5.0 2023-03-07 19:14:22,591 INFO [train2.py:809] (1/4) Epoch 6, batch 400, loss[ctc_loss=0.196, att_loss=0.2739, loss=0.2583, over 15651.00 frames. utt_duration=1694 frames, utt_pad_proportion=0.007742, over 37.00 utterances.], tot_loss[ctc_loss=0.1512, att_loss=0.2749, loss=0.2502, over 2825364.71 frames. utt_duration=1215 frames, utt_pad_proportion=0.06541, over 9313.48 utterances.], batch size: 37, lr: 1.85e-02, grad_scale: 8.0 2023-03-07 19:14:25,755 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=20322.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:15:42,855 INFO [train2.py:809] (1/4) Epoch 6, batch 450, loss[ctc_loss=0.1742, att_loss=0.2799, loss=0.2587, over 17366.00 frames. utt_duration=1104 frames, utt_pad_proportion=0.0348, over 63.00 utterances.], tot_loss[ctc_loss=0.1512, att_loss=0.2747, loss=0.25, over 2931131.57 frames. utt_duration=1210 frames, utt_pad_proportion=0.0638, over 9700.48 utterances.], batch size: 63, lr: 1.85e-02, grad_scale: 8.0 2023-03-07 19:15:51,721 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7260, 4.5881, 4.4252, 4.6823, 5.1194, 4.9563, 4.3854, 2.0523], device='cuda:1'), covar=tensor([0.0246, 0.0420, 0.0306, 0.0222, 0.0937, 0.0172, 0.0389, 0.2986], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0128, 0.0122, 0.0131, 0.0299, 0.0133, 0.0120, 0.0243], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-07 19:15:53,045 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20376.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:16:12,195 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6753, 2.8623, 3.9629, 3.0250, 3.8147, 4.7556, 4.4430, 3.3943], device='cuda:1'), covar=tensor([0.0364, 0.1825, 0.0736, 0.1247, 0.0797, 0.0555, 0.0443, 0.1306], device='cuda:1'), in_proj_covar=tensor([0.0207, 0.0212, 0.0204, 0.0190, 0.0216, 0.0214, 0.0174, 0.0201], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-03-07 19:16:12,222 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20388.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:16:48,763 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.120e+02 3.385e+02 4.079e+02 4.609e+02 1.123e+03, threshold=8.157e+02, percent-clipped=3.0 2023-03-07 19:17:00,091 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20418.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:17:02,924 INFO [train2.py:809] (1/4) Epoch 6, batch 500, loss[ctc_loss=0.1326, att_loss=0.2526, loss=0.2286, over 16272.00 frames. utt_duration=1515 frames, utt_pad_proportion=0.006527, over 43.00 utterances.], tot_loss[ctc_loss=0.1503, att_loss=0.2751, loss=0.2502, over 3017434.93 frames. utt_duration=1236 frames, utt_pad_proportion=0.05419, over 9776.69 utterances.], batch size: 43, lr: 1.85e-02, grad_scale: 8.0 2023-03-07 19:17:42,545 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20445.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:17:48,937 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20449.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:18:08,463 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.8310, 5.9599, 5.3646, 5.8734, 5.6302, 5.3089, 5.3851, 5.2763], device='cuda:1'), covar=tensor([0.1109, 0.0769, 0.0755, 0.0614, 0.0581, 0.1508, 0.2402, 0.2135], device='cuda:1'), in_proj_covar=tensor([0.0342, 0.0379, 0.0294, 0.0308, 0.0277, 0.0365, 0.0410, 0.0381], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-07 19:18:19,119 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-03-07 19:18:19,435 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20468.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:18:22,352 INFO [train2.py:809] (1/4) Epoch 6, batch 550, loss[ctc_loss=0.1388, att_loss=0.2615, loss=0.237, over 15962.00 frames. utt_duration=1559 frames, utt_pad_proportion=0.005771, over 41.00 utterances.], tot_loss[ctc_loss=0.1495, att_loss=0.2739, loss=0.249, over 3069789.77 frames. utt_duration=1247 frames, utt_pad_proportion=0.054, over 9857.95 utterances.], batch size: 41, lr: 1.84e-02, grad_scale: 8.0 2023-03-07 19:18:37,155 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20479.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:19:11,741 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.3764, 1.7558, 2.1806, 1.8250, 2.9722, 2.5277, 1.6455, 1.8908], device='cuda:1'), covar=tensor([0.0348, 0.2429, 0.1517, 0.1057, 0.0459, 0.0908, 0.2337, 0.1412], device='cuda:1'), in_proj_covar=tensor([0.0090, 0.0099, 0.0094, 0.0085, 0.0080, 0.0081, 0.0101, 0.0078], device='cuda:1'), out_proj_covar=tensor([4.0328e-05, 5.3581e-05, 5.1583e-05, 4.3037e-05, 3.8691e-05, 4.6301e-05, 5.2841e-05, 4.4490e-05], device='cuda:1') 2023-03-07 19:19:28,294 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.949e+02 3.263e+02 3.939e+02 5.001e+02 1.057e+03, threshold=7.878e+02, percent-clipped=3.0 2023-03-07 19:19:43,051 INFO [train2.py:809] (1/4) Epoch 6, batch 600, loss[ctc_loss=0.1067, att_loss=0.2257, loss=0.2019, over 14957.00 frames. utt_duration=1815 frames, utt_pad_proportion=0.02652, over 33.00 utterances.], tot_loss[ctc_loss=0.1474, att_loss=0.2725, loss=0.2475, over 3120461.35 frames. utt_duration=1271 frames, utt_pad_proportion=0.0471, over 9833.77 utterances.], batch size: 33, lr: 1.84e-02, grad_scale: 8.0 2023-03-07 19:19:45,691 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-03-07 19:20:47,292 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20560.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:20:58,193 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.8987, 3.7663, 3.9512, 2.7862, 3.8058, 3.6179, 3.6002, 2.4449], device='cuda:1'), covar=tensor([0.0151, 0.0156, 0.0125, 0.0876, 0.0097, 0.0309, 0.0269, 0.1412], device='cuda:1'), in_proj_covar=tensor([0.0055, 0.0066, 0.0054, 0.0100, 0.0058, 0.0073, 0.0081, 0.0103], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-07 19:21:03,189 INFO [train2.py:809] (1/4) Epoch 6, batch 650, loss[ctc_loss=0.1506, att_loss=0.2832, loss=0.2567, over 17023.00 frames. utt_duration=1337 frames, utt_pad_proportion=0.007642, over 51.00 utterances.], tot_loss[ctc_loss=0.1485, att_loss=0.2735, loss=0.2485, over 3159546.41 frames. utt_duration=1244 frames, utt_pad_proportion=0.05204, over 10174.45 utterances.], batch size: 51, lr: 1.84e-02, grad_scale: 8.0 2023-03-07 19:21:30,648 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.2922, 5.1022, 5.2009, 3.7033, 5.1058, 4.3790, 4.7970, 2.7116], device='cuda:1'), covar=tensor([0.0094, 0.0085, 0.0189, 0.0704, 0.0081, 0.0158, 0.0201, 0.1248], device='cuda:1'), in_proj_covar=tensor([0.0054, 0.0066, 0.0053, 0.0099, 0.0058, 0.0072, 0.0081, 0.0102], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-07 19:21:32,143 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20588.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:21:56,070 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.67 vs. limit=2.0 2023-03-07 19:22:08,669 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.746e+02 2.894e+02 3.503e+02 4.491e+02 1.231e+03, threshold=7.006e+02, percent-clipped=2.0 2023-03-07 19:22:23,251 INFO [train2.py:809] (1/4) Epoch 6, batch 700, loss[ctc_loss=0.1459, att_loss=0.2754, loss=0.2495, over 17399.00 frames. utt_duration=1106 frames, utt_pad_proportion=0.03299, over 63.00 utterances.], tot_loss[ctc_loss=0.148, att_loss=0.273, loss=0.248, over 3179577.71 frames. utt_duration=1250 frames, utt_pad_proportion=0.05301, over 10190.09 utterances.], batch size: 63, lr: 1.84e-02, grad_scale: 8.0 2023-03-07 19:22:59,968 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0918, 4.9725, 4.9535, 2.8577, 4.8078, 4.2865, 4.2023, 2.5201], device='cuda:1'), covar=tensor([0.0088, 0.0082, 0.0159, 0.1065, 0.0085, 0.0189, 0.0308, 0.1536], device='cuda:1'), in_proj_covar=tensor([0.0054, 0.0066, 0.0053, 0.0099, 0.0058, 0.0072, 0.0082, 0.0102], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-07 19:23:09,503 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20649.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:23:35,696 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.3567, 3.3113, 3.8573, 2.1820, 3.6611, 3.4936, 3.2497, 1.9635], device='cuda:1'), covar=tensor([0.1203, 0.0626, 0.0941, 0.5437, 0.0862, 0.2281, 0.0572, 0.8883], device='cuda:1'), in_proj_covar=tensor([0.0069, 0.0073, 0.0073, 0.0105, 0.0069, 0.0097, 0.0066, 0.0116], device='cuda:1'), out_proj_covar=tensor([5.3429e-05, 5.0739e-05, 5.4192e-05, 7.7123e-05, 5.1551e-05, 7.4608e-05, 4.9212e-05, 8.8723e-05], device='cuda:1') 2023-03-07 19:23:42,152 INFO [train2.py:809] (1/4) Epoch 6, batch 750, loss[ctc_loss=0.1984, att_loss=0.312, loss=0.2893, over 17050.00 frames. utt_duration=1313 frames, utt_pad_proportion=0.009115, over 52.00 utterances.], tot_loss[ctc_loss=0.1484, att_loss=0.2729, loss=0.248, over 3196650.12 frames. utt_duration=1269 frames, utt_pad_proportion=0.04979, over 10091.07 utterances.], batch size: 52, lr: 1.84e-02, grad_scale: 8.0 2023-03-07 19:23:51,702 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20676.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:24:47,042 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.294e+02 3.454e+02 4.265e+02 5.138e+02 1.250e+03, threshold=8.530e+02, percent-clipped=7.0 2023-03-07 19:25:01,489 INFO [train2.py:809] (1/4) Epoch 6, batch 800, loss[ctc_loss=0.1637, att_loss=0.283, loss=0.2591, over 17283.00 frames. utt_duration=1259 frames, utt_pad_proportion=0.01286, over 55.00 utterances.], tot_loss[ctc_loss=0.1488, att_loss=0.2736, loss=0.2486, over 3213165.53 frames. utt_duration=1256 frames, utt_pad_proportion=0.05322, over 10244.92 utterances.], batch size: 55, lr: 1.83e-02, grad_scale: 8.0 2023-03-07 19:25:07,725 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=20724.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:25:39,393 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20744.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:25:41,070 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20745.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:26:18,422 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20768.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:26:21,152 INFO [train2.py:809] (1/4) Epoch 6, batch 850, loss[ctc_loss=0.1754, att_loss=0.2951, loss=0.2711, over 16858.00 frames. utt_duration=682.7 frames, utt_pad_proportion=0.1423, over 99.00 utterances.], tot_loss[ctc_loss=0.1485, att_loss=0.2733, loss=0.2483, over 3226059.97 frames. utt_duration=1265 frames, utt_pad_proportion=0.05132, over 10210.98 utterances.], batch size: 99, lr: 1.83e-02, grad_scale: 8.0 2023-03-07 19:26:27,704 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20774.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:26:44,176 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.3365, 1.7319, 1.9522, 1.1289, 2.5344, 1.8382, 1.7801, 1.6362], device='cuda:1'), covar=tensor([0.0614, 0.3189, 0.2637, 0.2379, 0.1642, 0.1946, 0.2294, 0.2007], device='cuda:1'), in_proj_covar=tensor([0.0091, 0.0095, 0.0092, 0.0085, 0.0080, 0.0080, 0.0096, 0.0077], device='cuda:1'), out_proj_covar=tensor([3.9753e-05, 5.1994e-05, 5.0542e-05, 4.3622e-05, 3.8463e-05, 4.5723e-05, 5.1164e-05, 4.4079e-05], device='cuda:1') 2023-03-07 19:26:57,841 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=20793.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:27:04,733 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.86 vs. limit=2.0 2023-03-07 19:27:26,710 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.013e+02 3.098e+02 3.888e+02 4.985e+02 1.005e+03, threshold=7.777e+02, percent-clipped=2.0 2023-03-07 19:27:35,416 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=20816.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:27:41,728 INFO [train2.py:809] (1/4) Epoch 6, batch 900, loss[ctc_loss=0.1281, att_loss=0.2342, loss=0.213, over 15391.00 frames. utt_duration=1760 frames, utt_pad_proportion=0.009342, over 35.00 utterances.], tot_loss[ctc_loss=0.148, att_loss=0.2735, loss=0.2484, over 3242401.68 frames. utt_duration=1243 frames, utt_pad_proportion=0.05583, over 10444.68 utterances.], batch size: 35, lr: 1.83e-02, grad_scale: 8.0 2023-03-07 19:28:45,609 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20860.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:29:01,665 INFO [train2.py:809] (1/4) Epoch 6, batch 950, loss[ctc_loss=0.1486, att_loss=0.2529, loss=0.2321, over 15625.00 frames. utt_duration=1691 frames, utt_pad_proportion=0.01019, over 37.00 utterances.], tot_loss[ctc_loss=0.1477, att_loss=0.2728, loss=0.2478, over 3243891.38 frames. utt_duration=1249 frames, utt_pad_proportion=0.05695, over 10401.88 utterances.], batch size: 37, lr: 1.83e-02, grad_scale: 8.0 2023-03-07 19:30:01,041 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=20908.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:30:06,282 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.745e+02 3.096e+02 4.099e+02 5.928e+02 9.906e+02, threshold=8.198e+02, percent-clipped=5.0 2023-03-07 19:30:21,096 INFO [train2.py:809] (1/4) Epoch 6, batch 1000, loss[ctc_loss=0.1117, att_loss=0.2479, loss=0.2206, over 15953.00 frames. utt_duration=1558 frames, utt_pad_proportion=0.007026, over 41.00 utterances.], tot_loss[ctc_loss=0.1474, att_loss=0.2721, loss=0.2471, over 3235170.53 frames. utt_duration=1260 frames, utt_pad_proportion=0.05869, over 10280.08 utterances.], batch size: 41, lr: 1.83e-02, grad_scale: 8.0 2023-03-07 19:30:59,399 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20944.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:31:40,995 INFO [train2.py:809] (1/4) Epoch 6, batch 1050, loss[ctc_loss=0.1495, att_loss=0.2856, loss=0.2584, over 16972.00 frames. utt_duration=1359 frames, utt_pad_proportion=0.006998, over 50.00 utterances.], tot_loss[ctc_loss=0.147, att_loss=0.2725, loss=0.2474, over 3249190.20 frames. utt_duration=1270 frames, utt_pad_proportion=0.05263, over 10249.45 utterances.], batch size: 50, lr: 1.82e-02, grad_scale: 8.0 2023-03-07 19:31:45,396 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-03-07 19:32:47,826 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.831e+02 3.142e+02 3.611e+02 4.263e+02 9.074e+02, threshold=7.223e+02, percent-clipped=1.0 2023-03-07 19:33:02,163 INFO [train2.py:809] (1/4) Epoch 6, batch 1100, loss[ctc_loss=0.1564, att_loss=0.2957, loss=0.2678, over 17103.00 frames. utt_duration=1223 frames, utt_pad_proportion=0.01507, over 56.00 utterances.], tot_loss[ctc_loss=0.1474, att_loss=0.2733, loss=0.2482, over 3250988.15 frames. utt_duration=1255 frames, utt_pad_proportion=0.0549, over 10371.51 utterances.], batch size: 56, lr: 1.82e-02, grad_scale: 8.0 2023-03-07 19:33:12,126 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9560, 5.4377, 4.8227, 5.3716, 4.8662, 5.0727, 5.6132, 5.2836], device='cuda:1'), covar=tensor([0.0337, 0.0174, 0.0557, 0.0154, 0.0313, 0.0149, 0.0123, 0.0132], device='cuda:1'), in_proj_covar=tensor([0.0229, 0.0182, 0.0235, 0.0158, 0.0197, 0.0152, 0.0174, 0.0173], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-07 19:33:27,043 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-03-07 19:33:40,417 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21044.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:33:52,713 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21052.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:34:22,325 INFO [train2.py:809] (1/4) Epoch 6, batch 1150, loss[ctc_loss=0.183, att_loss=0.2833, loss=0.2633, over 16542.00 frames. utt_duration=1472 frames, utt_pad_proportion=0.006167, over 45.00 utterances.], tot_loss[ctc_loss=0.1489, att_loss=0.2739, loss=0.2489, over 3252930.32 frames. utt_duration=1231 frames, utt_pad_proportion=0.06046, over 10583.51 utterances.], batch size: 45, lr: 1.82e-02, grad_scale: 8.0 2023-03-07 19:34:28,552 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21074.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:34:57,174 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=21092.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:35:19,137 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2036, 5.1281, 4.9238, 2.6750, 1.9965, 2.8125, 5.0250, 3.8527], device='cuda:1'), covar=tensor([0.0537, 0.0168, 0.0266, 0.2898, 0.5849, 0.2389, 0.0177, 0.1847], device='cuda:1'), in_proj_covar=tensor([0.0283, 0.0177, 0.0204, 0.0179, 0.0358, 0.0333, 0.0186, 0.0326], device='cuda:1'), out_proj_covar=tensor([1.4387e-04, 7.6809e-05, 9.2539e-05, 8.1164e-05, 1.7315e-04, 1.5035e-04, 7.9397e-05, 1.5497e-04], device='cuda:1') 2023-03-07 19:35:27,185 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.084e+02 3.236e+02 4.040e+02 5.073e+02 1.159e+03, threshold=8.079e+02, percent-clipped=5.0 2023-03-07 19:35:30,942 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21113.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:35:39,370 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21118.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:35:41,978 INFO [train2.py:809] (1/4) Epoch 6, batch 1200, loss[ctc_loss=0.1223, att_loss=0.2316, loss=0.2098, over 15496.00 frames. utt_duration=1723 frames, utt_pad_proportion=0.008519, over 36.00 utterances.], tot_loss[ctc_loss=0.1484, att_loss=0.274, loss=0.2489, over 3261744.29 frames. utt_duration=1249 frames, utt_pad_proportion=0.05383, over 10459.38 utterances.], batch size: 36, lr: 1.82e-02, grad_scale: 8.0 2023-03-07 19:35:45,288 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=21122.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:35:47,150 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6425, 4.6808, 4.5799, 4.7112, 4.9926, 4.6675, 4.5624, 2.0914], device='cuda:1'), covar=tensor([0.0251, 0.0272, 0.0198, 0.0160, 0.1326, 0.0234, 0.0272, 0.2984], device='cuda:1'), in_proj_covar=tensor([0.0132, 0.0124, 0.0120, 0.0125, 0.0294, 0.0130, 0.0117, 0.0237], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-07 19:36:18,596 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7015, 5.0252, 4.9900, 4.8702, 5.1146, 5.0570, 4.7696, 4.5507], device='cuda:1'), covar=tensor([0.1066, 0.0499, 0.0196, 0.0470, 0.0290, 0.0296, 0.0315, 0.0324], device='cuda:1'), in_proj_covar=tensor([0.0390, 0.0227, 0.0166, 0.0206, 0.0260, 0.0291, 0.0222, 0.0250], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0003, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-07 19:37:02,263 INFO [train2.py:809] (1/4) Epoch 6, batch 1250, loss[ctc_loss=0.1613, att_loss=0.2842, loss=0.2596, over 17386.00 frames. utt_duration=1009 frames, utt_pad_proportion=0.04802, over 69.00 utterances.], tot_loss[ctc_loss=0.1481, att_loss=0.2742, loss=0.249, over 3269825.11 frames. utt_duration=1232 frames, utt_pad_proportion=0.05695, over 10627.92 utterances.], batch size: 69, lr: 1.82e-02, grad_scale: 8.0 2023-03-07 19:37:17,134 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21179.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:38:08,272 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.094e+02 3.230e+02 3.853e+02 4.858e+02 1.560e+03, threshold=7.707e+02, percent-clipped=4.0 2023-03-07 19:38:22,118 INFO [train2.py:809] (1/4) Epoch 6, batch 1300, loss[ctc_loss=0.1692, att_loss=0.3012, loss=0.2748, over 17022.00 frames. utt_duration=1337 frames, utt_pad_proportion=0.007642, over 51.00 utterances.], tot_loss[ctc_loss=0.1491, att_loss=0.2743, loss=0.2493, over 3255354.44 frames. utt_duration=1204 frames, utt_pad_proportion=0.06708, over 10832.68 utterances.], batch size: 51, lr: 1.81e-02, grad_scale: 8.0 2023-03-07 19:38:42,263 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-03-07 19:38:59,932 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21244.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:39:04,646 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9248, 5.3089, 4.8312, 5.4094, 4.7783, 5.0783, 5.5155, 5.2433], device='cuda:1'), covar=tensor([0.0417, 0.0409, 0.0682, 0.0162, 0.0456, 0.0185, 0.0196, 0.0169], device='cuda:1'), in_proj_covar=tensor([0.0235, 0.0187, 0.0241, 0.0160, 0.0199, 0.0154, 0.0177, 0.0176], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-07 19:39:42,382 INFO [train2.py:809] (1/4) Epoch 6, batch 1350, loss[ctc_loss=0.1215, att_loss=0.2479, loss=0.2226, over 15882.00 frames. utt_duration=1630 frames, utt_pad_proportion=0.009025, over 39.00 utterances.], tot_loss[ctc_loss=0.1484, att_loss=0.2743, loss=0.2491, over 3265720.23 frames. utt_duration=1216 frames, utt_pad_proportion=0.06138, over 10752.38 utterances.], batch size: 39, lr: 1.81e-02, grad_scale: 8.0 2023-03-07 19:40:17,767 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=21292.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:40:17,949 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6165, 5.0668, 4.8369, 5.0414, 5.1189, 4.7294, 3.8894, 5.0212], device='cuda:1'), covar=tensor([0.0074, 0.0115, 0.0077, 0.0071, 0.0072, 0.0087, 0.0422, 0.0161], device='cuda:1'), in_proj_covar=tensor([0.0057, 0.0055, 0.0061, 0.0042, 0.0042, 0.0052, 0.0074, 0.0072], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-03-07 19:40:48,697 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.207e+02 3.220e+02 3.890e+02 4.663e+02 1.481e+03, threshold=7.781e+02, percent-clipped=4.0 2023-03-07 19:40:50,633 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21312.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:40:56,889 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5080, 5.0251, 4.7436, 4.9739, 5.0878, 4.7957, 3.9130, 5.0717], device='cuda:1'), covar=tensor([0.0102, 0.0129, 0.0093, 0.0084, 0.0091, 0.0079, 0.0442, 0.0181], device='cuda:1'), in_proj_covar=tensor([0.0058, 0.0056, 0.0062, 0.0043, 0.0043, 0.0052, 0.0075, 0.0072], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-03-07 19:41:02,795 INFO [train2.py:809] (1/4) Epoch 6, batch 1400, loss[ctc_loss=0.1464, att_loss=0.2764, loss=0.2504, over 15947.00 frames. utt_duration=1557 frames, utt_pad_proportion=0.007508, over 41.00 utterances.], tot_loss[ctc_loss=0.1469, att_loss=0.2732, loss=0.248, over 3267292.53 frames. utt_duration=1245 frames, utt_pad_proportion=0.05567, over 10512.02 utterances.], batch size: 41, lr: 1.81e-02, grad_scale: 8.0 2023-03-07 19:41:03,310 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5006, 2.6305, 4.9577, 3.9776, 3.0673, 4.4827, 4.7713, 4.7880], device='cuda:1'), covar=tensor([0.0198, 0.1870, 0.0172, 0.1138, 0.2131, 0.0213, 0.0108, 0.0176], device='cuda:1'), in_proj_covar=tensor([0.0131, 0.0242, 0.0120, 0.0295, 0.0294, 0.0182, 0.0107, 0.0134], device='cuda:1'), out_proj_covar=tensor([1.1787e-04, 1.9704e-04, 1.0471e-04, 2.3977e-04, 2.4581e-04, 1.5711e-04, 9.6752e-05, 1.1863e-04], device='cuda:1') 2023-03-07 19:41:16,037 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21328.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:42:15,108 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-03-07 19:42:23,225 INFO [train2.py:809] (1/4) Epoch 6, batch 1450, loss[ctc_loss=0.1477, att_loss=0.2684, loss=0.2442, over 16177.00 frames. utt_duration=1580 frames, utt_pad_proportion=0.005848, over 41.00 utterances.], tot_loss[ctc_loss=0.1477, att_loss=0.2732, loss=0.2481, over 3252700.12 frames. utt_duration=1223 frames, utt_pad_proportion=0.06572, over 10648.84 utterances.], batch size: 41, lr: 1.81e-02, grad_scale: 8.0 2023-03-07 19:42:28,286 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21373.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:42:54,033 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21389.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:43:24,925 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21408.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:43:29,214 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.143e+02 3.096e+02 3.926e+02 4.783e+02 1.162e+03, threshold=7.851e+02, percent-clipped=3.0 2023-03-07 19:43:43,054 INFO [train2.py:809] (1/4) Epoch 6, batch 1500, loss[ctc_loss=0.1623, att_loss=0.2922, loss=0.2662, over 16780.00 frames. utt_duration=1400 frames, utt_pad_proportion=0.005726, over 48.00 utterances.], tot_loss[ctc_loss=0.149, att_loss=0.2745, loss=0.2494, over 3257478.26 frames. utt_duration=1186 frames, utt_pad_proportion=0.07381, over 11003.05 utterances.], batch size: 48, lr: 1.81e-02, grad_scale: 8.0 2023-03-07 19:44:31,963 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9677, 5.2816, 5.1950, 5.3544, 5.3963, 5.3166, 5.0197, 4.8223], device='cuda:1'), covar=tensor([0.1005, 0.0425, 0.0190, 0.0335, 0.0233, 0.0263, 0.0288, 0.0303], device='cuda:1'), in_proj_covar=tensor([0.0392, 0.0228, 0.0171, 0.0209, 0.0268, 0.0300, 0.0225, 0.0257], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-07 19:45:01,847 INFO [train2.py:809] (1/4) Epoch 6, batch 1550, loss[ctc_loss=0.1384, att_loss=0.2794, loss=0.2512, over 17355.00 frames. utt_duration=1104 frames, utt_pad_proportion=0.03278, over 63.00 utterances.], tot_loss[ctc_loss=0.1478, att_loss=0.273, loss=0.248, over 3257089.55 frames. utt_duration=1206 frames, utt_pad_proportion=0.06807, over 10815.03 utterances.], batch size: 63, lr: 1.80e-02, grad_scale: 8.0 2023-03-07 19:45:08,951 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21474.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:45:48,285 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([1.3629, 2.8798, 3.0071, 2.3446, 3.0864, 2.9653, 2.6604, 1.5907], device='cuda:1'), covar=tensor([0.1162, 0.0775, 0.4001, 0.5407, 0.2541, 0.2966, 0.0968, 0.8544], device='cuda:1'), in_proj_covar=tensor([0.0072, 0.0072, 0.0078, 0.0113, 0.0071, 0.0106, 0.0068, 0.0123], device='cuda:1'), out_proj_covar=tensor([5.7228e-05, 5.1888e-05, 5.8361e-05, 8.3015e-05, 5.4702e-05, 8.0728e-05, 5.1699e-05, 9.3511e-05], device='cuda:1') 2023-03-07 19:46:07,491 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.008e+02 3.352e+02 3.791e+02 4.457e+02 9.611e+02, threshold=7.583e+02, percent-clipped=1.0 2023-03-07 19:46:21,676 INFO [train2.py:809] (1/4) Epoch 6, batch 1600, loss[ctc_loss=0.1371, att_loss=0.2765, loss=0.2486, over 16608.00 frames. utt_duration=1415 frames, utt_pad_proportion=0.006201, over 47.00 utterances.], tot_loss[ctc_loss=0.1468, att_loss=0.2729, loss=0.2477, over 3260645.83 frames. utt_duration=1222 frames, utt_pad_proportion=0.06396, over 10688.32 utterances.], batch size: 47, lr: 1.80e-02, grad_scale: 8.0 2023-03-07 19:47:23,625 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.98 vs. limit=5.0 2023-03-07 19:47:41,024 INFO [train2.py:809] (1/4) Epoch 6, batch 1650, loss[ctc_loss=0.1258, att_loss=0.2448, loss=0.221, over 15638.00 frames. utt_duration=1692 frames, utt_pad_proportion=0.008566, over 37.00 utterances.], tot_loss[ctc_loss=0.1456, att_loss=0.2717, loss=0.2465, over 3255576.66 frames. utt_duration=1232 frames, utt_pad_proportion=0.06457, over 10585.21 utterances.], batch size: 37, lr: 1.80e-02, grad_scale: 8.0 2023-03-07 19:47:49,484 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.3599, 3.8032, 3.0326, 3.4468, 3.9045, 3.6080, 2.1408, 4.3993], device='cuda:1'), covar=tensor([0.1396, 0.0376, 0.1252, 0.0702, 0.0675, 0.0682, 0.1366, 0.0456], device='cuda:1'), in_proj_covar=tensor([0.0161, 0.0141, 0.0185, 0.0151, 0.0173, 0.0179, 0.0157, 0.0169], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-03-07 19:47:55,896 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9466, 4.5281, 4.5375, 2.2439, 2.0508, 2.6084, 4.1684, 3.4293], device='cuda:1'), covar=tensor([0.0551, 0.0158, 0.0172, 0.2982, 0.5601, 0.2514, 0.0291, 0.1885], device='cuda:1'), in_proj_covar=tensor([0.0288, 0.0175, 0.0206, 0.0177, 0.0356, 0.0329, 0.0189, 0.0322], device='cuda:1'), out_proj_covar=tensor([1.4486e-04, 7.5150e-05, 9.3308e-05, 7.9996e-05, 1.7076e-04, 1.4737e-04, 8.1067e-05, 1.5291e-04], device='cuda:1') 2023-03-07 19:48:13,311 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.99 vs. limit=5.0 2023-03-07 19:48:20,306 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21595.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:48:46,272 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5797, 2.1130, 4.9754, 3.8113, 3.1049, 4.4513, 4.8293, 4.7180], device='cuda:1'), covar=tensor([0.0203, 0.2100, 0.0179, 0.1385, 0.2179, 0.0247, 0.0108, 0.0220], device='cuda:1'), in_proj_covar=tensor([0.0130, 0.0238, 0.0120, 0.0296, 0.0293, 0.0181, 0.0106, 0.0134], device='cuda:1'), out_proj_covar=tensor([1.1607e-04, 1.9446e-04, 1.0633e-04, 2.3954e-04, 2.4546e-04, 1.5624e-04, 9.5608e-05, 1.1890e-04], device='cuda:1') 2023-03-07 19:48:47,374 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.750e+02 3.087e+02 4.072e+02 5.275e+02 1.721e+03, threshold=8.145e+02, percent-clipped=7.0 2023-03-07 19:48:50,773 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21613.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:49:01,645 INFO [train2.py:809] (1/4) Epoch 6, batch 1700, loss[ctc_loss=0.1073, att_loss=0.2326, loss=0.2075, over 14081.00 frames. utt_duration=1818 frames, utt_pad_proportion=0.04845, over 31.00 utterances.], tot_loss[ctc_loss=0.1457, att_loss=0.2725, loss=0.2471, over 3257690.60 frames. utt_duration=1228 frames, utt_pad_proportion=0.06541, over 10628.47 utterances.], batch size: 31, lr: 1.80e-02, grad_scale: 8.0 2023-03-07 19:50:00,443 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21656.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:50:05,652 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.98 vs. limit=2.0 2023-03-07 19:50:19,145 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21668.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:50:22,102 INFO [train2.py:809] (1/4) Epoch 6, batch 1750, loss[ctc_loss=0.1557, att_loss=0.2715, loss=0.2483, over 17128.00 frames. utt_duration=693.5 frames, utt_pad_proportion=0.1287, over 99.00 utterances.], tot_loss[ctc_loss=0.1459, att_loss=0.272, loss=0.2468, over 3255223.12 frames. utt_duration=1215 frames, utt_pad_proportion=0.06883, over 10731.25 utterances.], batch size: 99, lr: 1.80e-02, grad_scale: 8.0 2023-03-07 19:50:29,310 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21674.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:50:44,672 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21684.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:51:23,940 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21708.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:51:28,243 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.826e+02 3.408e+02 4.076e+02 5.243e+02 8.515e+02, threshold=8.152e+02, percent-clipped=2.0 2023-03-07 19:51:42,603 INFO [train2.py:809] (1/4) Epoch 6, batch 1800, loss[ctc_loss=0.1623, att_loss=0.2749, loss=0.2524, over 16416.00 frames. utt_duration=1494 frames, utt_pad_proportion=0.006049, over 44.00 utterances.], tot_loss[ctc_loss=0.1476, att_loss=0.2739, loss=0.2486, over 3272260.24 frames. utt_duration=1215 frames, utt_pad_proportion=0.06306, over 10784.61 utterances.], batch size: 44, lr: 1.79e-02, grad_scale: 8.0 2023-03-07 19:51:50,375 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4803, 4.8418, 4.6977, 5.0711, 4.7549, 4.6791, 3.6672, 4.6385], device='cuda:1'), covar=tensor([0.0093, 0.0128, 0.0096, 0.0064, 0.0112, 0.0077, 0.0559, 0.0247], device='cuda:1'), in_proj_covar=tensor([0.0056, 0.0055, 0.0062, 0.0044, 0.0043, 0.0052, 0.0076, 0.0072], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-03-07 19:52:39,971 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=21756.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:52:51,289 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21763.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:53:02,478 INFO [train2.py:809] (1/4) Epoch 6, batch 1850, loss[ctc_loss=0.1117, att_loss=0.2451, loss=0.2184, over 15860.00 frames. utt_duration=1628 frames, utt_pad_proportion=0.00967, over 39.00 utterances.], tot_loss[ctc_loss=0.1468, att_loss=0.2734, loss=0.2481, over 3271021.18 frames. utt_duration=1208 frames, utt_pad_proportion=0.06631, over 10845.03 utterances.], batch size: 39, lr: 1.79e-02, grad_scale: 8.0 2023-03-07 19:53:09,220 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21774.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:54:09,695 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.389e+02 3.296e+02 3.946e+02 5.007e+02 1.050e+03, threshold=7.891e+02, percent-clipped=3.0 2023-03-07 19:54:24,814 INFO [train2.py:809] (1/4) Epoch 6, batch 1900, loss[ctc_loss=0.1203, att_loss=0.2577, loss=0.2302, over 16275.00 frames. utt_duration=1515 frames, utt_pad_proportion=0.007661, over 43.00 utterances.], tot_loss[ctc_loss=0.1461, att_loss=0.2733, loss=0.2478, over 3275294.99 frames. utt_duration=1218 frames, utt_pad_proportion=0.06307, over 10767.16 utterances.], batch size: 43, lr: 1.79e-02, grad_scale: 8.0 2023-03-07 19:54:28,005 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=21822.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:54:31,354 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21824.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:55:07,382 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9314, 5.3970, 5.1011, 5.1580, 5.4054, 5.3456, 5.1796, 4.8297], device='cuda:1'), covar=tensor([0.1033, 0.0329, 0.0279, 0.0508, 0.0253, 0.0269, 0.0263, 0.0312], device='cuda:1'), in_proj_covar=tensor([0.0397, 0.0233, 0.0177, 0.0217, 0.0276, 0.0300, 0.0228, 0.0257], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-07 19:55:14,667 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21850.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:55:47,728 INFO [train2.py:809] (1/4) Epoch 6, batch 1950, loss[ctc_loss=0.1623, att_loss=0.2956, loss=0.2689, over 17467.00 frames. utt_duration=1014 frames, utt_pad_proportion=0.04412, over 69.00 utterances.], tot_loss[ctc_loss=0.1462, att_loss=0.273, loss=0.2477, over 3271535.03 frames. utt_duration=1233 frames, utt_pad_proportion=0.05924, over 10627.42 utterances.], batch size: 69, lr: 1.79e-02, grad_scale: 8.0 2023-03-07 19:55:51,433 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5934, 2.2707, 4.9458, 3.7117, 2.8355, 4.4077, 4.5328, 4.8134], device='cuda:1'), covar=tensor([0.0176, 0.1920, 0.0146, 0.1243, 0.2217, 0.0239, 0.0102, 0.0201], device='cuda:1'), in_proj_covar=tensor([0.0133, 0.0243, 0.0122, 0.0303, 0.0297, 0.0184, 0.0107, 0.0136], device='cuda:1'), out_proj_covar=tensor([1.1863e-04, 1.9873e-04, 1.0767e-04, 2.4535e-04, 2.4961e-04, 1.5985e-04, 9.6239e-05, 1.2156e-04], device='cuda:1') 2023-03-07 19:56:21,485 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-03-07 19:56:54,341 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.830e+02 3.406e+02 4.124e+02 5.166e+02 8.214e+02, threshold=8.249e+02, percent-clipped=4.0 2023-03-07 19:56:54,794 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21911.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:57:09,091 INFO [train2.py:809] (1/4) Epoch 6, batch 2000, loss[ctc_loss=0.122, att_loss=0.2527, loss=0.2265, over 16544.00 frames. utt_duration=1472 frames, utt_pad_proportion=0.006182, over 45.00 utterances.], tot_loss[ctc_loss=0.1463, att_loss=0.2731, loss=0.2477, over 3271749.46 frames. utt_duration=1242 frames, utt_pad_proportion=0.05823, over 10550.88 utterances.], batch size: 45, lr: 1.79e-02, grad_scale: 8.0 2023-03-07 19:57:27,922 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.6891, 5.8857, 5.3014, 5.8313, 5.5396, 5.2505, 5.2798, 5.1849], device='cuda:1'), covar=tensor([0.1152, 0.0874, 0.0837, 0.0642, 0.0632, 0.1357, 0.2450, 0.2052], device='cuda:1'), in_proj_covar=tensor([0.0348, 0.0390, 0.0306, 0.0321, 0.0282, 0.0372, 0.0432, 0.0392], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-03-07 19:57:59,923 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21951.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:58:27,680 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21968.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:58:29,178 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21969.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:58:30,580 INFO [train2.py:809] (1/4) Epoch 6, batch 2050, loss[ctc_loss=0.1116, att_loss=0.2431, loss=0.2168, over 16296.00 frames. utt_duration=1517 frames, utt_pad_proportion=0.005745, over 43.00 utterances.], tot_loss[ctc_loss=0.1453, att_loss=0.2726, loss=0.2472, over 3271640.22 frames. utt_duration=1243 frames, utt_pad_proportion=0.05622, over 10544.72 utterances.], batch size: 43, lr: 1.78e-02, grad_scale: 8.0 2023-03-07 19:58:30,977 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21970.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:58:31,723 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.64 vs. limit=2.0 2023-03-07 19:58:53,139 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21984.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:59:42,650 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.988e+02 3.148e+02 3.962e+02 4.824e+02 9.666e+02, threshold=7.923e+02, percent-clipped=1.0 2023-03-07 19:59:51,258 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=22016.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 19:59:57,611 INFO [train2.py:809] (1/4) Epoch 6, batch 2100, loss[ctc_loss=0.174, att_loss=0.2947, loss=0.2706, over 17132.00 frames. utt_duration=1225 frames, utt_pad_proportion=0.01375, over 56.00 utterances.], tot_loss[ctc_loss=0.1459, att_loss=0.2724, loss=0.2471, over 3261339.03 frames. utt_duration=1220 frames, utt_pad_proportion=0.06439, over 10707.69 utterances.], batch size: 56, lr: 1.78e-02, grad_scale: 8.0 2023-03-07 20:00:15,211 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22031.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:00:16,527 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=22032.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:01:20,714 INFO [train2.py:809] (1/4) Epoch 6, batch 2150, loss[ctc_loss=0.1506, att_loss=0.2716, loss=0.2474, over 17064.00 frames. utt_duration=1289 frames, utt_pad_proportion=0.008948, over 53.00 utterances.], tot_loss[ctc_loss=0.1465, att_loss=0.2727, loss=0.2475, over 3266791.35 frames. utt_duration=1213 frames, utt_pad_proportion=0.06463, over 10784.51 utterances.], batch size: 53, lr: 1.78e-02, grad_scale: 8.0 2023-03-07 20:01:30,774 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6017, 4.6238, 4.3361, 4.6300, 5.0502, 4.7259, 4.6519, 2.0255], device='cuda:1'), covar=tensor([0.0275, 0.0301, 0.0268, 0.0251, 0.1306, 0.0239, 0.0234, 0.3085], device='cuda:1'), in_proj_covar=tensor([0.0129, 0.0119, 0.0121, 0.0123, 0.0293, 0.0126, 0.0111, 0.0226], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-07 20:01:45,272 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4039, 4.8561, 4.6004, 4.2567, 2.1658, 4.6777, 2.6614, 2.0023], device='cuda:1'), covar=tensor([0.0274, 0.0116, 0.0702, 0.0284, 0.2612, 0.0182, 0.1572, 0.1779], device='cuda:1'), in_proj_covar=tensor([0.0109, 0.0099, 0.0249, 0.0110, 0.0223, 0.0100, 0.0222, 0.0206], device='cuda:1'), out_proj_covar=tensor([1.0831e-04, 9.8846e-05, 2.2071e-04, 1.0025e-04, 2.0283e-04, 9.6097e-05, 1.9631e-04, 1.8342e-04], device='cuda:1') 2023-03-07 20:02:28,365 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.159e+02 3.198e+02 3.917e+02 4.536e+02 8.271e+02, threshold=7.834e+02, percent-clipped=2.0 2023-03-07 20:02:28,599 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9224, 6.1298, 5.4917, 6.0537, 5.7573, 5.4362, 5.6541, 5.4754], device='cuda:1'), covar=tensor([0.1138, 0.0815, 0.0723, 0.0752, 0.0622, 0.1433, 0.2108, 0.2403], device='cuda:1'), in_proj_covar=tensor([0.0344, 0.0393, 0.0303, 0.0324, 0.0284, 0.0369, 0.0424, 0.0388], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-03-07 20:02:42,286 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22119.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:02:43,578 INFO [train2.py:809] (1/4) Epoch 6, batch 2200, loss[ctc_loss=0.1691, att_loss=0.2964, loss=0.2709, over 17373.00 frames. utt_duration=881.1 frames, utt_pad_proportion=0.07735, over 79.00 utterances.], tot_loss[ctc_loss=0.1458, att_loss=0.2722, loss=0.2469, over 3265266.63 frames. utt_duration=1222 frames, utt_pad_proportion=0.06314, over 10704.99 utterances.], batch size: 79, lr: 1.78e-02, grad_scale: 8.0 2023-03-07 20:02:47,697 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-03-07 20:02:59,614 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22130.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:03:39,971 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6021, 4.6772, 4.4947, 4.7608, 5.0704, 4.6519, 4.5526, 2.0346], device='cuda:1'), covar=tensor([0.0232, 0.0241, 0.0237, 0.0110, 0.1085, 0.0232, 0.0235, 0.2834], device='cuda:1'), in_proj_covar=tensor([0.0131, 0.0122, 0.0123, 0.0123, 0.0298, 0.0127, 0.0114, 0.0230], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-07 20:03:44,494 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3688, 4.6886, 4.6603, 4.8016, 4.7661, 4.5799, 3.5433, 4.5197], device='cuda:1'), covar=tensor([0.0110, 0.0152, 0.0087, 0.0095, 0.0114, 0.0088, 0.0548, 0.0266], device='cuda:1'), in_proj_covar=tensor([0.0056, 0.0055, 0.0062, 0.0043, 0.0043, 0.0051, 0.0075, 0.0072], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-03-07 20:04:03,530 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-03-07 20:04:05,341 INFO [train2.py:809] (1/4) Epoch 6, batch 2250, loss[ctc_loss=0.137, att_loss=0.2797, loss=0.2512, over 17363.00 frames. utt_duration=880.6 frames, utt_pad_proportion=0.07697, over 79.00 utterances.], tot_loss[ctc_loss=0.1465, att_loss=0.2732, loss=0.2479, over 3273098.04 frames. utt_duration=1210 frames, utt_pad_proportion=0.06422, over 10830.02 utterances.], batch size: 79, lr: 1.78e-02, grad_scale: 8.0 2023-03-07 20:04:14,619 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22176.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:04:38,858 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22191.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:05:00,116 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8283, 5.3022, 4.3559, 5.3909, 4.6360, 5.0638, 5.3217, 5.2052], device='cuda:1'), covar=tensor([0.0402, 0.0221, 0.0960, 0.0140, 0.0475, 0.0155, 0.0268, 0.0168], device='cuda:1'), in_proj_covar=tensor([0.0238, 0.0186, 0.0244, 0.0162, 0.0201, 0.0158, 0.0177, 0.0174], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-07 20:05:03,086 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22206.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:05:11,250 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.756e+02 2.999e+02 3.625e+02 4.758e+02 1.652e+03, threshold=7.250e+02, percent-clipped=4.0 2023-03-07 20:05:17,934 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22215.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:05:25,448 INFO [train2.py:809] (1/4) Epoch 6, batch 2300, loss[ctc_loss=0.116, att_loss=0.2516, loss=0.2245, over 15883.00 frames. utt_duration=1630 frames, utt_pad_proportion=0.008469, over 39.00 utterances.], tot_loss[ctc_loss=0.1474, att_loss=0.2737, loss=0.2484, over 3268684.37 frames. utt_duration=1190 frames, utt_pad_proportion=0.07032, over 11005.28 utterances.], batch size: 39, lr: 1.77e-02, grad_scale: 16.0 2023-03-07 20:05:52,494 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22237.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:05:52,520 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22237.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:06:16,001 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22251.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:06:44,807 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22269.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:06:44,927 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3782, 4.7173, 4.4837, 4.4127, 2.2874, 4.5330, 2.2767, 2.3227], device='cuda:1'), covar=tensor([0.0228, 0.0108, 0.0664, 0.0221, 0.2298, 0.0151, 0.1793, 0.1702], device='cuda:1'), in_proj_covar=tensor([0.0108, 0.0099, 0.0251, 0.0112, 0.0224, 0.0100, 0.0225, 0.0207], device='cuda:1'), out_proj_covar=tensor([1.0782e-04, 9.8855e-05, 2.2232e-04, 1.0236e-04, 2.0464e-04, 9.6391e-05, 1.9886e-04, 1.8499e-04], device='cuda:1') 2023-03-07 20:06:46,019 INFO [train2.py:809] (1/4) Epoch 6, batch 2350, loss[ctc_loss=0.1843, att_loss=0.2987, loss=0.2758, over 17049.00 frames. utt_duration=1313 frames, utt_pad_proportion=0.009013, over 52.00 utterances.], tot_loss[ctc_loss=0.1466, att_loss=0.2735, loss=0.2481, over 3271068.05 frames. utt_duration=1217 frames, utt_pad_proportion=0.06363, over 10767.29 utterances.], batch size: 52, lr: 1.77e-02, grad_scale: 16.0 2023-03-07 20:06:56,047 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22276.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:07:02,162 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1920, 5.1164, 5.2235, 3.7442, 5.0211, 4.4149, 4.5480, 2.8167], device='cuda:1'), covar=tensor([0.0097, 0.0093, 0.0109, 0.0698, 0.0088, 0.0164, 0.0255, 0.1389], device='cuda:1'), in_proj_covar=tensor([0.0055, 0.0067, 0.0055, 0.0101, 0.0060, 0.0077, 0.0082, 0.0102], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-07 20:07:10,454 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7578, 4.7917, 4.7785, 3.0844, 4.6093, 4.1936, 3.8283, 2.2544], device='cuda:1'), covar=tensor([0.0130, 0.0095, 0.0129, 0.0980, 0.0091, 0.0215, 0.0374, 0.1739], device='cuda:1'), in_proj_covar=tensor([0.0055, 0.0067, 0.0055, 0.0100, 0.0060, 0.0077, 0.0082, 0.0102], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-07 20:07:30,921 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22298.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:07:32,243 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=22299.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:07:36,848 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-03-07 20:07:52,172 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.236e+02 3.128e+02 3.825e+02 4.605e+02 9.373e+02, threshold=7.650e+02, percent-clipped=4.0 2023-03-07 20:08:01,633 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=22317.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:08:06,059 INFO [train2.py:809] (1/4) Epoch 6, batch 2400, loss[ctc_loss=0.1153, att_loss=0.2503, loss=0.2233, over 15994.00 frames. utt_duration=1601 frames, utt_pad_proportion=0.007548, over 40.00 utterances.], tot_loss[ctc_loss=0.1462, att_loss=0.2737, loss=0.2482, over 3280705.93 frames. utt_duration=1233 frames, utt_pad_proportion=0.05741, over 10659.33 utterances.], batch size: 40, lr: 1.77e-02, grad_scale: 16.0 2023-03-07 20:08:15,576 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22326.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:09:26,764 INFO [train2.py:809] (1/4) Epoch 6, batch 2450, loss[ctc_loss=0.09899, att_loss=0.2342, loss=0.2071, over 14539.00 frames. utt_duration=1819 frames, utt_pad_proportion=0.02995, over 32.00 utterances.], tot_loss[ctc_loss=0.146, att_loss=0.2735, loss=0.248, over 3270958.38 frames. utt_duration=1230 frames, utt_pad_proportion=0.05993, over 10653.23 utterances.], batch size: 32, lr: 1.77e-02, grad_scale: 16.0 2023-03-07 20:09:51,323 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22385.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:10:04,913 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 2023-03-07 20:10:05,666 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7435, 5.0472, 5.0337, 5.0282, 5.1660, 5.1679, 4.8678, 4.7788], device='cuda:1'), covar=tensor([0.1008, 0.0466, 0.0200, 0.0387, 0.0276, 0.0261, 0.0258, 0.0280], device='cuda:1'), in_proj_covar=tensor([0.0400, 0.0230, 0.0171, 0.0216, 0.0274, 0.0297, 0.0229, 0.0258], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-07 20:10:34,025 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.011e+02 3.249e+02 3.944e+02 4.920e+02 7.510e+02, threshold=7.889e+02, percent-clipped=0.0 2023-03-07 20:10:37,613 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6780, 4.8776, 4.2690, 4.8451, 4.4493, 4.2697, 4.4415, 4.2428], device='cuda:1'), covar=tensor([0.1110, 0.0987, 0.0905, 0.0793, 0.1038, 0.1370, 0.2047, 0.2120], device='cuda:1'), in_proj_covar=tensor([0.0349, 0.0399, 0.0306, 0.0327, 0.0287, 0.0369, 0.0436, 0.0389], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-03-07 20:10:47,147 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22419.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:10:48,451 INFO [train2.py:809] (1/4) Epoch 6, batch 2500, loss[ctc_loss=0.1804, att_loss=0.3061, loss=0.2809, over 17303.00 frames. utt_duration=1260 frames, utt_pad_proportion=0.01117, over 55.00 utterances.], tot_loss[ctc_loss=0.1458, att_loss=0.2734, loss=0.2479, over 3271020.33 frames. utt_duration=1231 frames, utt_pad_proportion=0.06003, over 10641.84 utterances.], batch size: 55, lr: 1.77e-02, grad_scale: 16.0 2023-03-07 20:11:02,769 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5185, 4.3804, 4.3418, 4.3625, 4.8974, 4.6277, 4.3145, 2.1280], device='cuda:1'), covar=tensor([0.0275, 0.0376, 0.0363, 0.0189, 0.0921, 0.0207, 0.0297, 0.2858], device='cuda:1'), in_proj_covar=tensor([0.0131, 0.0124, 0.0122, 0.0125, 0.0295, 0.0127, 0.0115, 0.0231], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-07 20:11:30,444 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22446.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:12:04,145 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=22467.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:12:08,536 INFO [train2.py:809] (1/4) Epoch 6, batch 2550, loss[ctc_loss=0.1338, att_loss=0.2771, loss=0.2484, over 16900.00 frames. utt_duration=1381 frames, utt_pad_proportion=0.005921, over 49.00 utterances.], tot_loss[ctc_loss=0.1465, att_loss=0.2736, loss=0.2482, over 3266146.07 frames. utt_duration=1217 frames, utt_pad_proportion=0.06406, over 10749.64 utterances.], batch size: 49, lr: 1.76e-02, grad_scale: 16.0 2023-03-07 20:12:34,602 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22486.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:12:43,681 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.10 vs. limit=5.0 2023-03-07 20:13:07,425 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22506.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:13:14,006 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.91 vs. limit=2.0 2023-03-07 20:13:14,818 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.039e+02 3.216e+02 3.975e+02 4.594e+02 1.058e+03, threshold=7.951e+02, percent-clipped=4.0 2023-03-07 20:13:28,856 INFO [train2.py:809] (1/4) Epoch 6, batch 2600, loss[ctc_loss=0.1399, att_loss=0.2776, loss=0.25, over 16482.00 frames. utt_duration=1435 frames, utt_pad_proportion=0.00586, over 46.00 utterances.], tot_loss[ctc_loss=0.1446, att_loss=0.2723, loss=0.2467, over 3265448.16 frames. utt_duration=1246 frames, utt_pad_proportion=0.0578, over 10499.20 utterances.], batch size: 46, lr: 1.76e-02, grad_scale: 16.0 2023-03-07 20:13:47,874 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22532.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:14:19,823 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-03-07 20:14:25,001 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=22554.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:14:51,057 INFO [train2.py:809] (1/4) Epoch 6, batch 2650, loss[ctc_loss=0.1246, att_loss=0.272, loss=0.2426, over 16882.00 frames. utt_duration=1380 frames, utt_pad_proportion=0.00737, over 49.00 utterances.], tot_loss[ctc_loss=0.1426, att_loss=0.2702, loss=0.2447, over 3259773.34 frames. utt_duration=1268 frames, utt_pad_proportion=0.05407, over 10291.72 utterances.], batch size: 49, lr: 1.76e-02, grad_scale: 16.0 2023-03-07 20:14:52,733 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22571.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:15:28,365 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22593.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:15:59,451 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.951e+02 3.146e+02 3.715e+02 4.789e+02 1.567e+03, threshold=7.430e+02, percent-clipped=3.0 2023-03-07 20:16:11,935 INFO [train2.py:809] (1/4) Epoch 6, batch 2700, loss[ctc_loss=0.1646, att_loss=0.2972, loss=0.2707, over 17016.00 frames. utt_duration=1336 frames, utt_pad_proportion=0.008035, over 51.00 utterances.], tot_loss[ctc_loss=0.1418, att_loss=0.269, loss=0.2436, over 3261851.28 frames. utt_duration=1296 frames, utt_pad_proportion=0.04725, over 10075.30 utterances.], batch size: 51, lr: 1.76e-02, grad_scale: 8.0 2023-03-07 20:16:14,513 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.93 vs. limit=5.0 2023-03-07 20:16:21,394 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22626.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:16:41,841 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22639.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:16:48,802 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3753, 2.6588, 3.6209, 2.7216, 3.7347, 4.5902, 4.3804, 3.2822], device='cuda:1'), covar=tensor([0.0444, 0.1646, 0.0913, 0.1369, 0.0798, 0.0395, 0.0445, 0.1191], device='cuda:1'), in_proj_covar=tensor([0.0206, 0.0206, 0.0210, 0.0190, 0.0214, 0.0219, 0.0171, 0.0198], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-03-07 20:17:02,444 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9957, 4.9422, 4.9092, 3.0154, 4.7142, 4.1834, 4.4097, 2.4911], device='cuda:1'), covar=tensor([0.0133, 0.0096, 0.0190, 0.0965, 0.0074, 0.0195, 0.0258, 0.1473], device='cuda:1'), in_proj_covar=tensor([0.0054, 0.0066, 0.0056, 0.0098, 0.0059, 0.0076, 0.0081, 0.0101], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-07 20:17:32,381 INFO [train2.py:809] (1/4) Epoch 6, batch 2750, loss[ctc_loss=0.1409, att_loss=0.2749, loss=0.2481, over 16555.00 frames. utt_duration=1473 frames, utt_pad_proportion=0.005462, over 45.00 utterances.], tot_loss[ctc_loss=0.1444, att_loss=0.2709, loss=0.2456, over 3262635.78 frames. utt_duration=1258 frames, utt_pad_proportion=0.05634, over 10386.32 utterances.], batch size: 45, lr: 1.76e-02, grad_scale: 8.0 2023-03-07 20:17:38,673 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=22674.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:17:42,035 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7219, 1.2030, 1.9147, 1.7055, 3.5688, 1.5873, 1.6651, 1.6267], device='cuda:1'), covar=tensor([0.0953, 0.4070, 0.3083, 0.2285, 0.0520, 0.2667, 0.2503, 0.2272], device='cuda:1'), in_proj_covar=tensor([0.0091, 0.0099, 0.0097, 0.0086, 0.0082, 0.0081, 0.0093, 0.0079], device='cuda:1'), out_proj_covar=tensor([4.1071e-05, 5.5241e-05, 5.3761e-05, 4.5201e-05, 3.9869e-05, 4.7320e-05, 5.2085e-05, 4.6527e-05], device='cuda:1') 2023-03-07 20:18:15,964 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.8882, 6.1113, 5.5207, 5.9856, 5.7484, 5.4651, 5.5640, 5.4127], device='cuda:1'), covar=tensor([0.1069, 0.0820, 0.0734, 0.0649, 0.0708, 0.1199, 0.2080, 0.1974], device='cuda:1'), in_proj_covar=tensor([0.0352, 0.0400, 0.0312, 0.0326, 0.0291, 0.0373, 0.0441, 0.0390], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-03-07 20:18:21,022 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22700.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 20:18:40,004 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.820e+02 3.125e+02 3.816e+02 4.658e+02 9.758e+02, threshold=7.632e+02, percent-clipped=3.0 2023-03-07 20:18:52,615 INFO [train2.py:809] (1/4) Epoch 6, batch 2800, loss[ctc_loss=0.1554, att_loss=0.2886, loss=0.262, over 16610.00 frames. utt_duration=1415 frames, utt_pad_proportion=0.006231, over 47.00 utterances.], tot_loss[ctc_loss=0.1439, att_loss=0.2711, loss=0.2456, over 3266672.37 frames. utt_duration=1264 frames, utt_pad_proportion=0.05319, over 10351.55 utterances.], batch size: 47, lr: 1.76e-02, grad_scale: 8.0 2023-03-07 20:19:26,386 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22741.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:20:11,902 INFO [train2.py:809] (1/4) Epoch 6, batch 2850, loss[ctc_loss=0.1418, att_loss=0.2723, loss=0.2462, over 16463.00 frames. utt_duration=1433 frames, utt_pad_proportion=0.006915, over 46.00 utterances.], tot_loss[ctc_loss=0.1437, att_loss=0.2704, loss=0.245, over 3262680.45 frames. utt_duration=1274 frames, utt_pad_proportion=0.05105, over 10258.50 utterances.], batch size: 46, lr: 1.75e-02, grad_scale: 8.0 2023-03-07 20:20:38,178 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22786.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:20:51,869 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22794.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:21:08,618 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.5121, 3.3664, 2.8220, 3.1817, 3.5194, 3.2751, 2.4663, 3.8140], device='cuda:1'), covar=tensor([0.1027, 0.0357, 0.0968, 0.0563, 0.0564, 0.0612, 0.0957, 0.0420], device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0141, 0.0187, 0.0154, 0.0176, 0.0183, 0.0157, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-03-07 20:21:20,768 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.881e+02 3.039e+02 3.742e+02 4.818e+02 9.718e+02, threshold=7.483e+02, percent-clipped=5.0 2023-03-07 20:21:33,091 INFO [train2.py:809] (1/4) Epoch 6, batch 2900, loss[ctc_loss=0.09047, att_loss=0.2245, loss=0.1977, over 15500.00 frames. utt_duration=1724 frames, utt_pad_proportion=0.008167, over 36.00 utterances.], tot_loss[ctc_loss=0.1433, att_loss=0.2705, loss=0.245, over 3264211.39 frames. utt_duration=1256 frames, utt_pad_proportion=0.05431, over 10411.51 utterances.], batch size: 36, lr: 1.75e-02, grad_scale: 8.0 2023-03-07 20:21:41,141 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22825.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:21:52,497 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22832.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:21:55,394 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=22834.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:22:30,791 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22855.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:22:40,686 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.20 vs. limit=5.0 2023-03-07 20:22:53,779 INFO [train2.py:809] (1/4) Epoch 6, batch 2950, loss[ctc_loss=0.1576, att_loss=0.2929, loss=0.2658, over 17314.00 frames. utt_duration=1261 frames, utt_pad_proportion=0.01114, over 55.00 utterances.], tot_loss[ctc_loss=0.1424, att_loss=0.2699, loss=0.2444, over 3261841.85 frames. utt_duration=1268 frames, utt_pad_proportion=0.05175, over 10302.02 utterances.], batch size: 55, lr: 1.75e-02, grad_scale: 8.0 2023-03-07 20:22:55,596 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22871.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:23:05,877 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([1.6299, 3.1479, 3.3208, 2.6901, 3.2930, 3.3787, 3.2684, 1.8042], device='cuda:1'), covar=tensor([0.1846, 0.1101, 0.2156, 0.5963, 0.1204, 0.2653, 0.0641, 1.0438], device='cuda:1'), in_proj_covar=tensor([0.0071, 0.0076, 0.0083, 0.0115, 0.0070, 0.0107, 0.0068, 0.0125], device='cuda:1'), out_proj_covar=tensor([5.8482e-05, 5.6018e-05, 6.4309e-05, 8.6129e-05, 5.6251e-05, 8.3060e-05, 5.2421e-05, 9.6307e-05], device='cuda:1') 2023-03-07 20:23:10,175 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=22880.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:23:20,697 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22886.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:23:32,150 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22893.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:24:02,046 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.974e+02 3.285e+02 3.915e+02 4.917e+02 1.190e+03, threshold=7.831e+02, percent-clipped=3.0 2023-03-07 20:24:12,915 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=22919.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:24:14,354 INFO [train2.py:809] (1/4) Epoch 6, batch 3000, loss[ctc_loss=0.172, att_loss=0.2897, loss=0.2662, over 16825.00 frames. utt_duration=681.3 frames, utt_pad_proportion=0.1442, over 99.00 utterances.], tot_loss[ctc_loss=0.1428, att_loss=0.2705, loss=0.2449, over 3271005.72 frames. utt_duration=1252 frames, utt_pad_proportion=0.05345, over 10465.79 utterances.], batch size: 99, lr: 1.75e-02, grad_scale: 8.0 2023-03-07 20:24:14,354 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-07 20:24:28,229 INFO [train2.py:843] (1/4) Epoch 6, validation: ctc_loss=0.06806, att_loss=0.2473, loss=0.2115, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-07 20:24:28,230 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-07 20:25:01,614 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=22941.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:25:18,813 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-03-07 20:25:47,896 INFO [train2.py:809] (1/4) Epoch 6, batch 3050, loss[ctc_loss=0.1535, att_loss=0.291, loss=0.2635, over 17371.00 frames. utt_duration=1104 frames, utt_pad_proportion=0.03372, over 63.00 utterances.], tot_loss[ctc_loss=0.1425, att_loss=0.2699, loss=0.2444, over 3253421.94 frames. utt_duration=1280 frames, utt_pad_proportion=0.05038, over 10179.21 utterances.], batch size: 63, lr: 1.75e-02, grad_scale: 8.0 2023-03-07 20:26:27,914 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22995.0, num_to_drop=1, layers_to_drop={2} 2023-03-07 20:26:42,787 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4939, 2.6118, 4.8946, 3.9145, 2.8555, 4.2129, 4.6125, 4.6938], device='cuda:1'), covar=tensor([0.0201, 0.1843, 0.0147, 0.1091, 0.2273, 0.0295, 0.0115, 0.0229], device='cuda:1'), in_proj_covar=tensor([0.0140, 0.0241, 0.0124, 0.0303, 0.0296, 0.0185, 0.0105, 0.0140], device='cuda:1'), out_proj_covar=tensor([1.2577e-04, 1.9750e-04, 1.1068e-04, 2.4632e-04, 2.4978e-04, 1.6103e-04, 9.4728e-05, 1.2590e-04], device='cuda:1') 2023-03-07 20:26:55,429 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.122e+02 3.380e+02 4.169e+02 5.152e+02 1.227e+03, threshold=8.338e+02, percent-clipped=8.0 2023-03-07 20:27:00,321 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2361, 1.4378, 2.1244, 1.5487, 2.9416, 2.4831, 1.5649, 1.1005], device='cuda:1'), covar=tensor([0.0334, 0.3011, 0.2585, 0.3290, 0.0524, 0.1218, 0.2368, 0.2274], device='cuda:1'), in_proj_covar=tensor([0.0083, 0.0092, 0.0093, 0.0081, 0.0077, 0.0078, 0.0090, 0.0076], device='cuda:1'), out_proj_covar=tensor([3.8586e-05, 5.1887e-05, 5.2019e-05, 4.3206e-05, 3.8253e-05, 4.5489e-05, 5.0286e-05, 4.4551e-05], device='cuda:1') 2023-03-07 20:27:07,732 INFO [train2.py:809] (1/4) Epoch 6, batch 3100, loss[ctc_loss=0.1558, att_loss=0.2812, loss=0.2561, over 17249.00 frames. utt_duration=874.7 frames, utt_pad_proportion=0.08218, over 79.00 utterances.], tot_loss[ctc_loss=0.1425, att_loss=0.2695, loss=0.2441, over 3257161.38 frames. utt_duration=1285 frames, utt_pad_proportion=0.04863, over 10150.14 utterances.], batch size: 79, lr: 1.74e-02, grad_scale: 8.0 2023-03-07 20:27:14,235 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.5635, 5.8316, 5.2362, 5.7616, 5.4077, 5.1360, 5.1303, 5.0907], device='cuda:1'), covar=tensor([0.1374, 0.0832, 0.0883, 0.0640, 0.0909, 0.1266, 0.2198, 0.2062], device='cuda:1'), in_proj_covar=tensor([0.0363, 0.0397, 0.0307, 0.0324, 0.0295, 0.0372, 0.0434, 0.0390], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-03-07 20:27:42,238 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23041.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:28:28,022 INFO [train2.py:809] (1/4) Epoch 6, batch 3150, loss[ctc_loss=0.1238, att_loss=0.2715, loss=0.2419, over 16893.00 frames. utt_duration=691 frames, utt_pad_proportion=0.1341, over 98.00 utterances.], tot_loss[ctc_loss=0.1432, att_loss=0.2707, loss=0.2452, over 3268455.64 frames. utt_duration=1256 frames, utt_pad_proportion=0.0509, over 10418.49 utterances.], batch size: 98, lr: 1.74e-02, grad_scale: 8.0 2023-03-07 20:28:59,006 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=23089.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:29:30,192 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23108.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:29:36,089 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.237e+02 3.296e+02 3.877e+02 4.841e+02 1.237e+03, threshold=7.753e+02, percent-clipped=8.0 2023-03-07 20:29:48,529 INFO [train2.py:809] (1/4) Epoch 6, batch 3200, loss[ctc_loss=0.198, att_loss=0.3126, loss=0.2897, over 16757.00 frames. utt_duration=1398 frames, utt_pad_proportion=0.00725, over 48.00 utterances.], tot_loss[ctc_loss=0.1427, att_loss=0.2702, loss=0.2447, over 3274187.47 frames. utt_duration=1269 frames, utt_pad_proportion=0.04651, over 10333.26 utterances.], batch size: 48, lr: 1.74e-02, grad_scale: 8.0 2023-03-07 20:30:36,879 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23150.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:31:07,712 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23169.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:31:08,932 INFO [train2.py:809] (1/4) Epoch 6, batch 3250, loss[ctc_loss=0.1642, att_loss=0.2898, loss=0.2647, over 17415.00 frames. utt_duration=1011 frames, utt_pad_proportion=0.04722, over 69.00 utterances.], tot_loss[ctc_loss=0.1424, att_loss=0.2701, loss=0.2445, over 3273922.79 frames. utt_duration=1257 frames, utt_pad_proportion=0.05177, over 10430.65 utterances.], batch size: 69, lr: 1.74e-02, grad_scale: 8.0 2023-03-07 20:31:20,309 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-03-07 20:31:26,726 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23181.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:32:18,017 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.115e+02 3.273e+02 3.918e+02 4.726e+02 1.280e+03, threshold=7.837e+02, percent-clipped=1.0 2023-03-07 20:32:26,189 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23217.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:32:30,550 INFO [train2.py:809] (1/4) Epoch 6, batch 3300, loss[ctc_loss=0.1526, att_loss=0.2825, loss=0.2565, over 16468.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.006433, over 46.00 utterances.], tot_loss[ctc_loss=0.1431, att_loss=0.2715, loss=0.2458, over 3279871.95 frames. utt_duration=1238 frames, utt_pad_proportion=0.05472, over 10611.94 utterances.], batch size: 46, lr: 1.74e-02, grad_scale: 8.0 2023-03-07 20:32:45,792 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23229.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:32:58,781 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8973, 4.8082, 4.8255, 2.6085, 4.6986, 4.2563, 4.2048, 2.3299], device='cuda:1'), covar=tensor([0.0107, 0.0077, 0.0195, 0.0955, 0.0077, 0.0153, 0.0249, 0.1476], device='cuda:1'), in_proj_covar=tensor([0.0053, 0.0067, 0.0057, 0.0098, 0.0059, 0.0075, 0.0081, 0.0100], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-07 20:33:28,437 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.5358, 3.8646, 3.0354, 3.2870, 3.9663, 3.6863, 2.3520, 4.4228], device='cuda:1'), covar=tensor([0.1089, 0.0398, 0.1019, 0.0637, 0.0473, 0.0560, 0.1071, 0.0372], device='cuda:1'), in_proj_covar=tensor([0.0163, 0.0142, 0.0188, 0.0152, 0.0175, 0.0186, 0.0159, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-03-07 20:33:51,768 INFO [train2.py:809] (1/4) Epoch 6, batch 3350, loss[ctc_loss=0.126, att_loss=0.2724, loss=0.2431, over 16624.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.005364, over 47.00 utterances.], tot_loss[ctc_loss=0.1432, att_loss=0.2714, loss=0.2458, over 3284461.71 frames. utt_duration=1238 frames, utt_pad_proportion=0.05336, over 10622.76 utterances.], batch size: 47, lr: 1.74e-02, grad_scale: 8.0 2023-03-07 20:34:05,051 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23278.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:34:24,923 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23290.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:34:32,808 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23295.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:34:33,712 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-03-07 20:34:34,258 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9735, 4.8010, 4.8281, 2.6857, 4.7046, 4.2907, 4.0890, 2.7368], device='cuda:1'), covar=tensor([0.0117, 0.0082, 0.0211, 0.0954, 0.0094, 0.0162, 0.0299, 0.1260], device='cuda:1'), in_proj_covar=tensor([0.0054, 0.0069, 0.0057, 0.0099, 0.0060, 0.0077, 0.0083, 0.0101], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-07 20:34:59,027 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.052e+02 3.226e+02 3.955e+02 4.943e+02 9.794e+02, threshold=7.910e+02, percent-clipped=2.0 2023-03-07 20:35:01,541 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=2.01 vs. limit=2.0 2023-03-07 20:35:11,507 INFO [train2.py:809] (1/4) Epoch 6, batch 3400, loss[ctc_loss=0.1106, att_loss=0.2442, loss=0.2175, over 15875.00 frames. utt_duration=1630 frames, utt_pad_proportion=0.009814, over 39.00 utterances.], tot_loss[ctc_loss=0.1434, att_loss=0.2714, loss=0.2458, over 3285393.78 frames. utt_duration=1253 frames, utt_pad_proportion=0.05032, over 10499.54 utterances.], batch size: 39, lr: 1.73e-02, grad_scale: 8.0 2023-03-07 20:35:49,199 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=23343.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:36:31,953 INFO [train2.py:809] (1/4) Epoch 6, batch 3450, loss[ctc_loss=0.1511, att_loss=0.2905, loss=0.2626, over 16958.00 frames. utt_duration=1358 frames, utt_pad_proportion=0.008123, over 50.00 utterances.], tot_loss[ctc_loss=0.1428, att_loss=0.2711, loss=0.2454, over 3284001.65 frames. utt_duration=1259 frames, utt_pad_proportion=0.04862, over 10449.80 utterances.], batch size: 50, lr: 1.73e-02, grad_scale: 8.0 2023-03-07 20:37:23,698 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.0034, 3.5276, 3.4610, 2.9349, 3.3029, 3.2401, 3.2277, 2.1325], device='cuda:1'), covar=tensor([0.1800, 0.0737, 0.1710, 0.3527, 0.0845, 0.6784, 0.0827, 0.6558], device='cuda:1'), in_proj_covar=tensor([0.0078, 0.0078, 0.0085, 0.0124, 0.0075, 0.0112, 0.0071, 0.0129], device='cuda:1'), out_proj_covar=tensor([6.2699e-05, 5.7747e-05, 6.6690e-05, 9.2966e-05, 5.9300e-05, 8.7793e-05, 5.4883e-05, 1.0069e-04], device='cuda:1') 2023-03-07 20:37:39,558 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.288e+02 3.391e+02 4.040e+02 5.091e+02 1.147e+03, threshold=8.079e+02, percent-clipped=5.0 2023-03-07 20:37:52,188 INFO [train2.py:809] (1/4) Epoch 6, batch 3500, loss[ctc_loss=0.1202, att_loss=0.2542, loss=0.2274, over 15946.00 frames. utt_duration=1557 frames, utt_pad_proportion=0.00686, over 41.00 utterances.], tot_loss[ctc_loss=0.1432, att_loss=0.271, loss=0.2454, over 3279835.79 frames. utt_duration=1264 frames, utt_pad_proportion=0.04812, over 10395.13 utterances.], batch size: 41, lr: 1.73e-02, grad_scale: 8.0 2023-03-07 20:38:03,680 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-03-07 20:38:40,696 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23450.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:39:03,021 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23464.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:39:12,997 INFO [train2.py:809] (1/4) Epoch 6, batch 3550, loss[ctc_loss=0.1111, att_loss=0.2494, loss=0.2217, over 15950.00 frames. utt_duration=1557 frames, utt_pad_proportion=0.006751, over 41.00 utterances.], tot_loss[ctc_loss=0.1426, att_loss=0.2707, loss=0.2451, over 3274332.09 frames. utt_duration=1274 frames, utt_pad_proportion=0.04774, over 10294.00 utterances.], batch size: 41, lr: 1.73e-02, grad_scale: 8.0 2023-03-07 20:39:30,631 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23481.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:39:35,306 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6083, 5.0862, 4.7771, 5.0845, 5.2558, 4.7350, 3.7811, 5.0818], device='cuda:1'), covar=tensor([0.0080, 0.0084, 0.0075, 0.0062, 0.0058, 0.0070, 0.0444, 0.0122], device='cuda:1'), in_proj_covar=tensor([0.0057, 0.0053, 0.0061, 0.0041, 0.0042, 0.0052, 0.0073, 0.0071], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002], device='cuda:1') 2023-03-07 20:39:42,272 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23488.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:39:57,397 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=23498.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:40:19,336 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.856e+02 3.330e+02 3.992e+02 4.756e+02 1.054e+03, threshold=7.983e+02, percent-clipped=2.0 2023-03-07 20:40:31,491 INFO [train2.py:809] (1/4) Epoch 6, batch 3600, loss[ctc_loss=0.1285, att_loss=0.2571, loss=0.2314, over 16002.00 frames. utt_duration=1601 frames, utt_pad_proportion=0.007776, over 40.00 utterances.], tot_loss[ctc_loss=0.1432, att_loss=0.271, loss=0.2454, over 3271132.63 frames. utt_duration=1272 frames, utt_pad_proportion=0.04905, over 10295.70 utterances.], batch size: 40, lr: 1.73e-02, grad_scale: 8.0 2023-03-07 20:40:46,283 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=23529.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:40:49,611 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-03-07 20:41:18,595 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23549.0, num_to_drop=1, layers_to_drop={3} 2023-03-07 20:41:51,579 INFO [train2.py:809] (1/4) Epoch 6, batch 3650, loss[ctc_loss=0.1511, att_loss=0.2809, loss=0.255, over 17047.00 frames. utt_duration=1338 frames, utt_pad_proportion=0.007215, over 51.00 utterances.], tot_loss[ctc_loss=0.1432, att_loss=0.2711, loss=0.2455, over 3275782.31 frames. utt_duration=1282 frames, utt_pad_proportion=0.04648, over 10236.63 utterances.], batch size: 51, lr: 1.72e-02, grad_scale: 8.0 2023-03-07 20:41:57,519 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23573.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:42:04,924 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-03-07 20:42:16,980 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23585.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:42:22,867 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.80 vs. limit=2.0 2023-03-07 20:42:59,613 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.210e+02 3.073e+02 3.816e+02 5.187e+02 1.095e+03, threshold=7.631e+02, percent-clipped=5.0 2023-03-07 20:43:12,939 INFO [train2.py:809] (1/4) Epoch 6, batch 3700, loss[ctc_loss=0.09395, att_loss=0.2261, loss=0.1997, over 15366.00 frames. utt_duration=1758 frames, utt_pad_proportion=0.01147, over 35.00 utterances.], tot_loss[ctc_loss=0.1442, att_loss=0.2711, loss=0.2457, over 3257804.46 frames. utt_duration=1254 frames, utt_pad_proportion=0.05698, over 10401.21 utterances.], batch size: 35, lr: 1.72e-02, grad_scale: 8.0 2023-03-07 20:44:32,075 INFO [train2.py:809] (1/4) Epoch 6, batch 3750, loss[ctc_loss=0.1327, att_loss=0.2662, loss=0.2395, over 17252.00 frames. utt_duration=875.1 frames, utt_pad_proportion=0.08269, over 79.00 utterances.], tot_loss[ctc_loss=0.144, att_loss=0.2709, loss=0.2455, over 3263279.42 frames. utt_duration=1250 frames, utt_pad_proportion=0.05502, over 10454.20 utterances.], batch size: 79, lr: 1.72e-02, grad_scale: 8.0 2023-03-07 20:44:44,505 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.53 vs. limit=5.0 2023-03-07 20:45:26,102 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-03-07 20:45:38,747 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.845e+02 3.181e+02 3.944e+02 5.174e+02 1.956e+03, threshold=7.887e+02, percent-clipped=9.0 2023-03-07 20:45:52,112 INFO [train2.py:809] (1/4) Epoch 6, batch 3800, loss[ctc_loss=0.1577, att_loss=0.2929, loss=0.2658, over 17122.00 frames. utt_duration=1224 frames, utt_pad_proportion=0.01412, over 56.00 utterances.], tot_loss[ctc_loss=0.1454, att_loss=0.2723, loss=0.2469, over 3265426.64 frames. utt_duration=1207 frames, utt_pad_proportion=0.06554, over 10830.82 utterances.], batch size: 56, lr: 1.72e-02, grad_scale: 8.0 2023-03-07 20:47:02,177 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23764.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:47:08,178 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.7492, 5.9025, 5.3446, 5.8076, 5.5828, 5.2323, 5.4240, 5.1571], device='cuda:1'), covar=tensor([0.1293, 0.0861, 0.0712, 0.0619, 0.0687, 0.1283, 0.2005, 0.1976], device='cuda:1'), in_proj_covar=tensor([0.0375, 0.0414, 0.0312, 0.0330, 0.0301, 0.0376, 0.0448, 0.0403], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-03-07 20:47:11,175 INFO [train2.py:809] (1/4) Epoch 6, batch 3850, loss[ctc_loss=0.1197, att_loss=0.2468, loss=0.2214, over 15865.00 frames. utt_duration=1629 frames, utt_pad_proportion=0.01055, over 39.00 utterances.], tot_loss[ctc_loss=0.1447, att_loss=0.2715, loss=0.2462, over 3264302.42 frames. utt_duration=1219 frames, utt_pad_proportion=0.06164, over 10722.15 utterances.], batch size: 39, lr: 1.72e-02, grad_scale: 8.0 2023-03-07 20:47:47,861 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.06 vs. limit=5.0 2023-03-07 20:48:18,382 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.854e+02 3.226e+02 4.048e+02 5.028e+02 1.314e+03, threshold=8.096e+02, percent-clipped=6.0 2023-03-07 20:48:18,493 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=23812.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:48:31,083 INFO [train2.py:809] (1/4) Epoch 6, batch 3900, loss[ctc_loss=0.1359, att_loss=0.2764, loss=0.2483, over 17417.00 frames. utt_duration=1011 frames, utt_pad_proportion=0.04494, over 69.00 utterances.], tot_loss[ctc_loss=0.1442, att_loss=0.2721, loss=0.2465, over 3279538.02 frames. utt_duration=1224 frames, utt_pad_proportion=0.0567, over 10727.73 utterances.], batch size: 69, lr: 1.72e-02, grad_scale: 8.0 2023-03-07 20:48:47,607 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-03-07 20:48:50,153 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.5140, 1.9757, 1.5637, 1.6065, 3.0701, 2.0687, 1.4820, 2.8536], device='cuda:1'), covar=tensor([0.0344, 0.2465, 0.3695, 0.1184, 0.0585, 0.1259, 0.2571, 0.0740], device='cuda:1'), in_proj_covar=tensor([0.0083, 0.0087, 0.0091, 0.0078, 0.0077, 0.0075, 0.0086, 0.0067], device='cuda:1'), out_proj_covar=tensor([3.8083e-05, 5.0754e-05, 5.1411e-05, 4.1936e-05, 3.7874e-05, 4.3910e-05, 4.8949e-05, 4.0718e-05], device='cuda:1') 2023-03-07 20:49:08,488 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23844.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 20:49:15,978 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9498, 5.3555, 4.4830, 5.4736, 4.7695, 5.0885, 5.4602, 5.1715], device='cuda:1'), covar=tensor([0.0445, 0.0246, 0.1100, 0.0166, 0.0475, 0.0185, 0.0262, 0.0230], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0191, 0.0249, 0.0172, 0.0210, 0.0158, 0.0183, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004], device='cuda:1') 2023-03-07 20:49:36,920 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([1.8979, 3.3683, 3.3356, 2.8428, 2.9197, 3.1820, 3.5231, 2.3219], device='cuda:1'), covar=tensor([0.1789, 0.1188, 0.2427, 0.4580, 0.3355, 0.2956, 0.0770, 0.7651], device='cuda:1'), in_proj_covar=tensor([0.0077, 0.0079, 0.0085, 0.0123, 0.0076, 0.0111, 0.0070, 0.0128], device='cuda:1'), out_proj_covar=tensor([6.2754e-05, 5.8536e-05, 6.7270e-05, 9.2719e-05, 6.0641e-05, 8.7333e-05, 5.3899e-05, 1.0009e-04], device='cuda:1') 2023-03-07 20:49:48,967 INFO [train2.py:809] (1/4) Epoch 6, batch 3950, loss[ctc_loss=0.1079, att_loss=0.2635, loss=0.2324, over 16635.00 frames. utt_duration=1417 frames, utt_pad_proportion=0.004781, over 47.00 utterances.], tot_loss[ctc_loss=0.1439, att_loss=0.2725, loss=0.2468, over 3276974.52 frames. utt_duration=1214 frames, utt_pad_proportion=0.06025, over 10814.36 utterances.], batch size: 47, lr: 1.71e-02, grad_scale: 8.0 2023-03-07 20:49:54,035 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23873.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:50:01,642 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3984, 4.6910, 4.6301, 5.1036, 2.4048, 4.7747, 2.4615, 2.0974], device='cuda:1'), covar=tensor([0.0226, 0.0133, 0.0704, 0.0150, 0.2411, 0.0139, 0.1819, 0.2021], device='cuda:1'), in_proj_covar=tensor([0.0115, 0.0102, 0.0246, 0.0108, 0.0218, 0.0097, 0.0217, 0.0204], device='cuda:1'), out_proj_covar=tensor([1.1335e-04, 1.0064e-04, 2.1975e-04, 9.9169e-05, 2.0156e-04, 9.4405e-05, 1.9460e-04, 1.8259e-04], device='cuda:1') 2023-03-07 20:50:12,174 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23885.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:51:06,490 INFO [train2.py:809] (1/4) Epoch 7, batch 0, loss[ctc_loss=0.127, att_loss=0.2601, loss=0.2335, over 16405.00 frames. utt_duration=1493 frames, utt_pad_proportion=0.007268, over 44.00 utterances.], tot_loss[ctc_loss=0.127, att_loss=0.2601, loss=0.2335, over 16405.00 frames. utt_duration=1493 frames, utt_pad_proportion=0.007268, over 44.00 utterances.], batch size: 44, lr: 1.61e-02, grad_scale: 8.0 2023-03-07 20:51:06,490 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-07 20:51:19,280 INFO [train2.py:843] (1/4) Epoch 7, validation: ctc_loss=0.06772, att_loss=0.2471, loss=0.2112, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-07 20:51:19,281 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-07 20:51:34,365 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.137e+02 3.227e+02 4.296e+02 5.385e+02 1.552e+03, threshold=8.591e+02, percent-clipped=5.0 2023-03-07 20:51:48,668 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=23921.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:52:07,301 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=23933.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:52:30,512 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0536, 5.0708, 5.0091, 2.5481, 1.8817, 2.5346, 4.5441, 3.6746], device='cuda:1'), covar=tensor([0.0565, 0.0177, 0.0199, 0.2837, 0.6037, 0.2765, 0.0379, 0.1915], device='cuda:1'), in_proj_covar=tensor([0.0296, 0.0191, 0.0214, 0.0177, 0.0350, 0.0333, 0.0202, 0.0329], device='cuda:1'), out_proj_covar=tensor([1.4531e-04, 7.7508e-05, 9.4048e-05, 8.1019e-05, 1.6523e-04, 1.4631e-04, 8.4685e-05, 1.5242e-04], device='cuda:1') 2023-03-07 20:52:38,298 INFO [train2.py:809] (1/4) Epoch 7, batch 50, loss[ctc_loss=0.1473, att_loss=0.288, loss=0.2598, over 17401.00 frames. utt_duration=1106 frames, utt_pad_proportion=0.03305, over 63.00 utterances.], tot_loss[ctc_loss=0.1422, att_loss=0.2705, loss=0.2448, over 739644.52 frames. utt_duration=1288 frames, utt_pad_proportion=0.0497, over 2300.01 utterances.], batch size: 63, lr: 1.60e-02, grad_scale: 8.0 2023-03-07 20:52:38,507 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.8429, 5.0917, 5.3354, 5.2573, 5.2040, 5.7562, 5.0461, 5.9291], device='cuda:1'), covar=tensor([0.0586, 0.0610, 0.0604, 0.0819, 0.1634, 0.0751, 0.0570, 0.0483], device='cuda:1'), in_proj_covar=tensor([0.0561, 0.0350, 0.0375, 0.0446, 0.0609, 0.0384, 0.0314, 0.0390], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-03-07 20:53:40,708 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.1221, 2.2868, 2.8433, 3.8493, 3.6457, 3.7818, 2.5644, 1.8482], device='cuda:1'), covar=tensor([0.0716, 0.2550, 0.1269, 0.0613, 0.0721, 0.0324, 0.1947, 0.2789], device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0197, 0.0195, 0.0167, 0.0147, 0.0126, 0.0192, 0.0180], 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') 2023-03-07 20:54:02,454 INFO [train2.py:809] (1/4) Epoch 7, batch 100, loss[ctc_loss=0.144, att_loss=0.2841, loss=0.2561, over 17260.00 frames. utt_duration=1172 frames, utt_pad_proportion=0.02606, over 59.00 utterances.], tot_loss[ctc_loss=0.1411, att_loss=0.2693, loss=0.2437, over 1304105.02 frames. utt_duration=1253 frames, utt_pad_proportion=0.05052, over 4169.32 utterances.], batch size: 59, lr: 1.60e-02, grad_scale: 8.0 2023-03-07 20:54:18,137 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.071e+02 3.045e+02 3.704e+02 4.782e+02 7.393e+02, threshold=7.408e+02, percent-clipped=0.0 2023-03-07 20:54:57,359 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24037.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:55:00,429 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.4326, 5.1405, 5.3676, 2.9303, 5.2688, 4.5977, 4.6353, 2.6064], device='cuda:1'), covar=tensor([0.0078, 0.0072, 0.0093, 0.0930, 0.0059, 0.0128, 0.0226, 0.1408], device='cuda:1'), in_proj_covar=tensor([0.0053, 0.0068, 0.0057, 0.0099, 0.0061, 0.0076, 0.0081, 0.0099], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-07 20:55:21,600 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0233, 5.2972, 5.2326, 5.1596, 5.4296, 5.3976, 5.1417, 4.8297], device='cuda:1'), covar=tensor([0.0887, 0.0359, 0.0217, 0.0393, 0.0206, 0.0216, 0.0197, 0.0293], device='cuda:1'), in_proj_covar=tensor([0.0394, 0.0237, 0.0180, 0.0218, 0.0278, 0.0302, 0.0225, 0.0261], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-07 20:55:23,558 INFO [train2.py:809] (1/4) Epoch 7, batch 150, loss[ctc_loss=0.09756, att_loss=0.2335, loss=0.2063, over 16181.00 frames. utt_duration=1580 frames, utt_pad_proportion=0.006305, over 41.00 utterances.], tot_loss[ctc_loss=0.1411, att_loss=0.2689, loss=0.2433, over 1743080.95 frames. utt_duration=1216 frames, utt_pad_proportion=0.05754, over 5742.20 utterances.], batch size: 41, lr: 1.60e-02, grad_scale: 8.0 2023-03-07 20:56:34,151 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24098.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:56:42,639 INFO [train2.py:809] (1/4) Epoch 7, batch 200, loss[ctc_loss=0.1193, att_loss=0.2479, loss=0.2222, over 16248.00 frames. utt_duration=1513 frames, utt_pad_proportion=0.009153, over 43.00 utterances.], tot_loss[ctc_loss=0.139, att_loss=0.2675, loss=0.2418, over 2068771.64 frames. utt_duration=1265 frames, utt_pad_proportion=0.05144, over 6549.51 utterances.], batch size: 43, lr: 1.60e-02, grad_scale: 8.0 2023-03-07 20:56:56,754 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.983e+02 2.953e+02 3.642e+02 4.338e+02 9.354e+02, threshold=7.285e+02, percent-clipped=1.0 2023-03-07 20:57:09,473 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2038, 1.3425, 1.9080, 0.9826, 3.7116, 2.2915, 1.5778, 2.3383], device='cuda:1'), covar=tensor([0.0417, 0.3271, 0.2432, 0.1990, 0.0278, 0.1358, 0.2513, 0.1081], device='cuda:1'), in_proj_covar=tensor([0.0082, 0.0088, 0.0088, 0.0078, 0.0076, 0.0075, 0.0089, 0.0071], device='cuda:1'), out_proj_covar=tensor([3.8733e-05, 5.1395e-05, 5.0511e-05, 4.2348e-05, 3.7698e-05, 4.4236e-05, 5.0340e-05, 4.2526e-05], device='cuda:1') 2023-03-07 20:57:37,386 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2968, 2.3684, 3.1717, 4.2268, 3.9544, 3.7967, 2.8937, 1.8844], device='cuda:1'), covar=tensor([0.0718, 0.2575, 0.1124, 0.0519, 0.0513, 0.0415, 0.1643, 0.2601], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0195, 0.0191, 0.0165, 0.0144, 0.0128, 0.0188, 0.0177], 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') 2023-03-07 20:57:47,333 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=24144.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 20:58:02,069 INFO [train2.py:809] (1/4) Epoch 7, batch 250, loss[ctc_loss=0.1188, att_loss=0.2547, loss=0.2275, over 16521.00 frames. utt_duration=1470 frames, utt_pad_proportion=0.007502, over 45.00 utterances.], tot_loss[ctc_loss=0.1383, att_loss=0.2673, loss=0.2415, over 2339890.71 frames. utt_duration=1258 frames, utt_pad_proportion=0.05022, over 7448.63 utterances.], batch size: 45, lr: 1.60e-02, grad_scale: 8.0 2023-03-07 20:59:02,761 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=24192.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 20:59:22,697 INFO [train2.py:809] (1/4) Epoch 7, batch 300, loss[ctc_loss=0.1238, att_loss=0.2468, loss=0.2222, over 11026.00 frames. utt_duration=1839 frames, utt_pad_proportion=0.1999, over 24.00 utterances.], tot_loss[ctc_loss=0.1391, att_loss=0.2682, loss=0.2424, over 2543983.66 frames. utt_duration=1226 frames, utt_pad_proportion=0.05806, over 8310.03 utterances.], batch size: 24, lr: 1.60e-02, grad_scale: 8.0 2023-03-07 20:59:36,411 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.964e+02 2.956e+02 3.932e+02 4.844e+02 1.089e+03, threshold=7.865e+02, percent-clipped=8.0 2023-03-07 21:00:42,739 INFO [train2.py:809] (1/4) Epoch 7, batch 350, loss[ctc_loss=0.1356, att_loss=0.2711, loss=0.244, over 17057.00 frames. utt_duration=1288 frames, utt_pad_proportion=0.008955, over 53.00 utterances.], tot_loss[ctc_loss=0.1397, att_loss=0.2691, loss=0.2432, over 2710638.50 frames. utt_duration=1208 frames, utt_pad_proportion=0.06155, over 8985.32 utterances.], batch size: 53, lr: 1.59e-02, grad_scale: 8.0 2023-03-07 21:01:23,134 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3441, 2.3473, 3.5215, 2.5962, 3.2659, 4.4935, 4.1534, 3.0181], device='cuda:1'), covar=tensor([0.0446, 0.2046, 0.0992, 0.1534, 0.1050, 0.0492, 0.0504, 0.1458], device='cuda:1'), in_proj_covar=tensor([0.0212, 0.0211, 0.0210, 0.0192, 0.0215, 0.0235, 0.0180, 0.0202], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-03-07 21:02:03,067 INFO [train2.py:809] (1/4) Epoch 7, batch 400, loss[ctc_loss=0.09219, att_loss=0.2254, loss=0.1988, over 15780.00 frames. utt_duration=1663 frames, utt_pad_proportion=0.007945, over 38.00 utterances.], tot_loss[ctc_loss=0.1386, att_loss=0.2682, loss=0.2422, over 2830721.63 frames. utt_duration=1206 frames, utt_pad_proportion=0.06506, over 9403.23 utterances.], batch size: 38, lr: 1.59e-02, grad_scale: 8.0 2023-03-07 21:02:16,567 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.699e+02 2.848e+02 3.496e+02 4.281e+02 6.983e+02, threshold=6.993e+02, percent-clipped=0.0 2023-03-07 21:02:31,236 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2826, 4.6554, 4.3078, 4.5736, 2.1687, 4.5639, 2.5761, 1.7559], device='cuda:1'), covar=tensor([0.0244, 0.0127, 0.0834, 0.0174, 0.2312, 0.0189, 0.1747, 0.1872], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0103, 0.0253, 0.0110, 0.0225, 0.0102, 0.0225, 0.0207], device='cuda:1'), out_proj_covar=tensor([1.1524e-04, 1.0300e-04, 2.2708e-04, 9.9973e-05, 2.0793e-04, 9.9952e-05, 2.0144e-04, 1.8630e-04], device='cuda:1') 2023-03-07 21:03:22,772 INFO [train2.py:809] (1/4) Epoch 7, batch 450, loss[ctc_loss=0.1686, att_loss=0.2834, loss=0.2604, over 17315.00 frames. utt_duration=1101 frames, utt_pad_proportion=0.0358, over 63.00 utterances.], tot_loss[ctc_loss=0.1381, att_loss=0.2682, loss=0.2422, over 2925851.88 frames. utt_duration=1224 frames, utt_pad_proportion=0.06125, over 9575.58 utterances.], batch size: 63, lr: 1.59e-02, grad_scale: 8.0 2023-03-07 21:03:23,230 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1137, 4.9331, 4.9326, 4.7376, 5.2733, 5.0636, 4.8588, 2.3488], device='cuda:1'), covar=tensor([0.0159, 0.0255, 0.0158, 0.0281, 0.1038, 0.0155, 0.0219, 0.2417], device='cuda:1'), in_proj_covar=tensor([0.0133, 0.0125, 0.0122, 0.0123, 0.0299, 0.0126, 0.0117, 0.0229], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-07 21:04:25,996 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=24393.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 21:04:42,795 INFO [train2.py:809] (1/4) Epoch 7, batch 500, loss[ctc_loss=0.09723, att_loss=0.2234, loss=0.1981, over 15653.00 frames. utt_duration=1694 frames, utt_pad_proportion=0.008434, over 37.00 utterances.], tot_loss[ctc_loss=0.1379, att_loss=0.2674, loss=0.2415, over 2997713.40 frames. utt_duration=1234 frames, utt_pad_proportion=0.06034, over 9727.28 utterances.], batch size: 37, lr: 1.59e-02, grad_scale: 8.0 2023-03-07 21:04:56,598 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.656e+02 3.293e+02 3.980e+02 5.188e+02 9.157e+02, threshold=7.960e+02, percent-clipped=6.0 2023-03-07 21:05:48,817 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7186, 3.8067, 3.0494, 3.4304, 3.9089, 3.6323, 2.6968, 4.4729], device='cuda:1'), covar=tensor([0.1168, 0.0392, 0.1162, 0.0578, 0.0484, 0.0625, 0.0948, 0.0338], device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0152, 0.0192, 0.0157, 0.0183, 0.0187, 0.0163, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-03-07 21:06:02,364 INFO [train2.py:809] (1/4) Epoch 7, batch 550, loss[ctc_loss=0.1505, att_loss=0.2726, loss=0.2482, over 16309.00 frames. utt_duration=1452 frames, utt_pad_proportion=0.007159, over 45.00 utterances.], tot_loss[ctc_loss=0.138, att_loss=0.2676, loss=0.2417, over 3060066.72 frames. utt_duration=1240 frames, utt_pad_proportion=0.05857, over 9884.24 utterances.], batch size: 45, lr: 1.59e-02, grad_scale: 8.0 2023-03-07 21:06:38,436 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.3239, 5.1766, 5.1032, 3.1098, 5.0651, 4.3110, 4.6208, 3.1087], device='cuda:1'), covar=tensor([0.0115, 0.0059, 0.0195, 0.0870, 0.0069, 0.0187, 0.0227, 0.1117], device='cuda:1'), in_proj_covar=tensor([0.0054, 0.0067, 0.0058, 0.0099, 0.0062, 0.0078, 0.0082, 0.0098], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-07 21:07:23,381 INFO [train2.py:809] (1/4) Epoch 7, batch 600, loss[ctc_loss=0.1466, att_loss=0.2758, loss=0.25, over 17043.00 frames. utt_duration=1288 frames, utt_pad_proportion=0.01012, over 53.00 utterances.], tot_loss[ctc_loss=0.1381, att_loss=0.2675, loss=0.2416, over 3107628.96 frames. utt_duration=1232 frames, utt_pad_proportion=0.05824, over 10105.75 utterances.], batch size: 53, lr: 1.59e-02, grad_scale: 8.0 2023-03-07 21:07:37,150 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.166e+02 3.070e+02 3.787e+02 4.448e+02 1.341e+03, threshold=7.574e+02, percent-clipped=2.0 2023-03-07 21:07:58,264 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.63 vs. limit=5.0 2023-03-07 21:08:33,825 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24547.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 21:08:43,728 INFO [train2.py:809] (1/4) Epoch 7, batch 650, loss[ctc_loss=0.1409, att_loss=0.2825, loss=0.2542, over 16611.00 frames. utt_duration=1415 frames, utt_pad_proportion=0.005428, over 47.00 utterances.], tot_loss[ctc_loss=0.1375, att_loss=0.2671, loss=0.2412, over 3146812.71 frames. utt_duration=1247 frames, utt_pad_proportion=0.05366, over 10106.26 utterances.], batch size: 47, lr: 1.59e-02, grad_scale: 8.0 2023-03-07 21:09:03,802 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=5.15 vs. limit=5.0 2023-03-07 21:09:08,931 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24569.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 21:09:17,155 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-03-07 21:10:04,224 INFO [train2.py:809] (1/4) Epoch 7, batch 700, loss[ctc_loss=0.1804, att_loss=0.2907, loss=0.2686, over 17310.00 frames. utt_duration=877.8 frames, utt_pad_proportion=0.08084, over 79.00 utterances.], tot_loss[ctc_loss=0.1369, att_loss=0.2666, loss=0.2407, over 3169312.47 frames. utt_duration=1251 frames, utt_pad_proportion=0.05463, over 10143.60 utterances.], batch size: 79, lr: 1.58e-02, grad_scale: 8.0 2023-03-07 21:10:12,531 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24608.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 21:10:18,242 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.862e+02 2.863e+02 3.503e+02 4.304e+02 9.533e+02, threshold=7.005e+02, percent-clipped=2.0 2023-03-07 21:10:46,392 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24630.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 21:11:23,260 INFO [train2.py:809] (1/4) Epoch 7, batch 750, loss[ctc_loss=0.1977, att_loss=0.3, loss=0.2795, over 17276.00 frames. utt_duration=876.1 frames, utt_pad_proportion=0.08259, over 79.00 utterances.], tot_loss[ctc_loss=0.1393, att_loss=0.2683, loss=0.2425, over 3193260.83 frames. utt_duration=1226 frames, utt_pad_proportion=0.05924, over 10429.03 utterances.], batch size: 79, lr: 1.58e-02, grad_scale: 16.0 2023-03-07 21:12:03,519 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.6925, 5.9559, 5.3136, 5.8482, 5.6016, 5.1770, 5.3199, 5.1632], device='cuda:1'), covar=tensor([0.1154, 0.0807, 0.0831, 0.0576, 0.0606, 0.1296, 0.2168, 0.1934], device='cuda:1'), in_proj_covar=tensor([0.0378, 0.0421, 0.0322, 0.0339, 0.0309, 0.0384, 0.0446, 0.0415], 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') 2023-03-07 21:12:05,212 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1279, 5.0009, 4.9317, 2.4808, 4.8142, 4.4065, 4.3199, 2.2577], device='cuda:1'), covar=tensor([0.0142, 0.0085, 0.0200, 0.1183, 0.0093, 0.0157, 0.0298, 0.1579], device='cuda:1'), in_proj_covar=tensor([0.0055, 0.0067, 0.0058, 0.0100, 0.0063, 0.0078, 0.0082, 0.0100], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-07 21:12:16,684 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.95 vs. limit=2.0 2023-03-07 21:12:27,409 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=24693.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 21:12:36,531 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8478, 3.5860, 3.9559, 3.2291, 3.9660, 4.8489, 4.5279, 3.7506], device='cuda:1'), covar=tensor([0.0320, 0.1153, 0.0918, 0.1086, 0.0748, 0.0606, 0.0477, 0.0939], device='cuda:1'), in_proj_covar=tensor([0.0208, 0.0212, 0.0216, 0.0189, 0.0214, 0.0234, 0.0180, 0.0200], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-03-07 21:12:43,247 INFO [train2.py:809] (1/4) Epoch 7, batch 800, loss[ctc_loss=0.1124, att_loss=0.273, loss=0.2409, over 17064.00 frames. utt_duration=1289 frames, utt_pad_proportion=0.009064, over 53.00 utterances.], tot_loss[ctc_loss=0.1398, att_loss=0.2697, loss=0.2438, over 3216349.10 frames. utt_duration=1207 frames, utt_pad_proportion=0.06245, over 10671.81 utterances.], batch size: 53, lr: 1.58e-02, grad_scale: 16.0 2023-03-07 21:12:57,268 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.749e+02 2.992e+02 3.996e+02 4.981e+02 1.398e+03, threshold=7.991e+02, percent-clipped=9.0 2023-03-07 21:12:58,181 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-03-07 21:13:08,444 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.8189, 3.8620, 3.1871, 3.3465, 3.8081, 3.5115, 2.2275, 4.4190], device='cuda:1'), covar=tensor([0.0946, 0.0395, 0.0901, 0.0643, 0.0627, 0.0629, 0.1160, 0.0336], device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0149, 0.0189, 0.0156, 0.0183, 0.0187, 0.0163, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-03-07 21:13:21,370 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3067, 5.2403, 5.0321, 2.8310, 2.1505, 2.8864, 4.7554, 3.8699], device='cuda:1'), covar=tensor([0.0511, 0.0179, 0.0249, 0.3134, 0.5492, 0.2293, 0.0407, 0.1723], device='cuda:1'), in_proj_covar=tensor([0.0295, 0.0184, 0.0210, 0.0173, 0.0347, 0.0333, 0.0201, 0.0324], device='cuda:1'), out_proj_covar=tensor([1.4501e-04, 7.4815e-05, 9.2674e-05, 7.9302e-05, 1.6289e-04, 1.4562e-04, 8.3413e-05, 1.5015e-04], device='cuda:1') 2023-03-07 21:13:27,073 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4362, 4.7825, 4.2364, 4.8246, 4.2544, 4.5325, 4.8987, 4.7091], device='cuda:1'), covar=tensor([0.0485, 0.0259, 0.0853, 0.0199, 0.0493, 0.0323, 0.0226, 0.0171], device='cuda:1'), in_proj_covar=tensor([0.0262, 0.0204, 0.0264, 0.0182, 0.0219, 0.0168, 0.0193, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0005, 0.0005, 0.0004, 0.0005, 0.0004, 0.0005, 0.0004], device='cuda:1') 2023-03-07 21:13:43,786 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=24741.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 21:14:03,647 INFO [train2.py:809] (1/4) Epoch 7, batch 850, loss[ctc_loss=0.1374, att_loss=0.256, loss=0.2322, over 15768.00 frames. utt_duration=1661 frames, utt_pad_proportion=0.008809, over 38.00 utterances.], tot_loss[ctc_loss=0.139, att_loss=0.2695, loss=0.2434, over 3231555.75 frames. utt_duration=1208 frames, utt_pad_proportion=0.06215, over 10711.22 utterances.], batch size: 38, lr: 1.58e-02, grad_scale: 16.0 2023-03-07 21:14:22,624 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24765.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 21:15:24,352 INFO [train2.py:809] (1/4) Epoch 7, batch 900, loss[ctc_loss=0.1441, att_loss=0.2815, loss=0.254, over 17038.00 frames. utt_duration=1338 frames, utt_pad_proportion=0.006725, over 51.00 utterances.], tot_loss[ctc_loss=0.138, att_loss=0.2686, loss=0.2425, over 3241941.16 frames. utt_duration=1239 frames, utt_pad_proportion=0.05436, over 10475.44 utterances.], batch size: 51, lr: 1.58e-02, grad_scale: 16.0 2023-03-07 21:15:38,323 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.099e+02 2.848e+02 3.280e+02 4.214e+02 1.122e+03, threshold=6.561e+02, percent-clipped=2.0 2023-03-07 21:16:00,324 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24826.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 21:16:05,068 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.3127, 1.4369, 2.0472, 2.1728, 3.3734, 1.8607, 1.8440, 2.4374], device='cuda:1'), covar=tensor([0.0579, 0.3858, 0.3065, 0.1185, 0.0552, 0.1751, 0.2378, 0.1375], device='cuda:1'), in_proj_covar=tensor([0.0079, 0.0086, 0.0086, 0.0077, 0.0077, 0.0074, 0.0084, 0.0068], device='cuda:1'), out_proj_covar=tensor([3.7443e-05, 5.0952e-05, 4.9862e-05, 4.2200e-05, 3.8120e-05, 4.3662e-05, 4.8614e-05, 4.1850e-05], device='cuda:1') 2023-03-07 21:16:36,221 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24848.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 21:16:44,175 INFO [train2.py:809] (1/4) Epoch 7, batch 950, loss[ctc_loss=0.1218, att_loss=0.259, loss=0.2316, over 15992.00 frames. utt_duration=1601 frames, utt_pad_proportion=0.008086, over 40.00 utterances.], tot_loss[ctc_loss=0.1364, att_loss=0.2681, loss=0.2418, over 3249869.28 frames. utt_duration=1264 frames, utt_pad_proportion=0.04816, over 10300.32 utterances.], batch size: 40, lr: 1.58e-02, grad_scale: 16.0 2023-03-07 21:16:58,014 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9008, 6.1089, 5.5022, 5.9250, 5.7521, 5.3864, 5.5671, 5.2208], device='cuda:1'), covar=tensor([0.1056, 0.0903, 0.0809, 0.0704, 0.0673, 0.1301, 0.2275, 0.2338], device='cuda:1'), in_proj_covar=tensor([0.0374, 0.0426, 0.0320, 0.0339, 0.0303, 0.0382, 0.0447, 0.0414], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-03-07 21:17:02,791 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.6013, 2.8675, 3.6046, 4.4403, 4.1679, 4.1489, 2.8153, 2.3618], device='cuda:1'), covar=tensor([0.0574, 0.2308, 0.1065, 0.0560, 0.0686, 0.0291, 0.1849, 0.2535], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0200, 0.0190, 0.0167, 0.0152, 0.0130, 0.0188, 0.0177], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-07 21:17:09,899 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-03-07 21:17:10,871 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8769, 4.6886, 4.7740, 4.8826, 5.1429, 5.1255, 4.6183, 2.0192], device='cuda:1'), covar=tensor([0.0174, 0.0186, 0.0145, 0.0069, 0.0769, 0.0120, 0.0231, 0.2805], device='cuda:1'), in_proj_covar=tensor([0.0132, 0.0122, 0.0122, 0.0122, 0.0301, 0.0128, 0.0117, 0.0228], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-07 21:17:28,091 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.79 vs. limit=2.0 2023-03-07 21:17:50,346 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.1565, 3.3404, 3.2477, 2.3816, 3.1819, 3.1233, 3.3944, 1.8225], device='cuda:1'), covar=tensor([0.1365, 0.1097, 0.2950, 0.7125, 0.1685, 0.2748, 0.0823, 1.0349], device='cuda:1'), in_proj_covar=tensor([0.0076, 0.0079, 0.0081, 0.0129, 0.0073, 0.0116, 0.0071, 0.0129], device='cuda:1'), out_proj_covar=tensor([6.2655e-05, 6.0010e-05, 6.6320e-05, 9.7749e-05, 5.9441e-05, 9.0688e-05, 5.4859e-05, 1.0126e-04], device='cuda:1') 2023-03-07 21:18:03,788 INFO [train2.py:809] (1/4) Epoch 7, batch 1000, loss[ctc_loss=0.1443, att_loss=0.2703, loss=0.2451, over 16553.00 frames. utt_duration=1473 frames, utt_pad_proportion=0.005582, over 45.00 utterances.], tot_loss[ctc_loss=0.1366, att_loss=0.2676, loss=0.2414, over 3245417.98 frames. utt_duration=1274 frames, utt_pad_proportion=0.04739, over 10201.84 utterances.], batch size: 45, lr: 1.57e-02, grad_scale: 16.0 2023-03-07 21:18:04,009 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=24903.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 21:18:13,389 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24909.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 21:18:17,630 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.779e+02 2.930e+02 3.688e+02 4.379e+02 1.333e+03, threshold=7.376e+02, percent-clipped=8.0 2023-03-07 21:18:37,656 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=24925.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 21:18:42,158 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9006, 5.2089, 5.0534, 5.0597, 5.2261, 5.2703, 4.9447, 4.6961], device='cuda:1'), covar=tensor([0.1025, 0.0359, 0.0193, 0.0511, 0.0268, 0.0253, 0.0284, 0.0320], device='cuda:1'), in_proj_covar=tensor([0.0398, 0.0237, 0.0182, 0.0223, 0.0281, 0.0306, 0.0230, 0.0263], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-07 21:19:22,726 INFO [train2.py:809] (1/4) Epoch 7, batch 1050, loss[ctc_loss=0.1528, att_loss=0.2882, loss=0.2611, over 16616.00 frames. utt_duration=672.8 frames, utt_pad_proportion=0.1558, over 99.00 utterances.], tot_loss[ctc_loss=0.1366, att_loss=0.2673, loss=0.2412, over 3253849.14 frames. utt_duration=1271 frames, utt_pad_proportion=0.04695, over 10252.63 utterances.], batch size: 99, lr: 1.57e-02, grad_scale: 16.0 2023-03-07 21:20:29,722 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24995.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 21:20:43,049 INFO [train2.py:809] (1/4) Epoch 7, batch 1100, loss[ctc_loss=0.1695, att_loss=0.2737, loss=0.2529, over 16387.00 frames. utt_duration=1492 frames, utt_pad_proportion=0.007621, over 44.00 utterances.], tot_loss[ctc_loss=0.1359, att_loss=0.2672, loss=0.2409, over 3260065.59 frames. utt_duration=1278 frames, utt_pad_proportion=0.04514, over 10215.45 utterances.], batch size: 44, lr: 1.57e-02, grad_scale: 16.0 2023-03-07 21:20:57,052 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.688e+02 2.910e+02 3.574e+02 4.609e+02 1.124e+03, threshold=7.148e+02, percent-clipped=6.0 2023-03-07 21:21:19,377 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.0299, 3.4309, 3.4148, 2.7111, 3.1959, 3.0143, 3.2675, 1.8833], device='cuda:1'), covar=tensor([0.1570, 0.0994, 0.1313, 0.5006, 0.1187, 0.5800, 0.0855, 0.9416], device='cuda:1'), in_proj_covar=tensor([0.0076, 0.0079, 0.0081, 0.0129, 0.0073, 0.0118, 0.0072, 0.0130], device='cuda:1'), out_proj_covar=tensor([6.3618e-05, 5.9889e-05, 6.6640e-05, 9.7891e-05, 5.9621e-05, 9.2269e-05, 5.5663e-05, 1.0206e-04], device='cuda:1') 2023-03-07 21:21:56,914 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6420, 2.7828, 3.6051, 2.9257, 3.4746, 4.6799, 4.4934, 3.1341], device='cuda:1'), covar=tensor([0.0315, 0.1763, 0.1209, 0.1442, 0.1074, 0.0666, 0.0445, 0.1473], device='cuda:1'), in_proj_covar=tensor([0.0210, 0.0214, 0.0218, 0.0192, 0.0216, 0.0240, 0.0184, 0.0204], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-03-07 21:22:03,894 INFO [train2.py:809] (1/4) Epoch 7, batch 1150, loss[ctc_loss=0.1202, att_loss=0.2685, loss=0.2389, over 16760.00 frames. utt_duration=1398 frames, utt_pad_proportion=0.006777, over 48.00 utterances.], tot_loss[ctc_loss=0.1364, att_loss=0.2673, loss=0.2411, over 3258464.20 frames. utt_duration=1242 frames, utt_pad_proportion=0.05666, over 10507.34 utterances.], batch size: 48, lr: 1.57e-02, grad_scale: 16.0 2023-03-07 21:22:08,944 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25056.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 21:22:52,687 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5223, 2.0389, 4.8261, 3.6165, 2.9448, 4.4514, 4.6924, 4.7033], device='cuda:1'), covar=tensor([0.0227, 0.2121, 0.0145, 0.1282, 0.2077, 0.0213, 0.0114, 0.0219], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0246, 0.0130, 0.0306, 0.0300, 0.0187, 0.0111, 0.0144], device='cuda:1'), out_proj_covar=tensor([1.3632e-04, 2.0187e-04, 1.1548e-04, 2.5089e-04, 2.5611e-04, 1.6369e-04, 9.9290e-05, 1.3086e-04], device='cuda:1') 2023-03-07 21:23:13,848 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.96 vs. limit=2.0 2023-03-07 21:23:24,556 INFO [train2.py:809] (1/4) Epoch 7, batch 1200, loss[ctc_loss=0.1456, att_loss=0.2584, loss=0.2359, over 15773.00 frames. utt_duration=1662 frames, utt_pad_proportion=0.008479, over 38.00 utterances.], tot_loss[ctc_loss=0.1359, att_loss=0.2667, loss=0.2406, over 3253470.58 frames. utt_duration=1236 frames, utt_pad_proportion=0.0613, over 10542.63 utterances.], batch size: 38, lr: 1.57e-02, grad_scale: 16.0 2023-03-07 21:23:38,666 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.736e+02 2.863e+02 3.570e+02 4.681e+02 1.368e+03, threshold=7.141e+02, percent-clipped=7.0 2023-03-07 21:23:52,654 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25121.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 21:23:56,919 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-03-07 21:24:44,725 INFO [train2.py:809] (1/4) Epoch 7, batch 1250, loss[ctc_loss=0.1018, att_loss=0.2297, loss=0.2041, over 15624.00 frames. utt_duration=1691 frames, utt_pad_proportion=0.009642, over 37.00 utterances.], tot_loss[ctc_loss=0.1365, att_loss=0.2674, loss=0.2413, over 3265263.24 frames. utt_duration=1230 frames, utt_pad_proportion=0.05998, over 10629.87 utterances.], batch size: 37, lr: 1.57e-02, grad_scale: 16.0 2023-03-07 21:25:48,352 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.99 vs. limit=5.0 2023-03-07 21:26:05,283 INFO [train2.py:809] (1/4) Epoch 7, batch 1300, loss[ctc_loss=0.1125, att_loss=0.2321, loss=0.2082, over 15898.00 frames. utt_duration=1632 frames, utt_pad_proportion=0.00835, over 39.00 utterances.], tot_loss[ctc_loss=0.1366, att_loss=0.2675, loss=0.2413, over 3268632.46 frames. utt_duration=1230 frames, utt_pad_proportion=0.0596, over 10642.09 utterances.], batch size: 39, lr: 1.57e-02, grad_scale: 16.0 2023-03-07 21:26:05,541 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25203.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 21:26:06,853 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25204.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 21:26:19,175 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.912e+02 2.854e+02 3.559e+02 4.625e+02 8.622e+02, threshold=7.119e+02, percent-clipped=5.0 2023-03-07 21:26:19,663 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9827, 4.3628, 4.0201, 4.4432, 2.0224, 3.9783, 2.4709, 2.2300], device='cuda:1'), covar=tensor([0.0318, 0.0135, 0.0943, 0.0157, 0.2580, 0.0267, 0.1774, 0.1472], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0103, 0.0253, 0.0109, 0.0225, 0.0105, 0.0227, 0.0203], device='cuda:1'), out_proj_covar=tensor([1.1652e-04, 1.0340e-04, 2.2704e-04, 9.9568e-05, 2.0824e-04, 1.0270e-04, 2.0354e-04, 1.8434e-04], device='cuda:1') 2023-03-07 21:26:39,743 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25225.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 21:26:40,626 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.55 vs. limit=5.0 2023-03-07 21:27:22,555 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=25251.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 21:27:25,583 INFO [train2.py:809] (1/4) Epoch 7, batch 1350, loss[ctc_loss=0.09938, att_loss=0.2288, loss=0.203, over 14472.00 frames. utt_duration=1811 frames, utt_pad_proportion=0.04296, over 32.00 utterances.], tot_loss[ctc_loss=0.1367, att_loss=0.2677, loss=0.2415, over 3265432.35 frames. utt_duration=1201 frames, utt_pad_proportion=0.06674, over 10887.07 utterances.], batch size: 32, lr: 1.56e-02, grad_scale: 16.0 2023-03-07 21:27:56,364 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=25273.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 21:28:45,923 INFO [train2.py:809] (1/4) Epoch 7, batch 1400, loss[ctc_loss=0.1945, att_loss=0.307, loss=0.2845, over 17516.00 frames. utt_duration=1002 frames, utt_pad_proportion=0.0517, over 70.00 utterances.], tot_loss[ctc_loss=0.1362, att_loss=0.2676, loss=0.2413, over 3269098.53 frames. utt_duration=1211 frames, utt_pad_proportion=0.06283, over 10810.84 utterances.], batch size: 70, lr: 1.56e-02, grad_scale: 16.0 2023-03-07 21:29:01,605 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.558e+02 2.920e+02 3.478e+02 4.487e+02 1.591e+03, threshold=6.957e+02, percent-clipped=5.0 2023-03-07 21:29:14,332 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4388, 4.3419, 4.2267, 4.4406, 4.6793, 4.5530, 4.2407, 2.1960], device='cuda:1'), covar=tensor([0.0261, 0.0323, 0.0280, 0.0114, 0.0999, 0.0214, 0.0314, 0.2566], device='cuda:1'), in_proj_covar=tensor([0.0132, 0.0123, 0.0121, 0.0120, 0.0304, 0.0127, 0.0116, 0.0229], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-07 21:30:03,020 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25351.0, num_to_drop=1, layers_to_drop={2} 2023-03-07 21:30:05,929 INFO [train2.py:809] (1/4) Epoch 7, batch 1450, loss[ctc_loss=0.1344, att_loss=0.2856, loss=0.2553, over 17048.00 frames. utt_duration=1339 frames, utt_pad_proportion=0.00704, over 51.00 utterances.], tot_loss[ctc_loss=0.1363, att_loss=0.268, loss=0.2416, over 3276060.23 frames. utt_duration=1202 frames, utt_pad_proportion=0.06325, over 10917.14 utterances.], batch size: 51, lr: 1.56e-02, grad_scale: 8.0 2023-03-07 21:30:50,057 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.14 vs. limit=5.0 2023-03-07 21:31:25,457 INFO [train2.py:809] (1/4) Epoch 7, batch 1500, loss[ctc_loss=0.1272, att_loss=0.2765, loss=0.2467, over 16862.00 frames. utt_duration=1378 frames, utt_pad_proportion=0.007949, over 49.00 utterances.], tot_loss[ctc_loss=0.1354, att_loss=0.2672, loss=0.2408, over 3267552.70 frames. utt_duration=1196 frames, utt_pad_proportion=0.06817, over 10938.87 utterances.], batch size: 49, lr: 1.56e-02, grad_scale: 8.0 2023-03-07 21:31:40,546 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.713e+02 2.966e+02 3.623e+02 4.720e+02 8.485e+02, threshold=7.246e+02, percent-clipped=4.0 2023-03-07 21:31:52,945 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25421.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 21:32:03,888 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=25428.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 21:32:44,990 INFO [train2.py:809] (1/4) Epoch 7, batch 1550, loss[ctc_loss=0.1381, att_loss=0.2566, loss=0.2329, over 15963.00 frames. utt_duration=1559 frames, utt_pad_proportion=0.006482, over 41.00 utterances.], tot_loss[ctc_loss=0.1354, att_loss=0.2669, loss=0.2406, over 3273235.05 frames. utt_duration=1226 frames, utt_pad_proportion=0.05908, over 10688.65 utterances.], batch size: 41, lr: 1.56e-02, grad_scale: 8.0 2023-03-07 21:33:10,092 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=25469.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 21:33:11,871 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4775, 5.0188, 4.6976, 4.8853, 5.0017, 4.6600, 3.6259, 4.8250], device='cuda:1'), covar=tensor([0.0104, 0.0089, 0.0093, 0.0109, 0.0082, 0.0088, 0.0496, 0.0204], device='cuda:1'), in_proj_covar=tensor([0.0061, 0.0056, 0.0068, 0.0045, 0.0045, 0.0055, 0.0078, 0.0076], 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') 2023-03-07 21:33:31,248 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.86 vs. limit=2.0 2023-03-07 21:33:42,827 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0367, 5.3701, 4.7997, 5.4364, 4.8364, 5.1221, 5.5792, 5.3233], device='cuda:1'), covar=tensor([0.0397, 0.0209, 0.0744, 0.0163, 0.0431, 0.0174, 0.0156, 0.0147], device='cuda:1'), in_proj_covar=tensor([0.0258, 0.0202, 0.0258, 0.0178, 0.0215, 0.0163, 0.0184, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0005, 0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-07 21:33:42,980 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25489.0, num_to_drop=1, layers_to_drop={2} 2023-03-07 21:33:43,029 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9538, 4.7660, 4.7093, 4.6120, 5.3619, 5.0706, 4.7247, 2.4386], device='cuda:1'), covar=tensor([0.0259, 0.0265, 0.0195, 0.0265, 0.0938, 0.0184, 0.0245, 0.2492], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0125, 0.0123, 0.0121, 0.0303, 0.0126, 0.0116, 0.0230], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-07 21:34:04,337 INFO [train2.py:809] (1/4) Epoch 7, batch 1600, loss[ctc_loss=0.1008, att_loss=0.2571, loss=0.2258, over 17046.00 frames. utt_duration=1288 frames, utt_pad_proportion=0.01014, over 53.00 utterances.], tot_loss[ctc_loss=0.1356, att_loss=0.2669, loss=0.2407, over 3271733.59 frames. utt_duration=1226 frames, utt_pad_proportion=0.06054, over 10683.72 utterances.], batch size: 53, lr: 1.56e-02, grad_scale: 8.0 2023-03-07 21:34:06,161 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25504.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 21:34:19,607 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.988e+02 2.911e+02 3.522e+02 4.081e+02 9.255e+02, threshold=7.044e+02, percent-clipped=3.0 2023-03-07 21:35:22,697 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=25552.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 21:35:24,112 INFO [train2.py:809] (1/4) Epoch 7, batch 1650, loss[ctc_loss=0.1308, att_loss=0.27, loss=0.2422, over 16972.00 frames. utt_duration=1359 frames, utt_pad_proportion=0.006998, over 50.00 utterances.], tot_loss[ctc_loss=0.1356, att_loss=0.2673, loss=0.241, over 3275375.47 frames. utt_duration=1234 frames, utt_pad_proportion=0.05817, over 10629.65 utterances.], batch size: 50, lr: 1.56e-02, grad_scale: 8.0 2023-03-07 21:36:44,062 INFO [train2.py:809] (1/4) Epoch 7, batch 1700, loss[ctc_loss=0.123, att_loss=0.2349, loss=0.2125, over 15366.00 frames. utt_duration=1758 frames, utt_pad_proportion=0.01151, over 35.00 utterances.], tot_loss[ctc_loss=0.1355, att_loss=0.2679, loss=0.2414, over 3281556.88 frames. utt_duration=1228 frames, utt_pad_proportion=0.05688, over 10699.61 utterances.], batch size: 35, lr: 1.55e-02, grad_scale: 8.0 2023-03-07 21:36:59,367 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.939e+02 2.859e+02 3.554e+02 4.579e+02 8.259e+02, threshold=7.107e+02, percent-clipped=6.0 2023-03-07 21:38:00,702 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25651.0, num_to_drop=1, layers_to_drop={2} 2023-03-07 21:38:03,530 INFO [train2.py:809] (1/4) Epoch 7, batch 1750, loss[ctc_loss=0.132, att_loss=0.271, loss=0.2432, over 16770.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.006407, over 48.00 utterances.], tot_loss[ctc_loss=0.1364, att_loss=0.2683, loss=0.2419, over 3283162.52 frames. utt_duration=1228 frames, utt_pad_proportion=0.05717, over 10709.58 utterances.], batch size: 48, lr: 1.55e-02, grad_scale: 8.0 2023-03-07 21:38:22,618 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7226, 5.3203, 5.0430, 5.2277, 5.2549, 4.9634, 4.2060, 5.2030], device='cuda:1'), covar=tensor([0.0105, 0.0076, 0.0078, 0.0065, 0.0082, 0.0090, 0.0385, 0.0190], device='cuda:1'), in_proj_covar=tensor([0.0061, 0.0056, 0.0070, 0.0045, 0.0046, 0.0056, 0.0079, 0.0077], 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') 2023-03-07 21:38:32,289 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.7619, 2.7309, 3.6295, 4.3545, 3.9480, 4.1998, 2.9970, 2.0749], device='cuda:1'), covar=tensor([0.0490, 0.2359, 0.0913, 0.0582, 0.0729, 0.0372, 0.1408, 0.2651], device='cuda:1'), in_proj_covar=tensor([0.0151, 0.0194, 0.0183, 0.0168, 0.0153, 0.0128, 0.0181, 0.0174], 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') 2023-03-07 21:39:05,725 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2534, 5.2902, 5.1270, 2.6412, 2.0746, 2.9880, 5.1963, 3.8428], device='cuda:1'), covar=tensor([0.0568, 0.0195, 0.0237, 0.3465, 0.6511, 0.2507, 0.0205, 0.1945], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0196, 0.0219, 0.0184, 0.0365, 0.0340, 0.0206, 0.0341], device='cuda:1'), out_proj_covar=tensor([1.5035e-04, 7.7798e-05, 9.6299e-05, 8.5326e-05, 1.6919e-04, 1.4798e-04, 8.3849e-05, 1.5578e-04], device='cuda:1') 2023-03-07 21:39:18,794 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=25699.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 21:39:20,522 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9461, 5.1952, 5.4723, 5.3847, 5.2888, 5.9237, 5.0474, 6.0525], device='cuda:1'), covar=tensor([0.0555, 0.0648, 0.0588, 0.0855, 0.1657, 0.0702, 0.0568, 0.0490], device='cuda:1'), in_proj_covar=tensor([0.0584, 0.0363, 0.0400, 0.0471, 0.0637, 0.0401, 0.0336, 0.0417], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-07 21:39:25,097 INFO [train2.py:809] (1/4) Epoch 7, batch 1800, loss[ctc_loss=0.1126, att_loss=0.2549, loss=0.2264, over 17011.00 frames. utt_duration=1336 frames, utt_pad_proportion=0.008341, over 51.00 utterances.], tot_loss[ctc_loss=0.1343, att_loss=0.267, loss=0.2404, over 3278001.94 frames. utt_duration=1229 frames, utt_pad_proportion=0.05821, over 10684.69 utterances.], batch size: 51, lr: 1.55e-02, grad_scale: 8.0 2023-03-07 21:39:31,806 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1756, 4.6838, 4.2561, 4.7134, 2.1434, 4.1845, 2.2241, 1.5393], device='cuda:1'), covar=tensor([0.0258, 0.0132, 0.0755, 0.0145, 0.2595, 0.0244, 0.1881, 0.2008], device='cuda:1'), in_proj_covar=tensor([0.0119, 0.0105, 0.0254, 0.0110, 0.0224, 0.0105, 0.0229, 0.0203], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-07 21:39:40,904 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.727e+02 2.893e+02 3.531e+02 4.524e+02 7.224e+02, threshold=7.063e+02, percent-clipped=1.0 2023-03-07 21:40:46,309 INFO [train2.py:809] (1/4) Epoch 7, batch 1850, loss[ctc_loss=0.1308, att_loss=0.2774, loss=0.2481, over 16775.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.006185, over 48.00 utterances.], tot_loss[ctc_loss=0.133, att_loss=0.2662, loss=0.2396, over 3274928.89 frames. utt_duration=1238 frames, utt_pad_proportion=0.05723, over 10596.77 utterances.], batch size: 48, lr: 1.55e-02, grad_scale: 8.0 2023-03-07 21:41:37,049 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25784.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 21:42:06,980 INFO [train2.py:809] (1/4) Epoch 7, batch 1900, loss[ctc_loss=0.122, att_loss=0.2368, loss=0.2139, over 15380.00 frames. utt_duration=1759 frames, utt_pad_proportion=0.01064, over 35.00 utterances.], tot_loss[ctc_loss=0.1331, att_loss=0.2667, loss=0.2399, over 3277026.54 frames. utt_duration=1250 frames, utt_pad_proportion=0.05298, over 10501.11 utterances.], batch size: 35, lr: 1.55e-02, grad_scale: 8.0 2023-03-07 21:42:22,403 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.824e+02 2.923e+02 3.414e+02 4.315e+02 6.616e+02, threshold=6.827e+02, percent-clipped=0.0 2023-03-07 21:42:22,777 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2133, 4.8859, 4.5487, 4.9463, 4.8580, 4.5659, 3.6450, 4.5577], device='cuda:1'), covar=tensor([0.0130, 0.0096, 0.0096, 0.0069, 0.0088, 0.0103, 0.0501, 0.0208], device='cuda:1'), in_proj_covar=tensor([0.0059, 0.0056, 0.0067, 0.0044, 0.0046, 0.0055, 0.0077, 0.0075], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-03-07 21:42:49,021 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.29 vs. limit=5.0 2023-03-07 21:43:27,377 INFO [train2.py:809] (1/4) Epoch 7, batch 1950, loss[ctc_loss=0.1336, att_loss=0.2755, loss=0.2471, over 17025.00 frames. utt_duration=1337 frames, utt_pad_proportion=0.007744, over 51.00 utterances.], tot_loss[ctc_loss=0.1342, att_loss=0.267, loss=0.2404, over 3284230.45 frames. utt_duration=1240 frames, utt_pad_proportion=0.0528, over 10609.90 utterances.], batch size: 51, lr: 1.55e-02, grad_scale: 8.0 2023-03-07 21:43:35,789 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=25858.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 21:43:40,228 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9523, 5.1334, 5.4102, 5.5364, 5.2776, 5.9316, 5.1474, 6.0784], device='cuda:1'), covar=tensor([0.0493, 0.0662, 0.0626, 0.0727, 0.1621, 0.0617, 0.0553, 0.0408], device='cuda:1'), in_proj_covar=tensor([0.0579, 0.0360, 0.0399, 0.0455, 0.0628, 0.0396, 0.0325, 0.0404], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-03-07 21:44:04,467 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1200, 4.2186, 4.1402, 4.6523, 2.3135, 4.2031, 2.5499, 1.9274], device='cuda:1'), covar=tensor([0.0270, 0.0170, 0.0734, 0.0108, 0.2186, 0.0191, 0.1689, 0.1805], device='cuda:1'), in_proj_covar=tensor([0.0119, 0.0105, 0.0252, 0.0107, 0.0220, 0.0103, 0.0226, 0.0201], device='cuda:1'), out_proj_covar=tensor([1.1842e-04, 1.0485e-04, 2.2664e-04, 9.8712e-05, 2.0399e-04, 1.0070e-04, 2.0293e-04, 1.8257e-04], device='cuda:1') 2023-03-07 21:44:48,784 INFO [train2.py:809] (1/4) Epoch 7, batch 2000, loss[ctc_loss=0.1219, att_loss=0.2773, loss=0.2462, over 17306.00 frames. utt_duration=1175 frames, utt_pad_proportion=0.02333, over 59.00 utterances.], tot_loss[ctc_loss=0.1327, att_loss=0.2657, loss=0.2391, over 3285134.78 frames. utt_duration=1256 frames, utt_pad_proportion=0.04829, over 10476.17 utterances.], batch size: 59, lr: 1.55e-02, grad_scale: 8.0 2023-03-07 21:44:50,633 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2450, 4.5425, 4.1289, 4.8103, 2.0015, 4.5922, 2.4939, 1.8310], device='cuda:1'), covar=tensor([0.0227, 0.0139, 0.1004, 0.0132, 0.2824, 0.0153, 0.1876, 0.2109], device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0105, 0.0255, 0.0108, 0.0222, 0.0104, 0.0228, 0.0203], device='cuda:1'), out_proj_covar=tensor([1.1929e-04, 1.0596e-04, 2.2904e-04, 9.9933e-05, 2.0621e-04, 1.0152e-04, 2.0420e-04, 1.8465e-04], device='cuda:1') 2023-03-07 21:45:04,084 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.703e+02 2.979e+02 3.606e+02 4.355e+02 1.257e+03, threshold=7.211e+02, percent-clipped=2.0 2023-03-07 21:45:14,775 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25919.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 21:45:57,437 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0067, 3.9541, 3.9489, 2.5315, 3.8092, 3.7730, 3.4294, 2.5569], device='cuda:1'), covar=tensor([0.0099, 0.0115, 0.0130, 0.0984, 0.0100, 0.0244, 0.0341, 0.1242], device='cuda:1'), in_proj_covar=tensor([0.0055, 0.0071, 0.0058, 0.0099, 0.0063, 0.0079, 0.0085, 0.0099], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-07 21:46:09,509 INFO [train2.py:809] (1/4) Epoch 7, batch 2050, loss[ctc_loss=0.1012, att_loss=0.2265, loss=0.2014, over 15650.00 frames. utt_duration=1693 frames, utt_pad_proportion=0.008624, over 37.00 utterances.], tot_loss[ctc_loss=0.1327, att_loss=0.2659, loss=0.2393, over 3283308.15 frames. utt_duration=1245 frames, utt_pad_proportion=0.05229, over 10561.75 utterances.], batch size: 37, lr: 1.54e-02, grad_scale: 8.0 2023-03-07 21:46:57,443 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0990, 5.4275, 5.7402, 5.6703, 5.4792, 6.0912, 5.2055, 6.1570], device='cuda:1'), covar=tensor([0.0568, 0.0538, 0.0566, 0.0911, 0.1576, 0.0618, 0.0547, 0.0592], device='cuda:1'), in_proj_covar=tensor([0.0602, 0.0366, 0.0406, 0.0463, 0.0640, 0.0412, 0.0333, 0.0418], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-07 21:47:33,893 INFO [train2.py:809] (1/4) Epoch 7, batch 2100, loss[ctc_loss=0.2169, att_loss=0.3073, loss=0.2892, over 14257.00 frames. utt_duration=389.6 frames, utt_pad_proportion=0.3177, over 147.00 utterances.], tot_loss[ctc_loss=0.1322, att_loss=0.2656, loss=0.2389, over 3283236.01 frames. utt_duration=1251 frames, utt_pad_proportion=0.0507, over 10510.17 utterances.], batch size: 147, lr: 1.54e-02, grad_scale: 8.0 2023-03-07 21:47:49,610 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.867e+02 2.961e+02 3.539e+02 4.278e+02 1.039e+03, threshold=7.077e+02, percent-clipped=2.0 2023-03-07 21:48:54,477 INFO [train2.py:809] (1/4) Epoch 7, batch 2150, loss[ctc_loss=0.1457, att_loss=0.292, loss=0.2627, over 17066.00 frames. utt_duration=1314 frames, utt_pad_proportion=0.008215, over 52.00 utterances.], tot_loss[ctc_loss=0.133, att_loss=0.266, loss=0.2394, over 3287753.86 frames. utt_duration=1252 frames, utt_pad_proportion=0.0495, over 10518.01 utterances.], batch size: 52, lr: 1.54e-02, grad_scale: 8.0 2023-03-07 21:49:11,032 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-03-07 21:49:44,676 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26084.0, num_to_drop=1, layers_to_drop={2} 2023-03-07 21:50:13,361 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26102.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 21:50:14,579 INFO [train2.py:809] (1/4) Epoch 7, batch 2200, loss[ctc_loss=0.131, att_loss=0.2538, loss=0.2293, over 16178.00 frames. utt_duration=1580 frames, utt_pad_proportion=0.006351, over 41.00 utterances.], tot_loss[ctc_loss=0.1314, att_loss=0.2647, loss=0.238, over 3288268.42 frames. utt_duration=1279 frames, utt_pad_proportion=0.04318, over 10295.19 utterances.], batch size: 41, lr: 1.54e-02, grad_scale: 8.0 2023-03-07 21:50:30,121 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.379e+02 2.999e+02 3.731e+02 5.039e+02 1.380e+03, threshold=7.462e+02, percent-clipped=10.0 2023-03-07 21:51:01,058 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=26132.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 21:51:31,741 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.4643, 2.2545, 3.2039, 4.2082, 3.9022, 4.0163, 2.6773, 1.8204], device='cuda:1'), covar=tensor([0.0604, 0.2649, 0.1095, 0.0715, 0.0648, 0.0389, 0.1656, 0.2676], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0198, 0.0187, 0.0174, 0.0154, 0.0124, 0.0184, 0.0174], 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') 2023-03-07 21:51:34,424 INFO [train2.py:809] (1/4) Epoch 7, batch 2250, loss[ctc_loss=0.159, att_loss=0.2844, loss=0.2593, over 17354.00 frames. utt_duration=1008 frames, utt_pad_proportion=0.05031, over 69.00 utterances.], tot_loss[ctc_loss=0.1328, att_loss=0.2658, loss=0.2392, over 3282616.08 frames. utt_duration=1255 frames, utt_pad_proportion=0.0497, over 10478.17 utterances.], batch size: 69, lr: 1.54e-02, grad_scale: 8.0 2023-03-07 21:51:50,732 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26163.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 21:51:59,353 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26168.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 21:52:16,860 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6478, 2.6209, 3.7373, 2.7294, 3.5206, 4.7083, 4.3889, 3.2269], device='cuda:1'), covar=tensor([0.0350, 0.1975, 0.1007, 0.1646, 0.0988, 0.0565, 0.0683, 0.1359], device='cuda:1'), in_proj_covar=tensor([0.0210, 0.0214, 0.0215, 0.0194, 0.0218, 0.0241, 0.0188, 0.0205], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-03-07 21:52:55,165 INFO [train2.py:809] (1/4) Epoch 7, batch 2300, loss[ctc_loss=0.1147, att_loss=0.2689, loss=0.238, over 16763.00 frames. utt_duration=1398 frames, utt_pad_proportion=0.006806, over 48.00 utterances.], tot_loss[ctc_loss=0.1318, att_loss=0.2655, loss=0.2387, over 3286439.80 frames. utt_duration=1283 frames, utt_pad_proportion=0.04273, over 10258.25 utterances.], batch size: 48, lr: 1.54e-02, grad_scale: 8.0 2023-03-07 21:53:10,518 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.696e+02 2.838e+02 3.640e+02 4.798e+02 9.404e+02, threshold=7.281e+02, percent-clipped=6.0 2023-03-07 21:53:12,387 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26214.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 21:53:25,365 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0336, 5.2280, 5.5751, 5.5883, 5.4085, 5.9985, 5.1670, 6.0547], device='cuda:1'), covar=tensor([0.0549, 0.0663, 0.0596, 0.0744, 0.1594, 0.0631, 0.0553, 0.0496], device='cuda:1'), in_proj_covar=tensor([0.0609, 0.0368, 0.0413, 0.0468, 0.0655, 0.0415, 0.0342, 0.0417], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-07 21:53:30,210 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.6646, 3.6231, 3.0018, 3.2414, 3.8216, 3.4420, 2.4387, 4.2312], device='cuda:1'), covar=tensor([0.1084, 0.0461, 0.1087, 0.0692, 0.0601, 0.0714, 0.1037, 0.0412], device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0154, 0.0188, 0.0160, 0.0188, 0.0187, 0.0163, 0.0198], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-07 21:53:37,059 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26229.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 21:54:11,190 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26250.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 21:54:15,534 INFO [train2.py:809] (1/4) Epoch 7, batch 2350, loss[ctc_loss=0.112, att_loss=0.2447, loss=0.2182, over 15996.00 frames. utt_duration=1601 frames, utt_pad_proportion=0.007946, over 40.00 utterances.], tot_loss[ctc_loss=0.1319, att_loss=0.2658, loss=0.2391, over 3288328.21 frames. utt_duration=1284 frames, utt_pad_proportion=0.04213, over 10254.31 utterances.], batch size: 40, lr: 1.54e-02, grad_scale: 8.0 2023-03-07 21:54:48,174 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-03-07 21:55:36,002 INFO [train2.py:809] (1/4) Epoch 7, batch 2400, loss[ctc_loss=0.189, att_loss=0.2913, loss=0.2709, over 14467.00 frames. utt_duration=400.4 frames, utt_pad_proportion=0.3036, over 145.00 utterances.], tot_loss[ctc_loss=0.1314, att_loss=0.2652, loss=0.2385, over 3284823.51 frames. utt_duration=1265 frames, utt_pad_proportion=0.04715, over 10399.95 utterances.], batch size: 145, lr: 1.53e-02, grad_scale: 8.0 2023-03-07 21:55:42,948 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-03-07 21:55:49,007 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26311.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 21:55:52,368 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.814e+02 3.013e+02 3.656e+02 4.621e+02 8.184e+02, threshold=7.313e+02, percent-clipped=1.0 2023-03-07 21:56:00,264 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9746, 6.1581, 5.6847, 6.0683, 5.8897, 5.4703, 5.6312, 5.5595], device='cuda:1'), covar=tensor([0.1099, 0.0873, 0.0684, 0.0668, 0.0620, 0.1379, 0.2089, 0.2246], device='cuda:1'), in_proj_covar=tensor([0.0371, 0.0423, 0.0320, 0.0331, 0.0305, 0.0384, 0.0436, 0.0400], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-07 21:56:56,977 INFO [train2.py:809] (1/4) Epoch 7, batch 2450, loss[ctc_loss=0.1687, att_loss=0.2899, loss=0.2656, over 17423.00 frames. utt_duration=1012 frames, utt_pad_proportion=0.04652, over 69.00 utterances.], tot_loss[ctc_loss=0.1313, att_loss=0.265, loss=0.2383, over 3285194.53 frames. utt_duration=1262 frames, utt_pad_proportion=0.04705, over 10427.24 utterances.], batch size: 69, lr: 1.53e-02, grad_scale: 8.0 2023-03-07 21:58:18,163 INFO [train2.py:809] (1/4) Epoch 7, batch 2500, loss[ctc_loss=0.1302, att_loss=0.2765, loss=0.2473, over 17350.00 frames. utt_duration=1178 frames, utt_pad_proportion=0.02088, over 59.00 utterances.], tot_loss[ctc_loss=0.1308, att_loss=0.2649, loss=0.2381, over 3286424.61 frames. utt_duration=1280 frames, utt_pad_proportion=0.04203, over 10282.75 utterances.], batch size: 59, lr: 1.53e-02, grad_scale: 8.0 2023-03-07 21:58:33,370 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.952e+02 2.816e+02 3.513e+02 4.106e+02 7.996e+02, threshold=7.027e+02, percent-clipped=1.0 2023-03-07 21:58:36,088 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26414.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 21:58:45,504 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4735, 4.7607, 4.7813, 4.8717, 2.4858, 4.7902, 2.5199, 2.2961], device='cuda:1'), covar=tensor([0.0175, 0.0118, 0.0549, 0.0156, 0.2060, 0.0198, 0.1661, 0.1581], device='cuda:1'), in_proj_covar=tensor([0.0115, 0.0102, 0.0246, 0.0109, 0.0218, 0.0103, 0.0225, 0.0201], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-07 21:59:24,487 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1639, 5.1312, 4.9754, 2.3999, 1.8738, 2.4990, 4.4686, 3.6291], device='cuda:1'), covar=tensor([0.0592, 0.0197, 0.0215, 0.3505, 0.6247, 0.2909, 0.0522, 0.2195], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0197, 0.0215, 0.0178, 0.0353, 0.0335, 0.0210, 0.0340], device='cuda:1'), out_proj_covar=tensor([1.4920e-04, 7.8248e-05, 9.5239e-05, 8.2446e-05, 1.6351e-04, 1.4453e-04, 8.5821e-05, 1.5414e-04], device='cuda:1') 2023-03-07 21:59:32,841 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-03-07 21:59:38,004 INFO [train2.py:809] (1/4) Epoch 7, batch 2550, loss[ctc_loss=0.09585, att_loss=0.2294, loss=0.2027, over 15642.00 frames. utt_duration=1693 frames, utt_pad_proportion=0.008487, over 37.00 utterances.], tot_loss[ctc_loss=0.1309, att_loss=0.2647, loss=0.238, over 3284904.99 frames. utt_duration=1280 frames, utt_pad_proportion=0.04358, over 10277.75 utterances.], batch size: 37, lr: 1.53e-02, grad_scale: 8.0 2023-03-07 21:59:45,992 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26458.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 21:59:58,141 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.84 vs. limit=2.0 2023-03-07 22:00:14,041 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26475.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 22:00:23,603 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9239, 5.2898, 4.6942, 5.3298, 4.7271, 5.0834, 5.3868, 5.2450], device='cuda:1'), covar=tensor([0.0508, 0.0220, 0.0909, 0.0192, 0.0517, 0.0157, 0.0207, 0.0150], device='cuda:1'), in_proj_covar=tensor([0.0263, 0.0209, 0.0261, 0.0185, 0.0217, 0.0164, 0.0193, 0.0192], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0005, 0.0005, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004], device='cuda:1') 2023-03-07 22:00:58,231 INFO [train2.py:809] (1/4) Epoch 7, batch 2600, loss[ctc_loss=0.1607, att_loss=0.2956, loss=0.2687, over 16972.00 frames. utt_duration=1359 frames, utt_pad_proportion=0.007159, over 50.00 utterances.], tot_loss[ctc_loss=0.1308, att_loss=0.2651, loss=0.2382, over 3286837.35 frames. utt_duration=1275 frames, utt_pad_proportion=0.04491, over 10320.22 utterances.], batch size: 50, lr: 1.53e-02, grad_scale: 8.0 2023-03-07 22:01:14,580 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.671e+02 3.263e+02 3.835e+02 5.237e+02 1.388e+03, threshold=7.671e+02, percent-clipped=11.0 2023-03-07 22:01:16,499 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26514.0, num_to_drop=1, layers_to_drop={2} 2023-03-07 22:01:31,415 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26524.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 22:02:18,265 INFO [train2.py:809] (1/4) Epoch 7, batch 2650, loss[ctc_loss=0.1136, att_loss=0.2393, loss=0.2141, over 16113.00 frames. utt_duration=1536 frames, utt_pad_proportion=0.00633, over 42.00 utterances.], tot_loss[ctc_loss=0.132, att_loss=0.2655, loss=0.2388, over 3286615.88 frames. utt_duration=1274 frames, utt_pad_proportion=0.04681, over 10329.29 utterances.], batch size: 42, lr: 1.53e-02, grad_scale: 8.0 2023-03-07 22:02:18,789 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1136, 5.1196, 5.0065, 2.3538, 1.9608, 2.4503, 4.3597, 3.7756], device='cuda:1'), covar=tensor([0.0578, 0.0161, 0.0213, 0.3095, 0.6002, 0.3025, 0.0540, 0.1824], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0197, 0.0217, 0.0177, 0.0355, 0.0339, 0.0210, 0.0339], device='cuda:1'), out_proj_covar=tensor([1.5005e-04, 7.8335e-05, 9.6051e-05, 8.1868e-05, 1.6414e-04, 1.4637e-04, 8.5830e-05, 1.5370e-04], device='cuda:1') 2023-03-07 22:02:32,523 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=26562.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 22:02:36,064 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.1078, 6.3203, 5.6912, 6.1547, 6.0157, 5.7338, 5.8490, 5.6600], device='cuda:1'), covar=tensor([0.1080, 0.0817, 0.0692, 0.0658, 0.0588, 0.1337, 0.1962, 0.2459], device='cuda:1'), in_proj_covar=tensor([0.0379, 0.0439, 0.0327, 0.0342, 0.0313, 0.0391, 0.0448, 0.0413], 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') 2023-03-07 22:02:42,010 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.7834, 6.0534, 5.3811, 5.8606, 5.6784, 5.4088, 5.5259, 5.3858], device='cuda:1'), covar=tensor([0.1179, 0.0831, 0.0810, 0.0674, 0.0672, 0.1420, 0.2162, 0.2397], device='cuda:1'), in_proj_covar=tensor([0.0380, 0.0439, 0.0327, 0.0342, 0.0313, 0.0391, 0.0449, 0.0413], 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') 2023-03-07 22:03:27,899 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-03-07 22:03:37,576 INFO [train2.py:809] (1/4) Epoch 7, batch 2700, loss[ctc_loss=0.1586, att_loss=0.2913, loss=0.2648, over 17279.00 frames. utt_duration=1258 frames, utt_pad_proportion=0.0124, over 55.00 utterances.], tot_loss[ctc_loss=0.1316, att_loss=0.265, loss=0.2383, over 3282381.31 frames. utt_duration=1291 frames, utt_pad_proportion=0.04383, over 10184.14 utterances.], batch size: 55, lr: 1.53e-02, grad_scale: 8.0 2023-03-07 22:03:42,444 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26606.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 22:03:53,739 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.795e+02 3.023e+02 3.591e+02 4.363e+02 1.241e+03, threshold=7.183e+02, percent-clipped=4.0 2023-03-07 22:04:36,472 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-03-07 22:04:56,685 INFO [train2.py:809] (1/4) Epoch 7, batch 2750, loss[ctc_loss=0.128, att_loss=0.2501, loss=0.2257, over 16108.00 frames. utt_duration=1536 frames, utt_pad_proportion=0.007249, over 42.00 utterances.], tot_loss[ctc_loss=0.1317, att_loss=0.2651, loss=0.2384, over 3288702.33 frames. utt_duration=1295 frames, utt_pad_proportion=0.0412, over 10166.18 utterances.], batch size: 42, lr: 1.52e-02, grad_scale: 8.0 2023-03-07 22:06:15,391 INFO [train2.py:809] (1/4) Epoch 7, batch 2800, loss[ctc_loss=0.0963, att_loss=0.2515, loss=0.2204, over 16765.00 frames. utt_duration=1398 frames, utt_pad_proportion=0.006836, over 48.00 utterances.], tot_loss[ctc_loss=0.1319, att_loss=0.2652, loss=0.2385, over 3288908.32 frames. utt_duration=1294 frames, utt_pad_proportion=0.04104, over 10180.14 utterances.], batch size: 48, lr: 1.52e-02, grad_scale: 8.0 2023-03-07 22:06:31,174 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.919e+02 3.082e+02 3.761e+02 4.456e+02 1.465e+03, threshold=7.521e+02, percent-clipped=2.0 2023-03-07 22:06:48,455 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-03-07 22:07:34,540 INFO [train2.py:809] (1/4) Epoch 7, batch 2850, loss[ctc_loss=0.1186, att_loss=0.2366, loss=0.213, over 15780.00 frames. utt_duration=1662 frames, utt_pad_proportion=0.008102, over 38.00 utterances.], tot_loss[ctc_loss=0.1333, att_loss=0.2658, loss=0.2393, over 3277194.78 frames. utt_duration=1266 frames, utt_pad_proportion=0.05147, over 10368.23 utterances.], batch size: 38, lr: 1.52e-02, grad_scale: 8.0 2023-03-07 22:07:34,767 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0555, 5.2791, 5.6359, 5.5588, 5.3426, 5.9421, 5.2035, 6.0573], device='cuda:1'), covar=tensor([0.0562, 0.0615, 0.0535, 0.0775, 0.1814, 0.0838, 0.0476, 0.0549], device='cuda:1'), in_proj_covar=tensor([0.0592, 0.0361, 0.0400, 0.0462, 0.0631, 0.0413, 0.0325, 0.0415], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-03-07 22:07:42,481 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26758.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 22:08:01,765 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26770.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 22:08:01,957 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.5668, 1.6487, 1.9777, 1.7106, 3.5602, 2.3855, 2.0675, 1.3777], device='cuda:1'), covar=tensor([0.0798, 0.4028, 0.4024, 0.1710, 0.0401, 0.1448, 0.2421, 0.2155], device='cuda:1'), in_proj_covar=tensor([0.0081, 0.0089, 0.0086, 0.0079, 0.0074, 0.0075, 0.0087, 0.0068], device='cuda:1'), out_proj_covar=tensor([4.0040e-05, 5.3329e-05, 5.0621e-05, 4.3766e-05, 3.8674e-05, 4.4898e-05, 5.1339e-05, 4.3141e-05], device='cuda:1') 2023-03-07 22:08:54,865 INFO [train2.py:809] (1/4) Epoch 7, batch 2900, loss[ctc_loss=0.1262, att_loss=0.2432, loss=0.2198, over 15491.00 frames. utt_duration=1722 frames, utt_pad_proportion=0.009568, over 36.00 utterances.], tot_loss[ctc_loss=0.1327, att_loss=0.2657, loss=0.2391, over 3280115.71 frames. utt_duration=1260 frames, utt_pad_proportion=0.05184, over 10428.54 utterances.], batch size: 36, lr: 1.52e-02, grad_scale: 8.0 2023-03-07 22:08:59,635 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=26806.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 22:09:11,158 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.815e+02 3.082e+02 3.745e+02 4.484e+02 1.236e+03, threshold=7.491e+02, percent-clipped=3.0 2023-03-07 22:09:30,291 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26824.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 22:09:48,461 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1631, 5.0980, 5.0794, 2.5975, 1.8719, 2.6848, 4.6780, 3.7619], device='cuda:1'), covar=tensor([0.0573, 0.0189, 0.0220, 0.3346, 0.6630, 0.2874, 0.0477, 0.2085], device='cuda:1'), in_proj_covar=tensor([0.0306, 0.0195, 0.0215, 0.0181, 0.0356, 0.0340, 0.0212, 0.0343], device='cuda:1'), out_proj_covar=tensor([1.4920e-04, 7.7570e-05, 9.5898e-05, 8.3726e-05, 1.6434e-04, 1.4676e-04, 8.6566e-05, 1.5538e-04], device='cuda:1') 2023-03-07 22:10:16,148 INFO [train2.py:809] (1/4) Epoch 7, batch 2950, loss[ctc_loss=0.1461, att_loss=0.2813, loss=0.2543, over 17358.00 frames. utt_duration=1178 frames, utt_pad_proportion=0.02151, over 59.00 utterances.], tot_loss[ctc_loss=0.1324, att_loss=0.2656, loss=0.2389, over 3271327.40 frames. utt_duration=1249 frames, utt_pad_proportion=0.05524, over 10487.60 utterances.], batch size: 59, lr: 1.52e-02, grad_scale: 8.0 2023-03-07 22:10:47,882 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=26872.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 22:11:07,785 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-03-07 22:11:36,460 INFO [train2.py:809] (1/4) Epoch 7, batch 3000, loss[ctc_loss=0.1176, att_loss=0.2613, loss=0.2326, over 16485.00 frames. utt_duration=1435 frames, utt_pad_proportion=0.006233, over 46.00 utterances.], tot_loss[ctc_loss=0.1318, att_loss=0.2646, loss=0.238, over 3260639.97 frames. utt_duration=1246 frames, utt_pad_proportion=0.05858, over 10477.25 utterances.], batch size: 46, lr: 1.52e-02, grad_scale: 8.0 2023-03-07 22:11:36,460 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-07 22:11:50,143 INFO [train2.py:843] (1/4) Epoch 7, validation: ctc_loss=0.06224, att_loss=0.2434, loss=0.2072, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-07 22:11:50,143 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-07 22:11:52,066 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26904.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 22:11:55,117 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26906.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 22:12:06,320 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.662e+02 2.833e+02 3.331e+02 4.099e+02 1.001e+03, threshold=6.663e+02, percent-clipped=1.0 2023-03-07 22:12:09,830 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3614, 5.2243, 5.1635, 2.3864, 2.0775, 2.2487, 4.7798, 3.8283], device='cuda:1'), covar=tensor([0.0414, 0.0183, 0.0215, 0.2964, 0.5641, 0.3089, 0.0348, 0.1925], device='cuda:1'), in_proj_covar=tensor([0.0311, 0.0198, 0.0221, 0.0186, 0.0361, 0.0345, 0.0216, 0.0346], device='cuda:1'), out_proj_covar=tensor([1.5132e-04, 7.9183e-05, 9.8817e-05, 8.6260e-05, 1.6660e-04, 1.4879e-04, 8.8154e-05, 1.5677e-04], device='cuda:1') 2023-03-07 22:13:05,198 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26950.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 22:13:09,496 INFO [train2.py:809] (1/4) Epoch 7, batch 3050, loss[ctc_loss=0.1903, att_loss=0.3042, loss=0.2814, over 13918.00 frames. utt_duration=382.9 frames, utt_pad_proportion=0.3307, over 146.00 utterances.], tot_loss[ctc_loss=0.1325, att_loss=0.2653, loss=0.2387, over 3263660.34 frames. utt_duration=1246 frames, utt_pad_proportion=0.05681, over 10494.16 utterances.], batch size: 146, lr: 1.52e-02, grad_scale: 8.0 2023-03-07 22:13:11,182 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=26954.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 22:13:29,163 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26965.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 22:13:52,128 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4913, 2.6536, 3.6511, 2.6052, 3.4637, 4.6816, 4.4485, 3.1868], device='cuda:1'), covar=tensor([0.0411, 0.1944, 0.1058, 0.1689, 0.1149, 0.0661, 0.0457, 0.1426], device='cuda:1'), in_proj_covar=tensor([0.0214, 0.0214, 0.0215, 0.0194, 0.0219, 0.0247, 0.0187, 0.0205], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-03-07 22:14:29,786 INFO [train2.py:809] (1/4) Epoch 7, batch 3100, loss[ctc_loss=0.1102, att_loss=0.2696, loss=0.2377, over 17058.00 frames. utt_duration=1289 frames, utt_pad_proportion=0.009412, over 53.00 utterances.], tot_loss[ctc_loss=0.1331, att_loss=0.2659, loss=0.2394, over 3264164.93 frames. utt_duration=1240 frames, utt_pad_proportion=0.05861, over 10538.92 utterances.], batch size: 53, lr: 1.51e-02, grad_scale: 8.0 2023-03-07 22:14:43,220 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=27011.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 22:14:45,949 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.079e+02 3.057e+02 3.764e+02 4.844e+02 1.301e+03, threshold=7.527e+02, percent-clipped=9.0 2023-03-07 22:15:49,016 INFO [train2.py:809] (1/4) Epoch 7, batch 3150, loss[ctc_loss=0.2039, att_loss=0.3131, loss=0.2913, over 14194.00 frames. utt_duration=387.7 frames, utt_pad_proportion=0.3199, over 147.00 utterances.], tot_loss[ctc_loss=0.1343, att_loss=0.2665, loss=0.24, over 3262328.49 frames. utt_duration=1224 frames, utt_pad_proportion=0.06426, over 10673.80 utterances.], batch size: 147, lr: 1.51e-02, grad_scale: 8.0 2023-03-07 22:16:17,127 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27070.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 22:17:09,174 INFO [train2.py:809] (1/4) Epoch 7, batch 3200, loss[ctc_loss=0.1107, att_loss=0.2646, loss=0.2338, over 16861.00 frames. utt_duration=1378 frames, utt_pad_proportion=0.008022, over 49.00 utterances.], tot_loss[ctc_loss=0.1331, att_loss=0.2657, loss=0.2392, over 3266663.74 frames. utt_duration=1217 frames, utt_pad_proportion=0.06419, over 10752.32 utterances.], batch size: 49, lr: 1.51e-02, grad_scale: 8.0 2023-03-07 22:17:09,378 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.1139, 5.3031, 5.7055, 5.5134, 5.4130, 6.0509, 5.2151, 6.1864], device='cuda:1'), covar=tensor([0.0583, 0.0608, 0.0552, 0.1089, 0.1838, 0.0753, 0.0473, 0.0511], device='cuda:1'), in_proj_covar=tensor([0.0606, 0.0366, 0.0411, 0.0474, 0.0639, 0.0418, 0.0336, 0.0415], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-07 22:17:25,184 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.984e+02 2.918e+02 3.542e+02 4.352e+02 6.869e+02, threshold=7.084e+02, percent-clipped=0.0 2023-03-07 22:17:32,942 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=27118.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 22:18:21,589 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=27148.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 22:18:29,008 INFO [train2.py:809] (1/4) Epoch 7, batch 3250, loss[ctc_loss=0.1395, att_loss=0.2762, loss=0.2488, over 17285.00 frames. utt_duration=1259 frames, utt_pad_proportion=0.01214, over 55.00 utterances.], tot_loss[ctc_loss=0.1331, att_loss=0.2654, loss=0.239, over 3258707.81 frames. utt_duration=1191 frames, utt_pad_proportion=0.07298, over 10955.99 utterances.], batch size: 55, lr: 1.51e-02, grad_scale: 8.0 2023-03-07 22:19:17,249 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.4252, 3.6175, 2.9926, 3.3075, 3.7342, 3.3817, 2.3770, 4.0755], device='cuda:1'), covar=tensor([0.1187, 0.0336, 0.0974, 0.0611, 0.0523, 0.0623, 0.1001, 0.0384], device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0160, 0.0194, 0.0163, 0.0196, 0.0194, 0.0167, 0.0205], 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') 2023-03-07 22:19:48,582 INFO [train2.py:809] (1/4) Epoch 7, batch 3300, loss[ctc_loss=0.09784, att_loss=0.221, loss=0.1964, over 15505.00 frames. utt_duration=1724 frames, utt_pad_proportion=0.008642, over 36.00 utterances.], tot_loss[ctc_loss=0.1332, att_loss=0.2655, loss=0.239, over 3261012.11 frames. utt_duration=1181 frames, utt_pad_proportion=0.07568, over 11058.37 utterances.], batch size: 36, lr: 1.51e-02, grad_scale: 8.0 2023-03-07 22:19:59,205 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=27209.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 22:20:05,128 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.634e+02 2.966e+02 3.461e+02 4.502e+02 1.067e+03, threshold=6.921e+02, percent-clipped=5.0 2023-03-07 22:20:33,928 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2174, 4.8562, 4.4197, 4.4856, 2.1390, 4.5337, 2.3153, 1.8210], device='cuda:1'), covar=tensor([0.0246, 0.0095, 0.0690, 0.0219, 0.2365, 0.0184, 0.1793, 0.1953], device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0101, 0.0253, 0.0114, 0.0223, 0.0103, 0.0225, 0.0207], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-07 22:21:08,402 INFO [train2.py:809] (1/4) Epoch 7, batch 3350, loss[ctc_loss=0.1422, att_loss=0.2797, loss=0.2522, over 16988.00 frames. utt_duration=688 frames, utt_pad_proportion=0.1346, over 99.00 utterances.], tot_loss[ctc_loss=0.1321, att_loss=0.265, loss=0.2384, over 3261871.93 frames. utt_duration=1199 frames, utt_pad_proportion=0.06816, over 10894.95 utterances.], batch size: 99, lr: 1.51e-02, grad_scale: 8.0 2023-03-07 22:21:20,196 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=27260.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 22:22:28,316 INFO [train2.py:809] (1/4) Epoch 7, batch 3400, loss[ctc_loss=0.1412, att_loss=0.2748, loss=0.2481, over 17119.00 frames. utt_duration=1224 frames, utt_pad_proportion=0.01501, over 56.00 utterances.], tot_loss[ctc_loss=0.1321, att_loss=0.2652, loss=0.2386, over 3265311.14 frames. utt_duration=1207 frames, utt_pad_proportion=0.06503, over 10830.94 utterances.], batch size: 56, lr: 1.51e-02, grad_scale: 8.0 2023-03-07 22:22:33,604 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=27306.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 22:22:43,940 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.844e+02 2.932e+02 3.516e+02 4.279e+02 1.181e+03, threshold=7.032e+02, percent-clipped=5.0 2023-03-07 22:23:02,974 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3382, 2.4263, 4.8473, 3.6636, 2.7795, 4.3971, 4.4482, 4.5746], device='cuda:1'), covar=tensor([0.0292, 0.1760, 0.0151, 0.1250, 0.2245, 0.0232, 0.0169, 0.0274], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0248, 0.0128, 0.0307, 0.0290, 0.0181, 0.0110, 0.0142], device='cuda:1'), out_proj_covar=tensor([1.3801e-04, 2.0558e-04, 1.1371e-04, 2.5120e-04, 2.5147e-04, 1.5949e-04, 9.9214e-05, 1.3014e-04], device='cuda:1') 2023-03-07 22:23:18,176 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2161, 4.9507, 5.0075, 2.3354, 1.8714, 2.5687, 4.0853, 3.8534], device='cuda:1'), covar=tensor([0.0543, 0.0170, 0.0219, 0.3828, 0.6031, 0.2729, 0.0805, 0.1783], device='cuda:1'), in_proj_covar=tensor([0.0307, 0.0195, 0.0218, 0.0182, 0.0354, 0.0342, 0.0214, 0.0344], device='cuda:1'), out_proj_covar=tensor([1.4837e-04, 7.8042e-05, 9.7508e-05, 8.4430e-05, 1.6349e-04, 1.4649e-04, 8.7325e-05, 1.5552e-04], device='cuda:1') 2023-03-07 22:23:43,642 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.46 vs. limit=5.0 2023-03-07 22:23:45,930 INFO [train2.py:809] (1/4) Epoch 7, batch 3450, loss[ctc_loss=0.1322, att_loss=0.2832, loss=0.253, over 16861.00 frames. utt_duration=1378 frames, utt_pad_proportion=0.008096, over 49.00 utterances.], tot_loss[ctc_loss=0.1334, att_loss=0.266, loss=0.2395, over 3264935.19 frames. utt_duration=1228 frames, utt_pad_proportion=0.06, over 10645.89 utterances.], batch size: 49, lr: 1.51e-02, grad_scale: 16.0 2023-03-07 22:24:22,760 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 2023-03-07 22:24:40,969 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.7775, 2.5560, 3.4853, 4.4229, 4.2684, 4.3831, 2.7915, 2.1819], device='cuda:1'), covar=tensor([0.0556, 0.2729, 0.1215, 0.0868, 0.0602, 0.0296, 0.1747, 0.2644], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0202, 0.0196, 0.0180, 0.0157, 0.0129, 0.0191, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-07 22:24:45,539 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.4368, 2.2299, 3.1495, 4.0930, 3.8689, 4.0618, 2.5030, 1.8356], device='cuda:1'), covar=tensor([0.0625, 0.2859, 0.1188, 0.0857, 0.0562, 0.0304, 0.1974, 0.2817], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0202, 0.0196, 0.0180, 0.0158, 0.0129, 0.0191, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-07 22:24:49,062 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-03-07 22:25:06,082 INFO [train2.py:809] (1/4) Epoch 7, batch 3500, loss[ctc_loss=0.08159, att_loss=0.2374, loss=0.2062, over 16556.00 frames. utt_duration=1473 frames, utt_pad_proportion=0.005312, over 45.00 utterances.], tot_loss[ctc_loss=0.1338, att_loss=0.2663, loss=0.2398, over 3265847.04 frames. utt_duration=1227 frames, utt_pad_proportion=0.06026, over 10661.36 utterances.], batch size: 45, lr: 1.50e-02, grad_scale: 16.0 2023-03-07 22:25:11,022 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.6796, 2.8722, 3.5334, 4.5721, 4.2797, 4.5040, 2.8906, 2.3864], device='cuda:1'), covar=tensor([0.0564, 0.2153, 0.1142, 0.0545, 0.0623, 0.0211, 0.1648, 0.2247], device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0201, 0.0195, 0.0180, 0.0158, 0.0129, 0.0190, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-07 22:25:22,143 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.067e+02 3.129e+02 3.937e+02 5.112e+02 1.151e+03, threshold=7.873e+02, percent-clipped=8.0 2023-03-07 22:26:26,695 INFO [train2.py:809] (1/4) Epoch 7, batch 3550, loss[ctc_loss=0.1365, att_loss=0.2445, loss=0.2229, over 15505.00 frames. utt_duration=1724 frames, utt_pad_proportion=0.00861, over 36.00 utterances.], tot_loss[ctc_loss=0.1338, att_loss=0.2659, loss=0.2395, over 3254564.44 frames. utt_duration=1213 frames, utt_pad_proportion=0.06684, over 10749.93 utterances.], batch size: 36, lr: 1.50e-02, grad_scale: 16.0 2023-03-07 22:27:08,285 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.65 vs. limit=2.0 2023-03-07 22:27:46,247 INFO [train2.py:809] (1/4) Epoch 7, batch 3600, loss[ctc_loss=0.1202, att_loss=0.2649, loss=0.2359, over 16328.00 frames. utt_duration=1453 frames, utt_pad_proportion=0.006354, over 45.00 utterances.], tot_loss[ctc_loss=0.1331, att_loss=0.266, loss=0.2394, over 3264281.49 frames. utt_duration=1211 frames, utt_pad_proportion=0.06465, over 10791.32 utterances.], batch size: 45, lr: 1.50e-02, grad_scale: 16.0 2023-03-07 22:27:48,071 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=27504.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 22:28:02,318 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.030e+02 2.874e+02 3.597e+02 4.562e+02 1.127e+03, threshold=7.194e+02, percent-clipped=4.0 2023-03-07 22:29:06,204 INFO [train2.py:809] (1/4) Epoch 7, batch 3650, loss[ctc_loss=0.1232, att_loss=0.2558, loss=0.2293, over 16269.00 frames. utt_duration=1515 frames, utt_pad_proportion=0.008072, over 43.00 utterances.], tot_loss[ctc_loss=0.1332, att_loss=0.2658, loss=0.2393, over 3260997.74 frames. utt_duration=1228 frames, utt_pad_proportion=0.05998, over 10635.11 utterances.], batch size: 43, lr: 1.50e-02, grad_scale: 16.0 2023-03-07 22:29:18,183 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27560.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 22:30:27,717 INFO [train2.py:809] (1/4) Epoch 7, batch 3700, loss[ctc_loss=0.1258, att_loss=0.2765, loss=0.2463, over 17349.00 frames. utt_duration=1178 frames, utt_pad_proportion=0.02185, over 59.00 utterances.], tot_loss[ctc_loss=0.1321, att_loss=0.2646, loss=0.2381, over 3257550.29 frames. utt_duration=1245 frames, utt_pad_proportion=0.05683, over 10482.68 utterances.], batch size: 59, lr: 1.50e-02, grad_scale: 16.0 2023-03-07 22:30:33,196 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27606.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 22:30:36,306 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=27608.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 22:30:44,494 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.774e+02 3.024e+02 4.004e+02 4.895e+02 1.157e+03, threshold=8.007e+02, percent-clipped=6.0 2023-03-07 22:31:47,384 INFO [train2.py:809] (1/4) Epoch 7, batch 3750, loss[ctc_loss=0.1099, att_loss=0.254, loss=0.2252, over 16324.00 frames. utt_duration=1452 frames, utt_pad_proportion=0.006521, over 45.00 utterances.], tot_loss[ctc_loss=0.1321, att_loss=0.2647, loss=0.2382, over 3266281.07 frames. utt_duration=1246 frames, utt_pad_proportion=0.05454, over 10501.01 utterances.], batch size: 45, lr: 1.50e-02, grad_scale: 16.0 2023-03-07 22:31:49,602 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=27654.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 22:32:43,041 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.37 vs. limit=5.0 2023-03-07 22:33:06,376 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9579, 5.2468, 5.1779, 5.1796, 5.3614, 5.3568, 5.1269, 4.7679], device='cuda:1'), covar=tensor([0.1005, 0.0482, 0.0239, 0.0378, 0.0247, 0.0241, 0.0221, 0.0288], device='cuda:1'), in_proj_covar=tensor([0.0419, 0.0256, 0.0198, 0.0234, 0.0293, 0.0319, 0.0245, 0.0278], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-07 22:33:07,749 INFO [train2.py:809] (1/4) Epoch 7, batch 3800, loss[ctc_loss=0.1181, att_loss=0.259, loss=0.2308, over 16116.00 frames. utt_duration=1536 frames, utt_pad_proportion=0.006361, over 42.00 utterances.], tot_loss[ctc_loss=0.1318, att_loss=0.2648, loss=0.2382, over 3266161.13 frames. utt_duration=1258 frames, utt_pad_proportion=0.05161, over 10397.69 utterances.], batch size: 42, lr: 1.50e-02, grad_scale: 16.0 2023-03-07 22:33:25,168 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.970e+02 3.032e+02 3.559e+02 4.389e+02 6.627e+02, threshold=7.117e+02, percent-clipped=0.0 2023-03-07 22:33:27,190 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=27714.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 22:33:28,603 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1967, 5.0686, 5.0661, 3.1261, 4.7631, 4.3344, 4.6801, 2.6907], device='cuda:1'), covar=tensor([0.0112, 0.0081, 0.0154, 0.0954, 0.0105, 0.0179, 0.0209, 0.1325], device='cuda:1'), in_proj_covar=tensor([0.0054, 0.0074, 0.0060, 0.0101, 0.0064, 0.0081, 0.0085, 0.0099], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-07 22:34:28,362 INFO [train2.py:809] (1/4) Epoch 7, batch 3850, loss[ctc_loss=0.1367, att_loss=0.2657, loss=0.2399, over 16404.00 frames. utt_duration=1493 frames, utt_pad_proportion=0.007525, over 44.00 utterances.], tot_loss[ctc_loss=0.1306, att_loss=0.264, loss=0.2373, over 3266027.41 frames. utt_duration=1264 frames, utt_pad_proportion=0.05056, over 10345.26 utterances.], batch size: 44, lr: 1.49e-02, grad_scale: 16.0 2023-03-07 22:35:03,329 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=27775.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 22:35:46,403 INFO [train2.py:809] (1/4) Epoch 7, batch 3900, loss[ctc_loss=0.1263, att_loss=0.2435, loss=0.22, over 15646.00 frames. utt_duration=1693 frames, utt_pad_proportion=0.008656, over 37.00 utterances.], tot_loss[ctc_loss=0.1314, att_loss=0.2646, loss=0.238, over 3272932.41 frames. utt_duration=1248 frames, utt_pad_proportion=0.05276, over 10499.84 utterances.], batch size: 37, lr: 1.49e-02, grad_scale: 16.0 2023-03-07 22:35:48,259 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27804.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 22:36:01,930 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.865e+02 3.007e+02 3.697e+02 4.485e+02 8.676e+02, threshold=7.394e+02, percent-clipped=2.0 2023-03-07 22:36:33,048 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2979, 4.6881, 4.6740, 4.6720, 2.5061, 4.6438, 2.9211, 1.5627], device='cuda:1'), covar=tensor([0.0197, 0.0124, 0.0607, 0.0164, 0.2100, 0.0152, 0.1479, 0.2038], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0097, 0.0249, 0.0112, 0.0218, 0.0098, 0.0222, 0.0202], device='cuda:1'), out_proj_covar=tensor([1.1716e-04, 1.0103e-04, 2.2519e-04, 1.0524e-04, 2.0507e-04, 9.6998e-05, 2.0112e-04, 1.8411e-04], device='cuda:1') 2023-03-07 22:37:02,134 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=27852.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 22:37:03,594 INFO [train2.py:809] (1/4) Epoch 7, batch 3950, loss[ctc_loss=0.1316, att_loss=0.2742, loss=0.2457, over 16619.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.004979, over 47.00 utterances.], tot_loss[ctc_loss=0.131, att_loss=0.2646, loss=0.2378, over 3275400.08 frames. utt_duration=1252 frames, utt_pad_proportion=0.05137, over 10479.42 utterances.], batch size: 47, lr: 1.49e-02, grad_scale: 16.0 2023-03-07 22:37:24,453 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 2023-03-07 22:37:35,355 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-03-07 22:38:22,295 INFO [train2.py:809] (1/4) Epoch 8, batch 0, loss[ctc_loss=0.1382, att_loss=0.2859, loss=0.2563, over 17414.00 frames. utt_duration=1107 frames, utt_pad_proportion=0.03226, over 63.00 utterances.], tot_loss[ctc_loss=0.1382, att_loss=0.2859, loss=0.2563, over 17414.00 frames. utt_duration=1107 frames, utt_pad_proportion=0.03226, over 63.00 utterances.], batch size: 63, lr: 1.40e-02, grad_scale: 8.0 2023-03-07 22:38:22,295 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-07 22:38:34,611 INFO [train2.py:843] (1/4) Epoch 8, validation: ctc_loss=0.06098, att_loss=0.2435, loss=0.207, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-07 22:38:34,612 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-07 22:39:19,192 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.946e+02 2.808e+02 3.476e+02 4.336e+02 8.495e+02, threshold=6.952e+02, percent-clipped=4.0 2023-03-07 22:39:49,753 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2332, 4.7848, 4.7032, 4.6582, 2.5095, 4.7737, 2.3428, 2.0754], device='cuda:1'), covar=tensor([0.0212, 0.0120, 0.0585, 0.0198, 0.2206, 0.0154, 0.1805, 0.1768], device='cuda:1'), in_proj_covar=tensor([0.0119, 0.0100, 0.0254, 0.0114, 0.0224, 0.0101, 0.0228, 0.0206], device='cuda:1'), out_proj_covar=tensor([1.2005e-04, 1.0292e-04, 2.3046e-04, 1.0715e-04, 2.1040e-04, 9.9420e-05, 2.0616e-04, 1.8771e-04], device='cuda:1') 2023-03-07 22:39:54,579 INFO [train2.py:809] (1/4) Epoch 8, batch 50, loss[ctc_loss=0.1294, att_loss=0.277, loss=0.2475, over 16966.00 frames. utt_duration=1359 frames, utt_pad_proportion=0.007524, over 50.00 utterances.], tot_loss[ctc_loss=0.129, att_loss=0.2652, loss=0.2379, over 744617.41 frames. utt_duration=1283 frames, utt_pad_proportion=0.03906, over 2324.14 utterances.], batch size: 50, lr: 1.40e-02, grad_scale: 4.0 2023-03-07 22:40:52,072 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0979, 5.0669, 4.9819, 2.6530, 1.9389, 2.9060, 4.3746, 3.8464], device='cuda:1'), covar=tensor([0.0627, 0.0147, 0.0174, 0.3021, 0.5882, 0.2234, 0.0521, 0.1727], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0194, 0.0219, 0.0182, 0.0350, 0.0340, 0.0212, 0.0339], device='cuda:1'), out_proj_covar=tensor([1.4831e-04, 7.6864e-05, 9.7208e-05, 8.3502e-05, 1.6112e-04, 1.4467e-04, 8.6274e-05, 1.5294e-04], device='cuda:1') 2023-03-07 22:41:14,414 INFO [train2.py:809] (1/4) Epoch 8, batch 100, loss[ctc_loss=0.12, att_loss=0.2576, loss=0.2301, over 16109.00 frames. utt_duration=1536 frames, utt_pad_proportion=0.007111, over 42.00 utterances.], tot_loss[ctc_loss=0.1318, att_loss=0.2667, loss=0.2397, over 1313319.44 frames. utt_duration=1244 frames, utt_pad_proportion=0.04772, over 4227.68 utterances.], batch size: 42, lr: 1.40e-02, grad_scale: 4.0 2023-03-07 22:42:02,729 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.935e+02 2.962e+02 3.730e+02 4.479e+02 1.131e+03, threshold=7.459e+02, percent-clipped=8.0 2023-03-07 22:42:13,356 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-03-07 22:42:39,310 INFO [train2.py:809] (1/4) Epoch 8, batch 150, loss[ctc_loss=0.08708, att_loss=0.2283, loss=0.2, over 16026.00 frames. utt_duration=1604 frames, utt_pad_proportion=0.006672, over 40.00 utterances.], tot_loss[ctc_loss=0.1295, att_loss=0.2649, loss=0.2379, over 1746415.56 frames. utt_duration=1237 frames, utt_pad_proportion=0.0527, over 5654.09 utterances.], batch size: 40, lr: 1.40e-02, grad_scale: 4.0 2023-03-07 22:43:28,915 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3538, 2.7426, 3.6149, 2.6440, 3.4614, 4.4778, 4.2801, 2.9943], device='cuda:1'), covar=tensor([0.0465, 0.1798, 0.1066, 0.1457, 0.0937, 0.0640, 0.0496, 0.1518], device='cuda:1'), in_proj_covar=tensor([0.0216, 0.0212, 0.0219, 0.0196, 0.0222, 0.0252, 0.0194, 0.0209], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-03-07 22:43:31,912 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28070.0, num_to_drop=1, layers_to_drop={2} 2023-03-07 22:44:00,281 INFO [train2.py:809] (1/4) Epoch 8, batch 200, loss[ctc_loss=0.1199, att_loss=0.2507, loss=0.2245, over 16131.00 frames. utt_duration=1538 frames, utt_pad_proportion=0.005864, over 42.00 utterances.], tot_loss[ctc_loss=0.1308, att_loss=0.267, loss=0.2398, over 2095964.93 frames. utt_duration=1214 frames, utt_pad_proportion=0.05607, over 6911.78 utterances.], batch size: 42, lr: 1.40e-02, grad_scale: 4.0 2023-03-07 22:44:44,918 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.776e+02 2.763e+02 3.553e+02 4.276e+02 6.955e+02, threshold=7.106e+02, percent-clipped=0.0 2023-03-07 22:45:03,283 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.3124, 5.1986, 5.2246, 3.0780, 4.9300, 4.7128, 4.6335, 2.8282], device='cuda:1'), covar=tensor([0.0096, 0.0079, 0.0149, 0.0906, 0.0086, 0.0124, 0.0229, 0.1187], device='cuda:1'), in_proj_covar=tensor([0.0053, 0.0072, 0.0060, 0.0098, 0.0063, 0.0079, 0.0084, 0.0097], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-07 22:45:08,588 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0119, 4.6336, 4.2979, 4.3135, 2.4040, 4.1863, 2.5855, 1.9720], device='cuda:1'), covar=tensor([0.0282, 0.0081, 0.0599, 0.0162, 0.2029, 0.0171, 0.1586, 0.1753], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0095, 0.0251, 0.0111, 0.0218, 0.0098, 0.0222, 0.0202], device='cuda:1'), out_proj_covar=tensor([1.1909e-04, 9.9304e-05, 2.2780e-04, 1.0387e-04, 2.0574e-04, 9.6700e-05, 2.0106e-04, 1.8472e-04], device='cuda:1') 2023-03-07 22:45:21,332 INFO [train2.py:809] (1/4) Epoch 8, batch 250, loss[ctc_loss=0.1173, att_loss=0.2414, loss=0.2166, over 15516.00 frames. utt_duration=1725 frames, utt_pad_proportion=0.007779, over 36.00 utterances.], tot_loss[ctc_loss=0.13, att_loss=0.2655, loss=0.2384, over 2360319.48 frames. utt_duration=1222 frames, utt_pad_proportion=0.05566, over 7737.23 utterances.], batch size: 36, lr: 1.40e-02, grad_scale: 4.0 2023-03-07 22:45:26,912 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9242, 5.1838, 5.4405, 5.4285, 5.2498, 5.8171, 5.1223, 5.9982], device='cuda:1'), covar=tensor([0.0640, 0.0738, 0.0701, 0.0942, 0.1902, 0.0939, 0.0575, 0.0527], device='cuda:1'), in_proj_covar=tensor([0.0604, 0.0367, 0.0420, 0.0478, 0.0650, 0.0426, 0.0344, 0.0417], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-07 22:46:31,876 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-03-07 22:46:41,308 INFO [train2.py:809] (1/4) Epoch 8, batch 300, loss[ctc_loss=0.1273, att_loss=0.2555, loss=0.2299, over 16690.00 frames. utt_duration=1453 frames, utt_pad_proportion=0.006275, over 46.00 utterances.], tot_loss[ctc_loss=0.1295, att_loss=0.2654, loss=0.2382, over 2565181.83 frames. utt_duration=1229 frames, utt_pad_proportion=0.05522, over 8356.43 utterances.], batch size: 46, lr: 1.40e-02, grad_scale: 4.0 2023-03-07 22:47:27,066 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.936e+02 2.948e+02 3.357e+02 4.007e+02 1.001e+03, threshold=6.714e+02, percent-clipped=3.0 2023-03-07 22:47:51,007 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28229.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 22:48:03,884 INFO [train2.py:809] (1/4) Epoch 8, batch 350, loss[ctc_loss=0.1245, att_loss=0.2748, loss=0.2447, over 16483.00 frames. utt_duration=1435 frames, utt_pad_proportion=0.005695, over 46.00 utterances.], tot_loss[ctc_loss=0.1287, att_loss=0.2637, loss=0.2367, over 2712611.27 frames. utt_duration=1225 frames, utt_pad_proportion=0.06089, over 8867.56 utterances.], batch size: 46, lr: 1.40e-02, grad_scale: 4.0 2023-03-07 22:48:14,437 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.6052, 5.8342, 5.1498, 5.7530, 5.4930, 5.0939, 5.2680, 5.1763], device='cuda:1'), covar=tensor([0.1033, 0.0838, 0.0858, 0.0627, 0.0786, 0.1380, 0.2117, 0.2140], device='cuda:1'), in_proj_covar=tensor([0.0385, 0.0433, 0.0333, 0.0342, 0.0318, 0.0392, 0.0453, 0.0414], 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') 2023-03-07 22:48:25,646 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7831, 5.1022, 4.9645, 4.9899, 5.1496, 5.1458, 4.8178, 4.6029], device='cuda:1'), covar=tensor([0.1123, 0.0513, 0.0303, 0.0562, 0.0309, 0.0344, 0.0298, 0.0389], device='cuda:1'), in_proj_covar=tensor([0.0429, 0.0257, 0.0203, 0.0246, 0.0301, 0.0328, 0.0253, 0.0284], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-07 22:48:33,617 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28255.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 22:49:25,166 INFO [train2.py:809] (1/4) Epoch 8, batch 400, loss[ctc_loss=0.137, att_loss=0.2434, loss=0.2222, over 15395.00 frames. utt_duration=1761 frames, utt_pad_proportion=0.00932, over 35.00 utterances.], tot_loss[ctc_loss=0.1285, att_loss=0.2634, loss=0.2364, over 2836093.88 frames. utt_duration=1233 frames, utt_pad_proportion=0.05796, over 9210.19 utterances.], batch size: 35, lr: 1.40e-02, grad_scale: 8.0 2023-03-07 22:49:25,530 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28287.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 22:49:30,657 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28290.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 22:49:35,260 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1823, 5.1888, 5.0692, 2.2553, 1.8631, 2.6307, 4.1204, 3.5980], device='cuda:1'), covar=tensor([0.0558, 0.0142, 0.0176, 0.4310, 0.6043, 0.2800, 0.0715, 0.2253], device='cuda:1'), in_proj_covar=tensor([0.0319, 0.0195, 0.0221, 0.0188, 0.0357, 0.0348, 0.0217, 0.0348], device='cuda:1'), out_proj_covar=tensor([1.5348e-04, 7.7407e-05, 9.7042e-05, 8.6974e-05, 1.6408e-04, 1.4822e-04, 8.7514e-05, 1.5660e-04], device='cuda:1') 2023-03-07 22:49:49,480 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.1185, 3.9042, 3.2953, 3.7618, 4.1033, 3.6876, 2.8246, 4.5572], device='cuda:1'), covar=tensor([0.0890, 0.0530, 0.0993, 0.0540, 0.0522, 0.0590, 0.0972, 0.0294], device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0162, 0.0191, 0.0164, 0.0197, 0.0196, 0.0169, 0.0208], 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') 2023-03-07 22:50:09,511 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.675e+02 2.601e+02 3.135e+02 4.187e+02 9.825e+02, threshold=6.270e+02, percent-clipped=5.0 2023-03-07 22:50:11,477 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28316.0, num_to_drop=1, layers_to_drop={1} 2023-03-07 22:50:35,819 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0432, 4.7755, 4.3144, 4.4991, 2.6965, 4.4091, 2.3455, 1.6704], device='cuda:1'), covar=tensor([0.0287, 0.0091, 0.0721, 0.0167, 0.2042, 0.0145, 0.1925, 0.2050], device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0097, 0.0253, 0.0111, 0.0219, 0.0099, 0.0225, 0.0203], device='cuda:1'), out_proj_covar=tensor([1.2054e-04, 1.0085e-04, 2.2940e-04, 1.0408e-04, 2.0676e-04, 9.7024e-05, 2.0371e-04, 1.8611e-04], device='cuda:1') 2023-03-07 22:50:45,305 INFO [train2.py:809] (1/4) Epoch 8, batch 450, loss[ctc_loss=0.1461, att_loss=0.2658, loss=0.2419, over 16387.00 frames. utt_duration=1492 frames, utt_pad_proportion=0.006329, over 44.00 utterances.], tot_loss[ctc_loss=0.128, att_loss=0.2628, loss=0.2359, over 2926785.08 frames. utt_duration=1257 frames, utt_pad_proportion=0.05382, over 9321.75 utterances.], batch size: 44, lr: 1.39e-02, grad_scale: 8.0 2023-03-07 22:51:03,311 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28348.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 22:51:37,015 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28370.0, num_to_drop=1, layers_to_drop={2} 2023-03-07 22:51:41,803 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.5470, 3.4784, 2.7980, 3.1995, 3.5128, 3.2717, 2.1867, 3.7145], device='cuda:1'), covar=tensor([0.1019, 0.0403, 0.1134, 0.0588, 0.0636, 0.0628, 0.1114, 0.0441], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0160, 0.0192, 0.0163, 0.0199, 0.0196, 0.0169, 0.0209], 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') 2023-03-07 22:52:04,595 INFO [train2.py:809] (1/4) Epoch 8, batch 500, loss[ctc_loss=0.1467, att_loss=0.2753, loss=0.2495, over 17347.00 frames. utt_duration=1177 frames, utt_pad_proportion=0.0204, over 59.00 utterances.], tot_loss[ctc_loss=0.1287, att_loss=0.2627, loss=0.2359, over 2988818.24 frames. utt_duration=1239 frames, utt_pad_proportion=0.06298, over 9662.82 utterances.], batch size: 59, lr: 1.39e-02, grad_scale: 8.0 2023-03-07 22:52:49,177 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.916e+02 2.823e+02 3.661e+02 4.475e+02 7.657e+02, threshold=7.322e+02, percent-clipped=5.0 2023-03-07 22:52:54,254 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=28418.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 22:53:25,909 INFO [train2.py:809] (1/4) Epoch 8, batch 550, loss[ctc_loss=0.1031, att_loss=0.2457, loss=0.2172, over 16466.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.006584, over 46.00 utterances.], tot_loss[ctc_loss=0.1278, att_loss=0.2623, loss=0.2354, over 3051740.49 frames. utt_duration=1215 frames, utt_pad_proportion=0.06641, over 10056.81 utterances.], batch size: 46, lr: 1.39e-02, grad_scale: 8.0 2023-03-07 22:53:31,310 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9083, 4.5529, 4.5637, 2.3392, 2.1550, 2.6576, 2.9301, 3.6308], device='cuda:1'), covar=tensor([0.0623, 0.0168, 0.0215, 0.3265, 0.5460, 0.2662, 0.1458, 0.1701], device='cuda:1'), in_proj_covar=tensor([0.0312, 0.0195, 0.0220, 0.0184, 0.0352, 0.0339, 0.0215, 0.0342], device='cuda:1'), out_proj_covar=tensor([1.4952e-04, 7.6320e-05, 9.6562e-05, 8.5292e-05, 1.6110e-04, 1.4408e-04, 8.7045e-05, 1.5358e-04], device='cuda:1') 2023-03-07 22:54:01,454 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.4668, 3.6681, 3.5149, 2.9936, 3.3423, 3.4562, 3.3361, 1.9529], device='cuda:1'), covar=tensor([0.1252, 0.0985, 0.2448, 0.7248, 0.2699, 0.4251, 0.1338, 1.2346], device='cuda:1'), in_proj_covar=tensor([0.0076, 0.0089, 0.0090, 0.0147, 0.0085, 0.0135, 0.0082, 0.0142], device='cuda:1'), out_proj_covar=tensor([6.8573e-05, 7.0450e-05, 7.6987e-05, 1.1445e-04, 7.1930e-05, 1.0692e-04, 6.6573e-05, 1.1327e-04], device='cuda:1') 2023-03-07 22:54:43,122 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28486.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 22:54:44,263 INFO [train2.py:809] (1/4) Epoch 8, batch 600, loss[ctc_loss=0.1118, att_loss=0.2578, loss=0.2286, over 16635.00 frames. utt_duration=1417 frames, utt_pad_proportion=0.004901, over 47.00 utterances.], tot_loss[ctc_loss=0.1266, att_loss=0.2614, loss=0.2345, over 3104302.16 frames. utt_duration=1246 frames, utt_pad_proportion=0.05725, over 9974.36 utterances.], batch size: 47, lr: 1.39e-02, grad_scale: 8.0 2023-03-07 22:54:52,837 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28492.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 22:55:25,033 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3209, 4.8154, 4.7479, 4.5334, 2.3773, 4.4434, 2.5651, 2.2524], device='cuda:1'), covar=tensor([0.0235, 0.0114, 0.0594, 0.0191, 0.2458, 0.0141, 0.1834, 0.1834], device='cuda:1'), in_proj_covar=tensor([0.0123, 0.0100, 0.0257, 0.0113, 0.0224, 0.0102, 0.0229, 0.0206], device='cuda:1'), out_proj_covar=tensor([1.2350e-04, 1.0350e-04, 2.3331e-04, 1.0671e-04, 2.1105e-04, 9.9685e-05, 2.0777e-04, 1.8879e-04], device='cuda:1') 2023-03-07 22:55:27,684 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.904e+02 2.636e+02 3.439e+02 4.247e+02 1.142e+03, threshold=6.879e+02, percent-clipped=2.0 2023-03-07 22:56:03,435 INFO [train2.py:809] (1/4) Epoch 8, batch 650, loss[ctc_loss=0.1183, att_loss=0.2654, loss=0.236, over 16634.00 frames. utt_duration=1417 frames, utt_pad_proportion=0.004931, over 47.00 utterances.], tot_loss[ctc_loss=0.1272, att_loss=0.2619, loss=0.235, over 3142672.19 frames. utt_duration=1251 frames, utt_pad_proportion=0.05606, over 10059.15 utterances.], batch size: 47, lr: 1.39e-02, grad_scale: 8.0 2023-03-07 22:56:19,719 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28547.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 22:56:29,204 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28553.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 22:57:21,180 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28585.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 22:57:24,655 INFO [train2.py:809] (1/4) Epoch 8, batch 700, loss[ctc_loss=0.1371, att_loss=0.2771, loss=0.2491, over 17468.00 frames. utt_duration=1014 frames, utt_pad_proportion=0.04208, over 69.00 utterances.], tot_loss[ctc_loss=0.1273, att_loss=0.2616, loss=0.2348, over 3170057.90 frames. utt_duration=1231 frames, utt_pad_proportion=0.06107, over 10314.11 utterances.], batch size: 69, lr: 1.39e-02, grad_scale: 8.0 2023-03-07 22:58:02,977 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28611.0, num_to_drop=1, layers_to_drop={3} 2023-03-07 22:58:08,929 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.741e+02 2.921e+02 3.565e+02 4.512e+02 9.961e+02, threshold=7.130e+02, percent-clipped=4.0 2023-03-07 22:58:44,786 INFO [train2.py:809] (1/4) Epoch 8, batch 750, loss[ctc_loss=0.1788, att_loss=0.3086, loss=0.2827, over 17224.00 frames. utt_duration=873.6 frames, utt_pad_proportion=0.08328, over 79.00 utterances.], tot_loss[ctc_loss=0.1275, att_loss=0.262, loss=0.2351, over 3186304.60 frames. utt_duration=1220 frames, utt_pad_proportion=0.06482, over 10460.22 utterances.], batch size: 79, lr: 1.39e-02, grad_scale: 8.0 2023-03-07 22:58:55,041 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28643.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 22:59:10,012 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-03-07 23:00:04,388 INFO [train2.py:809] (1/4) Epoch 8, batch 800, loss[ctc_loss=0.1381, att_loss=0.2648, loss=0.2395, over 16285.00 frames. utt_duration=1516 frames, utt_pad_proportion=0.00696, over 43.00 utterances.], tot_loss[ctc_loss=0.1274, att_loss=0.2624, loss=0.2354, over 3211676.81 frames. utt_duration=1234 frames, utt_pad_proportion=0.05767, over 10422.71 utterances.], batch size: 43, lr: 1.39e-02, grad_scale: 8.0 2023-03-07 23:00:47,880 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.805e+02 2.816e+02 3.750e+02 4.467e+02 1.351e+03, threshold=7.500e+02, percent-clipped=6.0 2023-03-07 23:01:24,436 INFO [train2.py:809] (1/4) Epoch 8, batch 850, loss[ctc_loss=0.1262, att_loss=0.2647, loss=0.237, over 17399.00 frames. utt_duration=1106 frames, utt_pad_proportion=0.03225, over 63.00 utterances.], tot_loss[ctc_loss=0.1272, att_loss=0.2622, loss=0.2352, over 3217410.51 frames. utt_duration=1203 frames, utt_pad_proportion=0.06795, over 10709.58 utterances.], batch size: 63, lr: 1.38e-02, grad_scale: 8.0 2023-03-07 23:01:38,875 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.95 vs. limit=2.0 2023-03-07 23:02:43,216 INFO [train2.py:809] (1/4) Epoch 8, batch 900, loss[ctc_loss=0.09006, att_loss=0.2305, loss=0.2024, over 14458.00 frames. utt_duration=1809 frames, utt_pad_proportion=0.04696, over 32.00 utterances.], tot_loss[ctc_loss=0.1268, att_loss=0.2618, loss=0.2348, over 3220839.84 frames. utt_duration=1227 frames, utt_pad_proportion=0.06449, over 10516.51 utterances.], batch size: 32, lr: 1.38e-02, grad_scale: 8.0 2023-03-07 23:03:27,129 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.754e+02 2.976e+02 3.721e+02 4.391e+02 9.491e+02, threshold=7.443e+02, percent-clipped=5.0 2023-03-07 23:04:03,682 INFO [train2.py:809] (1/4) Epoch 8, batch 950, loss[ctc_loss=0.09416, att_loss=0.2522, loss=0.2206, over 16334.00 frames. utt_duration=1453 frames, utt_pad_proportion=0.005943, over 45.00 utterances.], tot_loss[ctc_loss=0.1272, att_loss=0.2625, loss=0.2355, over 3231769.13 frames. utt_duration=1224 frames, utt_pad_proportion=0.06257, over 10572.73 utterances.], batch size: 45, lr: 1.38e-02, grad_scale: 8.0 2023-03-07 23:04:11,995 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28842.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:04:21,448 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28848.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:04:23,306 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1524, 4.9546, 5.0223, 2.2797, 1.9242, 2.5903, 3.6898, 3.7888], device='cuda:1'), covar=tensor([0.0566, 0.0161, 0.0165, 0.3352, 0.5904, 0.2707, 0.1039, 0.1688], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0196, 0.0216, 0.0179, 0.0347, 0.0330, 0.0210, 0.0338], device='cuda:1'), out_proj_covar=tensor([1.4806e-04, 7.6887e-05, 9.5054e-05, 8.2954e-05, 1.5859e-04, 1.4068e-04, 8.5520e-05, 1.5135e-04], device='cuda:1') 2023-03-07 23:04:42,032 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28861.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:05:21,455 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28885.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:05:24,288 INFO [train2.py:809] (1/4) Epoch 8, batch 1000, loss[ctc_loss=0.1348, att_loss=0.2712, loss=0.244, over 16760.00 frames. utt_duration=1398 frames, utt_pad_proportion=0.007058, over 48.00 utterances.], tot_loss[ctc_loss=0.1267, att_loss=0.2626, loss=0.2354, over 3247888.25 frames. utt_duration=1237 frames, utt_pad_proportion=0.05733, over 10514.19 utterances.], batch size: 48, lr: 1.38e-02, grad_scale: 8.0 2023-03-07 23:05:26,152 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6323, 5.2198, 4.8460, 5.2703, 5.3830, 4.7757, 3.6790, 5.1566], device='cuda:1'), covar=tensor([0.0101, 0.0082, 0.0124, 0.0068, 0.0048, 0.0100, 0.0510, 0.0173], device='cuda:1'), in_proj_covar=tensor([0.0066, 0.0060, 0.0072, 0.0047, 0.0048, 0.0059, 0.0083, 0.0081], 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') 2023-03-07 23:05:37,285 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 2023-03-07 23:05:58,550 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6592, 2.1340, 5.0173, 3.8606, 3.0322, 4.6205, 4.7350, 4.7654], device='cuda:1'), covar=tensor([0.0169, 0.1894, 0.0116, 0.1084, 0.1937, 0.0176, 0.0087, 0.0163], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0239, 0.0121, 0.0299, 0.0277, 0.0180, 0.0104, 0.0140], device='cuda:1'), out_proj_covar=tensor([1.3329e-04, 1.9845e-04, 1.0870e-04, 2.4555e-04, 2.4344e-04, 1.6015e-04, 9.4241e-05, 1.2935e-04], device='cuda:1') 2023-03-07 23:06:01,431 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28911.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 23:06:07,294 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.707e+02 2.634e+02 3.393e+02 4.335e+02 1.819e+03, threshold=6.785e+02, percent-clipped=6.0 2023-03-07 23:06:20,091 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28922.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:06:36,981 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=28933.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:06:43,110 INFO [train2.py:809] (1/4) Epoch 8, batch 1050, loss[ctc_loss=0.106, att_loss=0.2595, loss=0.2288, over 16783.00 frames. utt_duration=1400 frames, utt_pad_proportion=0.005726, over 48.00 utterances.], tot_loss[ctc_loss=0.1257, att_loss=0.2615, loss=0.2343, over 3253158.63 frames. utt_duration=1274 frames, utt_pad_proportion=0.04925, over 10223.86 utterances.], batch size: 48, lr: 1.38e-02, grad_scale: 8.0 2023-03-07 23:06:53,371 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28943.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:07:01,112 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9517, 5.3558, 4.7174, 5.3567, 4.6526, 4.9662, 5.4530, 5.2726], device='cuda:1'), covar=tensor([0.0468, 0.0216, 0.0812, 0.0206, 0.0402, 0.0224, 0.0165, 0.0150], device='cuda:1'), in_proj_covar=tensor([0.0278, 0.0211, 0.0274, 0.0198, 0.0225, 0.0174, 0.0199, 0.0196], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0005, 0.0005, 0.0005, 0.0004, 0.0004, 0.0005, 0.0004], device='cuda:1') 2023-03-07 23:07:18,000 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=28959.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 23:07:25,920 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7146, 3.2791, 3.8376, 3.1712, 3.5939, 4.7442, 4.5274, 3.7370], device='cuda:1'), covar=tensor([0.0309, 0.1395, 0.0895, 0.1257, 0.0908, 0.0595, 0.0449, 0.0997], device='cuda:1'), in_proj_covar=tensor([0.0214, 0.0213, 0.0221, 0.0196, 0.0222, 0.0254, 0.0193, 0.0205], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-03-07 23:08:03,727 INFO [train2.py:809] (1/4) Epoch 8, batch 1100, loss[ctc_loss=0.09652, att_loss=0.2245, loss=0.1989, over 15772.00 frames. utt_duration=1662 frames, utt_pad_proportion=0.008589, over 38.00 utterances.], tot_loss[ctc_loss=0.1251, att_loss=0.261, loss=0.2338, over 3259357.13 frames. utt_duration=1277 frames, utt_pad_proportion=0.04815, over 10224.62 utterances.], batch size: 38, lr: 1.38e-02, grad_scale: 8.0 2023-03-07 23:08:10,585 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=28991.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:08:28,614 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29002.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:08:48,757 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.591e+02 2.718e+02 3.264e+02 3.994e+02 7.593e+02, threshold=6.528e+02, percent-clipped=2.0 2023-03-07 23:09:08,377 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.6801, 5.9501, 5.3310, 5.7727, 5.6183, 5.2491, 5.3524, 5.0719], device='cuda:1'), covar=tensor([0.1185, 0.0835, 0.0828, 0.0658, 0.0702, 0.1220, 0.2230, 0.2558], device='cuda:1'), in_proj_covar=tensor([0.0380, 0.0436, 0.0336, 0.0343, 0.0322, 0.0387, 0.0465, 0.0417], 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') 2023-03-07 23:09:25,644 INFO [train2.py:809] (1/4) Epoch 8, batch 1150, loss[ctc_loss=0.1085, att_loss=0.2528, loss=0.224, over 16770.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.005641, over 48.00 utterances.], tot_loss[ctc_loss=0.124, att_loss=0.2608, loss=0.2334, over 3265521.70 frames. utt_duration=1270 frames, utt_pad_proportion=0.04843, over 10295.55 utterances.], batch size: 48, lr: 1.38e-02, grad_scale: 8.0 2023-03-07 23:10:06,623 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29063.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:10:18,143 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1980, 5.2418, 5.1034, 2.3555, 1.9695, 2.9326, 3.9651, 3.9640], device='cuda:1'), covar=tensor([0.0561, 0.0185, 0.0185, 0.4074, 0.6283, 0.2340, 0.0927, 0.1742], device='cuda:1'), in_proj_covar=tensor([0.0318, 0.0201, 0.0225, 0.0184, 0.0359, 0.0342, 0.0217, 0.0346], device='cuda:1'), out_proj_covar=tensor([1.5160e-04, 7.8418e-05, 9.9001e-05, 8.5732e-05, 1.6324e-04, 1.4530e-04, 8.7780e-05, 1.5484e-04], device='cuda:1') 2023-03-07 23:10:46,212 INFO [train2.py:809] (1/4) Epoch 8, batch 1200, loss[ctc_loss=0.1033, att_loss=0.229, loss=0.2038, over 15404.00 frames. utt_duration=1762 frames, utt_pad_proportion=0.008538, over 35.00 utterances.], tot_loss[ctc_loss=0.1239, att_loss=0.2603, loss=0.233, over 3252473.51 frames. utt_duration=1245 frames, utt_pad_proportion=0.05822, over 10466.29 utterances.], batch size: 35, lr: 1.38e-02, grad_scale: 8.0 2023-03-07 23:10:54,160 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7930, 5.2158, 4.9125, 5.0766, 5.2532, 5.2033, 4.9194, 4.6815], device='cuda:1'), covar=tensor([0.1098, 0.0464, 0.0353, 0.0512, 0.0328, 0.0318, 0.0257, 0.0336], device='cuda:1'), in_proj_covar=tensor([0.0425, 0.0260, 0.0201, 0.0240, 0.0301, 0.0330, 0.0249, 0.0282], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-07 23:11:30,217 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.598e+02 2.794e+02 3.105e+02 3.921e+02 8.857e+02, threshold=6.210e+02, percent-clipped=2.0 2023-03-07 23:12:01,757 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9609, 5.3894, 4.7391, 5.4201, 4.7216, 5.1129, 5.4742, 5.2666], device='cuda:1'), covar=tensor([0.0508, 0.0267, 0.0898, 0.0222, 0.0431, 0.0204, 0.0254, 0.0167], device='cuda:1'), in_proj_covar=tensor([0.0277, 0.0211, 0.0276, 0.0200, 0.0227, 0.0174, 0.0201, 0.0196], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0005, 0.0005, 0.0005, 0.0004, 0.0004, 0.0005, 0.0004], device='cuda:1') 2023-03-07 23:12:06,183 INFO [train2.py:809] (1/4) Epoch 8, batch 1250, loss[ctc_loss=0.1032, att_loss=0.2515, loss=0.2219, over 16264.00 frames. utt_duration=1514 frames, utt_pad_proportion=0.007635, over 43.00 utterances.], tot_loss[ctc_loss=0.1245, att_loss=0.2604, loss=0.2332, over 3243216.78 frames. utt_duration=1224 frames, utt_pad_proportion=0.06696, over 10612.53 utterances.], batch size: 43, lr: 1.38e-02, grad_scale: 8.0 2023-03-07 23:12:06,589 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3113, 4.8637, 4.4287, 4.6528, 2.4519, 4.6479, 2.9918, 2.0381], device='cuda:1'), covar=tensor([0.0281, 0.0112, 0.0708, 0.0170, 0.2024, 0.0144, 0.1414, 0.1689], device='cuda:1'), in_proj_covar=tensor([0.0121, 0.0099, 0.0251, 0.0113, 0.0217, 0.0101, 0.0224, 0.0200], device='cuda:1'), out_proj_covar=tensor([1.2135e-04, 1.0245e-04, 2.2897e-04, 1.0704e-04, 2.0581e-04, 9.9895e-05, 2.0373e-04, 1.8297e-04], device='cuda:1') 2023-03-07 23:12:14,237 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29142.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:12:23,499 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29148.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:13:13,960 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-03-07 23:13:25,931 INFO [train2.py:809] (1/4) Epoch 8, batch 1300, loss[ctc_loss=0.1353, att_loss=0.2835, loss=0.2539, over 17337.00 frames. utt_duration=1177 frames, utt_pad_proportion=0.02276, over 59.00 utterances.], tot_loss[ctc_loss=0.1231, att_loss=0.2597, loss=0.2323, over 3247801.80 frames. utt_duration=1228 frames, utt_pad_proportion=0.06527, over 10593.45 utterances.], batch size: 59, lr: 1.37e-02, grad_scale: 8.0 2023-03-07 23:13:30,546 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=29190.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:13:39,855 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=29196.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:13:45,292 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.85 vs. limit=2.0 2023-03-07 23:14:09,582 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.817e+02 2.738e+02 3.321e+02 4.360e+02 1.126e+03, threshold=6.642e+02, percent-clipped=8.0 2023-03-07 23:14:13,601 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29217.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:14:44,874 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29236.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:14:46,055 INFO [train2.py:809] (1/4) Epoch 8, batch 1350, loss[ctc_loss=0.1221, att_loss=0.2705, loss=0.2408, over 17054.00 frames. utt_duration=1313 frames, utt_pad_proportion=0.008708, over 52.00 utterances.], tot_loss[ctc_loss=0.1249, att_loss=0.2614, loss=0.2341, over 3257587.56 frames. utt_duration=1203 frames, utt_pad_proportion=0.06827, over 10844.70 utterances.], batch size: 52, lr: 1.37e-02, grad_scale: 8.0 2023-03-07 23:14:46,949 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.71 vs. limit=5.0 2023-03-07 23:14:47,414 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-03-07 23:15:41,852 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-03-07 23:16:06,780 INFO [train2.py:809] (1/4) Epoch 8, batch 1400, loss[ctc_loss=0.1241, att_loss=0.2744, loss=0.2443, over 17467.00 frames. utt_duration=1014 frames, utt_pad_proportion=0.0442, over 69.00 utterances.], tot_loss[ctc_loss=0.1248, att_loss=0.2612, loss=0.2339, over 3266538.86 frames. utt_duration=1220 frames, utt_pad_proportion=0.06244, over 10720.69 utterances.], batch size: 69, lr: 1.37e-02, grad_scale: 8.0 2023-03-07 23:16:23,182 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29297.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:16:51,999 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.467e+02 2.810e+02 3.377e+02 4.369e+02 8.474e+02, threshold=6.754e+02, percent-clipped=4.0 2023-03-07 23:17:14,610 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7775, 4.7821, 4.5952, 4.7745, 5.2334, 5.2062, 4.5663, 2.3442], device='cuda:1'), covar=tensor([0.0213, 0.0237, 0.0250, 0.0156, 0.0976, 0.0148, 0.0277, 0.2312], device='cuda:1'), in_proj_covar=tensor([0.0130, 0.0122, 0.0128, 0.0125, 0.0309, 0.0124, 0.0114, 0.0227], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-07 23:17:27,562 INFO [train2.py:809] (1/4) Epoch 8, batch 1450, loss[ctc_loss=0.1922, att_loss=0.2975, loss=0.2764, over 13931.00 frames. utt_duration=385.7 frames, utt_pad_proportion=0.3304, over 145.00 utterances.], tot_loss[ctc_loss=0.1254, att_loss=0.2613, loss=0.2341, over 3261446.71 frames. utt_duration=1200 frames, utt_pad_proportion=0.06935, over 10880.61 utterances.], batch size: 145, lr: 1.37e-02, grad_scale: 8.0 2023-03-07 23:17:56,005 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.2681, 5.0996, 4.9917, 2.6945, 4.9931, 4.5666, 4.4956, 2.4147], device='cuda:1'), covar=tensor([0.0076, 0.0088, 0.0180, 0.1013, 0.0078, 0.0155, 0.0258, 0.1389], device='cuda:1'), in_proj_covar=tensor([0.0054, 0.0073, 0.0061, 0.0098, 0.0063, 0.0081, 0.0084, 0.0097], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-07 23:18:00,628 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29358.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:18:47,772 INFO [train2.py:809] (1/4) Epoch 8, batch 1500, loss[ctc_loss=0.09726, att_loss=0.2445, loss=0.2151, over 16625.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.005319, over 47.00 utterances.], tot_loss[ctc_loss=0.1239, att_loss=0.2606, loss=0.2333, over 3263382.18 frames. utt_duration=1233 frames, utt_pad_proportion=0.0605, over 10601.09 utterances.], batch size: 47, lr: 1.37e-02, grad_scale: 8.0 2023-03-07 23:19:30,619 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.613e+02 2.592e+02 3.615e+02 4.521e+02 1.886e+03, threshold=7.229e+02, percent-clipped=4.0 2023-03-07 23:19:55,167 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.77 vs. limit=2.0 2023-03-07 23:20:07,836 INFO [train2.py:809] (1/4) Epoch 8, batch 1550, loss[ctc_loss=0.1146, att_loss=0.2373, loss=0.2127, over 15764.00 frames. utt_duration=1661 frames, utt_pad_proportion=0.009028, over 38.00 utterances.], tot_loss[ctc_loss=0.1246, att_loss=0.2605, loss=0.2334, over 3259423.59 frames. utt_duration=1229 frames, utt_pad_proportion=0.06314, over 10622.96 utterances.], batch size: 38, lr: 1.37e-02, grad_scale: 8.0 2023-03-07 23:21:12,283 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4883, 2.6541, 5.0140, 3.8952, 2.9033, 4.4956, 4.8054, 4.6843], device='cuda:1'), covar=tensor([0.0254, 0.1667, 0.0141, 0.0935, 0.2002, 0.0211, 0.0103, 0.0253], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0242, 0.0122, 0.0298, 0.0280, 0.0184, 0.0107, 0.0143], device='cuda:1'), out_proj_covar=tensor([1.3725e-04, 2.0173e-04, 1.0905e-04, 2.4604e-04, 2.4622e-04, 1.6271e-04, 9.6383e-05, 1.3266e-04], device='cuda:1') 2023-03-07 23:21:28,331 INFO [train2.py:809] (1/4) Epoch 8, batch 1600, loss[ctc_loss=0.1571, att_loss=0.2869, loss=0.2609, over 17048.00 frames. utt_duration=1313 frames, utt_pad_proportion=0.008381, over 52.00 utterances.], tot_loss[ctc_loss=0.1241, att_loss=0.261, loss=0.2336, over 3261298.22 frames. utt_duration=1223 frames, utt_pad_proportion=0.0629, over 10677.05 utterances.], batch size: 52, lr: 1.37e-02, grad_scale: 8.0 2023-03-07 23:22:11,636 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.666e+02 2.839e+02 3.566e+02 4.248e+02 8.470e+02, threshold=7.133e+02, percent-clipped=2.0 2023-03-07 23:22:15,565 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29517.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:22:48,443 INFO [train2.py:809] (1/4) Epoch 8, batch 1650, loss[ctc_loss=0.1355, att_loss=0.2721, loss=0.2448, over 17262.00 frames. utt_duration=1002 frames, utt_pad_proportion=0.05536, over 69.00 utterances.], tot_loss[ctc_loss=0.1246, att_loss=0.261, loss=0.2337, over 3264474.28 frames. utt_duration=1209 frames, utt_pad_proportion=0.06574, over 10815.95 utterances.], batch size: 69, lr: 1.37e-02, grad_scale: 8.0 2023-03-07 23:23:10,607 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29551.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:23:32,755 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=29565.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:24:08,551 INFO [train2.py:809] (1/4) Epoch 8, batch 1700, loss[ctc_loss=0.1247, att_loss=0.2679, loss=0.2393, over 17405.00 frames. utt_duration=1106 frames, utt_pad_proportion=0.03201, over 63.00 utterances.], tot_loss[ctc_loss=0.1257, att_loss=0.262, loss=0.2348, over 3268968.98 frames. utt_duration=1189 frames, utt_pad_proportion=0.06998, over 11008.67 utterances.], batch size: 63, lr: 1.36e-02, grad_scale: 8.0 2023-03-07 23:24:16,289 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29592.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:24:47,180 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29612.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:24:52,437 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.13 vs. limit=5.0 2023-03-07 23:24:52,831 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.894e+02 2.662e+02 3.188e+02 4.272e+02 1.218e+03, threshold=6.376e+02, percent-clipped=1.0 2023-03-07 23:25:28,579 INFO [train2.py:809] (1/4) Epoch 8, batch 1750, loss[ctc_loss=0.1196, att_loss=0.2679, loss=0.2382, over 16541.00 frames. utt_duration=1472 frames, utt_pad_proportion=0.006212, over 45.00 utterances.], tot_loss[ctc_loss=0.1254, att_loss=0.2618, loss=0.2346, over 3270499.69 frames. utt_duration=1223 frames, utt_pad_proportion=0.06082, over 10706.91 utterances.], batch size: 45, lr: 1.36e-02, grad_scale: 8.0 2023-03-07 23:26:01,585 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29658.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:26:07,747 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29662.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:26:24,164 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-03-07 23:26:48,812 INFO [train2.py:809] (1/4) Epoch 8, batch 1800, loss[ctc_loss=0.1144, att_loss=0.2476, loss=0.2209, over 16393.00 frames. utt_duration=1492 frames, utt_pad_proportion=0.00693, over 44.00 utterances.], tot_loss[ctc_loss=0.1251, att_loss=0.2615, loss=0.2342, over 3262863.72 frames. utt_duration=1218 frames, utt_pad_proportion=0.06475, over 10732.01 utterances.], batch size: 44, lr: 1.36e-02, grad_scale: 8.0 2023-03-07 23:27:19,005 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=29706.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:27:33,856 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.769e+02 3.006e+02 3.671e+02 4.359e+02 7.069e+02, threshold=7.341e+02, percent-clipped=3.0 2023-03-07 23:27:47,104 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29723.0, num_to_drop=1, layers_to_drop={3} 2023-03-07 23:28:08,902 INFO [train2.py:809] (1/4) Epoch 8, batch 1850, loss[ctc_loss=0.1203, att_loss=0.2302, loss=0.2082, over 15646.00 frames. utt_duration=1693 frames, utt_pad_proportion=0.008703, over 37.00 utterances.], tot_loss[ctc_loss=0.1249, att_loss=0.2613, loss=0.2341, over 3263629.36 frames. utt_duration=1216 frames, utt_pad_proportion=0.06429, over 10748.45 utterances.], batch size: 37, lr: 1.36e-02, grad_scale: 8.0 2023-03-07 23:29:29,188 INFO [train2.py:809] (1/4) Epoch 8, batch 1900, loss[ctc_loss=0.1193, att_loss=0.2715, loss=0.2411, over 16759.00 frames. utt_duration=1398 frames, utt_pad_proportion=0.00688, over 48.00 utterances.], tot_loss[ctc_loss=0.1233, att_loss=0.2607, loss=0.2332, over 3256497.48 frames. utt_duration=1241 frames, utt_pad_proportion=0.05974, over 10512.14 utterances.], batch size: 48, lr: 1.36e-02, grad_scale: 8.0 2023-03-07 23:29:53,379 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29802.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:30:02,546 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.6512, 3.6836, 3.0912, 3.2295, 3.8946, 3.4357, 2.8370, 4.2815], device='cuda:1'), covar=tensor([0.1117, 0.0496, 0.1071, 0.0764, 0.0598, 0.0704, 0.0929, 0.0413], device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0160, 0.0193, 0.0166, 0.0205, 0.0195, 0.0168, 0.0217], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-07 23:30:14,329 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.846e+02 2.848e+02 3.489e+02 4.183e+02 9.529e+02, threshold=6.977e+02, percent-clipped=3.0 2023-03-07 23:30:49,804 INFO [train2.py:809] (1/4) Epoch 8, batch 1950, loss[ctc_loss=0.1205, att_loss=0.2733, loss=0.2427, over 16860.00 frames. utt_duration=1378 frames, utt_pad_proportion=0.00719, over 49.00 utterances.], tot_loss[ctc_loss=0.1243, att_loss=0.2609, loss=0.2335, over 3255906.12 frames. utt_duration=1237 frames, utt_pad_proportion=0.06026, over 10544.73 utterances.], batch size: 49, lr: 1.36e-02, grad_scale: 8.0 2023-03-07 23:31:32,140 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29863.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:32:10,085 INFO [train2.py:809] (1/4) Epoch 8, batch 2000, loss[ctc_loss=0.128, att_loss=0.2695, loss=0.2412, over 16630.00 frames. utt_duration=1417 frames, utt_pad_proportion=0.005095, over 47.00 utterances.], tot_loss[ctc_loss=0.1257, att_loss=0.2618, loss=0.2346, over 3263160.54 frames. utt_duration=1201 frames, utt_pad_proportion=0.06729, over 10883.15 utterances.], batch size: 47, lr: 1.36e-02, grad_scale: 8.0 2023-03-07 23:32:18,042 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29892.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:32:41,260 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29907.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:32:54,605 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.764e+02 2.836e+02 3.233e+02 3.942e+02 8.276e+02, threshold=6.465e+02, percent-clipped=1.0 2023-03-07 23:33:30,289 INFO [train2.py:809] (1/4) Epoch 8, batch 2050, loss[ctc_loss=0.1271, att_loss=0.2605, loss=0.2338, over 16284.00 frames. utt_duration=1516 frames, utt_pad_proportion=0.006507, over 43.00 utterances.], tot_loss[ctc_loss=0.1253, att_loss=0.262, loss=0.2347, over 3259846.09 frames. utt_duration=1206 frames, utt_pad_proportion=0.06717, over 10825.13 utterances.], batch size: 43, lr: 1.36e-02, grad_scale: 8.0 2023-03-07 23:33:34,848 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=29940.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:34:21,296 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1768, 4.8675, 4.6641, 4.4731, 2.5375, 4.8774, 2.2681, 1.7422], device='cuda:1'), covar=tensor([0.0278, 0.0091, 0.0579, 0.0214, 0.1946, 0.0120, 0.1856, 0.1840], device='cuda:1'), in_proj_covar=tensor([0.0122, 0.0098, 0.0250, 0.0110, 0.0220, 0.0099, 0.0226, 0.0198], device='cuda:1'), out_proj_covar=tensor([1.2221e-04, 1.0183e-04, 2.2801e-04, 1.0487e-04, 2.0835e-04, 9.7778e-05, 2.0524e-04, 1.8134e-04], device='cuda:1') 2023-03-07 23:34:26,051 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0960, 4.7629, 4.4420, 4.4356, 2.4078, 4.6483, 2.5085, 1.5026], device='cuda:1'), covar=tensor([0.0303, 0.0096, 0.0709, 0.0213, 0.2248, 0.0149, 0.1761, 0.2105], device='cuda:1'), in_proj_covar=tensor([0.0122, 0.0098, 0.0250, 0.0110, 0.0220, 0.0099, 0.0226, 0.0198], device='cuda:1'), out_proj_covar=tensor([1.2217e-04, 1.0180e-04, 2.2816e-04, 1.0477e-04, 2.0827e-04, 9.7828e-05, 2.0539e-04, 1.8137e-04], device='cuda:1') 2023-03-07 23:34:51,167 INFO [train2.py:809] (1/4) Epoch 8, batch 2100, loss[ctc_loss=0.1179, att_loss=0.239, loss=0.2148, over 15365.00 frames. utt_duration=1758 frames, utt_pad_proportion=0.01135, over 35.00 utterances.], tot_loss[ctc_loss=0.125, att_loss=0.2623, loss=0.2348, over 3263386.97 frames. utt_duration=1193 frames, utt_pad_proportion=0.07054, over 10958.70 utterances.], batch size: 35, lr: 1.36e-02, grad_scale: 8.0 2023-03-07 23:35:40,751 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-03-07 23:35:43,065 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.761e+02 2.671e+02 3.420e+02 4.365e+02 1.034e+03, threshold=6.841e+02, percent-clipped=5.0 2023-03-07 23:35:47,043 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30018.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 23:36:16,108 INFO [train2.py:809] (1/4) Epoch 8, batch 2150, loss[ctc_loss=0.1299, att_loss=0.2575, loss=0.2319, over 16257.00 frames. utt_duration=1514 frames, utt_pad_proportion=0.008681, over 43.00 utterances.], tot_loss[ctc_loss=0.1255, att_loss=0.2627, loss=0.2352, over 3279063.73 frames. utt_duration=1201 frames, utt_pad_proportion=0.06338, over 10933.64 utterances.], batch size: 43, lr: 1.35e-02, grad_scale: 8.0 2023-03-07 23:36:18,684 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.79 vs. limit=5.0 2023-03-07 23:36:26,724 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.78 vs. limit=2.0 2023-03-07 23:37:20,694 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4805, 4.8756, 4.7036, 4.7718, 4.7901, 4.5218, 3.5676, 4.6407], device='cuda:1'), covar=tensor([0.0095, 0.0103, 0.0099, 0.0094, 0.0111, 0.0110, 0.0594, 0.0195], device='cuda:1'), in_proj_covar=tensor([0.0066, 0.0062, 0.0072, 0.0047, 0.0049, 0.0058, 0.0083, 0.0081], 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') 2023-03-07 23:37:36,074 INFO [train2.py:809] (1/4) Epoch 8, batch 2200, loss[ctc_loss=0.1644, att_loss=0.283, loss=0.2592, over 17400.00 frames. utt_duration=1010 frames, utt_pad_proportion=0.04718, over 69.00 utterances.], tot_loss[ctc_loss=0.1252, att_loss=0.2624, loss=0.235, over 3273716.98 frames. utt_duration=1220 frames, utt_pad_proportion=0.06063, over 10745.39 utterances.], batch size: 69, lr: 1.35e-02, grad_scale: 8.0 2023-03-07 23:38:22,431 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.913e+02 2.675e+02 3.232e+02 4.079e+02 9.910e+02, threshold=6.464e+02, percent-clipped=1.0 2023-03-07 23:38:37,607 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30125.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:38:55,633 INFO [train2.py:809] (1/4) Epoch 8, batch 2250, loss[ctc_loss=0.1108, att_loss=0.2606, loss=0.2306, over 16687.00 frames. utt_duration=1452 frames, utt_pad_proportion=0.006026, over 46.00 utterances.], tot_loss[ctc_loss=0.1247, att_loss=0.2623, loss=0.2348, over 3278834.79 frames. utt_duration=1224 frames, utt_pad_proportion=0.05785, over 10726.88 utterances.], batch size: 46, lr: 1.35e-02, grad_scale: 8.0 2023-03-07 23:39:00,731 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6456, 2.3889, 4.9615, 3.7602, 2.7808, 4.4641, 4.6571, 4.5892], device='cuda:1'), covar=tensor([0.0196, 0.1808, 0.0125, 0.1122, 0.2143, 0.0208, 0.0119, 0.0232], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0237, 0.0125, 0.0299, 0.0281, 0.0180, 0.0108, 0.0144], device='cuda:1'), out_proj_covar=tensor([1.3631e-04, 1.9795e-04, 1.1063e-04, 2.4756e-04, 2.4801e-04, 1.5926e-04, 9.7718e-05, 1.3408e-04], device='cuda:1') 2023-03-07 23:39:28,583 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9958, 5.1487, 5.5356, 5.3494, 5.2999, 5.9032, 5.1501, 6.0191], device='cuda:1'), covar=tensor([0.0545, 0.0607, 0.0534, 0.0898, 0.1696, 0.0760, 0.0504, 0.0493], device='cuda:1'), in_proj_covar=tensor([0.0614, 0.0371, 0.0427, 0.0484, 0.0661, 0.0430, 0.0339, 0.0420], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-03-07 23:39:28,598 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30158.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:40:13,313 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30186.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:40:14,496 INFO [train2.py:809] (1/4) Epoch 8, batch 2300, loss[ctc_loss=0.1273, att_loss=0.2683, loss=0.2401, over 16324.00 frames. utt_duration=1453 frames, utt_pad_proportion=0.006445, over 45.00 utterances.], tot_loss[ctc_loss=0.1256, att_loss=0.2628, loss=0.2354, over 3279787.17 frames. utt_duration=1197 frames, utt_pad_proportion=0.06481, over 10975.10 utterances.], batch size: 45, lr: 1.35e-02, grad_scale: 8.0 2023-03-07 23:40:46,668 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30207.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:40:50,433 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30209.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:41:00,842 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.528e+02 2.649e+02 3.337e+02 4.141e+02 1.009e+03, threshold=6.675e+02, percent-clipped=1.0 2023-03-07 23:41:34,431 INFO [train2.py:809] (1/4) Epoch 8, batch 2350, loss[ctc_loss=0.1145, att_loss=0.2635, loss=0.2337, over 16768.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.006496, over 48.00 utterances.], tot_loss[ctc_loss=0.1253, att_loss=0.2627, loss=0.2352, over 3288472.01 frames. utt_duration=1204 frames, utt_pad_proportion=0.06036, over 10937.67 utterances.], batch size: 48, lr: 1.35e-02, grad_scale: 8.0 2023-03-07 23:42:02,810 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=30255.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:42:15,076 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.25 vs. limit=5.0 2023-03-07 23:42:28,177 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30270.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:42:54,198 INFO [train2.py:809] (1/4) Epoch 8, batch 2400, loss[ctc_loss=0.1495, att_loss=0.287, loss=0.2595, over 17024.00 frames. utt_duration=1287 frames, utt_pad_proportion=0.009586, over 53.00 utterances.], tot_loss[ctc_loss=0.1252, att_loss=0.2631, loss=0.2355, over 3288832.52 frames. utt_duration=1213 frames, utt_pad_proportion=0.05869, over 10860.42 utterances.], batch size: 53, lr: 1.35e-02, grad_scale: 8.0 2023-03-07 23:43:40,599 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.713e+02 2.833e+02 3.441e+02 4.475e+02 1.318e+03, threshold=6.883e+02, percent-clipped=7.0 2023-03-07 23:43:44,647 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30318.0, num_to_drop=1, layers_to_drop={0} 2023-03-07 23:44:13,410 INFO [train2.py:809] (1/4) Epoch 8, batch 2450, loss[ctc_loss=0.1404, att_loss=0.2575, loss=0.2341, over 15955.00 frames. utt_duration=1558 frames, utt_pad_proportion=0.006378, over 41.00 utterances.], tot_loss[ctc_loss=0.1245, att_loss=0.2623, loss=0.2347, over 3286769.61 frames. utt_duration=1224 frames, utt_pad_proportion=0.05736, over 10755.42 utterances.], batch size: 41, lr: 1.35e-02, grad_scale: 8.0 2023-03-07 23:44:50,954 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30360.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:45:00,763 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=30366.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:45:34,118 INFO [train2.py:809] (1/4) Epoch 8, batch 2500, loss[ctc_loss=0.145, att_loss=0.2809, loss=0.2537, over 17314.00 frames. utt_duration=1175 frames, utt_pad_proportion=0.02288, over 59.00 utterances.], tot_loss[ctc_loss=0.1238, att_loss=0.2614, loss=0.2339, over 3278766.77 frames. utt_duration=1242 frames, utt_pad_proportion=0.05429, over 10571.70 utterances.], batch size: 59, lr: 1.35e-02, grad_scale: 8.0 2023-03-07 23:46:21,433 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.959e+02 2.650e+02 3.302e+02 4.215e+02 1.043e+03, threshold=6.603e+02, percent-clipped=3.0 2023-03-07 23:46:30,129 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30421.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:46:54,580 INFO [train2.py:809] (1/4) Epoch 8, batch 2550, loss[ctc_loss=0.1331, att_loss=0.2572, loss=0.2324, over 16141.00 frames. utt_duration=1539 frames, utt_pad_proportion=0.004728, over 42.00 utterances.], tot_loss[ctc_loss=0.1247, att_loss=0.262, loss=0.2345, over 3278934.00 frames. utt_duration=1227 frames, utt_pad_proportion=0.05739, over 10703.16 utterances.], batch size: 42, lr: 1.35e-02, grad_scale: 8.0 2023-03-07 23:47:18,687 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5570, 2.6177, 3.6366, 2.7776, 3.4937, 4.5852, 4.2715, 3.2967], device='cuda:1'), covar=tensor([0.0333, 0.1754, 0.1051, 0.1368, 0.0907, 0.0571, 0.0524, 0.1234], device='cuda:1'), in_proj_covar=tensor([0.0218, 0.0217, 0.0228, 0.0196, 0.0224, 0.0259, 0.0197, 0.0208], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-03-07 23:47:28,433 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30458.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:48:04,550 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30481.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:48:13,627 INFO [train2.py:809] (1/4) Epoch 8, batch 2600, loss[ctc_loss=0.1853, att_loss=0.2979, loss=0.2754, over 14448.00 frames. utt_duration=397.4 frames, utt_pad_proportion=0.3076, over 146.00 utterances.], tot_loss[ctc_loss=0.1249, att_loss=0.2616, loss=0.2343, over 3270517.54 frames. utt_duration=1212 frames, utt_pad_proportion=0.06151, over 10806.65 utterances.], batch size: 146, lr: 1.35e-02, grad_scale: 8.0 2023-03-07 23:48:18,687 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.4880, 1.6902, 1.9541, 1.8626, 3.1098, 1.7684, 1.8704, 2.4532], device='cuda:1'), covar=tensor([0.0513, 0.4754, 0.4093, 0.1491, 0.0635, 0.1854, 0.2554, 0.1372], device='cuda:1'), in_proj_covar=tensor([0.0079, 0.0087, 0.0090, 0.0077, 0.0075, 0.0071, 0.0082, 0.0068], device='cuda:1'), out_proj_covar=tensor([4.2488e-05, 5.4279e-05, 5.4701e-05, 4.5323e-05, 4.1037e-05, 4.5702e-05, 5.1480e-05, 4.4322e-05], device='cuda:1') 2023-03-07 23:48:44,080 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=30506.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:49:00,818 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.127e+02 2.736e+02 3.358e+02 3.851e+02 8.673e+02, threshold=6.715e+02, percent-clipped=1.0 2023-03-07 23:49:33,582 INFO [train2.py:809] (1/4) Epoch 8, batch 2650, loss[ctc_loss=0.1153, att_loss=0.2459, loss=0.2198, over 15517.00 frames. utt_duration=1726 frames, utt_pad_proportion=0.007747, over 36.00 utterances.], tot_loss[ctc_loss=0.1236, att_loss=0.2608, loss=0.2334, over 3262693.63 frames. utt_duration=1219 frames, utt_pad_proportion=0.06287, over 10715.81 utterances.], batch size: 36, lr: 1.34e-02, grad_scale: 8.0 2023-03-07 23:49:50,151 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2023-03-07 23:50:18,929 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30565.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:50:32,697 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-03-07 23:50:52,903 INFO [train2.py:809] (1/4) Epoch 8, batch 2700, loss[ctc_loss=0.1425, att_loss=0.2498, loss=0.2283, over 15356.00 frames. utt_duration=1756 frames, utt_pad_proportion=0.01167, over 35.00 utterances.], tot_loss[ctc_loss=0.1234, att_loss=0.2609, loss=0.2334, over 3271428.00 frames. utt_duration=1219 frames, utt_pad_proportion=0.06166, over 10744.45 utterances.], batch size: 35, lr: 1.34e-02, grad_scale: 8.0 2023-03-07 23:51:21,835 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-03-07 23:51:39,696 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.853e+02 2.974e+02 3.669e+02 4.665e+02 1.299e+03, threshold=7.338e+02, percent-clipped=5.0 2023-03-07 23:52:12,703 INFO [train2.py:809] (1/4) Epoch 8, batch 2750, loss[ctc_loss=0.1649, att_loss=0.2886, loss=0.2638, over 16919.00 frames. utt_duration=685.1 frames, utt_pad_proportion=0.1371, over 99.00 utterances.], tot_loss[ctc_loss=0.1231, att_loss=0.2605, loss=0.233, over 3275079.43 frames. utt_duration=1227 frames, utt_pad_proportion=0.0583, over 10686.47 utterances.], batch size: 99, lr: 1.34e-02, grad_scale: 8.0 2023-03-07 23:52:27,676 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.60 vs. limit=5.0 2023-03-07 23:52:31,279 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0065, 6.1665, 5.4124, 5.9998, 5.7594, 5.4264, 5.6485, 5.4459], device='cuda:1'), covar=tensor([0.0990, 0.0802, 0.0829, 0.0647, 0.0796, 0.1265, 0.1889, 0.2046], device='cuda:1'), in_proj_covar=tensor([0.0396, 0.0459, 0.0353, 0.0361, 0.0334, 0.0396, 0.0476, 0.0425], 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') 2023-03-07 23:52:40,550 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8840, 5.2649, 5.1492, 5.1068, 5.2827, 5.2623, 4.9797, 4.7483], device='cuda:1'), covar=tensor([0.1064, 0.0357, 0.0224, 0.0452, 0.0248, 0.0250, 0.0325, 0.0347], device='cuda:1'), in_proj_covar=tensor([0.0434, 0.0265, 0.0213, 0.0250, 0.0308, 0.0335, 0.0249, 0.0288], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-07 23:52:42,256 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30655.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:53:33,180 INFO [train2.py:809] (1/4) Epoch 8, batch 2800, loss[ctc_loss=0.08595, att_loss=0.2313, loss=0.2022, over 15371.00 frames. utt_duration=1758 frames, utt_pad_proportion=0.01056, over 35.00 utterances.], tot_loss[ctc_loss=0.1216, att_loss=0.26, loss=0.2323, over 3275775.90 frames. utt_duration=1260 frames, utt_pad_proportion=0.05017, over 10408.78 utterances.], batch size: 35, lr: 1.34e-02, grad_scale: 8.0 2023-03-07 23:54:21,209 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.747e+02 2.729e+02 3.286e+02 4.018e+02 8.527e+02, threshold=6.573e+02, percent-clipped=3.0 2023-03-07 23:54:21,438 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30716.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:54:21,631 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30716.0, num_to_drop=1, layers_to_drop={3} 2023-03-07 23:54:53,844 INFO [train2.py:809] (1/4) Epoch 8, batch 2850, loss[ctc_loss=0.1301, att_loss=0.2633, loss=0.2366, over 17113.00 frames. utt_duration=1224 frames, utt_pad_proportion=0.01533, over 56.00 utterances.], tot_loss[ctc_loss=0.121, att_loss=0.2594, loss=0.2317, over 3270433.90 frames. utt_duration=1254 frames, utt_pad_proportion=0.05286, over 10442.94 utterances.], batch size: 56, lr: 1.34e-02, grad_scale: 8.0 2023-03-07 23:55:24,223 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.1544, 3.9921, 3.4375, 3.7822, 4.1789, 3.8642, 3.5379, 4.6263], device='cuda:1'), covar=tensor([0.0941, 0.0348, 0.0885, 0.0571, 0.0441, 0.0584, 0.0694, 0.0340], device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0167, 0.0192, 0.0165, 0.0208, 0.0195, 0.0170, 0.0221], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-07 23:56:04,090 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30781.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:56:13,170 INFO [train2.py:809] (1/4) Epoch 8, batch 2900, loss[ctc_loss=0.113, att_loss=0.2575, loss=0.2286, over 17501.00 frames. utt_duration=887.7 frames, utt_pad_proportion=0.07048, over 79.00 utterances.], tot_loss[ctc_loss=0.122, att_loss=0.26, loss=0.2324, over 3266269.80 frames. utt_duration=1219 frames, utt_pad_proportion=0.06341, over 10729.19 utterances.], batch size: 79, lr: 1.34e-02, grad_scale: 8.0 2023-03-07 23:56:24,167 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5233, 2.3374, 4.7884, 3.7315, 2.9842, 4.2586, 4.4363, 4.4566], device='cuda:1'), covar=tensor([0.0144, 0.1834, 0.0084, 0.1064, 0.1926, 0.0234, 0.0118, 0.0195], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0242, 0.0124, 0.0305, 0.0284, 0.0182, 0.0110, 0.0143], device='cuda:1'), out_proj_covar=tensor([1.3758e-04, 2.0268e-04, 1.1066e-04, 2.5164e-04, 2.5115e-04, 1.6212e-04, 9.9556e-05, 1.3446e-04], device='cuda:1') 2023-03-07 23:56:42,964 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-03-07 23:56:56,975 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.4335, 3.7732, 3.0608, 3.3242, 4.0257, 3.6222, 2.4776, 4.1946], device='cuda:1'), covar=tensor([0.1316, 0.0415, 0.1012, 0.0644, 0.0519, 0.0567, 0.1152, 0.0503], device='cuda:1'), in_proj_covar=tensor([0.0175, 0.0168, 0.0194, 0.0165, 0.0209, 0.0195, 0.0171, 0.0222], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-07 23:56:59,560 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.721e+02 2.826e+02 3.398e+02 4.214e+02 7.798e+02, threshold=6.796e+02, percent-clipped=3.0 2023-03-07 23:57:09,855 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1627, 4.6957, 4.6781, 4.7381, 2.3507, 4.7684, 2.6856, 2.0906], device='cuda:1'), covar=tensor([0.0299, 0.0153, 0.0691, 0.0174, 0.2346, 0.0141, 0.1674, 0.1690], device='cuda:1'), in_proj_covar=tensor([0.0124, 0.0100, 0.0253, 0.0109, 0.0223, 0.0099, 0.0226, 0.0199], device='cuda:1'), out_proj_covar=tensor([1.2452e-04, 1.0439e-04, 2.3153e-04, 1.0478e-04, 2.1096e-04, 9.8613e-05, 2.0677e-04, 1.8276e-04], device='cuda:1') 2023-03-07 23:57:20,122 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=30829.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:57:32,269 INFO [train2.py:809] (1/4) Epoch 8, batch 2950, loss[ctc_loss=0.1091, att_loss=0.2586, loss=0.2287, over 17030.00 frames. utt_duration=1338 frames, utt_pad_proportion=0.006973, over 51.00 utterances.], tot_loss[ctc_loss=0.1222, att_loss=0.2598, loss=0.2323, over 3269029.49 frames. utt_duration=1223 frames, utt_pad_proportion=0.06258, over 10701.62 utterances.], batch size: 51, lr: 1.34e-02, grad_scale: 8.0 2023-03-07 23:57:57,935 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-03-07 23:58:17,835 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30865.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:58:38,855 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5387, 2.2164, 4.9257, 3.7930, 2.9402, 4.3971, 4.6246, 4.6420], device='cuda:1'), covar=tensor([0.0183, 0.1987, 0.0113, 0.1208, 0.2134, 0.0255, 0.0116, 0.0179], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0240, 0.0124, 0.0300, 0.0281, 0.0181, 0.0109, 0.0142], device='cuda:1'), out_proj_covar=tensor([1.3596e-04, 2.0095e-04, 1.0985e-04, 2.4763e-04, 2.4829e-04, 1.6057e-04, 9.8624e-05, 1.3327e-04], device='cuda:1') 2023-03-07 23:58:52,464 INFO [train2.py:809] (1/4) Epoch 8, batch 3000, loss[ctc_loss=0.1215, att_loss=0.2856, loss=0.2528, over 16965.00 frames. utt_duration=1359 frames, utt_pad_proportion=0.007641, over 50.00 utterances.], tot_loss[ctc_loss=0.1214, att_loss=0.2596, loss=0.232, over 3273573.61 frames. utt_duration=1254 frames, utt_pad_proportion=0.05393, over 10450.89 utterances.], batch size: 50, lr: 1.34e-02, grad_scale: 8.0 2023-03-07 23:58:52,464 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-07 23:59:06,434 INFO [train2.py:843] (1/4) Epoch 8, validation: ctc_loss=0.05812, att_loss=0.2422, loss=0.2054, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-07 23:59:06,435 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-07 23:59:32,920 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30903.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:59:48,144 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=30913.0, num_to_drop=0, layers_to_drop=set() 2023-03-07 23:59:53,216 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.005e+02 2.876e+02 3.540e+02 4.400e+02 7.249e+02, threshold=7.079e+02, percent-clipped=3.0 2023-03-08 00:00:07,406 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30925.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:00:21,859 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30934.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:00:26,210 INFO [train2.py:809] (1/4) Epoch 8, batch 3050, loss[ctc_loss=0.1374, att_loss=0.2821, loss=0.2532, over 17477.00 frames. utt_duration=1015 frames, utt_pad_proportion=0.04286, over 69.00 utterances.], tot_loss[ctc_loss=0.1211, att_loss=0.2595, loss=0.2319, over 3264978.83 frames. utt_duration=1244 frames, utt_pad_proportion=0.05789, over 10513.17 utterances.], batch size: 69, lr: 1.34e-02, grad_scale: 8.0 2023-03-08 00:00:30,360 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.31 vs. limit=5.0 2023-03-08 00:00:50,339 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.6621, 5.9415, 5.4068, 5.7656, 5.5362, 5.2162, 5.3083, 5.2327], device='cuda:1'), covar=tensor([0.1239, 0.0881, 0.0776, 0.0716, 0.0851, 0.1474, 0.2178, 0.1993], device='cuda:1'), in_proj_covar=tensor([0.0398, 0.0446, 0.0347, 0.0359, 0.0331, 0.0397, 0.0468, 0.0420], 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') 2023-03-08 00:01:01,326 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-03-08 00:01:10,385 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30964.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:01:26,235 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.59 vs. limit=5.0 2023-03-08 00:01:44,853 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30986.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:01:46,066 INFO [train2.py:809] (1/4) Epoch 8, batch 3100, loss[ctc_loss=0.1082, att_loss=0.2686, loss=0.2366, over 16958.00 frames. utt_duration=1358 frames, utt_pad_proportion=0.007193, over 50.00 utterances.], tot_loss[ctc_loss=0.1214, att_loss=0.2605, loss=0.2327, over 3281117.45 frames. utt_duration=1256 frames, utt_pad_proportion=0.0498, over 10461.16 utterances.], batch size: 50, lr: 1.33e-02, grad_scale: 8.0 2023-03-08 00:01:59,003 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30995.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:02:24,742 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31011.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 00:02:24,889 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31011.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 00:02:32,584 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.786e+02 2.541e+02 3.093e+02 3.936e+02 7.180e+02, threshold=6.186e+02, percent-clipped=1.0 2023-03-08 00:02:32,935 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31016.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:02:54,580 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=2.01 vs. limit=2.0 2023-03-08 00:03:05,876 INFO [train2.py:809] (1/4) Epoch 8, batch 3150, loss[ctc_loss=0.1279, att_loss=0.2612, loss=0.2346, over 16555.00 frames. utt_duration=1473 frames, utt_pad_proportion=0.005282, over 45.00 utterances.], tot_loss[ctc_loss=0.1206, att_loss=0.2595, loss=0.2317, over 3277443.56 frames. utt_duration=1259 frames, utt_pad_proportion=0.04923, over 10428.57 utterances.], batch size: 45, lr: 1.33e-02, grad_scale: 8.0 2023-03-08 00:03:49,191 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=31064.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:04:01,732 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31072.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 00:04:13,023 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9923, 5.0964, 5.5143, 5.4096, 5.2704, 5.8993, 5.0564, 5.9765], device='cuda:1'), covar=tensor([0.0559, 0.0625, 0.0670, 0.0906, 0.1908, 0.0796, 0.0594, 0.0552], device='cuda:1'), in_proj_covar=tensor([0.0616, 0.0377, 0.0432, 0.0490, 0.0662, 0.0432, 0.0350, 0.0429], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-03-08 00:04:25,395 INFO [train2.py:809] (1/4) Epoch 8, batch 3200, loss[ctc_loss=0.1253, att_loss=0.2756, loss=0.2456, over 17070.00 frames. utt_duration=1290 frames, utt_pad_proportion=0.008513, over 53.00 utterances.], tot_loss[ctc_loss=0.1218, att_loss=0.2607, loss=0.2329, over 3287558.93 frames. utt_duration=1248 frames, utt_pad_proportion=0.04906, over 10546.37 utterances.], batch size: 53, lr: 1.33e-02, grad_scale: 8.0 2023-03-08 00:05:05,558 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.1266, 5.3565, 5.6837, 5.5782, 5.5264, 6.0747, 5.2096, 6.1600], device='cuda:1'), covar=tensor([0.0639, 0.0645, 0.0650, 0.0827, 0.1816, 0.0752, 0.0562, 0.0530], device='cuda:1'), in_proj_covar=tensor([0.0619, 0.0378, 0.0431, 0.0490, 0.0663, 0.0431, 0.0351, 0.0430], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-03-08 00:05:11,435 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.012e+02 2.783e+02 3.320e+02 3.859e+02 7.988e+02, threshold=6.640e+02, percent-clipped=5.0 2023-03-08 00:05:44,553 INFO [train2.py:809] (1/4) Epoch 8, batch 3250, loss[ctc_loss=0.1346, att_loss=0.2776, loss=0.249, over 17288.00 frames. utt_duration=1173 frames, utt_pad_proportion=0.02541, over 59.00 utterances.], tot_loss[ctc_loss=0.1224, att_loss=0.2608, loss=0.2332, over 3289741.09 frames. utt_duration=1224 frames, utt_pad_proportion=0.05241, over 10766.74 utterances.], batch size: 59, lr: 1.33e-02, grad_scale: 8.0 2023-03-08 00:06:20,637 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-03-08 00:06:58,061 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2270, 5.1966, 5.0634, 2.6295, 2.0754, 2.6474, 3.8199, 3.9018], device='cuda:1'), covar=tensor([0.0579, 0.0215, 0.0277, 0.3371, 0.6229, 0.3020, 0.1055, 0.1945], device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0206, 0.0229, 0.0191, 0.0361, 0.0345, 0.0221, 0.0353], device='cuda:1'), out_proj_covar=tensor([1.5551e-04, 8.1364e-05, 1.0161e-04, 8.8288e-05, 1.6344e-04, 1.4635e-04, 8.9039e-05, 1.5656e-04], device='cuda:1') 2023-03-08 00:07:03,808 INFO [train2.py:809] (1/4) Epoch 8, batch 3300, loss[ctc_loss=0.09454, att_loss=0.2297, loss=0.2027, over 15861.00 frames. utt_duration=1628 frames, utt_pad_proportion=0.01067, over 39.00 utterances.], tot_loss[ctc_loss=0.1229, att_loss=0.2608, loss=0.2332, over 3286883.01 frames. utt_duration=1220 frames, utt_pad_proportion=0.05479, over 10785.81 utterances.], batch size: 39, lr: 1.33e-02, grad_scale: 8.0 2023-03-08 00:07:51,302 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.702e+02 2.610e+02 3.147e+02 4.010e+02 7.285e+02, threshold=6.294e+02, percent-clipped=3.0 2023-03-08 00:08:24,607 INFO [train2.py:809] (1/4) Epoch 8, batch 3350, loss[ctc_loss=0.109, att_loss=0.2464, loss=0.2189, over 16176.00 frames. utt_duration=1580 frames, utt_pad_proportion=0.005925, over 41.00 utterances.], tot_loss[ctc_loss=0.1216, att_loss=0.2605, loss=0.2327, over 3289269.56 frames. utt_duration=1230 frames, utt_pad_proportion=0.05211, over 10707.41 utterances.], batch size: 41, lr: 1.33e-02, grad_scale: 8.0 2023-03-08 00:08:41,521 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.6712, 5.9438, 5.3616, 5.8115, 5.5921, 5.2253, 5.4114, 5.2670], device='cuda:1'), covar=tensor([0.1270, 0.1010, 0.0917, 0.0780, 0.0717, 0.1503, 0.2537, 0.2524], device='cuda:1'), in_proj_covar=tensor([0.0399, 0.0454, 0.0348, 0.0359, 0.0330, 0.0399, 0.0474, 0.0427], 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') 2023-03-08 00:09:00,716 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31259.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:09:35,265 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31281.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:09:44,165 INFO [train2.py:809] (1/4) Epoch 8, batch 3400, loss[ctc_loss=0.1115, att_loss=0.261, loss=0.2311, over 16623.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.005349, over 47.00 utterances.], tot_loss[ctc_loss=0.1237, att_loss=0.262, loss=0.2343, over 3282071.14 frames. utt_duration=1197 frames, utt_pad_proportion=0.06428, over 10982.71 utterances.], batch size: 47, lr: 1.33e-02, grad_scale: 8.0 2023-03-08 00:09:49,561 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31290.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:10:22,997 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31311.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:10:30,543 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.874e+02 2.871e+02 3.632e+02 4.520e+02 1.169e+03, threshold=7.263e+02, percent-clipped=5.0 2023-03-08 00:11:04,306 INFO [train2.py:809] (1/4) Epoch 8, batch 3450, loss[ctc_loss=0.1436, att_loss=0.2767, loss=0.2501, over 16957.00 frames. utt_duration=1358 frames, utt_pad_proportion=0.007904, over 50.00 utterances.], tot_loss[ctc_loss=0.1231, att_loss=0.2614, loss=0.2338, over 3284897.26 frames. utt_duration=1222 frames, utt_pad_proportion=0.05718, over 10762.16 utterances.], batch size: 50, lr: 1.33e-02, grad_scale: 8.0 2023-03-08 00:11:40,442 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=31359.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:11:40,549 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.8499, 5.0295, 5.3340, 5.3250, 5.1922, 5.8659, 5.0626, 5.9558], device='cuda:1'), covar=tensor([0.0652, 0.0719, 0.0683, 0.0902, 0.1832, 0.0764, 0.0581, 0.0531], device='cuda:1'), in_proj_covar=tensor([0.0634, 0.0384, 0.0435, 0.0498, 0.0670, 0.0432, 0.0355, 0.0431], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 00:11:52,614 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31367.0, num_to_drop=1, layers_to_drop={3} 2023-03-08 00:12:25,032 INFO [train2.py:809] (1/4) Epoch 8, batch 3500, loss[ctc_loss=0.104, att_loss=0.2622, loss=0.2306, over 17017.00 frames. utt_duration=1336 frames, utt_pad_proportion=0.00805, over 51.00 utterances.], tot_loss[ctc_loss=0.1228, att_loss=0.261, loss=0.2333, over 3281830.33 frames. utt_duration=1217 frames, utt_pad_proportion=0.05909, over 10804.18 utterances.], batch size: 51, lr: 1.33e-02, grad_scale: 8.0 2023-03-08 00:12:37,379 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.8909, 6.1765, 5.5929, 5.9818, 5.8513, 5.4291, 5.5579, 5.4316], device='cuda:1'), covar=tensor([0.1370, 0.0873, 0.0756, 0.0737, 0.0693, 0.1293, 0.2111, 0.2229], device='cuda:1'), in_proj_covar=tensor([0.0400, 0.0455, 0.0346, 0.0362, 0.0329, 0.0398, 0.0472, 0.0428], 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') 2023-03-08 00:12:40,836 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.8101, 1.6319, 1.9764, 1.8096, 2.4988, 2.2199, 1.5534, 2.7460], device='cuda:1'), covar=tensor([0.0616, 0.4224, 0.2522, 0.0994, 0.0972, 0.1390, 0.2706, 0.0810], device='cuda:1'), in_proj_covar=tensor([0.0074, 0.0086, 0.0086, 0.0075, 0.0072, 0.0069, 0.0080, 0.0064], device='cuda:1'), out_proj_covar=tensor([4.1335e-05, 5.3498e-05, 5.2784e-05, 4.5267e-05, 4.0273e-05, 4.4725e-05, 5.0188e-05, 4.2837e-05], device='cuda:1') 2023-03-08 00:13:02,237 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31409.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:13:03,552 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.3380, 5.6498, 5.0699, 5.4777, 5.3082, 4.9808, 5.0665, 4.8665], device='cuda:1'), covar=tensor([0.1226, 0.0953, 0.0719, 0.0805, 0.0605, 0.1151, 0.2074, 0.2462], device='cuda:1'), in_proj_covar=tensor([0.0398, 0.0455, 0.0346, 0.0361, 0.0328, 0.0397, 0.0472, 0.0427], 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') 2023-03-08 00:13:12,728 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.860e+02 2.623e+02 3.276e+02 4.275e+02 9.794e+02, threshold=6.553e+02, percent-clipped=4.0 2023-03-08 00:13:47,409 INFO [train2.py:809] (1/4) Epoch 8, batch 3550, loss[ctc_loss=0.09526, att_loss=0.2334, loss=0.2058, over 16172.00 frames. utt_duration=1579 frames, utt_pad_proportion=0.007282, over 41.00 utterances.], tot_loss[ctc_loss=0.1224, att_loss=0.2602, loss=0.2327, over 3277564.46 frames. utt_duration=1199 frames, utt_pad_proportion=0.06434, over 10950.24 utterances.], batch size: 41, lr: 1.33e-02, grad_scale: 8.0 2023-03-08 00:14:07,772 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.4217, 2.1191, 2.8399, 4.0227, 3.7724, 3.8202, 2.5540, 1.9133], device='cuda:1'), covar=tensor([0.0562, 0.2698, 0.1155, 0.0679, 0.0727, 0.0356, 0.1674, 0.2464], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0201, 0.0184, 0.0178, 0.0170, 0.0136, 0.0185, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 00:14:40,197 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31470.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:15:08,417 INFO [train2.py:809] (1/4) Epoch 8, batch 3600, loss[ctc_loss=0.1218, att_loss=0.2742, loss=0.2437, over 16776.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.006111, over 48.00 utterances.], tot_loss[ctc_loss=0.1223, att_loss=0.2602, loss=0.2326, over 3275009.27 frames. utt_duration=1213 frames, utt_pad_proportion=0.06098, over 10812.22 utterances.], batch size: 48, lr: 1.32e-02, grad_scale: 8.0 2023-03-08 00:15:14,936 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8351, 5.1525, 5.1474, 4.9790, 5.2471, 5.2436, 4.9240, 4.6422], device='cuda:1'), covar=tensor([0.1220, 0.0450, 0.0252, 0.0591, 0.0278, 0.0282, 0.0276, 0.0353], device='cuda:1'), in_proj_covar=tensor([0.0428, 0.0267, 0.0213, 0.0253, 0.0308, 0.0329, 0.0249, 0.0288], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 00:15:15,053 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31491.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 00:15:54,465 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.411e+02 2.609e+02 2.998e+02 4.010e+02 6.869e+02, threshold=5.995e+02, percent-clipped=2.0 2023-03-08 00:16:29,577 INFO [train2.py:809] (1/4) Epoch 8, batch 3650, loss[ctc_loss=0.1185, att_loss=0.2664, loss=0.2368, over 17284.00 frames. utt_duration=1004 frames, utt_pad_proportion=0.05131, over 69.00 utterances.], tot_loss[ctc_loss=0.1211, att_loss=0.2596, loss=0.2319, over 3281900.02 frames. utt_duration=1240 frames, utt_pad_proportion=0.05293, over 10598.34 utterances.], batch size: 69, lr: 1.32e-02, grad_scale: 8.0 2023-03-08 00:16:45,753 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1398, 5.1873, 5.1410, 2.4164, 2.0538, 2.6314, 3.8457, 3.8302], device='cuda:1'), covar=tensor([0.0615, 0.0190, 0.0215, 0.3844, 0.6569, 0.3182, 0.1118, 0.1932], device='cuda:1'), in_proj_covar=tensor([0.0321, 0.0208, 0.0229, 0.0187, 0.0355, 0.0341, 0.0218, 0.0349], device='cuda:1'), out_proj_covar=tensor([1.5229e-04, 8.0968e-05, 1.0043e-04, 8.6030e-05, 1.6063e-04, 1.4402e-04, 8.7068e-05, 1.5350e-04], device='cuda:1') 2023-03-08 00:16:54,280 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31552.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 00:16:56,487 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-03-08 00:16:58,979 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7038, 2.4664, 5.0745, 4.0593, 2.9754, 4.6446, 5.0122, 4.7430], device='cuda:1'), covar=tensor([0.0191, 0.1784, 0.0175, 0.1077, 0.2097, 0.0192, 0.0090, 0.0195], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0244, 0.0126, 0.0303, 0.0282, 0.0184, 0.0111, 0.0145], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-03-08 00:17:04,961 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31559.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:17:40,200 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31581.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:17:50,150 INFO [train2.py:809] (1/4) Epoch 8, batch 3700, loss[ctc_loss=0.08941, att_loss=0.2282, loss=0.2004, over 15509.00 frames. utt_duration=1725 frames, utt_pad_proportion=0.00821, over 36.00 utterances.], tot_loss[ctc_loss=0.1211, att_loss=0.2604, loss=0.2326, over 3288712.51 frames. utt_duration=1236 frames, utt_pad_proportion=0.0531, over 10653.88 utterances.], batch size: 36, lr: 1.32e-02, grad_scale: 8.0 2023-03-08 00:17:55,131 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31590.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:18:22,258 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=31607.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:18:35,885 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.785e+02 2.776e+02 3.358e+02 4.277e+02 8.903e+02, threshold=6.717e+02, percent-clipped=4.0 2023-03-08 00:18:57,078 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=31629.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:19:10,109 INFO [train2.py:809] (1/4) Epoch 8, batch 3750, loss[ctc_loss=0.1118, att_loss=0.2681, loss=0.2369, over 17072.00 frames. utt_duration=1290 frames, utt_pad_proportion=0.008644, over 53.00 utterances.], tot_loss[ctc_loss=0.1212, att_loss=0.2601, loss=0.2323, over 3281188.99 frames. utt_duration=1224 frames, utt_pad_proportion=0.05737, over 10735.27 utterances.], batch size: 53, lr: 1.32e-02, grad_scale: 8.0 2023-03-08 00:19:11,678 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=31638.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:19:51,326 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0610, 5.2022, 4.9594, 2.6986, 2.0205, 2.7885, 3.9091, 3.6174], device='cuda:1'), covar=tensor([0.0621, 0.0179, 0.0234, 0.2888, 0.6155, 0.2628, 0.1142, 0.2142], device='cuda:1'), in_proj_covar=tensor([0.0319, 0.0205, 0.0226, 0.0183, 0.0352, 0.0337, 0.0216, 0.0346], device='cuda:1'), out_proj_covar=tensor([1.5127e-04, 7.9786e-05, 9.9624e-05, 8.4226e-05, 1.5896e-04, 1.4265e-04, 8.6366e-05, 1.5224e-04], device='cuda:1') 2023-03-08 00:19:57,351 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31667.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 00:20:29,569 INFO [train2.py:809] (1/4) Epoch 8, batch 3800, loss[ctc_loss=0.1178, att_loss=0.2626, loss=0.2337, over 17046.00 frames. utt_duration=1313 frames, utt_pad_proportion=0.008482, over 52.00 utterances.], tot_loss[ctc_loss=0.1205, att_loss=0.2596, loss=0.2318, over 3271184.54 frames. utt_duration=1246 frames, utt_pad_proportion=0.05184, over 10512.11 utterances.], batch size: 52, lr: 1.32e-02, grad_scale: 8.0 2023-03-08 00:20:42,513 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1976, 4.6722, 4.7264, 4.6172, 2.6288, 5.1041, 3.0298, 1.8860], device='cuda:1'), covar=tensor([0.0332, 0.0183, 0.0701, 0.0166, 0.2010, 0.0070, 0.1449, 0.1948], device='cuda:1'), in_proj_covar=tensor([0.0123, 0.0101, 0.0250, 0.0107, 0.0218, 0.0098, 0.0223, 0.0196], device='cuda:1'), out_proj_covar=tensor([1.2411e-04, 1.0425e-04, 2.2878e-04, 1.0367e-04, 2.0703e-04, 9.7446e-05, 2.0457e-04, 1.8116e-04], device='cuda:1') 2023-03-08 00:20:56,792 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31704.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:21:13,675 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=31715.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 00:21:14,931 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.603e+02 2.740e+02 3.298e+02 4.235e+02 1.019e+03, threshold=6.595e+02, percent-clipped=5.0 2023-03-08 00:21:23,091 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2407, 4.7557, 4.5726, 4.8987, 4.8495, 4.5019, 3.6000, 4.6959], device='cuda:1'), covar=tensor([0.0117, 0.0114, 0.0105, 0.0057, 0.0079, 0.0116, 0.0534, 0.0193], device='cuda:1'), in_proj_covar=tensor([0.0067, 0.0063, 0.0073, 0.0046, 0.0050, 0.0060, 0.0083, 0.0084], 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') 2023-03-08 00:21:49,310 INFO [train2.py:809] (1/4) Epoch 8, batch 3850, loss[ctc_loss=0.1526, att_loss=0.2861, loss=0.2594, over 17344.00 frames. utt_duration=1177 frames, utt_pad_proportion=0.02206, over 59.00 utterances.], tot_loss[ctc_loss=0.1202, att_loss=0.2596, loss=0.2317, over 3278429.23 frames. utt_duration=1238 frames, utt_pad_proportion=0.05331, over 10609.67 utterances.], batch size: 59, lr: 1.32e-02, grad_scale: 8.0 2023-03-08 00:21:49,509 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9761, 5.2902, 5.5092, 5.4661, 5.3320, 5.9267, 5.1457, 6.0306], device='cuda:1'), covar=tensor([0.0578, 0.0605, 0.0624, 0.0879, 0.1669, 0.0721, 0.0532, 0.0533], device='cuda:1'), in_proj_covar=tensor([0.0631, 0.0385, 0.0438, 0.0501, 0.0674, 0.0437, 0.0354, 0.0429], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0002, 0.0003], device='cuda:1') 2023-03-08 00:22:32,470 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31765.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:22:32,665 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31765.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:22:58,794 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.6416, 4.9065, 5.1370, 5.0620, 5.0276, 5.5559, 4.8730, 5.7139], device='cuda:1'), covar=tensor([0.0597, 0.0639, 0.0632, 0.0897, 0.1795, 0.0782, 0.0757, 0.0506], device='cuda:1'), in_proj_covar=tensor([0.0634, 0.0387, 0.0438, 0.0501, 0.0671, 0.0437, 0.0356, 0.0431], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 00:23:05,943 INFO [train2.py:809] (1/4) Epoch 8, batch 3900, loss[ctc_loss=0.08268, att_loss=0.2276, loss=0.1986, over 15627.00 frames. utt_duration=1691 frames, utt_pad_proportion=0.009373, over 37.00 utterances.], tot_loss[ctc_loss=0.1205, att_loss=0.2595, loss=0.2317, over 3281468.74 frames. utt_duration=1242 frames, utt_pad_proportion=0.05169, over 10578.35 utterances.], batch size: 37, lr: 1.32e-02, grad_scale: 8.0 2023-03-08 00:23:07,806 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31788.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:23:24,503 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=5.00 vs. limit=5.0 2023-03-08 00:23:50,996 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.828e+02 2.672e+02 3.345e+02 4.002e+02 8.198e+02, threshold=6.690e+02, percent-clipped=4.0 2023-03-08 00:24:01,914 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31823.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:24:22,885 INFO [train2.py:809] (1/4) Epoch 8, batch 3950, loss[ctc_loss=0.1293, att_loss=0.271, loss=0.2426, over 17029.00 frames. utt_duration=1311 frames, utt_pad_proportion=0.009456, over 52.00 utterances.], tot_loss[ctc_loss=0.1208, att_loss=0.2602, loss=0.2323, over 3282616.75 frames. utt_duration=1222 frames, utt_pad_proportion=0.05708, over 10759.11 utterances.], batch size: 52, lr: 1.32e-02, grad_scale: 8.0 2023-03-08 00:24:38,342 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31847.0, num_to_drop=1, layers_to_drop={3} 2023-03-08 00:24:41,462 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31849.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:25:10,293 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.03 vs. limit=5.0 2023-03-08 00:25:39,113 INFO [train2.py:809] (1/4) Epoch 9, batch 0, loss[ctc_loss=0.1322, att_loss=0.2704, loss=0.2428, over 16962.00 frames. utt_duration=1358 frames, utt_pad_proportion=0.007699, over 50.00 utterances.], tot_loss[ctc_loss=0.1322, att_loss=0.2704, loss=0.2428, over 16962.00 frames. utt_duration=1358 frames, utt_pad_proportion=0.007699, over 50.00 utterances.], batch size: 50, lr: 1.25e-02, grad_scale: 8.0 2023-03-08 00:25:39,113 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-08 00:25:51,862 INFO [train2.py:843] (1/4) Epoch 9, validation: ctc_loss=0.05701, att_loss=0.2418, loss=0.2048, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-08 00:25:51,867 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-08 00:26:05,406 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31878.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:26:10,054 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3611, 2.4813, 3.6382, 2.6244, 3.3906, 4.5285, 4.3433, 3.2817], device='cuda:1'), covar=tensor([0.0489, 0.2021, 0.0958, 0.1652, 0.1097, 0.0623, 0.0445, 0.1346], device='cuda:1'), in_proj_covar=tensor([0.0224, 0.0219, 0.0227, 0.0197, 0.0230, 0.0262, 0.0199, 0.0214], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-03-08 00:26:14,990 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31884.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:27:04,436 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31915.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:27:05,620 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.515e+02 2.740e+02 3.193e+02 4.001e+02 8.553e+02, threshold=6.386e+02, percent-clipped=3.0 2023-03-08 00:27:12,466 INFO [train2.py:809] (1/4) Epoch 9, batch 50, loss[ctc_loss=0.1221, att_loss=0.2702, loss=0.2405, over 16965.00 frames. utt_duration=1359 frames, utt_pad_proportion=0.007611, over 50.00 utterances.], tot_loss[ctc_loss=0.1168, att_loss=0.2579, loss=0.2297, over 745404.43 frames. utt_duration=1299 frames, utt_pad_proportion=0.03019, over 2297.94 utterances.], batch size: 50, lr: 1.25e-02, grad_scale: 16.0 2023-03-08 00:27:43,180 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31939.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:28:00,030 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31950.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:28:08,427 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5983, 2.6190, 4.9692, 3.9487, 2.8380, 4.5415, 4.8484, 4.7271], device='cuda:1'), covar=tensor([0.0228, 0.1652, 0.0159, 0.1039, 0.2267, 0.0211, 0.0105, 0.0215], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0244, 0.0129, 0.0307, 0.0286, 0.0185, 0.0110, 0.0147], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0003, 0.0003, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-03-08 00:28:12,840 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5488, 2.3907, 4.9171, 3.8237, 2.8650, 4.4316, 4.7253, 4.6531], device='cuda:1'), covar=tensor([0.0215, 0.1789, 0.0124, 0.1116, 0.2168, 0.0243, 0.0103, 0.0224], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0243, 0.0129, 0.0306, 0.0285, 0.0185, 0.0110, 0.0146], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0003, 0.0003, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-03-08 00:28:31,528 INFO [train2.py:809] (1/4) Epoch 9, batch 100, loss[ctc_loss=0.09012, att_loss=0.2274, loss=0.1999, over 15509.00 frames. utt_duration=1725 frames, utt_pad_proportion=0.00829, over 36.00 utterances.], tot_loss[ctc_loss=0.1167, att_loss=0.2569, loss=0.2288, over 1310812.26 frames. utt_duration=1322 frames, utt_pad_proportion=0.02776, over 3971.26 utterances.], batch size: 36, lr: 1.24e-02, grad_scale: 16.0 2023-03-08 00:28:41,654 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31976.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:29:11,601 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4280, 4.8586, 4.7043, 4.6523, 2.7079, 4.3077, 2.9313, 2.2241], device='cuda:1'), covar=tensor([0.0207, 0.0109, 0.0610, 0.0261, 0.2209, 0.0213, 0.1695, 0.1827], device='cuda:1'), in_proj_covar=tensor([0.0123, 0.0103, 0.0250, 0.0108, 0.0222, 0.0103, 0.0227, 0.0199], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 00:29:41,553 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=32011.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:29:48,728 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.872e+02 3.075e+02 3.496e+02 4.231e+02 7.573e+02, threshold=6.993e+02, percent-clipped=4.0 2023-03-08 00:29:55,000 INFO [train2.py:809] (1/4) Epoch 9, batch 150, loss[ctc_loss=0.09742, att_loss=0.2529, loss=0.2218, over 16774.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.006141, over 48.00 utterances.], tot_loss[ctc_loss=0.1179, att_loss=0.2575, loss=0.2296, over 1748784.16 frames. utt_duration=1262 frames, utt_pad_proportion=0.0435, over 5548.14 utterances.], batch size: 48, lr: 1.24e-02, grad_scale: 16.0 2023-03-08 00:30:43,804 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=5.03 vs. limit=5.0 2023-03-08 00:30:59,454 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32060.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:31:07,683 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32065.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:31:15,196 INFO [train2.py:809] (1/4) Epoch 9, batch 200, loss[ctc_loss=0.09134, att_loss=0.234, loss=0.2055, over 15774.00 frames. utt_duration=1662 frames, utt_pad_proportion=0.008385, over 38.00 utterances.], tot_loss[ctc_loss=0.1174, att_loss=0.2567, loss=0.2288, over 2088190.58 frames. utt_duration=1279 frames, utt_pad_proportion=0.04215, over 6539.98 utterances.], batch size: 38, lr: 1.24e-02, grad_scale: 16.0 2023-03-08 00:32:23,945 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=32113.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:32:28,423 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.810e+02 2.737e+02 3.348e+02 3.979e+02 1.144e+03, threshold=6.695e+02, percent-clipped=2.0 2023-03-08 00:32:34,682 INFO [train2.py:809] (1/4) Epoch 9, batch 250, loss[ctc_loss=0.093, att_loss=0.2415, loss=0.2118, over 16126.00 frames. utt_duration=1537 frames, utt_pad_proportion=0.006326, over 42.00 utterances.], tot_loss[ctc_loss=0.1183, att_loss=0.2575, loss=0.2296, over 2348903.72 frames. utt_duration=1239 frames, utt_pad_proportion=0.05179, over 7591.33 utterances.], batch size: 42, lr: 1.24e-02, grad_scale: 16.0 2023-03-08 00:33:13,082 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32144.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:33:17,796 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32147.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 00:33:54,277 INFO [train2.py:809] (1/4) Epoch 9, batch 300, loss[ctc_loss=0.09529, att_loss=0.2298, loss=0.2029, over 15868.00 frames. utt_duration=1629 frames, utt_pad_proportion=0.01042, over 39.00 utterances.], tot_loss[ctc_loss=0.1174, att_loss=0.2565, loss=0.2287, over 2552112.17 frames. utt_duration=1258 frames, utt_pad_proportion=0.0492, over 8121.91 utterances.], batch size: 39, lr: 1.24e-02, grad_scale: 16.0 2023-03-08 00:34:09,517 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32179.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:34:21,162 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.80 vs. limit=5.0 2023-03-08 00:34:34,215 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=32195.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 00:35:08,279 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.478e+02 2.524e+02 3.099e+02 4.059e+02 8.225e+02, threshold=6.197e+02, percent-clipped=2.0 2023-03-08 00:35:14,579 INFO [train2.py:809] (1/4) Epoch 9, batch 350, loss[ctc_loss=0.09995, att_loss=0.2393, loss=0.2114, over 16179.00 frames. utt_duration=1580 frames, utt_pad_proportion=0.007067, over 41.00 utterances.], tot_loss[ctc_loss=0.1167, att_loss=0.2561, loss=0.2282, over 2709360.60 frames. utt_duration=1261 frames, utt_pad_proportion=0.05028, over 8602.40 utterances.], batch size: 41, lr: 1.24e-02, grad_scale: 16.0 2023-03-08 00:35:36,444 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.5079, 2.3876, 3.2637, 4.2894, 3.8968, 3.8386, 2.7548, 1.8609], device='cuda:1'), covar=tensor([0.0695, 0.2413, 0.1133, 0.0540, 0.0762, 0.0402, 0.1620, 0.2815], device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0198, 0.0185, 0.0177, 0.0173, 0.0138, 0.0188, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 00:35:37,872 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32234.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:36:35,197 INFO [train2.py:809] (1/4) Epoch 9, batch 400, loss[ctc_loss=0.1149, att_loss=0.2612, loss=0.2319, over 17274.00 frames. utt_duration=1258 frames, utt_pad_proportion=0.01343, over 55.00 utterances.], tot_loss[ctc_loss=0.1169, att_loss=0.2572, loss=0.2292, over 2845866.32 frames. utt_duration=1243 frames, utt_pad_proportion=0.05006, over 9168.28 utterances.], batch size: 55, lr: 1.24e-02, grad_scale: 16.0 2023-03-08 00:36:36,909 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32271.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:37:32,242 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32306.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:37:48,587 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.723e+02 2.567e+02 3.123e+02 3.914e+02 5.946e+02, threshold=6.245e+02, percent-clipped=0.0 2023-03-08 00:37:54,621 INFO [train2.py:809] (1/4) Epoch 9, batch 450, loss[ctc_loss=0.08556, att_loss=0.2341, loss=0.2044, over 16029.00 frames. utt_duration=1604 frames, utt_pad_proportion=0.006041, over 40.00 utterances.], tot_loss[ctc_loss=0.1181, att_loss=0.2583, loss=0.2302, over 2943056.79 frames. utt_duration=1208 frames, utt_pad_proportion=0.05857, over 9754.72 utterances.], batch size: 40, lr: 1.24e-02, grad_scale: 16.0 2023-03-08 00:38:59,120 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32360.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:39:14,593 INFO [train2.py:809] (1/4) Epoch 9, batch 500, loss[ctc_loss=0.1153, att_loss=0.2434, loss=0.2178, over 15647.00 frames. utt_duration=1693 frames, utt_pad_proportion=0.007948, over 37.00 utterances.], tot_loss[ctc_loss=0.1175, att_loss=0.2575, loss=0.2295, over 3017124.32 frames. utt_duration=1250 frames, utt_pad_proportion=0.04969, over 9665.49 utterances.], batch size: 37, lr: 1.24e-02, grad_scale: 16.0 2023-03-08 00:40:15,776 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=32408.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:40:28,732 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.910e+02 2.692e+02 3.196e+02 3.958e+02 9.983e+02, threshold=6.392e+02, percent-clipped=1.0 2023-03-08 00:40:35,576 INFO [train2.py:809] (1/4) Epoch 9, batch 550, loss[ctc_loss=0.1527, att_loss=0.2876, loss=0.2606, over 17031.00 frames. utt_duration=1312 frames, utt_pad_proportion=0.00862, over 52.00 utterances.], tot_loss[ctc_loss=0.1169, att_loss=0.2568, loss=0.2288, over 3077499.67 frames. utt_duration=1267 frames, utt_pad_proportion=0.04617, over 9727.89 utterances.], batch size: 52, lr: 1.24e-02, grad_scale: 16.0 2023-03-08 00:41:04,751 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.0437, 3.3481, 3.3430, 2.9003, 3.2392, 3.4442, 3.2481, 2.0777], device='cuda:1'), covar=tensor([0.1589, 0.2125, 0.2895, 0.5958, 0.4299, 0.3833, 0.1191, 0.9360], device='cuda:1'), in_proj_covar=tensor([0.0088, 0.0099, 0.0105, 0.0168, 0.0089, 0.0152, 0.0088, 0.0157], device='cuda:1'), out_proj_covar=tensor([8.1302e-05, 8.2725e-05, 9.1554e-05, 1.3318e-04, 7.9138e-05, 1.2285e-04, 7.4586e-05, 1.2540e-04], device='cuda:1') 2023-03-08 00:41:13,844 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32444.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:41:35,380 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.56 vs. limit=5.0 2023-03-08 00:41:55,743 INFO [train2.py:809] (1/4) Epoch 9, batch 600, loss[ctc_loss=0.09972, att_loss=0.2552, loss=0.2241, over 16478.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.005981, over 46.00 utterances.], tot_loss[ctc_loss=0.1167, att_loss=0.2568, loss=0.2288, over 3128753.07 frames. utt_duration=1285 frames, utt_pad_proportion=0.04056, over 9751.98 utterances.], batch size: 46, lr: 1.23e-02, grad_scale: 16.0 2023-03-08 00:42:10,433 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32479.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:42:29,840 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=32492.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:43:07,804 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.381e+02 2.723e+02 3.556e+02 4.400e+02 8.502e+02, threshold=7.112e+02, percent-clipped=9.0 2023-03-08 00:43:14,505 INFO [train2.py:809] (1/4) Epoch 9, batch 650, loss[ctc_loss=0.1236, att_loss=0.2486, loss=0.2236, over 16116.00 frames. utt_duration=1536 frames, utt_pad_proportion=0.006206, over 42.00 utterances.], tot_loss[ctc_loss=0.1172, att_loss=0.2566, loss=0.2287, over 3159329.88 frames. utt_duration=1282 frames, utt_pad_proportion=0.04258, over 9866.72 utterances.], batch size: 42, lr: 1.23e-02, grad_scale: 16.0 2023-03-08 00:43:25,726 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=32527.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:43:36,639 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32534.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:44:17,157 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-03-08 00:44:33,111 INFO [train2.py:809] (1/4) Epoch 9, batch 700, loss[ctc_loss=0.1395, att_loss=0.2526, loss=0.23, over 15985.00 frames. utt_duration=1600 frames, utt_pad_proportion=0.007951, over 40.00 utterances.], tot_loss[ctc_loss=0.1182, att_loss=0.2572, loss=0.2294, over 3175731.52 frames. utt_duration=1265 frames, utt_pad_proportion=0.04976, over 10055.71 utterances.], batch size: 40, lr: 1.23e-02, grad_scale: 16.0 2023-03-08 00:44:35,660 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32571.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:44:52,242 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=32582.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:45:31,390 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32606.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:45:47,001 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.724e+02 2.817e+02 3.384e+02 3.932e+02 1.069e+03, threshold=6.768e+02, percent-clipped=2.0 2023-03-08 00:45:51,660 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=32619.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:45:53,046 INFO [train2.py:809] (1/4) Epoch 9, batch 750, loss[ctc_loss=0.114, att_loss=0.244, loss=0.218, over 15929.00 frames. utt_duration=1636 frames, utt_pad_proportion=0.006371, over 39.00 utterances.], tot_loss[ctc_loss=0.1185, att_loss=0.2579, loss=0.23, over 3206331.00 frames. utt_duration=1269 frames, utt_pad_proportion=0.04686, over 10116.15 utterances.], batch size: 39, lr: 1.23e-02, grad_scale: 16.0 2023-03-08 00:46:47,260 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=32654.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:47:12,887 INFO [train2.py:809] (1/4) Epoch 9, batch 800, loss[ctc_loss=0.1058, att_loss=0.2428, loss=0.2154, over 16427.00 frames. utt_duration=1495 frames, utt_pad_proportion=0.005398, over 44.00 utterances.], tot_loss[ctc_loss=0.1188, att_loss=0.2591, loss=0.231, over 3230438.68 frames. utt_duration=1252 frames, utt_pad_proportion=0.04934, over 10337.38 utterances.], batch size: 44, lr: 1.23e-02, grad_scale: 16.0 2023-03-08 00:47:40,584 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0053, 4.1612, 4.0258, 4.1288, 4.4564, 4.1903, 4.0042, 2.2092], device='cuda:1'), covar=tensor([0.0352, 0.0301, 0.0286, 0.0215, 0.1075, 0.0296, 0.0318, 0.2244], device='cuda:1'), in_proj_covar=tensor([0.0125, 0.0121, 0.0124, 0.0129, 0.0308, 0.0125, 0.0114, 0.0220], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 00:48:28,947 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.641e+02 2.761e+02 3.783e+02 4.545e+02 8.141e+02, threshold=7.567e+02, percent-clipped=8.0 2023-03-08 00:48:33,637 INFO [train2.py:809] (1/4) Epoch 9, batch 850, loss[ctc_loss=0.09811, att_loss=0.2396, loss=0.2113, over 15908.00 frames. utt_duration=1633 frames, utt_pad_proportion=0.007295, over 39.00 utterances.], tot_loss[ctc_loss=0.1188, att_loss=0.2586, loss=0.2306, over 3232507.19 frames. utt_duration=1233 frames, utt_pad_proportion=0.05472, over 10496.52 utterances.], batch size: 39, lr: 1.23e-02, grad_scale: 8.0 2023-03-08 00:49:07,072 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6084, 1.9949, 5.0142, 3.7920, 2.9494, 4.4332, 4.6292, 4.6736], device='cuda:1'), covar=tensor([0.0216, 0.2108, 0.0120, 0.1099, 0.1931, 0.0195, 0.0127, 0.0223], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0243, 0.0126, 0.0300, 0.0280, 0.0182, 0.0111, 0.0145], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-03-08 00:49:54,783 INFO [train2.py:809] (1/4) Epoch 9, batch 900, loss[ctc_loss=0.144, att_loss=0.2531, loss=0.2313, over 15621.00 frames. utt_duration=1690 frames, utt_pad_proportion=0.01025, over 37.00 utterances.], tot_loss[ctc_loss=0.1183, att_loss=0.258, loss=0.2301, over 3240576.12 frames. utt_duration=1236 frames, utt_pad_proportion=0.0553, over 10496.12 utterances.], batch size: 37, lr: 1.23e-02, grad_scale: 8.0 2023-03-08 00:51:11,047 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.656e+02 2.630e+02 3.149e+02 3.930e+02 7.487e+02, threshold=6.299e+02, percent-clipped=0.0 2023-03-08 00:51:15,681 INFO [train2.py:809] (1/4) Epoch 9, batch 950, loss[ctc_loss=0.1224, att_loss=0.2635, loss=0.2352, over 16482.00 frames. utt_duration=1435 frames, utt_pad_proportion=0.006308, over 46.00 utterances.], tot_loss[ctc_loss=0.1179, att_loss=0.2575, loss=0.2296, over 3243164.19 frames. utt_duration=1255 frames, utt_pad_proportion=0.05403, over 10348.55 utterances.], batch size: 46, lr: 1.23e-02, grad_scale: 8.0 2023-03-08 00:51:58,902 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4991, 4.3873, 4.6299, 4.7117, 5.1256, 4.7020, 4.5635, 2.1567], device='cuda:1'), covar=tensor([0.0283, 0.0312, 0.0190, 0.0176, 0.1099, 0.0222, 0.0270, 0.2468], device='cuda:1'), in_proj_covar=tensor([0.0125, 0.0121, 0.0123, 0.0129, 0.0307, 0.0123, 0.0115, 0.0219], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 00:52:00,479 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1618, 4.9710, 4.8827, 2.4209, 1.9874, 2.8113, 3.2015, 3.7572], device='cuda:1'), covar=tensor([0.0560, 0.0149, 0.0176, 0.3574, 0.5677, 0.2361, 0.1578, 0.1611], device='cuda:1'), in_proj_covar=tensor([0.0329, 0.0210, 0.0238, 0.0193, 0.0360, 0.0345, 0.0226, 0.0355], device='cuda:1'), out_proj_covar=tensor([1.5513e-04, 8.0067e-05, 1.0458e-04, 8.9652e-05, 1.6219e-04, 1.4467e-04, 8.9980e-05, 1.5531e-04], device='cuda:1') 2023-03-08 00:52:11,792 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0878, 5.2259, 5.0366, 2.6192, 1.9075, 2.9240, 4.0055, 3.8170], device='cuda:1'), covar=tensor([0.0673, 0.0174, 0.0203, 0.3136, 0.6211, 0.2402, 0.0950, 0.1750], device='cuda:1'), in_proj_covar=tensor([0.0328, 0.0209, 0.0238, 0.0192, 0.0359, 0.0343, 0.0225, 0.0353], device='cuda:1'), out_proj_covar=tensor([1.5464e-04, 7.9742e-05, 1.0436e-04, 8.9017e-05, 1.6155e-04, 1.4413e-04, 8.9728e-05, 1.5469e-04], device='cuda:1') 2023-03-08 00:52:35,535 INFO [train2.py:809] (1/4) Epoch 9, batch 1000, loss[ctc_loss=0.126, att_loss=0.2662, loss=0.2382, over 16692.00 frames. utt_duration=1453 frames, utt_pad_proportion=0.005535, over 46.00 utterances.], tot_loss[ctc_loss=0.1177, att_loss=0.2576, loss=0.2296, over 3256309.36 frames. utt_duration=1253 frames, utt_pad_proportion=0.05193, over 10407.26 utterances.], batch size: 46, lr: 1.23e-02, grad_scale: 8.0 2023-03-08 00:53:42,795 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6290, 2.4521, 5.0616, 4.0652, 3.0490, 4.4581, 4.7858, 4.7949], device='cuda:1'), covar=tensor([0.0228, 0.1795, 0.0142, 0.1007, 0.1900, 0.0212, 0.0108, 0.0182], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0239, 0.0124, 0.0294, 0.0274, 0.0179, 0.0109, 0.0143], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-03-08 00:53:50,523 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.538e+02 2.656e+02 3.223e+02 3.962e+02 9.648e+02, threshold=6.446e+02, percent-clipped=5.0 2023-03-08 00:53:55,324 INFO [train2.py:809] (1/4) Epoch 9, batch 1050, loss[ctc_loss=0.1111, att_loss=0.2512, loss=0.2232, over 15946.00 frames. utt_duration=1557 frames, utt_pad_proportion=0.007539, over 41.00 utterances.], tot_loss[ctc_loss=0.1173, att_loss=0.2578, loss=0.2297, over 3266942.13 frames. utt_duration=1246 frames, utt_pad_proportion=0.05152, over 10502.31 utterances.], batch size: 41, lr: 1.23e-02, grad_scale: 8.0 2023-03-08 00:54:38,717 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5386, 2.7895, 3.7061, 2.7630, 3.6015, 4.6708, 4.5244, 3.3450], device='cuda:1'), covar=tensor([0.0396, 0.1665, 0.1009, 0.1532, 0.0978, 0.0648, 0.0336, 0.1300], device='cuda:1'), in_proj_covar=tensor([0.0222, 0.0221, 0.0233, 0.0200, 0.0229, 0.0272, 0.0204, 0.0214], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:1') 2023-03-08 00:55:10,096 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=32966.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:55:14,528 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9739, 5.1396, 5.4950, 5.3942, 5.3014, 5.9214, 5.1114, 6.0107], device='cuda:1'), covar=tensor([0.0608, 0.0727, 0.0639, 0.1058, 0.2015, 0.0830, 0.0638, 0.0552], device='cuda:1'), in_proj_covar=tensor([0.0648, 0.0395, 0.0451, 0.0513, 0.0691, 0.0454, 0.0365, 0.0445], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 00:55:15,879 INFO [train2.py:809] (1/4) Epoch 9, batch 1100, loss[ctc_loss=0.1107, att_loss=0.2571, loss=0.2278, over 16472.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.007031, over 46.00 utterances.], tot_loss[ctc_loss=0.1161, att_loss=0.2565, loss=0.2284, over 3260944.38 frames. utt_duration=1264 frames, utt_pad_proportion=0.0509, over 10335.45 utterances.], batch size: 46, lr: 1.23e-02, grad_scale: 8.0 2023-03-08 00:55:33,778 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.92 vs. limit=5.0 2023-03-08 00:56:31,654 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.884e+02 3.040e+02 3.595e+02 4.615e+02 9.742e+02, threshold=7.190e+02, percent-clipped=7.0 2023-03-08 00:56:36,388 INFO [train2.py:809] (1/4) Epoch 9, batch 1150, loss[ctc_loss=0.1112, att_loss=0.2541, loss=0.2255, over 16389.00 frames. utt_duration=1491 frames, utt_pad_proportion=0.008372, over 44.00 utterances.], tot_loss[ctc_loss=0.1173, att_loss=0.2573, loss=0.2293, over 3259130.31 frames. utt_duration=1238 frames, utt_pad_proportion=0.05745, over 10538.83 utterances.], batch size: 44, lr: 1.22e-02, grad_scale: 8.0 2023-03-08 00:56:40,624 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-03-08 00:56:47,287 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.6131, 5.9324, 5.2525, 5.7186, 5.5369, 5.1848, 5.2907, 5.1584], device='cuda:1'), covar=tensor([0.1267, 0.0850, 0.0916, 0.0791, 0.0888, 0.1414, 0.2289, 0.2403], device='cuda:1'), in_proj_covar=tensor([0.0400, 0.0448, 0.0348, 0.0361, 0.0331, 0.0400, 0.0475, 0.0426], 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') 2023-03-08 00:56:47,519 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33027.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:57:54,913 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33069.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 00:57:56,105 INFO [train2.py:809] (1/4) Epoch 9, batch 1200, loss[ctc_loss=0.1023, att_loss=0.2447, loss=0.2162, over 16265.00 frames. utt_duration=1514 frames, utt_pad_proportion=0.007574, over 43.00 utterances.], tot_loss[ctc_loss=0.118, att_loss=0.2579, loss=0.2299, over 3266517.97 frames. utt_duration=1249 frames, utt_pad_proportion=0.05363, over 10474.26 utterances.], batch size: 43, lr: 1.22e-02, grad_scale: 8.0 2023-03-08 00:58:28,568 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.4600, 2.5934, 3.3297, 4.2688, 3.8985, 3.9207, 2.7182, 1.8909], device='cuda:1'), covar=tensor([0.0569, 0.2239, 0.1000, 0.0502, 0.0689, 0.0378, 0.1646, 0.2513], device='cuda:1'), in_proj_covar=tensor([0.0161, 0.0201, 0.0185, 0.0177, 0.0172, 0.0137, 0.0187, 0.0173], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 00:59:12,584 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.824e+02 2.465e+02 3.116e+02 3.703e+02 8.772e+02, threshold=6.232e+02, percent-clipped=2.0 2023-03-08 00:59:17,165 INFO [train2.py:809] (1/4) Epoch 9, batch 1250, loss[ctc_loss=0.1131, att_loss=0.2647, loss=0.2344, over 16693.00 frames. utt_duration=1453 frames, utt_pad_proportion=0.005431, over 46.00 utterances.], tot_loss[ctc_loss=0.1185, att_loss=0.2582, loss=0.2303, over 3269088.19 frames. utt_duration=1223 frames, utt_pad_proportion=0.05937, over 10703.34 utterances.], batch size: 46, lr: 1.22e-02, grad_scale: 8.0 2023-03-08 00:59:32,911 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33130.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:00:14,103 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33155.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:00:36,807 INFO [train2.py:809] (1/4) Epoch 9, batch 1300, loss[ctc_loss=0.1211, att_loss=0.2558, loss=0.2288, over 16020.00 frames. utt_duration=1603 frames, utt_pad_proportion=0.006537, over 40.00 utterances.], tot_loss[ctc_loss=0.119, att_loss=0.2587, loss=0.2308, over 3272844.21 frames. utt_duration=1236 frames, utt_pad_proportion=0.05548, over 10605.79 utterances.], batch size: 40, lr: 1.22e-02, grad_scale: 8.0 2023-03-08 01:00:57,111 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.24 vs. limit=5.0 2023-03-08 01:01:51,068 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.3672, 3.7233, 3.4756, 3.1207, 3.7035, 3.6197, 3.5096, 2.3853], device='cuda:1'), covar=tensor([0.1391, 0.1898, 0.4104, 0.6297, 0.1123, 0.4391, 0.0821, 0.8749], device='cuda:1'), in_proj_covar=tensor([0.0088, 0.0100, 0.0107, 0.0173, 0.0090, 0.0157, 0.0090, 0.0157], device='cuda:1'), out_proj_covar=tensor([8.1823e-05, 8.4019e-05, 9.3892e-05, 1.3711e-04, 8.1110e-05, 1.2691e-04, 7.7063e-05, 1.2635e-04], device='cuda:1') 2023-03-08 01:01:51,088 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33216.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:01:52,243 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.789e+02 2.474e+02 3.066e+02 3.872e+02 7.426e+02, threshold=6.132e+02, percent-clipped=4.0 2023-03-08 01:01:56,794 INFO [train2.py:809] (1/4) Epoch 9, batch 1350, loss[ctc_loss=0.1216, att_loss=0.2362, loss=0.2133, over 15882.00 frames. utt_duration=1630 frames, utt_pad_proportion=0.009487, over 39.00 utterances.], tot_loss[ctc_loss=0.1188, att_loss=0.258, loss=0.2301, over 3268188.67 frames. utt_duration=1222 frames, utt_pad_proportion=0.0604, over 10710.64 utterances.], batch size: 39, lr: 1.22e-02, grad_scale: 8.0 2023-03-08 01:03:16,721 INFO [train2.py:809] (1/4) Epoch 9, batch 1400, loss[ctc_loss=0.1317, att_loss=0.263, loss=0.2367, over 16289.00 frames. utt_duration=1517 frames, utt_pad_proportion=0.006747, over 43.00 utterances.], tot_loss[ctc_loss=0.1184, att_loss=0.2581, loss=0.2302, over 3264172.72 frames. utt_duration=1198 frames, utt_pad_proportion=0.06774, over 10911.95 utterances.], batch size: 43, lr: 1.22e-02, grad_scale: 8.0 2023-03-08 01:03:24,731 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7066, 5.1126, 4.9746, 5.0022, 5.1273, 5.1418, 4.8175, 4.5972], device='cuda:1'), covar=tensor([0.1260, 0.0431, 0.0269, 0.0419, 0.0277, 0.0280, 0.0298, 0.0322], device='cuda:1'), in_proj_covar=tensor([0.0432, 0.0269, 0.0213, 0.0255, 0.0320, 0.0339, 0.0258, 0.0296], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 01:03:34,897 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2116, 5.2430, 5.1317, 2.8671, 2.1391, 2.9315, 3.7220, 3.9394], device='cuda:1'), covar=tensor([0.0587, 0.0291, 0.0256, 0.3246, 0.5532, 0.2452, 0.1315, 0.1724], device='cuda:1'), in_proj_covar=tensor([0.0319, 0.0207, 0.0230, 0.0186, 0.0349, 0.0336, 0.0219, 0.0346], device='cuda:1'), out_proj_covar=tensor([1.5028e-04, 7.9596e-05, 1.0097e-04, 8.5056e-05, 1.5680e-04, 1.4056e-04, 8.7551e-05, 1.5090e-04], device='cuda:1') 2023-03-08 01:04:32,132 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.539e+02 2.521e+02 2.872e+02 3.748e+02 7.311e+02, threshold=5.745e+02, percent-clipped=4.0 2023-03-08 01:04:37,047 INFO [train2.py:809] (1/4) Epoch 9, batch 1450, loss[ctc_loss=0.1918, att_loss=0.3047, loss=0.2821, over 13633.00 frames. utt_duration=375.1 frames, utt_pad_proportion=0.3465, over 146.00 utterances.], tot_loss[ctc_loss=0.1174, att_loss=0.2579, loss=0.2298, over 3268221.05 frames. utt_duration=1205 frames, utt_pad_proportion=0.0643, over 10862.87 utterances.], batch size: 146, lr: 1.22e-02, grad_scale: 8.0 2023-03-08 01:04:40,495 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33322.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:05:49,125 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0023, 5.4092, 4.7644, 5.4639, 4.7323, 5.0664, 5.5107, 5.2824], device='cuda:1'), covar=tensor([0.0507, 0.0280, 0.0888, 0.0188, 0.0453, 0.0195, 0.0179, 0.0157], device='cuda:1'), in_proj_covar=tensor([0.0304, 0.0234, 0.0297, 0.0223, 0.0242, 0.0190, 0.0218, 0.0211], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0005, 0.0006, 0.0005, 0.0005, 0.0004, 0.0005, 0.0004], device='cuda:1') 2023-03-08 01:05:56,599 INFO [train2.py:809] (1/4) Epoch 9, batch 1500, loss[ctc_loss=0.1214, att_loss=0.2706, loss=0.2408, over 17405.00 frames. utt_duration=1010 frames, utt_pad_proportion=0.04681, over 69.00 utterances.], tot_loss[ctc_loss=0.1171, att_loss=0.2574, loss=0.2293, over 3264599.83 frames. utt_duration=1211 frames, utt_pad_proportion=0.06546, over 10792.68 utterances.], batch size: 69, lr: 1.22e-02, grad_scale: 8.0 2023-03-08 01:07:00,591 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-03-08 01:07:11,748 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.661e+02 2.556e+02 3.254e+02 4.022e+02 8.037e+02, threshold=6.507e+02, percent-clipped=3.0 2023-03-08 01:07:16,447 INFO [train2.py:809] (1/4) Epoch 9, batch 1550, loss[ctc_loss=0.08567, att_loss=0.2367, loss=0.2065, over 15766.00 frames. utt_duration=1661 frames, utt_pad_proportion=0.008761, over 38.00 utterances.], tot_loss[ctc_loss=0.1183, att_loss=0.2588, loss=0.2307, over 3272651.69 frames. utt_duration=1199 frames, utt_pad_proportion=0.06714, over 10931.71 utterances.], batch size: 38, lr: 1.22e-02, grad_scale: 8.0 2023-03-08 01:07:24,575 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33425.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:07:29,954 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5531, 2.7202, 4.9874, 3.9790, 3.0097, 4.5642, 4.9151, 4.7544], device='cuda:1'), covar=tensor([0.0231, 0.1592, 0.0170, 0.1046, 0.2001, 0.0213, 0.0097, 0.0184], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0239, 0.0124, 0.0295, 0.0271, 0.0178, 0.0107, 0.0143], device='cuda:1'), out_proj_covar=tensor([1.3202e-04, 2.0171e-04, 1.1119e-04, 2.4591e-04, 2.4177e-04, 1.5939e-04, 9.9075e-05, 1.3651e-04], device='cuda:1') 2023-03-08 01:08:37,542 INFO [train2.py:809] (1/4) Epoch 9, batch 1600, loss[ctc_loss=0.1476, att_loss=0.2813, loss=0.2545, over 17670.00 frames. utt_duration=1011 frames, utt_pad_proportion=0.0416, over 70.00 utterances.], tot_loss[ctc_loss=0.1182, att_loss=0.2587, loss=0.2306, over 3274091.05 frames. utt_duration=1198 frames, utt_pad_proportion=0.06723, over 10948.76 utterances.], batch size: 70, lr: 1.22e-02, grad_scale: 8.0 2023-03-08 01:08:40,895 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33472.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:09:43,899 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33511.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:09:53,210 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.888e+02 2.746e+02 3.273e+02 4.611e+02 8.460e+02, threshold=6.545e+02, percent-clipped=8.0 2023-03-08 01:09:55,262 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.4020, 1.5702, 1.7944, 1.8439, 1.8932, 1.9908, 1.8163, 2.9357], device='cuda:1'), covar=tensor([0.0922, 0.4039, 0.3521, 0.2183, 0.1849, 0.2139, 0.3773, 0.1242], device='cuda:1'), in_proj_covar=tensor([0.0078, 0.0085, 0.0089, 0.0074, 0.0076, 0.0072, 0.0084, 0.0065], device='cuda:1'), out_proj_covar=tensor([4.4629e-05, 5.4800e-05, 5.5126e-05, 4.5979e-05, 4.3563e-05, 4.6880e-05, 5.3253e-05, 4.3882e-05], device='cuda:1') 2023-03-08 01:09:57,951 INFO [train2.py:809] (1/4) Epoch 9, batch 1650, loss[ctc_loss=0.1057, att_loss=0.2554, loss=0.2255, over 17027.00 frames. utt_duration=1337 frames, utt_pad_proportion=0.007438, over 51.00 utterances.], tot_loss[ctc_loss=0.1181, att_loss=0.2589, loss=0.2307, over 3275927.59 frames. utt_duration=1231 frames, utt_pad_proportion=0.05813, over 10655.79 utterances.], batch size: 51, lr: 1.22e-02, grad_scale: 8.0 2023-03-08 01:10:18,994 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33533.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:11:18,160 INFO [train2.py:809] (1/4) Epoch 9, batch 1700, loss[ctc_loss=0.1016, att_loss=0.2522, loss=0.2221, over 16763.00 frames. utt_duration=1398 frames, utt_pad_proportion=0.006806, over 48.00 utterances.], tot_loss[ctc_loss=0.1165, att_loss=0.2574, loss=0.2292, over 3273848.46 frames. utt_duration=1243 frames, utt_pad_proportion=0.05686, over 10549.54 utterances.], batch size: 48, lr: 1.21e-02, grad_scale: 8.0 2023-03-08 01:12:35,922 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.622e+02 2.664e+02 3.328e+02 4.191e+02 7.435e+02, threshold=6.656e+02, percent-clipped=3.0 2023-03-08 01:12:39,100 INFO [train2.py:809] (1/4) Epoch 9, batch 1750, loss[ctc_loss=0.1122, att_loss=0.2626, loss=0.2325, over 16869.00 frames. utt_duration=1378 frames, utt_pad_proportion=0.007581, over 49.00 utterances.], tot_loss[ctc_loss=0.1163, att_loss=0.2575, loss=0.2293, over 3269382.99 frames. utt_duration=1242 frames, utt_pad_proportion=0.05759, over 10544.44 utterances.], batch size: 49, lr: 1.21e-02, grad_scale: 4.0 2023-03-08 01:12:42,473 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=33622.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:13:58,114 INFO [train2.py:809] (1/4) Epoch 9, batch 1800, loss[ctc_loss=0.118, att_loss=0.2618, loss=0.2331, over 16308.00 frames. utt_duration=1519 frames, utt_pad_proportion=0.005452, over 43.00 utterances.], tot_loss[ctc_loss=0.1151, att_loss=0.2571, loss=0.2287, over 3277053.04 frames. utt_duration=1259 frames, utt_pad_proportion=0.0509, over 10421.50 utterances.], batch size: 43, lr: 1.21e-02, grad_scale: 4.0 2023-03-08 01:13:58,229 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=33670.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:14:06,896 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.65 vs. limit=2.0 2023-03-08 01:14:09,855 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33677.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:15:12,709 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-03-08 01:15:14,676 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.479e+02 2.773e+02 3.319e+02 4.110e+02 8.879e+02, threshold=6.639e+02, percent-clipped=6.0 2023-03-08 01:15:17,757 INFO [train2.py:809] (1/4) Epoch 9, batch 1850, loss[ctc_loss=0.1205, att_loss=0.267, loss=0.2377, over 17035.00 frames. utt_duration=1287 frames, utt_pad_proportion=0.009869, over 53.00 utterances.], tot_loss[ctc_loss=0.1158, att_loss=0.2577, loss=0.2293, over 3277772.38 frames. utt_duration=1248 frames, utt_pad_proportion=0.05343, over 10521.06 utterances.], batch size: 53, lr: 1.21e-02, grad_scale: 4.0 2023-03-08 01:15:26,847 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=33725.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:15:47,531 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33738.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 01:16:37,509 INFO [train2.py:809] (1/4) Epoch 9, batch 1900, loss[ctc_loss=0.1123, att_loss=0.2694, loss=0.238, over 16952.00 frames. utt_duration=1358 frames, utt_pad_proportion=0.008181, over 50.00 utterances.], tot_loss[ctc_loss=0.1161, att_loss=0.2573, loss=0.2291, over 3271238.51 frames. utt_duration=1233 frames, utt_pad_proportion=0.06044, over 10625.21 utterances.], batch size: 50, lr: 1.21e-02, grad_scale: 4.0 2023-03-08 01:16:42,297 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=33773.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:17:29,772 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2086, 5.1355, 5.0846, 2.3051, 2.0089, 2.4955, 3.4358, 3.8133], device='cuda:1'), covar=tensor([0.0548, 0.0184, 0.0204, 0.3480, 0.5778, 0.2994, 0.1247, 0.1855], device='cuda:1'), in_proj_covar=tensor([0.0318, 0.0206, 0.0228, 0.0188, 0.0350, 0.0338, 0.0219, 0.0347], device='cuda:1'), out_proj_covar=tensor([1.4980e-04, 7.8922e-05, 9.9996e-05, 8.5784e-05, 1.5641e-04, 1.4129e-04, 8.7303e-05, 1.5115e-04], device='cuda:1') 2023-03-08 01:17:44,861 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=33811.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:17:55,252 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.768e+02 2.711e+02 3.222e+02 4.216e+02 1.081e+03, threshold=6.444e+02, percent-clipped=7.0 2023-03-08 01:17:58,264 INFO [train2.py:809] (1/4) Epoch 9, batch 1950, loss[ctc_loss=0.08154, att_loss=0.2244, loss=0.1958, over 15769.00 frames. utt_duration=1662 frames, utt_pad_proportion=0.008589, over 38.00 utterances.], tot_loss[ctc_loss=0.1165, att_loss=0.2578, loss=0.2295, over 3274756.76 frames. utt_duration=1213 frames, utt_pad_proportion=0.06426, over 10811.44 utterances.], batch size: 38, lr: 1.21e-02, grad_scale: 4.0 2023-03-08 01:18:11,631 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33828.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:18:52,919 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-03-08 01:19:01,159 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=33859.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:19:17,451 INFO [train2.py:809] (1/4) Epoch 9, batch 2000, loss[ctc_loss=0.109, att_loss=0.2552, loss=0.2259, over 16127.00 frames. utt_duration=1537 frames, utt_pad_proportion=0.00559, over 42.00 utterances.], tot_loss[ctc_loss=0.115, att_loss=0.2563, loss=0.228, over 3260280.91 frames. utt_duration=1233 frames, utt_pad_proportion=0.0628, over 10589.83 utterances.], batch size: 42, lr: 1.21e-02, grad_scale: 8.0 2023-03-08 01:19:37,882 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-03-08 01:20:10,001 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0194, 5.1430, 4.9907, 2.2604, 2.0132, 2.6151, 3.4709, 3.8594], device='cuda:1'), covar=tensor([0.0623, 0.0187, 0.0212, 0.4187, 0.6006, 0.2892, 0.1402, 0.1756], device='cuda:1'), in_proj_covar=tensor([0.0319, 0.0208, 0.0227, 0.0189, 0.0350, 0.0337, 0.0220, 0.0347], device='cuda:1'), out_proj_covar=tensor([1.5014e-04, 7.8933e-05, 9.9087e-05, 8.6000e-05, 1.5599e-04, 1.4069e-04, 8.7286e-05, 1.5075e-04], device='cuda:1') 2023-03-08 01:20:34,522 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.651e+02 2.627e+02 3.033e+02 3.756e+02 1.034e+03, threshold=6.065e+02, percent-clipped=4.0 2023-03-08 01:20:35,323 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.63 vs. limit=5.0 2023-03-08 01:20:37,624 INFO [train2.py:809] (1/4) Epoch 9, batch 2050, loss[ctc_loss=0.1022, att_loss=0.2252, loss=0.2006, over 14140.00 frames. utt_duration=1826 frames, utt_pad_proportion=0.04632, over 31.00 utterances.], tot_loss[ctc_loss=0.1152, att_loss=0.2568, loss=0.2285, over 3271430.90 frames. utt_duration=1232 frames, utt_pad_proportion=0.059, over 10638.39 utterances.], batch size: 31, lr: 1.21e-02, grad_scale: 8.0 2023-03-08 01:20:56,031 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7594, 4.6923, 4.6605, 4.7472, 5.1345, 4.8732, 4.6301, 2.2649], device='cuda:1'), covar=tensor([0.0171, 0.0229, 0.0169, 0.0160, 0.0851, 0.0153, 0.0209, 0.2271], device='cuda:1'), in_proj_covar=tensor([0.0129, 0.0126, 0.0125, 0.0131, 0.0316, 0.0124, 0.0116, 0.0221], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 01:21:03,742 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.0314, 2.5931, 2.4985, 2.2487, 2.4443, 2.3448, 2.4660, 1.7984], device='cuda:1'), covar=tensor([0.1620, 0.1676, 0.3238, 0.8257, 0.3039, 0.4345, 0.1460, 0.7666], device='cuda:1'), in_proj_covar=tensor([0.0083, 0.0100, 0.0105, 0.0169, 0.0089, 0.0153, 0.0089, 0.0154], device='cuda:1'), out_proj_covar=tensor([7.8662e-05, 8.4550e-05, 9.2661e-05, 1.3517e-04, 8.0382e-05, 1.2539e-04, 7.6775e-05, 1.2400e-04], device='cuda:1') 2023-03-08 01:21:08,885 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33939.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:21:49,029 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6866, 2.5108, 5.0500, 3.9283, 3.2127, 4.5409, 4.9236, 4.7668], device='cuda:1'), covar=tensor([0.0236, 0.1682, 0.0205, 0.1147, 0.1859, 0.0195, 0.0099, 0.0203], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0248, 0.0129, 0.0307, 0.0284, 0.0183, 0.0113, 0.0147], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0003, 0.0003, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-03-08 01:21:58,314 INFO [train2.py:809] (1/4) Epoch 9, batch 2100, loss[ctc_loss=0.1095, att_loss=0.252, loss=0.2235, over 16262.00 frames. utt_duration=1514 frames, utt_pad_proportion=0.00762, over 43.00 utterances.], tot_loss[ctc_loss=0.1167, att_loss=0.2579, loss=0.2297, over 3272227.74 frames. utt_duration=1214 frames, utt_pad_proportion=0.06471, over 10791.91 utterances.], batch size: 43, lr: 1.21e-02, grad_scale: 8.0 2023-03-08 01:22:33,415 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7270, 2.9614, 3.7382, 3.0656, 3.5671, 4.8303, 4.5978, 3.5580], device='cuda:1'), covar=tensor([0.0366, 0.1531, 0.1064, 0.1375, 0.1091, 0.0719, 0.0425, 0.1228], device='cuda:1'), in_proj_covar=tensor([0.0224, 0.0221, 0.0233, 0.0204, 0.0235, 0.0276, 0.0206, 0.0213], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:1') 2023-03-08 01:22:50,624 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34000.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:23:18,327 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.708e+02 2.606e+02 3.129e+02 4.184e+02 1.323e+03, threshold=6.258e+02, percent-clipped=9.0 2023-03-08 01:23:22,287 INFO [train2.py:809] (1/4) Epoch 9, batch 2150, loss[ctc_loss=0.1555, att_loss=0.2812, loss=0.256, over 16637.00 frames. utt_duration=673.8 frames, utt_pad_proportion=0.1535, over 99.00 utterances.], tot_loss[ctc_loss=0.1164, att_loss=0.2569, loss=0.2288, over 3265527.81 frames. utt_duration=1231 frames, utt_pad_proportion=0.0614, over 10620.97 utterances.], batch size: 99, lr: 1.21e-02, grad_scale: 8.0 2023-03-08 01:23:42,965 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34033.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 01:24:02,274 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34044.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:24:17,922 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4771, 5.0942, 4.8096, 4.9408, 5.0456, 4.6861, 3.5464, 5.0661], device='cuda:1'), covar=tensor([0.0096, 0.0097, 0.0101, 0.0085, 0.0097, 0.0103, 0.0599, 0.0174], device='cuda:1'), in_proj_covar=tensor([0.0065, 0.0064, 0.0076, 0.0049, 0.0052, 0.0062, 0.0084, 0.0083], 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') 2023-03-08 01:24:42,987 INFO [train2.py:809] (1/4) Epoch 9, batch 2200, loss[ctc_loss=0.1267, att_loss=0.2635, loss=0.2361, over 17448.00 frames. utt_duration=1013 frames, utt_pad_proportion=0.04238, over 69.00 utterances.], tot_loss[ctc_loss=0.1163, att_loss=0.2568, loss=0.2287, over 3263895.90 frames. utt_duration=1234 frames, utt_pad_proportion=0.06112, over 10589.12 utterances.], batch size: 69, lr: 1.21e-02, grad_scale: 8.0 2023-03-08 01:25:09,228 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-03-08 01:25:32,249 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.5046, 3.8773, 3.0579, 3.6763, 4.0162, 3.5667, 2.5433, 4.3151], device='cuda:1'), covar=tensor([0.1275, 0.0433, 0.1199, 0.0604, 0.0577, 0.0703, 0.1112, 0.0528], device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0173, 0.0200, 0.0168, 0.0218, 0.0206, 0.0176, 0.0234], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-08 01:25:38,253 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34105.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:25:58,666 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.871e+02 2.581e+02 3.160e+02 3.967e+02 1.179e+03, threshold=6.319e+02, percent-clipped=5.0 2023-03-08 01:26:01,794 INFO [train2.py:809] (1/4) Epoch 9, batch 2250, loss[ctc_loss=0.1012, att_loss=0.2427, loss=0.2144, over 16757.00 frames. utt_duration=1398 frames, utt_pad_proportion=0.007176, over 48.00 utterances.], tot_loss[ctc_loss=0.1156, att_loss=0.2565, loss=0.2283, over 3262569.48 frames. utt_duration=1246 frames, utt_pad_proportion=0.05754, over 10484.11 utterances.], batch size: 48, lr: 1.21e-02, grad_scale: 8.0 2023-03-08 01:26:14,008 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1508, 5.2518, 5.1951, 3.1589, 4.9453, 4.6670, 4.6773, 2.7597], device='cuda:1'), covar=tensor([0.0205, 0.0080, 0.0149, 0.0883, 0.0094, 0.0176, 0.0238, 0.1376], device='cuda:1'), in_proj_covar=tensor([0.0058, 0.0078, 0.0066, 0.0101, 0.0066, 0.0090, 0.0088, 0.0098], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-08 01:26:14,018 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34128.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:26:42,340 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34145.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:27:21,330 INFO [train2.py:809] (1/4) Epoch 9, batch 2300, loss[ctc_loss=0.1138, att_loss=0.2509, loss=0.2235, over 16397.00 frames. utt_duration=1492 frames, utt_pad_proportion=0.007213, over 44.00 utterances.], tot_loss[ctc_loss=0.115, att_loss=0.2561, loss=0.2279, over 3258625.64 frames. utt_duration=1270 frames, utt_pad_proportion=0.05187, over 10277.74 utterances.], batch size: 44, lr: 1.20e-02, grad_scale: 8.0 2023-03-08 01:27:30,966 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=34176.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:28:00,245 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-03-08 01:28:05,724 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3774, 4.5147, 4.6175, 5.3230, 2.5156, 5.0046, 2.4406, 1.8485], device='cuda:1'), covar=tensor([0.0201, 0.0177, 0.0804, 0.0073, 0.2059, 0.0119, 0.1985, 0.2006], device='cuda:1'), in_proj_covar=tensor([0.0129, 0.0106, 0.0260, 0.0108, 0.0225, 0.0104, 0.0231, 0.0206], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 01:28:16,086 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-03-08 01:28:20,208 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34206.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:28:32,499 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8269, 5.2115, 5.0947, 5.0903, 5.2461, 5.2272, 4.8902, 4.6540], device='cuda:1'), covar=tensor([0.1219, 0.0494, 0.0223, 0.0436, 0.0284, 0.0301, 0.0307, 0.0360], device='cuda:1'), in_proj_covar=tensor([0.0429, 0.0270, 0.0212, 0.0253, 0.0317, 0.0343, 0.0258, 0.0296], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 01:28:39,055 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.844e+02 2.559e+02 3.119e+02 3.946e+02 1.160e+03, threshold=6.239e+02, percent-clipped=6.0 2023-03-08 01:28:42,212 INFO [train2.py:809] (1/4) Epoch 9, batch 2350, loss[ctc_loss=0.101, att_loss=0.2528, loss=0.2224, over 16546.00 frames. utt_duration=1472 frames, utt_pad_proportion=0.005942, over 45.00 utterances.], tot_loss[ctc_loss=0.1158, att_loss=0.2568, loss=0.2286, over 3257960.99 frames. utt_duration=1239 frames, utt_pad_proportion=0.06031, over 10527.31 utterances.], batch size: 45, lr: 1.20e-02, grad_scale: 8.0 2023-03-08 01:29:09,296 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3272, 4.5773, 4.5330, 4.5851, 4.6672, 4.6416, 4.3597, 4.2130], device='cuda:1'), covar=tensor([0.1076, 0.0621, 0.0247, 0.0421, 0.0298, 0.0324, 0.0295, 0.0359], device='cuda:1'), in_proj_covar=tensor([0.0432, 0.0274, 0.0215, 0.0255, 0.0320, 0.0346, 0.0261, 0.0300], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 01:29:26,602 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-03-08 01:30:03,504 INFO [train2.py:809] (1/4) Epoch 9, batch 2400, loss[ctc_loss=0.1074, att_loss=0.2455, loss=0.2179, over 16181.00 frames. utt_duration=1580 frames, utt_pad_proportion=0.006213, over 41.00 utterances.], tot_loss[ctc_loss=0.1163, att_loss=0.2576, loss=0.2294, over 3266306.63 frames. utt_duration=1231 frames, utt_pad_proportion=0.0603, over 10623.47 utterances.], batch size: 41, lr: 1.20e-02, grad_scale: 8.0 2023-03-08 01:30:14,759 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([1.8681, 3.4880, 3.4119, 2.7881, 3.2923, 3.2211, 3.1994, 2.0440], device='cuda:1'), covar=tensor([0.1435, 0.1035, 0.1289, 0.5044, 0.1452, 0.2725, 0.0937, 0.8243], device='cuda:1'), in_proj_covar=tensor([0.0084, 0.0100, 0.0105, 0.0169, 0.0091, 0.0155, 0.0089, 0.0156], device='cuda:1'), out_proj_covar=tensor([7.9200e-05, 8.4778e-05, 9.2464e-05, 1.3472e-04, 8.1857e-05, 1.2644e-04, 7.7230e-05, 1.2534e-04], device='cuda:1') 2023-03-08 01:30:19,343 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4167, 5.1601, 4.8919, 5.1978, 5.0871, 4.8092, 3.5098, 4.9196], device='cuda:1'), covar=tensor([0.0140, 0.0135, 0.0113, 0.0062, 0.0110, 0.0108, 0.0683, 0.0256], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0067, 0.0080, 0.0050, 0.0054, 0.0064, 0.0087, 0.0086], 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') 2023-03-08 01:30:45,163 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34295.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:31:21,641 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.665e+02 2.539e+02 3.254e+02 3.979e+02 8.947e+02, threshold=6.508e+02, percent-clipped=3.0 2023-03-08 01:31:24,811 INFO [train2.py:809] (1/4) Epoch 9, batch 2450, loss[ctc_loss=0.104, att_loss=0.2452, loss=0.217, over 16386.00 frames. utt_duration=1491 frames, utt_pad_proportion=0.007666, over 44.00 utterances.], tot_loss[ctc_loss=0.1178, att_loss=0.2593, loss=0.231, over 3278773.59 frames. utt_duration=1197 frames, utt_pad_proportion=0.06405, over 10968.20 utterances.], batch size: 44, lr: 1.20e-02, grad_scale: 8.0 2023-03-08 01:31:47,050 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34333.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:31:54,729 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-03-08 01:32:12,571 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7211, 4.1764, 4.4692, 4.2974, 4.2852, 4.6566, 4.2808, 4.7331], device='cuda:1'), covar=tensor([0.0824, 0.0730, 0.0608, 0.1015, 0.1610, 0.0873, 0.1873, 0.0651], device='cuda:1'), in_proj_covar=tensor([0.0652, 0.0393, 0.0447, 0.0514, 0.0688, 0.0457, 0.0370, 0.0444], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 01:32:45,731 INFO [train2.py:809] (1/4) Epoch 9, batch 2500, loss[ctc_loss=0.09214, att_loss=0.2367, loss=0.2078, over 16171.00 frames. utt_duration=1579 frames, utt_pad_proportion=0.006842, over 41.00 utterances.], tot_loss[ctc_loss=0.1169, att_loss=0.2583, loss=0.23, over 3278854.22 frames. utt_duration=1209 frames, utt_pad_proportion=0.06246, over 10862.61 utterances.], batch size: 41, lr: 1.20e-02, grad_scale: 8.0 2023-03-08 01:32:54,278 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-03-08 01:33:03,135 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=34381.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:33:35,233 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34400.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:34:03,750 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.632e+02 2.575e+02 3.136e+02 3.976e+02 7.316e+02, threshold=6.272e+02, percent-clipped=1.0 2023-03-08 01:34:06,912 INFO [train2.py:809] (1/4) Epoch 9, batch 2550, loss[ctc_loss=0.1129, att_loss=0.2413, loss=0.2156, over 15989.00 frames. utt_duration=1600 frames, utt_pad_proportion=0.007781, over 40.00 utterances.], tot_loss[ctc_loss=0.1175, att_loss=0.2588, loss=0.2305, over 3276301.73 frames. utt_duration=1184 frames, utt_pad_proportion=0.06977, over 11080.77 utterances.], batch size: 40, lr: 1.20e-02, grad_scale: 8.0 2023-03-08 01:34:07,319 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.9370, 3.7047, 3.1081, 3.4967, 3.7334, 3.5790, 2.6905, 4.1720], device='cuda:1'), covar=tensor([0.0951, 0.0443, 0.1081, 0.0599, 0.0722, 0.0645, 0.0976, 0.0429], device='cuda:1'), in_proj_covar=tensor([0.0178, 0.0173, 0.0198, 0.0167, 0.0216, 0.0205, 0.0176, 0.0234], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-08 01:34:18,220 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-03-08 01:35:28,486 INFO [train2.py:809] (1/4) Epoch 9, batch 2600, loss[ctc_loss=0.1214, att_loss=0.266, loss=0.2371, over 16619.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.005708, over 47.00 utterances.], tot_loss[ctc_loss=0.1171, att_loss=0.2585, loss=0.2302, over 3277248.58 frames. utt_duration=1181 frames, utt_pad_proportion=0.07057, over 11117.35 utterances.], batch size: 47, lr: 1.20e-02, grad_scale: 8.0 2023-03-08 01:35:30,315 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.6726, 3.5710, 2.9199, 3.2498, 3.7174, 3.4314, 2.4725, 4.0541], device='cuda:1'), covar=tensor([0.1031, 0.0462, 0.1088, 0.0642, 0.0546, 0.0653, 0.0972, 0.0377], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0171, 0.0197, 0.0166, 0.0212, 0.0203, 0.0173, 0.0231], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-08 01:36:19,409 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34501.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:36:46,607 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.469e+02 2.544e+02 3.202e+02 3.743e+02 1.159e+03, threshold=6.403e+02, percent-clipped=3.0 2023-03-08 01:36:49,761 INFO [train2.py:809] (1/4) Epoch 9, batch 2650, loss[ctc_loss=0.09553, att_loss=0.2224, loss=0.197, over 15763.00 frames. utt_duration=1661 frames, utt_pad_proportion=0.009044, over 38.00 utterances.], tot_loss[ctc_loss=0.1165, att_loss=0.2583, loss=0.23, over 3281366.99 frames. utt_duration=1200 frames, utt_pad_proportion=0.0649, over 10954.69 utterances.], batch size: 38, lr: 1.20e-02, grad_scale: 8.0 2023-03-08 01:36:59,525 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.8191, 3.9029, 3.7180, 2.9527, 3.6818, 3.8238, 3.6258, 2.3907], device='cuda:1'), covar=tensor([0.1370, 0.1304, 0.3315, 1.0410, 0.2201, 0.4488, 0.1475, 1.0698], device='cuda:1'), in_proj_covar=tensor([0.0085, 0.0105, 0.0110, 0.0175, 0.0093, 0.0158, 0.0092, 0.0157], device='cuda:1'), out_proj_covar=tensor([8.1040e-05, 8.9103e-05, 9.7408e-05, 1.3975e-04, 8.3158e-05, 1.2962e-04, 7.9755e-05, 1.2728e-04], device='cuda:1') 2023-03-08 01:37:37,580 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0533, 5.0082, 4.9476, 2.4092, 1.9231, 2.8023, 3.1232, 3.7563], device='cuda:1'), covar=tensor([0.0604, 0.0160, 0.0200, 0.3303, 0.5321, 0.2404, 0.1727, 0.1681], device='cuda:1'), in_proj_covar=tensor([0.0316, 0.0206, 0.0221, 0.0185, 0.0343, 0.0329, 0.0219, 0.0342], device='cuda:1'), out_proj_covar=tensor([1.4801e-04, 7.8726e-05, 9.6671e-05, 8.4002e-05, 1.5272e-04, 1.3690e-04, 8.7118e-05, 1.4828e-04], device='cuda:1') 2023-03-08 01:37:51,496 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.8569, 5.2155, 5.3455, 5.3113, 5.2842, 5.8516, 5.0165, 5.9442], device='cuda:1'), covar=tensor([0.0631, 0.0636, 0.0725, 0.0974, 0.1775, 0.0953, 0.0610, 0.0662], device='cuda:1'), in_proj_covar=tensor([0.0652, 0.0396, 0.0455, 0.0516, 0.0695, 0.0459, 0.0373, 0.0444], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 01:38:10,644 INFO [train2.py:809] (1/4) Epoch 9, batch 2700, loss[ctc_loss=0.1229, att_loss=0.2467, loss=0.222, over 16407.00 frames. utt_duration=1493 frames, utt_pad_proportion=0.006789, over 44.00 utterances.], tot_loss[ctc_loss=0.117, att_loss=0.2583, loss=0.2301, over 3280596.31 frames. utt_duration=1202 frames, utt_pad_proportion=0.06495, over 10934.49 utterances.], batch size: 44, lr: 1.20e-02, grad_scale: 8.0 2023-03-08 01:38:47,936 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34593.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:38:51,088 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34595.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:39:27,959 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 2.068e+02 2.571e+02 3.320e+02 3.912e+02 1.043e+03, threshold=6.640e+02, percent-clipped=3.0 2023-03-08 01:39:30,984 INFO [train2.py:809] (1/4) Epoch 9, batch 2750, loss[ctc_loss=0.1211, att_loss=0.264, loss=0.2354, over 17045.00 frames. utt_duration=1313 frames, utt_pad_proportion=0.009303, over 52.00 utterances.], tot_loss[ctc_loss=0.1169, att_loss=0.2581, loss=0.2299, over 3277020.17 frames. utt_duration=1191 frames, utt_pad_proportion=0.06759, over 11020.04 utterances.], batch size: 52, lr: 1.20e-02, grad_scale: 8.0 2023-03-08 01:39:34,475 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34622.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:39:41,367 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34626.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:39:53,498 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0481, 5.4443, 5.3263, 5.2407, 5.4746, 5.4474, 5.1555, 4.9690], device='cuda:1'), covar=tensor([0.1058, 0.0390, 0.0175, 0.0413, 0.0254, 0.0251, 0.0256, 0.0252], device='cuda:1'), in_proj_covar=tensor([0.0435, 0.0273, 0.0218, 0.0259, 0.0320, 0.0344, 0.0262, 0.0299], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 01:40:08,114 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=34643.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:40:11,877 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.27 vs. limit=5.0 2023-03-08 01:40:25,686 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34654.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:40:50,937 INFO [train2.py:809] (1/4) Epoch 9, batch 2800, loss[ctc_loss=0.129, att_loss=0.2718, loss=0.2432, over 17054.00 frames. utt_duration=1313 frames, utt_pad_proportion=0.008911, over 52.00 utterances.], tot_loss[ctc_loss=0.1165, att_loss=0.258, loss=0.2297, over 3285216.33 frames. utt_duration=1227 frames, utt_pad_proportion=0.05744, over 10725.99 utterances.], batch size: 52, lr: 1.20e-02, grad_scale: 8.0 2023-03-08 01:40:52,853 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0334, 5.4298, 5.5252, 5.4828, 5.4145, 6.0264, 5.2225, 6.0561], device='cuda:1'), covar=tensor([0.0625, 0.0616, 0.0684, 0.0859, 0.1820, 0.0670, 0.0587, 0.0590], device='cuda:1'), in_proj_covar=tensor([0.0658, 0.0402, 0.0459, 0.0523, 0.0702, 0.0468, 0.0376, 0.0450], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 01:40:56,125 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3652, 4.9129, 4.6075, 5.0249, 4.9580, 4.5415, 3.3572, 4.7764], device='cuda:1'), covar=tensor([0.0126, 0.0135, 0.0143, 0.0072, 0.0112, 0.0138, 0.0732, 0.0298], device='cuda:1'), in_proj_covar=tensor([0.0068, 0.0067, 0.0079, 0.0049, 0.0053, 0.0064, 0.0088, 0.0086], 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') 2023-03-08 01:41:12,996 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34683.0, num_to_drop=1, layers_to_drop={3} 2023-03-08 01:41:19,733 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34687.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:41:33,718 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.8437, 1.5523, 2.5923, 2.5171, 3.3443, 2.3474, 1.7985, 3.0438], device='cuda:1'), covar=tensor([0.0490, 0.4346, 0.2554, 0.1576, 0.0657, 0.1351, 0.2467, 0.0784], device='cuda:1'), in_proj_covar=tensor([0.0073, 0.0085, 0.0083, 0.0073, 0.0072, 0.0068, 0.0080, 0.0059], device='cuda:1'), out_proj_covar=tensor([4.3229e-05, 5.4883e-05, 5.3455e-05, 4.5627e-05, 4.2426e-05, 4.5215e-05, 5.1694e-05, 4.0946e-05], device='cuda:1') 2023-03-08 01:41:40,026 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34700.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:41:55,192 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34710.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 01:42:07,902 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.706e+02 2.689e+02 3.286e+02 4.031e+02 7.048e+02, threshold=6.572e+02, percent-clipped=5.0 2023-03-08 01:42:11,095 INFO [train2.py:809] (1/4) Epoch 9, batch 2850, loss[ctc_loss=0.1046, att_loss=0.2433, loss=0.2155, over 15389.00 frames. utt_duration=1760 frames, utt_pad_proportion=0.01004, over 35.00 utterances.], tot_loss[ctc_loss=0.1172, att_loss=0.2581, loss=0.2299, over 3281214.75 frames. utt_duration=1210 frames, utt_pad_proportion=0.06201, over 10856.00 utterances.], batch size: 35, lr: 1.20e-02, grad_scale: 8.0 2023-03-08 01:42:16,055 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.1508, 2.4224, 2.8368, 3.9801, 3.6929, 3.7418, 2.7069, 1.9298], device='cuda:1'), covar=tensor([0.0660, 0.2199, 0.1185, 0.0569, 0.0664, 0.0381, 0.1492, 0.2431], device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0200, 0.0186, 0.0178, 0.0174, 0.0139, 0.0188, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 01:42:28,316 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.6333, 1.4243, 2.3011, 1.9215, 3.0567, 2.1012, 1.9909, 2.6952], device='cuda:1'), covar=tensor([0.0350, 0.4371, 0.2859, 0.1719, 0.0607, 0.1079, 0.2004, 0.0852], device='cuda:1'), in_proj_covar=tensor([0.0073, 0.0086, 0.0084, 0.0074, 0.0072, 0.0068, 0.0080, 0.0060], device='cuda:1'), out_proj_covar=tensor([4.3549e-05, 5.5151e-05, 5.3823e-05, 4.6299e-05, 4.2672e-05, 4.5330e-05, 5.2060e-05, 4.1349e-05], device='cuda:1') 2023-03-08 01:42:56,385 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=34748.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:43:31,252 INFO [train2.py:809] (1/4) Epoch 9, batch 2900, loss[ctc_loss=0.1021, att_loss=0.2505, loss=0.2208, over 16387.00 frames. utt_duration=1491 frames, utt_pad_proportion=0.006632, over 44.00 utterances.], tot_loss[ctc_loss=0.1164, att_loss=0.2577, loss=0.2294, over 3278315.50 frames. utt_duration=1225 frames, utt_pad_proportion=0.05972, over 10714.06 utterances.], batch size: 44, lr: 1.19e-02, grad_scale: 8.0 2023-03-08 01:43:33,243 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34771.0, num_to_drop=1, layers_to_drop={3} 2023-03-08 01:44:22,048 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34801.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:44:48,786 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.507e+02 2.823e+02 3.310e+02 4.171e+02 9.231e+02, threshold=6.619e+02, percent-clipped=4.0 2023-03-08 01:44:51,953 INFO [train2.py:809] (1/4) Epoch 9, batch 2950, loss[ctc_loss=0.1089, att_loss=0.2352, loss=0.2099, over 15771.00 frames. utt_duration=1662 frames, utt_pad_proportion=0.008432, over 38.00 utterances.], tot_loss[ctc_loss=0.1157, att_loss=0.2565, loss=0.2283, over 3270347.21 frames. utt_duration=1237 frames, utt_pad_proportion=0.05871, over 10590.08 utterances.], batch size: 38, lr: 1.19e-02, grad_scale: 8.0 2023-03-08 01:45:40,212 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=34849.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:45:45,036 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5251, 4.9281, 4.7088, 4.9403, 5.0368, 4.6976, 3.3154, 4.8171], device='cuda:1'), covar=tensor([0.0095, 0.0123, 0.0112, 0.0082, 0.0093, 0.0087, 0.0675, 0.0224], device='cuda:1'), in_proj_covar=tensor([0.0066, 0.0066, 0.0078, 0.0048, 0.0052, 0.0062, 0.0086, 0.0084], 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') 2023-03-08 01:46:13,926 INFO [train2.py:809] (1/4) Epoch 9, batch 3000, loss[ctc_loss=0.1045, att_loss=0.2597, loss=0.2287, over 16644.00 frames. utt_duration=1418 frames, utt_pad_proportion=0.004243, over 47.00 utterances.], tot_loss[ctc_loss=0.1147, att_loss=0.2567, loss=0.2283, over 3274814.60 frames. utt_duration=1241 frames, utt_pad_proportion=0.05616, over 10568.85 utterances.], batch size: 47, lr: 1.19e-02, grad_scale: 8.0 2023-03-08 01:46:13,927 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-08 01:46:32,381 INFO [train2.py:843] (1/4) Epoch 9, validation: ctc_loss=0.05401, att_loss=0.2408, loss=0.2035, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-08 01:46:32,383 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-08 01:47:16,960 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2308, 1.8935, 1.9808, 2.2651, 2.8794, 1.8071, 2.1202, 3.0209], device='cuda:1'), covar=tensor([0.0616, 0.3898, 0.3512, 0.1459, 0.0789, 0.1754, 0.2192, 0.0843], device='cuda:1'), in_proj_covar=tensor([0.0073, 0.0084, 0.0086, 0.0073, 0.0071, 0.0068, 0.0081, 0.0060], device='cuda:1'), out_proj_covar=tensor([4.3634e-05, 5.4308e-05, 5.4433e-05, 4.5908e-05, 4.2325e-05, 4.5695e-05, 5.2418e-05, 4.1556e-05], device='cuda:1') 2023-03-08 01:47:49,369 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.566e+02 2.760e+02 3.196e+02 3.997e+02 8.726e+02, threshold=6.392e+02, percent-clipped=4.0 2023-03-08 01:47:52,532 INFO [train2.py:809] (1/4) Epoch 9, batch 3050, loss[ctc_loss=0.09838, att_loss=0.2495, loss=0.2193, over 16114.00 frames. utt_duration=1536 frames, utt_pad_proportion=0.006849, over 42.00 utterances.], tot_loss[ctc_loss=0.1159, att_loss=0.2573, loss=0.229, over 3277678.71 frames. utt_duration=1227 frames, utt_pad_proportion=0.05908, over 10695.25 utterances.], batch size: 42, lr: 1.19e-02, grad_scale: 8.0 2023-03-08 01:48:26,657 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8623, 5.2172, 4.7649, 5.3410, 4.6137, 4.9595, 5.3866, 5.1218], device='cuda:1'), covar=tensor([0.0518, 0.0296, 0.0711, 0.0214, 0.0458, 0.0214, 0.0178, 0.0188], device='cuda:1'), in_proj_covar=tensor([0.0308, 0.0238, 0.0298, 0.0231, 0.0244, 0.0190, 0.0220, 0.0216], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 01:48:33,172 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.9039, 3.8139, 2.9922, 3.5558, 3.8919, 3.6950, 2.6656, 4.3348], device='cuda:1'), covar=tensor([0.1003, 0.0480, 0.1164, 0.0569, 0.0670, 0.0554, 0.0908, 0.0400], device='cuda:1'), in_proj_covar=tensor([0.0182, 0.0175, 0.0201, 0.0169, 0.0220, 0.0206, 0.0178, 0.0237], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-08 01:48:40,006 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34949.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:48:52,512 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0728, 5.1755, 4.9981, 2.3415, 1.9166, 2.7813, 3.8824, 3.7985], device='cuda:1'), covar=tensor([0.0716, 0.0217, 0.0229, 0.4397, 0.6223, 0.2789, 0.1087, 0.1938], device='cuda:1'), in_proj_covar=tensor([0.0326, 0.0209, 0.0227, 0.0194, 0.0353, 0.0336, 0.0224, 0.0352], device='cuda:1'), out_proj_covar=tensor([1.5278e-04, 8.0334e-05, 9.9257e-05, 8.8281e-05, 1.5693e-04, 1.3962e-04, 8.9283e-05, 1.5256e-04], device='cuda:1') 2023-03-08 01:48:55,402 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.1789, 3.1000, 2.9111, 2.3607, 2.8193, 2.8541, 2.7939, 2.0130], device='cuda:1'), covar=tensor([0.1453, 0.2080, 0.4954, 0.9569, 0.8457, 0.7447, 0.1804, 0.9836], device='cuda:1'), in_proj_covar=tensor([0.0087, 0.0107, 0.0113, 0.0177, 0.0096, 0.0164, 0.0093, 0.0160], device='cuda:1'), out_proj_covar=tensor([8.2856e-05, 9.0941e-05, 1.0006e-04, 1.4239e-04, 8.6435e-05, 1.3363e-04, 8.0617e-05, 1.2993e-04], device='cuda:1') 2023-03-08 01:49:12,666 INFO [train2.py:809] (1/4) Epoch 9, batch 3100, loss[ctc_loss=0.09249, att_loss=0.2476, loss=0.2166, over 16544.00 frames. utt_duration=1472 frames, utt_pad_proportion=0.006077, over 45.00 utterances.], tot_loss[ctc_loss=0.1156, att_loss=0.2569, loss=0.2286, over 3272961.55 frames. utt_duration=1238 frames, utt_pad_proportion=0.05828, over 10585.20 utterances.], batch size: 45, lr: 1.19e-02, grad_scale: 8.0 2023-03-08 01:49:20,702 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6767, 2.8352, 5.0348, 3.7069, 2.8917, 4.4773, 4.9927, 4.7463], device='cuda:1'), covar=tensor([0.0227, 0.1625, 0.0199, 0.1238, 0.2264, 0.0234, 0.0089, 0.0196], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0245, 0.0127, 0.0304, 0.0280, 0.0182, 0.0110, 0.0145], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-03-08 01:49:25,033 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34978.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 01:49:32,536 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34982.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:50:30,869 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.761e+02 2.700e+02 3.373e+02 4.220e+02 7.587e+02, threshold=6.745e+02, percent-clipped=2.0 2023-03-08 01:50:34,018 INFO [train2.py:809] (1/4) Epoch 9, batch 3150, loss[ctc_loss=0.09801, att_loss=0.2382, loss=0.2101, over 16300.00 frames. utt_duration=1518 frames, utt_pad_proportion=0.006107, over 43.00 utterances.], tot_loss[ctc_loss=0.1154, att_loss=0.2565, loss=0.2283, over 3264781.17 frames. utt_duration=1220 frames, utt_pad_proportion=0.06445, over 10714.08 utterances.], batch size: 43, lr: 1.19e-02, grad_scale: 8.0 2023-03-08 01:51:48,855 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35066.0, num_to_drop=1, layers_to_drop={3} 2023-03-08 01:51:54,783 INFO [train2.py:809] (1/4) Epoch 9, batch 3200, loss[ctc_loss=0.13, att_loss=0.2712, loss=0.243, over 16495.00 frames. utt_duration=1436 frames, utt_pad_proportion=0.005555, over 46.00 utterances.], tot_loss[ctc_loss=0.114, att_loss=0.2558, loss=0.2275, over 3274131.01 frames. utt_duration=1243 frames, utt_pad_proportion=0.05624, over 10548.60 utterances.], batch size: 46, lr: 1.19e-02, grad_scale: 8.0 2023-03-08 01:52:31,405 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35092.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:52:51,111 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6150, 2.6859, 5.1021, 3.9757, 3.0029, 4.4034, 4.9164, 4.5709], device='cuda:1'), covar=tensor([0.0252, 0.1673, 0.0163, 0.1066, 0.2045, 0.0234, 0.0099, 0.0215], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0245, 0.0128, 0.0304, 0.0279, 0.0182, 0.0110, 0.0146], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-03-08 01:53:13,426 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.527e+02 2.431e+02 2.913e+02 3.671e+02 9.523e+02, threshold=5.826e+02, percent-clipped=2.0 2023-03-08 01:53:16,523 INFO [train2.py:809] (1/4) Epoch 9, batch 3250, loss[ctc_loss=0.1221, att_loss=0.2547, loss=0.2282, over 16333.00 frames. utt_duration=1453 frames, utt_pad_proportion=0.006156, over 45.00 utterances.], tot_loss[ctc_loss=0.114, att_loss=0.2558, loss=0.2275, over 3270461.71 frames. utt_duration=1237 frames, utt_pad_proportion=0.05906, over 10591.57 utterances.], batch size: 45, lr: 1.19e-02, grad_scale: 8.0 2023-03-08 01:54:10,642 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35153.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:54:37,147 INFO [train2.py:809] (1/4) Epoch 9, batch 3300, loss[ctc_loss=0.1332, att_loss=0.2706, loss=0.2431, over 16480.00 frames. utt_duration=1435 frames, utt_pad_proportion=0.00586, over 46.00 utterances.], tot_loss[ctc_loss=0.1143, att_loss=0.2559, loss=0.2276, over 3260820.32 frames. utt_duration=1223 frames, utt_pad_proportion=0.06339, over 10675.02 utterances.], batch size: 46, lr: 1.19e-02, grad_scale: 8.0 2023-03-08 01:55:15,920 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1233, 5.4442, 4.8539, 5.4594, 4.8483, 5.1363, 5.6042, 5.2778], device='cuda:1'), covar=tensor([0.0436, 0.0201, 0.0819, 0.0201, 0.0415, 0.0159, 0.0210, 0.0166], device='cuda:1'), in_proj_covar=tensor([0.0309, 0.0242, 0.0305, 0.0234, 0.0246, 0.0194, 0.0222, 0.0216], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0005, 0.0006, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 01:55:25,201 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7996, 4.8670, 4.6117, 4.8280, 5.2489, 4.7664, 4.6931, 2.5267], device='cuda:1'), covar=tensor([0.0182, 0.0197, 0.0245, 0.0182, 0.0789, 0.0171, 0.0276, 0.2096], device='cuda:1'), in_proj_covar=tensor([0.0125, 0.0123, 0.0126, 0.0131, 0.0318, 0.0122, 0.0115, 0.0215], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 01:55:35,823 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-03-08 01:55:55,871 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.545e+02 2.550e+02 3.164e+02 3.869e+02 8.354e+02, threshold=6.328e+02, percent-clipped=5.0 2023-03-08 01:55:58,993 INFO [train2.py:809] (1/4) Epoch 9, batch 3350, loss[ctc_loss=0.1266, att_loss=0.269, loss=0.2405, over 17061.00 frames. utt_duration=1314 frames, utt_pad_proportion=0.008258, over 52.00 utterances.], tot_loss[ctc_loss=0.1142, att_loss=0.2565, loss=0.228, over 3270953.18 frames. utt_duration=1233 frames, utt_pad_proportion=0.05879, over 10620.13 utterances.], batch size: 52, lr: 1.19e-02, grad_scale: 8.0 2023-03-08 01:56:28,811 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-03-08 01:56:46,163 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35249.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:57:18,720 INFO [train2.py:809] (1/4) Epoch 9, batch 3400, loss[ctc_loss=0.118, att_loss=0.2378, loss=0.2138, over 15341.00 frames. utt_duration=1755 frames, utt_pad_proportion=0.01199, over 35.00 utterances.], tot_loss[ctc_loss=0.116, att_loss=0.2575, loss=0.2292, over 3264965.79 frames. utt_duration=1217 frames, utt_pad_proportion=0.0636, over 10744.39 utterances.], batch size: 35, lr: 1.19e-02, grad_scale: 8.0 2023-03-08 01:57:32,786 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35278.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 01:57:38,867 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35282.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:58:03,244 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=35297.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:58:35,854 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-03-08 01:58:36,339 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.421e+02 2.749e+02 3.298e+02 4.168e+02 1.046e+03, threshold=6.595e+02, percent-clipped=7.0 2023-03-08 01:58:39,567 INFO [train2.py:809] (1/4) Epoch 9, batch 3450, loss[ctc_loss=0.1123, att_loss=0.2463, loss=0.2195, over 16011.00 frames. utt_duration=1603 frames, utt_pad_proportion=0.007477, over 40.00 utterances.], tot_loss[ctc_loss=0.1154, att_loss=0.2562, loss=0.228, over 3259349.76 frames. utt_duration=1228 frames, utt_pad_proportion=0.06285, over 10631.90 utterances.], batch size: 40, lr: 1.19e-02, grad_scale: 8.0 2023-03-08 01:58:50,429 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=35326.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:58:56,845 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=35330.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 01:59:21,341 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35345.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 01:59:55,015 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35366.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 02:00:01,143 INFO [train2.py:809] (1/4) Epoch 9, batch 3500, loss[ctc_loss=0.1805, att_loss=0.2955, loss=0.2725, over 14124.00 frames. utt_duration=388.4 frames, utt_pad_proportion=0.321, over 146.00 utterances.], tot_loss[ctc_loss=0.1164, att_loss=0.2571, loss=0.229, over 3262911.82 frames. utt_duration=1204 frames, utt_pad_proportion=0.06882, over 10856.10 utterances.], batch size: 146, lr: 1.18e-02, grad_scale: 8.0 2023-03-08 02:00:09,029 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35374.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:00:59,464 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35406.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 02:01:12,461 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=35414.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 02:01:18,494 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.648e+02 2.480e+02 3.195e+02 3.962e+02 8.803e+02, threshold=6.389e+02, percent-clipped=2.0 2023-03-08 02:01:21,712 INFO [train2.py:809] (1/4) Epoch 9, batch 3550, loss[ctc_loss=0.09878, att_loss=0.2518, loss=0.2212, over 16128.00 frames. utt_duration=1538 frames, utt_pad_proportion=0.006079, over 42.00 utterances.], tot_loss[ctc_loss=0.1167, att_loss=0.2576, loss=0.2294, over 3269786.72 frames. utt_duration=1210 frames, utt_pad_proportion=0.06614, over 10818.54 utterances.], batch size: 42, lr: 1.18e-02, grad_scale: 8.0 2023-03-08 02:01:46,291 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35435.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:02:06,779 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35448.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:02:41,171 INFO [train2.py:809] (1/4) Epoch 9, batch 3600, loss[ctc_loss=0.1106, att_loss=0.2615, loss=0.2313, over 17305.00 frames. utt_duration=1175 frames, utt_pad_proportion=0.0236, over 59.00 utterances.], tot_loss[ctc_loss=0.1171, att_loss=0.2579, loss=0.2297, over 3272512.74 frames. utt_duration=1217 frames, utt_pad_proportion=0.06213, over 10769.50 utterances.], batch size: 59, lr: 1.18e-02, grad_scale: 8.0 2023-03-08 02:02:52,479 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0424, 5.0408, 4.8709, 2.5594, 1.9819, 2.6326, 2.9716, 3.7739], device='cuda:1'), covar=tensor([0.0666, 0.0236, 0.0251, 0.3619, 0.5895, 0.2811, 0.2064, 0.1824], device='cuda:1'), in_proj_covar=tensor([0.0331, 0.0221, 0.0235, 0.0196, 0.0358, 0.0338, 0.0230, 0.0355], device='cuda:1'), out_proj_covar=tensor([1.5392e-04, 8.3622e-05, 1.0193e-04, 8.9365e-05, 1.5808e-04, 1.3966e-04, 9.0944e-05, 1.5284e-04], device='cuda:1') 2023-03-08 02:03:27,256 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2578, 5.2583, 5.1274, 2.6614, 2.1347, 2.9300, 3.5995, 3.8442], device='cuda:1'), covar=tensor([0.0570, 0.0269, 0.0205, 0.3914, 0.5715, 0.2447, 0.1426, 0.1878], device='cuda:1'), in_proj_covar=tensor([0.0328, 0.0219, 0.0233, 0.0195, 0.0355, 0.0337, 0.0228, 0.0351], device='cuda:1'), out_proj_covar=tensor([1.5268e-04, 8.2844e-05, 1.0065e-04, 8.8790e-05, 1.5708e-04, 1.3893e-04, 9.0267e-05, 1.5131e-04], device='cuda:1') 2023-03-08 02:03:58,105 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.720e+02 2.714e+02 3.545e+02 4.370e+02 8.238e+02, threshold=7.091e+02, percent-clipped=8.0 2023-03-08 02:04:01,308 INFO [train2.py:809] (1/4) Epoch 9, batch 3650, loss[ctc_loss=0.1152, att_loss=0.2678, loss=0.2373, over 17351.00 frames. utt_duration=1103 frames, utt_pad_proportion=0.03573, over 63.00 utterances.], tot_loss[ctc_loss=0.1162, att_loss=0.2572, loss=0.229, over 3265117.34 frames. utt_duration=1227 frames, utt_pad_proportion=0.06092, over 10657.04 utterances.], batch size: 63, lr: 1.18e-02, grad_scale: 8.0 2023-03-08 02:04:05,860 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.5555, 2.4210, 3.0984, 4.3337, 4.0238, 4.0097, 3.0085, 1.9278], device='cuda:1'), covar=tensor([0.0525, 0.2201, 0.1113, 0.0420, 0.0500, 0.0365, 0.1220, 0.2330], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0199, 0.0184, 0.0180, 0.0171, 0.0140, 0.0188, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 02:04:18,698 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35530.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:05:22,723 INFO [train2.py:809] (1/4) Epoch 9, batch 3700, loss[ctc_loss=0.1277, att_loss=0.273, loss=0.2439, over 16617.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.005738, over 47.00 utterances.], tot_loss[ctc_loss=0.116, att_loss=0.2567, loss=0.2285, over 3264593.46 frames. utt_duration=1257 frames, utt_pad_proportion=0.05357, over 10399.94 utterances.], batch size: 47, lr: 1.18e-02, grad_scale: 8.0 2023-03-08 02:05:31,635 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=5.03 vs. limit=5.0 2023-03-08 02:05:57,170 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35591.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:06:00,917 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2873, 4.7330, 4.7016, 4.9203, 2.2842, 4.6812, 3.3838, 1.5790], device='cuda:1'), covar=tensor([0.0268, 0.0133, 0.0614, 0.0133, 0.2327, 0.0166, 0.1325, 0.2034], device='cuda:1'), in_proj_covar=tensor([0.0129, 0.0106, 0.0254, 0.0108, 0.0222, 0.0105, 0.0224, 0.0201], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 02:06:09,801 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8886, 4.8457, 4.8491, 2.6273, 4.6334, 4.4514, 3.9785, 2.4511], device='cuda:1'), covar=tensor([0.0139, 0.0086, 0.0199, 0.1140, 0.0097, 0.0197, 0.0341, 0.1497], device='cuda:1'), in_proj_covar=tensor([0.0059, 0.0078, 0.0067, 0.0101, 0.0066, 0.0091, 0.0088, 0.0100], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-08 02:06:40,036 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.872e+02 2.486e+02 3.166e+02 4.039e+02 7.024e+02, threshold=6.332e+02, percent-clipped=0.0 2023-03-08 02:06:43,373 INFO [train2.py:809] (1/4) Epoch 9, batch 3750, loss[ctc_loss=0.1163, att_loss=0.2799, loss=0.2472, over 17450.00 frames. utt_duration=1110 frames, utt_pad_proportion=0.03093, over 63.00 utterances.], tot_loss[ctc_loss=0.1144, att_loss=0.256, loss=0.2277, over 3267995.80 frames. utt_duration=1270 frames, utt_pad_proportion=0.04854, over 10301.35 utterances.], batch size: 63, lr: 1.18e-02, grad_scale: 16.0 2023-03-08 02:07:36,790 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5233, 2.3538, 5.0650, 3.9659, 3.1014, 4.5388, 5.0239, 4.6485], device='cuda:1'), covar=tensor([0.0236, 0.1830, 0.0191, 0.0970, 0.1901, 0.0209, 0.0091, 0.0212], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0244, 0.0132, 0.0303, 0.0280, 0.0183, 0.0111, 0.0147], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-03-08 02:07:37,168 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.97 vs. limit=2.0 2023-03-08 02:08:04,809 INFO [train2.py:809] (1/4) Epoch 9, batch 3800, loss[ctc_loss=0.1286, att_loss=0.2409, loss=0.2185, over 15485.00 frames. utt_duration=1722 frames, utt_pad_proportion=0.009792, over 36.00 utterances.], tot_loss[ctc_loss=0.1131, att_loss=0.2553, loss=0.2268, over 3272735.11 frames. utt_duration=1291 frames, utt_pad_proportion=0.04255, over 10151.21 utterances.], batch size: 36, lr: 1.18e-02, grad_scale: 16.0 2023-03-08 02:08:56,066 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35701.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 02:09:23,355 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.337e+02 2.475e+02 3.126e+02 3.794e+02 8.902e+02, threshold=6.253e+02, percent-clipped=4.0 2023-03-08 02:09:26,478 INFO [train2.py:809] (1/4) Epoch 9, batch 3850, loss[ctc_loss=0.09368, att_loss=0.2405, loss=0.2111, over 16148.00 frames. utt_duration=1577 frames, utt_pad_proportion=0.007613, over 41.00 utterances.], tot_loss[ctc_loss=0.1126, att_loss=0.2555, loss=0.2269, over 3265709.52 frames. utt_duration=1277 frames, utt_pad_proportion=0.0463, over 10242.09 utterances.], batch size: 41, lr: 1.18e-02, grad_scale: 16.0 2023-03-08 02:09:42,257 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35730.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:09:53,194 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.6478, 1.3974, 2.3062, 1.7119, 2.3336, 2.2405, 2.5861, 2.0218], device='cuda:1'), covar=tensor([0.0546, 0.4325, 0.2798, 0.2188, 0.1451, 0.1313, 0.1723, 0.1662], device='cuda:1'), in_proj_covar=tensor([0.0073, 0.0085, 0.0089, 0.0075, 0.0075, 0.0069, 0.0080, 0.0063], device='cuda:1'), out_proj_covar=tensor([4.4322e-05, 5.5480e-05, 5.6826e-05, 4.7699e-05, 4.4867e-05, 4.6571e-05, 5.2929e-05, 4.4035e-05], device='cuda:1') 2023-03-08 02:10:10,755 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35748.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:10:45,209 INFO [train2.py:809] (1/4) Epoch 9, batch 3900, loss[ctc_loss=0.1209, att_loss=0.2392, loss=0.2155, over 12314.00 frames. utt_duration=1826 frames, utt_pad_proportion=0.1523, over 27.00 utterances.], tot_loss[ctc_loss=0.1127, att_loss=0.2552, loss=0.2267, over 3261341.67 frames. utt_duration=1295 frames, utt_pad_proportion=0.04368, over 10087.99 utterances.], batch size: 27, lr: 1.18e-02, grad_scale: 16.0 2023-03-08 02:10:47,096 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35771.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:11:26,179 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=35796.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:11:50,451 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35811.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:11:52,005 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7775, 4.5178, 4.5549, 4.4820, 4.9595, 4.7866, 4.4713, 2.0802], device='cuda:1'), covar=tensor([0.0166, 0.0287, 0.0213, 0.0207, 0.1052, 0.0164, 0.0238, 0.2452], device='cuda:1'), in_proj_covar=tensor([0.0127, 0.0123, 0.0128, 0.0134, 0.0323, 0.0122, 0.0115, 0.0220], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 02:12:00,559 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.615e+02 2.688e+02 3.135e+02 3.909e+02 9.741e+02, threshold=6.269e+02, percent-clipped=3.0 2023-03-08 02:12:03,703 INFO [train2.py:809] (1/4) Epoch 9, batch 3950, loss[ctc_loss=0.1163, att_loss=0.2403, loss=0.2155, over 15353.00 frames. utt_duration=1756 frames, utt_pad_proportion=0.01053, over 35.00 utterances.], tot_loss[ctc_loss=0.1125, att_loss=0.2553, loss=0.2267, over 3269868.71 frames. utt_duration=1294 frames, utt_pad_proportion=0.04264, over 10116.88 utterances.], batch size: 35, lr: 1.18e-02, grad_scale: 16.0 2023-03-08 02:12:22,932 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35832.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:13:22,886 INFO [train2.py:809] (1/4) Epoch 10, batch 0, loss[ctc_loss=0.1105, att_loss=0.2538, loss=0.2251, over 16278.00 frames. utt_duration=1516 frames, utt_pad_proportion=0.007356, over 43.00 utterances.], tot_loss[ctc_loss=0.1105, att_loss=0.2538, loss=0.2251, over 16278.00 frames. utt_duration=1516 frames, utt_pad_proportion=0.007356, over 43.00 utterances.], batch size: 43, lr: 1.12e-02, grad_scale: 16.0 2023-03-08 02:13:22,886 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-08 02:13:35,376 INFO [train2.py:843] (1/4) Epoch 10, validation: ctc_loss=0.0538, att_loss=0.2416, loss=0.204, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-08 02:13:35,377 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-08 02:14:04,788 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35872.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:14:25,939 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35886.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:14:44,588 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9621, 4.3521, 4.1337, 4.4796, 2.4903, 4.4333, 2.5860, 1.8981], device='cuda:1'), covar=tensor([0.0378, 0.0154, 0.0731, 0.0133, 0.1934, 0.0129, 0.1594, 0.1690], device='cuda:1'), in_proj_covar=tensor([0.0129, 0.0108, 0.0253, 0.0108, 0.0221, 0.0101, 0.0224, 0.0202], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 02:14:54,638 INFO [train2.py:809] (1/4) Epoch 10, batch 50, loss[ctc_loss=0.12, att_loss=0.2457, loss=0.2206, over 16015.00 frames. utt_duration=1603 frames, utt_pad_proportion=0.006568, over 40.00 utterances.], tot_loss[ctc_loss=0.1146, att_loss=0.257, loss=0.2285, over 740445.24 frames. utt_duration=1146 frames, utt_pad_proportion=0.0769, over 2587.93 utterances.], batch size: 40, lr: 1.12e-02, grad_scale: 16.0 2023-03-08 02:15:18,101 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.853e+02 2.508e+02 2.998e+02 4.058e+02 7.580e+02, threshold=5.995e+02, percent-clipped=4.0 2023-03-08 02:16:05,804 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.8546, 2.4174, 2.6751, 3.2817, 3.0267, 3.2500, 2.4692, 2.2164], device='cuda:1'), covar=tensor([0.0585, 0.1846, 0.0951, 0.0666, 0.0722, 0.0396, 0.1423, 0.1888], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0201, 0.0183, 0.0180, 0.0172, 0.0141, 0.0189, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 02:16:15,462 INFO [train2.py:809] (1/4) Epoch 10, batch 100, loss[ctc_loss=0.1483, att_loss=0.2733, loss=0.2483, over 16777.00 frames. utt_duration=1400 frames, utt_pad_proportion=0.005845, over 48.00 utterances.], tot_loss[ctc_loss=0.1157, att_loss=0.2582, loss=0.2297, over 1305744.33 frames. utt_duration=1175 frames, utt_pad_proportion=0.06902, over 4450.33 utterances.], batch size: 48, lr: 1.12e-02, grad_scale: 16.0 2023-03-08 02:17:36,814 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36001.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 02:17:41,614 INFO [train2.py:809] (1/4) Epoch 10, batch 150, loss[ctc_loss=0.08482, att_loss=0.2512, loss=0.2179, over 16327.00 frames. utt_duration=1453 frames, utt_pad_proportion=0.006369, over 45.00 utterances.], tot_loss[ctc_loss=0.1152, att_loss=0.2586, loss=0.2299, over 1743707.10 frames. utt_duration=1161 frames, utt_pad_proportion=0.0747, over 6016.23 utterances.], batch size: 45, lr: 1.12e-02, grad_scale: 16.0 2023-03-08 02:18:03,935 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.622e+02 2.601e+02 3.072e+02 3.945e+02 7.612e+02, threshold=6.144e+02, percent-clipped=5.0 2023-03-08 02:18:22,432 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36030.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:18:52,121 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=36049.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 02:19:00,382 INFO [train2.py:809] (1/4) Epoch 10, batch 200, loss[ctc_loss=0.1181, att_loss=0.2589, loss=0.2308, over 16938.00 frames. utt_duration=1357 frames, utt_pad_proportion=0.009116, over 50.00 utterances.], tot_loss[ctc_loss=0.1166, att_loss=0.2582, loss=0.2299, over 2080832.60 frames. utt_duration=1149 frames, utt_pad_proportion=0.07701, over 7253.46 utterances.], batch size: 50, lr: 1.12e-02, grad_scale: 16.0 2023-03-08 02:19:38,217 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=36078.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:19:43,047 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1153, 5.1833, 5.0457, 2.8097, 4.8696, 4.6829, 4.5090, 2.4551], device='cuda:1'), covar=tensor([0.0168, 0.0078, 0.0167, 0.1059, 0.0090, 0.0162, 0.0253, 0.1483], device='cuda:1'), in_proj_covar=tensor([0.0056, 0.0075, 0.0065, 0.0096, 0.0062, 0.0087, 0.0085, 0.0095], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-08 02:20:11,497 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.8345, 2.5122, 4.0804, 3.6647, 2.9917, 3.8274, 3.6770, 3.8635], device='cuda:1'), covar=tensor([0.0168, 0.1217, 0.0087, 0.0628, 0.1373, 0.0186, 0.0137, 0.0182], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0243, 0.0130, 0.0302, 0.0278, 0.0183, 0.0111, 0.0145], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-03-08 02:20:19,275 INFO [train2.py:809] (1/4) Epoch 10, batch 250, loss[ctc_loss=0.08505, att_loss=0.2259, loss=0.1977, over 14540.00 frames. utt_duration=1819 frames, utt_pad_proportion=0.03205, over 32.00 utterances.], tot_loss[ctc_loss=0.116, att_loss=0.2571, loss=0.2289, over 2338438.01 frames. utt_duration=1156 frames, utt_pad_proportion=0.07744, over 8101.82 utterances.], batch size: 32, lr: 1.11e-02, grad_scale: 16.0 2023-03-08 02:20:42,219 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.582e+02 2.693e+02 3.162e+02 4.102e+02 9.052e+02, threshold=6.324e+02, percent-clipped=7.0 2023-03-08 02:20:56,469 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36127.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:20:59,642 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7170, 5.0532, 4.9925, 4.9512, 5.0407, 5.0937, 4.7537, 4.4904], device='cuda:1'), covar=tensor([0.0980, 0.0481, 0.0225, 0.0411, 0.0262, 0.0261, 0.0313, 0.0336], device='cuda:1'), in_proj_covar=tensor([0.0429, 0.0275, 0.0222, 0.0261, 0.0325, 0.0342, 0.0269, 0.0299], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 02:21:39,367 INFO [train2.py:809] (1/4) Epoch 10, batch 300, loss[ctc_loss=0.09359, att_loss=0.2422, loss=0.2124, over 16114.00 frames. utt_duration=1536 frames, utt_pad_proportion=0.006957, over 42.00 utterances.], tot_loss[ctc_loss=0.1151, att_loss=0.2569, loss=0.2285, over 2545390.34 frames. utt_duration=1170 frames, utt_pad_proportion=0.07525, over 8712.66 utterances.], batch size: 42, lr: 1.11e-02, grad_scale: 16.0 2023-03-08 02:21:53,800 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0519, 6.2306, 5.6839, 5.9573, 5.9133, 5.5182, 5.7133, 5.4697], device='cuda:1'), covar=tensor([0.0970, 0.0783, 0.0628, 0.0788, 0.0719, 0.1220, 0.2040, 0.2379], device='cuda:1'), in_proj_covar=tensor([0.0418, 0.0489, 0.0363, 0.0378, 0.0355, 0.0413, 0.0499, 0.0448], 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') 2023-03-08 02:22:01,541 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36167.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:22:23,336 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36181.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:22:31,042 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36186.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:23:01,511 INFO [train2.py:809] (1/4) Epoch 10, batch 350, loss[ctc_loss=0.106, att_loss=0.2385, loss=0.212, over 15873.00 frames. utt_duration=1629 frames, utt_pad_proportion=0.009555, over 39.00 utterances.], tot_loss[ctc_loss=0.1141, att_loss=0.2564, loss=0.2279, over 2709187.14 frames. utt_duration=1193 frames, utt_pad_proportion=0.06723, over 9091.31 utterances.], batch size: 39, lr: 1.11e-02, grad_scale: 16.0 2023-03-08 02:23:24,551 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.349e+02 2.621e+02 3.186e+02 4.132e+02 7.508e+02, threshold=6.371e+02, percent-clipped=3.0 2023-03-08 02:23:28,135 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5966, 2.9776, 3.6218, 2.8581, 3.4667, 4.6789, 4.3796, 3.4243], device='cuda:1'), covar=tensor([0.0347, 0.1574, 0.1068, 0.1422, 0.1133, 0.0634, 0.0546, 0.1209], device='cuda:1'), in_proj_covar=tensor([0.0218, 0.0217, 0.0232, 0.0198, 0.0228, 0.0277, 0.0206, 0.0209], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:1') 2023-03-08 02:23:50,173 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=36234.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:24:03,305 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36242.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:24:17,260 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1603, 4.6044, 4.5595, 4.6976, 4.7930, 4.4352, 2.9655, 4.4934], device='cuda:1'), covar=tensor([0.0147, 0.0164, 0.0124, 0.0115, 0.0101, 0.0127, 0.0916, 0.0321], device='cuda:1'), in_proj_covar=tensor([0.0069, 0.0068, 0.0079, 0.0050, 0.0054, 0.0064, 0.0086, 0.0087], 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') 2023-03-08 02:24:24,488 INFO [train2.py:809] (1/4) Epoch 10, batch 400, loss[ctc_loss=0.1024, att_loss=0.2325, loss=0.2065, over 15662.00 frames. utt_duration=1695 frames, utt_pad_proportion=0.007833, over 37.00 utterances.], tot_loss[ctc_loss=0.1129, att_loss=0.2557, loss=0.2272, over 2835363.42 frames. utt_duration=1233 frames, utt_pad_proportion=0.05733, over 9212.17 utterances.], batch size: 37, lr: 1.11e-02, grad_scale: 16.0 2023-03-08 02:24:31,346 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3160, 2.7810, 3.5455, 2.7024, 3.3537, 4.5045, 4.2797, 3.0180], device='cuda:1'), covar=tensor([0.0441, 0.1677, 0.1069, 0.1514, 0.1082, 0.0738, 0.0535, 0.1499], device='cuda:1'), in_proj_covar=tensor([0.0219, 0.0215, 0.0232, 0.0197, 0.0227, 0.0276, 0.0206, 0.0209], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:1') 2023-03-08 02:25:35,567 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5908, 2.5735, 5.0028, 3.9769, 2.9873, 4.5227, 4.9489, 4.6374], device='cuda:1'), covar=tensor([0.0232, 0.1546, 0.0217, 0.0938, 0.1881, 0.0187, 0.0101, 0.0243], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0245, 0.0131, 0.0303, 0.0280, 0.0184, 0.0111, 0.0146], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0003, 0.0003, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-03-08 02:25:45,286 INFO [train2.py:809] (1/4) Epoch 10, batch 450, loss[ctc_loss=0.108, att_loss=0.2478, loss=0.2199, over 16403.00 frames. utt_duration=1493 frames, utt_pad_proportion=0.00751, over 44.00 utterances.], tot_loss[ctc_loss=0.1126, att_loss=0.2553, loss=0.2267, over 2931129.16 frames. utt_duration=1227 frames, utt_pad_proportion=0.05909, over 9569.10 utterances.], batch size: 44, lr: 1.11e-02, grad_scale: 16.0 2023-03-08 02:26:06,977 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.425e+02 2.409e+02 2.979e+02 3.960e+02 1.144e+03, threshold=5.958e+02, percent-clipped=2.0 2023-03-08 02:26:46,016 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.0423, 2.3904, 2.4542, 0.9391, 2.9258, 2.5063, 1.7452, 3.0126], device='cuda:1'), covar=tensor([0.0976, 0.2641, 0.3283, 0.3565, 0.1410, 0.1171, 0.2877, 0.1535], device='cuda:1'), in_proj_covar=tensor([0.0074, 0.0082, 0.0090, 0.0076, 0.0073, 0.0068, 0.0082, 0.0063], device='cuda:1'), out_proj_covar=tensor([4.5232e-05, 5.4681e-05, 5.7971e-05, 4.8335e-05, 4.4467e-05, 4.6345e-05, 5.4129e-05, 4.3871e-05], device='cuda:1') 2023-03-08 02:27:03,698 INFO [train2.py:809] (1/4) Epoch 10, batch 500, loss[ctc_loss=0.09176, att_loss=0.2475, loss=0.2163, over 16008.00 frames. utt_duration=1602 frames, utt_pad_proportion=0.007141, over 40.00 utterances.], tot_loss[ctc_loss=0.1114, att_loss=0.2544, loss=0.2258, over 3011646.34 frames. utt_duration=1258 frames, utt_pad_proportion=0.05051, over 9590.08 utterances.], batch size: 40, lr: 1.11e-02, grad_scale: 16.0 2023-03-08 02:27:14,107 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0435, 6.2555, 5.7061, 6.0480, 5.9598, 5.4843, 5.7542, 5.4207], device='cuda:1'), covar=tensor([0.1198, 0.0821, 0.0654, 0.0758, 0.0646, 0.1366, 0.1766, 0.2196], device='cuda:1'), in_proj_covar=tensor([0.0420, 0.0481, 0.0361, 0.0376, 0.0353, 0.0409, 0.0497, 0.0445], 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') 2023-03-08 02:27:51,241 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-03-08 02:28:22,569 INFO [train2.py:809] (1/4) Epoch 10, batch 550, loss[ctc_loss=0.1097, att_loss=0.2561, loss=0.2268, over 16333.00 frames. utt_duration=1453 frames, utt_pad_proportion=0.006202, over 45.00 utterances.], tot_loss[ctc_loss=0.1123, att_loss=0.2543, loss=0.2259, over 3061450.75 frames. utt_duration=1246 frames, utt_pad_proportion=0.05643, over 9840.24 utterances.], batch size: 45, lr: 1.11e-02, grad_scale: 16.0 2023-03-08 02:28:45,406 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.595e+02 2.390e+02 3.015e+02 3.600e+02 8.483e+02, threshold=6.031e+02, percent-clipped=3.0 2023-03-08 02:28:59,589 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36427.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:29:41,951 INFO [train2.py:809] (1/4) Epoch 10, batch 600, loss[ctc_loss=0.07728, att_loss=0.2228, loss=0.1937, over 13669.00 frames. utt_duration=1824 frames, utt_pad_proportion=0.07787, over 30.00 utterances.], tot_loss[ctc_loss=0.1107, att_loss=0.2539, loss=0.2253, over 3112959.74 frames. utt_duration=1255 frames, utt_pad_proportion=0.05307, over 9932.46 utterances.], batch size: 30, lr: 1.11e-02, grad_scale: 16.0 2023-03-08 02:30:03,570 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36467.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:30:15,894 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=36475.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:30:33,146 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.8136, 6.0478, 5.4967, 5.8023, 5.6547, 5.3487, 5.4890, 5.2379], device='cuda:1'), covar=tensor([0.1007, 0.0870, 0.0767, 0.0830, 0.0822, 0.1507, 0.2238, 0.2427], device='cuda:1'), in_proj_covar=tensor([0.0421, 0.0487, 0.0364, 0.0381, 0.0350, 0.0410, 0.0503, 0.0444], 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') 2023-03-08 02:30:49,747 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-03-08 02:31:02,501 INFO [train2.py:809] (1/4) Epoch 10, batch 650, loss[ctc_loss=0.1296, att_loss=0.2807, loss=0.2505, over 17042.00 frames. utt_duration=1312 frames, utt_pad_proportion=0.008802, over 52.00 utterances.], tot_loss[ctc_loss=0.1107, att_loss=0.2544, loss=0.2257, over 3155062.18 frames. utt_duration=1234 frames, utt_pad_proportion=0.05672, over 10237.67 utterances.], batch size: 52, lr: 1.11e-02, grad_scale: 16.0 2023-03-08 02:31:20,041 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=36515.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:31:24,437 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.680e+02 2.550e+02 3.093e+02 3.802e+02 8.175e+02, threshold=6.187e+02, percent-clipped=1.0 2023-03-08 02:31:53,763 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36537.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:32:21,151 INFO [train2.py:809] (1/4) Epoch 10, batch 700, loss[ctc_loss=0.1191, att_loss=0.2612, loss=0.2328, over 16870.00 frames. utt_duration=1379 frames, utt_pad_proportion=0.007493, over 49.00 utterances.], tot_loss[ctc_loss=0.1108, att_loss=0.2541, loss=0.2254, over 3171875.49 frames. utt_duration=1227 frames, utt_pad_proportion=0.06217, over 10350.08 utterances.], batch size: 49, lr: 1.11e-02, grad_scale: 16.0 2023-03-08 02:32:23,601 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.78 vs. limit=5.0 2023-03-08 02:33:39,961 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0371, 5.1460, 4.9938, 2.3242, 1.8826, 2.6098, 3.2982, 3.8295], device='cuda:1'), covar=tensor([0.0714, 0.0203, 0.0218, 0.4265, 0.6702, 0.3015, 0.1905, 0.1807], device='cuda:1'), in_proj_covar=tensor([0.0333, 0.0218, 0.0233, 0.0200, 0.0359, 0.0340, 0.0232, 0.0352], device='cuda:1'), out_proj_covar=tensor([1.5465e-04, 8.2807e-05, 1.0051e-04, 9.0701e-05, 1.5846e-04, 1.3999e-04, 9.2039e-05, 1.5180e-04], device='cuda:1') 2023-03-08 02:33:41,079 INFO [train2.py:809] (1/4) Epoch 10, batch 750, loss[ctc_loss=0.0961, att_loss=0.2608, loss=0.2278, over 16625.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.005379, over 47.00 utterances.], tot_loss[ctc_loss=0.1111, att_loss=0.2543, loss=0.2257, over 3189598.87 frames. utt_duration=1239 frames, utt_pad_proportion=0.0589, over 10309.23 utterances.], batch size: 47, lr: 1.11e-02, grad_scale: 16.0 2023-03-08 02:33:43,643 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36605.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:33:48,468 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36608.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:33:55,261 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36612.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:34:04,092 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.731e+02 2.573e+02 3.133e+02 3.667e+02 8.255e+02, threshold=6.267e+02, percent-clipped=5.0 2023-03-08 02:35:01,775 INFO [train2.py:809] (1/4) Epoch 10, batch 800, loss[ctc_loss=0.1591, att_loss=0.2837, loss=0.2588, over 17009.00 frames. utt_duration=1335 frames, utt_pad_proportion=0.009309, over 51.00 utterances.], tot_loss[ctc_loss=0.1099, att_loss=0.253, loss=0.2244, over 3208109.43 frames. utt_duration=1272 frames, utt_pad_proportion=0.05077, over 10098.80 utterances.], batch size: 51, lr: 1.11e-02, grad_scale: 16.0 2023-03-08 02:35:21,854 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36666.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:35:26,448 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36669.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:35:32,472 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36673.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:36:08,647 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.8388, 1.5714, 2.0731, 1.6859, 2.6913, 1.8076, 1.8865, 2.5407], device='cuda:1'), covar=tensor([0.0910, 0.3166, 0.2784, 0.2142, 0.1385, 0.1766, 0.2703, 0.1030], device='cuda:1'), in_proj_covar=tensor([0.0074, 0.0081, 0.0086, 0.0074, 0.0072, 0.0070, 0.0082, 0.0063], device='cuda:1'), out_proj_covar=tensor([4.5293e-05, 5.4232e-05, 5.6370e-05, 4.7700e-05, 4.4171e-05, 4.7227e-05, 5.3821e-05, 4.3802e-05], device='cuda:1') 2023-03-08 02:36:15,400 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36700.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 02:36:16,140 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.93 vs. limit=5.0 2023-03-08 02:36:21,096 INFO [train2.py:809] (1/4) Epoch 10, batch 850, loss[ctc_loss=0.1147, att_loss=0.2685, loss=0.2377, over 17064.00 frames. utt_duration=1314 frames, utt_pad_proportion=0.00764, over 52.00 utterances.], tot_loss[ctc_loss=0.1101, att_loss=0.2534, loss=0.2247, over 3229934.99 frames. utt_duration=1285 frames, utt_pad_proportion=0.04524, over 10067.58 utterances.], batch size: 52, lr: 1.11e-02, grad_scale: 16.0 2023-03-08 02:36:33,621 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3382, 4.6741, 4.6381, 5.0852, 2.7844, 4.9129, 2.8833, 1.9014], device='cuda:1'), covar=tensor([0.0208, 0.0125, 0.0545, 0.0094, 0.1736, 0.0091, 0.1533, 0.1933], device='cuda:1'), in_proj_covar=tensor([0.0130, 0.0107, 0.0251, 0.0107, 0.0220, 0.0102, 0.0226, 0.0205], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 02:36:43,992 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.977e+02 2.594e+02 3.302e+02 4.019e+02 1.127e+03, threshold=6.604e+02, percent-clipped=9.0 2023-03-08 02:37:41,456 INFO [train2.py:809] (1/4) Epoch 10, batch 900, loss[ctc_loss=0.1053, att_loss=0.2322, loss=0.2068, over 15528.00 frames. utt_duration=1727 frames, utt_pad_proportion=0.006964, over 36.00 utterances.], tot_loss[ctc_loss=0.1121, att_loss=0.254, loss=0.2256, over 3238888.20 frames. utt_duration=1252 frames, utt_pad_proportion=0.0532, over 10361.16 utterances.], batch size: 36, lr: 1.11e-02, grad_scale: 16.0 2023-03-08 02:37:53,202 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36761.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 02:38:56,829 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2274, 4.6262, 4.6218, 4.5818, 5.1584, 4.6422, 4.6893, 2.0677], device='cuda:1'), covar=tensor([0.0278, 0.0292, 0.0224, 0.0210, 0.0736, 0.0182, 0.0195, 0.2327], device='cuda:1'), in_proj_covar=tensor([0.0126, 0.0125, 0.0129, 0.0132, 0.0318, 0.0121, 0.0114, 0.0219], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 02:39:01,055 INFO [train2.py:809] (1/4) Epoch 10, batch 950, loss[ctc_loss=0.1015, att_loss=0.2581, loss=0.2268, over 17016.00 frames. utt_duration=1336 frames, utt_pad_proportion=0.008123, over 51.00 utterances.], tot_loss[ctc_loss=0.1119, att_loss=0.2545, loss=0.226, over 3233814.32 frames. utt_duration=1243 frames, utt_pad_proportion=0.05701, over 10416.06 utterances.], batch size: 51, lr: 1.10e-02, grad_scale: 16.0 2023-03-08 02:39:03,640 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36805.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 02:39:23,925 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.872e+02 2.661e+02 3.552e+02 4.499e+02 1.125e+03, threshold=7.103e+02, percent-clipped=5.0 2023-03-08 02:39:54,060 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36837.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:40:22,145 INFO [train2.py:809] (1/4) Epoch 10, batch 1000, loss[ctc_loss=0.1514, att_loss=0.2508, loss=0.2309, over 11396.00 frames. utt_duration=1825 frames, utt_pad_proportion=0.1883, over 25.00 utterances.], tot_loss[ctc_loss=0.1119, att_loss=0.2543, loss=0.2258, over 3238345.90 frames. utt_duration=1234 frames, utt_pad_proportion=0.06006, over 10509.67 utterances.], batch size: 25, lr: 1.10e-02, grad_scale: 16.0 2023-03-08 02:40:35,882 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1861, 4.6060, 4.4692, 4.6933, 2.5211, 4.4553, 2.6404, 1.8164], device='cuda:1'), covar=tensor([0.0264, 0.0132, 0.0615, 0.0158, 0.1880, 0.0144, 0.1561, 0.1791], device='cuda:1'), in_proj_covar=tensor([0.0130, 0.0106, 0.0247, 0.0108, 0.0216, 0.0101, 0.0224, 0.0203], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 02:40:41,840 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36866.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 02:41:11,558 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=36885.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:41:43,593 INFO [train2.py:809] (1/4) Epoch 10, batch 1050, loss[ctc_loss=0.1033, att_loss=0.2365, loss=0.2098, over 15944.00 frames. utt_duration=1557 frames, utt_pad_proportion=0.007046, over 41.00 utterances.], tot_loss[ctc_loss=0.1113, att_loss=0.2536, loss=0.2252, over 3242234.73 frames. utt_duration=1232 frames, utt_pad_proportion=0.062, over 10542.03 utterances.], batch size: 41, lr: 1.10e-02, grad_scale: 16.0 2023-03-08 02:41:55,879 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36911.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:42:06,546 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.560e+02 2.544e+02 2.981e+02 3.460e+02 1.238e+03, threshold=5.962e+02, percent-clipped=1.0 2023-03-08 02:43:05,657 INFO [train2.py:809] (1/4) Epoch 10, batch 1100, loss[ctc_loss=0.1126, att_loss=0.2602, loss=0.2307, over 16842.00 frames. utt_duration=1377 frames, utt_pad_proportion=0.008904, over 49.00 utterances.], tot_loss[ctc_loss=0.1119, att_loss=0.2544, loss=0.2259, over 3254898.78 frames. utt_duration=1199 frames, utt_pad_proportion=0.06683, over 10876.29 utterances.], batch size: 49, lr: 1.10e-02, grad_scale: 16.0 2023-03-08 02:43:08,402 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.95 vs. limit=5.0 2023-03-08 02:43:12,957 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36958.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:43:17,529 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36961.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:43:22,072 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36964.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:43:28,263 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36968.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:43:34,572 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36972.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:44:26,438 INFO [train2.py:809] (1/4) Epoch 10, batch 1150, loss[ctc_loss=0.124, att_loss=0.2616, loss=0.2341, over 17069.00 frames. utt_duration=1289 frames, utt_pad_proportion=0.008847, over 53.00 utterances.], tot_loss[ctc_loss=0.1116, att_loss=0.2547, loss=0.2261, over 3267289.74 frames. utt_duration=1213 frames, utt_pad_proportion=0.06102, over 10784.69 utterances.], batch size: 53, lr: 1.10e-02, grad_scale: 16.0 2023-03-08 02:44:26,797 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37004.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:44:48,824 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.581e+02 2.521e+02 2.891e+02 3.394e+02 5.463e+02, threshold=5.783e+02, percent-clipped=0.0 2023-03-08 02:44:50,904 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37019.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:45:15,780 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6332, 5.2502, 4.8713, 4.9617, 5.0793, 4.7431, 3.7839, 5.0805], device='cuda:1'), covar=tensor([0.0099, 0.0105, 0.0108, 0.0092, 0.0112, 0.0109, 0.0544, 0.0181], device='cuda:1'), in_proj_covar=tensor([0.0069, 0.0068, 0.0081, 0.0050, 0.0055, 0.0064, 0.0087, 0.0086], 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') 2023-03-08 02:45:46,919 INFO [train2.py:809] (1/4) Epoch 10, batch 1200, loss[ctc_loss=0.1129, att_loss=0.2692, loss=0.238, over 17127.00 frames. utt_duration=1225 frames, utt_pad_proportion=0.01455, over 56.00 utterances.], tot_loss[ctc_loss=0.1117, att_loss=0.2546, loss=0.226, over 3274669.26 frames. utt_duration=1213 frames, utt_pad_proportion=0.05995, over 10808.22 utterances.], batch size: 56, lr: 1.10e-02, grad_scale: 16.0 2023-03-08 02:45:50,268 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37056.0, num_to_drop=1, layers_to_drop={3} 2023-03-08 02:46:04,756 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37065.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:46:41,958 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.2754, 5.4684, 5.8826, 5.7723, 5.7067, 6.2306, 5.3036, 6.3229], device='cuda:1'), covar=tensor([0.0492, 0.0625, 0.0666, 0.0693, 0.1434, 0.0781, 0.0575, 0.0567], device='cuda:1'), in_proj_covar=tensor([0.0665, 0.0408, 0.0469, 0.0526, 0.0704, 0.0475, 0.0383, 0.0463], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 02:47:03,999 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.9301, 3.8548, 3.1979, 3.4619, 3.9812, 3.7603, 2.7271, 4.3968], device='cuda:1'), covar=tensor([0.0973, 0.0398, 0.1008, 0.0627, 0.0555, 0.0554, 0.0908, 0.0397], device='cuda:1'), in_proj_covar=tensor([0.0178, 0.0173, 0.0198, 0.0166, 0.0220, 0.0203, 0.0176, 0.0239], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-08 02:47:06,733 INFO [train2.py:809] (1/4) Epoch 10, batch 1250, loss[ctc_loss=0.1062, att_loss=0.2569, loss=0.2268, over 17375.00 frames. utt_duration=1105 frames, utt_pad_proportion=0.03445, over 63.00 utterances.], tot_loss[ctc_loss=0.1117, att_loss=0.2544, loss=0.2259, over 3269603.55 frames. utt_duration=1229 frames, utt_pad_proportion=0.05755, over 10655.31 utterances.], batch size: 63, lr: 1.10e-02, grad_scale: 16.0 2023-03-08 02:47:29,342 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.464e+02 2.541e+02 3.066e+02 3.675e+02 8.173e+02, threshold=6.132e+02, percent-clipped=8.0 2023-03-08 02:48:28,601 INFO [train2.py:809] (1/4) Epoch 10, batch 1300, loss[ctc_loss=0.1143, att_loss=0.2531, loss=0.2253, over 16298.00 frames. utt_duration=1518 frames, utt_pad_proportion=0.006183, over 43.00 utterances.], tot_loss[ctc_loss=0.1113, att_loss=0.2544, loss=0.2258, over 3274396.63 frames. utt_duration=1233 frames, utt_pad_proportion=0.05638, over 10637.83 utterances.], batch size: 43, lr: 1.10e-02, grad_scale: 16.0 2023-03-08 02:48:40,427 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37161.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 02:49:49,919 INFO [train2.py:809] (1/4) Epoch 10, batch 1350, loss[ctc_loss=0.103, att_loss=0.2358, loss=0.2092, over 15672.00 frames. utt_duration=1696 frames, utt_pad_proportion=0.006618, over 37.00 utterances.], tot_loss[ctc_loss=0.1108, att_loss=0.2538, loss=0.2252, over 3270604.72 frames. utt_duration=1241 frames, utt_pad_proportion=0.05614, over 10554.11 utterances.], batch size: 37, lr: 1.10e-02, grad_scale: 16.0 2023-03-08 02:50:12,585 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.658e+02 2.454e+02 3.108e+02 4.121e+02 7.457e+02, threshold=6.215e+02, percent-clipped=5.0 2023-03-08 02:50:26,858 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.8540, 4.0015, 4.1278, 4.0870, 4.1744, 4.1717, 3.9498, 3.8405], device='cuda:1'), covar=tensor([0.1100, 0.0813, 0.0276, 0.0480, 0.0366, 0.0363, 0.0363, 0.0361], device='cuda:1'), in_proj_covar=tensor([0.0444, 0.0286, 0.0231, 0.0267, 0.0334, 0.0357, 0.0281, 0.0310], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 02:51:05,369 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9916, 5.3234, 5.3331, 5.2356, 5.4661, 5.3693, 5.1344, 4.8794], device='cuda:1'), covar=tensor([0.1070, 0.0520, 0.0219, 0.0436, 0.0271, 0.0322, 0.0297, 0.0300], device='cuda:1'), in_proj_covar=tensor([0.0445, 0.0288, 0.0232, 0.0269, 0.0335, 0.0359, 0.0283, 0.0311], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0004, 0.0004, 0.0005, 0.0003, 0.0004], device='cuda:1') 2023-03-08 02:51:11,266 INFO [train2.py:809] (1/4) Epoch 10, batch 1400, loss[ctc_loss=0.07072, att_loss=0.2123, loss=0.184, over 14530.00 frames. utt_duration=1818 frames, utt_pad_proportion=0.03417, over 32.00 utterances.], tot_loss[ctc_loss=0.1097, att_loss=0.2531, loss=0.2245, over 3258361.61 frames. utt_duration=1257 frames, utt_pad_proportion=0.05345, over 10377.71 utterances.], batch size: 32, lr: 1.10e-02, grad_scale: 16.0 2023-03-08 02:51:20,930 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7440, 4.4652, 4.3990, 4.6030, 5.1043, 4.7531, 4.5901, 2.1624], device='cuda:1'), covar=tensor([0.0139, 0.0336, 0.0266, 0.0236, 0.0796, 0.0124, 0.0224, 0.2173], device='cuda:1'), in_proj_covar=tensor([0.0125, 0.0123, 0.0130, 0.0133, 0.0319, 0.0121, 0.0112, 0.0216], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 02:51:23,151 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37261.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:51:27,867 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37264.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:51:32,341 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37267.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:51:33,965 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37268.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:51:34,028 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37268.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:52:31,799 INFO [train2.py:809] (1/4) Epoch 10, batch 1450, loss[ctc_loss=0.1184, att_loss=0.2594, loss=0.2312, over 17366.00 frames. utt_duration=1104 frames, utt_pad_proportion=0.03383, over 63.00 utterances.], tot_loss[ctc_loss=0.1106, att_loss=0.2541, loss=0.2254, over 3264728.41 frames. utt_duration=1258 frames, utt_pad_proportion=0.05304, over 10389.43 utterances.], batch size: 63, lr: 1.10e-02, grad_scale: 16.0 2023-03-08 02:52:39,564 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=37309.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:52:44,936 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=37312.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:52:48,262 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37314.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:52:51,388 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=37316.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:52:54,387 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.647e+02 2.516e+02 3.005e+02 3.782e+02 1.128e+03, threshold=6.010e+02, percent-clipped=1.0 2023-03-08 02:53:11,882 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37329.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:53:51,707 INFO [train2.py:809] (1/4) Epoch 10, batch 1500, loss[ctc_loss=0.1104, att_loss=0.2732, loss=0.2406, over 17061.00 frames. utt_duration=1314 frames, utt_pad_proportion=0.00836, over 52.00 utterances.], tot_loss[ctc_loss=0.1121, att_loss=0.2557, loss=0.227, over 3279575.60 frames. utt_duration=1267 frames, utt_pad_proportion=0.04693, over 10367.33 utterances.], batch size: 52, lr: 1.10e-02, grad_scale: 16.0 2023-03-08 02:53:55,188 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37356.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 02:54:02,000 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37360.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:54:08,239 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37364.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:54:32,339 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-03-08 02:55:12,297 INFO [train2.py:809] (1/4) Epoch 10, batch 1550, loss[ctc_loss=0.1165, att_loss=0.2283, loss=0.2059, over 14027.00 frames. utt_duration=1812 frames, utt_pad_proportion=0.05598, over 31.00 utterances.], tot_loss[ctc_loss=0.1119, att_loss=0.2554, loss=0.2267, over 3279473.51 frames. utt_duration=1255 frames, utt_pad_proportion=0.04976, over 10463.53 utterances.], batch size: 31, lr: 1.10e-02, grad_scale: 16.0 2023-03-08 02:55:12,400 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=37404.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 02:55:34,598 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.660e+02 2.658e+02 3.084e+02 3.993e+02 7.629e+02, threshold=6.167e+02, percent-clipped=4.0 2023-03-08 02:55:45,945 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37425.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 02:56:32,396 INFO [train2.py:809] (1/4) Epoch 10, batch 1600, loss[ctc_loss=0.1463, att_loss=0.273, loss=0.2476, over 17090.00 frames. utt_duration=691.9 frames, utt_pad_proportion=0.1285, over 99.00 utterances.], tot_loss[ctc_loss=0.1126, att_loss=0.2564, loss=0.2276, over 3286689.16 frames. utt_duration=1222 frames, utt_pad_proportion=0.05639, over 10773.34 utterances.], batch size: 99, lr: 1.09e-02, grad_scale: 16.0 2023-03-08 02:56:43,938 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37461.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 02:57:53,647 INFO [train2.py:809] (1/4) Epoch 10, batch 1650, loss[ctc_loss=0.1147, att_loss=0.2636, loss=0.2338, over 16640.00 frames. utt_duration=1417 frames, utt_pad_proportion=0.004572, over 47.00 utterances.], tot_loss[ctc_loss=0.1116, att_loss=0.256, loss=0.2271, over 3286144.77 frames. utt_duration=1224 frames, utt_pad_proportion=0.05643, over 10754.52 utterances.], batch size: 47, lr: 1.09e-02, grad_scale: 16.0 2023-03-08 02:58:01,292 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=37509.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 02:58:14,778 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.538e+02 2.486e+02 2.891e+02 3.571e+02 1.016e+03, threshold=5.782e+02, percent-clipped=2.0 2023-03-08 02:59:12,748 INFO [train2.py:809] (1/4) Epoch 10, batch 1700, loss[ctc_loss=0.1026, att_loss=0.268, loss=0.2349, over 16461.00 frames. utt_duration=1433 frames, utt_pad_proportion=0.007005, over 46.00 utterances.], tot_loss[ctc_loss=0.111, att_loss=0.2557, loss=0.2267, over 3284988.73 frames. utt_duration=1233 frames, utt_pad_proportion=0.05426, over 10670.19 utterances.], batch size: 46, lr: 1.09e-02, grad_scale: 16.0 2023-03-08 02:59:33,293 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37567.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:00:33,882 INFO [train2.py:809] (1/4) Epoch 10, batch 1750, loss[ctc_loss=0.1008, att_loss=0.2512, loss=0.2211, over 16537.00 frames. utt_duration=1472 frames, utt_pad_proportion=0.005736, over 45.00 utterances.], tot_loss[ctc_loss=0.1104, att_loss=0.2554, loss=0.2264, over 3285417.00 frames. utt_duration=1248 frames, utt_pad_proportion=0.05085, over 10538.78 utterances.], batch size: 45, lr: 1.09e-02, grad_scale: 16.0 2023-03-08 03:00:49,709 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37614.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:00:51,061 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=37615.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:00:55,646 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.443e+02 2.432e+02 2.940e+02 3.870e+02 6.231e+02, threshold=5.879e+02, percent-clipped=2.0 2023-03-08 03:01:05,270 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37624.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:01:53,792 INFO [train2.py:809] (1/4) Epoch 10, batch 1800, loss[ctc_loss=0.1134, att_loss=0.2665, loss=0.2359, over 16873.00 frames. utt_duration=1379 frames, utt_pad_proportion=0.008046, over 49.00 utterances.], tot_loss[ctc_loss=0.1095, att_loss=0.2544, loss=0.2254, over 3278492.24 frames. utt_duration=1258 frames, utt_pad_proportion=0.05118, over 10433.88 utterances.], batch size: 49, lr: 1.09e-02, grad_scale: 32.0 2023-03-08 03:02:03,003 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37660.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:02:05,720 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=37662.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:03:13,562 INFO [train2.py:809] (1/4) Epoch 10, batch 1850, loss[ctc_loss=0.09773, att_loss=0.2443, loss=0.215, over 16328.00 frames. utt_duration=1453 frames, utt_pad_proportion=0.006202, over 45.00 utterances.], tot_loss[ctc_loss=0.1097, att_loss=0.2547, loss=0.2257, over 3281788.69 frames. utt_duration=1243 frames, utt_pad_proportion=0.05362, over 10570.90 utterances.], batch size: 45, lr: 1.09e-02, grad_scale: 16.0 2023-03-08 03:03:19,696 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=37708.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:03:28,038 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.92 vs. limit=2.0 2023-03-08 03:03:36,152 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.525e+02 2.414e+02 3.142e+02 4.365e+02 1.152e+03, threshold=6.284e+02, percent-clipped=6.0 2023-03-08 03:03:37,924 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37720.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:04:21,669 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.3639, 5.2423, 5.1534, 2.7640, 5.0189, 4.7629, 4.7091, 2.5546], device='cuda:1'), covar=tensor([0.0091, 0.0072, 0.0170, 0.1029, 0.0087, 0.0163, 0.0235, 0.1377], device='cuda:1'), in_proj_covar=tensor([0.0058, 0.0079, 0.0068, 0.0100, 0.0067, 0.0089, 0.0088, 0.0098], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-08 03:04:33,114 INFO [train2.py:809] (1/4) Epoch 10, batch 1900, loss[ctc_loss=0.1034, att_loss=0.2332, loss=0.2072, over 12297.00 frames. utt_duration=1823 frames, utt_pad_proportion=0.0323, over 27.00 utterances.], tot_loss[ctc_loss=0.1092, att_loss=0.254, loss=0.225, over 3272487.72 frames. utt_duration=1256 frames, utt_pad_proportion=0.05155, over 10433.45 utterances.], batch size: 27, lr: 1.09e-02, grad_scale: 16.0 2023-03-08 03:04:36,510 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0352, 2.4492, 3.3370, 2.6411, 3.2727, 4.1878, 4.0677, 3.0454], device='cuda:1'), covar=tensor([0.0570, 0.2109, 0.1166, 0.1562, 0.1099, 0.0933, 0.0526, 0.1446], device='cuda:1'), in_proj_covar=tensor([0.0229, 0.0227, 0.0243, 0.0206, 0.0237, 0.0293, 0.0213, 0.0220], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 03:05:21,953 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6010, 4.7877, 4.7142, 4.6855, 5.1284, 4.6412, 4.6193, 2.3952], device='cuda:1'), covar=tensor([0.0257, 0.0267, 0.0199, 0.0205, 0.1166, 0.0235, 0.0236, 0.2249], device='cuda:1'), in_proj_covar=tensor([0.0124, 0.0123, 0.0129, 0.0132, 0.0318, 0.0120, 0.0111, 0.0216], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 03:05:53,428 INFO [train2.py:809] (1/4) Epoch 10, batch 1950, loss[ctc_loss=0.1291, att_loss=0.2708, loss=0.2425, over 17397.00 frames. utt_duration=1106 frames, utt_pad_proportion=0.0322, over 63.00 utterances.], tot_loss[ctc_loss=0.1101, att_loss=0.2547, loss=0.2258, over 3273172.95 frames. utt_duration=1216 frames, utt_pad_proportion=0.061, over 10783.11 utterances.], batch size: 63, lr: 1.09e-02, grad_scale: 16.0 2023-03-08 03:06:02,950 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1796, 5.2338, 5.0144, 2.6219, 2.0709, 2.9883, 3.6955, 3.9111], device='cuda:1'), covar=tensor([0.0627, 0.0300, 0.0286, 0.4331, 0.5865, 0.2491, 0.1446, 0.1876], device='cuda:1'), in_proj_covar=tensor([0.0327, 0.0214, 0.0230, 0.0195, 0.0346, 0.0328, 0.0223, 0.0346], device='cuda:1'), out_proj_covar=tensor([1.5013e-04, 8.1570e-05, 9.9626e-05, 8.8439e-05, 1.5299e-04, 1.3462e-04, 8.7793e-05, 1.4807e-04], device='cuda:1') 2023-03-08 03:06:16,462 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.575e+02 2.508e+02 3.059e+02 3.528e+02 6.330e+02, threshold=6.117e+02, percent-clipped=1.0 2023-03-08 03:07:12,865 INFO [train2.py:809] (1/4) Epoch 10, batch 2000, loss[ctc_loss=0.1061, att_loss=0.269, loss=0.2364, over 17077.00 frames. utt_duration=1290 frames, utt_pad_proportion=0.008252, over 53.00 utterances.], tot_loss[ctc_loss=0.1107, att_loss=0.2551, loss=0.2262, over 3271596.28 frames. utt_duration=1196 frames, utt_pad_proportion=0.06589, over 10958.26 utterances.], batch size: 53, lr: 1.09e-02, grad_scale: 16.0 2023-03-08 03:08:32,707 INFO [train2.py:809] (1/4) Epoch 10, batch 2050, loss[ctc_loss=0.09688, att_loss=0.2343, loss=0.2068, over 15875.00 frames. utt_duration=1630 frames, utt_pad_proportion=0.009861, over 39.00 utterances.], tot_loss[ctc_loss=0.1107, att_loss=0.2551, loss=0.2262, over 3270889.60 frames. utt_duration=1196 frames, utt_pad_proportion=0.06583, over 10952.32 utterances.], batch size: 39, lr: 1.09e-02, grad_scale: 16.0 2023-03-08 03:08:55,897 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.379e+02 2.394e+02 2.856e+02 3.386e+02 7.118e+02, threshold=5.711e+02, percent-clipped=3.0 2023-03-08 03:09:04,090 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37924.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:09:41,466 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9758, 6.1984, 5.5934, 6.0218, 5.8186, 5.5783, 5.6469, 5.4918], device='cuda:1'), covar=tensor([0.0970, 0.0747, 0.0725, 0.0608, 0.0729, 0.1000, 0.2155, 0.2077], device='cuda:1'), in_proj_covar=tensor([0.0416, 0.0482, 0.0365, 0.0373, 0.0346, 0.0405, 0.0491, 0.0449], 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') 2023-03-08 03:09:52,780 INFO [train2.py:809] (1/4) Epoch 10, batch 2100, loss[ctc_loss=0.08264, att_loss=0.2377, loss=0.2067, over 16536.00 frames. utt_duration=1471 frames, utt_pad_proportion=0.006497, over 45.00 utterances.], tot_loss[ctc_loss=0.1105, att_loss=0.2551, loss=0.2262, over 3275210.79 frames. utt_duration=1208 frames, utt_pad_proportion=0.06332, over 10862.82 utterances.], batch size: 45, lr: 1.09e-02, grad_scale: 8.0 2023-03-08 03:10:19,826 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.5043, 1.3272, 2.0828, 1.9108, 2.4542, 1.5017, 1.8635, 2.4073], device='cuda:1'), covar=tensor([0.1177, 0.4805, 0.2937, 0.1830, 0.1123, 0.2165, 0.2771, 0.1973], device='cuda:1'), in_proj_covar=tensor([0.0075, 0.0084, 0.0089, 0.0074, 0.0072, 0.0068, 0.0081, 0.0063], device='cuda:1'), out_proj_covar=tensor([4.6786e-05, 5.6174e-05, 5.8426e-05, 4.8419e-05, 4.5031e-05, 4.6747e-05, 5.4048e-05, 4.4347e-05], device='cuda:1') 2023-03-08 03:10:21,706 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=37972.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:10:33,447 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37979.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:10:38,939 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.89 vs. limit=5.0 2023-03-08 03:11:18,141 INFO [train2.py:809] (1/4) Epoch 10, batch 2150, loss[ctc_loss=0.1119, att_loss=0.2565, loss=0.2276, over 16381.00 frames. utt_duration=1491 frames, utt_pad_proportion=0.00746, over 44.00 utterances.], tot_loss[ctc_loss=0.111, att_loss=0.2558, loss=0.2268, over 3282778.09 frames. utt_duration=1212 frames, utt_pad_proportion=0.06069, over 10845.77 utterances.], batch size: 44, lr: 1.09e-02, grad_scale: 8.0 2023-03-08 03:11:42,430 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.632e+02 2.860e+02 3.437e+02 3.901e+02 7.219e+02, threshold=6.874e+02, percent-clipped=5.0 2023-03-08 03:11:42,755 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=38020.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:11:51,232 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1080, 4.5902, 4.2802, 4.5392, 4.5584, 4.3631, 3.1665, 4.5541], device='cuda:1'), covar=tensor([0.0122, 0.0122, 0.0157, 0.0086, 0.0105, 0.0109, 0.0691, 0.0190], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0070, 0.0082, 0.0051, 0.0055, 0.0066, 0.0088, 0.0087], 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') 2023-03-08 03:12:15,389 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=38040.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:12:28,321 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-03-08 03:12:28,956 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.2439, 5.3205, 5.0842, 3.0816, 5.1467, 4.9374, 4.4037, 2.5699], device='cuda:1'), covar=tensor([0.0132, 0.0058, 0.0215, 0.0908, 0.0073, 0.0124, 0.0293, 0.1478], device='cuda:1'), in_proj_covar=tensor([0.0059, 0.0079, 0.0069, 0.0100, 0.0067, 0.0090, 0.0089, 0.0099], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-08 03:12:37,087 INFO [train2.py:809] (1/4) Epoch 10, batch 2200, loss[ctc_loss=0.1099, att_loss=0.2646, loss=0.2337, over 17315.00 frames. utt_duration=1175 frames, utt_pad_proportion=0.02372, over 59.00 utterances.], tot_loss[ctc_loss=0.1114, att_loss=0.2554, loss=0.2266, over 3285199.12 frames. utt_duration=1223 frames, utt_pad_proportion=0.05733, over 10761.81 utterances.], batch size: 59, lr: 1.09e-02, grad_scale: 8.0 2023-03-08 03:12:58,883 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=38068.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:13:56,129 INFO [train2.py:809] (1/4) Epoch 10, batch 2250, loss[ctc_loss=0.1057, att_loss=0.2546, loss=0.2248, over 16391.00 frames. utt_duration=1492 frames, utt_pad_proportion=0.008281, over 44.00 utterances.], tot_loss[ctc_loss=0.1112, att_loss=0.2544, loss=0.2258, over 3274729.15 frames. utt_duration=1237 frames, utt_pad_proportion=0.05759, over 10600.91 utterances.], batch size: 44, lr: 1.09e-02, grad_scale: 8.0 2023-03-08 03:14:21,121 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.523e+02 2.484e+02 3.268e+02 3.999e+02 7.414e+02, threshold=6.535e+02, percent-clipped=1.0 2023-03-08 03:14:40,895 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2325, 4.6648, 4.6198, 4.3160, 2.5148, 4.9180, 2.3615, 1.6560], device='cuda:1'), covar=tensor([0.0211, 0.0168, 0.0554, 0.0191, 0.1692, 0.0094, 0.1553, 0.1665], device='cuda:1'), in_proj_covar=tensor([0.0127, 0.0104, 0.0249, 0.0107, 0.0213, 0.0099, 0.0219, 0.0199], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 03:15:14,535 INFO [train2.py:809] (1/4) Epoch 10, batch 2300, loss[ctc_loss=0.1209, att_loss=0.2663, loss=0.2372, over 17048.00 frames. utt_duration=1313 frames, utt_pad_proportion=0.008999, over 52.00 utterances.], tot_loss[ctc_loss=0.1106, att_loss=0.2538, loss=0.2252, over 3264858.69 frames. utt_duration=1247 frames, utt_pad_proportion=0.05758, over 10482.23 utterances.], batch size: 52, lr: 1.09e-02, grad_scale: 8.0 2023-03-08 03:16:33,797 INFO [train2.py:809] (1/4) Epoch 10, batch 2350, loss[ctc_loss=0.08908, att_loss=0.2558, loss=0.2225, over 17022.00 frames. utt_duration=1337 frames, utt_pad_proportion=0.007599, over 51.00 utterances.], tot_loss[ctc_loss=0.1094, att_loss=0.2531, loss=0.2243, over 3272359.99 frames. utt_duration=1273 frames, utt_pad_proportion=0.0498, over 10296.38 utterances.], batch size: 51, lr: 1.08e-02, grad_scale: 8.0 2023-03-08 03:16:59,114 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.810e+02 2.404e+02 3.254e+02 4.140e+02 7.234e+02, threshold=6.508e+02, percent-clipped=3.0 2023-03-08 03:17:02,710 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2069, 4.4876, 4.5893, 4.3758, 2.5252, 4.4649, 2.0489, 1.5307], device='cuda:1'), covar=tensor([0.0258, 0.0182, 0.0550, 0.0215, 0.1800, 0.0216, 0.1949, 0.1853], device='cuda:1'), in_proj_covar=tensor([0.0127, 0.0104, 0.0248, 0.0107, 0.0213, 0.0101, 0.0220, 0.0198], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 03:17:53,623 INFO [train2.py:809] (1/4) Epoch 10, batch 2400, loss[ctc_loss=0.09755, att_loss=0.2299, loss=0.2035, over 15654.00 frames. utt_duration=1694 frames, utt_pad_proportion=0.008482, over 37.00 utterances.], tot_loss[ctc_loss=0.1096, att_loss=0.253, loss=0.2244, over 3270774.28 frames. utt_duration=1254 frames, utt_pad_proportion=0.05471, over 10448.89 utterances.], batch size: 37, lr: 1.08e-02, grad_scale: 8.0 2023-03-08 03:19:14,659 INFO [train2.py:809] (1/4) Epoch 10, batch 2450, loss[ctc_loss=0.06934, att_loss=0.2135, loss=0.1847, over 12321.00 frames. utt_duration=1827 frames, utt_pad_proportion=0.1251, over 27.00 utterances.], tot_loss[ctc_loss=0.1086, att_loss=0.2532, loss=0.2243, over 3276828.58 frames. utt_duration=1275 frames, utt_pad_proportion=0.04788, over 10295.86 utterances.], batch size: 27, lr: 1.08e-02, grad_scale: 8.0 2023-03-08 03:19:40,304 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.402e+02 2.470e+02 3.021e+02 3.879e+02 8.501e+02, threshold=6.042e+02, percent-clipped=3.0 2023-03-08 03:20:05,101 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=38335.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:20:34,413 INFO [train2.py:809] (1/4) Epoch 10, batch 2500, loss[ctc_loss=0.09854, att_loss=0.2449, loss=0.2157, over 16393.00 frames. utt_duration=1492 frames, utt_pad_proportion=0.007949, over 44.00 utterances.], tot_loss[ctc_loss=0.11, att_loss=0.2545, loss=0.2256, over 3279345.47 frames. utt_duration=1217 frames, utt_pad_proportion=0.06109, over 10793.55 utterances.], batch size: 44, lr: 1.08e-02, grad_scale: 8.0 2023-03-08 03:20:36,777 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-03-08 03:20:43,857 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=38360.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:21:43,332 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-03-08 03:21:53,305 INFO [train2.py:809] (1/4) Epoch 10, batch 2550, loss[ctc_loss=0.1175, att_loss=0.2674, loss=0.2375, over 17292.00 frames. utt_duration=1174 frames, utt_pad_proportion=0.02483, over 59.00 utterances.], tot_loss[ctc_loss=0.1103, att_loss=0.2543, loss=0.2255, over 3273754.28 frames. utt_duration=1213 frames, utt_pad_proportion=0.06376, over 10804.50 utterances.], batch size: 59, lr: 1.08e-02, grad_scale: 8.0 2023-03-08 03:22:18,520 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.579e+02 2.719e+02 3.198e+02 3.977e+02 1.146e+03, threshold=6.397e+02, percent-clipped=5.0 2023-03-08 03:22:21,688 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=38421.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:23:12,912 INFO [train2.py:809] (1/4) Epoch 10, batch 2600, loss[ctc_loss=0.1143, att_loss=0.2694, loss=0.2384, over 17229.00 frames. utt_duration=873.7 frames, utt_pad_proportion=0.08508, over 79.00 utterances.], tot_loss[ctc_loss=0.1085, att_loss=0.2525, loss=0.2237, over 3266837.78 frames. utt_duration=1236 frames, utt_pad_proportion=0.05959, over 10581.95 utterances.], batch size: 79, lr: 1.08e-02, grad_scale: 8.0 2023-03-08 03:24:31,644 INFO [train2.py:809] (1/4) Epoch 10, batch 2650, loss[ctc_loss=0.07636, att_loss=0.2425, loss=0.2093, over 16256.00 frames. utt_duration=1514 frames, utt_pad_proportion=0.007925, over 43.00 utterances.], tot_loss[ctc_loss=0.1084, att_loss=0.2528, loss=0.2239, over 3274152.78 frames. utt_duration=1267 frames, utt_pad_proportion=0.05088, over 10346.90 utterances.], batch size: 43, lr: 1.08e-02, grad_scale: 8.0 2023-03-08 03:24:46,527 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0417, 4.4566, 4.3041, 4.3946, 4.9583, 4.3612, 4.5390, 2.1080], device='cuda:1'), covar=tensor([0.0330, 0.0373, 0.0331, 0.0277, 0.0967, 0.0243, 0.0280, 0.2476], device='cuda:1'), in_proj_covar=tensor([0.0124, 0.0127, 0.0132, 0.0136, 0.0327, 0.0120, 0.0117, 0.0220], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 03:24:58,438 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.587e+02 2.727e+02 3.206e+02 4.393e+02 9.975e+02, threshold=6.412e+02, percent-clipped=4.0 2023-03-08 03:25:51,260 INFO [train2.py:809] (1/4) Epoch 10, batch 2700, loss[ctc_loss=0.09103, att_loss=0.2249, loss=0.1981, over 14536.00 frames. utt_duration=1819 frames, utt_pad_proportion=0.03922, over 32.00 utterances.], tot_loss[ctc_loss=0.1086, att_loss=0.2525, loss=0.2238, over 3272082.42 frames. utt_duration=1260 frames, utt_pad_proportion=0.0528, over 10396.94 utterances.], batch size: 32, lr: 1.08e-02, grad_scale: 8.0 2023-03-08 03:26:53,180 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1317, 5.4009, 4.8925, 5.5020, 4.8876, 5.1837, 5.6198, 5.3928], device='cuda:1'), covar=tensor([0.0477, 0.0347, 0.0806, 0.0275, 0.0382, 0.0172, 0.0187, 0.0149], device='cuda:1'), in_proj_covar=tensor([0.0325, 0.0254, 0.0314, 0.0246, 0.0257, 0.0197, 0.0238, 0.0230], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0005, 0.0006, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 03:27:11,536 INFO [train2.py:809] (1/4) Epoch 10, batch 2750, loss[ctc_loss=0.08299, att_loss=0.2178, loss=0.1909, over 15607.00 frames. utt_duration=1689 frames, utt_pad_proportion=0.009432, over 37.00 utterances.], tot_loss[ctc_loss=0.1088, att_loss=0.2524, loss=0.2236, over 3273913.30 frames. utt_duration=1237 frames, utt_pad_proportion=0.05737, over 10595.67 utterances.], batch size: 37, lr: 1.08e-02, grad_scale: 8.0 2023-03-08 03:27:38,145 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.499e+02 2.648e+02 3.075e+02 3.913e+02 9.612e+02, threshold=6.151e+02, percent-clipped=3.0 2023-03-08 03:28:01,971 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=38635.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:28:31,008 INFO [train2.py:809] (1/4) Epoch 10, batch 2800, loss[ctc_loss=0.09026, att_loss=0.2368, loss=0.2075, over 16003.00 frames. utt_duration=1602 frames, utt_pad_proportion=0.007621, over 40.00 utterances.], tot_loss[ctc_loss=0.1103, att_loss=0.2531, loss=0.2246, over 3263945.24 frames. utt_duration=1186 frames, utt_pad_proportion=0.07348, over 11024.66 utterances.], batch size: 40, lr: 1.08e-02, grad_scale: 8.0 2023-03-08 03:29:18,057 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=38683.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:29:31,406 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.37 vs. limit=5.0 2023-03-08 03:29:51,020 INFO [train2.py:809] (1/4) Epoch 10, batch 2850, loss[ctc_loss=0.1087, att_loss=0.2425, loss=0.2157, over 15958.00 frames. utt_duration=1558 frames, utt_pad_proportion=0.006747, over 41.00 utterances.], tot_loss[ctc_loss=0.1097, att_loss=0.2529, loss=0.2243, over 3268280.46 frames. utt_duration=1207 frames, utt_pad_proportion=0.067, over 10844.81 utterances.], batch size: 41, lr: 1.08e-02, grad_scale: 8.0 2023-03-08 03:29:55,922 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-03-08 03:30:11,790 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=38716.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:30:17,873 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.492e+02 2.597e+02 3.110e+02 4.075e+02 8.061e+02, threshold=6.220e+02, percent-clipped=4.0 2023-03-08 03:31:10,905 INFO [train2.py:809] (1/4) Epoch 10, batch 2900, loss[ctc_loss=0.07383, att_loss=0.2258, loss=0.1954, over 16397.00 frames. utt_duration=1492 frames, utt_pad_proportion=0.007092, over 44.00 utterances.], tot_loss[ctc_loss=0.1096, att_loss=0.2533, loss=0.2246, over 3276676.08 frames. utt_duration=1209 frames, utt_pad_proportion=0.0642, over 10856.70 utterances.], batch size: 44, lr: 1.08e-02, grad_scale: 8.0 2023-03-08 03:31:52,376 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-03-08 03:32:25,520 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.15 vs. limit=5.0 2023-03-08 03:32:30,745 INFO [train2.py:809] (1/4) Epoch 10, batch 2950, loss[ctc_loss=0.09583, att_loss=0.259, loss=0.2264, over 16623.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.005409, over 47.00 utterances.], tot_loss[ctc_loss=0.1103, att_loss=0.2542, loss=0.2254, over 3282556.55 frames. utt_duration=1214 frames, utt_pad_proportion=0.06168, over 10827.01 utterances.], batch size: 47, lr: 1.08e-02, grad_scale: 8.0 2023-03-08 03:32:57,342 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.796e+02 2.729e+02 3.322e+02 4.124e+02 1.749e+03, threshold=6.643e+02, percent-clipped=6.0 2023-03-08 03:33:02,374 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=38823.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:33:29,149 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5056, 4.9290, 4.7985, 4.9172, 5.0332, 4.7004, 3.3110, 4.8623], device='cuda:1'), covar=tensor([0.0095, 0.0118, 0.0097, 0.0078, 0.0091, 0.0105, 0.0716, 0.0184], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0069, 0.0082, 0.0051, 0.0055, 0.0066, 0.0088, 0.0088], 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') 2023-03-08 03:33:50,299 INFO [train2.py:809] (1/4) Epoch 10, batch 3000, loss[ctc_loss=0.08222, att_loss=0.2198, loss=0.1923, over 15880.00 frames. utt_duration=1630 frames, utt_pad_proportion=0.009009, over 39.00 utterances.], tot_loss[ctc_loss=0.1092, att_loss=0.2532, loss=0.2244, over 3270654.65 frames. utt_duration=1213 frames, utt_pad_proportion=0.06622, over 10799.26 utterances.], batch size: 39, lr: 1.08e-02, grad_scale: 8.0 2023-03-08 03:33:50,300 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-08 03:34:06,507 INFO [train2.py:843] (1/4) Epoch 10, validation: ctc_loss=0.0536, att_loss=0.2406, loss=0.2032, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-08 03:34:06,507 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-08 03:34:22,869 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-03-08 03:34:54,845 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=38884.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:35:26,806 INFO [train2.py:809] (1/4) Epoch 10, batch 3050, loss[ctc_loss=0.09498, att_loss=0.2399, loss=0.2109, over 15875.00 frames. utt_duration=1630 frames, utt_pad_proportion=0.009923, over 39.00 utterances.], tot_loss[ctc_loss=0.1089, att_loss=0.2529, loss=0.2241, over 3260355.17 frames. utt_duration=1215 frames, utt_pad_proportion=0.06892, over 10743.88 utterances.], batch size: 39, lr: 1.07e-02, grad_scale: 8.0 2023-03-08 03:35:34,780 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-03-08 03:35:53,206 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.607e+02 2.391e+02 2.901e+02 3.646e+02 9.557e+02, threshold=5.803e+02, percent-clipped=2.0 2023-03-08 03:36:22,108 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-03-08 03:36:47,157 INFO [train2.py:809] (1/4) Epoch 10, batch 3100, loss[ctc_loss=0.1665, att_loss=0.2896, loss=0.265, over 13861.00 frames. utt_duration=378.6 frames, utt_pad_proportion=0.3369, over 147.00 utterances.], tot_loss[ctc_loss=0.1092, att_loss=0.253, loss=0.2243, over 3249124.65 frames. utt_duration=1189 frames, utt_pad_proportion=0.07716, over 10946.05 utterances.], batch size: 147, lr: 1.07e-02, grad_scale: 8.0 2023-03-08 03:36:56,730 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-03-08 03:38:08,096 INFO [train2.py:809] (1/4) Epoch 10, batch 3150, loss[ctc_loss=0.1098, att_loss=0.2467, loss=0.2193, over 16627.00 frames. utt_duration=1417 frames, utt_pad_proportion=0.005215, over 47.00 utterances.], tot_loss[ctc_loss=0.1091, att_loss=0.2531, loss=0.2243, over 3250512.94 frames. utt_duration=1188 frames, utt_pad_proportion=0.07693, over 10962.15 utterances.], batch size: 47, lr: 1.07e-02, grad_scale: 8.0 2023-03-08 03:38:28,470 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39016.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:38:34,152 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.681e+02 2.629e+02 3.270e+02 4.032e+02 9.981e+02, threshold=6.541e+02, percent-clipped=6.0 2023-03-08 03:38:35,273 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.94 vs. limit=5.0 2023-03-08 03:39:03,419 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9830, 5.3326, 5.2556, 5.1486, 5.3818, 5.2795, 5.0540, 4.8564], device='cuda:1'), covar=tensor([0.0968, 0.0340, 0.0211, 0.0478, 0.0222, 0.0291, 0.0278, 0.0276], device='cuda:1'), in_proj_covar=tensor([0.0441, 0.0281, 0.0237, 0.0269, 0.0333, 0.0356, 0.0278, 0.0315], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 03:39:27,537 INFO [train2.py:809] (1/4) Epoch 10, batch 3200, loss[ctc_loss=0.08089, att_loss=0.2471, loss=0.2138, over 16877.00 frames. utt_duration=1379 frames, utt_pad_proportion=0.007214, over 49.00 utterances.], tot_loss[ctc_loss=0.1079, att_loss=0.2519, loss=0.2231, over 3249510.48 frames. utt_duration=1212 frames, utt_pad_proportion=0.07184, over 10738.79 utterances.], batch size: 49, lr: 1.07e-02, grad_scale: 8.0 2023-03-08 03:39:44,531 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=39064.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:40:48,179 INFO [train2.py:809] (1/4) Epoch 10, batch 3250, loss[ctc_loss=0.1026, att_loss=0.2664, loss=0.2336, over 16610.00 frames. utt_duration=1415 frames, utt_pad_proportion=0.006081, over 47.00 utterances.], tot_loss[ctc_loss=0.1087, att_loss=0.2528, loss=0.2239, over 3253935.81 frames. utt_duration=1195 frames, utt_pad_proportion=0.07481, over 10903.78 utterances.], batch size: 47, lr: 1.07e-02, grad_scale: 8.0 2023-03-08 03:40:58,762 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0157, 4.9911, 4.9729, 2.4172, 2.0427, 2.7481, 2.5480, 3.7399], device='cuda:1'), covar=tensor([0.0658, 0.0187, 0.0187, 0.3673, 0.5992, 0.2653, 0.2869, 0.1908], device='cuda:1'), in_proj_covar=tensor([0.0337, 0.0217, 0.0236, 0.0206, 0.0356, 0.0342, 0.0234, 0.0359], device='cuda:1'), out_proj_covar=tensor([1.5401e-04, 8.2153e-05, 1.0169e-04, 9.3467e-05, 1.5672e-04, 1.3922e-04, 9.2590e-05, 1.5313e-04], device='cuda:1') 2023-03-08 03:41:13,727 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.722e+02 2.584e+02 3.081e+02 4.035e+02 1.100e+03, threshold=6.162e+02, percent-clipped=4.0 2023-03-08 03:42:08,484 INFO [train2.py:809] (1/4) Epoch 10, batch 3300, loss[ctc_loss=0.09013, att_loss=0.2233, loss=0.1966, over 15627.00 frames. utt_duration=1691 frames, utt_pad_proportion=0.009326, over 37.00 utterances.], tot_loss[ctc_loss=0.1089, att_loss=0.253, loss=0.2242, over 3262684.48 frames. utt_duration=1216 frames, utt_pad_proportion=0.06725, over 10746.53 utterances.], batch size: 37, lr: 1.07e-02, grad_scale: 8.0 2023-03-08 03:42:48,935 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39179.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:42:58,400 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39185.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:43:29,381 INFO [train2.py:809] (1/4) Epoch 10, batch 3350, loss[ctc_loss=0.1463, att_loss=0.2812, loss=0.2542, over 17056.00 frames. utt_duration=1314 frames, utt_pad_proportion=0.008534, over 52.00 utterances.], tot_loss[ctc_loss=0.1096, att_loss=0.2537, loss=0.2249, over 3265411.90 frames. utt_duration=1195 frames, utt_pad_proportion=0.07006, over 10943.89 utterances.], batch size: 52, lr: 1.07e-02, grad_scale: 8.0 2023-03-08 03:43:54,350 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.24 vs. limit=5.0 2023-03-08 03:43:54,949 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.724e+02 2.443e+02 2.910e+02 3.809e+02 9.404e+02, threshold=5.819e+02, percent-clipped=2.0 2023-03-08 03:44:29,206 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6892, 2.4062, 5.0434, 4.0944, 2.9737, 4.4423, 4.8624, 4.6684], device='cuda:1'), covar=tensor([0.0224, 0.1814, 0.0158, 0.0957, 0.1903, 0.0220, 0.0080, 0.0206], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0247, 0.0132, 0.0306, 0.0275, 0.0186, 0.0113, 0.0148], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-03-08 03:44:37,013 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39246.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:44:39,563 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.08 vs. limit=5.0 2023-03-08 03:44:49,736 INFO [train2.py:809] (1/4) Epoch 10, batch 3400, loss[ctc_loss=0.1917, att_loss=0.3003, loss=0.2786, over 14169.00 frames. utt_duration=392.6 frames, utt_pad_proportion=0.3172, over 145.00 utterances.], tot_loss[ctc_loss=0.1094, att_loss=0.2536, loss=0.2248, over 3257193.33 frames. utt_duration=1209 frames, utt_pad_proportion=0.06767, over 10790.54 utterances.], batch size: 145, lr: 1.07e-02, grad_scale: 8.0 2023-03-08 03:45:37,900 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39284.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:46:11,199 INFO [train2.py:809] (1/4) Epoch 10, batch 3450, loss[ctc_loss=0.1263, att_loss=0.2607, loss=0.2338, over 17298.00 frames. utt_duration=877.2 frames, utt_pad_proportion=0.08142, over 79.00 utterances.], tot_loss[ctc_loss=0.1092, att_loss=0.2536, loss=0.2247, over 3256194.24 frames. utt_duration=1178 frames, utt_pad_proportion=0.07547, over 11066.61 utterances.], batch size: 79, lr: 1.07e-02, grad_scale: 8.0 2023-03-08 03:46:25,902 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-03-08 03:46:36,009 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.709e+02 2.326e+02 2.969e+02 3.835e+02 9.960e+02, threshold=5.938e+02, percent-clipped=3.0 2023-03-08 03:46:43,636 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-03-08 03:46:50,427 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.4675, 3.5652, 3.6236, 3.0648, 3.6216, 3.6081, 3.4673, 2.5101], device='cuda:1'), covar=tensor([0.1219, 0.1435, 0.1397, 0.5289, 0.0774, 0.3367, 0.1032, 0.7622], device='cuda:1'), in_proj_covar=tensor([0.0095, 0.0117, 0.0123, 0.0190, 0.0098, 0.0176, 0.0103, 0.0173], device='cuda:1'), out_proj_covar=tensor([9.1190e-05, 1.0170e-04, 1.1052e-04, 1.5432e-04, 9.1559e-05, 1.4568e-04, 9.1353e-05, 1.4087e-04], device='cuda:1') 2023-03-08 03:47:16,367 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39345.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:47:21,068 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0004, 5.3658, 4.8249, 5.4610, 4.7234, 5.1172, 5.5231, 5.2854], device='cuda:1'), covar=tensor([0.0488, 0.0220, 0.0787, 0.0180, 0.0442, 0.0175, 0.0171, 0.0165], device='cuda:1'), in_proj_covar=tensor([0.0325, 0.0253, 0.0318, 0.0247, 0.0264, 0.0201, 0.0239, 0.0234], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0005, 0.0006, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 03:47:21,266 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9896, 4.8267, 4.5797, 4.7659, 5.2027, 4.7680, 4.7382, 2.2997], device='cuda:1'), covar=tensor([0.0139, 0.0228, 0.0246, 0.0175, 0.0937, 0.0177, 0.0196, 0.2222], device='cuda:1'), in_proj_covar=tensor([0.0122, 0.0126, 0.0133, 0.0135, 0.0327, 0.0122, 0.0117, 0.0217], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 03:47:31,577 INFO [train2.py:809] (1/4) Epoch 10, batch 3500, loss[ctc_loss=0.1113, att_loss=0.2349, loss=0.2102, over 15344.00 frames. utt_duration=1755 frames, utt_pad_proportion=0.01228, over 35.00 utterances.], tot_loss[ctc_loss=0.1084, att_loss=0.2532, loss=0.2243, over 3259782.98 frames. utt_duration=1208 frames, utt_pad_proportion=0.06748, over 10805.53 utterances.], batch size: 35, lr: 1.07e-02, grad_scale: 8.0 2023-03-08 03:48:32,248 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-03-08 03:48:52,043 INFO [train2.py:809] (1/4) Epoch 10, batch 3550, loss[ctc_loss=0.08157, att_loss=0.2235, loss=0.1951, over 15781.00 frames. utt_duration=1663 frames, utt_pad_proportion=0.007164, over 38.00 utterances.], tot_loss[ctc_loss=0.1084, att_loss=0.2533, loss=0.2243, over 3264381.19 frames. utt_duration=1219 frames, utt_pad_proportion=0.06473, over 10724.59 utterances.], batch size: 38, lr: 1.07e-02, grad_scale: 8.0 2023-03-08 03:49:17,422 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.387e+02 2.370e+02 2.861e+02 3.451e+02 5.590e+02, threshold=5.722e+02, percent-clipped=0.0 2023-03-08 03:49:36,493 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9455, 5.3500, 4.8673, 5.4243, 4.7703, 5.1138, 5.5188, 5.2907], device='cuda:1'), covar=tensor([0.0575, 0.0246, 0.0717, 0.0190, 0.0416, 0.0174, 0.0162, 0.0152], device='cuda:1'), in_proj_covar=tensor([0.0323, 0.0250, 0.0312, 0.0243, 0.0259, 0.0198, 0.0236, 0.0230], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0005, 0.0006, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 03:49:38,129 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2008, 4.6281, 4.6440, 4.8668, 2.7620, 4.4476, 2.9487, 1.7354], device='cuda:1'), covar=tensor([0.0305, 0.0156, 0.0510, 0.0119, 0.1692, 0.0180, 0.1310, 0.1785], device='cuda:1'), in_proj_covar=tensor([0.0131, 0.0104, 0.0246, 0.0105, 0.0215, 0.0104, 0.0218, 0.0198], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 03:50:11,855 INFO [train2.py:809] (1/4) Epoch 10, batch 3600, loss[ctc_loss=0.08971, att_loss=0.2394, loss=0.2094, over 16331.00 frames. utt_duration=1453 frames, utt_pad_proportion=0.00605, over 45.00 utterances.], tot_loss[ctc_loss=0.1084, att_loss=0.2529, loss=0.224, over 3261828.13 frames. utt_duration=1230 frames, utt_pad_proportion=0.06167, over 10618.29 utterances.], batch size: 45, lr: 1.07e-02, grad_scale: 8.0 2023-03-08 03:50:27,144 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39463.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:50:27,249 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7338, 2.6976, 4.9967, 3.8868, 3.0466, 4.3979, 4.9001, 4.6303], device='cuda:1'), covar=tensor([0.0179, 0.1718, 0.0125, 0.0992, 0.1804, 0.0225, 0.0087, 0.0201], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0246, 0.0131, 0.0305, 0.0273, 0.0186, 0.0113, 0.0148], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-03-08 03:50:52,020 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39479.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:51:32,919 INFO [train2.py:809] (1/4) Epoch 10, batch 3650, loss[ctc_loss=0.1055, att_loss=0.2597, loss=0.2289, over 17393.00 frames. utt_duration=1181 frames, utt_pad_proportion=0.01933, over 59.00 utterances.], tot_loss[ctc_loss=0.108, att_loss=0.2532, loss=0.2241, over 3272790.03 frames. utt_duration=1244 frames, utt_pad_proportion=0.05627, over 10539.40 utterances.], batch size: 59, lr: 1.07e-02, grad_scale: 8.0 2023-03-08 03:51:57,848 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.423e+02 2.407e+02 3.049e+02 3.792e+02 1.207e+03, threshold=6.099e+02, percent-clipped=6.0 2023-03-08 03:52:04,552 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39524.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:52:09,216 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=39527.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:52:13,106 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.85 vs. limit=2.0 2023-03-08 03:52:31,742 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39541.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:52:53,644 INFO [train2.py:809] (1/4) Epoch 10, batch 3700, loss[ctc_loss=0.102, att_loss=0.229, loss=0.2036, over 15370.00 frames. utt_duration=1758 frames, utt_pad_proportion=0.01056, over 35.00 utterances.], tot_loss[ctc_loss=0.1069, att_loss=0.2513, loss=0.2224, over 3256103.43 frames. utt_duration=1262 frames, utt_pad_proportion=0.05619, over 10331.53 utterances.], batch size: 35, lr: 1.07e-02, grad_scale: 8.0 2023-03-08 03:53:35,912 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4651, 4.9049, 4.6311, 4.8242, 4.9848, 4.6111, 3.3610, 4.8166], device='cuda:1'), covar=tensor([0.0102, 0.0101, 0.0134, 0.0091, 0.0081, 0.0106, 0.0717, 0.0210], device='cuda:1'), in_proj_covar=tensor([0.0072, 0.0070, 0.0084, 0.0052, 0.0056, 0.0067, 0.0089, 0.0089], 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') 2023-03-08 03:54:12,931 INFO [train2.py:809] (1/4) Epoch 10, batch 3750, loss[ctc_loss=0.1227, att_loss=0.2691, loss=0.2398, over 17363.00 frames. utt_duration=1179 frames, utt_pad_proportion=0.0202, over 59.00 utterances.], tot_loss[ctc_loss=0.107, att_loss=0.2515, loss=0.2226, over 3262883.86 frames. utt_duration=1290 frames, utt_pad_proportion=0.04776, over 10131.86 utterances.], batch size: 59, lr: 1.07e-02, grad_scale: 8.0 2023-03-08 03:54:38,432 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.452e+02 2.346e+02 2.722e+02 3.425e+02 7.090e+02, threshold=5.444e+02, percent-clipped=1.0 2023-03-08 03:54:56,810 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-03-08 03:55:08,110 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39639.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 03:55:09,567 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39640.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:55:33,235 INFO [train2.py:809] (1/4) Epoch 10, batch 3800, loss[ctc_loss=0.1078, att_loss=0.2616, loss=0.2309, over 17422.00 frames. utt_duration=1012 frames, utt_pad_proportion=0.04575, over 69.00 utterances.], tot_loss[ctc_loss=0.1074, att_loss=0.2523, loss=0.2233, over 3269579.36 frames. utt_duration=1254 frames, utt_pad_proportion=0.05411, over 10439.64 utterances.], batch size: 69, lr: 1.06e-02, grad_scale: 8.0 2023-03-08 03:55:39,724 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2700, 1.8400, 2.2440, 2.1456, 3.0365, 2.3082, 2.0266, 2.6046], device='cuda:1'), covar=tensor([0.2045, 0.7808, 0.4830, 0.4015, 0.2129, 0.2891, 0.5061, 0.2173], device='cuda:1'), in_proj_covar=tensor([0.0075, 0.0088, 0.0088, 0.0074, 0.0072, 0.0071, 0.0083, 0.0061], device='cuda:1'), out_proj_covar=tensor([4.7656e-05, 5.9178e-05, 5.8810e-05, 4.9602e-05, 4.6564e-05, 4.8957e-05, 5.6097e-05, 4.4011e-05], device='cuda:1') 2023-03-08 03:56:03,965 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.7377, 5.9478, 5.2231, 5.7440, 5.5292, 5.1980, 5.3627, 5.0473], device='cuda:1'), covar=tensor([0.1205, 0.0997, 0.0957, 0.0742, 0.0799, 0.1212, 0.2046, 0.2175], device='cuda:1'), in_proj_covar=tensor([0.0423, 0.0499, 0.0364, 0.0380, 0.0353, 0.0408, 0.0500, 0.0450], 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') 2023-03-08 03:56:46,651 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39700.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 03:56:53,084 INFO [train2.py:809] (1/4) Epoch 10, batch 3850, loss[ctc_loss=0.0996, att_loss=0.2625, loss=0.2299, over 17145.00 frames. utt_duration=1226 frames, utt_pad_proportion=0.01366, over 56.00 utterances.], tot_loss[ctc_loss=0.1066, att_loss=0.2519, loss=0.2229, over 3272491.98 frames. utt_duration=1285 frames, utt_pad_proportion=0.0462, over 10198.35 utterances.], batch size: 56, lr: 1.06e-02, grad_scale: 8.0 2023-03-08 03:57:07,138 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6088, 5.0498, 4.7833, 4.9627, 5.0837, 4.8009, 3.9648, 4.9602], device='cuda:1'), covar=tensor([0.0109, 0.0089, 0.0116, 0.0075, 0.0081, 0.0094, 0.0504, 0.0188], device='cuda:1'), in_proj_covar=tensor([0.0073, 0.0071, 0.0086, 0.0053, 0.0057, 0.0068, 0.0091, 0.0091], 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') 2023-03-08 03:57:18,322 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.719e+02 2.275e+02 2.627e+02 3.517e+02 7.871e+02, threshold=5.254e+02, percent-clipped=5.0 2023-03-08 03:57:51,425 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-03-08 03:58:10,237 INFO [train2.py:809] (1/4) Epoch 10, batch 3900, loss[ctc_loss=0.09887, att_loss=0.2235, loss=0.1985, over 16005.00 frames. utt_duration=1602 frames, utt_pad_proportion=0.00745, over 40.00 utterances.], tot_loss[ctc_loss=0.1068, att_loss=0.2519, loss=0.2229, over 3274616.65 frames. utt_duration=1289 frames, utt_pad_proportion=0.04461, over 10176.60 utterances.], batch size: 40, lr: 1.06e-02, grad_scale: 8.0 2023-03-08 03:58:11,877 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.2839, 5.6268, 4.8642, 5.4679, 5.2552, 4.9577, 5.0566, 4.8867], device='cuda:1'), covar=tensor([0.1581, 0.1021, 0.1137, 0.0806, 0.0939, 0.1489, 0.2280, 0.2380], device='cuda:1'), in_proj_covar=tensor([0.0432, 0.0508, 0.0370, 0.0384, 0.0358, 0.0417, 0.0505, 0.0457], 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') 2023-03-08 03:58:29,041 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.8532, 2.7155, 3.0450, 4.5649, 4.1346, 4.2837, 3.0659, 2.1528], device='cuda:1'), covar=tensor([0.0434, 0.2352, 0.1438, 0.0500, 0.0763, 0.0278, 0.1312, 0.2411], device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0210, 0.0189, 0.0191, 0.0187, 0.0148, 0.0194, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 03:58:30,703 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39767.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:58:31,464 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-03-08 03:59:07,363 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1959, 4.6749, 4.3908, 4.8645, 2.2375, 4.7492, 2.4785, 1.6243], device='cuda:1'), covar=tensor([0.0296, 0.0118, 0.0748, 0.0116, 0.2318, 0.0129, 0.1792, 0.2010], device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0107, 0.0254, 0.0107, 0.0224, 0.0107, 0.0225, 0.0205], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 03:59:26,946 INFO [train2.py:809] (1/4) Epoch 10, batch 3950, loss[ctc_loss=0.1006, att_loss=0.2296, loss=0.2038, over 15337.00 frames. utt_duration=1754 frames, utt_pad_proportion=0.01286, over 35.00 utterances.], tot_loss[ctc_loss=0.1069, att_loss=0.2523, loss=0.2232, over 3274647.35 frames. utt_duration=1282 frames, utt_pad_proportion=0.04523, over 10229.58 utterances.], batch size: 35, lr: 1.06e-02, grad_scale: 8.0 2023-03-08 03:59:44,049 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39815.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:59:50,030 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39819.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 03:59:51,183 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.448e+02 2.608e+02 3.275e+02 3.896e+02 7.798e+02, threshold=6.551e+02, percent-clipped=9.0 2023-03-08 04:00:04,078 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39828.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:00:45,405 INFO [train2.py:809] (1/4) Epoch 11, batch 0, loss[ctc_loss=0.1179, att_loss=0.2704, loss=0.2399, over 17221.00 frames. utt_duration=697.3 frames, utt_pad_proportion=0.124, over 99.00 utterances.], tot_loss[ctc_loss=0.1179, att_loss=0.2704, loss=0.2399, over 17221.00 frames. utt_duration=697.3 frames, utt_pad_proportion=0.124, over 99.00 utterances.], batch size: 99, lr: 1.01e-02, grad_scale: 8.0 2023-03-08 04:00:45,405 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-08 04:00:57,590 INFO [train2.py:843] (1/4) Epoch 11, validation: ctc_loss=0.05063, att_loss=0.2383, loss=0.2008, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-08 04:00:57,591 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-08 04:01:02,348 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39841.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:01:12,694 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-03-08 04:01:45,589 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.6163, 3.0082, 3.3983, 4.5450, 4.0388, 4.0439, 2.9341, 2.4052], device='cuda:1'), covar=tensor([0.0522, 0.1992, 0.0988, 0.0472, 0.0715, 0.0387, 0.1421, 0.2131], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0208, 0.0186, 0.0189, 0.0183, 0.0147, 0.0192, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 04:01:51,799 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-03-08 04:01:58,656 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39876.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:02:18,274 INFO [train2.py:809] (1/4) Epoch 11, batch 50, loss[ctc_loss=0.1249, att_loss=0.2735, loss=0.2438, over 17304.00 frames. utt_duration=1100 frames, utt_pad_proportion=0.03672, over 63.00 utterances.], tot_loss[ctc_loss=0.1049, att_loss=0.2544, loss=0.2245, over 746478.08 frames. utt_duration=1255 frames, utt_pad_proportion=0.04289, over 2382.62 utterances.], batch size: 63, lr: 1.01e-02, grad_scale: 8.0 2023-03-08 04:02:19,878 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=39889.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:03:08,052 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.374e+02 2.626e+02 3.075e+02 3.847e+02 1.291e+03, threshold=6.150e+02, percent-clipped=4.0 2023-03-08 04:03:37,055 INFO [train2.py:809] (1/4) Epoch 11, batch 100, loss[ctc_loss=0.1131, att_loss=0.2665, loss=0.2358, over 17015.00 frames. utt_duration=1336 frames, utt_pad_proportion=0.008582, over 51.00 utterances.], tot_loss[ctc_loss=0.1051, att_loss=0.2516, loss=0.2223, over 1302232.23 frames. utt_duration=1292 frames, utt_pad_proportion=0.03758, over 4034.95 utterances.], batch size: 51, lr: 1.01e-02, grad_scale: 8.0 2023-03-08 04:03:40,255 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39940.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:04:55,827 INFO [train2.py:809] (1/4) Epoch 11, batch 150, loss[ctc_loss=0.1052, att_loss=0.255, loss=0.225, over 16835.00 frames. utt_duration=681.5 frames, utt_pad_proportion=0.1417, over 99.00 utterances.], tot_loss[ctc_loss=0.1074, att_loss=0.2537, loss=0.2244, over 1749737.02 frames. utt_duration=1249 frames, utt_pad_proportion=0.04467, over 5612.53 utterances.], batch size: 99, lr: 1.01e-02, grad_scale: 16.0 2023-03-08 04:04:55,936 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=39988.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:05:07,701 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39995.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 04:05:51,853 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.648e+02 2.498e+02 3.139e+02 4.301e+02 8.550e+02, threshold=6.278e+02, percent-clipped=5.0 2023-03-08 04:06:01,236 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.9910, 3.8357, 3.1222, 3.3459, 3.8261, 3.5902, 2.7206, 4.3430], device='cuda:1'), covar=tensor([0.1086, 0.0471, 0.1208, 0.0792, 0.0739, 0.0791, 0.1003, 0.0470], device='cuda:1'), in_proj_covar=tensor([0.0183, 0.0179, 0.0205, 0.0174, 0.0232, 0.0209, 0.0182, 0.0249], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-08 04:06:20,491 INFO [train2.py:809] (1/4) Epoch 11, batch 200, loss[ctc_loss=0.1524, att_loss=0.2821, loss=0.2562, over 17235.00 frames. utt_duration=874.2 frames, utt_pad_proportion=0.08368, over 79.00 utterances.], tot_loss[ctc_loss=0.108, att_loss=0.254, loss=0.2248, over 2088660.15 frames. utt_duration=1233 frames, utt_pad_proportion=0.05083, over 6784.11 utterances.], batch size: 79, lr: 1.01e-02, grad_scale: 16.0 2023-03-08 04:07:39,778 INFO [train2.py:809] (1/4) Epoch 11, batch 250, loss[ctc_loss=0.09376, att_loss=0.254, loss=0.222, over 17025.00 frames. utt_duration=1337 frames, utt_pad_proportion=0.007599, over 51.00 utterances.], tot_loss[ctc_loss=0.1082, att_loss=0.2545, loss=0.2252, over 2354127.90 frames. utt_duration=1199 frames, utt_pad_proportion=0.05996, over 7862.92 utterances.], batch size: 51, lr: 1.01e-02, grad_scale: 16.0 2023-03-08 04:08:20,320 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1058, 5.2831, 5.0432, 2.6737, 5.1354, 4.9481, 4.4733, 2.7963], device='cuda:1'), covar=tensor([0.0121, 0.0075, 0.0266, 0.1263, 0.0072, 0.0129, 0.0318, 0.1540], device='cuda:1'), in_proj_covar=tensor([0.0060, 0.0083, 0.0075, 0.0104, 0.0069, 0.0094, 0.0093, 0.0101], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 04:08:28,139 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40119.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:08:29,482 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.404e+02 2.494e+02 2.990e+02 3.866e+02 8.648e+02, threshold=5.979e+02, percent-clipped=3.0 2023-03-08 04:08:34,913 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40123.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:08:59,716 INFO [train2.py:809] (1/4) Epoch 11, batch 300, loss[ctc_loss=0.09603, att_loss=0.2548, loss=0.2231, over 17011.00 frames. utt_duration=1336 frames, utt_pad_proportion=0.008989, over 51.00 utterances.], tot_loss[ctc_loss=0.1062, att_loss=0.2532, loss=0.2238, over 2557485.46 frames. utt_duration=1221 frames, utt_pad_proportion=0.05615, over 8391.72 utterances.], batch size: 51, lr: 1.01e-02, grad_scale: 16.0 2023-03-08 04:09:14,406 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.3546, 2.5063, 2.7913, 4.2783, 3.9796, 3.9331, 2.7334, 1.6826], device='cuda:1'), covar=tensor([0.0548, 0.2152, 0.1199, 0.0459, 0.0507, 0.0332, 0.1428, 0.2449], device='cuda:1'), in_proj_covar=tensor([0.0163, 0.0207, 0.0186, 0.0187, 0.0180, 0.0148, 0.0190, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 04:09:45,571 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=40167.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:09:52,682 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40171.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:10:05,885 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.5969, 3.6454, 2.9867, 3.3830, 3.7853, 3.4753, 2.7716, 4.1814], device='cuda:1'), covar=tensor([0.1211, 0.0572, 0.1098, 0.0694, 0.0735, 0.0740, 0.0924, 0.0583], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0181, 0.0207, 0.0174, 0.0236, 0.0211, 0.0183, 0.0249], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-08 04:10:06,026 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0341, 5.0544, 4.9655, 2.5133, 1.9706, 2.6401, 3.3037, 3.8237], device='cuda:1'), covar=tensor([0.0700, 0.0186, 0.0185, 0.3769, 0.5829, 0.2877, 0.1904, 0.1821], device='cuda:1'), in_proj_covar=tensor([0.0334, 0.0218, 0.0232, 0.0210, 0.0353, 0.0338, 0.0231, 0.0356], device='cuda:1'), out_proj_covar=tensor([1.5204e-04, 8.1497e-05, 9.9498e-05, 9.4452e-05, 1.5503e-04, 1.3746e-04, 9.1453e-05, 1.5161e-04], device='cuda:1') 2023-03-08 04:10:20,965 INFO [train2.py:809] (1/4) Epoch 11, batch 350, loss[ctc_loss=0.07355, att_loss=0.2181, loss=0.1892, over 15353.00 frames. utt_duration=1756 frames, utt_pad_proportion=0.01191, over 35.00 utterances.], tot_loss[ctc_loss=0.1066, att_loss=0.2532, loss=0.2238, over 2712679.17 frames. utt_duration=1208 frames, utt_pad_proportion=0.06175, over 8996.50 utterances.], batch size: 35, lr: 1.01e-02, grad_scale: 16.0 2023-03-08 04:10:31,939 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40195.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:11:08,026 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40218.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:11:11,921 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.687e+02 2.384e+02 2.929e+02 3.608e+02 1.665e+03, threshold=5.857e+02, percent-clipped=2.0 2023-03-08 04:11:41,141 INFO [train2.py:809] (1/4) Epoch 11, batch 400, loss[ctc_loss=0.1063, att_loss=0.2578, loss=0.2275, over 17352.00 frames. utt_duration=1103 frames, utt_pad_proportion=0.03635, over 63.00 utterances.], tot_loss[ctc_loss=0.1064, att_loss=0.2529, loss=0.2236, over 2839140.02 frames. utt_duration=1206 frames, utt_pad_proportion=0.06164, over 9431.04 utterances.], batch size: 63, lr: 1.01e-02, grad_scale: 16.0 2023-03-08 04:12:09,441 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40256.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:12:36,177 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40272.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:12:47,310 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40279.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:13:01,946 INFO [train2.py:809] (1/4) Epoch 11, batch 450, loss[ctc_loss=0.1173, att_loss=0.274, loss=0.2427, over 17057.00 frames. utt_duration=1288 frames, utt_pad_proportion=0.008853, over 53.00 utterances.], tot_loss[ctc_loss=0.1055, att_loss=0.2519, loss=0.2226, over 2934717.53 frames. utt_duration=1232 frames, utt_pad_proportion=0.05503, over 9538.96 utterances.], batch size: 53, lr: 1.01e-02, grad_scale: 16.0 2023-03-08 04:13:12,967 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40295.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 04:13:52,365 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.652e+02 2.417e+02 2.903e+02 3.339e+02 6.124e+02, threshold=5.806e+02, percent-clipped=1.0 2023-03-08 04:14:08,012 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40330.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:14:13,298 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40333.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:14:21,407 INFO [train2.py:809] (1/4) Epoch 11, batch 500, loss[ctc_loss=0.07956, att_loss=0.2216, loss=0.1932, over 15632.00 frames. utt_duration=1691 frames, utt_pad_proportion=0.00912, over 37.00 utterances.], tot_loss[ctc_loss=0.1055, att_loss=0.2519, loss=0.2226, over 3014245.65 frames. utt_duration=1234 frames, utt_pad_proportion=0.05373, over 9784.17 utterances.], batch size: 37, lr: 1.01e-02, grad_scale: 16.0 2023-03-08 04:14:29,168 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=40343.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 04:15:41,805 INFO [train2.py:809] (1/4) Epoch 11, batch 550, loss[ctc_loss=0.09919, att_loss=0.2491, loss=0.2191, over 16411.00 frames. utt_duration=1493 frames, utt_pad_proportion=0.007117, over 44.00 utterances.], tot_loss[ctc_loss=0.1057, att_loss=0.2518, loss=0.2226, over 3080211.68 frames. utt_duration=1252 frames, utt_pad_proportion=0.04772, over 9851.86 utterances.], batch size: 44, lr: 1.01e-02, grad_scale: 16.0 2023-03-08 04:15:46,927 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40391.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:16:31,901 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9833, 5.3494, 4.7892, 5.4382, 4.7614, 5.0041, 5.5053, 5.2666], device='cuda:1'), covar=tensor([0.0624, 0.0307, 0.0862, 0.0235, 0.0456, 0.0246, 0.0220, 0.0219], device='cuda:1'), in_proj_covar=tensor([0.0331, 0.0255, 0.0317, 0.0249, 0.0264, 0.0201, 0.0240, 0.0234], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0005, 0.0006, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 04:16:33,184 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.575e+02 2.316e+02 2.892e+02 3.547e+02 1.294e+03, threshold=5.784e+02, percent-clipped=4.0 2023-03-08 04:16:38,003 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40423.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:17:02,140 INFO [train2.py:809] (1/4) Epoch 11, batch 600, loss[ctc_loss=0.1478, att_loss=0.2813, loss=0.2546, over 17317.00 frames. utt_duration=1261 frames, utt_pad_proportion=0.01013, over 55.00 utterances.], tot_loss[ctc_loss=0.1061, att_loss=0.2522, loss=0.2229, over 3123077.35 frames. utt_duration=1244 frames, utt_pad_proportion=0.05225, over 10054.16 utterances.], batch size: 55, lr: 1.01e-02, grad_scale: 16.0 2023-03-08 04:17:54,375 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=40471.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:17:54,589 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40471.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:18:22,138 INFO [train2.py:809] (1/4) Epoch 11, batch 650, loss[ctc_loss=0.1084, att_loss=0.2421, loss=0.2154, over 15953.00 frames. utt_duration=1558 frames, utt_pad_proportion=0.006346, over 41.00 utterances.], tot_loss[ctc_loss=0.1066, att_loss=0.2522, loss=0.2231, over 3158397.00 frames. utt_duration=1257 frames, utt_pad_proportion=0.04932, over 10064.24 utterances.], batch size: 41, lr: 1.01e-02, grad_scale: 16.0 2023-03-08 04:18:27,258 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.2668, 3.3971, 3.3600, 2.9426, 3.3479, 3.4182, 3.2946, 2.4605], device='cuda:1'), covar=tensor([0.1291, 0.1922, 0.5012, 0.6982, 0.9888, 0.4120, 0.1626, 0.7724], device='cuda:1'), in_proj_covar=tensor([0.0094, 0.0118, 0.0125, 0.0191, 0.0098, 0.0178, 0.0103, 0.0169], device='cuda:1'), out_proj_covar=tensor([9.0743e-05, 1.0315e-04, 1.1245e-04, 1.5529e-04, 9.1639e-05, 1.4755e-04, 9.1173e-05, 1.3907e-04], device='cuda:1') 2023-03-08 04:18:28,805 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40492.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:19:11,083 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=40519.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:19:12,491 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.622e+02 2.380e+02 3.105e+02 3.848e+02 6.391e+02, threshold=6.210e+02, percent-clipped=2.0 2023-03-08 04:19:41,596 INFO [train2.py:809] (1/4) Epoch 11, batch 700, loss[ctc_loss=0.1282, att_loss=0.2664, loss=0.2387, over 17176.00 frames. utt_duration=871.4 frames, utt_pad_proportion=0.08467, over 79.00 utterances.], tot_loss[ctc_loss=0.1068, att_loss=0.2526, loss=0.2235, over 3184905.82 frames. utt_duration=1255 frames, utt_pad_proportion=0.05056, over 10164.77 utterances.], batch size: 79, lr: 1.01e-02, grad_scale: 16.0 2023-03-08 04:20:01,442 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40551.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:20:04,676 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40553.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:20:17,760 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-03-08 04:20:20,822 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-03-08 04:20:38,011 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40574.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:21:00,395 INFO [train2.py:809] (1/4) Epoch 11, batch 750, loss[ctc_loss=0.1413, att_loss=0.2761, loss=0.2492, over 17503.00 frames. utt_duration=887.7 frames, utt_pad_proportion=0.07142, over 79.00 utterances.], tot_loss[ctc_loss=0.1063, att_loss=0.2522, loss=0.223, over 3202787.90 frames. utt_duration=1263 frames, utt_pad_proportion=0.0502, over 10151.80 utterances.], batch size: 79, lr: 1.00e-02, grad_scale: 16.0 2023-03-08 04:21:21,304 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.6881, 3.5713, 3.5358, 3.0661, 3.6103, 3.7206, 3.4451, 2.6710], device='cuda:1'), covar=tensor([0.0828, 0.1450, 0.3033, 0.7230, 0.2098, 0.2925, 0.1208, 0.7645], device='cuda:1'), in_proj_covar=tensor([0.0093, 0.0118, 0.0124, 0.0191, 0.0099, 0.0178, 0.0103, 0.0169], device='cuda:1'), out_proj_covar=tensor([9.0389e-05, 1.0340e-04, 1.1214e-04, 1.5560e-04, 9.2466e-05, 1.4755e-04, 9.1122e-05, 1.3891e-04], device='cuda:1') 2023-03-08 04:21:21,828 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=2.00 vs. limit=2.0 2023-03-08 04:21:51,405 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.282e+02 2.391e+02 2.736e+02 3.441e+02 6.786e+02, threshold=5.472e+02, percent-clipped=2.0 2023-03-08 04:22:03,978 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40628.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:22:20,791 INFO [train2.py:809] (1/4) Epoch 11, batch 800, loss[ctc_loss=0.1191, att_loss=0.2573, loss=0.2296, over 16882.00 frames. utt_duration=1380 frames, utt_pad_proportion=0.007326, over 49.00 utterances.], tot_loss[ctc_loss=0.107, att_loss=0.2533, loss=0.224, over 3228061.85 frames. utt_duration=1247 frames, utt_pad_proportion=0.05182, over 10368.73 utterances.], batch size: 49, lr: 1.00e-02, grad_scale: 16.0 2023-03-08 04:22:41,723 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40651.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:23:38,259 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40686.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:23:41,161 INFO [train2.py:809] (1/4) Epoch 11, batch 850, loss[ctc_loss=0.1049, att_loss=0.2651, loss=0.233, over 16623.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.005528, over 47.00 utterances.], tot_loss[ctc_loss=0.1063, att_loss=0.2536, loss=0.2242, over 3243231.27 frames. utt_duration=1248 frames, utt_pad_proportion=0.05207, over 10409.27 utterances.], batch size: 47, lr: 1.00e-02, grad_scale: 16.0 2023-03-08 04:24:18,958 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40712.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:24:31,775 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.342e+02 2.347e+02 2.849e+02 3.646e+02 6.696e+02, threshold=5.699e+02, percent-clipped=1.0 2023-03-08 04:25:00,584 INFO [train2.py:809] (1/4) Epoch 11, batch 900, loss[ctc_loss=0.1001, att_loss=0.25, loss=0.22, over 16373.00 frames. utt_duration=1490 frames, utt_pad_proportion=0.008513, over 44.00 utterances.], tot_loss[ctc_loss=0.1057, att_loss=0.2529, loss=0.2235, over 3248262.28 frames. utt_duration=1240 frames, utt_pad_proportion=0.05445, over 10488.47 utterances.], batch size: 44, lr: 1.00e-02, grad_scale: 16.0 2023-03-08 04:25:08,660 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.8029, 6.0816, 5.4276, 5.8823, 5.7196, 5.2640, 5.4244, 5.2320], device='cuda:1'), covar=tensor([0.1173, 0.0875, 0.0737, 0.0727, 0.0788, 0.1370, 0.1984, 0.2210], device='cuda:1'), in_proj_covar=tensor([0.0433, 0.0499, 0.0366, 0.0380, 0.0355, 0.0413, 0.0508, 0.0453], 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') 2023-03-08 04:25:21,148 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.8318, 3.6266, 3.6363, 3.1111, 3.5956, 3.7357, 3.5010, 2.6459], device='cuda:1'), covar=tensor([0.0973, 0.1662, 0.3933, 0.7078, 0.1908, 0.6563, 0.1358, 0.7480], device='cuda:1'), in_proj_covar=tensor([0.0093, 0.0117, 0.0123, 0.0189, 0.0098, 0.0177, 0.0102, 0.0167], device='cuda:1'), out_proj_covar=tensor([8.9807e-05, 1.0299e-04, 1.1079e-04, 1.5440e-04, 9.1757e-05, 1.4687e-04, 9.0530e-05, 1.3775e-04], device='cuda:1') 2023-03-08 04:26:21,118 INFO [train2.py:809] (1/4) Epoch 11, batch 950, loss[ctc_loss=0.07186, att_loss=0.2207, loss=0.1909, over 15891.00 frames. utt_duration=1631 frames, utt_pad_proportion=0.008402, over 39.00 utterances.], tot_loss[ctc_loss=0.1052, att_loss=0.2527, loss=0.2232, over 3252480.62 frames. utt_duration=1243 frames, utt_pad_proportion=0.05473, over 10476.09 utterances.], batch size: 39, lr: 1.00e-02, grad_scale: 16.0 2023-03-08 04:27:12,368 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.493e+02 2.301e+02 2.865e+02 3.345e+02 1.136e+03, threshold=5.731e+02, percent-clipped=4.0 2023-03-08 04:27:41,336 INFO [train2.py:809] (1/4) Epoch 11, batch 1000, loss[ctc_loss=0.1187, att_loss=0.2599, loss=0.2316, over 16900.00 frames. utt_duration=1381 frames, utt_pad_proportion=0.005833, over 49.00 utterances.], tot_loss[ctc_loss=0.1052, att_loss=0.2529, loss=0.2233, over 3264153.50 frames. utt_duration=1245 frames, utt_pad_proportion=0.05112, over 10501.42 utterances.], batch size: 49, lr: 1.00e-02, grad_scale: 16.0 2023-03-08 04:27:45,838 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.88 vs. limit=5.0 2023-03-08 04:27:56,997 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40848.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:28:01,770 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40851.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:28:25,264 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40865.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:28:27,258 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-03-08 04:28:39,129 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40874.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:28:41,824 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.87 vs. limit=2.0 2023-03-08 04:28:48,591 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.46 vs. limit=5.0 2023-03-08 04:29:01,314 INFO [train2.py:809] (1/4) Epoch 11, batch 1050, loss[ctc_loss=0.1142, att_loss=0.2687, loss=0.2378, over 17010.00 frames. utt_duration=1336 frames, utt_pad_proportion=0.008399, over 51.00 utterances.], tot_loss[ctc_loss=0.1045, att_loss=0.252, loss=0.2225, over 3262710.08 frames. utt_duration=1261 frames, utt_pad_proportion=0.04867, over 10358.70 utterances.], batch size: 51, lr: 1.00e-02, grad_scale: 16.0 2023-03-08 04:29:18,878 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=40899.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:29:20,780 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40900.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:29:52,390 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.885e+02 2.443e+02 2.925e+02 3.809e+02 9.259e+02, threshold=5.849e+02, percent-clipped=3.0 2023-03-08 04:29:56,100 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=40922.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:30:03,307 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40926.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:30:06,355 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40928.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:30:22,024 INFO [train2.py:809] (1/4) Epoch 11, batch 1100, loss[ctc_loss=0.104, att_loss=0.2349, loss=0.2087, over 15881.00 frames. utt_duration=1630 frames, utt_pad_proportion=0.009471, over 39.00 utterances.], tot_loss[ctc_loss=0.1036, att_loss=0.2505, loss=0.2211, over 3259852.12 frames. utt_duration=1290 frames, utt_pad_proportion=0.04398, over 10123.43 utterances.], batch size: 39, lr: 1.00e-02, grad_scale: 16.0 2023-03-08 04:30:59,244 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40961.0, num_to_drop=1, layers_to_drop={3} 2023-03-08 04:31:02,569 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4701, 4.4811, 4.5050, 4.6406, 5.0028, 4.5983, 4.4242, 2.4406], device='cuda:1'), covar=tensor([0.0278, 0.0276, 0.0238, 0.0162, 0.0993, 0.0201, 0.0283, 0.1974], device='cuda:1'), in_proj_covar=tensor([0.0123, 0.0128, 0.0133, 0.0135, 0.0320, 0.0120, 0.0121, 0.0216], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 04:31:23,532 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=40976.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:31:39,184 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40986.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:31:42,049 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-03-08 04:31:42,183 INFO [train2.py:809] (1/4) Epoch 11, batch 1150, loss[ctc_loss=0.1172, att_loss=0.2688, loss=0.2385, over 17031.00 frames. utt_duration=1337 frames, utt_pad_proportion=0.007307, over 51.00 utterances.], tot_loss[ctc_loss=0.1047, att_loss=0.2506, loss=0.2215, over 3254763.07 frames. utt_duration=1274 frames, utt_pad_proportion=0.05025, over 10229.97 utterances.], batch size: 51, lr: 1.00e-02, grad_scale: 16.0 2023-03-08 04:31:56,056 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40997.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:32:12,920 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41007.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:32:34,184 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.489e+02 2.552e+02 3.086e+02 3.791e+02 8.496e+02, threshold=6.172e+02, percent-clipped=3.0 2023-03-08 04:32:56,345 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=41034.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:33:02,227 INFO [train2.py:809] (1/4) Epoch 11, batch 1200, loss[ctc_loss=0.1202, att_loss=0.2751, loss=0.2441, over 17124.00 frames. utt_duration=1225 frames, utt_pad_proportion=0.01477, over 56.00 utterances.], tot_loss[ctc_loss=0.1047, att_loss=0.2506, loss=0.2214, over 3257357.36 frames. utt_duration=1292 frames, utt_pad_proportion=0.04649, over 10095.16 utterances.], batch size: 56, lr: 9.99e-03, grad_scale: 16.0 2023-03-08 04:33:13,629 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41045.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:33:34,443 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41058.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:34:22,828 INFO [train2.py:809] (1/4) Epoch 11, batch 1250, loss[ctc_loss=0.06807, att_loss=0.2154, loss=0.1859, over 15367.00 frames. utt_duration=1758 frames, utt_pad_proportion=0.01126, over 35.00 utterances.], tot_loss[ctc_loss=0.1035, att_loss=0.2498, loss=0.2206, over 3266167.77 frames. utt_duration=1290 frames, utt_pad_proportion=0.04403, over 10139.71 utterances.], batch size: 35, lr: 9.99e-03, grad_scale: 16.0 2023-03-08 04:34:45,093 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41102.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:34:51,808 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41106.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:35:14,658 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.546e+02 2.278e+02 2.660e+02 3.579e+02 5.724e+02, threshold=5.320e+02, percent-clipped=1.0 2023-03-08 04:35:42,808 INFO [train2.py:809] (1/4) Epoch 11, batch 1300, loss[ctc_loss=0.168, att_loss=0.2978, loss=0.2718, over 17131.00 frames. utt_duration=1225 frames, utt_pad_proportion=0.01353, over 56.00 utterances.], tot_loss[ctc_loss=0.1035, att_loss=0.2499, loss=0.2206, over 3263659.96 frames. utt_duration=1281 frames, utt_pad_proportion=0.04686, over 10205.55 utterances.], batch size: 56, lr: 9.98e-03, grad_scale: 16.0 2023-03-08 04:35:58,636 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41148.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:36:23,567 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41163.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:37:02,503 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.66 vs. limit=2.0 2023-03-08 04:37:03,119 INFO [train2.py:809] (1/4) Epoch 11, batch 1350, loss[ctc_loss=0.09523, att_loss=0.2556, loss=0.2235, over 17051.00 frames. utt_duration=1313 frames, utt_pad_proportion=0.008955, over 52.00 utterances.], tot_loss[ctc_loss=0.104, att_loss=0.2501, loss=0.2209, over 3262016.56 frames. utt_duration=1273 frames, utt_pad_proportion=0.04975, over 10258.23 utterances.], batch size: 52, lr: 9.98e-03, grad_scale: 16.0 2023-03-08 04:37:15,569 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=41196.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:37:55,022 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.568e+02 2.465e+02 2.945e+02 3.733e+02 9.191e+02, threshold=5.889e+02, percent-clipped=5.0 2023-03-08 04:37:56,866 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41221.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:38:16,338 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7526, 5.3012, 5.0581, 5.2510, 5.2472, 4.8992, 3.6053, 5.1083], device='cuda:1'), covar=tensor([0.0092, 0.0075, 0.0083, 0.0059, 0.0071, 0.0089, 0.0625, 0.0154], device='cuda:1'), in_proj_covar=tensor([0.0072, 0.0070, 0.0086, 0.0051, 0.0057, 0.0067, 0.0089, 0.0089], 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') 2023-03-08 04:38:23,863 INFO [train2.py:809] (1/4) Epoch 11, batch 1400, loss[ctc_loss=0.08269, att_loss=0.2415, loss=0.2097, over 16942.00 frames. utt_duration=1357 frames, utt_pad_proportion=0.009028, over 50.00 utterances.], tot_loss[ctc_loss=0.1033, att_loss=0.2498, loss=0.2205, over 3262664.26 frames. utt_duration=1254 frames, utt_pad_proportion=0.05369, over 10415.97 utterances.], batch size: 50, lr: 9.97e-03, grad_scale: 16.0 2023-03-08 04:38:52,754 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41256.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 04:39:44,638 INFO [train2.py:809] (1/4) Epoch 11, batch 1450, loss[ctc_loss=0.1012, att_loss=0.2523, loss=0.2221, over 16689.00 frames. utt_duration=1452 frames, utt_pad_proportion=0.006557, over 46.00 utterances.], tot_loss[ctc_loss=0.1048, att_loss=0.2506, loss=0.2215, over 3259582.93 frames. utt_duration=1219 frames, utt_pad_proportion=0.06237, over 10709.54 utterances.], batch size: 46, lr: 9.96e-03, grad_scale: 16.0 2023-03-08 04:40:15,976 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41307.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:40:37,477 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.706e+02 2.447e+02 2.872e+02 3.322e+02 6.594e+02, threshold=5.743e+02, percent-clipped=2.0 2023-03-08 04:40:56,425 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6358, 4.6202, 4.5162, 4.5764, 5.1661, 4.7272, 4.4935, 2.2156], device='cuda:1'), covar=tensor([0.0195, 0.0264, 0.0273, 0.0228, 0.0885, 0.0179, 0.0288, 0.2181], device='cuda:1'), in_proj_covar=tensor([0.0127, 0.0132, 0.0137, 0.0139, 0.0329, 0.0122, 0.0124, 0.0222], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 04:41:05,352 INFO [train2.py:809] (1/4) Epoch 11, batch 1500, loss[ctc_loss=0.0905, att_loss=0.2436, loss=0.213, over 16136.00 frames. utt_duration=1538 frames, utt_pad_proportion=0.005864, over 42.00 utterances.], tot_loss[ctc_loss=0.1052, att_loss=0.2514, loss=0.2222, over 3269263.18 frames. utt_duration=1199 frames, utt_pad_proportion=0.06519, over 10919.54 utterances.], batch size: 42, lr: 9.96e-03, grad_scale: 16.0 2023-03-08 04:41:23,701 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7439, 4.7392, 4.5607, 4.6652, 5.2184, 4.8123, 4.5381, 2.1864], device='cuda:1'), covar=tensor([0.0203, 0.0228, 0.0268, 0.0219, 0.0809, 0.0173, 0.0283, 0.2286], device='cuda:1'), in_proj_covar=tensor([0.0127, 0.0132, 0.0138, 0.0139, 0.0329, 0.0121, 0.0124, 0.0221], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 04:41:30,424 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41353.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:41:33,406 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=41355.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:42:25,612 INFO [train2.py:809] (1/4) Epoch 11, batch 1550, loss[ctc_loss=0.1029, att_loss=0.2561, loss=0.2255, over 17303.00 frames. utt_duration=1260 frames, utt_pad_proportion=0.01115, over 55.00 utterances.], tot_loss[ctc_loss=0.1034, att_loss=0.2503, loss=0.2209, over 3273293.93 frames. utt_duration=1241 frames, utt_pad_proportion=0.05518, over 10560.46 utterances.], batch size: 55, lr: 9.95e-03, grad_scale: 16.0 2023-03-08 04:42:29,138 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.6738, 3.9878, 3.8510, 3.8937, 3.9515, 3.7732, 3.0035, 3.8358], device='cuda:1'), covar=tensor([0.0118, 0.0103, 0.0125, 0.0089, 0.0086, 0.0114, 0.0622, 0.0222], device='cuda:1'), in_proj_covar=tensor([0.0071, 0.0068, 0.0084, 0.0051, 0.0056, 0.0066, 0.0087, 0.0088], 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') 2023-03-08 04:42:47,302 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41401.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:43:19,828 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.785e+02 2.586e+02 2.982e+02 3.761e+02 9.691e+02, threshold=5.963e+02, percent-clipped=6.0 2023-03-08 04:43:32,652 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-03-08 04:43:45,719 INFO [train2.py:809] (1/4) Epoch 11, batch 1600, loss[ctc_loss=0.1175, att_loss=0.267, loss=0.2371, over 17280.00 frames. utt_duration=1258 frames, utt_pad_proportion=0.01305, over 55.00 utterances.], tot_loss[ctc_loss=0.1041, att_loss=0.2504, loss=0.2211, over 3272489.07 frames. utt_duration=1229 frames, utt_pad_proportion=0.05768, over 10661.11 utterances.], batch size: 55, lr: 9.95e-03, grad_scale: 8.0 2023-03-08 04:44:17,908 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41458.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:45:02,536 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5562, 4.5529, 4.3894, 4.6331, 5.0620, 4.6797, 4.5049, 2.2327], device='cuda:1'), covar=tensor([0.0229, 0.0283, 0.0326, 0.0196, 0.0922, 0.0189, 0.0282, 0.2258], device='cuda:1'), in_proj_covar=tensor([0.0127, 0.0133, 0.0139, 0.0139, 0.0330, 0.0122, 0.0124, 0.0222], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 04:45:05,419 INFO [train2.py:809] (1/4) Epoch 11, batch 1650, loss[ctc_loss=0.09197, att_loss=0.2615, loss=0.2276, over 16329.00 frames. utt_duration=1453 frames, utt_pad_proportion=0.006338, over 45.00 utterances.], tot_loss[ctc_loss=0.1031, att_loss=0.2498, loss=0.2205, over 3273336.67 frames. utt_duration=1250 frames, utt_pad_proportion=0.05312, over 10484.84 utterances.], batch size: 45, lr: 9.94e-03, grad_scale: 8.0 2023-03-08 04:45:15,961 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-03-08 04:45:58,326 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.357e+02 2.532e+02 2.866e+02 3.682e+02 7.453e+02, threshold=5.731e+02, percent-clipped=3.0 2023-03-08 04:45:58,630 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41521.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:46:25,299 INFO [train2.py:809] (1/4) Epoch 11, batch 1700, loss[ctc_loss=0.09929, att_loss=0.2603, loss=0.2281, over 17317.00 frames. utt_duration=1176 frames, utt_pad_proportion=0.02344, over 59.00 utterances.], tot_loss[ctc_loss=0.1028, att_loss=0.2502, loss=0.2207, over 3276688.07 frames. utt_duration=1249 frames, utt_pad_proportion=0.05328, over 10508.97 utterances.], batch size: 59, lr: 9.93e-03, grad_scale: 8.0 2023-03-08 04:46:55,400 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41556.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 04:47:11,672 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7831, 5.0773, 5.1092, 5.0496, 5.1624, 5.1223, 4.7667, 4.6579], device='cuda:1'), covar=tensor([0.1194, 0.0665, 0.0298, 0.0457, 0.0313, 0.0291, 0.0360, 0.0371], device='cuda:1'), in_proj_covar=tensor([0.0455, 0.0287, 0.0243, 0.0279, 0.0344, 0.0359, 0.0288, 0.0321], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 04:47:16,318 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=41569.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:47:23,133 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-03-08 04:47:46,032 INFO [train2.py:809] (1/4) Epoch 11, batch 1750, loss[ctc_loss=0.106, att_loss=0.2705, loss=0.2376, over 17399.00 frames. utt_duration=1106 frames, utt_pad_proportion=0.03205, over 63.00 utterances.], tot_loss[ctc_loss=0.1044, att_loss=0.2514, loss=0.222, over 3282000.90 frames. utt_duration=1239 frames, utt_pad_proportion=0.05478, over 10612.86 utterances.], batch size: 63, lr: 9.93e-03, grad_scale: 8.0 2023-03-08 04:48:12,443 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=41604.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:48:39,823 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.573e+02 2.288e+02 2.927e+02 3.480e+02 6.791e+02, threshold=5.853e+02, percent-clipped=3.0 2023-03-08 04:48:41,886 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6232, 2.6869, 5.0383, 4.0930, 3.2349, 4.6334, 5.0489, 4.7723], device='cuda:1'), covar=tensor([0.0263, 0.1659, 0.0259, 0.0988, 0.1726, 0.0204, 0.0111, 0.0217], device='cuda:1'), in_proj_covar=tensor([0.0148, 0.0243, 0.0133, 0.0305, 0.0276, 0.0185, 0.0117, 0.0154], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001], device='cuda:1') 2023-03-08 04:49:06,235 INFO [train2.py:809] (1/4) Epoch 11, batch 1800, loss[ctc_loss=0.1175, att_loss=0.2682, loss=0.2381, over 17041.00 frames. utt_duration=1312 frames, utt_pad_proportion=0.008686, over 52.00 utterances.], tot_loss[ctc_loss=0.1038, att_loss=0.2507, loss=0.2213, over 3272921.33 frames. utt_duration=1276 frames, utt_pad_proportion=0.0475, over 10271.92 utterances.], batch size: 52, lr: 9.92e-03, grad_scale: 8.0 2023-03-08 04:49:11,238 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0538, 5.3477, 4.8331, 5.4221, 4.7998, 5.0054, 5.5143, 5.2679], device='cuda:1'), covar=tensor([0.0512, 0.0275, 0.0835, 0.0209, 0.0397, 0.0207, 0.0193, 0.0156], device='cuda:1'), in_proj_covar=tensor([0.0331, 0.0255, 0.0317, 0.0250, 0.0260, 0.0203, 0.0242, 0.0233], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0005, 0.0006, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 04:49:31,685 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41653.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:49:33,273 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41654.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:50:27,077 INFO [train2.py:809] (1/4) Epoch 11, batch 1850, loss[ctc_loss=0.06542, att_loss=0.2101, loss=0.1812, over 15498.00 frames. utt_duration=1724 frames, utt_pad_proportion=0.008785, over 36.00 utterances.], tot_loss[ctc_loss=0.1038, att_loss=0.2504, loss=0.2211, over 3262539.60 frames. utt_duration=1247 frames, utt_pad_proportion=0.05821, over 10477.11 utterances.], batch size: 36, lr: 9.92e-03, grad_scale: 8.0 2023-03-08 04:50:48,885 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=41701.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:50:49,104 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41701.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:51:12,127 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41715.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 04:51:20,775 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.644e+02 2.415e+02 2.840e+02 3.452e+02 8.289e+02, threshold=5.680e+02, percent-clipped=4.0 2023-03-08 04:51:31,193 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-03-08 04:51:46,880 INFO [train2.py:809] (1/4) Epoch 11, batch 1900, loss[ctc_loss=0.1051, att_loss=0.2483, loss=0.2197, over 15960.00 frames. utt_duration=1558 frames, utt_pad_proportion=0.006144, over 41.00 utterances.], tot_loss[ctc_loss=0.1035, att_loss=0.25, loss=0.2207, over 3258862.85 frames. utt_duration=1256 frames, utt_pad_proportion=0.05726, over 10389.46 utterances.], batch size: 41, lr: 9.91e-03, grad_scale: 8.0 2023-03-08 04:52:04,786 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=41749.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:52:19,244 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41758.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:53:07,401 INFO [train2.py:809] (1/4) Epoch 11, batch 1950, loss[ctc_loss=0.08775, att_loss=0.2276, loss=0.1997, over 15785.00 frames. utt_duration=1663 frames, utt_pad_proportion=0.007741, over 38.00 utterances.], tot_loss[ctc_loss=0.1041, att_loss=0.2511, loss=0.2217, over 3265042.34 frames. utt_duration=1248 frames, utt_pad_proportion=0.05675, over 10473.96 utterances.], batch size: 38, lr: 9.91e-03, grad_scale: 8.0 2023-03-08 04:53:37,478 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=41806.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:53:44,157 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.8396, 3.7376, 2.9671, 3.3821, 3.9061, 3.5021, 2.6321, 4.3021], device='cuda:1'), covar=tensor([0.0962, 0.0505, 0.1162, 0.0635, 0.0585, 0.0668, 0.0963, 0.0430], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0182, 0.0202, 0.0173, 0.0235, 0.0215, 0.0184, 0.0252], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-08 04:54:00,607 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.413e+02 2.550e+02 3.116e+02 4.012e+02 6.949e+02, threshold=6.232e+02, percent-clipped=6.0 2023-03-08 04:54:16,276 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6792, 5.0504, 4.7685, 5.0678, 4.9884, 4.7113, 3.7638, 5.0374], device='cuda:1'), covar=tensor([0.0083, 0.0096, 0.0127, 0.0057, 0.0093, 0.0110, 0.0583, 0.0175], device='cuda:1'), in_proj_covar=tensor([0.0072, 0.0071, 0.0087, 0.0052, 0.0058, 0.0069, 0.0091, 0.0092], 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') 2023-03-08 04:54:27,440 INFO [train2.py:809] (1/4) Epoch 11, batch 2000, loss[ctc_loss=0.09791, att_loss=0.2298, loss=0.2034, over 15371.00 frames. utt_duration=1758 frames, utt_pad_proportion=0.01106, over 35.00 utterances.], tot_loss[ctc_loss=0.1042, att_loss=0.2513, loss=0.2219, over 3271103.96 frames. utt_duration=1239 frames, utt_pad_proportion=0.05731, over 10576.48 utterances.], batch size: 35, lr: 9.90e-03, grad_scale: 8.0 2023-03-08 04:55:02,117 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1104, 5.2279, 5.0774, 2.5249, 1.9665, 2.8359, 2.9823, 3.8469], device='cuda:1'), covar=tensor([0.0634, 0.0248, 0.0208, 0.4070, 0.6027, 0.2463, 0.2073, 0.1896], device='cuda:1'), in_proj_covar=tensor([0.0333, 0.0222, 0.0238, 0.0210, 0.0352, 0.0337, 0.0231, 0.0355], device='cuda:1'), out_proj_covar=tensor([1.5061e-04, 8.2990e-05, 1.0192e-04, 9.4490e-05, 1.5395e-04, 1.3689e-04, 9.1349e-05, 1.5048e-04], device='cuda:1') 2023-03-08 04:55:12,757 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1046, 4.6280, 4.2371, 4.7593, 2.1508, 4.5916, 2.2539, 1.7136], device='cuda:1'), covar=tensor([0.0329, 0.0126, 0.0925, 0.0108, 0.2628, 0.0140, 0.1959, 0.2067], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0109, 0.0257, 0.0109, 0.0219, 0.0109, 0.0222, 0.0201], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 04:55:29,649 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6566, 4.6775, 4.6405, 2.6238, 4.4638, 4.3752, 3.9815, 2.3751], device='cuda:1'), covar=tensor([0.0098, 0.0090, 0.0142, 0.1075, 0.0091, 0.0211, 0.0328, 0.1541], device='cuda:1'), in_proj_covar=tensor([0.0060, 0.0083, 0.0075, 0.0104, 0.0069, 0.0094, 0.0092, 0.0100], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 04:55:47,163 INFO [train2.py:809] (1/4) Epoch 11, batch 2050, loss[ctc_loss=0.1054, att_loss=0.2649, loss=0.233, over 17049.00 frames. utt_duration=1288 frames, utt_pad_proportion=0.009746, over 53.00 utterances.], tot_loss[ctc_loss=0.1057, att_loss=0.2521, loss=0.2228, over 3274992.08 frames. utt_duration=1226 frames, utt_pad_proportion=0.05885, over 10696.70 utterances.], batch size: 53, lr: 9.89e-03, grad_scale: 8.0 2023-03-08 04:56:05,532 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.4234, 2.8714, 3.0792, 4.2498, 3.9131, 4.0040, 2.9919, 2.1255], device='cuda:1'), covar=tensor([0.0683, 0.2132, 0.1167, 0.0623, 0.0758, 0.0375, 0.1460, 0.2543], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0204, 0.0184, 0.0188, 0.0188, 0.0148, 0.0192, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 04:56:13,327 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0940, 5.1297, 5.0284, 2.3771, 1.8380, 2.7569, 2.7531, 3.6642], device='cuda:1'), covar=tensor([0.0640, 0.0207, 0.0200, 0.4174, 0.6047, 0.2640, 0.2239, 0.2091], device='cuda:1'), in_proj_covar=tensor([0.0328, 0.0219, 0.0235, 0.0207, 0.0345, 0.0332, 0.0227, 0.0352], device='cuda:1'), out_proj_covar=tensor([1.4823e-04, 8.1713e-05, 1.0056e-04, 9.2994e-05, 1.5106e-04, 1.3480e-04, 8.9898e-05, 1.4891e-04], device='cuda:1') 2023-03-08 04:56:17,832 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41906.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:56:40,437 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.467e+02 2.492e+02 2.943e+02 3.583e+02 1.498e+03, threshold=5.886e+02, percent-clipped=5.0 2023-03-08 04:57:07,334 INFO [train2.py:809] (1/4) Epoch 11, batch 2100, loss[ctc_loss=0.1095, att_loss=0.2613, loss=0.231, over 17377.00 frames. utt_duration=881.5 frames, utt_pad_proportion=0.07696, over 79.00 utterances.], tot_loss[ctc_loss=0.1045, att_loss=0.2518, loss=0.2224, over 3277735.95 frames. utt_duration=1244 frames, utt_pad_proportion=0.05443, over 10554.58 utterances.], batch size: 79, lr: 9.89e-03, grad_scale: 8.0 2023-03-08 04:57:23,993 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41948.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 04:57:54,637 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41967.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 04:58:19,394 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-03-08 04:58:27,971 INFO [train2.py:809] (1/4) Epoch 11, batch 2150, loss[ctc_loss=0.1061, att_loss=0.2633, loss=0.2318, over 17042.00 frames. utt_duration=1288 frames, utt_pad_proportion=0.01017, over 53.00 utterances.], tot_loss[ctc_loss=0.1053, att_loss=0.2523, loss=0.2229, over 3282732.74 frames. utt_duration=1220 frames, utt_pad_proportion=0.05768, over 10774.17 utterances.], batch size: 53, lr: 9.88e-03, grad_scale: 8.0 2023-03-08 04:59:07,021 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42009.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 04:59:08,239 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42010.0, num_to_drop=1, layers_to_drop={3} 2023-03-08 04:59:25,317 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.495e+02 2.469e+02 2.816e+02 3.466e+02 1.172e+03, threshold=5.632e+02, percent-clipped=3.0 2023-03-08 04:59:52,147 INFO [train2.py:809] (1/4) Epoch 11, batch 2200, loss[ctc_loss=0.1316, att_loss=0.261, loss=0.2351, over 16543.00 frames. utt_duration=1471 frames, utt_pad_proportion=0.006422, over 45.00 utterances.], tot_loss[ctc_loss=0.1039, att_loss=0.2506, loss=0.2212, over 3273498.37 frames. utt_duration=1258 frames, utt_pad_proportion=0.05063, over 10418.09 utterances.], batch size: 45, lr: 9.88e-03, grad_scale: 8.0 2023-03-08 05:01:12,043 INFO [train2.py:809] (1/4) Epoch 11, batch 2250, loss[ctc_loss=0.1041, att_loss=0.2391, loss=0.2121, over 16017.00 frames. utt_duration=1603 frames, utt_pad_proportion=0.006691, over 40.00 utterances.], tot_loss[ctc_loss=0.1048, att_loss=0.2508, loss=0.2216, over 3260744.50 frames. utt_duration=1218 frames, utt_pad_proportion=0.06538, over 10720.31 utterances.], batch size: 40, lr: 9.87e-03, grad_scale: 8.0 2023-03-08 05:02:04,982 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.420e+02 2.270e+02 2.801e+02 3.546e+02 6.959e+02, threshold=5.601e+02, percent-clipped=2.0 2023-03-08 05:02:32,246 INFO [train2.py:809] (1/4) Epoch 11, batch 2300, loss[ctc_loss=0.08592, att_loss=0.2484, loss=0.2159, over 16679.00 frames. utt_duration=1452 frames, utt_pad_proportion=0.006354, over 46.00 utterances.], tot_loss[ctc_loss=0.104, att_loss=0.2507, loss=0.2214, over 3263303.91 frames. utt_duration=1220 frames, utt_pad_proportion=0.06423, over 10715.40 utterances.], batch size: 46, lr: 9.86e-03, grad_scale: 8.0 2023-03-08 05:03:52,294 INFO [train2.py:809] (1/4) Epoch 11, batch 2350, loss[ctc_loss=0.1162, att_loss=0.2694, loss=0.2388, over 17315.00 frames. utt_duration=1261 frames, utt_pad_proportion=0.01119, over 55.00 utterances.], tot_loss[ctc_loss=0.1046, att_loss=0.2517, loss=0.2223, over 3272968.13 frames. utt_duration=1224 frames, utt_pad_proportion=0.06129, over 10704.67 utterances.], batch size: 55, lr: 9.86e-03, grad_scale: 8.0 2023-03-08 05:04:33,927 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7709, 2.7479, 5.1978, 4.0454, 2.9502, 4.5630, 4.9246, 4.8011], device='cuda:1'), covar=tensor([0.0254, 0.1714, 0.0198, 0.1016, 0.2001, 0.0221, 0.0121, 0.0250], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0245, 0.0135, 0.0305, 0.0279, 0.0187, 0.0120, 0.0155], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0003, 0.0003, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 05:04:45,693 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.909e+02 2.492e+02 2.954e+02 3.764e+02 8.189e+02, threshold=5.908e+02, percent-clipped=4.0 2023-03-08 05:05:13,513 INFO [train2.py:809] (1/4) Epoch 11, batch 2400, loss[ctc_loss=0.1079, att_loss=0.272, loss=0.2392, over 17425.00 frames. utt_duration=1108 frames, utt_pad_proportion=0.02921, over 63.00 utterances.], tot_loss[ctc_loss=0.1039, att_loss=0.2514, loss=0.2219, over 3271573.09 frames. utt_duration=1224 frames, utt_pad_proportion=0.06265, over 10701.16 utterances.], batch size: 63, lr: 9.85e-03, grad_scale: 8.0 2023-03-08 05:05:30,272 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-03-08 05:05:52,550 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42262.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:06:30,226 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-03-08 05:06:34,469 INFO [train2.py:809] (1/4) Epoch 11, batch 2450, loss[ctc_loss=0.1207, att_loss=0.2693, loss=0.2396, over 16765.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.006718, over 48.00 utterances.], tot_loss[ctc_loss=0.1036, att_loss=0.2511, loss=0.2216, over 3269608.09 frames. utt_duration=1236 frames, utt_pad_proportion=0.06097, over 10592.69 utterances.], batch size: 48, lr: 9.85e-03, grad_scale: 8.0 2023-03-08 05:06:59,997 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42304.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 05:07:09,509 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42310.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 05:07:26,991 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.597e+02 2.356e+02 2.915e+02 3.557e+02 7.393e+02, threshold=5.830e+02, percent-clipped=4.0 2023-03-08 05:07:54,612 INFO [train2.py:809] (1/4) Epoch 11, batch 2500, loss[ctc_loss=0.1306, att_loss=0.2496, loss=0.2258, over 14544.00 frames. utt_duration=1820 frames, utt_pad_proportion=0.04183, over 32.00 utterances.], tot_loss[ctc_loss=0.1034, att_loss=0.2508, loss=0.2214, over 3263434.81 frames. utt_duration=1237 frames, utt_pad_proportion=0.06215, over 10566.90 utterances.], batch size: 32, lr: 9.84e-03, grad_scale: 8.0 2023-03-08 05:08:26,081 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=42358.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:08:59,973 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42379.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:09:14,577 INFO [train2.py:809] (1/4) Epoch 11, batch 2550, loss[ctc_loss=0.1175, att_loss=0.2372, loss=0.2133, over 15414.00 frames. utt_duration=1763 frames, utt_pad_proportion=0.008404, over 35.00 utterances.], tot_loss[ctc_loss=0.1029, att_loss=0.251, loss=0.2214, over 3273349.27 frames. utt_duration=1241 frames, utt_pad_proportion=0.05776, over 10562.06 utterances.], batch size: 35, lr: 9.84e-03, grad_scale: 8.0 2023-03-08 05:10:06,487 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.521e+02 2.376e+02 2.981e+02 3.759e+02 6.397e+02, threshold=5.961e+02, percent-clipped=1.0 2023-03-08 05:10:34,788 INFO [train2.py:809] (1/4) Epoch 11, batch 2600, loss[ctc_loss=0.08555, att_loss=0.2273, loss=0.1989, over 16120.00 frames. utt_duration=1537 frames, utt_pad_proportion=0.006772, over 42.00 utterances.], tot_loss[ctc_loss=0.1028, att_loss=0.2512, loss=0.2215, over 3279721.69 frames. utt_duration=1240 frames, utt_pad_proportion=0.05538, over 10588.53 utterances.], batch size: 42, lr: 9.83e-03, grad_scale: 8.0 2023-03-08 05:10:38,274 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42440.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:11:55,322 INFO [train2.py:809] (1/4) Epoch 11, batch 2650, loss[ctc_loss=0.1021, att_loss=0.2525, loss=0.2224, over 16617.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.005663, over 47.00 utterances.], tot_loss[ctc_loss=0.1028, att_loss=0.251, loss=0.2213, over 3284061.49 frames. utt_duration=1259 frames, utt_pad_proportion=0.04945, over 10447.08 utterances.], batch size: 47, lr: 9.82e-03, grad_scale: 8.0 2023-03-08 05:12:20,793 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.6258, 3.7732, 3.5489, 3.6640, 4.0030, 3.6268, 3.4851, 2.4539], device='cuda:1'), covar=tensor([0.0261, 0.0331, 0.0340, 0.0228, 0.0761, 0.0257, 0.0408, 0.1612], device='cuda:1'), in_proj_covar=tensor([0.0124, 0.0130, 0.0137, 0.0140, 0.0327, 0.0119, 0.0121, 0.0216], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 05:12:23,795 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42506.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:12:47,503 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.508e+02 2.235e+02 2.761e+02 3.286e+02 8.509e+02, threshold=5.523e+02, percent-clipped=3.0 2023-03-08 05:13:03,859 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0614, 5.1147, 4.9296, 2.4615, 1.9645, 2.8741, 2.7317, 3.8556], device='cuda:1'), covar=tensor([0.0668, 0.0217, 0.0213, 0.3989, 0.5675, 0.2449, 0.2254, 0.1696], device='cuda:1'), in_proj_covar=tensor([0.0332, 0.0223, 0.0239, 0.0210, 0.0349, 0.0334, 0.0228, 0.0353], device='cuda:1'), out_proj_covar=tensor([1.5012e-04, 8.3564e-05, 1.0224e-04, 9.4001e-05, 1.5249e-04, 1.3562e-04, 9.0233e-05, 1.4919e-04], device='cuda:1') 2023-03-08 05:13:15,976 INFO [train2.py:809] (1/4) Epoch 11, batch 2700, loss[ctc_loss=0.08095, att_loss=0.2452, loss=0.2123, over 16624.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.005424, over 47.00 utterances.], tot_loss[ctc_loss=0.1028, att_loss=0.2512, loss=0.2215, over 3280815.24 frames. utt_duration=1259 frames, utt_pad_proportion=0.05059, over 10437.14 utterances.], batch size: 47, lr: 9.82e-03, grad_scale: 8.0 2023-03-08 05:13:42,558 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7755, 5.1208, 5.0177, 5.2038, 5.2681, 4.6948, 3.6176, 5.1655], device='cuda:1'), covar=tensor([0.0089, 0.0106, 0.0104, 0.0062, 0.0076, 0.0130, 0.0642, 0.0153], device='cuda:1'), in_proj_covar=tensor([0.0074, 0.0072, 0.0087, 0.0053, 0.0059, 0.0070, 0.0091, 0.0091], 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') 2023-03-08 05:13:53,875 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42562.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:14:02,126 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42567.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:14:25,071 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.1703, 3.8932, 3.2571, 3.4643, 4.1192, 3.7815, 3.0504, 4.4584], device='cuda:1'), covar=tensor([0.0867, 0.0485, 0.1063, 0.0696, 0.0613, 0.0606, 0.0856, 0.0416], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0182, 0.0201, 0.0173, 0.0234, 0.0211, 0.0181, 0.0251], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-08 05:14:37,197 INFO [train2.py:809] (1/4) Epoch 11, batch 2750, loss[ctc_loss=0.1345, att_loss=0.2737, loss=0.2459, over 17312.00 frames. utt_duration=1260 frames, utt_pad_proportion=0.01073, over 55.00 utterances.], tot_loss[ctc_loss=0.1026, att_loss=0.2507, loss=0.2211, over 3281923.10 frames. utt_duration=1273 frames, utt_pad_proportion=0.04749, over 10324.80 utterances.], batch size: 55, lr: 9.81e-03, grad_scale: 8.0 2023-03-08 05:15:02,400 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42604.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 05:15:12,235 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=42610.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:15:30,221 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.628e+02 2.426e+02 2.955e+02 3.528e+02 6.434e+02, threshold=5.910e+02, percent-clipped=2.0 2023-03-08 05:15:57,767 INFO [train2.py:809] (1/4) Epoch 11, batch 2800, loss[ctc_loss=0.1279, att_loss=0.2639, loss=0.2367, over 16977.00 frames. utt_duration=687.5 frames, utt_pad_proportion=0.1373, over 99.00 utterances.], tot_loss[ctc_loss=0.1027, att_loss=0.251, loss=0.2213, over 3276478.69 frames. utt_duration=1259 frames, utt_pad_proportion=0.05245, over 10420.29 utterances.], batch size: 99, lr: 9.81e-03, grad_scale: 8.0 2023-03-08 05:16:15,460 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42649.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:16:19,867 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=42652.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 05:17:17,562 INFO [train2.py:809] (1/4) Epoch 11, batch 2850, loss[ctc_loss=0.1129, att_loss=0.2628, loss=0.2328, over 17042.00 frames. utt_duration=1312 frames, utt_pad_proportion=0.0087, over 52.00 utterances.], tot_loss[ctc_loss=0.1028, att_loss=0.2511, loss=0.2214, over 3278111.33 frames. utt_duration=1258 frames, utt_pad_proportion=0.05242, over 10432.42 utterances.], batch size: 52, lr: 9.80e-03, grad_scale: 8.0 2023-03-08 05:17:52,683 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42710.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:18:10,594 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.499e+02 2.468e+02 2.915e+02 3.973e+02 1.035e+03, threshold=5.829e+02, percent-clipped=4.0 2023-03-08 05:18:33,536 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42735.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:18:38,144 INFO [train2.py:809] (1/4) Epoch 11, batch 2900, loss[ctc_loss=0.1545, att_loss=0.2806, loss=0.2554, over 14042.00 frames. utt_duration=386.2 frames, utt_pad_proportion=0.3271, over 146.00 utterances.], tot_loss[ctc_loss=0.1031, att_loss=0.2513, loss=0.2216, over 3278204.52 frames. utt_duration=1227 frames, utt_pad_proportion=0.06021, over 10696.18 utterances.], batch size: 146, lr: 9.80e-03, grad_scale: 8.0 2023-03-08 05:19:58,449 INFO [train2.py:809] (1/4) Epoch 11, batch 2950, loss[ctc_loss=0.08265, att_loss=0.217, loss=0.1901, over 15376.00 frames. utt_duration=1759 frames, utt_pad_proportion=0.01078, over 35.00 utterances.], tot_loss[ctc_loss=0.1021, att_loss=0.2502, loss=0.2206, over 3276775.30 frames. utt_duration=1253 frames, utt_pad_proportion=0.05478, over 10469.44 utterances.], batch size: 35, lr: 9.79e-03, grad_scale: 8.0 2023-03-08 05:20:52,199 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.678e+02 2.273e+02 2.678e+02 3.532e+02 7.141e+02, threshold=5.356e+02, percent-clipped=1.0 2023-03-08 05:21:19,018 INFO [train2.py:809] (1/4) Epoch 11, batch 3000, loss[ctc_loss=0.08516, att_loss=0.2252, loss=0.1972, over 15959.00 frames. utt_duration=1559 frames, utt_pad_proportion=0.006051, over 41.00 utterances.], tot_loss[ctc_loss=0.1022, att_loss=0.25, loss=0.2205, over 3276029.06 frames. utt_duration=1254 frames, utt_pad_proportion=0.05493, over 10466.12 utterances.], batch size: 41, lr: 9.78e-03, grad_scale: 8.0 2023-03-08 05:21:19,018 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-08 05:21:32,741 INFO [train2.py:843] (1/4) Epoch 11, validation: ctc_loss=0.04985, att_loss=0.2382, loss=0.2006, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-08 05:21:32,742 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-08 05:21:58,020 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7102, 3.7123, 3.5656, 3.0997, 3.5642, 3.6562, 3.6242, 2.5880], device='cuda:1'), covar=tensor([0.1067, 0.1407, 0.3046, 0.7333, 0.2356, 0.3094, 0.0951, 0.7874], device='cuda:1'), in_proj_covar=tensor([0.0101, 0.0119, 0.0129, 0.0195, 0.0102, 0.0180, 0.0105, 0.0170], device='cuda:1'), out_proj_covar=tensor([9.7174e-05, 1.0513e-04, 1.1610e-04, 1.5974e-04, 9.5323e-05, 1.5001e-04, 9.3085e-05, 1.4083e-04], device='cuda:1') 2023-03-08 05:21:59,812 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1542, 5.1779, 4.9599, 2.5834, 1.9955, 3.2180, 3.4176, 3.9647], device='cuda:1'), covar=tensor([0.0601, 0.0264, 0.0265, 0.3953, 0.5881, 0.2171, 0.1726, 0.1771], device='cuda:1'), in_proj_covar=tensor([0.0337, 0.0225, 0.0244, 0.0214, 0.0352, 0.0337, 0.0231, 0.0359], device='cuda:1'), out_proj_covar=tensor([1.5174e-04, 8.4112e-05, 1.0430e-04, 9.6280e-05, 1.5385e-04, 1.3687e-04, 9.0624e-05, 1.5156e-04], device='cuda:1') 2023-03-08 05:22:05,851 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42859.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:22:10,434 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42862.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:22:37,873 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42879.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:22:52,096 INFO [train2.py:809] (1/4) Epoch 11, batch 3050, loss[ctc_loss=0.0966, att_loss=0.264, loss=0.2305, over 16341.00 frames. utt_duration=1454 frames, utt_pad_proportion=0.005396, over 45.00 utterances.], tot_loss[ctc_loss=0.1014, att_loss=0.2494, loss=0.2198, over 3270450.96 frames. utt_duration=1275 frames, utt_pad_proportion=0.05118, over 10275.34 utterances.], batch size: 45, lr: 9.78e-03, grad_scale: 8.0 2023-03-08 05:23:43,391 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42920.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:23:44,614 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.578e+02 2.391e+02 3.019e+02 3.669e+02 9.404e+02, threshold=6.038e+02, percent-clipped=6.0 2023-03-08 05:24:11,644 INFO [train2.py:809] (1/4) Epoch 11, batch 3100, loss[ctc_loss=0.09798, att_loss=0.2362, loss=0.2085, over 15870.00 frames. utt_duration=1629 frames, utt_pad_proportion=0.008953, over 39.00 utterances.], tot_loss[ctc_loss=0.1018, att_loss=0.2495, loss=0.22, over 3269109.95 frames. utt_duration=1276 frames, utt_pad_proportion=0.05017, over 10262.83 utterances.], batch size: 39, lr: 9.77e-03, grad_scale: 8.0 2023-03-08 05:24:15,756 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42940.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:24:22,003 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2989, 2.8412, 3.6768, 2.7414, 3.5013, 4.4743, 4.1660, 3.3021], device='cuda:1'), covar=tensor([0.0367, 0.1627, 0.0974, 0.1541, 0.0861, 0.0642, 0.0645, 0.1180], device='cuda:1'), in_proj_covar=tensor([0.0225, 0.0226, 0.0246, 0.0203, 0.0237, 0.0301, 0.0217, 0.0219], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 05:25:08,416 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0333, 5.3389, 4.8067, 5.3789, 4.7352, 4.9609, 5.4484, 5.2478], device='cuda:1'), covar=tensor([0.0481, 0.0280, 0.0835, 0.0218, 0.0413, 0.0189, 0.0199, 0.0172], device='cuda:1'), in_proj_covar=tensor([0.0337, 0.0262, 0.0318, 0.0254, 0.0265, 0.0204, 0.0249, 0.0237], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0005, 0.0006, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 05:25:21,596 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1677, 4.9328, 4.6753, 4.7024, 2.5215, 4.5553, 3.0604, 1.7071], device='cuda:1'), covar=tensor([0.0298, 0.0121, 0.0562, 0.0133, 0.1801, 0.0153, 0.1217, 0.1755], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0110, 0.0252, 0.0107, 0.0216, 0.0107, 0.0219, 0.0197], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 05:25:32,676 INFO [train2.py:809] (1/4) Epoch 11, batch 3150, loss[ctc_loss=0.1071, att_loss=0.2671, loss=0.2351, over 17108.00 frames. utt_duration=1223 frames, utt_pad_proportion=0.01579, over 56.00 utterances.], tot_loss[ctc_loss=0.1014, att_loss=0.2493, loss=0.2197, over 3276617.53 frames. utt_duration=1287 frames, utt_pad_proportion=0.04507, over 10193.54 utterances.], batch size: 56, lr: 9.77e-03, grad_scale: 8.0 2023-03-08 05:25:42,897 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0601, 4.1924, 4.0036, 4.1393, 4.5417, 4.2287, 4.0473, 2.3082], device='cuda:1'), covar=tensor([0.0200, 0.0332, 0.0358, 0.0202, 0.0778, 0.0178, 0.0270, 0.1959], device='cuda:1'), in_proj_covar=tensor([0.0124, 0.0130, 0.0136, 0.0141, 0.0329, 0.0119, 0.0119, 0.0214], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 05:26:00,839 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43005.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:26:26,643 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.352e+02 2.340e+02 2.878e+02 3.301e+02 6.036e+02, threshold=5.756e+02, percent-clipped=0.0 2023-03-08 05:26:49,534 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43035.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:26:53,986 INFO [train2.py:809] (1/4) Epoch 11, batch 3200, loss[ctc_loss=0.07198, att_loss=0.2223, loss=0.1922, over 15487.00 frames. utt_duration=1722 frames, utt_pad_proportion=0.0096, over 36.00 utterances.], tot_loss[ctc_loss=0.1014, att_loss=0.2491, loss=0.2196, over 3270537.67 frames. utt_duration=1251 frames, utt_pad_proportion=0.05624, over 10470.13 utterances.], batch size: 36, lr: 9.76e-03, grad_scale: 8.0 2023-03-08 05:27:12,156 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5679, 3.0591, 3.7878, 3.0551, 3.7061, 4.7311, 4.4796, 3.7795], device='cuda:1'), covar=tensor([0.0403, 0.1680, 0.1012, 0.1397, 0.0875, 0.0787, 0.0509, 0.0964], device='cuda:1'), in_proj_covar=tensor([0.0229, 0.0231, 0.0251, 0.0207, 0.0242, 0.0307, 0.0221, 0.0222], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 05:28:06,592 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=43083.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:28:14,941 INFO [train2.py:809] (1/4) Epoch 11, batch 3250, loss[ctc_loss=0.1044, att_loss=0.2563, loss=0.2259, over 17337.00 frames. utt_duration=1177 frames, utt_pad_proportion=0.02266, over 59.00 utterances.], tot_loss[ctc_loss=0.1026, att_loss=0.2497, loss=0.2203, over 3266079.79 frames. utt_duration=1220 frames, utt_pad_proportion=0.06434, over 10720.01 utterances.], batch size: 59, lr: 9.76e-03, grad_scale: 8.0 2023-03-08 05:29:07,588 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.575e+02 2.382e+02 2.901e+02 3.870e+02 8.262e+02, threshold=5.801e+02, percent-clipped=4.0 2023-03-08 05:29:35,238 INFO [train2.py:809] (1/4) Epoch 11, batch 3300, loss[ctc_loss=0.1207, att_loss=0.2516, loss=0.2254, over 16118.00 frames. utt_duration=1536 frames, utt_pad_proportion=0.006849, over 42.00 utterances.], tot_loss[ctc_loss=0.1037, att_loss=0.25, loss=0.2208, over 3253500.39 frames. utt_duration=1187 frames, utt_pad_proportion=0.07754, over 10978.21 utterances.], batch size: 42, lr: 9.75e-03, grad_scale: 8.0 2023-03-08 05:30:12,590 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43162.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:30:55,500 INFO [train2.py:809] (1/4) Epoch 11, batch 3350, loss[ctc_loss=0.1135, att_loss=0.2695, loss=0.2383, over 17343.00 frames. utt_duration=1177 frames, utt_pad_proportion=0.02137, over 59.00 utterances.], tot_loss[ctc_loss=0.1039, att_loss=0.2501, loss=0.2209, over 3257423.69 frames. utt_duration=1202 frames, utt_pad_proportion=0.07229, over 10850.63 utterances.], batch size: 59, lr: 9.75e-03, grad_scale: 8.0 2023-03-08 05:31:30,913 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=43210.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:31:39,584 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43215.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:31:48,997 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-03-08 05:31:49,462 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.559e+02 2.322e+02 2.901e+02 3.775e+02 1.145e+03, threshold=5.802e+02, percent-clipped=9.0 2023-03-08 05:32:12,666 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43235.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:32:17,048 INFO [train2.py:809] (1/4) Epoch 11, batch 3400, loss[ctc_loss=0.09067, att_loss=0.2528, loss=0.2204, over 16324.00 frames. utt_duration=1452 frames, utt_pad_proportion=0.006536, over 45.00 utterances.], tot_loss[ctc_loss=0.103, att_loss=0.2501, loss=0.2207, over 3253370.81 frames. utt_duration=1221 frames, utt_pad_proportion=0.06807, over 10669.16 utterances.], batch size: 45, lr: 9.74e-03, grad_scale: 8.0 2023-03-08 05:33:10,604 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.00 vs. limit=5.0 2023-03-08 05:33:37,403 INFO [train2.py:809] (1/4) Epoch 11, batch 3450, loss[ctc_loss=0.1127, att_loss=0.2644, loss=0.2341, over 17042.00 frames. utt_duration=1288 frames, utt_pad_proportion=0.01037, over 53.00 utterances.], tot_loss[ctc_loss=0.103, att_loss=0.2502, loss=0.2208, over 3247401.70 frames. utt_duration=1195 frames, utt_pad_proportion=0.07638, over 10887.72 utterances.], batch size: 53, lr: 9.73e-03, grad_scale: 8.0 2023-03-08 05:34:03,785 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43305.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:34:29,499 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.748e+02 2.188e+02 2.793e+02 3.558e+02 5.760e+02, threshold=5.586e+02, percent-clipped=0.0 2023-03-08 05:34:57,080 INFO [train2.py:809] (1/4) Epoch 11, batch 3500, loss[ctc_loss=0.1001, att_loss=0.2323, loss=0.2058, over 15636.00 frames. utt_duration=1692 frames, utt_pad_proportion=0.008381, over 37.00 utterances.], tot_loss[ctc_loss=0.1029, att_loss=0.2505, loss=0.221, over 3255372.38 frames. utt_duration=1200 frames, utt_pad_proportion=0.0729, over 10865.74 utterances.], batch size: 37, lr: 9.73e-03, grad_scale: 8.0 2023-03-08 05:35:20,536 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=43353.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:36:02,215 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-03-08 05:36:17,030 INFO [train2.py:809] (1/4) Epoch 11, batch 3550, loss[ctc_loss=0.1111, att_loss=0.2648, loss=0.2341, over 17102.00 frames. utt_duration=1223 frames, utt_pad_proportion=0.0152, over 56.00 utterances.], tot_loss[ctc_loss=0.1028, att_loss=0.2505, loss=0.221, over 3260330.50 frames. utt_duration=1212 frames, utt_pad_proportion=0.06882, over 10770.00 utterances.], batch size: 56, lr: 9.72e-03, grad_scale: 4.0 2023-03-08 05:37:11,205 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.765e+02 2.366e+02 2.941e+02 3.430e+02 6.532e+02, threshold=5.882e+02, percent-clipped=4.0 2023-03-08 05:37:37,561 INFO [train2.py:809] (1/4) Epoch 11, batch 3600, loss[ctc_loss=0.1256, att_loss=0.2649, loss=0.237, over 17419.00 frames. utt_duration=883.7 frames, utt_pad_proportion=0.07559, over 79.00 utterances.], tot_loss[ctc_loss=0.1025, att_loss=0.2502, loss=0.2207, over 3267746.45 frames. utt_duration=1218 frames, utt_pad_proportion=0.06421, over 10748.73 utterances.], batch size: 79, lr: 9.72e-03, grad_scale: 8.0 2023-03-08 05:38:03,415 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-03-08 05:38:58,672 INFO [train2.py:809] (1/4) Epoch 11, batch 3650, loss[ctc_loss=0.1239, att_loss=0.2676, loss=0.2389, over 14512.00 frames. utt_duration=393.7 frames, utt_pad_proportion=0.3045, over 148.00 utterances.], tot_loss[ctc_loss=0.1034, att_loss=0.2513, loss=0.2217, over 3273623.73 frames. utt_duration=1200 frames, utt_pad_proportion=0.06706, over 10922.32 utterances.], batch size: 148, lr: 9.71e-03, grad_scale: 8.0 2023-03-08 05:39:41,234 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43514.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:39:42,800 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43515.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:39:48,354 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-03-08 05:39:53,255 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.496e+02 2.334e+02 2.934e+02 3.698e+02 7.886e+02, threshold=5.869e+02, percent-clipped=6.0 2023-03-08 05:40:14,886 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43535.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:40:19,099 INFO [train2.py:809] (1/4) Epoch 11, batch 3700, loss[ctc_loss=0.1058, att_loss=0.2615, loss=0.2303, over 17106.00 frames. utt_duration=1223 frames, utt_pad_proportion=0.01572, over 56.00 utterances.], tot_loss[ctc_loss=0.1032, att_loss=0.251, loss=0.2214, over 3279619.15 frames. utt_duration=1206 frames, utt_pad_proportion=0.06286, over 10890.35 utterances.], batch size: 56, lr: 9.71e-03, grad_scale: 8.0 2023-03-08 05:40:59,607 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=43563.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:41:19,030 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43575.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:41:30,971 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=43583.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:41:38,577 INFO [train2.py:809] (1/4) Epoch 11, batch 3750, loss[ctc_loss=0.1056, att_loss=0.2644, loss=0.2326, over 17344.00 frames. utt_duration=1102 frames, utt_pad_proportion=0.03551, over 63.00 utterances.], tot_loss[ctc_loss=0.1027, att_loss=0.2511, loss=0.2215, over 3288514.74 frames. utt_duration=1229 frames, utt_pad_proportion=0.05479, over 10714.11 utterances.], batch size: 63, lr: 9.70e-03, grad_scale: 8.0 2023-03-08 05:42:11,820 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-03-08 05:42:32,237 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.570e+02 2.236e+02 2.829e+02 3.523e+02 7.340e+02, threshold=5.658e+02, percent-clipped=1.0 2023-03-08 05:42:58,804 INFO [train2.py:809] (1/4) Epoch 11, batch 3800, loss[ctc_loss=0.1413, att_loss=0.2763, loss=0.2493, over 17117.00 frames. utt_duration=1224 frames, utt_pad_proportion=0.01439, over 56.00 utterances.], tot_loss[ctc_loss=0.1025, att_loss=0.2509, loss=0.2212, over 3276249.56 frames. utt_duration=1238 frames, utt_pad_proportion=0.05566, over 10598.50 utterances.], batch size: 56, lr: 9.70e-03, grad_scale: 8.0 2023-03-08 05:44:18,444 INFO [train2.py:809] (1/4) Epoch 11, batch 3850, loss[ctc_loss=0.06825, att_loss=0.2157, loss=0.1862, over 15617.00 frames. utt_duration=1690 frames, utt_pad_proportion=0.009959, over 37.00 utterances.], tot_loss[ctc_loss=0.1018, att_loss=0.2507, loss=0.2209, over 3282823.08 frames. utt_duration=1244 frames, utt_pad_proportion=0.05239, over 10569.98 utterances.], batch size: 37, lr: 9.69e-03, grad_scale: 8.0 2023-03-08 05:44:44,651 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.4880, 3.7220, 3.3796, 3.0652, 3.6449, 3.6459, 3.4372, 2.3840], device='cuda:1'), covar=tensor([0.0984, 0.1491, 0.2568, 0.6300, 0.1113, 0.4455, 0.0956, 0.7577], device='cuda:1'), in_proj_covar=tensor([0.0104, 0.0126, 0.0137, 0.0208, 0.0107, 0.0192, 0.0112, 0.0181], device='cuda:1'), out_proj_covar=tensor([1.0024e-04, 1.1145e-04, 1.2335e-04, 1.7043e-04, 1.0059e-04, 1.5911e-04, 9.9224e-05, 1.4963e-04], device='cuda:1') 2023-03-08 05:45:10,392 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.366e+02 2.441e+02 3.005e+02 4.132e+02 9.940e+02, threshold=6.009e+02, percent-clipped=5.0 2023-03-08 05:45:35,984 INFO [train2.py:809] (1/4) Epoch 11, batch 3900, loss[ctc_loss=0.09196, att_loss=0.2348, loss=0.2062, over 15974.00 frames. utt_duration=1560 frames, utt_pad_proportion=0.005721, over 41.00 utterances.], tot_loss[ctc_loss=0.1015, att_loss=0.2502, loss=0.2205, over 3279928.69 frames. utt_duration=1262 frames, utt_pad_proportion=0.04914, over 10412.31 utterances.], batch size: 41, lr: 9.69e-03, grad_scale: 8.0 2023-03-08 05:46:39,729 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8717, 4.8013, 4.7888, 2.6119, 4.5838, 4.5124, 4.2778, 2.6108], device='cuda:1'), covar=tensor([0.0129, 0.0098, 0.0177, 0.1093, 0.0103, 0.0197, 0.0286, 0.1417], device='cuda:1'), in_proj_covar=tensor([0.0063, 0.0086, 0.0078, 0.0107, 0.0072, 0.0097, 0.0095, 0.0103], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 05:46:51,860 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43787.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:46:53,041 INFO [train2.py:809] (1/4) Epoch 11, batch 3950, loss[ctc_loss=0.1018, att_loss=0.2642, loss=0.2317, over 16483.00 frames. utt_duration=1435 frames, utt_pad_proportion=0.005619, over 46.00 utterances.], tot_loss[ctc_loss=0.1021, att_loss=0.2504, loss=0.2207, over 3275098.47 frames. utt_duration=1248 frames, utt_pad_proportion=0.05229, over 10508.61 utterances.], batch size: 46, lr: 9.68e-03, grad_scale: 8.0 2023-03-08 05:47:05,594 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.4788, 3.6434, 3.3753, 2.9895, 3.5260, 3.6282, 3.3924, 2.4028], device='cuda:1'), covar=tensor([0.1169, 0.1575, 0.2537, 0.5962, 0.1955, 0.3089, 0.1126, 0.7480], device='cuda:1'), in_proj_covar=tensor([0.0103, 0.0125, 0.0137, 0.0206, 0.0107, 0.0190, 0.0110, 0.0178], device='cuda:1'), out_proj_covar=tensor([9.9713e-05, 1.1060e-04, 1.2301e-04, 1.6852e-04, 1.0008e-04, 1.5784e-04, 9.8297e-05, 1.4762e-04], device='cuda:1') 2023-03-08 05:48:13,303 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.357e+02 2.409e+02 2.914e+02 3.960e+02 1.049e+03, threshold=5.829e+02, percent-clipped=5.0 2023-03-08 05:48:13,361 INFO [train2.py:809] (1/4) Epoch 12, batch 0, loss[ctc_loss=0.07865, att_loss=0.2192, loss=0.1911, over 15472.00 frames. utt_duration=1720 frames, utt_pad_proportion=0.01018, over 36.00 utterances.], tot_loss[ctc_loss=0.07865, att_loss=0.2192, loss=0.1911, over 15472.00 frames. utt_duration=1720 frames, utt_pad_proportion=0.01018, over 36.00 utterances.], batch size: 36, lr: 9.27e-03, grad_scale: 8.0 2023-03-08 05:48:13,362 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-08 05:48:17,054 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3902, 2.8828, 3.5920, 2.9826, 3.4851, 4.4298, 4.2137, 3.4225], device='cuda:1'), covar=tensor([0.0317, 0.1657, 0.1095, 0.1389, 0.1021, 0.0815, 0.0534, 0.1221], device='cuda:1'), in_proj_covar=tensor([0.0226, 0.0225, 0.0245, 0.0203, 0.0235, 0.0301, 0.0215, 0.0219], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 05:48:25,650 INFO [train2.py:843] (1/4) Epoch 12, validation: ctc_loss=0.04785, att_loss=0.2384, loss=0.2003, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-08 05:48:25,651 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-08 05:48:38,275 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.33 vs. limit=5.0 2023-03-08 05:49:09,125 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43848.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:49:12,383 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.8222, 3.6766, 3.0014, 3.1613, 3.8658, 3.5534, 2.6122, 4.2137], device='cuda:1'), covar=tensor([0.1188, 0.0514, 0.1130, 0.0832, 0.0721, 0.0663, 0.1020, 0.0499], device='cuda:1'), in_proj_covar=tensor([0.0183, 0.0182, 0.0202, 0.0173, 0.0236, 0.0208, 0.0181, 0.0248], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-08 05:49:45,197 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43870.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:49:47,997 INFO [train2.py:809] (1/4) Epoch 12, batch 50, loss[ctc_loss=0.109, att_loss=0.2605, loss=0.2302, over 17315.00 frames. utt_duration=1261 frames, utt_pad_proportion=0.01055, over 55.00 utterances.], tot_loss[ctc_loss=0.09844, att_loss=0.2486, loss=0.2186, over 742990.97 frames. utt_duration=1265 frames, utt_pad_proportion=0.04329, over 2352.25 utterances.], batch size: 55, lr: 9.26e-03, grad_scale: 8.0 2023-03-08 05:50:48,042 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43909.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:50:52,791 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.0789, 3.9365, 3.2251, 3.5395, 4.0235, 3.6129, 2.7866, 4.4423], device='cuda:1'), covar=tensor([0.0935, 0.0389, 0.1014, 0.0623, 0.0613, 0.0625, 0.0936, 0.0380], device='cuda:1'), in_proj_covar=tensor([0.0182, 0.0182, 0.0201, 0.0172, 0.0236, 0.0207, 0.0181, 0.0247], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-08 05:51:09,635 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.511e+02 2.447e+02 2.836e+02 3.349e+02 4.947e+02, threshold=5.672e+02, percent-clipped=0.0 2023-03-08 05:51:09,679 INFO [train2.py:809] (1/4) Epoch 12, batch 100, loss[ctc_loss=0.08271, att_loss=0.2254, loss=0.1969, over 15884.00 frames. utt_duration=1631 frames, utt_pad_proportion=0.009393, over 39.00 utterances.], tot_loss[ctc_loss=0.09896, att_loss=0.2499, loss=0.2197, over 1302082.72 frames. utt_duration=1259 frames, utt_pad_proportion=0.04964, over 4141.56 utterances.], batch size: 39, lr: 9.26e-03, grad_scale: 8.0 2023-03-08 05:52:09,325 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43959.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:52:28,387 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43970.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:52:31,100 INFO [train2.py:809] (1/4) Epoch 12, batch 150, loss[ctc_loss=0.1215, att_loss=0.2691, loss=0.2396, over 16469.00 frames. utt_duration=1433 frames, utt_pad_proportion=0.006674, over 46.00 utterances.], tot_loss[ctc_loss=0.09934, att_loss=0.2499, loss=0.2198, over 1735180.17 frames. utt_duration=1225 frames, utt_pad_proportion=0.05783, over 5674.50 utterances.], batch size: 46, lr: 9.25e-03, grad_scale: 8.0 2023-03-08 05:52:40,879 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2431, 5.3115, 5.0065, 2.6409, 2.2135, 2.8239, 3.4793, 3.9867], device='cuda:1'), covar=tensor([0.0633, 0.0282, 0.0264, 0.4167, 0.5780, 0.2602, 0.1812, 0.1736], device='cuda:1'), in_proj_covar=tensor([0.0337, 0.0227, 0.0243, 0.0214, 0.0351, 0.0337, 0.0231, 0.0358], device='cuda:1'), out_proj_covar=tensor([1.5087e-04, 8.3629e-05, 1.0381e-04, 9.5496e-05, 1.5257e-04, 1.3638e-04, 9.0665e-05, 1.5074e-04], device='cuda:1') 2023-03-08 05:53:40,895 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2859, 3.9942, 3.3068, 3.6646, 4.1311, 3.8457, 2.9692, 4.4971], device='cuda:1'), covar=tensor([0.0848, 0.0416, 0.1032, 0.0636, 0.0626, 0.0551, 0.0823, 0.0425], device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0186, 0.0204, 0.0175, 0.0239, 0.0211, 0.0184, 0.0253], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-08 05:53:52,669 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=44020.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:53:55,400 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.396e+02 2.379e+02 2.873e+02 3.449e+02 9.028e+02, threshold=5.746e+02, percent-clipped=3.0 2023-03-08 05:53:55,443 INFO [train2.py:809] (1/4) Epoch 12, batch 200, loss[ctc_loss=0.08817, att_loss=0.2331, loss=0.2041, over 15778.00 frames. utt_duration=1662 frames, utt_pad_proportion=0.007494, over 38.00 utterances.], tot_loss[ctc_loss=0.0986, att_loss=0.2478, loss=0.2179, over 2068096.81 frames. utt_duration=1244 frames, utt_pad_proportion=0.05682, over 6658.55 utterances.], batch size: 38, lr: 9.25e-03, grad_scale: 8.0 2023-03-08 05:54:37,376 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0813, 4.5568, 4.4891, 4.6354, 2.7263, 4.3251, 2.4468, 2.0466], device='cuda:1'), covar=tensor([0.0357, 0.0129, 0.0626, 0.0147, 0.1656, 0.0158, 0.1567, 0.1580], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0113, 0.0256, 0.0111, 0.0218, 0.0108, 0.0221, 0.0199], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 05:55:14,812 INFO [train2.py:809] (1/4) Epoch 12, batch 250, loss[ctc_loss=0.09471, att_loss=0.2631, loss=0.2294, over 17062.00 frames. utt_duration=1314 frames, utt_pad_proportion=0.008316, over 52.00 utterances.], tot_loss[ctc_loss=0.09871, att_loss=0.2491, loss=0.219, over 2345733.36 frames. utt_duration=1252 frames, utt_pad_proportion=0.05091, over 7504.81 utterances.], batch size: 52, lr: 9.24e-03, grad_scale: 8.0 2023-03-08 05:55:21,585 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2611, 5.2990, 5.0050, 2.4901, 2.0972, 2.9370, 3.1988, 3.9801], device='cuda:1'), covar=tensor([0.0545, 0.0258, 0.0264, 0.4248, 0.5840, 0.2415, 0.2116, 0.1782], device='cuda:1'), in_proj_covar=tensor([0.0332, 0.0224, 0.0240, 0.0212, 0.0347, 0.0334, 0.0229, 0.0353], device='cuda:1'), out_proj_covar=tensor([1.4860e-04, 8.2703e-05, 1.0301e-04, 9.4142e-05, 1.5109e-04, 1.3503e-04, 9.0197e-05, 1.4926e-04], device='cuda:1') 2023-03-08 05:55:54,449 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.22 vs. limit=5.0 2023-03-08 05:56:35,433 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.521e+02 2.301e+02 2.863e+02 3.737e+02 8.031e+02, threshold=5.726e+02, percent-clipped=4.0 2023-03-08 05:56:35,477 INFO [train2.py:809] (1/4) Epoch 12, batch 300, loss[ctc_loss=0.1188, att_loss=0.2538, loss=0.2268, over 16529.00 frames. utt_duration=1471 frames, utt_pad_proportion=0.006737, over 45.00 utterances.], tot_loss[ctc_loss=0.099, att_loss=0.249, loss=0.219, over 2555729.40 frames. utt_duration=1276 frames, utt_pad_proportion=0.0443, over 8024.05 utterances.], batch size: 45, lr: 9.24e-03, grad_scale: 8.0 2023-03-08 05:57:10,204 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44143.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:57:32,705 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-03-08 05:57:53,920 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44170.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:57:56,718 INFO [train2.py:809] (1/4) Epoch 12, batch 350, loss[ctc_loss=0.1023, att_loss=0.2666, loss=0.2337, over 17076.00 frames. utt_duration=1315 frames, utt_pad_proportion=0.007678, over 52.00 utterances.], tot_loss[ctc_loss=0.09776, att_loss=0.2482, loss=0.2181, over 2717141.86 frames. utt_duration=1312 frames, utt_pad_proportion=0.03645, over 8295.91 utterances.], batch size: 52, lr: 9.23e-03, grad_scale: 8.0 2023-03-08 05:59:10,309 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=44218.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 05:59:12,118 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2143, 3.8495, 3.1684, 3.5611, 3.9338, 3.6237, 2.8555, 4.3937], device='cuda:1'), covar=tensor([0.0960, 0.0487, 0.1148, 0.0622, 0.0710, 0.0700, 0.0897, 0.0441], device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0185, 0.0205, 0.0175, 0.0241, 0.0214, 0.0184, 0.0254], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-08 05:59:17,106 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.518e+02 2.261e+02 2.765e+02 3.364e+02 1.191e+03, threshold=5.531e+02, percent-clipped=3.0 2023-03-08 05:59:17,151 INFO [train2.py:809] (1/4) Epoch 12, batch 400, loss[ctc_loss=0.07446, att_loss=0.2432, loss=0.2095, over 16487.00 frames. utt_duration=1435 frames, utt_pad_proportion=0.006218, over 46.00 utterances.], tot_loss[ctc_loss=0.0972, att_loss=0.2477, loss=0.2176, over 2834395.08 frames. utt_duration=1323 frames, utt_pad_proportion=0.03448, over 8575.87 utterances.], batch size: 46, lr: 9.23e-03, grad_scale: 8.0 2023-03-08 06:00:00,166 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0893, 5.1753, 4.9435, 2.1683, 1.9009, 2.8355, 2.9245, 3.7891], device='cuda:1'), covar=tensor([0.0614, 0.0191, 0.0208, 0.5146, 0.6019, 0.2440, 0.2335, 0.1763], device='cuda:1'), in_proj_covar=tensor([0.0331, 0.0222, 0.0239, 0.0212, 0.0345, 0.0333, 0.0230, 0.0353], device='cuda:1'), out_proj_covar=tensor([1.4849e-04, 8.1967e-05, 1.0252e-04, 9.4315e-05, 1.5035e-04, 1.3443e-04, 9.0390e-05, 1.4885e-04], device='cuda:1') 2023-03-08 06:00:24,785 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44265.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:00:36,685 INFO [train2.py:809] (1/4) Epoch 12, batch 450, loss[ctc_loss=0.1131, att_loss=0.2526, loss=0.2247, over 16475.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.006116, over 46.00 utterances.], tot_loss[ctc_loss=0.09802, att_loss=0.2477, loss=0.2178, over 2936393.65 frames. utt_duration=1339 frames, utt_pad_proportion=0.03026, over 8778.63 utterances.], batch size: 46, lr: 9.22e-03, grad_scale: 8.0 2023-03-08 06:01:44,250 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44315.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:01:56,076 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.632e+02 2.333e+02 2.897e+02 3.791e+02 1.244e+03, threshold=5.793e+02, percent-clipped=7.0 2023-03-08 06:01:56,120 INFO [train2.py:809] (1/4) Epoch 12, batch 500, loss[ctc_loss=0.108, att_loss=0.2373, loss=0.2114, over 14602.00 frames. utt_duration=1827 frames, utt_pad_proportion=0.0305, over 32.00 utterances.], tot_loss[ctc_loss=0.09935, att_loss=0.2487, loss=0.2188, over 3009083.84 frames. utt_duration=1298 frames, utt_pad_proportion=0.04286, over 9284.11 utterances.], batch size: 32, lr: 9.22e-03, grad_scale: 8.0 2023-03-08 06:02:19,705 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-03-08 06:03:15,766 INFO [train2.py:809] (1/4) Epoch 12, batch 550, loss[ctc_loss=0.09217, att_loss=0.2614, loss=0.2276, over 17009.00 frames. utt_duration=1285 frames, utt_pad_proportion=0.01228, over 53.00 utterances.], tot_loss[ctc_loss=0.09902, att_loss=0.2488, loss=0.2189, over 3074386.77 frames. utt_duration=1295 frames, utt_pad_proportion=0.04043, over 9504.95 utterances.], batch size: 53, lr: 9.21e-03, grad_scale: 8.0 2023-03-08 06:03:44,163 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.53 vs. limit=5.0 2023-03-08 06:03:54,687 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.71 vs. limit=2.0 2023-03-08 06:04:34,992 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.625e+02 2.230e+02 2.908e+02 3.570e+02 8.469e+02, threshold=5.817e+02, percent-clipped=3.0 2023-03-08 06:04:35,036 INFO [train2.py:809] (1/4) Epoch 12, batch 600, loss[ctc_loss=0.08664, att_loss=0.2295, loss=0.201, over 15381.00 frames. utt_duration=1759 frames, utt_pad_proportion=0.009459, over 35.00 utterances.], tot_loss[ctc_loss=0.09874, att_loss=0.2484, loss=0.2185, over 3114701.54 frames. utt_duration=1293 frames, utt_pad_proportion=0.04282, over 9645.05 utterances.], batch size: 35, lr: 9.21e-03, grad_scale: 8.0 2023-03-08 06:05:08,634 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44443.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:05:45,772 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.6912, 3.7726, 3.5633, 3.0904, 3.6442, 3.7766, 3.5521, 2.5911], device='cuda:1'), covar=tensor([0.1204, 0.1633, 0.2349, 0.5673, 0.2633, 0.3724, 0.1023, 0.7852], device='cuda:1'), in_proj_covar=tensor([0.0100, 0.0121, 0.0133, 0.0200, 0.0105, 0.0184, 0.0108, 0.0173], device='cuda:1'), out_proj_covar=tensor([9.7020e-05, 1.0770e-04, 1.1990e-04, 1.6381e-04, 9.8628e-05, 1.5315e-04, 9.6142e-05, 1.4374e-04], device='cuda:1') 2023-03-08 06:05:47,255 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9831, 5.1490, 4.9219, 2.6116, 4.9285, 4.7176, 4.2001, 2.5950], device='cuda:1'), covar=tensor([0.0142, 0.0083, 0.0254, 0.1314, 0.0085, 0.0188, 0.0389, 0.1675], device='cuda:1'), in_proj_covar=tensor([0.0062, 0.0084, 0.0078, 0.0106, 0.0072, 0.0096, 0.0095, 0.0101], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 06:05:53,594 INFO [train2.py:809] (1/4) Epoch 12, batch 650, loss[ctc_loss=0.09514, att_loss=0.2568, loss=0.2245, over 16999.00 frames. utt_duration=1335 frames, utt_pad_proportion=0.008821, over 51.00 utterances.], tot_loss[ctc_loss=0.09896, att_loss=0.2483, loss=0.2184, over 3150269.83 frames. utt_duration=1279 frames, utt_pad_proportion=0.04669, over 9865.86 utterances.], batch size: 51, lr: 9.20e-03, grad_scale: 8.0 2023-03-08 06:05:58,458 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1853, 5.2417, 5.1398, 3.0190, 5.0245, 4.7808, 4.6927, 3.1068], device='cuda:1'), covar=tensor([0.0095, 0.0060, 0.0190, 0.0924, 0.0072, 0.0155, 0.0239, 0.1150], device='cuda:1'), in_proj_covar=tensor([0.0062, 0.0084, 0.0078, 0.0106, 0.0072, 0.0096, 0.0095, 0.0101], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 06:06:23,476 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=44491.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:06:51,569 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-03-08 06:07:07,774 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9118, 5.2544, 4.7212, 5.2924, 4.6211, 4.9654, 5.3519, 5.1169], device='cuda:1'), covar=tensor([0.0513, 0.0267, 0.0823, 0.0234, 0.0459, 0.0216, 0.0198, 0.0171], device='cuda:1'), in_proj_covar=tensor([0.0336, 0.0262, 0.0316, 0.0259, 0.0269, 0.0205, 0.0250, 0.0239], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 06:07:12,356 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.481e+02 2.522e+02 3.198e+02 3.829e+02 8.235e+02, threshold=6.395e+02, percent-clipped=2.0 2023-03-08 06:07:12,400 INFO [train2.py:809] (1/4) Epoch 12, batch 700, loss[ctc_loss=0.1038, att_loss=0.2552, loss=0.2249, over 16272.00 frames. utt_duration=1515 frames, utt_pad_proportion=0.007661, over 43.00 utterances.], tot_loss[ctc_loss=0.09961, att_loss=0.2486, loss=0.2188, over 3181786.98 frames. utt_duration=1274 frames, utt_pad_proportion=0.04684, over 10001.87 utterances.], batch size: 43, lr: 9.20e-03, grad_scale: 8.0 2023-03-08 06:07:20,664 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=44527.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:07:57,880 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=44550.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:08:21,771 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44565.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:08:32,510 INFO [train2.py:809] (1/4) Epoch 12, batch 750, loss[ctc_loss=0.1014, att_loss=0.2375, loss=0.2102, over 16283.00 frames. utt_duration=1516 frames, utt_pad_proportion=0.007341, over 43.00 utterances.], tot_loss[ctc_loss=0.1, att_loss=0.2494, loss=0.2195, over 3203557.97 frames. utt_duration=1262 frames, utt_pad_proportion=0.04956, over 10162.41 utterances.], batch size: 43, lr: 9.19e-03, grad_scale: 8.0 2023-03-08 06:08:58,581 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=44588.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:09:34,894 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=44611.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:09:39,010 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=44613.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:09:42,274 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44615.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:09:48,336 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9771, 5.3777, 4.8714, 5.4444, 4.7659, 5.1044, 5.5043, 5.2525], device='cuda:1'), covar=tensor([0.0524, 0.0240, 0.0796, 0.0243, 0.0395, 0.0207, 0.0228, 0.0182], device='cuda:1'), in_proj_covar=tensor([0.0337, 0.0264, 0.0314, 0.0259, 0.0270, 0.0206, 0.0249, 0.0241], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 06:09:52,668 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.593e+02 2.480e+02 2.940e+02 3.562e+02 6.927e+02, threshold=5.880e+02, percent-clipped=2.0 2023-03-08 06:09:52,713 INFO [train2.py:809] (1/4) Epoch 12, batch 800, loss[ctc_loss=0.1037, att_loss=0.2275, loss=0.2027, over 15342.00 frames. utt_duration=1755 frames, utt_pad_proportion=0.01027, over 35.00 utterances.], tot_loss[ctc_loss=0.1003, att_loss=0.2493, loss=0.2195, over 3212099.48 frames. utt_duration=1241 frames, utt_pad_proportion=0.05781, over 10369.69 utterances.], batch size: 35, lr: 9.19e-03, grad_scale: 8.0 2023-03-08 06:10:01,243 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5419, 4.7855, 4.8150, 4.7481, 4.8600, 4.8373, 4.5852, 4.3649], device='cuda:1'), covar=tensor([0.1044, 0.0528, 0.0244, 0.0547, 0.0290, 0.0328, 0.0314, 0.0349], device='cuda:1'), in_proj_covar=tensor([0.0463, 0.0295, 0.0254, 0.0292, 0.0352, 0.0362, 0.0292, 0.0328], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 06:10:57,837 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=44663.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:11:08,001 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2685, 2.2912, 3.3737, 2.5389, 3.2513, 4.5036, 4.4060, 2.5128], device='cuda:1'), covar=tensor([0.0588, 0.2738, 0.1265, 0.1999, 0.1213, 0.0770, 0.0446, 0.2468], device='cuda:1'), in_proj_covar=tensor([0.0228, 0.0226, 0.0249, 0.0203, 0.0238, 0.0302, 0.0217, 0.0220], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 06:11:12,233 INFO [train2.py:809] (1/4) Epoch 12, batch 850, loss[ctc_loss=0.08334, att_loss=0.2455, loss=0.2131, over 16620.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.005558, over 47.00 utterances.], tot_loss[ctc_loss=0.1001, att_loss=0.2489, loss=0.2192, over 3225832.76 frames. utt_duration=1252 frames, utt_pad_proportion=0.05412, over 10315.13 utterances.], batch size: 47, lr: 9.18e-03, grad_scale: 8.0 2023-03-08 06:11:31,209 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-03-08 06:12:30,574 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.599e+02 2.371e+02 2.719e+02 3.452e+02 7.248e+02, threshold=5.437e+02, percent-clipped=1.0 2023-03-08 06:12:30,618 INFO [train2.py:809] (1/4) Epoch 12, batch 900, loss[ctc_loss=0.08684, att_loss=0.2266, loss=0.1986, over 16001.00 frames. utt_duration=1602 frames, utt_pad_proportion=0.007698, over 40.00 utterances.], tot_loss[ctc_loss=0.09947, att_loss=0.2481, loss=0.2184, over 3231715.88 frames. utt_duration=1279 frames, utt_pad_proportion=0.04928, over 10120.46 utterances.], batch size: 40, lr: 9.18e-03, grad_scale: 8.0 2023-03-08 06:12:54,828 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.3928, 2.6831, 3.2423, 4.2851, 3.9612, 3.9868, 3.0989, 2.1745], device='cuda:1'), covar=tensor([0.0764, 0.2291, 0.1157, 0.0639, 0.0696, 0.0394, 0.1369, 0.2446], device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0206, 0.0187, 0.0195, 0.0189, 0.0153, 0.0196, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 06:13:49,827 INFO [train2.py:809] (1/4) Epoch 12, batch 950, loss[ctc_loss=0.1261, att_loss=0.2745, loss=0.2448, over 17106.00 frames. utt_duration=1223 frames, utt_pad_proportion=0.01426, over 56.00 utterances.], tot_loss[ctc_loss=0.09957, att_loss=0.2483, loss=0.2186, over 3244106.00 frames. utt_duration=1281 frames, utt_pad_proportion=0.04745, over 10142.94 utterances.], batch size: 56, lr: 9.17e-03, grad_scale: 8.0 2023-03-08 06:14:20,526 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8317, 5.1308, 5.1395, 5.0856, 5.1840, 5.1429, 4.8750, 4.6026], device='cuda:1'), covar=tensor([0.1059, 0.0506, 0.0259, 0.0482, 0.0303, 0.0305, 0.0350, 0.0348], device='cuda:1'), in_proj_covar=tensor([0.0456, 0.0293, 0.0254, 0.0288, 0.0351, 0.0364, 0.0288, 0.0325], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 06:15:09,798 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.510e+02 2.480e+02 3.005e+02 3.455e+02 6.503e+02, threshold=6.009e+02, percent-clipped=2.0 2023-03-08 06:15:09,842 INFO [train2.py:809] (1/4) Epoch 12, batch 1000, loss[ctc_loss=0.08279, att_loss=0.2272, loss=0.1983, over 15934.00 frames. utt_duration=1556 frames, utt_pad_proportion=0.007066, over 41.00 utterances.], tot_loss[ctc_loss=0.09947, att_loss=0.2491, loss=0.2192, over 3258192.89 frames. utt_duration=1266 frames, utt_pad_proportion=0.04809, over 10304.35 utterances.], batch size: 41, lr: 9.17e-03, grad_scale: 8.0 2023-03-08 06:16:28,097 INFO [train2.py:809] (1/4) Epoch 12, batch 1050, loss[ctc_loss=0.1285, att_loss=0.269, loss=0.2409, over 17391.00 frames. utt_duration=1010 frames, utt_pad_proportion=0.04657, over 69.00 utterances.], tot_loss[ctc_loss=0.09931, att_loss=0.2477, loss=0.218, over 3248703.58 frames. utt_duration=1277 frames, utt_pad_proportion=0.04981, over 10191.35 utterances.], batch size: 69, lr: 9.16e-03, grad_scale: 8.0 2023-03-08 06:16:46,083 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44883.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:17:22,293 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44906.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:17:47,209 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.611e+02 2.358e+02 3.023e+02 3.796e+02 1.133e+03, threshold=6.045e+02, percent-clipped=5.0 2023-03-08 06:17:47,253 INFO [train2.py:809] (1/4) Epoch 12, batch 1100, loss[ctc_loss=0.08395, att_loss=0.2169, loss=0.1903, over 15767.00 frames. utt_duration=1661 frames, utt_pad_proportion=0.008871, over 38.00 utterances.], tot_loss[ctc_loss=0.1005, att_loss=0.2484, loss=0.2189, over 3258094.08 frames. utt_duration=1245 frames, utt_pad_proportion=0.05593, over 10481.35 utterances.], batch size: 38, lr: 9.16e-03, grad_scale: 8.0 2023-03-08 06:18:06,352 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.60 vs. limit=5.0 2023-03-08 06:18:42,127 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.7654, 2.9394, 3.7546, 4.3814, 4.2138, 4.0996, 3.0530, 2.4844], device='cuda:1'), covar=tensor([0.0505, 0.2011, 0.0800, 0.0617, 0.0615, 0.0356, 0.1371, 0.2004], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0209, 0.0187, 0.0196, 0.0188, 0.0153, 0.0196, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 06:19:06,840 INFO [train2.py:809] (1/4) Epoch 12, batch 1150, loss[ctc_loss=0.1059, att_loss=0.2657, loss=0.2338, over 17145.00 frames. utt_duration=1226 frames, utt_pad_proportion=0.01275, over 56.00 utterances.], tot_loss[ctc_loss=0.1, att_loss=0.2484, loss=0.2187, over 3260430.47 frames. utt_duration=1247 frames, utt_pad_proportion=0.05623, over 10473.37 utterances.], batch size: 56, lr: 9.15e-03, grad_scale: 8.0 2023-03-08 06:19:22,251 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.7305, 3.8787, 3.9587, 2.4318, 2.4347, 2.8803, 2.6208, 3.5360], device='cuda:1'), covar=tensor([0.0693, 0.0321, 0.0331, 0.3161, 0.4187, 0.1979, 0.1927, 0.1365], device='cuda:1'), in_proj_covar=tensor([0.0342, 0.0228, 0.0244, 0.0217, 0.0354, 0.0342, 0.0235, 0.0360], device='cuda:1'), out_proj_covar=tensor([1.5269e-04, 8.4462e-05, 1.0492e-04, 9.6734e-05, 1.5390e-04, 1.3804e-04, 9.2455e-05, 1.5150e-04], device='cuda:1') 2023-03-08 06:20:27,105 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.801e+02 2.531e+02 3.116e+02 3.796e+02 1.058e+03, threshold=6.232e+02, percent-clipped=4.0 2023-03-08 06:20:27,148 INFO [train2.py:809] (1/4) Epoch 12, batch 1200, loss[ctc_loss=0.09417, att_loss=0.2522, loss=0.2206, over 17314.00 frames. utt_duration=1261 frames, utt_pad_proportion=0.01036, over 55.00 utterances.], tot_loss[ctc_loss=0.1005, att_loss=0.2485, loss=0.2189, over 3263363.47 frames. utt_duration=1239 frames, utt_pad_proportion=0.05857, over 10552.02 utterances.], batch size: 55, lr: 9.15e-03, grad_scale: 8.0 2023-03-08 06:21:17,085 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.05 vs. limit=5.0 2023-03-08 06:21:22,290 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45056.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:21:47,504 INFO [train2.py:809] (1/4) Epoch 12, batch 1250, loss[ctc_loss=0.1359, att_loss=0.2757, loss=0.2477, over 14722.00 frames. utt_duration=405 frames, utt_pad_proportion=0.2944, over 146.00 utterances.], tot_loss[ctc_loss=0.1014, att_loss=0.2503, loss=0.2205, over 3281647.49 frames. utt_duration=1220 frames, utt_pad_proportion=0.058, over 10771.62 utterances.], batch size: 146, lr: 9.14e-03, grad_scale: 8.0 2023-03-08 06:22:02,260 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45081.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:22:58,404 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45117.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 06:23:01,957 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5388, 4.4383, 4.5888, 4.5642, 5.1274, 4.6999, 4.5983, 2.3971], device='cuda:1'), covar=tensor([0.0193, 0.0335, 0.0229, 0.0240, 0.0795, 0.0167, 0.0230, 0.2041], device='cuda:1'), in_proj_covar=tensor([0.0126, 0.0134, 0.0137, 0.0147, 0.0334, 0.0121, 0.0122, 0.0215], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 06:23:05,869 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.644e+02 2.248e+02 2.874e+02 3.449e+02 8.124e+02, threshold=5.748e+02, percent-clipped=2.0 2023-03-08 06:23:05,913 INFO [train2.py:809] (1/4) Epoch 12, batch 1300, loss[ctc_loss=0.1375, att_loss=0.2745, loss=0.2471, over 17461.00 frames. utt_duration=1014 frames, utt_pad_proportion=0.04094, over 69.00 utterances.], tot_loss[ctc_loss=0.1021, att_loss=0.2502, loss=0.2206, over 3269219.76 frames. utt_duration=1198 frames, utt_pad_proportion=0.06723, over 10927.14 utterances.], batch size: 69, lr: 9.14e-03, grad_scale: 8.0 2023-03-08 06:23:37,486 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45142.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:24:24,487 INFO [train2.py:809] (1/4) Epoch 12, batch 1350, loss[ctc_loss=0.1186, att_loss=0.2586, loss=0.2306, over 16198.00 frames. utt_duration=1582 frames, utt_pad_proportion=0.005795, over 41.00 utterances.], tot_loss[ctc_loss=0.1014, att_loss=0.2499, loss=0.2202, over 3277191.08 frames. utt_duration=1211 frames, utt_pad_proportion=0.06191, over 10836.29 utterances.], batch size: 41, lr: 9.13e-03, grad_scale: 8.0 2023-03-08 06:24:42,122 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45183.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:25:18,424 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45206.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:25:43,526 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.529e+02 2.386e+02 2.782e+02 3.300e+02 6.204e+02, threshold=5.564e+02, percent-clipped=3.0 2023-03-08 06:25:43,571 INFO [train2.py:809] (1/4) Epoch 12, batch 1400, loss[ctc_loss=0.1153, att_loss=0.2729, loss=0.2414, over 17056.00 frames. utt_duration=1314 frames, utt_pad_proportion=0.00865, over 52.00 utterances.], tot_loss[ctc_loss=0.1019, att_loss=0.2501, loss=0.2205, over 3279337.46 frames. utt_duration=1209 frames, utt_pad_proportion=0.06114, over 10860.28 utterances.], batch size: 52, lr: 9.13e-03, grad_scale: 8.0 2023-03-08 06:25:58,482 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=45231.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:26:34,488 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=45254.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:26:40,800 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-03-08 06:27:04,203 INFO [train2.py:809] (1/4) Epoch 12, batch 1450, loss[ctc_loss=0.1127, att_loss=0.2658, loss=0.2352, over 17289.00 frames. utt_duration=1259 frames, utt_pad_proportion=0.01273, over 55.00 utterances.], tot_loss[ctc_loss=0.1017, att_loss=0.2502, loss=0.2205, over 3279700.81 frames. utt_duration=1232 frames, utt_pad_proportion=0.05446, over 10656.95 utterances.], batch size: 55, lr: 9.12e-03, grad_scale: 8.0 2023-03-08 06:27:13,500 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-03-08 06:27:35,415 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.92 vs. limit=2.0 2023-03-08 06:28:24,614 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.115e+02 2.356e+02 2.763e+02 3.457e+02 6.962e+02, threshold=5.526e+02, percent-clipped=5.0 2023-03-08 06:28:24,657 INFO [train2.py:809] (1/4) Epoch 12, batch 1500, loss[ctc_loss=0.09617, att_loss=0.2309, loss=0.204, over 15631.00 frames. utt_duration=1691 frames, utt_pad_proportion=0.009684, over 37.00 utterances.], tot_loss[ctc_loss=0.1006, att_loss=0.2492, loss=0.2195, over 3279372.71 frames. utt_duration=1247 frames, utt_pad_proportion=0.05138, over 10528.79 utterances.], batch size: 37, lr: 9.12e-03, grad_scale: 8.0 2023-03-08 06:29:06,057 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45348.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 06:29:45,020 INFO [train2.py:809] (1/4) Epoch 12, batch 1550, loss[ctc_loss=0.09981, att_loss=0.2645, loss=0.2315, over 16964.00 frames. utt_duration=1359 frames, utt_pad_proportion=0.007421, over 50.00 utterances.], tot_loss[ctc_loss=0.1004, att_loss=0.2493, loss=0.2195, over 3275977.66 frames. utt_duration=1240 frames, utt_pad_proportion=0.05356, over 10578.91 utterances.], batch size: 50, lr: 9.11e-03, grad_scale: 8.0 2023-03-08 06:30:44,802 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45409.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 06:30:49,196 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45412.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 06:31:05,228 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.391e+02 2.353e+02 2.825e+02 3.392e+02 6.329e+02, threshold=5.650e+02, percent-clipped=2.0 2023-03-08 06:31:05,272 INFO [train2.py:809] (1/4) Epoch 12, batch 1600, loss[ctc_loss=0.08525, att_loss=0.2557, loss=0.2216, over 17037.00 frames. utt_duration=1287 frames, utt_pad_proportion=0.01056, over 53.00 utterances.], tot_loss[ctc_loss=0.1004, att_loss=0.2497, loss=0.2198, over 3284380.01 frames. utt_duration=1246 frames, utt_pad_proportion=0.05042, over 10553.10 utterances.], batch size: 53, lr: 9.11e-03, grad_scale: 16.0 2023-03-08 06:31:28,625 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45437.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:32:24,618 INFO [train2.py:809] (1/4) Epoch 12, batch 1650, loss[ctc_loss=0.08997, att_loss=0.2323, loss=0.2038, over 15653.00 frames. utt_duration=1694 frames, utt_pad_proportion=0.007774, over 37.00 utterances.], tot_loss[ctc_loss=0.101, att_loss=0.2491, loss=0.2195, over 3266512.92 frames. utt_duration=1215 frames, utt_pad_proportion=0.06331, over 10767.39 utterances.], batch size: 37, lr: 9.10e-03, grad_scale: 16.0 2023-03-08 06:33:43,905 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.495e+02 2.369e+02 2.877e+02 3.430e+02 6.900e+02, threshold=5.754e+02, percent-clipped=3.0 2023-03-08 06:33:43,949 INFO [train2.py:809] (1/4) Epoch 12, batch 1700, loss[ctc_loss=0.1231, att_loss=0.262, loss=0.2342, over 17296.00 frames. utt_duration=1259 frames, utt_pad_proportion=0.01222, over 55.00 utterances.], tot_loss[ctc_loss=0.101, att_loss=0.2496, loss=0.2199, over 3273113.61 frames. utt_duration=1227 frames, utt_pad_proportion=0.05873, over 10685.93 utterances.], batch size: 55, lr: 9.10e-03, grad_scale: 16.0 2023-03-08 06:34:49,093 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.24 vs. limit=5.0 2023-03-08 06:34:57,099 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.84 vs. limit=2.0 2023-03-08 06:35:04,019 INFO [train2.py:809] (1/4) Epoch 12, batch 1750, loss[ctc_loss=0.08893, att_loss=0.2237, loss=0.1967, over 15875.00 frames. utt_duration=1630 frames, utt_pad_proportion=0.009056, over 39.00 utterances.], tot_loss[ctc_loss=0.1009, att_loss=0.2495, loss=0.2198, over 3270823.26 frames. utt_duration=1203 frames, utt_pad_proportion=0.06588, over 10887.62 utterances.], batch size: 39, lr: 9.09e-03, grad_scale: 16.0 2023-03-08 06:36:24,188 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.564e+02 2.369e+02 2.874e+02 3.424e+02 1.036e+03, threshold=5.747e+02, percent-clipped=2.0 2023-03-08 06:36:24,232 INFO [train2.py:809] (1/4) Epoch 12, batch 1800, loss[ctc_loss=0.08416, att_loss=0.2274, loss=0.1988, over 15955.00 frames. utt_duration=1558 frames, utt_pad_proportion=0.006191, over 41.00 utterances.], tot_loss[ctc_loss=0.09977, att_loss=0.2489, loss=0.2191, over 3266946.71 frames. utt_duration=1193 frames, utt_pad_proportion=0.07057, over 10967.86 utterances.], batch size: 41, lr: 9.09e-03, grad_scale: 16.0 2023-03-08 06:37:06,120 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5418, 5.0014, 4.7920, 4.9791, 5.0430, 4.5872, 3.6097, 4.9038], device='cuda:1'), covar=tensor([0.0100, 0.0107, 0.0112, 0.0072, 0.0070, 0.0113, 0.0593, 0.0183], device='cuda:1'), in_proj_covar=tensor([0.0075, 0.0070, 0.0087, 0.0053, 0.0059, 0.0070, 0.0091, 0.0090], 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') 2023-03-08 06:37:12,893 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5073, 4.9559, 4.6995, 4.8809, 4.9031, 4.5466, 3.3386, 4.8553], device='cuda:1'), covar=tensor([0.0107, 0.0130, 0.0148, 0.0083, 0.0117, 0.0131, 0.0721, 0.0210], device='cuda:1'), in_proj_covar=tensor([0.0074, 0.0070, 0.0087, 0.0053, 0.0059, 0.0069, 0.0091, 0.0090], 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') 2023-03-08 06:37:43,872 INFO [train2.py:809] (1/4) Epoch 12, batch 1850, loss[ctc_loss=0.1334, att_loss=0.278, loss=0.2491, over 17029.00 frames. utt_duration=1311 frames, utt_pad_proportion=0.008939, over 52.00 utterances.], tot_loss[ctc_loss=0.09914, att_loss=0.2484, loss=0.2185, over 3264337.85 frames. utt_duration=1216 frames, utt_pad_proportion=0.06469, over 10749.39 utterances.], batch size: 52, lr: 9.08e-03, grad_scale: 8.0 2023-03-08 06:38:32,209 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45702.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:38:35,057 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45704.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 06:38:48,027 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45712.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 06:39:02,765 INFO [train2.py:809] (1/4) Epoch 12, batch 1900, loss[ctc_loss=0.09425, att_loss=0.2549, loss=0.2228, over 17054.00 frames. utt_duration=1314 frames, utt_pad_proportion=0.0078, over 52.00 utterances.], tot_loss[ctc_loss=0.0992, att_loss=0.2482, loss=0.2184, over 3265355.91 frames. utt_duration=1222 frames, utt_pad_proportion=0.0618, over 10697.44 utterances.], batch size: 52, lr: 9.08e-03, grad_scale: 8.0 2023-03-08 06:39:04,237 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.442e+02 2.369e+02 2.834e+02 3.743e+02 8.835e+02, threshold=5.667e+02, percent-clipped=7.0 2023-03-08 06:39:26,600 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45737.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:40:02,984 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=45760.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:40:07,743 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45763.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:40:21,062 INFO [train2.py:809] (1/4) Epoch 12, batch 1950, loss[ctc_loss=0.09772, att_loss=0.2629, loss=0.2299, over 17082.00 frames. utt_duration=1315 frames, utt_pad_proportion=0.0073, over 52.00 utterances.], tot_loss[ctc_loss=0.09979, att_loss=0.2483, loss=0.2186, over 3267007.13 frames. utt_duration=1209 frames, utt_pad_proportion=0.06546, over 10822.19 utterances.], batch size: 52, lr: 9.07e-03, grad_scale: 8.0 2023-03-08 06:40:24,521 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45774.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:40:35,594 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9685, 4.9967, 4.8043, 2.8046, 4.7436, 4.5703, 4.2003, 2.4033], device='cuda:1'), covar=tensor([0.0109, 0.0076, 0.0207, 0.1065, 0.0084, 0.0188, 0.0345, 0.1550], device='cuda:1'), in_proj_covar=tensor([0.0060, 0.0083, 0.0077, 0.0102, 0.0070, 0.0094, 0.0093, 0.0098], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 06:40:41,622 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=45785.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:41:41,125 INFO [train2.py:809] (1/4) Epoch 12, batch 2000, loss[ctc_loss=0.07542, att_loss=0.2241, loss=0.1943, over 15939.00 frames. utt_duration=1557 frames, utt_pad_proportion=0.006631, over 41.00 utterances.], tot_loss[ctc_loss=0.09965, att_loss=0.2478, loss=0.2182, over 3261266.52 frames. utt_duration=1231 frames, utt_pad_proportion=0.06207, over 10607.91 utterances.], batch size: 41, lr: 9.07e-03, grad_scale: 8.0 2023-03-08 06:41:42,586 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.549e+02 2.558e+02 3.002e+02 3.965e+02 8.183e+02, threshold=6.004e+02, percent-clipped=7.0 2023-03-08 06:42:02,189 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45835.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:42:16,826 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0930, 5.3394, 5.1659, 5.2136, 5.3715, 5.3176, 4.9827, 4.8289], device='cuda:1'), covar=tensor([0.0947, 0.0411, 0.0317, 0.0556, 0.0285, 0.0273, 0.0344, 0.0286], device='cuda:1'), in_proj_covar=tensor([0.0464, 0.0299, 0.0261, 0.0292, 0.0356, 0.0364, 0.0297, 0.0328], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 06:42:41,083 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45859.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:43:00,501 INFO [train2.py:809] (1/4) Epoch 12, batch 2050, loss[ctc_loss=0.1136, att_loss=0.2589, loss=0.2298, over 17399.00 frames. utt_duration=1106 frames, utt_pad_proportion=0.03155, over 63.00 utterances.], tot_loss[ctc_loss=0.09891, att_loss=0.2476, loss=0.2179, over 3266188.48 frames. utt_duration=1254 frames, utt_pad_proportion=0.05607, over 10428.42 utterances.], batch size: 63, lr: 9.06e-03, grad_scale: 8.0 2023-03-08 06:44:18,290 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45920.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:44:21,035 INFO [train2.py:809] (1/4) Epoch 12, batch 2100, loss[ctc_loss=0.1257, att_loss=0.2695, loss=0.2407, over 16696.00 frames. utt_duration=675.9 frames, utt_pad_proportion=0.1498, over 99.00 utterances.], tot_loss[ctc_loss=0.09981, att_loss=0.2488, loss=0.219, over 3274549.18 frames. utt_duration=1247 frames, utt_pad_proportion=0.05608, over 10515.19 utterances.], batch size: 99, lr: 9.06e-03, grad_scale: 8.0 2023-03-08 06:44:22,595 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.410e+02 2.414e+02 3.055e+02 3.692e+02 7.937e+02, threshold=6.110e+02, percent-clipped=4.0 2023-03-08 06:45:41,050 INFO [train2.py:809] (1/4) Epoch 12, batch 2150, loss[ctc_loss=0.08537, att_loss=0.2423, loss=0.2109, over 16183.00 frames. utt_duration=1581 frames, utt_pad_proportion=0.006408, over 41.00 utterances.], tot_loss[ctc_loss=0.09944, att_loss=0.2482, loss=0.2184, over 3261172.28 frames. utt_duration=1258 frames, utt_pad_proportion=0.0561, over 10380.27 utterances.], batch size: 41, lr: 9.05e-03, grad_scale: 8.0 2023-03-08 06:46:37,271 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46004.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 06:47:04,791 INFO [train2.py:809] (1/4) Epoch 12, batch 2200, loss[ctc_loss=0.09158, att_loss=0.265, loss=0.2303, over 17300.00 frames. utt_duration=1174 frames, utt_pad_proportion=0.02471, over 59.00 utterances.], tot_loss[ctc_loss=0.09958, att_loss=0.2488, loss=0.219, over 3264838.03 frames. utt_duration=1245 frames, utt_pad_proportion=0.05875, over 10505.49 utterances.], batch size: 59, lr: 9.05e-03, grad_scale: 8.0 2023-03-08 06:47:06,226 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.436e+02 2.315e+02 2.803e+02 3.534e+02 8.357e+02, threshold=5.606e+02, percent-clipped=3.0 2023-03-08 06:47:53,439 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=46052.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 06:48:02,425 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46058.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:48:16,124 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9163, 5.3755, 4.2956, 5.5699, 4.7570, 5.1069, 5.3941, 5.3028], device='cuda:1'), covar=tensor([0.0670, 0.0317, 0.1502, 0.0283, 0.0470, 0.0220, 0.0346, 0.0198], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0265, 0.0321, 0.0261, 0.0271, 0.0204, 0.0254, 0.0239], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0005, 0.0006, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 06:48:23,313 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.03 vs. limit=2.0 2023-03-08 06:48:23,929 INFO [train2.py:809] (1/4) Epoch 12, batch 2250, loss[ctc_loss=0.1434, att_loss=0.2732, loss=0.2473, over 13679.00 frames. utt_duration=373.8 frames, utt_pad_proportion=0.3453, over 147.00 utterances.], tot_loss[ctc_loss=0.09941, att_loss=0.2489, loss=0.219, over 3260855.46 frames. utt_duration=1244 frames, utt_pad_proportion=0.0606, over 10495.13 utterances.], batch size: 147, lr: 9.04e-03, grad_scale: 8.0 2023-03-08 06:49:42,482 INFO [train2.py:809] (1/4) Epoch 12, batch 2300, loss[ctc_loss=0.1143, att_loss=0.2593, loss=0.2303, over 16773.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.005553, over 48.00 utterances.], tot_loss[ctc_loss=0.0995, att_loss=0.2492, loss=0.2193, over 3271709.24 frames. utt_duration=1268 frames, utt_pad_proportion=0.05178, over 10330.89 utterances.], batch size: 48, lr: 9.04e-03, grad_scale: 8.0 2023-03-08 06:49:44,022 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.648e+02 2.436e+02 3.067e+02 3.650e+02 8.781e+02, threshold=6.133e+02, percent-clipped=4.0 2023-03-08 06:49:55,625 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46130.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:51:01,257 INFO [train2.py:809] (1/4) Epoch 12, batch 2350, loss[ctc_loss=0.1259, att_loss=0.2661, loss=0.2381, over 17026.00 frames. utt_duration=1337 frames, utt_pad_proportion=0.00738, over 51.00 utterances.], tot_loss[ctc_loss=0.09897, att_loss=0.248, loss=0.2182, over 3266797.72 frames. utt_duration=1287 frames, utt_pad_proportion=0.04869, over 10164.60 utterances.], batch size: 51, lr: 9.03e-03, grad_scale: 8.0 2023-03-08 06:52:09,344 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46215.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:52:19,911 INFO [train2.py:809] (1/4) Epoch 12, batch 2400, loss[ctc_loss=0.07798, att_loss=0.2313, loss=0.2006, over 14056.00 frames. utt_duration=1815 frames, utt_pad_proportion=0.06252, over 31.00 utterances.], tot_loss[ctc_loss=0.0992, att_loss=0.2485, loss=0.2186, over 3269111.60 frames. utt_duration=1276 frames, utt_pad_proportion=0.05044, over 10257.23 utterances.], batch size: 31, lr: 9.03e-03, grad_scale: 8.0 2023-03-08 06:52:21,344 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.579e+02 2.431e+02 2.918e+02 3.900e+02 7.255e+02, threshold=5.836e+02, percent-clipped=5.0 2023-03-08 06:53:08,939 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-03-08 06:53:35,574 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-03-08 06:53:39,241 INFO [train2.py:809] (1/4) Epoch 12, batch 2450, loss[ctc_loss=0.08088, att_loss=0.2224, loss=0.1941, over 15362.00 frames. utt_duration=1757 frames, utt_pad_proportion=0.01165, over 35.00 utterances.], tot_loss[ctc_loss=0.09918, att_loss=0.2486, loss=0.2187, over 3272928.15 frames. utt_duration=1271 frames, utt_pad_proportion=0.05149, over 10312.01 utterances.], batch size: 35, lr: 9.02e-03, grad_scale: 8.0 2023-03-08 06:53:46,236 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46276.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:54:24,121 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0674, 4.4588, 4.1046, 4.4411, 2.6396, 4.5611, 2.4518, 1.8061], device='cuda:1'), covar=tensor([0.0362, 0.0121, 0.0828, 0.0147, 0.2014, 0.0137, 0.1902, 0.2044], device='cuda:1'), in_proj_covar=tensor([0.0144, 0.0117, 0.0256, 0.0115, 0.0221, 0.0109, 0.0225, 0.0201], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 06:54:28,633 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1530, 5.1595, 4.9664, 3.0354, 4.8307, 4.6589, 4.4151, 2.7426], device='cuda:1'), covar=tensor([0.0094, 0.0069, 0.0226, 0.1011, 0.0088, 0.0188, 0.0291, 0.1402], device='cuda:1'), in_proj_covar=tensor([0.0060, 0.0083, 0.0076, 0.0102, 0.0070, 0.0094, 0.0093, 0.0098], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 06:54:44,598 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5488, 2.5706, 5.1287, 3.9162, 2.9063, 4.2764, 4.9546, 4.6716], device='cuda:1'), covar=tensor([0.0234, 0.1712, 0.0166, 0.1012, 0.1977, 0.0249, 0.0104, 0.0223], device='cuda:1'), in_proj_covar=tensor([0.0151, 0.0245, 0.0138, 0.0307, 0.0272, 0.0186, 0.0126, 0.0158], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 06:54:58,104 INFO [train2.py:809] (1/4) Epoch 12, batch 2500, loss[ctc_loss=0.1339, att_loss=0.2774, loss=0.2487, over 17035.00 frames. utt_duration=1338 frames, utt_pad_proportion=0.006958, over 51.00 utterances.], tot_loss[ctc_loss=0.09895, att_loss=0.2484, loss=0.2185, over 3279858.07 frames. utt_duration=1271 frames, utt_pad_proportion=0.04932, over 10333.38 utterances.], batch size: 51, lr: 9.02e-03, grad_scale: 8.0 2023-03-08 06:55:00,272 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.509e+02 2.157e+02 2.613e+02 3.362e+02 8.530e+02, threshold=5.227e+02, percent-clipped=3.0 2023-03-08 06:55:22,635 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46337.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:55:56,111 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46358.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:56:09,752 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.96 vs. limit=5.0 2023-03-08 06:56:18,495 INFO [train2.py:809] (1/4) Epoch 12, batch 2550, loss[ctc_loss=0.1147, att_loss=0.2643, loss=0.2344, over 17052.00 frames. utt_duration=1313 frames, utt_pad_proportion=0.008824, over 52.00 utterances.], tot_loss[ctc_loss=0.09954, att_loss=0.2488, loss=0.219, over 3278867.85 frames. utt_duration=1230 frames, utt_pad_proportion=0.05816, over 10675.82 utterances.], batch size: 52, lr: 9.01e-03, grad_scale: 8.0 2023-03-08 06:56:28,135 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.7476, 3.2958, 3.6836, 4.4893, 3.9289, 4.1720, 2.9558, 2.4653], device='cuda:1'), covar=tensor([0.0583, 0.1824, 0.0857, 0.0700, 0.0892, 0.0372, 0.1563, 0.2103], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0210, 0.0187, 0.0195, 0.0193, 0.0154, 0.0198, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 06:56:57,549 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.68 vs. limit=5.0 2023-03-08 06:57:12,243 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=46406.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:57:38,531 INFO [train2.py:809] (1/4) Epoch 12, batch 2600, loss[ctc_loss=0.09823, att_loss=0.2351, loss=0.2077, over 15628.00 frames. utt_duration=1691 frames, utt_pad_proportion=0.009262, over 37.00 utterances.], tot_loss[ctc_loss=0.09979, att_loss=0.2493, loss=0.2194, over 3284083.15 frames. utt_duration=1217 frames, utt_pad_proportion=0.05917, over 10803.59 utterances.], batch size: 37, lr: 9.01e-03, grad_scale: 8.0 2023-03-08 06:57:39,864 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.782e+02 2.587e+02 3.130e+02 3.974e+02 7.202e+02, threshold=6.259e+02, percent-clipped=6.0 2023-03-08 06:57:51,765 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46430.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:58:57,575 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-03-08 06:58:58,113 INFO [train2.py:809] (1/4) Epoch 12, batch 2650, loss[ctc_loss=0.09371, att_loss=0.2503, loss=0.219, over 17116.00 frames. utt_duration=693 frames, utt_pad_proportion=0.1304, over 99.00 utterances.], tot_loss[ctc_loss=0.09957, att_loss=0.2496, loss=0.2196, over 3280361.41 frames. utt_duration=1219 frames, utt_pad_proportion=0.05783, over 10774.93 utterances.], batch size: 99, lr: 9.00e-03, grad_scale: 8.0 2023-03-08 06:59:07,495 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=46478.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 06:59:48,711 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-03-08 07:00:06,995 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46515.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:00:11,540 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-03-08 07:00:18,194 INFO [train2.py:809] (1/4) Epoch 12, batch 2700, loss[ctc_loss=0.1207, att_loss=0.2808, loss=0.2488, over 17386.00 frames. utt_duration=1106 frames, utt_pad_proportion=0.03024, over 63.00 utterances.], tot_loss[ctc_loss=0.09953, att_loss=0.2495, loss=0.2195, over 3279361.75 frames. utt_duration=1231 frames, utt_pad_proportion=0.05569, over 10666.30 utterances.], batch size: 63, lr: 9.00e-03, grad_scale: 8.0 2023-03-08 07:00:19,663 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.175e+02 2.444e+02 2.933e+02 3.610e+02 6.469e+02, threshold=5.867e+02, percent-clipped=1.0 2023-03-08 07:01:03,089 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.6845, 3.7196, 3.6044, 3.1366, 3.7680, 3.7514, 3.5302, 2.8131], device='cuda:1'), covar=tensor([0.1505, 0.1552, 0.2920, 0.6969, 0.3035, 0.5715, 0.1380, 0.7318], device='cuda:1'), in_proj_covar=tensor([0.0108, 0.0126, 0.0140, 0.0209, 0.0109, 0.0194, 0.0116, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 07:01:23,501 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=46563.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:01:37,872 INFO [train2.py:809] (1/4) Epoch 12, batch 2750, loss[ctc_loss=0.07231, att_loss=0.2237, loss=0.1934, over 15525.00 frames. utt_duration=1726 frames, utt_pad_proportion=0.006809, over 36.00 utterances.], tot_loss[ctc_loss=0.09891, att_loss=0.2492, loss=0.2191, over 3277563.97 frames. utt_duration=1224 frames, utt_pad_proportion=0.05838, over 10724.62 utterances.], batch size: 36, lr: 9.00e-03, grad_scale: 8.0 2023-03-08 07:01:48,624 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.53 vs. limit=5.0 2023-03-08 07:02:15,215 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46595.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:02:26,662 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46602.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:02:52,430 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.86 vs. limit=2.0 2023-03-08 07:02:53,557 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7657, 4.9770, 4.5815, 5.1494, 4.4920, 4.8671, 5.2498, 4.9905], device='cuda:1'), covar=tensor([0.0554, 0.0332, 0.0871, 0.0259, 0.0479, 0.0245, 0.0226, 0.0176], device='cuda:1'), in_proj_covar=tensor([0.0339, 0.0268, 0.0323, 0.0265, 0.0273, 0.0207, 0.0254, 0.0241], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0005, 0.0006, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 07:02:57,842 INFO [train2.py:809] (1/4) Epoch 12, batch 2800, loss[ctc_loss=0.09708, att_loss=0.2361, loss=0.2083, over 15525.00 frames. utt_duration=1726 frames, utt_pad_proportion=0.007364, over 36.00 utterances.], tot_loss[ctc_loss=0.09863, att_loss=0.2485, loss=0.2185, over 3268478.07 frames. utt_duration=1220 frames, utt_pad_proportion=0.06125, over 10726.69 utterances.], batch size: 36, lr: 8.99e-03, grad_scale: 8.0 2023-03-08 07:02:59,237 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.251e+02 2.286e+02 2.805e+02 3.463e+02 9.248e+02, threshold=5.611e+02, percent-clipped=3.0 2023-03-08 07:03:14,282 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46632.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:03:52,221 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-03-08 07:03:52,978 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46656.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 07:04:04,653 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46663.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:04:18,470 INFO [train2.py:809] (1/4) Epoch 12, batch 2850, loss[ctc_loss=0.1043, att_loss=0.2609, loss=0.2296, over 16533.00 frames. utt_duration=1471 frames, utt_pad_proportion=0.005946, over 45.00 utterances.], tot_loss[ctc_loss=0.09866, att_loss=0.2482, loss=0.2183, over 3263633.50 frames. utt_duration=1218 frames, utt_pad_proportion=0.06401, over 10735.16 utterances.], batch size: 45, lr: 8.99e-03, grad_scale: 8.0 2023-03-08 07:04:57,087 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-03-08 07:05:39,964 INFO [train2.py:809] (1/4) Epoch 12, batch 2900, loss[ctc_loss=0.08518, att_loss=0.2253, loss=0.1973, over 15501.00 frames. utt_duration=1724 frames, utt_pad_proportion=0.008279, over 36.00 utterances.], tot_loss[ctc_loss=0.09883, att_loss=0.2485, loss=0.2185, over 3267488.43 frames. utt_duration=1224 frames, utt_pad_proportion=0.06191, over 10693.66 utterances.], batch size: 36, lr: 8.98e-03, grad_scale: 8.0 2023-03-08 07:05:41,464 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.596e+02 2.133e+02 2.612e+02 3.383e+02 7.389e+02, threshold=5.223e+02, percent-clipped=3.0 2023-03-08 07:06:13,925 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6831, 5.1620, 5.0011, 5.0780, 5.2475, 4.8142, 3.8091, 5.0724], device='cuda:1'), covar=tensor([0.0098, 0.0081, 0.0101, 0.0071, 0.0111, 0.0091, 0.0592, 0.0193], device='cuda:1'), in_proj_covar=tensor([0.0078, 0.0073, 0.0090, 0.0055, 0.0061, 0.0072, 0.0095, 0.0094], 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') 2023-03-08 07:07:00,901 INFO [train2.py:809] (1/4) Epoch 12, batch 2950, loss[ctc_loss=0.08576, att_loss=0.2383, loss=0.2078, over 16399.00 frames. utt_duration=1492 frames, utt_pad_proportion=0.007752, over 44.00 utterances.], tot_loss[ctc_loss=0.0999, att_loss=0.2487, loss=0.2189, over 3259639.74 frames. utt_duration=1207 frames, utt_pad_proportion=0.0673, over 10816.08 utterances.], batch size: 44, lr: 8.98e-03, grad_scale: 8.0 2023-03-08 07:08:03,478 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46810.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:08:23,020 INFO [train2.py:809] (1/4) Epoch 12, batch 3000, loss[ctc_loss=0.1051, att_loss=0.2694, loss=0.2366, over 17157.00 frames. utt_duration=1227 frames, utt_pad_proportion=0.01274, over 56.00 utterances.], tot_loss[ctc_loss=0.09913, att_loss=0.2485, loss=0.2186, over 3267368.03 frames. utt_duration=1223 frames, utt_pad_proportion=0.0615, over 10699.33 utterances.], batch size: 56, lr: 8.97e-03, grad_scale: 8.0 2023-03-08 07:08:23,020 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-08 07:08:35,035 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.2476, 4.7809, 4.8540, 4.7210, 4.6625, 5.1676, 5.0088, 5.2656], device='cuda:1'), covar=tensor([0.0696, 0.0699, 0.0799, 0.1079, 0.1906, 0.0918, 0.0356, 0.0628], device='cuda:1'), in_proj_covar=tensor([0.0734, 0.0429, 0.0509, 0.0562, 0.0742, 0.0515, 0.0410, 0.0498], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 07:08:37,754 INFO [train2.py:843] (1/4) Epoch 12, validation: ctc_loss=0.04782, att_loss=0.2378, loss=0.1998, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-08 07:08:37,754 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-08 07:08:39,277 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.604e+02 2.287e+02 2.757e+02 3.651e+02 8.301e+02, threshold=5.514e+02, percent-clipped=2.0 2023-03-08 07:09:56,632 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46871.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:09:57,798 INFO [train2.py:809] (1/4) Epoch 12, batch 3050, loss[ctc_loss=0.1059, att_loss=0.2565, loss=0.2264, over 16862.00 frames. utt_duration=682.6 frames, utt_pad_proportion=0.1424, over 99.00 utterances.], tot_loss[ctc_loss=0.1, att_loss=0.2493, loss=0.2195, over 3272374.38 frames. utt_duration=1212 frames, utt_pad_proportion=0.06367, over 10817.47 utterances.], batch size: 99, lr: 8.97e-03, grad_scale: 8.0 2023-03-08 07:10:14,083 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9013, 5.1462, 5.1021, 5.0154, 5.2141, 5.1977, 4.7923, 4.6203], device='cuda:1'), covar=tensor([0.0975, 0.0542, 0.0252, 0.0536, 0.0264, 0.0255, 0.0376, 0.0331], device='cuda:1'), in_proj_covar=tensor([0.0468, 0.0305, 0.0265, 0.0297, 0.0355, 0.0371, 0.0303, 0.0337], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 07:10:34,290 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.8322, 6.0883, 5.5551, 5.8728, 5.7682, 5.3854, 5.4943, 5.3601], device='cuda:1'), covar=tensor([0.1141, 0.0825, 0.0715, 0.0673, 0.0893, 0.1317, 0.2153, 0.2018], device='cuda:1'), in_proj_covar=tensor([0.0449, 0.0513, 0.0386, 0.0388, 0.0376, 0.0429, 0.0532, 0.0460], 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') 2023-03-08 07:10:54,517 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.54 vs. limit=5.0 2023-03-08 07:11:17,290 INFO [train2.py:809] (1/4) Epoch 12, batch 3100, loss[ctc_loss=0.09085, att_loss=0.2486, loss=0.217, over 16687.00 frames. utt_duration=1453 frames, utt_pad_proportion=0.005773, over 46.00 utterances.], tot_loss[ctc_loss=0.09985, att_loss=0.2486, loss=0.2188, over 3270189.66 frames. utt_duration=1217 frames, utt_pad_proportion=0.06389, over 10760.63 utterances.], batch size: 46, lr: 8.96e-03, grad_scale: 8.0 2023-03-08 07:11:18,788 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.489e+02 2.293e+02 2.705e+02 3.311e+02 9.468e+02, threshold=5.410e+02, percent-clipped=5.0 2023-03-08 07:11:31,496 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1147, 5.3666, 4.9372, 5.4935, 4.8511, 5.1507, 5.5406, 5.2952], device='cuda:1'), covar=tensor([0.0509, 0.0265, 0.0838, 0.0207, 0.0460, 0.0173, 0.0224, 0.0190], device='cuda:1'), in_proj_covar=tensor([0.0341, 0.0269, 0.0328, 0.0267, 0.0276, 0.0208, 0.0258, 0.0243], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0005, 0.0006, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 07:11:32,973 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46932.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:12:02,252 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46951.0, num_to_drop=1, layers_to_drop={3} 2023-03-08 07:12:13,037 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46958.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:12:35,421 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.71 vs. limit=2.0 2023-03-08 07:12:36,183 INFO [train2.py:809] (1/4) Epoch 12, batch 3150, loss[ctc_loss=0.06478, att_loss=0.2106, loss=0.1814, over 15509.00 frames. utt_duration=1725 frames, utt_pad_proportion=0.008226, over 36.00 utterances.], tot_loss[ctc_loss=0.09914, att_loss=0.2479, loss=0.2182, over 3268484.96 frames. utt_duration=1231 frames, utt_pad_proportion=0.06007, over 10634.60 utterances.], batch size: 36, lr: 8.96e-03, grad_scale: 8.0 2023-03-08 07:12:48,215 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=46980.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:12:48,344 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5151, 4.7614, 4.7608, 4.6841, 4.8433, 4.8155, 4.4567, 4.2980], device='cuda:1'), covar=tensor([0.0995, 0.0496, 0.0283, 0.0515, 0.0331, 0.0326, 0.0376, 0.0398], device='cuda:1'), in_proj_covar=tensor([0.0473, 0.0304, 0.0265, 0.0298, 0.0357, 0.0373, 0.0306, 0.0339], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 07:12:52,265 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-03-08 07:13:34,475 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9163, 4.5270, 4.5502, 2.2215, 2.1099, 2.7958, 2.3538, 3.7336], device='cuda:1'), covar=tensor([0.0677, 0.0201, 0.0218, 0.4451, 0.5259, 0.2422, 0.2742, 0.1412], device='cuda:1'), in_proj_covar=tensor([0.0332, 0.0221, 0.0238, 0.0213, 0.0343, 0.0331, 0.0230, 0.0353], device='cuda:1'), out_proj_covar=tensor([1.4779e-04, 8.2796e-05, 1.0323e-04, 9.4109e-05, 1.4877e-04, 1.3340e-04, 9.0634e-05, 1.4811e-04], device='cuda:1') 2023-03-08 07:13:55,731 INFO [train2.py:809] (1/4) Epoch 12, batch 3200, loss[ctc_loss=0.1337, att_loss=0.2674, loss=0.2406, over 17067.00 frames. utt_duration=691.1 frames, utt_pad_proportion=0.1329, over 99.00 utterances.], tot_loss[ctc_loss=0.09866, att_loss=0.2482, loss=0.2183, over 3280828.39 frames. utt_duration=1240 frames, utt_pad_proportion=0.05358, over 10592.32 utterances.], batch size: 99, lr: 8.95e-03, grad_scale: 8.0 2023-03-08 07:13:57,262 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.636e+02 2.280e+02 2.660e+02 3.739e+02 9.855e+02, threshold=5.321e+02, percent-clipped=6.0 2023-03-08 07:14:03,910 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7976, 2.3745, 2.6227, 3.2423, 3.0803, 3.2601, 2.5304, 2.1740], device='cuda:1'), covar=tensor([0.0688, 0.1821, 0.0963, 0.0729, 0.0715, 0.0401, 0.1386, 0.1822], device='cuda:1'), in_proj_covar=tensor([0.0176, 0.0213, 0.0192, 0.0197, 0.0193, 0.0157, 0.0199, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 07:14:27,334 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.74 vs. limit=2.0 2023-03-08 07:14:45,444 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7141, 4.5767, 4.6967, 4.5820, 5.1507, 4.8483, 4.6801, 2.3678], device='cuda:1'), covar=tensor([0.0185, 0.0257, 0.0169, 0.0197, 0.0698, 0.0158, 0.0201, 0.2035], device='cuda:1'), in_proj_covar=tensor([0.0126, 0.0137, 0.0137, 0.0150, 0.0339, 0.0123, 0.0125, 0.0214], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 07:15:16,023 INFO [train2.py:809] (1/4) Epoch 12, batch 3250, loss[ctc_loss=0.08613, att_loss=0.2429, loss=0.2115, over 16560.00 frames. utt_duration=1473 frames, utt_pad_proportion=0.005147, over 45.00 utterances.], tot_loss[ctc_loss=0.09847, att_loss=0.2481, loss=0.2182, over 3278677.76 frames. utt_duration=1229 frames, utt_pad_proportion=0.0579, over 10684.48 utterances.], batch size: 45, lr: 8.95e-03, grad_scale: 8.0 2023-03-08 07:15:42,035 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-03-08 07:15:58,877 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.80 vs. limit=2.0 2023-03-08 07:16:07,674 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.3239, 2.4610, 3.3381, 4.2420, 3.7114, 3.8876, 2.7413, 2.1099], device='cuda:1'), covar=tensor([0.0755, 0.2377, 0.0926, 0.0657, 0.0805, 0.0400, 0.1735, 0.2306], device='cuda:1'), in_proj_covar=tensor([0.0176, 0.0211, 0.0190, 0.0197, 0.0194, 0.0156, 0.0198, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 07:16:34,791 INFO [train2.py:809] (1/4) Epoch 12, batch 3300, loss[ctc_loss=0.0931, att_loss=0.2343, loss=0.2061, over 15632.00 frames. utt_duration=1692 frames, utt_pad_proportion=0.009589, over 37.00 utterances.], tot_loss[ctc_loss=0.09786, att_loss=0.2483, loss=0.2182, over 3284408.22 frames. utt_duration=1237 frames, utt_pad_proportion=0.05533, over 10631.48 utterances.], batch size: 37, lr: 8.94e-03, grad_scale: 8.0 2023-03-08 07:16:36,329 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.338e+02 2.270e+02 2.752e+02 3.318e+02 7.003e+02, threshold=5.505e+02, percent-clipped=5.0 2023-03-08 07:17:43,755 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=47166.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:17:53,907 INFO [train2.py:809] (1/4) Epoch 12, batch 3350, loss[ctc_loss=0.09232, att_loss=0.2317, loss=0.2038, over 16015.00 frames. utt_duration=1603 frames, utt_pad_proportion=0.007048, over 40.00 utterances.], tot_loss[ctc_loss=0.09833, att_loss=0.2488, loss=0.2187, over 3286821.46 frames. utt_duration=1244 frames, utt_pad_proportion=0.05254, over 10581.63 utterances.], batch size: 40, lr: 8.94e-03, grad_scale: 8.0 2023-03-08 07:18:27,670 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7861, 5.0473, 4.9731, 4.9884, 5.1371, 5.1028, 4.7584, 4.5140], device='cuda:1'), covar=tensor([0.0994, 0.0533, 0.0332, 0.0496, 0.0265, 0.0271, 0.0375, 0.0354], device='cuda:1'), in_proj_covar=tensor([0.0476, 0.0305, 0.0266, 0.0298, 0.0358, 0.0373, 0.0307, 0.0340], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 07:19:14,296 INFO [train2.py:809] (1/4) Epoch 12, batch 3400, loss[ctc_loss=0.1443, att_loss=0.2759, loss=0.2495, over 14067.00 frames. utt_duration=386.8 frames, utt_pad_proportion=0.3261, over 146.00 utterances.], tot_loss[ctc_loss=0.09817, att_loss=0.2488, loss=0.2187, over 3285559.19 frames. utt_duration=1243 frames, utt_pad_proportion=0.05351, over 10583.68 utterances.], batch size: 146, lr: 8.93e-03, grad_scale: 8.0 2023-03-08 07:19:15,802 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.372e+02 2.335e+02 2.724e+02 3.467e+02 7.189e+02, threshold=5.448e+02, percent-clipped=4.0 2023-03-08 07:19:44,368 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47241.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:19:59,843 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47251.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:19:59,993 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1679, 4.1739, 3.9085, 4.0318, 4.4742, 4.1529, 3.9404, 2.5175], device='cuda:1'), covar=tensor([0.0277, 0.0367, 0.0387, 0.0257, 0.0949, 0.0266, 0.0337, 0.1768], device='cuda:1'), in_proj_covar=tensor([0.0128, 0.0137, 0.0139, 0.0151, 0.0342, 0.0123, 0.0126, 0.0216], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0002, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 07:20:10,661 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47258.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:20:33,849 INFO [train2.py:809] (1/4) Epoch 12, batch 3450, loss[ctc_loss=0.09347, att_loss=0.2599, loss=0.2266, over 17324.00 frames. utt_duration=1102 frames, utt_pad_proportion=0.03546, over 63.00 utterances.], tot_loss[ctc_loss=0.09894, att_loss=0.2495, loss=0.2194, over 3285586.13 frames. utt_duration=1226 frames, utt_pad_proportion=0.05659, over 10736.49 utterances.], batch size: 63, lr: 8.93e-03, grad_scale: 8.0 2023-03-08 07:21:15,943 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=47299.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:21:20,712 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47302.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:21:26,636 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=47306.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:21:39,900 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-03-08 07:21:53,258 INFO [train2.py:809] (1/4) Epoch 12, batch 3500, loss[ctc_loss=0.08908, att_loss=0.2582, loss=0.2244, over 17275.00 frames. utt_duration=1258 frames, utt_pad_proportion=0.01349, over 55.00 utterances.], tot_loss[ctc_loss=0.09977, att_loss=0.2496, loss=0.2196, over 3280793.79 frames. utt_duration=1213 frames, utt_pad_proportion=0.06036, over 10830.00 utterances.], batch size: 55, lr: 8.92e-03, grad_scale: 8.0 2023-03-08 07:21:53,654 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3257, 2.7445, 3.4438, 2.7022, 3.4201, 4.4742, 4.2939, 3.0997], device='cuda:1'), covar=tensor([0.0391, 0.1737, 0.1094, 0.1408, 0.1033, 0.0691, 0.0493, 0.1374], device='cuda:1'), in_proj_covar=tensor([0.0226, 0.0228, 0.0249, 0.0204, 0.0239, 0.0310, 0.0222, 0.0220], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 07:21:54,808 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.587e+02 2.392e+02 2.957e+02 3.491e+02 6.570e+02, threshold=5.913e+02, percent-clipped=5.0 2023-03-08 07:22:44,734 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.12 vs. limit=5.0 2023-03-08 07:22:46,960 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0397, 5.2735, 5.5562, 5.3435, 5.4003, 5.9492, 5.2594, 6.0327], device='cuda:1'), covar=tensor([0.0634, 0.0810, 0.0705, 0.1159, 0.1842, 0.0926, 0.0607, 0.0682], device='cuda:1'), in_proj_covar=tensor([0.0743, 0.0439, 0.0513, 0.0576, 0.0756, 0.0519, 0.0416, 0.0501], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 07:23:13,808 INFO [train2.py:809] (1/4) Epoch 12, batch 3550, loss[ctc_loss=0.09149, att_loss=0.2635, loss=0.2291, over 17067.00 frames. utt_duration=1221 frames, utt_pad_proportion=0.01627, over 56.00 utterances.], tot_loss[ctc_loss=0.0985, att_loss=0.2491, loss=0.219, over 3279651.93 frames. utt_duration=1219 frames, utt_pad_proportion=0.05964, over 10777.89 utterances.], batch size: 56, lr: 8.92e-03, grad_scale: 8.0 2023-03-08 07:23:25,937 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47379.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:23:44,658 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.0737, 3.8488, 3.2318, 3.5772, 4.1013, 3.6071, 3.0259, 4.4601], device='cuda:1'), covar=tensor([0.1087, 0.0466, 0.1041, 0.0673, 0.0570, 0.0751, 0.0820, 0.0418], device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0190, 0.0208, 0.0180, 0.0245, 0.0218, 0.0185, 0.0261], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-08 07:24:23,985 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-03-08 07:24:34,830 INFO [train2.py:809] (1/4) Epoch 12, batch 3600, loss[ctc_loss=0.1359, att_loss=0.2767, loss=0.2486, over 17494.00 frames. utt_duration=1016 frames, utt_pad_proportion=0.04282, over 69.00 utterances.], tot_loss[ctc_loss=0.0986, att_loss=0.2493, loss=0.2191, over 3280788.60 frames. utt_duration=1232 frames, utt_pad_proportion=0.05572, over 10665.19 utterances.], batch size: 69, lr: 8.92e-03, grad_scale: 8.0 2023-03-08 07:24:36,351 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.815e+02 2.371e+02 2.874e+02 3.552e+02 5.141e+02, threshold=5.747e+02, percent-clipped=0.0 2023-03-08 07:24:49,003 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.30 vs. limit=5.0 2023-03-08 07:24:56,388 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6826, 3.1979, 3.6892, 3.1329, 3.7571, 4.7166, 4.5649, 3.6184], device='cuda:1'), covar=tensor([0.0265, 0.1359, 0.0945, 0.1182, 0.0881, 0.0664, 0.0454, 0.1061], device='cuda:1'), in_proj_covar=tensor([0.0225, 0.0228, 0.0248, 0.0204, 0.0240, 0.0311, 0.0222, 0.0220], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 07:25:04,379 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47440.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:25:09,779 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.35 vs. limit=5.0 2023-03-08 07:25:47,348 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47466.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:25:56,421 INFO [train2.py:809] (1/4) Epoch 12, batch 3650, loss[ctc_loss=0.1115, att_loss=0.2615, loss=0.2315, over 17344.00 frames. utt_duration=1103 frames, utt_pad_proportion=0.0352, over 63.00 utterances.], tot_loss[ctc_loss=0.09766, att_loss=0.2485, loss=0.2183, over 3276012.63 frames. utt_duration=1243 frames, utt_pad_proportion=0.0532, over 10556.41 utterances.], batch size: 63, lr: 8.91e-03, grad_scale: 8.0 2023-03-08 07:25:56,772 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.4906, 2.7133, 3.4494, 4.4202, 3.9311, 4.0443, 2.9048, 1.9519], device='cuda:1'), covar=tensor([0.0661, 0.2368, 0.0892, 0.0649, 0.0819, 0.0380, 0.1563, 0.2810], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0214, 0.0190, 0.0198, 0.0194, 0.0160, 0.0200, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 07:26:04,959 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5337, 5.0780, 4.9478, 4.9790, 5.0585, 4.7358, 3.3522, 4.9558], device='cuda:1'), covar=tensor([0.0134, 0.0137, 0.0130, 0.0115, 0.0107, 0.0125, 0.0845, 0.0275], device='cuda:1'), in_proj_covar=tensor([0.0079, 0.0074, 0.0092, 0.0056, 0.0062, 0.0074, 0.0096, 0.0095], 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') 2023-03-08 07:26:34,343 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.2280, 5.1827, 5.0736, 3.1146, 5.0270, 4.7918, 4.6962, 3.0523], device='cuda:1'), covar=tensor([0.0087, 0.0067, 0.0183, 0.0918, 0.0071, 0.0138, 0.0201, 0.1117], device='cuda:1'), in_proj_covar=tensor([0.0062, 0.0087, 0.0081, 0.0106, 0.0073, 0.0096, 0.0094, 0.0101], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 07:27:03,208 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=47514.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:27:16,059 INFO [train2.py:809] (1/4) Epoch 12, batch 3700, loss[ctc_loss=0.08956, att_loss=0.2292, loss=0.2012, over 15634.00 frames. utt_duration=1692 frames, utt_pad_proportion=0.009041, over 37.00 utterances.], tot_loss[ctc_loss=0.09726, att_loss=0.2481, loss=0.2179, over 3274986.18 frames. utt_duration=1256 frames, utt_pad_proportion=0.04912, over 10441.36 utterances.], batch size: 37, lr: 8.91e-03, grad_scale: 8.0 2023-03-08 07:27:17,595 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.684e+02 2.382e+02 2.831e+02 3.423e+02 6.375e+02, threshold=5.662e+02, percent-clipped=1.0 2023-03-08 07:28:38,161 INFO [train2.py:809] (1/4) Epoch 12, batch 3750, loss[ctc_loss=0.08473, att_loss=0.2352, loss=0.2051, over 15777.00 frames. utt_duration=1662 frames, utt_pad_proportion=0.008353, over 38.00 utterances.], tot_loss[ctc_loss=0.09859, att_loss=0.2489, loss=0.2188, over 3276479.31 frames. utt_duration=1230 frames, utt_pad_proportion=0.05668, over 10671.43 utterances.], batch size: 38, lr: 8.90e-03, grad_scale: 8.0 2023-03-08 07:29:17,214 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=47597.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:29:58,051 INFO [train2.py:809] (1/4) Epoch 12, batch 3800, loss[ctc_loss=0.07915, att_loss=0.2574, loss=0.2217, over 17025.00 frames. utt_duration=1337 frames, utt_pad_proportion=0.00816, over 51.00 utterances.], tot_loss[ctc_loss=0.09866, att_loss=0.2491, loss=0.219, over 3279631.47 frames. utt_duration=1245 frames, utt_pad_proportion=0.05246, over 10545.63 utterances.], batch size: 51, lr: 8.90e-03, grad_scale: 8.0 2023-03-08 07:29:59,587 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.758e+02 2.386e+02 2.816e+02 3.593e+02 7.143e+02, threshold=5.632e+02, percent-clipped=4.0 2023-03-08 07:30:03,866 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.12 vs. limit=5.0 2023-03-08 07:31:17,522 INFO [train2.py:809] (1/4) Epoch 12, batch 3850, loss[ctc_loss=0.1019, att_loss=0.2672, loss=0.2341, over 17138.00 frames. utt_duration=1226 frames, utt_pad_proportion=0.01401, over 56.00 utterances.], tot_loss[ctc_loss=0.09766, att_loss=0.2485, loss=0.2184, over 3284659.70 frames. utt_duration=1266 frames, utt_pad_proportion=0.04621, over 10392.77 utterances.], batch size: 56, lr: 8.89e-03, grad_scale: 16.0 2023-03-08 07:32:34,727 INFO [train2.py:809] (1/4) Epoch 12, batch 3900, loss[ctc_loss=0.06037, att_loss=0.212, loss=0.1817, over 16006.00 frames. utt_duration=1602 frames, utt_pad_proportion=0.007988, over 40.00 utterances.], tot_loss[ctc_loss=0.09711, att_loss=0.2482, loss=0.218, over 3290054.26 frames. utt_duration=1282 frames, utt_pad_proportion=0.04112, over 10280.70 utterances.], batch size: 40, lr: 8.89e-03, grad_scale: 16.0 2023-03-08 07:32:36,203 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.572e+02 2.278e+02 2.685e+02 3.452e+02 6.474e+02, threshold=5.369e+02, percent-clipped=2.0 2023-03-08 07:32:55,405 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=47735.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:33:00,347 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47738.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:33:52,999 INFO [train2.py:809] (1/4) Epoch 12, batch 3950, loss[ctc_loss=0.08551, att_loss=0.2216, loss=0.1944, over 15622.00 frames. utt_duration=1691 frames, utt_pad_proportion=0.009611, over 37.00 utterances.], tot_loss[ctc_loss=0.0969, att_loss=0.2483, loss=0.218, over 3288030.93 frames. utt_duration=1277 frames, utt_pad_proportion=0.04422, over 10312.89 utterances.], batch size: 37, lr: 8.88e-03, grad_scale: 8.0 2023-03-08 07:34:17,716 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47788.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:34:35,096 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47799.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:35:12,421 INFO [train2.py:809] (1/4) Epoch 13, batch 0, loss[ctc_loss=0.1236, att_loss=0.2698, loss=0.2405, over 17024.00 frames. utt_duration=1286 frames, utt_pad_proportion=0.0112, over 53.00 utterances.], tot_loss[ctc_loss=0.1236, att_loss=0.2698, loss=0.2405, over 17024.00 frames. utt_duration=1286 frames, utt_pad_proportion=0.0112, over 53.00 utterances.], batch size: 53, lr: 8.53e-03, grad_scale: 8.0 2023-03-08 07:35:12,421 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-08 07:35:24,623 INFO [train2.py:843] (1/4) Epoch 13, validation: ctc_loss=0.04799, att_loss=0.2379, loss=0.2, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-08 07:35:24,624 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-08 07:35:29,634 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47809.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 07:35:54,071 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.302e+02 2.314e+02 2.717e+02 3.534e+02 1.025e+03, threshold=5.435e+02, percent-clipped=1.0 2023-03-08 07:36:34,346 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47849.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:36:45,286 INFO [train2.py:809] (1/4) Epoch 13, batch 50, loss[ctc_loss=0.09359, att_loss=0.2597, loss=0.2265, over 16688.00 frames. utt_duration=1453 frames, utt_pad_proportion=0.006394, over 46.00 utterances.], tot_loss[ctc_loss=0.09737, att_loss=0.2471, loss=0.2171, over 733337.24 frames. utt_duration=1332 frames, utt_pad_proportion=0.03771, over 2204.87 utterances.], batch size: 46, lr: 8.53e-03, grad_scale: 8.0 2023-03-08 07:37:05,464 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.8237, 2.7894, 3.4157, 2.6029, 3.2700, 3.9824, 3.9113, 2.9055], device='cuda:1'), covar=tensor([0.0408, 0.1486, 0.0971, 0.1346, 0.1007, 0.0819, 0.0525, 0.1316], device='cuda:1'), in_proj_covar=tensor([0.0230, 0.0236, 0.0255, 0.0209, 0.0248, 0.0319, 0.0229, 0.0227], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 07:37:08,552 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47870.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 07:37:36,197 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.78 vs. limit=2.0 2023-03-08 07:37:51,326 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47897.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:37:53,377 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.73 vs. limit=2.0 2023-03-08 07:38:05,791 INFO [train2.py:809] (1/4) Epoch 13, batch 100, loss[ctc_loss=0.08412, att_loss=0.2509, loss=0.2175, over 16768.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.00657, over 48.00 utterances.], tot_loss[ctc_loss=0.09674, att_loss=0.2475, loss=0.2173, over 1298614.12 frames. utt_duration=1235 frames, utt_pad_proportion=0.05714, over 4211.08 utterances.], batch size: 48, lr: 8.52e-03, grad_scale: 8.0 2023-03-08 07:38:35,372 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.519e+02 2.449e+02 2.869e+02 3.545e+02 8.286e+02, threshold=5.739e+02, percent-clipped=3.0 2023-03-08 07:39:08,783 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=47945.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:39:26,427 INFO [train2.py:809] (1/4) Epoch 13, batch 150, loss[ctc_loss=0.08055, att_loss=0.2485, loss=0.2149, over 17009.00 frames. utt_duration=1336 frames, utt_pad_proportion=0.008385, over 51.00 utterances.], tot_loss[ctc_loss=0.09589, att_loss=0.2473, loss=0.217, over 1738322.99 frames. utt_duration=1199 frames, utt_pad_proportion=0.06505, over 5806.89 utterances.], batch size: 51, lr: 8.52e-03, grad_scale: 8.0 2023-03-08 07:40:51,747 INFO [train2.py:809] (1/4) Epoch 13, batch 200, loss[ctc_loss=0.1183, att_loss=0.2758, loss=0.2443, over 17275.00 frames. utt_duration=1258 frames, utt_pad_proportion=0.01335, over 55.00 utterances.], tot_loss[ctc_loss=0.09532, att_loss=0.247, loss=0.2167, over 2082620.04 frames. utt_duration=1258 frames, utt_pad_proportion=0.04969, over 6632.53 utterances.], batch size: 55, lr: 8.52e-03, grad_scale: 8.0 2023-03-08 07:41:07,835 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-03-08 07:41:20,953 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.382e+02 2.209e+02 2.594e+02 3.149e+02 6.379e+02, threshold=5.187e+02, percent-clipped=1.0 2023-03-08 07:41:38,788 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48035.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:41:38,958 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4184, 2.3140, 4.7474, 3.7550, 2.9514, 4.1922, 4.3737, 4.3767], device='cuda:1'), covar=tensor([0.0179, 0.1806, 0.0098, 0.1006, 0.1755, 0.0216, 0.0129, 0.0235], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0241, 0.0138, 0.0305, 0.0265, 0.0182, 0.0125, 0.0157], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 07:42:10,164 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0112, 4.9403, 4.7125, 2.7031, 4.6811, 4.6074, 4.1892, 2.6560], device='cuda:1'), covar=tensor([0.0117, 0.0113, 0.0286, 0.1323, 0.0119, 0.0209, 0.0378, 0.1601], device='cuda:1'), in_proj_covar=tensor([0.0063, 0.0087, 0.0082, 0.0106, 0.0073, 0.0095, 0.0095, 0.0101], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 07:42:12,097 INFO [train2.py:809] (1/4) Epoch 13, batch 250, loss[ctc_loss=0.09135, att_loss=0.2592, loss=0.2257, over 16775.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.006096, over 48.00 utterances.], tot_loss[ctc_loss=0.09545, att_loss=0.2471, loss=0.2167, over 2343882.96 frames. utt_duration=1240 frames, utt_pad_proportion=0.05408, over 7567.49 utterances.], batch size: 48, lr: 8.51e-03, grad_scale: 8.0 2023-03-08 07:42:38,655 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48072.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:42:55,997 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=48083.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:43:12,986 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48094.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:43:28,635 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7774, 5.2260, 5.0275, 5.2133, 5.3173, 4.8473, 3.7584, 5.1625], device='cuda:1'), covar=tensor([0.0107, 0.0096, 0.0105, 0.0071, 0.0061, 0.0107, 0.0625, 0.0158], device='cuda:1'), in_proj_covar=tensor([0.0080, 0.0074, 0.0093, 0.0057, 0.0062, 0.0075, 0.0097, 0.0096], 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') 2023-03-08 07:43:31,448 INFO [train2.py:809] (1/4) Epoch 13, batch 300, loss[ctc_loss=0.08847, att_loss=0.2295, loss=0.2013, over 15759.00 frames. utt_duration=1661 frames, utt_pad_proportion=0.008468, over 38.00 utterances.], tot_loss[ctc_loss=0.09541, att_loss=0.2464, loss=0.2162, over 2547978.37 frames. utt_duration=1266 frames, utt_pad_proportion=0.04963, over 8060.69 utterances.], batch size: 38, lr: 8.51e-03, grad_scale: 8.0 2023-03-08 07:44:01,029 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.586e+02 2.171e+02 2.716e+02 3.489e+02 9.070e+02, threshold=5.431e+02, percent-clipped=6.0 2023-03-08 07:44:16,026 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48133.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:44:32,569 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48144.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:44:51,566 INFO [train2.py:809] (1/4) Epoch 13, batch 350, loss[ctc_loss=0.0802, att_loss=0.2349, loss=0.2039, over 16403.00 frames. utt_duration=1492 frames, utt_pad_proportion=0.007001, over 44.00 utterances.], tot_loss[ctc_loss=0.09626, att_loss=0.2468, loss=0.2167, over 2706969.92 frames. utt_duration=1230 frames, utt_pad_proportion=0.05816, over 8811.34 utterances.], batch size: 44, lr: 8.50e-03, grad_scale: 8.0 2023-03-08 07:45:04,057 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-03-08 07:45:06,288 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48165.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 07:45:35,908 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9158, 6.1337, 5.5607, 5.9159, 5.7920, 5.3628, 5.5500, 5.3687], device='cuda:1'), covar=tensor([0.1215, 0.0872, 0.0860, 0.0738, 0.0785, 0.1579, 0.2520, 0.2442], device='cuda:1'), in_proj_covar=tensor([0.0450, 0.0521, 0.0388, 0.0384, 0.0373, 0.0426, 0.0534, 0.0466], 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') 2023-03-08 07:46:10,722 INFO [train2.py:809] (1/4) Epoch 13, batch 400, loss[ctc_loss=0.07469, att_loss=0.2253, loss=0.1952, over 16008.00 frames. utt_duration=1603 frames, utt_pad_proportion=0.007048, over 40.00 utterances.], tot_loss[ctc_loss=0.09563, att_loss=0.2461, loss=0.216, over 2827736.10 frames. utt_duration=1240 frames, utt_pad_proportion=0.05843, over 9134.66 utterances.], batch size: 40, lr: 8.50e-03, grad_scale: 8.0 2023-03-08 07:46:21,616 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48212.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 07:46:40,273 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.388e+02 2.275e+02 2.732e+02 3.494e+02 6.607e+02, threshold=5.464e+02, percent-clipped=2.0 2023-03-08 07:46:48,654 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4565, 2.1756, 5.0127, 3.6167, 2.6578, 4.2807, 4.7659, 4.4627], device='cuda:1'), covar=tensor([0.0212, 0.1990, 0.0112, 0.1192, 0.2018, 0.0224, 0.0097, 0.0235], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0239, 0.0139, 0.0303, 0.0264, 0.0181, 0.0125, 0.0156], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 07:47:03,914 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-03-08 07:47:29,598 INFO [train2.py:809] (1/4) Epoch 13, batch 450, loss[ctc_loss=0.06216, att_loss=0.2149, loss=0.1844, over 15390.00 frames. utt_duration=1761 frames, utt_pad_proportion=0.009229, over 35.00 utterances.], tot_loss[ctc_loss=0.09459, att_loss=0.2455, loss=0.2153, over 2924318.56 frames. utt_duration=1261 frames, utt_pad_proportion=0.05395, over 9286.23 utterances.], batch size: 35, lr: 8.49e-03, grad_scale: 8.0 2023-03-08 07:47:56,914 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48273.0, num_to_drop=1, layers_to_drop={3} 2023-03-08 07:48:48,218 INFO [train2.py:809] (1/4) Epoch 13, batch 500, loss[ctc_loss=0.08844, att_loss=0.2382, loss=0.2083, over 16410.00 frames. utt_duration=1493 frames, utt_pad_proportion=0.007027, over 44.00 utterances.], tot_loss[ctc_loss=0.09522, att_loss=0.246, loss=0.2158, over 3002288.05 frames. utt_duration=1238 frames, utt_pad_proportion=0.05932, over 9712.95 utterances.], batch size: 44, lr: 8.49e-03, grad_scale: 8.0 2023-03-08 07:49:10,183 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.90 vs. limit=2.0 2023-03-08 07:49:16,720 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.612e+02 2.184e+02 2.674e+02 3.467e+02 7.635e+02, threshold=5.348e+02, percent-clipped=2.0 2023-03-08 07:49:17,124 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48324.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:50:07,390 INFO [train2.py:809] (1/4) Epoch 13, batch 550, loss[ctc_loss=0.1141, att_loss=0.2647, loss=0.2346, over 17117.00 frames. utt_duration=1224 frames, utt_pad_proportion=0.01523, over 56.00 utterances.], tot_loss[ctc_loss=0.09536, att_loss=0.2465, loss=0.2163, over 3066061.99 frames. utt_duration=1220 frames, utt_pad_proportion=0.06246, over 10064.09 utterances.], batch size: 56, lr: 8.49e-03, grad_scale: 8.0 2023-03-08 07:50:14,179 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7146, 4.7235, 4.7100, 4.7294, 5.3086, 4.8623, 4.7598, 2.5730], device='cuda:1'), covar=tensor([0.0196, 0.0281, 0.0221, 0.0239, 0.0729, 0.0173, 0.0244, 0.1941], device='cuda:1'), in_proj_covar=tensor([0.0128, 0.0138, 0.0141, 0.0154, 0.0340, 0.0124, 0.0127, 0.0215], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0002, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 07:50:54,948 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48385.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:51:08,855 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48394.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:51:13,513 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48397.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:51:27,705 INFO [train2.py:809] (1/4) Epoch 13, batch 600, loss[ctc_loss=0.0752, att_loss=0.2309, loss=0.1998, over 15971.00 frames. utt_duration=1560 frames, utt_pad_proportion=0.006016, over 41.00 utterances.], tot_loss[ctc_loss=0.0951, att_loss=0.2464, loss=0.2162, over 3116172.44 frames. utt_duration=1218 frames, utt_pad_proportion=0.06177, over 10250.29 utterances.], batch size: 41, lr: 8.48e-03, grad_scale: 8.0 2023-03-08 07:51:47,791 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.81 vs. limit=5.0 2023-03-08 07:51:56,294 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48423.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:51:57,442 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.401e+02 2.243e+02 2.721e+02 3.200e+02 7.268e+02, threshold=5.442e+02, percent-clipped=2.0 2023-03-08 07:52:04,398 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48428.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:52:25,659 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=48442.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:52:28,968 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48444.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:52:29,404 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.05 vs. limit=5.0 2023-03-08 07:52:47,830 INFO [train2.py:809] (1/4) Epoch 13, batch 650, loss[ctc_loss=0.08341, att_loss=0.2459, loss=0.2134, over 16385.00 frames. utt_duration=1492 frames, utt_pad_proportion=0.007561, over 44.00 utterances.], tot_loss[ctc_loss=0.09497, att_loss=0.2469, loss=0.2165, over 3159070.67 frames. utt_duration=1215 frames, utt_pad_proportion=0.06021, over 10409.83 utterances.], batch size: 44, lr: 8.48e-03, grad_scale: 8.0 2023-03-08 07:52:51,845 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48458.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:53:03,108 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48465.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 07:53:33,519 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48484.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:53:45,421 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=48492.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:54:07,435 INFO [train2.py:809] (1/4) Epoch 13, batch 700, loss[ctc_loss=0.08956, att_loss=0.2298, loss=0.2018, over 15955.00 frames. utt_duration=1558 frames, utt_pad_proportion=0.006222, over 41.00 utterances.], tot_loss[ctc_loss=0.09527, att_loss=0.2469, loss=0.2166, over 3187372.76 frames. utt_duration=1219 frames, utt_pad_proportion=0.0587, over 10475.01 utterances.], batch size: 41, lr: 8.47e-03, grad_scale: 8.0 2023-03-08 07:54:19,503 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=48513.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 07:54:36,145 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.318e+02 2.306e+02 2.831e+02 3.598e+02 7.957e+02, threshold=5.663e+02, percent-clipped=6.0 2023-03-08 07:55:13,380 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.13 vs. limit=5.0 2023-03-08 07:55:26,624 INFO [train2.py:809] (1/4) Epoch 13, batch 750, loss[ctc_loss=0.08492, att_loss=0.226, loss=0.1978, over 16011.00 frames. utt_duration=1603 frames, utt_pad_proportion=0.007709, over 40.00 utterances.], tot_loss[ctc_loss=0.09466, att_loss=0.2463, loss=0.216, over 3206957.03 frames. utt_duration=1228 frames, utt_pad_proportion=0.05748, over 10463.04 utterances.], batch size: 40, lr: 8.47e-03, grad_scale: 8.0 2023-03-08 07:55:46,618 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48568.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 07:56:46,006 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.7743, 5.0259, 5.2729, 5.1979, 5.2628, 5.7550, 5.0699, 5.8673], device='cuda:1'), covar=tensor([0.0688, 0.0684, 0.0758, 0.1097, 0.1703, 0.0831, 0.0681, 0.0626], device='cuda:1'), in_proj_covar=tensor([0.0744, 0.0441, 0.0512, 0.0572, 0.0763, 0.0519, 0.0416, 0.0510], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 07:56:47,920 INFO [train2.py:809] (1/4) Epoch 13, batch 800, loss[ctc_loss=0.1198, att_loss=0.2692, loss=0.2393, over 17098.00 frames. utt_duration=1223 frames, utt_pad_proportion=0.01629, over 56.00 utterances.], tot_loss[ctc_loss=0.09544, att_loss=0.2472, loss=0.2168, over 3232834.12 frames. utt_duration=1237 frames, utt_pad_proportion=0.05192, over 10466.86 utterances.], batch size: 56, lr: 8.46e-03, grad_scale: 8.0 2023-03-08 07:57:16,742 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.553e+02 2.319e+02 2.822e+02 3.507e+02 7.468e+02, threshold=5.644e+02, percent-clipped=2.0 2023-03-08 07:58:07,624 INFO [train2.py:809] (1/4) Epoch 13, batch 850, loss[ctc_loss=0.07531, att_loss=0.2259, loss=0.1958, over 15376.00 frames. utt_duration=1759 frames, utt_pad_proportion=0.01077, over 35.00 utterances.], tot_loss[ctc_loss=0.09457, att_loss=0.2463, loss=0.2159, over 3233749.73 frames. utt_duration=1245 frames, utt_pad_proportion=0.05335, over 10399.45 utterances.], batch size: 35, lr: 8.46e-03, grad_scale: 8.0 2023-03-08 07:58:18,027 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0428, 2.4636, 3.3879, 2.5032, 3.2918, 4.1913, 3.9966, 3.0283], device='cuda:1'), covar=tensor([0.0464, 0.1938, 0.1112, 0.1568, 0.1045, 0.0912, 0.0539, 0.1361], device='cuda:1'), in_proj_covar=tensor([0.0231, 0.0234, 0.0252, 0.0207, 0.0245, 0.0319, 0.0228, 0.0225], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 07:58:46,032 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48680.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 07:59:28,741 INFO [train2.py:809] (1/4) Epoch 13, batch 900, loss[ctc_loss=0.09378, att_loss=0.2524, loss=0.2207, over 17307.00 frames. utt_duration=1175 frames, utt_pad_proportion=0.02428, over 59.00 utterances.], tot_loss[ctc_loss=0.09492, att_loss=0.2467, loss=0.2163, over 3241152.01 frames. utt_duration=1225 frames, utt_pad_proportion=0.05931, over 10592.29 utterances.], batch size: 59, lr: 8.45e-03, grad_scale: 8.0 2023-03-08 07:59:56,781 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.412e+02 2.153e+02 2.503e+02 3.048e+02 6.271e+02, threshold=5.007e+02, percent-clipped=1.0 2023-03-08 08:00:04,000 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48728.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:00:43,574 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48753.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:00:49,762 INFO [train2.py:809] (1/4) Epoch 13, batch 950, loss[ctc_loss=0.07034, att_loss=0.2122, loss=0.1839, over 15635.00 frames. utt_duration=1692 frames, utt_pad_proportion=0.009057, over 37.00 utterances.], tot_loss[ctc_loss=0.09524, att_loss=0.247, loss=0.2167, over 3241821.50 frames. utt_duration=1220 frames, utt_pad_proportion=0.06276, over 10643.19 utterances.], batch size: 37, lr: 8.45e-03, grad_scale: 8.0 2023-03-08 08:01:00,399 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-03-08 08:01:16,739 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4087, 2.7181, 3.5004, 2.7315, 3.4688, 4.4894, 4.2300, 3.2580], device='cuda:1'), covar=tensor([0.0396, 0.1975, 0.1240, 0.1399, 0.1069, 0.0762, 0.0587, 0.1269], device='cuda:1'), in_proj_covar=tensor([0.0227, 0.0231, 0.0249, 0.0202, 0.0243, 0.0314, 0.0225, 0.0221], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 08:01:21,801 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=48776.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:01:26,501 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48779.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:02:10,410 INFO [train2.py:809] (1/4) Epoch 13, batch 1000, loss[ctc_loss=0.08718, att_loss=0.2306, loss=0.2019, over 15996.00 frames. utt_duration=1601 frames, utt_pad_proportion=0.0079, over 40.00 utterances.], tot_loss[ctc_loss=0.09645, att_loss=0.2479, loss=0.2176, over 3248633.15 frames. utt_duration=1211 frames, utt_pad_proportion=0.06388, over 10745.52 utterances.], batch size: 40, lr: 8.45e-03, grad_scale: 8.0 2023-03-08 08:02:33,386 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48821.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:02:37,582 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.313e+02 2.228e+02 2.748e+02 3.245e+02 7.781e+02, threshold=5.496e+02, percent-clipped=5.0 2023-03-08 08:03:00,453 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48838.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:03:30,179 INFO [train2.py:809] (1/4) Epoch 13, batch 1050, loss[ctc_loss=0.09237, att_loss=0.2465, loss=0.2156, over 16408.00 frames. utt_duration=1493 frames, utt_pad_proportion=0.007238, over 44.00 utterances.], tot_loss[ctc_loss=0.0961, att_loss=0.248, loss=0.2176, over 3255514.55 frames. utt_duration=1222 frames, utt_pad_proportion=0.06201, over 10665.16 utterances.], batch size: 44, lr: 8.44e-03, grad_scale: 8.0 2023-03-08 08:03:49,023 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48868.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 08:04:11,831 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48882.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:04:39,019 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48899.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:04:50,179 INFO [train2.py:809] (1/4) Epoch 13, batch 1100, loss[ctc_loss=0.08869, att_loss=0.2325, loss=0.2037, over 16163.00 frames. utt_duration=1578 frames, utt_pad_proportion=0.007987, over 41.00 utterances.], tot_loss[ctc_loss=0.09676, att_loss=0.2481, loss=0.2178, over 3254656.93 frames. utt_duration=1213 frames, utt_pad_proportion=0.06608, over 10748.57 utterances.], batch size: 41, lr: 8.44e-03, grad_scale: 8.0 2023-03-08 08:05:05,418 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=48916.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 08:05:18,175 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.628e+02 2.462e+02 3.131e+02 3.917e+02 9.784e+02, threshold=6.261e+02, percent-clipped=6.0 2023-03-08 08:06:09,391 INFO [train2.py:809] (1/4) Epoch 13, batch 1150, loss[ctc_loss=0.08519, att_loss=0.2228, loss=0.1953, over 16017.00 frames. utt_duration=1603 frames, utt_pad_proportion=0.007152, over 40.00 utterances.], tot_loss[ctc_loss=0.09692, att_loss=0.2478, loss=0.2176, over 3256168.81 frames. utt_duration=1207 frames, utt_pad_proportion=0.06757, over 10804.32 utterances.], batch size: 40, lr: 8.43e-03, grad_scale: 8.0 2023-03-08 08:06:14,369 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7695, 3.0803, 3.7495, 3.1251, 3.7252, 4.8618, 4.6379, 3.6845], device='cuda:1'), covar=tensor([0.0339, 0.1683, 0.0992, 0.1198, 0.0898, 0.0616, 0.0463, 0.1116], device='cuda:1'), in_proj_covar=tensor([0.0233, 0.0237, 0.0256, 0.0207, 0.0249, 0.0323, 0.0230, 0.0228], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 08:06:43,513 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.92 vs. limit=5.0 2023-03-08 08:06:47,193 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48980.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:07:28,989 INFO [train2.py:809] (1/4) Epoch 13, batch 1200, loss[ctc_loss=0.1093, att_loss=0.2651, loss=0.2339, over 17342.00 frames. utt_duration=1103 frames, utt_pad_proportion=0.0362, over 63.00 utterances.], tot_loss[ctc_loss=0.0964, att_loss=0.2479, loss=0.2176, over 3272892.07 frames. utt_duration=1233 frames, utt_pad_proportion=0.0574, over 10628.15 utterances.], batch size: 63, lr: 8.43e-03, grad_scale: 8.0 2023-03-08 08:07:51,078 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9142, 5.2082, 4.6815, 5.3118, 4.6212, 5.0001, 5.3477, 5.1162], device='cuda:1'), covar=tensor([0.0528, 0.0295, 0.0904, 0.0230, 0.0443, 0.0192, 0.0201, 0.0188], device='cuda:1'), in_proj_covar=tensor([0.0339, 0.0270, 0.0324, 0.0269, 0.0276, 0.0208, 0.0253, 0.0243], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 08:07:57,676 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.513e+02 2.494e+02 2.872e+02 3.566e+02 1.058e+03, threshold=5.743e+02, percent-clipped=4.0 2023-03-08 08:08:04,025 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=49028.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:08:44,307 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49053.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:08:48,630 INFO [train2.py:809] (1/4) Epoch 13, batch 1250, loss[ctc_loss=0.09357, att_loss=0.2556, loss=0.2232, over 16626.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.005349, over 47.00 utterances.], tot_loss[ctc_loss=0.09633, att_loss=0.2478, loss=0.2175, over 3273774.26 frames. utt_duration=1225 frames, utt_pad_proportion=0.06003, over 10702.57 utterances.], batch size: 47, lr: 8.42e-03, grad_scale: 8.0 2023-03-08 08:09:25,458 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49079.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:10:01,128 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=49101.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:10:08,582 INFO [train2.py:809] (1/4) Epoch 13, batch 1300, loss[ctc_loss=0.06597, att_loss=0.206, loss=0.178, over 15504.00 frames. utt_duration=1724 frames, utt_pad_proportion=0.008402, over 36.00 utterances.], tot_loss[ctc_loss=0.09559, att_loss=0.2473, loss=0.2169, over 3268308.90 frames. utt_duration=1224 frames, utt_pad_proportion=0.06142, over 10696.05 utterances.], batch size: 36, lr: 8.42e-03, grad_scale: 8.0 2023-03-08 08:10:23,396 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-03-08 08:10:36,806 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.295e+02 2.317e+02 2.735e+02 3.523e+02 8.419e+02, threshold=5.469e+02, percent-clipped=3.0 2023-03-08 08:10:41,555 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=49127.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:10:47,750 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0974, 4.5881, 4.4751, 4.7200, 2.5537, 4.3327, 2.8455, 2.1277], device='cuda:1'), covar=tensor([0.0313, 0.0166, 0.0657, 0.0132, 0.1882, 0.0198, 0.1458, 0.1654], device='cuda:1'), in_proj_covar=tensor([0.0153, 0.0124, 0.0259, 0.0119, 0.0224, 0.0114, 0.0228, 0.0202], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 08:11:28,001 INFO [train2.py:809] (1/4) Epoch 13, batch 1350, loss[ctc_loss=0.09942, att_loss=0.2275, loss=0.2019, over 16173.00 frames. utt_duration=1580 frames, utt_pad_proportion=0.006581, over 41.00 utterances.], tot_loss[ctc_loss=0.09429, att_loss=0.2459, loss=0.2156, over 3269455.54 frames. utt_duration=1243 frames, utt_pad_proportion=0.05712, over 10537.51 utterances.], batch size: 41, lr: 8.42e-03, grad_scale: 8.0 2023-03-08 08:12:01,179 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49177.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:12:29,065 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49194.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:12:37,889 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-03-08 08:12:47,667 INFO [train2.py:809] (1/4) Epoch 13, batch 1400, loss[ctc_loss=0.09475, att_loss=0.2584, loss=0.2256, over 16486.00 frames. utt_duration=1435 frames, utt_pad_proportion=0.006112, over 46.00 utterances.], tot_loss[ctc_loss=0.09455, att_loss=0.2462, loss=0.2159, over 3270841.69 frames. utt_duration=1228 frames, utt_pad_proportion=0.05968, over 10671.42 utterances.], batch size: 46, lr: 8.41e-03, grad_scale: 8.0 2023-03-08 08:13:15,976 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.537e+02 2.165e+02 2.664e+02 3.367e+02 6.809e+02, threshold=5.329e+02, percent-clipped=2.0 2023-03-08 08:13:35,667 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49236.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:13:45,295 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.1817, 6.3353, 5.8876, 6.1028, 6.0070, 5.6094, 5.8262, 5.7019], device='cuda:1'), covar=tensor([0.1138, 0.0760, 0.0701, 0.0779, 0.0822, 0.1537, 0.1947, 0.1985], device='cuda:1'), in_proj_covar=tensor([0.0458, 0.0531, 0.0394, 0.0397, 0.0374, 0.0426, 0.0539, 0.0476], 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') 2023-03-08 08:14:07,334 INFO [train2.py:809] (1/4) Epoch 13, batch 1450, loss[ctc_loss=0.09259, att_loss=0.2557, loss=0.2231, over 16628.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.005274, over 47.00 utterances.], tot_loss[ctc_loss=0.09486, att_loss=0.247, loss=0.2166, over 3277948.95 frames. utt_duration=1241 frames, utt_pad_proportion=0.05482, over 10581.18 utterances.], batch size: 47, lr: 8.41e-03, grad_scale: 8.0 2023-03-08 08:14:56,268 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.8115, 6.0634, 5.4481, 5.8672, 5.7269, 5.3988, 5.4692, 5.2367], device='cuda:1'), covar=tensor([0.1249, 0.0856, 0.0847, 0.0700, 0.0790, 0.1373, 0.2332, 0.2258], device='cuda:1'), in_proj_covar=tensor([0.0456, 0.0532, 0.0395, 0.0396, 0.0374, 0.0428, 0.0540, 0.0476], 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') 2023-03-08 08:15:13,299 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49297.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:15:26,866 INFO [train2.py:809] (1/4) Epoch 13, batch 1500, loss[ctc_loss=0.1513, att_loss=0.2697, loss=0.246, over 13886.00 frames. utt_duration=384.6 frames, utt_pad_proportion=0.3323, over 145.00 utterances.], tot_loss[ctc_loss=0.09428, att_loss=0.2461, loss=0.2158, over 3268147.62 frames. utt_duration=1241 frames, utt_pad_proportion=0.05676, over 10544.18 utterances.], batch size: 145, lr: 8.40e-03, grad_scale: 8.0 2023-03-08 08:15:55,304 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.604e+02 2.233e+02 2.652e+02 3.112e+02 8.079e+02, threshold=5.303e+02, percent-clipped=1.0 2023-03-08 08:16:47,515 INFO [train2.py:809] (1/4) Epoch 13, batch 1550, loss[ctc_loss=0.1022, att_loss=0.2598, loss=0.2283, over 17425.00 frames. utt_duration=1012 frames, utt_pad_proportion=0.04648, over 69.00 utterances.], tot_loss[ctc_loss=0.09405, att_loss=0.2458, loss=0.2154, over 3265981.80 frames. utt_duration=1248 frames, utt_pad_proportion=0.05573, over 10479.62 utterances.], batch size: 69, lr: 8.40e-03, grad_scale: 8.0 2023-03-08 08:17:31,567 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-03-08 08:17:42,742 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.42 vs. limit=5.0 2023-03-08 08:18:08,485 INFO [train2.py:809] (1/4) Epoch 13, batch 1600, loss[ctc_loss=0.08699, att_loss=0.2594, loss=0.2249, over 17081.00 frames. utt_duration=1291 frames, utt_pad_proportion=0.008034, over 53.00 utterances.], tot_loss[ctc_loss=0.09503, att_loss=0.2463, loss=0.2161, over 3267853.76 frames. utt_duration=1236 frames, utt_pad_proportion=0.05921, over 10590.31 utterances.], batch size: 53, lr: 8.40e-03, grad_scale: 8.0 2023-03-08 08:18:34,209 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49422.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:18:36,893 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.458e+02 2.327e+02 2.714e+02 3.320e+02 9.456e+02, threshold=5.428e+02, percent-clipped=3.0 2023-03-08 08:18:38,850 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.5755, 2.7295, 3.4013, 4.4231, 3.8453, 4.0016, 3.0408, 2.1232], device='cuda:1'), covar=tensor([0.0554, 0.2044, 0.0921, 0.0543, 0.0807, 0.0372, 0.1390, 0.2362], device='cuda:1'), in_proj_covar=tensor([0.0169, 0.0207, 0.0186, 0.0195, 0.0192, 0.0159, 0.0193, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 08:18:50,256 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49432.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 08:19:28,468 INFO [train2.py:809] (1/4) Epoch 13, batch 1650, loss[ctc_loss=0.1012, att_loss=0.233, loss=0.2066, over 16018.00 frames. utt_duration=1603 frames, utt_pad_proportion=0.007198, over 40.00 utterances.], tot_loss[ctc_loss=0.09444, att_loss=0.246, loss=0.2157, over 3264444.35 frames. utt_duration=1217 frames, utt_pad_proportion=0.06538, over 10740.27 utterances.], batch size: 40, lr: 8.39e-03, grad_scale: 8.0 2023-03-08 08:20:03,213 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49477.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:20:12,596 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49483.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:20:14,638 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.6292, 2.7331, 3.6956, 4.6475, 4.0241, 4.0143, 3.0421, 2.1649], device='cuda:1'), covar=tensor([0.0630, 0.2127, 0.0869, 0.0436, 0.0715, 0.0377, 0.1416, 0.2315], device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0207, 0.0187, 0.0195, 0.0193, 0.0159, 0.0195, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 08:20:18,952 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.79 vs. limit=2.0 2023-03-08 08:20:29,278 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49493.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 08:20:30,721 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49494.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:20:32,471 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5401, 2.0468, 4.9759, 3.7480, 2.7617, 4.2483, 4.6641, 4.4408], device='cuda:1'), covar=tensor([0.0247, 0.2065, 0.0153, 0.1064, 0.1985, 0.0254, 0.0127, 0.0290], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0244, 0.0144, 0.0309, 0.0270, 0.0186, 0.0131, 0.0160], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 08:20:48,666 INFO [train2.py:809] (1/4) Epoch 13, batch 1700, loss[ctc_loss=0.1451, att_loss=0.2798, loss=0.2528, over 16769.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.006555, over 48.00 utterances.], tot_loss[ctc_loss=0.09461, att_loss=0.2461, loss=0.2158, over 3265342.75 frames. utt_duration=1227 frames, utt_pad_proportion=0.06335, over 10659.15 utterances.], batch size: 48, lr: 8.39e-03, grad_scale: 8.0 2023-03-08 08:21:16,982 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.458e+02 2.324e+02 2.788e+02 3.752e+02 8.319e+02, threshold=5.576e+02, percent-clipped=5.0 2023-03-08 08:21:18,734 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=49525.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:21:32,308 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6620, 5.2710, 5.0303, 5.1895, 5.2359, 4.8407, 3.7475, 5.1976], device='cuda:1'), covar=tensor([0.0106, 0.0085, 0.0109, 0.0070, 0.0085, 0.0098, 0.0629, 0.0170], device='cuda:1'), in_proj_covar=tensor([0.0078, 0.0073, 0.0091, 0.0057, 0.0062, 0.0072, 0.0093, 0.0094], 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') 2023-03-08 08:21:46,000 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=49542.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:21:49,181 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.0882, 3.8718, 3.2586, 3.6378, 4.0303, 3.6464, 2.8568, 4.3972], device='cuda:1'), covar=tensor([0.0999, 0.0463, 0.1051, 0.0567, 0.0675, 0.0644, 0.0895, 0.0395], device='cuda:1'), in_proj_covar=tensor([0.0191, 0.0194, 0.0212, 0.0181, 0.0248, 0.0221, 0.0188, 0.0261], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-08 08:22:07,062 INFO [train2.py:809] (1/4) Epoch 13, batch 1750, loss[ctc_loss=0.1045, att_loss=0.2585, loss=0.2277, over 17104.00 frames. utt_duration=692.7 frames, utt_pad_proportion=0.132, over 99.00 utterances.], tot_loss[ctc_loss=0.09478, att_loss=0.246, loss=0.2158, over 3260470.35 frames. utt_duration=1218 frames, utt_pad_proportion=0.06608, over 10724.57 utterances.], batch size: 99, lr: 8.38e-03, grad_scale: 8.0 2023-03-08 08:22:43,634 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7247, 5.0241, 4.2217, 5.1853, 4.5486, 4.8081, 5.1045, 4.9096], device='cuda:1'), covar=tensor([0.0588, 0.0334, 0.1370, 0.0300, 0.0437, 0.0288, 0.0317, 0.0248], device='cuda:1'), in_proj_covar=tensor([0.0343, 0.0271, 0.0330, 0.0275, 0.0279, 0.0210, 0.0259, 0.0245], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0005, 0.0006, 0.0005, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 08:23:03,831 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49592.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:23:25,898 INFO [train2.py:809] (1/4) Epoch 13, batch 1800, loss[ctc_loss=0.08417, att_loss=0.2201, loss=0.1929, over 15495.00 frames. utt_duration=1723 frames, utt_pad_proportion=0.009121, over 36.00 utterances.], tot_loss[ctc_loss=0.09507, att_loss=0.2463, loss=0.2161, over 3265305.24 frames. utt_duration=1216 frames, utt_pad_proportion=0.06626, over 10754.84 utterances.], batch size: 36, lr: 8.38e-03, grad_scale: 8.0 2023-03-08 08:23:54,567 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.478e+02 2.431e+02 2.885e+02 3.492e+02 8.603e+02, threshold=5.770e+02, percent-clipped=6.0 2023-03-08 08:24:45,025 INFO [train2.py:809] (1/4) Epoch 13, batch 1850, loss[ctc_loss=0.08971, att_loss=0.2241, loss=0.1972, over 15487.00 frames. utt_duration=1722 frames, utt_pad_proportion=0.009568, over 36.00 utterances.], tot_loss[ctc_loss=0.09429, att_loss=0.2458, loss=0.2155, over 3266200.27 frames. utt_duration=1256 frames, utt_pad_proportion=0.05641, over 10416.63 utterances.], batch size: 36, lr: 8.37e-03, grad_scale: 8.0 2023-03-08 08:25:28,019 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3859, 4.8276, 4.6778, 4.8615, 5.0239, 4.5171, 3.5805, 4.7959], device='cuda:1'), covar=tensor([0.0110, 0.0111, 0.0136, 0.0084, 0.0076, 0.0103, 0.0633, 0.0193], device='cuda:1'), in_proj_covar=tensor([0.0078, 0.0074, 0.0092, 0.0057, 0.0062, 0.0072, 0.0094, 0.0094], 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') 2023-03-08 08:25:46,042 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.2338, 5.4796, 5.4513, 5.3984, 5.5102, 5.4836, 5.2410, 4.9667], device='cuda:1'), covar=tensor([0.1044, 0.0468, 0.0226, 0.0488, 0.0268, 0.0288, 0.0298, 0.0298], device='cuda:1'), in_proj_covar=tensor([0.0468, 0.0303, 0.0269, 0.0296, 0.0356, 0.0376, 0.0304, 0.0338], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 08:26:04,420 INFO [train2.py:809] (1/4) Epoch 13, batch 1900, loss[ctc_loss=0.1148, att_loss=0.2633, loss=0.2336, over 16494.00 frames. utt_duration=1436 frames, utt_pad_proportion=0.005676, over 46.00 utterances.], tot_loss[ctc_loss=0.09386, att_loss=0.2454, loss=0.2151, over 3257473.81 frames. utt_duration=1278 frames, utt_pad_proportion=0.05086, over 10207.60 utterances.], batch size: 46, lr: 8.37e-03, grad_scale: 8.0 2023-03-08 08:26:33,119 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.532e+02 2.157e+02 2.753e+02 3.358e+02 9.197e+02, threshold=5.506e+02, percent-clipped=6.0 2023-03-08 08:27:24,890 INFO [train2.py:809] (1/4) Epoch 13, batch 1950, loss[ctc_loss=0.08416, att_loss=0.2436, loss=0.2117, over 16614.00 frames. utt_duration=1415 frames, utt_pad_proportion=0.006126, over 47.00 utterances.], tot_loss[ctc_loss=0.09422, att_loss=0.246, loss=0.2156, over 3260586.60 frames. utt_duration=1276 frames, utt_pad_proportion=0.04971, over 10236.60 utterances.], batch size: 47, lr: 8.37e-03, grad_scale: 16.0 2023-03-08 08:27:41,114 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49766.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:27:59,751 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49778.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:28:17,536 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49788.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 08:28:39,334 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-03-08 08:28:45,895 INFO [train2.py:809] (1/4) Epoch 13, batch 2000, loss[ctc_loss=0.08899, att_loss=0.23, loss=0.2018, over 11485.00 frames. utt_duration=1839 frames, utt_pad_proportion=0.183, over 25.00 utterances.], tot_loss[ctc_loss=0.09387, att_loss=0.2453, loss=0.215, over 3257543.84 frames. utt_duration=1303 frames, utt_pad_proportion=0.04391, over 10011.63 utterances.], batch size: 25, lr: 8.36e-03, grad_scale: 16.0 2023-03-08 08:29:14,829 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.359e+02 2.246e+02 2.552e+02 3.168e+02 7.222e+02, threshold=5.105e+02, percent-clipped=2.0 2023-03-08 08:29:19,813 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49827.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:30:05,632 INFO [train2.py:809] (1/4) Epoch 13, batch 2050, loss[ctc_loss=0.1005, att_loss=0.2308, loss=0.2047, over 15876.00 frames. utt_duration=1630 frames, utt_pad_proportion=0.009845, over 39.00 utterances.], tot_loss[ctc_loss=0.09436, att_loss=0.2455, loss=0.2153, over 3257295.76 frames. utt_duration=1283 frames, utt_pad_proportion=0.04991, over 10164.53 utterances.], batch size: 39, lr: 8.36e-03, grad_scale: 16.0 2023-03-08 08:31:03,767 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49892.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:31:25,304 INFO [train2.py:809] (1/4) Epoch 13, batch 2100, loss[ctc_loss=0.0957, att_loss=0.2389, loss=0.2102, over 16289.00 frames. utt_duration=1516 frames, utt_pad_proportion=0.005642, over 43.00 utterances.], tot_loss[ctc_loss=0.09592, att_loss=0.2466, loss=0.2164, over 3265037.70 frames. utt_duration=1236 frames, utt_pad_proportion=0.05813, over 10583.07 utterances.], batch size: 43, lr: 8.35e-03, grad_scale: 16.0 2023-03-08 08:31:53,560 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.781e+02 2.361e+02 2.979e+02 3.749e+02 1.127e+03, threshold=5.957e+02, percent-clipped=6.0 2023-03-08 08:32:08,317 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5588, 4.9987, 4.8003, 4.9613, 5.0538, 4.6081, 3.8164, 4.9465], device='cuda:1'), covar=tensor([0.0103, 0.0103, 0.0127, 0.0076, 0.0091, 0.0116, 0.0534, 0.0199], device='cuda:1'), in_proj_covar=tensor([0.0078, 0.0074, 0.0092, 0.0057, 0.0063, 0.0073, 0.0094, 0.0095], 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') 2023-03-08 08:32:19,979 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=49940.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:32:44,498 INFO [train2.py:809] (1/4) Epoch 13, batch 2150, loss[ctc_loss=0.07612, att_loss=0.2277, loss=0.1974, over 16179.00 frames. utt_duration=1580 frames, utt_pad_proportion=0.006776, over 41.00 utterances.], tot_loss[ctc_loss=0.09624, att_loss=0.2469, loss=0.2168, over 3262632.52 frames. utt_duration=1214 frames, utt_pad_proportion=0.06459, over 10762.50 utterances.], batch size: 41, lr: 8.35e-03, grad_scale: 16.0 2023-03-08 08:33:03,774 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.9245, 1.7326, 2.4703, 2.6520, 2.6913, 2.6858, 2.4912, 2.8231], device='cuda:1'), covar=tensor([0.1384, 0.5784, 0.3840, 0.1417, 0.1760, 0.1486, 0.2949, 0.0993], device='cuda:1'), in_proj_covar=tensor([0.0085, 0.0094, 0.0097, 0.0080, 0.0087, 0.0075, 0.0095, 0.0065], device='cuda:1'), out_proj_covar=tensor([5.9242e-05, 6.6963e-05, 6.9282e-05, 5.7555e-05, 5.9209e-05, 5.6446e-05, 6.6945e-05, 5.0167e-05], device='cuda:1') 2023-03-08 08:34:08,271 INFO [train2.py:809] (1/4) Epoch 13, batch 2200, loss[ctc_loss=0.07895, att_loss=0.2469, loss=0.2133, over 17297.00 frames. utt_duration=1259 frames, utt_pad_proportion=0.01087, over 55.00 utterances.], tot_loss[ctc_loss=0.09582, att_loss=0.2466, loss=0.2164, over 3270549.69 frames. utt_duration=1223 frames, utt_pad_proportion=0.06176, over 10709.52 utterances.], batch size: 55, lr: 8.35e-03, grad_scale: 16.0 2023-03-08 08:34:35,763 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.558e+02 2.313e+02 2.828e+02 3.809e+02 8.151e+02, threshold=5.657e+02, percent-clipped=5.0 2023-03-08 08:35:25,943 INFO [train2.py:809] (1/4) Epoch 13, batch 2250, loss[ctc_loss=0.09809, att_loss=0.2559, loss=0.2243, over 16626.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.005289, over 47.00 utterances.], tot_loss[ctc_loss=0.0964, att_loss=0.2472, loss=0.217, over 3272188.88 frames. utt_duration=1204 frames, utt_pad_proportion=0.06663, over 10885.43 utterances.], batch size: 47, lr: 8.34e-03, grad_scale: 16.0 2023-03-08 08:36:00,902 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50078.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:36:17,715 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50088.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 08:36:45,462 INFO [train2.py:809] (1/4) Epoch 13, batch 2300, loss[ctc_loss=0.1083, att_loss=0.2391, loss=0.2129, over 16417.00 frames. utt_duration=1494 frames, utt_pad_proportion=0.00683, over 44.00 utterances.], tot_loss[ctc_loss=0.09701, att_loss=0.2475, loss=0.2174, over 3261630.34 frames. utt_duration=1217 frames, utt_pad_proportion=0.0654, over 10732.68 utterances.], batch size: 44, lr: 8.34e-03, grad_scale: 8.0 2023-03-08 08:37:10,662 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50122.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:37:15,134 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.321e+02 2.441e+02 2.731e+02 3.266e+02 6.421e+02, threshold=5.462e+02, percent-clipped=3.0 2023-03-08 08:37:17,315 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=50126.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:37:20,760 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50128.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:37:34,135 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=50136.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 08:38:04,744 INFO [train2.py:809] (1/4) Epoch 13, batch 2350, loss[ctc_loss=0.08745, att_loss=0.237, loss=0.2071, over 16329.00 frames. utt_duration=1453 frames, utt_pad_proportion=0.006247, over 45.00 utterances.], tot_loss[ctc_loss=0.09622, att_loss=0.2469, loss=0.2168, over 3264564.81 frames. utt_duration=1245 frames, utt_pad_proportion=0.05693, over 10501.59 utterances.], batch size: 45, lr: 8.33e-03, grad_scale: 8.0 2023-03-08 08:38:24,136 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.7341, 5.0089, 5.3003, 5.2010, 5.1770, 5.6455, 5.1088, 5.7555], device='cuda:1'), covar=tensor([0.0683, 0.0726, 0.0731, 0.1194, 0.1872, 0.0878, 0.0741, 0.0652], device='cuda:1'), in_proj_covar=tensor([0.0758, 0.0448, 0.0532, 0.0586, 0.0772, 0.0529, 0.0428, 0.0526], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 08:38:53,851 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0649, 5.3371, 5.6541, 5.4741, 5.4899, 6.0031, 5.3401, 6.1077], device='cuda:1'), covar=tensor([0.0691, 0.0697, 0.0687, 0.1209, 0.1901, 0.0871, 0.0577, 0.0608], device='cuda:1'), in_proj_covar=tensor([0.0761, 0.0449, 0.0532, 0.0587, 0.0774, 0.0530, 0.0429, 0.0526], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 08:38:58,507 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50189.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:39:10,483 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50197.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 08:39:24,034 INFO [train2.py:809] (1/4) Epoch 13, batch 2400, loss[ctc_loss=0.1086, att_loss=0.267, loss=0.2353, over 17201.00 frames. utt_duration=696.4 frames, utt_pad_proportion=0.1251, over 99.00 utterances.], tot_loss[ctc_loss=0.09597, att_loss=0.2461, loss=0.2161, over 3262977.86 frames. utt_duration=1234 frames, utt_pad_proportion=0.05986, over 10590.59 utterances.], batch size: 99, lr: 8.33e-03, grad_scale: 8.0 2023-03-08 08:39:27,339 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9973, 5.2200, 5.5566, 5.4068, 5.4330, 5.9751, 5.1963, 6.0623], device='cuda:1'), covar=tensor([0.0672, 0.0677, 0.0706, 0.1137, 0.1727, 0.0808, 0.0636, 0.0570], device='cuda:1'), in_proj_covar=tensor([0.0759, 0.0448, 0.0529, 0.0585, 0.0771, 0.0529, 0.0429, 0.0523], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 08:39:54,444 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.486e+02 2.261e+02 2.728e+02 3.440e+02 5.730e+02, threshold=5.455e+02, percent-clipped=1.0 2023-03-08 08:40:42,742 INFO [train2.py:809] (1/4) Epoch 13, batch 2450, loss[ctc_loss=0.1143, att_loss=0.2557, loss=0.2274, over 17209.00 frames. utt_duration=696.7 frames, utt_pad_proportion=0.1258, over 99.00 utterances.], tot_loss[ctc_loss=0.09629, att_loss=0.2467, loss=0.2166, over 3274110.14 frames. utt_duration=1228 frames, utt_pad_proportion=0.05745, over 10676.91 utterances.], batch size: 99, lr: 8.32e-03, grad_scale: 8.0 2023-03-08 08:40:46,199 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50258.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 08:40:51,817 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-03-08 08:40:57,563 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0029, 4.5522, 4.6458, 2.5023, 2.0833, 2.6058, 2.4606, 3.6031], device='cuda:1'), covar=tensor([0.0632, 0.0204, 0.0215, 0.3439, 0.5567, 0.2791, 0.2592, 0.1688], device='cuda:1'), in_proj_covar=tensor([0.0342, 0.0233, 0.0241, 0.0219, 0.0350, 0.0333, 0.0236, 0.0356], device='cuda:1'), out_proj_covar=tensor([1.5101e-04, 8.5999e-05, 1.0439e-04, 9.5786e-05, 1.5012e-04, 1.3337e-04, 9.3454e-05, 1.4854e-04], device='cuda:1') 2023-03-08 08:41:35,795 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-03-08 08:42:02,224 INFO [train2.py:809] (1/4) Epoch 13, batch 2500, loss[ctc_loss=0.09004, att_loss=0.2526, loss=0.2201, over 17343.00 frames. utt_duration=879.6 frames, utt_pad_proportion=0.07803, over 79.00 utterances.], tot_loss[ctc_loss=0.09638, att_loss=0.2469, loss=0.2168, over 3274095.89 frames. utt_duration=1210 frames, utt_pad_proportion=0.06189, over 10833.40 utterances.], batch size: 79, lr: 8.32e-03, grad_scale: 8.0 2023-03-08 08:42:33,321 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.348e+02 2.276e+02 2.773e+02 3.463e+02 7.694e+02, threshold=5.547e+02, percent-clipped=7.0 2023-03-08 08:43:21,287 INFO [train2.py:809] (1/4) Epoch 13, batch 2550, loss[ctc_loss=0.09043, att_loss=0.2348, loss=0.2059, over 16121.00 frames. utt_duration=1537 frames, utt_pad_proportion=0.00576, over 42.00 utterances.], tot_loss[ctc_loss=0.09572, att_loss=0.2468, loss=0.2166, over 3284218.29 frames. utt_duration=1231 frames, utt_pad_proportion=0.05433, over 10683.45 utterances.], batch size: 42, lr: 8.32e-03, grad_scale: 8.0 2023-03-08 08:43:26,863 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-03-08 08:43:34,094 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9216, 4.1562, 4.0126, 4.2399, 2.6093, 4.1010, 2.6915, 1.8860], device='cuda:1'), covar=tensor([0.0393, 0.0172, 0.0638, 0.0180, 0.1605, 0.0170, 0.1363, 0.1594], device='cuda:1'), in_proj_covar=tensor([0.0152, 0.0124, 0.0253, 0.0118, 0.0218, 0.0111, 0.0224, 0.0199], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 08:43:34,402 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.67 vs. limit=5.0 2023-03-08 08:44:01,390 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-03-08 08:44:40,230 INFO [train2.py:809] (1/4) Epoch 13, batch 2600, loss[ctc_loss=0.06762, att_loss=0.2244, loss=0.1931, over 16390.00 frames. utt_duration=1492 frames, utt_pad_proportion=0.008145, over 44.00 utterances.], tot_loss[ctc_loss=0.09504, att_loss=0.2462, loss=0.216, over 3282350.74 frames. utt_duration=1255 frames, utt_pad_proportion=0.04882, over 10474.15 utterances.], batch size: 44, lr: 8.31e-03, grad_scale: 8.0 2023-03-08 08:45:06,219 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50422.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:45:10,504 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.653e+02 2.288e+02 2.850e+02 3.496e+02 8.572e+02, threshold=5.699e+02, percent-clipped=4.0 2023-03-08 08:45:59,192 INFO [train2.py:809] (1/4) Epoch 13, batch 2650, loss[ctc_loss=0.08561, att_loss=0.2484, loss=0.2158, over 16902.00 frames. utt_duration=684.4 frames, utt_pad_proportion=0.1402, over 99.00 utterances.], tot_loss[ctc_loss=0.09476, att_loss=0.2464, loss=0.2161, over 3269145.92 frames. utt_duration=1227 frames, utt_pad_proportion=0.05726, over 10666.11 utterances.], batch size: 99, lr: 8.31e-03, grad_scale: 8.0 2023-03-08 08:46:22,030 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=50470.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:46:44,839 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50484.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:47:17,951 INFO [train2.py:809] (1/4) Epoch 13, batch 2700, loss[ctc_loss=0.07692, att_loss=0.2313, loss=0.2004, over 14581.00 frames. utt_duration=1824 frames, utt_pad_proportion=0.03449, over 32.00 utterances.], tot_loss[ctc_loss=0.09419, att_loss=0.2456, loss=0.2154, over 3273261.84 frames. utt_duration=1252 frames, utt_pad_proportion=0.05099, over 10473.68 utterances.], batch size: 32, lr: 8.30e-03, grad_scale: 8.0 2023-03-08 08:47:49,716 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.590e+02 2.365e+02 2.788e+02 3.367e+02 6.677e+02, threshold=5.576e+02, percent-clipped=4.0 2023-03-08 08:48:06,650 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-03-08 08:48:23,021 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50546.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:48:33,796 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50553.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 08:48:38,309 INFO [train2.py:809] (1/4) Epoch 13, batch 2750, loss[ctc_loss=0.1562, att_loss=0.286, loss=0.2601, over 14691.00 frames. utt_duration=406.8 frames, utt_pad_proportion=0.2937, over 145.00 utterances.], tot_loss[ctc_loss=0.09366, att_loss=0.2452, loss=0.2149, over 3270062.22 frames. utt_duration=1242 frames, utt_pad_proportion=0.05411, over 10545.21 utterances.], batch size: 145, lr: 8.30e-03, grad_scale: 8.0 2023-03-08 08:48:46,040 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1413, 5.4514, 4.8260, 5.2695, 5.0376, 4.6703, 4.9163, 4.6456], device='cuda:1'), covar=tensor([0.1378, 0.0916, 0.0918, 0.0727, 0.0949, 0.1512, 0.2166, 0.2294], device='cuda:1'), in_proj_covar=tensor([0.0453, 0.0528, 0.0395, 0.0396, 0.0379, 0.0431, 0.0546, 0.0476], 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') 2023-03-08 08:49:22,687 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-03-08 08:49:26,729 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7004, 4.6526, 4.7145, 4.5938, 5.2483, 4.7080, 4.7655, 2.3928], device='cuda:1'), covar=tensor([0.0186, 0.0283, 0.0214, 0.0251, 0.0780, 0.0201, 0.0207, 0.1931], device='cuda:1'), in_proj_covar=tensor([0.0130, 0.0138, 0.0145, 0.0156, 0.0343, 0.0126, 0.0128, 0.0213], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0002, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 08:49:57,335 INFO [train2.py:809] (1/4) Epoch 13, batch 2800, loss[ctc_loss=0.06578, att_loss=0.2114, loss=0.1822, over 15350.00 frames. utt_duration=1756 frames, utt_pad_proportion=0.01242, over 35.00 utterances.], tot_loss[ctc_loss=0.09378, att_loss=0.2454, loss=0.2151, over 3272768.41 frames. utt_duration=1228 frames, utt_pad_proportion=0.05672, over 10671.20 utterances.], batch size: 35, lr: 8.30e-03, grad_scale: 8.0 2023-03-08 08:49:59,312 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50607.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:50:29,967 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.488e+02 2.143e+02 2.586e+02 3.162e+02 8.446e+02, threshold=5.172e+02, percent-clipped=2.0 2023-03-08 08:51:04,292 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2122, 4.7983, 4.3381, 4.8298, 2.6890, 4.6493, 2.4179, 2.1302], device='cuda:1'), covar=tensor([0.0370, 0.0152, 0.0792, 0.0179, 0.1969, 0.0170, 0.1771, 0.1658], device='cuda:1'), in_proj_covar=tensor([0.0154, 0.0124, 0.0255, 0.0120, 0.0221, 0.0113, 0.0226, 0.0201], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 08:51:17,763 INFO [train2.py:809] (1/4) Epoch 13, batch 2850, loss[ctc_loss=0.0734, att_loss=0.2208, loss=0.1913, over 15626.00 frames. utt_duration=1691 frames, utt_pad_proportion=0.009231, over 37.00 utterances.], tot_loss[ctc_loss=0.09358, att_loss=0.2452, loss=0.2149, over 3266992.29 frames. utt_duration=1232 frames, utt_pad_proportion=0.05616, over 10618.18 utterances.], batch size: 37, lr: 8.29e-03, grad_scale: 8.0 2023-03-08 08:52:27,023 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0602, 5.3728, 4.7918, 5.4127, 4.7140, 5.0243, 5.4897, 5.2403], device='cuda:1'), covar=tensor([0.0553, 0.0248, 0.0954, 0.0262, 0.0452, 0.0264, 0.0210, 0.0184], device='cuda:1'), in_proj_covar=tensor([0.0347, 0.0276, 0.0329, 0.0280, 0.0281, 0.0213, 0.0260, 0.0245], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0005, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 08:52:37,437 INFO [train2.py:809] (1/4) Epoch 13, batch 2900, loss[ctc_loss=0.1075, att_loss=0.2355, loss=0.2099, over 15369.00 frames. utt_duration=1758 frames, utt_pad_proportion=0.01131, over 35.00 utterances.], tot_loss[ctc_loss=0.09325, att_loss=0.2452, loss=0.2148, over 3270187.01 frames. utt_duration=1233 frames, utt_pad_proportion=0.05491, over 10622.62 utterances.], batch size: 35, lr: 8.29e-03, grad_scale: 8.0 2023-03-08 08:53:08,849 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.567e+02 2.231e+02 2.674e+02 3.196e+02 5.262e+02, threshold=5.348e+02, percent-clipped=1.0 2023-03-08 08:53:56,556 INFO [train2.py:809] (1/4) Epoch 13, batch 2950, loss[ctc_loss=0.08361, att_loss=0.2407, loss=0.2092, over 16336.00 frames. utt_duration=1454 frames, utt_pad_proportion=0.005715, over 45.00 utterances.], tot_loss[ctc_loss=0.09371, att_loss=0.2454, loss=0.2151, over 3276839.52 frames. utt_duration=1244 frames, utt_pad_proportion=0.05154, over 10552.73 utterances.], batch size: 45, lr: 8.28e-03, grad_scale: 8.0 2023-03-08 08:54:39,285 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50782.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:54:42,248 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50784.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:54:49,932 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.6541, 3.5733, 3.5040, 2.7011, 3.6117, 3.6457, 3.5367, 1.9237], device='cuda:1'), covar=tensor([0.1020, 0.1335, 0.2363, 1.0696, 0.1674, 0.3706, 0.1162, 1.2909], device='cuda:1'), in_proj_covar=tensor([0.0111, 0.0135, 0.0147, 0.0217, 0.0110, 0.0200, 0.0122, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 08:55:05,342 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3681, 3.0255, 3.5663, 2.8952, 3.4190, 4.5582, 4.2896, 3.0889], device='cuda:1'), covar=tensor([0.0424, 0.1659, 0.1085, 0.1419, 0.1051, 0.0714, 0.0597, 0.1517], device='cuda:1'), in_proj_covar=tensor([0.0231, 0.0233, 0.0252, 0.0206, 0.0247, 0.0320, 0.0232, 0.0225], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 08:55:16,326 INFO [train2.py:809] (1/4) Epoch 13, batch 3000, loss[ctc_loss=0.07582, att_loss=0.2297, loss=0.199, over 16172.00 frames. utt_duration=1580 frames, utt_pad_proportion=0.00655, over 41.00 utterances.], tot_loss[ctc_loss=0.0945, att_loss=0.2458, loss=0.2155, over 3278510.82 frames. utt_duration=1247 frames, utt_pad_proportion=0.05139, over 10530.19 utterances.], batch size: 41, lr: 8.28e-03, grad_scale: 8.0 2023-03-08 08:55:16,326 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-08 08:55:30,014 INFO [train2.py:843] (1/4) Epoch 13, validation: ctc_loss=0.04571, att_loss=0.2368, loss=0.1986, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-08 08:55:30,015 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-08 08:55:33,392 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50808.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 08:56:00,976 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.736e+02 2.441e+02 3.101e+02 4.045e+02 6.345e+02, threshold=6.202e+02, percent-clipped=8.0 2023-03-08 08:56:12,079 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=50832.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:56:29,172 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50843.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:56:30,142 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-03-08 08:56:44,459 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50853.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 08:56:48,706 INFO [train2.py:809] (1/4) Epoch 13, batch 3050, loss[ctc_loss=0.09872, att_loss=0.2381, loss=0.2102, over 16538.00 frames. utt_duration=1471 frames, utt_pad_proportion=0.006542, over 45.00 utterances.], tot_loss[ctc_loss=0.09604, att_loss=0.2465, loss=0.2164, over 3273247.17 frames. utt_duration=1218 frames, utt_pad_proportion=0.05995, over 10760.45 utterances.], batch size: 45, lr: 8.28e-03, grad_scale: 8.0 2023-03-08 08:56:59,869 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8241, 5.3682, 5.3377, 5.3148, 5.3068, 5.3614, 5.0082, 4.8082], device='cuda:1'), covar=tensor([0.1437, 0.0525, 0.0302, 0.0573, 0.0470, 0.0408, 0.0405, 0.0480], device='cuda:1'), in_proj_covar=tensor([0.0468, 0.0307, 0.0275, 0.0304, 0.0364, 0.0384, 0.0309, 0.0345], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 08:57:02,280 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-03-08 08:57:10,216 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50869.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 08:57:32,626 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.70 vs. limit=2.0 2023-03-08 08:58:00,755 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=50901.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 08:58:02,420 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50902.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:58:08,260 INFO [train2.py:809] (1/4) Epoch 13, batch 3100, loss[ctc_loss=0.09671, att_loss=0.245, loss=0.2154, over 17069.00 frames. utt_duration=1290 frames, utt_pad_proportion=0.008571, over 53.00 utterances.], tot_loss[ctc_loss=0.09507, att_loss=0.2454, loss=0.2154, over 3270371.19 frames. utt_duration=1242 frames, utt_pad_proportion=0.05576, over 10541.36 utterances.], batch size: 53, lr: 8.27e-03, grad_scale: 8.0 2023-03-08 08:58:39,136 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.514e+02 2.262e+02 2.755e+02 3.486e+02 1.234e+03, threshold=5.511e+02, percent-clipped=2.0 2023-03-08 08:59:11,634 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50946.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 08:59:26,546 INFO [train2.py:809] (1/4) Epoch 13, batch 3150, loss[ctc_loss=0.07571, att_loss=0.2171, loss=0.1888, over 15632.00 frames. utt_duration=1692 frames, utt_pad_proportion=0.009025, over 37.00 utterances.], tot_loss[ctc_loss=0.09387, att_loss=0.2444, loss=0.2143, over 3265471.76 frames. utt_duration=1246 frames, utt_pad_proportion=0.05631, over 10499.40 utterances.], batch size: 37, lr: 8.27e-03, grad_scale: 8.0 2023-03-08 09:00:46,289 INFO [train2.py:809] (1/4) Epoch 13, batch 3200, loss[ctc_loss=0.08122, att_loss=0.2502, loss=0.2164, over 17364.00 frames. utt_duration=1104 frames, utt_pad_proportion=0.03592, over 63.00 utterances.], tot_loss[ctc_loss=0.09408, att_loss=0.2446, loss=0.2145, over 3261890.48 frames. utt_duration=1242 frames, utt_pad_proportion=0.05935, over 10519.15 utterances.], batch size: 63, lr: 8.26e-03, grad_scale: 8.0 2023-03-08 09:00:48,209 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51007.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:01:17,932 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.429e+02 2.247e+02 2.628e+02 3.429e+02 8.157e+02, threshold=5.255e+02, percent-clipped=5.0 2023-03-08 09:01:42,558 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0068, 4.5531, 4.4062, 4.7191, 3.0360, 4.4814, 2.4681, 1.7242], device='cuda:1'), covar=tensor([0.0394, 0.0188, 0.0730, 0.0151, 0.1588, 0.0180, 0.1687, 0.1829], device='cuda:1'), in_proj_covar=tensor([0.0156, 0.0125, 0.0257, 0.0120, 0.0221, 0.0113, 0.0227, 0.0202], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 09:02:05,327 INFO [train2.py:809] (1/4) Epoch 13, batch 3250, loss[ctc_loss=0.08281, att_loss=0.2174, loss=0.1905, over 14907.00 frames. utt_duration=1808 frames, utt_pad_proportion=0.02457, over 33.00 utterances.], tot_loss[ctc_loss=0.09354, att_loss=0.2441, loss=0.214, over 3262159.25 frames. utt_duration=1267 frames, utt_pad_proportion=0.05433, over 10314.80 utterances.], batch size: 33, lr: 8.26e-03, grad_scale: 8.0 2023-03-08 09:02:14,820 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51062.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:02:39,848 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1346, 5.0734, 4.9748, 2.4912, 2.0272, 2.8213, 2.6552, 3.7224], device='cuda:1'), covar=tensor([0.0627, 0.0235, 0.0198, 0.3988, 0.5906, 0.2594, 0.3018, 0.1837], device='cuda:1'), in_proj_covar=tensor([0.0338, 0.0235, 0.0241, 0.0220, 0.0347, 0.0333, 0.0237, 0.0354], device='cuda:1'), out_proj_covar=tensor([1.4952e-04, 8.6750e-05, 1.0413e-04, 9.6515e-05, 1.4892e-04, 1.3369e-04, 9.3869e-05, 1.4779e-04], device='cuda:1') 2023-03-08 09:02:50,689 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.3844, 3.5588, 2.7064, 3.0463, 3.6577, 3.3087, 2.3367, 3.7427], device='cuda:1'), covar=tensor([0.1222, 0.0466, 0.1091, 0.0710, 0.0618, 0.0663, 0.1066, 0.0552], device='cuda:1'), in_proj_covar=tensor([0.0188, 0.0194, 0.0210, 0.0181, 0.0243, 0.0222, 0.0187, 0.0261], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-08 09:02:55,126 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.6928, 3.6301, 2.9277, 3.2259, 3.7664, 3.3730, 2.4571, 3.9363], device='cuda:1'), covar=tensor([0.1110, 0.0433, 0.1065, 0.0766, 0.0665, 0.0756, 0.1063, 0.0562], device='cuda:1'), in_proj_covar=tensor([0.0188, 0.0194, 0.0210, 0.0181, 0.0243, 0.0222, 0.0187, 0.0261], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-08 09:02:58,077 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.6025, 1.9025, 1.9065, 2.4274, 2.8151, 2.5419, 1.8250, 2.5893], device='cuda:1'), covar=tensor([0.1590, 0.4965, 0.3641, 0.1515, 0.1437, 0.1287, 0.4001, 0.1172], device='cuda:1'), in_proj_covar=tensor([0.0081, 0.0090, 0.0091, 0.0077, 0.0081, 0.0070, 0.0093, 0.0063], device='cuda:1'), out_proj_covar=tensor([5.6943e-05, 6.4591e-05, 6.6343e-05, 5.5695e-05, 5.6550e-05, 5.3751e-05, 6.5206e-05, 4.8720e-05], device='cuda:1') 2023-03-08 09:03:23,231 INFO [train2.py:809] (1/4) Epoch 13, batch 3300, loss[ctc_loss=0.06924, att_loss=0.2119, loss=0.1834, over 15513.00 frames. utt_duration=1725 frames, utt_pad_proportion=0.007859, over 36.00 utterances.], tot_loss[ctc_loss=0.09469, att_loss=0.2454, loss=0.2152, over 3274424.09 frames. utt_duration=1260 frames, utt_pad_proportion=0.05207, over 10405.30 utterances.], batch size: 36, lr: 8.26e-03, grad_scale: 8.0 2023-03-08 09:03:46,524 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-03-08 09:03:51,904 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51123.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:03:54,635 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.818e+02 2.303e+02 2.804e+02 3.427e+02 6.539e+02, threshold=5.609e+02, percent-clipped=6.0 2023-03-08 09:04:14,234 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51138.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:04:41,663 INFO [train2.py:809] (1/4) Epoch 13, batch 3350, loss[ctc_loss=0.09298, att_loss=0.227, loss=0.2002, over 15789.00 frames. utt_duration=1663 frames, utt_pad_proportion=0.008176, over 38.00 utterances.], tot_loss[ctc_loss=0.09497, att_loss=0.2454, loss=0.2153, over 3272764.64 frames. utt_duration=1242 frames, utt_pad_proportion=0.05731, over 10554.95 utterances.], batch size: 38, lr: 8.25e-03, grad_scale: 8.0 2023-03-08 09:04:55,051 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51164.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 09:05:08,581 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51172.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 09:05:45,950 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1801, 4.4646, 4.4814, 4.8707, 2.7487, 4.5265, 2.6571, 1.9241], device='cuda:1'), covar=tensor([0.0315, 0.0197, 0.0712, 0.0119, 0.1771, 0.0156, 0.1593, 0.1765], device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0127, 0.0258, 0.0120, 0.0224, 0.0115, 0.0228, 0.0204], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 09:05:55,378 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51202.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:06:01,137 INFO [train2.py:809] (1/4) Epoch 13, batch 3400, loss[ctc_loss=0.1114, att_loss=0.2509, loss=0.223, over 16120.00 frames. utt_duration=1537 frames, utt_pad_proportion=0.006541, over 42.00 utterances.], tot_loss[ctc_loss=0.09627, att_loss=0.2465, loss=0.2164, over 3270205.26 frames. utt_duration=1224 frames, utt_pad_proportion=0.06316, over 10702.51 utterances.], batch size: 42, lr: 8.25e-03, grad_scale: 8.0 2023-03-08 09:06:06,703 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.28 vs. limit=5.0 2023-03-08 09:06:33,019 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.585e+02 2.487e+02 2.909e+02 3.549e+02 7.479e+02, threshold=5.817e+02, percent-clipped=4.0 2023-03-08 09:06:41,109 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7725, 5.2219, 4.9910, 5.1582, 5.2785, 4.9812, 4.0558, 5.1556], device='cuda:1'), covar=tensor([0.0098, 0.0085, 0.0119, 0.0065, 0.0060, 0.0079, 0.0485, 0.0141], device='cuda:1'), in_proj_covar=tensor([0.0077, 0.0073, 0.0091, 0.0056, 0.0061, 0.0072, 0.0091, 0.0093], 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') 2023-03-08 09:06:42,683 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.3683, 2.6516, 3.3759, 4.3683, 3.8113, 3.9284, 2.7589, 2.0042], device='cuda:1'), covar=tensor([0.0674, 0.2267, 0.0954, 0.0525, 0.0823, 0.0393, 0.1660, 0.2365], device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0213, 0.0191, 0.0201, 0.0202, 0.0160, 0.0198, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 09:06:45,893 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51233.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 09:07:11,551 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=51250.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:07:20,300 INFO [train2.py:809] (1/4) Epoch 13, batch 3450, loss[ctc_loss=0.09996, att_loss=0.2422, loss=0.2138, over 16897.00 frames. utt_duration=1381 frames, utt_pad_proportion=0.00598, over 49.00 utterances.], tot_loss[ctc_loss=0.09569, att_loss=0.246, loss=0.2159, over 3270718.17 frames. utt_duration=1217 frames, utt_pad_proportion=0.06371, over 10759.11 utterances.], batch size: 49, lr: 8.24e-03, grad_scale: 8.0 2023-03-08 09:07:23,271 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-03-08 09:07:42,999 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9338, 5.2207, 5.5179, 5.3660, 5.3018, 5.8880, 5.1886, 5.9907], device='cuda:1'), covar=tensor([0.0826, 0.0776, 0.0785, 0.1344, 0.1980, 0.0978, 0.0655, 0.0711], device='cuda:1'), in_proj_covar=tensor([0.0765, 0.0454, 0.0535, 0.0591, 0.0787, 0.0538, 0.0435, 0.0522], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 09:08:33,554 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51302.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:08:39,553 INFO [train2.py:809] (1/4) Epoch 13, batch 3500, loss[ctc_loss=0.09983, att_loss=0.2563, loss=0.225, over 17109.00 frames. utt_duration=1223 frames, utt_pad_proportion=0.01572, over 56.00 utterances.], tot_loss[ctc_loss=0.09434, att_loss=0.2455, loss=0.2153, over 3272725.08 frames. utt_duration=1245 frames, utt_pad_proportion=0.05583, over 10529.76 utterances.], batch size: 56, lr: 8.24e-03, grad_scale: 8.0 2023-03-08 09:09:10,404 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.552e+02 2.305e+02 2.663e+02 3.285e+02 9.490e+02, threshold=5.326e+02, percent-clipped=4.0 2023-03-08 09:09:48,817 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6004, 2.3236, 4.9992, 4.0079, 2.9444, 4.4551, 4.9447, 4.6941], device='cuda:1'), covar=tensor([0.0234, 0.1840, 0.0186, 0.1002, 0.1944, 0.0226, 0.0102, 0.0232], device='cuda:1'), in_proj_covar=tensor([0.0156, 0.0242, 0.0146, 0.0307, 0.0271, 0.0188, 0.0131, 0.0163], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 09:09:52,558 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-03-08 09:09:54,933 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.1939, 2.5404, 3.2750, 4.3008, 3.8234, 3.9410, 2.7084, 1.8940], device='cuda:1'), covar=tensor([0.0839, 0.2416, 0.1006, 0.0566, 0.0769, 0.0401, 0.1693, 0.2595], device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0214, 0.0191, 0.0202, 0.0202, 0.0161, 0.0198, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 09:09:57,629 INFO [train2.py:809] (1/4) Epoch 13, batch 3550, loss[ctc_loss=0.1183, att_loss=0.2664, loss=0.2367, over 16887.00 frames. utt_duration=683.8 frames, utt_pad_proportion=0.1367, over 99.00 utterances.], tot_loss[ctc_loss=0.09394, att_loss=0.2453, loss=0.215, over 3276048.26 frames. utt_duration=1251 frames, utt_pad_proportion=0.0532, over 10485.26 utterances.], batch size: 99, lr: 8.24e-03, grad_scale: 8.0 2023-03-08 09:10:22,461 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-03-08 09:10:39,749 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51382.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:10:54,964 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51392.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:11:16,643 INFO [train2.py:809] (1/4) Epoch 13, batch 3600, loss[ctc_loss=0.1187, att_loss=0.271, loss=0.2406, over 17065.00 frames. utt_duration=1290 frames, utt_pad_proportion=0.008832, over 53.00 utterances.], tot_loss[ctc_loss=0.09451, att_loss=0.2457, loss=0.2155, over 3277064.83 frames. utt_duration=1250 frames, utt_pad_proportion=0.05328, over 10499.75 utterances.], batch size: 53, lr: 8.23e-03, grad_scale: 8.0 2023-03-08 09:11:36,220 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51418.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:11:36,717 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.85 vs. limit=2.0 2023-03-08 09:11:46,270 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.606e+02 2.412e+02 2.884e+02 3.551e+02 7.030e+02, threshold=5.767e+02, percent-clipped=4.0 2023-03-08 09:12:05,896 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51438.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:12:13,304 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51443.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:12:29,021 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51453.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:12:33,509 INFO [train2.py:809] (1/4) Epoch 13, batch 3650, loss[ctc_loss=0.07905, att_loss=0.233, loss=0.2022, over 15639.00 frames. utt_duration=1692 frames, utt_pad_proportion=0.009162, over 37.00 utterances.], tot_loss[ctc_loss=0.09552, att_loss=0.2462, loss=0.2161, over 3285371.06 frames. utt_duration=1262 frames, utt_pad_proportion=0.04744, over 10427.11 utterances.], batch size: 37, lr: 8.23e-03, grad_scale: 8.0 2023-03-08 09:12:47,031 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51464.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 09:13:20,394 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=51486.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:13:34,864 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51495.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:13:52,936 INFO [train2.py:809] (1/4) Epoch 13, batch 3700, loss[ctc_loss=0.08829, att_loss=0.2625, loss=0.2277, over 17040.00 frames. utt_duration=1287 frames, utt_pad_proportion=0.01054, over 53.00 utterances.], tot_loss[ctc_loss=0.09515, att_loss=0.2462, loss=0.216, over 3278312.05 frames. utt_duration=1269 frames, utt_pad_proportion=0.04682, over 10349.18 utterances.], batch size: 53, lr: 8.22e-03, grad_scale: 8.0 2023-03-08 09:14:03,445 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=51512.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 09:14:23,031 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.524e+02 2.328e+02 2.806e+02 3.830e+02 9.896e+02, threshold=5.611e+02, percent-clipped=3.0 2023-03-08 09:14:28,091 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51528.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 09:15:13,029 INFO [train2.py:809] (1/4) Epoch 13, batch 3750, loss[ctc_loss=0.0955, att_loss=0.2361, loss=0.208, over 15874.00 frames. utt_duration=1630 frames, utt_pad_proportion=0.008656, over 39.00 utterances.], tot_loss[ctc_loss=0.09426, att_loss=0.2453, loss=0.2151, over 3265061.35 frames. utt_duration=1259 frames, utt_pad_proportion=0.05301, over 10389.88 utterances.], batch size: 39, lr: 8.22e-03, grad_scale: 8.0 2023-03-08 09:15:13,454 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51556.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:16:26,955 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51602.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:16:32,781 INFO [train2.py:809] (1/4) Epoch 13, batch 3800, loss[ctc_loss=0.1185, att_loss=0.2643, loss=0.2351, over 16892.00 frames. utt_duration=1381 frames, utt_pad_proportion=0.005936, over 49.00 utterances.], tot_loss[ctc_loss=0.0947, att_loss=0.2456, loss=0.2155, over 3254553.09 frames. utt_duration=1225 frames, utt_pad_proportion=0.06289, over 10643.20 utterances.], batch size: 49, lr: 8.22e-03, grad_scale: 8.0 2023-03-08 09:17:02,532 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.490e+02 2.194e+02 2.674e+02 3.110e+02 7.500e+02, threshold=5.349e+02, percent-clipped=1.0 2023-03-08 09:17:41,362 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=51650.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:17:51,575 INFO [train2.py:809] (1/4) Epoch 13, batch 3850, loss[ctc_loss=0.1229, att_loss=0.2635, loss=0.2354, over 17373.00 frames. utt_duration=1105 frames, utt_pad_proportion=0.03434, over 63.00 utterances.], tot_loss[ctc_loss=0.09477, att_loss=0.2458, loss=0.2156, over 3256825.71 frames. utt_duration=1208 frames, utt_pad_proportion=0.06731, over 10793.66 utterances.], batch size: 63, lr: 8.21e-03, grad_scale: 8.0 2023-03-08 09:18:36,313 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0724, 5.2801, 5.3158, 5.2700, 5.3938, 5.3315, 5.0649, 4.8652], device='cuda:1'), covar=tensor([0.0929, 0.0556, 0.0231, 0.0489, 0.0276, 0.0266, 0.0311, 0.0317], device='cuda:1'), in_proj_covar=tensor([0.0472, 0.0312, 0.0279, 0.0308, 0.0367, 0.0384, 0.0311, 0.0346], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 09:19:07,770 INFO [train2.py:809] (1/4) Epoch 13, batch 3900, loss[ctc_loss=0.08417, att_loss=0.2403, loss=0.2091, over 16413.00 frames. utt_duration=1494 frames, utt_pad_proportion=0.006185, over 44.00 utterances.], tot_loss[ctc_loss=0.09397, att_loss=0.2451, loss=0.2149, over 3256649.95 frames. utt_duration=1205 frames, utt_pad_proportion=0.06832, over 10820.32 utterances.], batch size: 44, lr: 8.21e-03, grad_scale: 8.0 2023-03-08 09:19:09,517 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0874, 5.3560, 5.6029, 5.3944, 5.4336, 6.0990, 5.2455, 6.1247], device='cuda:1'), covar=tensor([0.0707, 0.0678, 0.0718, 0.1255, 0.2070, 0.0795, 0.0567, 0.0789], device='cuda:1'), in_proj_covar=tensor([0.0765, 0.0450, 0.0536, 0.0589, 0.0781, 0.0538, 0.0435, 0.0522], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 09:19:26,176 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51718.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:19:36,508 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.485e+02 2.131e+02 2.698e+02 3.249e+02 8.076e+02, threshold=5.395e+02, percent-clipped=6.0 2023-03-08 09:19:56,079 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51738.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:19:57,919 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.7967, 2.4973, 3.2968, 2.5368, 3.2680, 3.9888, 3.8600, 2.6896], device='cuda:1'), covar=tensor([0.0535, 0.2060, 0.1294, 0.1477, 0.1072, 0.1007, 0.0644, 0.1638], device='cuda:1'), in_proj_covar=tensor([0.0233, 0.0234, 0.0254, 0.0209, 0.0251, 0.0326, 0.0232, 0.0226], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 09:20:10,170 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51747.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:20:11,540 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51748.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:20:23,592 INFO [train2.py:809] (1/4) Epoch 13, batch 3950, loss[ctc_loss=0.0832, att_loss=0.2427, loss=0.2108, over 16462.00 frames. utt_duration=1433 frames, utt_pad_proportion=0.007347, over 46.00 utterances.], tot_loss[ctc_loss=0.09432, att_loss=0.2459, loss=0.2156, over 3262479.32 frames. utt_duration=1223 frames, utt_pad_proportion=0.06117, over 10685.85 utterances.], batch size: 46, lr: 8.20e-03, grad_scale: 8.0 2023-03-08 09:20:27,099 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0438, 4.7660, 4.9038, 2.2883, 2.0412, 2.2659, 2.2485, 3.4560], device='cuda:1'), covar=tensor([0.0781, 0.0239, 0.0225, 0.3828, 0.6163, 0.3615, 0.3244, 0.2142], device='cuda:1'), in_proj_covar=tensor([0.0337, 0.0233, 0.0241, 0.0218, 0.0342, 0.0332, 0.0232, 0.0350], device='cuda:1'), out_proj_covar=tensor([1.4870e-04, 8.6226e-05, 1.0392e-04, 9.5344e-05, 1.4666e-04, 1.3282e-04, 9.2236e-05, 1.4586e-04], device='cuda:1') 2023-03-08 09:20:38,817 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=51766.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:21:41,016 INFO [train2.py:809] (1/4) Epoch 14, batch 0, loss[ctc_loss=0.1011, att_loss=0.2552, loss=0.2244, over 17044.00 frames. utt_duration=1313 frames, utt_pad_proportion=0.008439, over 52.00 utterances.], tot_loss[ctc_loss=0.1011, att_loss=0.2552, loss=0.2244, over 17044.00 frames. utt_duration=1313 frames, utt_pad_proportion=0.008439, over 52.00 utterances.], batch size: 52, lr: 7.90e-03, grad_scale: 8.0 2023-03-08 09:21:41,016 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-08 09:21:52,751 INFO [train2.py:843] (1/4) Epoch 14, validation: ctc_loss=0.04501, att_loss=0.2367, loss=0.1984, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-08 09:21:52,753 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-08 09:22:08,896 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8277, 5.1682, 4.6572, 5.2939, 4.5536, 4.9735, 5.3511, 5.1072], device='cuda:1'), covar=tensor([0.0672, 0.0353, 0.0890, 0.0307, 0.0512, 0.0261, 0.0247, 0.0205], device='cuda:1'), in_proj_covar=tensor([0.0344, 0.0271, 0.0323, 0.0278, 0.0279, 0.0211, 0.0258, 0.0240], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005, 0.0004], device='cuda:1') 2023-03-08 09:22:20,790 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51808.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:22:46,826 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.359e+02 2.322e+02 2.890e+02 3.740e+02 8.433e+02, threshold=5.780e+02, percent-clipped=5.0 2023-03-08 09:22:52,362 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51828.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 09:23:11,219 INFO [train2.py:809] (1/4) Epoch 14, batch 50, loss[ctc_loss=0.1002, att_loss=0.2629, loss=0.2304, over 17020.00 frames. utt_duration=1286 frames, utt_pad_proportion=0.0115, over 53.00 utterances.], tot_loss[ctc_loss=0.08903, att_loss=0.2426, loss=0.2119, over 735109.22 frames. utt_duration=1348 frames, utt_pad_proportion=0.03672, over 2183.64 utterances.], batch size: 53, lr: 7.90e-03, grad_scale: 8.0 2023-03-08 09:23:29,032 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51851.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:23:45,563 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-03-08 09:24:08,122 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=51876.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 09:24:29,806 INFO [train2.py:809] (1/4) Epoch 14, batch 100, loss[ctc_loss=0.0896, att_loss=0.2545, loss=0.2215, over 17028.00 frames. utt_duration=1337 frames, utt_pad_proportion=0.007264, over 51.00 utterances.], tot_loss[ctc_loss=0.08913, att_loss=0.2432, loss=0.2124, over 1294907.62 frames. utt_duration=1282 frames, utt_pad_proportion=0.04772, over 4044.81 utterances.], batch size: 51, lr: 7.90e-03, grad_scale: 8.0 2023-03-08 09:24:43,076 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2884, 5.1969, 5.1602, 2.7451, 2.2189, 2.9815, 3.1418, 3.9010], device='cuda:1'), covar=tensor([0.0685, 0.0421, 0.0241, 0.4219, 0.5646, 0.2539, 0.2354, 0.1959], device='cuda:1'), in_proj_covar=tensor([0.0347, 0.0240, 0.0247, 0.0226, 0.0353, 0.0342, 0.0239, 0.0359], device='cuda:1'), out_proj_covar=tensor([1.5320e-04, 8.8866e-05, 1.0694e-04, 9.9075e-05, 1.5127e-04, 1.3660e-04, 9.4740e-05, 1.4969e-04], device='cuda:1') 2023-03-08 09:24:58,811 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6551, 2.8896, 3.5790, 3.1769, 3.6953, 4.6733, 4.4784, 3.3646], device='cuda:1'), covar=tensor([0.0414, 0.1875, 0.1228, 0.1296, 0.0950, 0.0791, 0.0511, 0.1261], device='cuda:1'), in_proj_covar=tensor([0.0230, 0.0234, 0.0252, 0.0206, 0.0249, 0.0324, 0.0231, 0.0223], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 09:25:11,380 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-03-08 09:25:24,995 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.531e+02 2.393e+02 2.937e+02 3.541e+02 1.195e+03, threshold=5.873e+02, percent-clipped=2.0 2023-03-08 09:25:48,793 INFO [train2.py:809] (1/4) Epoch 14, batch 150, loss[ctc_loss=0.1271, att_loss=0.2723, loss=0.2433, over 17043.00 frames. utt_duration=1313 frames, utt_pad_proportion=0.008468, over 52.00 utterances.], tot_loss[ctc_loss=0.09176, att_loss=0.2445, loss=0.214, over 1730056.53 frames. utt_duration=1236 frames, utt_pad_proportion=0.05973, over 5606.94 utterances.], batch size: 52, lr: 7.89e-03, grad_scale: 8.0 2023-03-08 09:25:56,842 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4413, 5.0317, 4.7546, 4.9246, 4.9490, 4.6194, 3.5605, 4.8307], device='cuda:1'), covar=tensor([0.0128, 0.0113, 0.0124, 0.0091, 0.0106, 0.0137, 0.0624, 0.0233], device='cuda:1'), in_proj_covar=tensor([0.0078, 0.0076, 0.0093, 0.0058, 0.0063, 0.0074, 0.0094, 0.0095], 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') 2023-03-08 09:27:07,211 INFO [train2.py:809] (1/4) Epoch 14, batch 200, loss[ctc_loss=0.2039, att_loss=0.3037, loss=0.2837, over 14113.00 frames. utt_duration=383 frames, utt_pad_proportion=0.3245, over 148.00 utterances.], tot_loss[ctc_loss=0.09335, att_loss=0.2457, loss=0.2152, over 2066964.46 frames. utt_duration=1187 frames, utt_pad_proportion=0.07232, over 6974.36 utterances.], batch size: 148, lr: 7.89e-03, grad_scale: 8.0 2023-03-08 09:27:39,695 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9983, 5.3263, 4.7745, 5.3785, 4.7251, 4.9887, 5.4324, 5.2047], device='cuda:1'), covar=tensor([0.0545, 0.0233, 0.0844, 0.0238, 0.0438, 0.0249, 0.0197, 0.0181], device='cuda:1'), in_proj_covar=tensor([0.0344, 0.0269, 0.0323, 0.0278, 0.0278, 0.0210, 0.0259, 0.0240], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0005, 0.0004], device='cuda:1') 2023-03-08 09:27:50,735 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.4236, 1.6941, 1.9011, 2.2634, 2.0722, 2.1524, 1.6398, 2.6619], device='cuda:1'), covar=tensor([0.1009, 0.3025, 0.2925, 0.0743, 0.1666, 0.1124, 0.1801, 0.0678], device='cuda:1'), in_proj_covar=tensor([0.0085, 0.0093, 0.0095, 0.0080, 0.0085, 0.0074, 0.0098, 0.0066], device='cuda:1'), out_proj_covar=tensor([5.9345e-05, 6.6993e-05, 6.9470e-05, 5.8412e-05, 5.9244e-05, 5.6661e-05, 6.8265e-05, 5.0988e-05], device='cuda:1') 2023-03-08 09:28:06,027 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.576e+02 2.279e+02 2.700e+02 3.291e+02 7.221e+02, threshold=5.400e+02, percent-clipped=2.0 2023-03-08 09:28:27,689 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52038.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:28:30,296 INFO [train2.py:809] (1/4) Epoch 14, batch 250, loss[ctc_loss=0.1253, att_loss=0.2746, loss=0.2447, over 17300.00 frames. utt_duration=1260 frames, utt_pad_proportion=0.01209, over 55.00 utterances.], tot_loss[ctc_loss=0.09167, att_loss=0.2444, loss=0.2138, over 2334382.79 frames. utt_duration=1201 frames, utt_pad_proportion=0.06793, over 7781.73 utterances.], batch size: 55, lr: 7.88e-03, grad_scale: 8.0 2023-03-08 09:28:30,665 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52040.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:28:42,583 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52048.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:29:30,439 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.81 vs. limit=2.0 2023-03-08 09:29:42,239 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=52086.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:29:48,099 INFO [train2.py:809] (1/4) Epoch 14, batch 300, loss[ctc_loss=0.124, att_loss=0.2589, loss=0.2319, over 17355.00 frames. utt_duration=1007 frames, utt_pad_proportion=0.0496, over 69.00 utterances.], tot_loss[ctc_loss=0.09383, att_loss=0.2457, loss=0.2153, over 2539377.97 frames. utt_duration=1175 frames, utt_pad_proportion=0.07469, over 8657.81 utterances.], batch size: 69, lr: 7.88e-03, grad_scale: 16.0 2023-03-08 09:29:57,155 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=52096.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:30:05,844 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52101.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:30:08,640 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52103.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:30:20,608 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.6190, 3.6217, 3.3570, 3.0220, 3.4685, 3.4816, 3.5627, 2.3822], device='cuda:1'), covar=tensor([0.0938, 0.1543, 0.2948, 0.5932, 0.1420, 0.6032, 0.0946, 0.6667], device='cuda:1'), in_proj_covar=tensor([0.0114, 0.0137, 0.0152, 0.0220, 0.0113, 0.0203, 0.0124, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 09:30:42,751 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.393e+02 2.191e+02 2.821e+02 3.545e+02 1.097e+03, threshold=5.643e+02, percent-clipped=10.0 2023-03-08 09:31:06,327 INFO [train2.py:809] (1/4) Epoch 14, batch 350, loss[ctc_loss=0.1055, att_loss=0.2686, loss=0.236, over 17276.00 frames. utt_duration=1258 frames, utt_pad_proportion=0.01286, over 55.00 utterances.], tot_loss[ctc_loss=0.09307, att_loss=0.2458, loss=0.2153, over 2709473.29 frames. utt_duration=1196 frames, utt_pad_proportion=0.06451, over 9074.97 utterances.], batch size: 55, lr: 7.88e-03, grad_scale: 16.0 2023-03-08 09:31:23,513 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52151.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:31:39,704 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.0102, 3.8051, 3.0720, 3.4686, 3.9589, 3.6475, 2.9245, 4.3089], device='cuda:1'), covar=tensor([0.0954, 0.0400, 0.1069, 0.0611, 0.0594, 0.0614, 0.0808, 0.0384], device='cuda:1'), in_proj_covar=tensor([0.0190, 0.0196, 0.0211, 0.0184, 0.0247, 0.0223, 0.0187, 0.0263], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-08 09:32:16,369 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.65 vs. limit=5.0 2023-03-08 09:32:25,898 INFO [train2.py:809] (1/4) Epoch 14, batch 400, loss[ctc_loss=0.1012, att_loss=0.2439, loss=0.2153, over 16292.00 frames. utt_duration=1517 frames, utt_pad_proportion=0.006594, over 43.00 utterances.], tot_loss[ctc_loss=0.09398, att_loss=0.2462, loss=0.2158, over 2831848.02 frames. utt_duration=1182 frames, utt_pad_proportion=0.0702, over 9597.02 utterances.], batch size: 43, lr: 7.87e-03, grad_scale: 16.0 2023-03-08 09:32:31,129 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52193.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:32:40,136 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=52199.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:33:21,767 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.597e+02 2.286e+02 2.626e+02 3.493e+02 1.383e+03, threshold=5.251e+02, percent-clipped=5.0 2023-03-08 09:33:46,076 INFO [train2.py:809] (1/4) Epoch 14, batch 450, loss[ctc_loss=0.08328, att_loss=0.2306, loss=0.2012, over 16278.00 frames. utt_duration=1516 frames, utt_pad_proportion=0.007234, over 43.00 utterances.], tot_loss[ctc_loss=0.09374, att_loss=0.2456, loss=0.2152, over 2925668.35 frames. utt_duration=1188 frames, utt_pad_proportion=0.07067, over 9860.19 utterances.], batch size: 43, lr: 7.87e-03, grad_scale: 16.0 2023-03-08 09:34:09,183 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52254.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:35:03,002 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-03-08 09:35:04,913 INFO [train2.py:809] (1/4) Epoch 14, batch 500, loss[ctc_loss=0.1061, att_loss=0.2548, loss=0.2251, over 17422.00 frames. utt_duration=883.6 frames, utt_pad_proportion=0.07572, over 79.00 utterances.], tot_loss[ctc_loss=0.09392, att_loss=0.2465, loss=0.216, over 3012519.05 frames. utt_duration=1194 frames, utt_pad_proportion=0.06663, over 10104.85 utterances.], batch size: 79, lr: 7.87e-03, grad_scale: 16.0 2023-03-08 09:35:39,567 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52312.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:36:01,140 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.299e+02 2.275e+02 2.805e+02 3.649e+02 6.948e+02, threshold=5.610e+02, percent-clipped=5.0 2023-03-08 09:36:05,233 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0926, 2.6390, 3.4394, 2.6516, 3.3532, 4.2268, 4.0577, 2.9480], device='cuda:1'), covar=tensor([0.0415, 0.1967, 0.1237, 0.1435, 0.1034, 0.0925, 0.0565, 0.1503], device='cuda:1'), in_proj_covar=tensor([0.0230, 0.0232, 0.0253, 0.0207, 0.0250, 0.0324, 0.0231, 0.0224], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 09:36:23,208 INFO [train2.py:809] (1/4) Epoch 14, batch 550, loss[ctc_loss=0.08724, att_loss=0.2612, loss=0.2264, over 17049.00 frames. utt_duration=1288 frames, utt_pad_proportion=0.009702, over 53.00 utterances.], tot_loss[ctc_loss=0.09401, att_loss=0.2465, loss=0.216, over 3077067.83 frames. utt_duration=1213 frames, utt_pad_proportion=0.05936, over 10156.35 utterances.], batch size: 53, lr: 7.86e-03, grad_scale: 8.0 2023-03-08 09:37:15,139 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52373.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:37:42,123 INFO [train2.py:809] (1/4) Epoch 14, batch 600, loss[ctc_loss=0.1102, att_loss=0.2524, loss=0.224, over 16320.00 frames. utt_duration=1452 frames, utt_pad_proportion=0.00681, over 45.00 utterances.], tot_loss[ctc_loss=0.09378, att_loss=0.2464, loss=0.2159, over 3120095.17 frames. utt_duration=1205 frames, utt_pad_proportion=0.06395, over 10371.46 utterances.], batch size: 45, lr: 7.86e-03, grad_scale: 8.0 2023-03-08 09:37:51,403 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52396.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:38:03,148 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52403.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:38:39,774 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.434e+02 2.147e+02 2.714e+02 3.419e+02 5.544e+02, threshold=5.429e+02, percent-clipped=0.0 2023-03-08 09:39:01,291 INFO [train2.py:809] (1/4) Epoch 14, batch 650, loss[ctc_loss=0.1093, att_loss=0.2597, loss=0.2296, over 17358.00 frames. utt_duration=1008 frames, utt_pad_proportion=0.05027, over 69.00 utterances.], tot_loss[ctc_loss=0.09259, att_loss=0.2456, loss=0.215, over 3155120.53 frames. utt_duration=1229 frames, utt_pad_proportion=0.05851, over 10278.40 utterances.], batch size: 69, lr: 7.85e-03, grad_scale: 8.0 2023-03-08 09:39:18,704 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=52451.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:39:51,710 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2216, 5.2653, 5.0955, 2.4935, 2.1675, 3.0008, 2.8046, 3.8677], device='cuda:1'), covar=tensor([0.0709, 0.0363, 0.0256, 0.5121, 0.5581, 0.2461, 0.2725, 0.1842], device='cuda:1'), in_proj_covar=tensor([0.0342, 0.0237, 0.0245, 0.0223, 0.0348, 0.0335, 0.0231, 0.0353], device='cuda:1'), out_proj_covar=tensor([1.5068e-04, 8.8237e-05, 1.0640e-04, 9.7536e-05, 1.4882e-04, 1.3395e-04, 9.1641e-05, 1.4740e-04], device='cuda:1') 2023-03-08 09:40:19,854 INFO [train2.py:809] (1/4) Epoch 14, batch 700, loss[ctc_loss=0.07647, att_loss=0.2273, loss=0.1972, over 15985.00 frames. utt_duration=1600 frames, utt_pad_proportion=0.009257, over 40.00 utterances.], tot_loss[ctc_loss=0.09259, att_loss=0.2457, loss=0.2151, over 3187936.16 frames. utt_duration=1241 frames, utt_pad_proportion=0.05305, over 10291.43 utterances.], batch size: 40, lr: 7.85e-03, grad_scale: 8.0 2023-03-08 09:41:15,851 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.392e+02 1.961e+02 2.440e+02 3.056e+02 5.134e+02, threshold=4.881e+02, percent-clipped=0.0 2023-03-08 09:41:22,236 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-03-08 09:41:27,586 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4373, 4.6864, 4.6892, 5.0929, 2.8758, 4.9288, 2.5250, 2.2176], device='cuda:1'), covar=tensor([0.0267, 0.0180, 0.0638, 0.0089, 0.1595, 0.0105, 0.1671, 0.1522], device='cuda:1'), in_proj_covar=tensor([0.0156, 0.0130, 0.0258, 0.0120, 0.0224, 0.0115, 0.0227, 0.0202], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 09:41:37,732 INFO [train2.py:809] (1/4) Epoch 14, batch 750, loss[ctc_loss=0.07626, att_loss=0.2255, loss=0.1957, over 15946.00 frames. utt_duration=1557 frames, utt_pad_proportion=0.006938, over 41.00 utterances.], tot_loss[ctc_loss=0.09204, att_loss=0.2451, loss=0.2145, over 3204981.13 frames. utt_duration=1253 frames, utt_pad_proportion=0.0521, over 10244.29 utterances.], batch size: 41, lr: 7.85e-03, grad_scale: 8.0 2023-03-08 09:41:52,171 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52549.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:41:58,960 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8030, 5.0662, 4.5698, 5.1344, 4.4964, 4.8203, 5.1909, 4.9586], device='cuda:1'), covar=tensor([0.0572, 0.0298, 0.0819, 0.0289, 0.0412, 0.0257, 0.0248, 0.0195], device='cuda:1'), in_proj_covar=tensor([0.0350, 0.0274, 0.0327, 0.0284, 0.0282, 0.0213, 0.0260, 0.0244], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0005, 0.0005, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 09:42:02,147 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52555.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 09:42:56,737 INFO [train2.py:809] (1/4) Epoch 14, batch 800, loss[ctc_loss=0.09349, att_loss=0.2539, loss=0.2218, over 16525.00 frames. utt_duration=1471 frames, utt_pad_proportion=0.007052, over 45.00 utterances.], tot_loss[ctc_loss=0.09181, att_loss=0.2449, loss=0.2143, over 3218702.35 frames. utt_duration=1264 frames, utt_pad_proportion=0.04976, over 10198.83 utterances.], batch size: 45, lr: 7.84e-03, grad_scale: 8.0 2023-03-08 09:43:38,064 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52616.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 09:43:39,423 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5268, 3.0225, 3.7079, 3.0030, 3.6403, 4.6081, 4.4373, 3.4330], device='cuda:1'), covar=tensor([0.0398, 0.1739, 0.1093, 0.1415, 0.0888, 0.0920, 0.0581, 0.1156], device='cuda:1'), in_proj_covar=tensor([0.0232, 0.0232, 0.0253, 0.0208, 0.0248, 0.0325, 0.0232, 0.0224], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 09:43:43,824 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-03-08 09:43:54,265 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.482e+02 2.071e+02 2.560e+02 3.132e+02 4.899e+02, threshold=5.119e+02, percent-clipped=1.0 2023-03-08 09:44:16,109 INFO [train2.py:809] (1/4) Epoch 14, batch 850, loss[ctc_loss=0.09776, att_loss=0.2459, loss=0.2163, over 15958.00 frames. utt_duration=1558 frames, utt_pad_proportion=0.007073, over 41.00 utterances.], tot_loss[ctc_loss=0.09242, att_loss=0.2451, loss=0.2145, over 3231904.57 frames. utt_duration=1255 frames, utt_pad_proportion=0.05207, over 10312.35 utterances.], batch size: 41, lr: 7.84e-03, grad_scale: 8.0 2023-03-08 09:44:20,931 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6382, 4.8942, 4.8434, 4.8340, 4.9539, 4.9188, 4.6427, 4.4921], device='cuda:1'), covar=tensor([0.0953, 0.0563, 0.0325, 0.0493, 0.0311, 0.0304, 0.0382, 0.0341], device='cuda:1'), in_proj_covar=tensor([0.0475, 0.0314, 0.0279, 0.0307, 0.0360, 0.0378, 0.0311, 0.0346], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 09:45:00,846 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52668.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:45:32,767 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.5754, 3.8277, 3.4050, 3.7743, 2.7399, 3.6175, 2.7662, 2.3924], device='cuda:1'), covar=tensor([0.0410, 0.0205, 0.0770, 0.0273, 0.1521, 0.0209, 0.1316, 0.1285], device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0130, 0.0259, 0.0122, 0.0226, 0.0116, 0.0229, 0.0202], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 09:45:35,372 INFO [train2.py:809] (1/4) Epoch 14, batch 900, loss[ctc_loss=0.07973, att_loss=0.2378, loss=0.2062, over 17592.00 frames. utt_duration=1006 frames, utt_pad_proportion=0.04705, over 70.00 utterances.], tot_loss[ctc_loss=0.09235, att_loss=0.245, loss=0.2145, over 3245054.64 frames. utt_duration=1231 frames, utt_pad_proportion=0.05729, over 10560.29 utterances.], batch size: 70, lr: 7.84e-03, grad_scale: 8.0 2023-03-08 09:45:44,741 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52696.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:46:32,364 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.422e+02 2.184e+02 2.605e+02 3.188e+02 5.544e+02, threshold=5.210e+02, percent-clipped=2.0 2023-03-08 09:46:53,847 INFO [train2.py:809] (1/4) Epoch 14, batch 950, loss[ctc_loss=0.07838, att_loss=0.2237, loss=0.1946, over 15878.00 frames. utt_duration=1630 frames, utt_pad_proportion=0.008547, over 39.00 utterances.], tot_loss[ctc_loss=0.09216, att_loss=0.2449, loss=0.2143, over 3245562.76 frames. utt_duration=1222 frames, utt_pad_proportion=0.06122, over 10635.08 utterances.], batch size: 39, lr: 7.83e-03, grad_scale: 8.0 2023-03-08 09:47:00,130 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=52744.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:47:12,389 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52752.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:48:04,933 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4215, 2.7947, 3.5418, 2.7826, 3.3197, 4.5454, 4.2776, 2.9861], device='cuda:1'), covar=tensor([0.0365, 0.1899, 0.1297, 0.1535, 0.1242, 0.0699, 0.0637, 0.1494], device='cuda:1'), in_proj_covar=tensor([0.0233, 0.0234, 0.0254, 0.0210, 0.0250, 0.0326, 0.0233, 0.0224], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 09:48:12,277 INFO [train2.py:809] (1/4) Epoch 14, batch 1000, loss[ctc_loss=0.0862, att_loss=0.2352, loss=0.2054, over 16316.00 frames. utt_duration=1452 frames, utt_pad_proportion=0.006916, over 45.00 utterances.], tot_loss[ctc_loss=0.0921, att_loss=0.2454, loss=0.2148, over 3246212.72 frames. utt_duration=1215 frames, utt_pad_proportion=0.06354, over 10701.20 utterances.], batch size: 45, lr: 7.83e-03, grad_scale: 8.0 2023-03-08 09:48:48,542 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52813.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:49:09,713 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.532e+02 2.243e+02 2.663e+02 3.401e+02 6.618e+02, threshold=5.327e+02, percent-clipped=3.0 2023-03-08 09:49:31,216 INFO [train2.py:809] (1/4) Epoch 14, batch 1050, loss[ctc_loss=0.1404, att_loss=0.2803, loss=0.2523, over 14070.00 frames. utt_duration=384.4 frames, utt_pad_proportion=0.3245, over 147.00 utterances.], tot_loss[ctc_loss=0.09207, att_loss=0.2457, loss=0.2149, over 3254157.54 frames. utt_duration=1204 frames, utt_pad_proportion=0.06575, over 10826.94 utterances.], batch size: 147, lr: 7.82e-03, grad_scale: 4.0 2023-03-08 09:49:45,251 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52849.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:50:49,487 INFO [train2.py:809] (1/4) Epoch 14, batch 1100, loss[ctc_loss=0.1106, att_loss=0.277, loss=0.2437, over 16611.00 frames. utt_duration=1415 frames, utt_pad_proportion=0.00626, over 47.00 utterances.], tot_loss[ctc_loss=0.09197, att_loss=0.2459, loss=0.2151, over 3257433.72 frames. utt_duration=1185 frames, utt_pad_proportion=0.06972, over 11010.39 utterances.], batch size: 47, lr: 7.82e-03, grad_scale: 4.0 2023-03-08 09:51:00,108 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=52897.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:51:23,351 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52911.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 09:51:47,352 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.87 vs. limit=2.0 2023-03-08 09:51:48,092 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.396e+02 2.277e+02 2.760e+02 3.306e+02 6.438e+02, threshold=5.521e+02, percent-clipped=4.0 2023-03-08 09:52:07,947 INFO [train2.py:809] (1/4) Epoch 14, batch 1150, loss[ctc_loss=0.09786, att_loss=0.2468, loss=0.217, over 16866.00 frames. utt_duration=1379 frames, utt_pad_proportion=0.007464, over 49.00 utterances.], tot_loss[ctc_loss=0.09299, att_loss=0.2463, loss=0.2157, over 3265670.47 frames. utt_duration=1204 frames, utt_pad_proportion=0.06323, over 10861.49 utterances.], batch size: 49, lr: 7.82e-03, grad_scale: 4.0 2023-03-08 09:52:52,778 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52968.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:53:26,828 INFO [train2.py:809] (1/4) Epoch 14, batch 1200, loss[ctc_loss=0.1269, att_loss=0.2633, loss=0.236, over 17305.00 frames. utt_duration=1260 frames, utt_pad_proportion=0.01164, over 55.00 utterances.], tot_loss[ctc_loss=0.09246, att_loss=0.2464, loss=0.2156, over 3276452.40 frames. utt_duration=1216 frames, utt_pad_proportion=0.05831, over 10788.56 utterances.], batch size: 55, lr: 7.81e-03, grad_scale: 8.0 2023-03-08 09:53:35,400 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-03-08 09:54:08,666 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=53016.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:54:25,773 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.353e+02 2.184e+02 2.648e+02 3.563e+02 1.260e+03, threshold=5.296e+02, percent-clipped=9.0 2023-03-08 09:54:45,537 INFO [train2.py:809] (1/4) Epoch 14, batch 1250, loss[ctc_loss=0.09709, att_loss=0.2556, loss=0.2239, over 17308.00 frames. utt_duration=1175 frames, utt_pad_proportion=0.02417, over 59.00 utterances.], tot_loss[ctc_loss=0.09138, att_loss=0.2451, loss=0.2144, over 3276586.56 frames. utt_duration=1251 frames, utt_pad_proportion=0.05072, over 10490.45 utterances.], batch size: 59, lr: 7.81e-03, grad_scale: 8.0 2023-03-08 09:55:00,086 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1514, 5.2106, 5.1460, 2.3678, 1.8752, 2.6555, 2.6620, 3.8702], device='cuda:1'), covar=tensor([0.0607, 0.0242, 0.0190, 0.4235, 0.6109, 0.2819, 0.2548, 0.1700], device='cuda:1'), in_proj_covar=tensor([0.0347, 0.0244, 0.0251, 0.0229, 0.0354, 0.0343, 0.0237, 0.0362], device='cuda:1'), out_proj_covar=tensor([1.5239e-04, 9.0386e-05, 1.0838e-04, 1.0079e-04, 1.5138e-04, 1.3669e-04, 9.4238e-05, 1.5071e-04], device='cuda:1') 2023-03-08 09:55:10,516 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53056.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:56:03,885 INFO [train2.py:809] (1/4) Epoch 14, batch 1300, loss[ctc_loss=0.08161, att_loss=0.2477, loss=0.2145, over 16466.00 frames. utt_duration=1433 frames, utt_pad_proportion=0.006704, over 46.00 utterances.], tot_loss[ctc_loss=0.09153, att_loss=0.2453, loss=0.2146, over 3278839.82 frames. utt_duration=1255 frames, utt_pad_proportion=0.05025, over 10461.89 utterances.], batch size: 46, lr: 7.81e-03, grad_scale: 8.0 2023-03-08 09:56:32,021 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53108.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:56:47,513 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53117.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 09:57:03,198 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.477e+02 2.168e+02 2.572e+02 3.596e+02 7.210e+02, threshold=5.144e+02, percent-clipped=5.0 2023-03-08 09:57:12,984 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.4263, 2.9809, 2.8651, 2.5633, 2.8959, 2.9507, 2.9399, 2.1126], device='cuda:1'), covar=tensor([0.1158, 0.1757, 0.2880, 0.5997, 0.1501, 0.3386, 0.1625, 0.6783], device='cuda:1'), in_proj_covar=tensor([0.0115, 0.0138, 0.0149, 0.0221, 0.0114, 0.0205, 0.0127, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 09:57:23,235 INFO [train2.py:809] (1/4) Epoch 14, batch 1350, loss[ctc_loss=0.06537, att_loss=0.221, loss=0.1899, over 16284.00 frames. utt_duration=1516 frames, utt_pad_proportion=0.007082, over 43.00 utterances.], tot_loss[ctc_loss=0.09094, att_loss=0.2449, loss=0.2141, over 3276838.42 frames. utt_duration=1249 frames, utt_pad_proportion=0.05174, over 10510.63 utterances.], batch size: 43, lr: 7.80e-03, grad_scale: 8.0 2023-03-08 09:58:26,003 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-03-08 09:58:41,979 INFO [train2.py:809] (1/4) Epoch 14, batch 1400, loss[ctc_loss=0.0762, att_loss=0.2475, loss=0.2133, over 16774.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.006481, over 48.00 utterances.], tot_loss[ctc_loss=0.09063, att_loss=0.2452, loss=0.2143, over 3283714.75 frames. utt_duration=1252 frames, utt_pad_proportion=0.04931, over 10500.20 utterances.], batch size: 48, lr: 7.80e-03, grad_scale: 8.0 2023-03-08 09:59:16,391 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53211.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 09:59:41,364 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.524e+02 2.260e+02 2.643e+02 3.236e+02 5.232e+02, threshold=5.286e+02, percent-clipped=1.0 2023-03-08 09:59:52,546 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.5184, 3.0387, 3.2728, 4.4977, 4.1597, 4.1990, 3.1434, 2.2928], device='cuda:1'), covar=tensor([0.0658, 0.1942, 0.1078, 0.0465, 0.0704, 0.0315, 0.1203, 0.2185], device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0214, 0.0191, 0.0206, 0.0206, 0.0163, 0.0199, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 10:00:00,308 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6627, 4.5915, 4.5769, 4.7191, 5.1945, 4.6502, 4.7252, 2.2320], device='cuda:1'), covar=tensor([0.0189, 0.0234, 0.0264, 0.0205, 0.0702, 0.0189, 0.0203, 0.2081], device='cuda:1'), in_proj_covar=tensor([0.0133, 0.0143, 0.0150, 0.0160, 0.0349, 0.0129, 0.0133, 0.0214], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0002, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 10:00:01,539 INFO [train2.py:809] (1/4) Epoch 14, batch 1450, loss[ctc_loss=0.06757, att_loss=0.2421, loss=0.2072, over 16887.00 frames. utt_duration=1380 frames, utt_pad_proportion=0.006347, over 49.00 utterances.], tot_loss[ctc_loss=0.09039, att_loss=0.2443, loss=0.2135, over 3272894.42 frames. utt_duration=1258 frames, utt_pad_proportion=0.05083, over 10420.73 utterances.], batch size: 49, lr: 7.80e-03, grad_scale: 8.0 2023-03-08 10:00:03,448 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53241.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 10:00:31,932 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=53259.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 10:00:46,115 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-03-08 10:01:21,618 INFO [train2.py:809] (1/4) Epoch 14, batch 1500, loss[ctc_loss=0.08555, att_loss=0.2356, loss=0.2056, over 16271.00 frames. utt_duration=1515 frames, utt_pad_proportion=0.007117, over 43.00 utterances.], tot_loss[ctc_loss=0.09042, att_loss=0.2444, loss=0.2136, over 3264956.09 frames. utt_duration=1233 frames, utt_pad_proportion=0.05987, over 10606.64 utterances.], batch size: 43, lr: 7.79e-03, grad_scale: 8.0 2023-03-08 10:01:41,544 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53302.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 10:01:44,576 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53304.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 10:01:55,769 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.2282, 5.4472, 5.7131, 5.5458, 5.6889, 6.1567, 5.3549, 6.2321], device='cuda:1'), covar=tensor([0.0613, 0.0680, 0.0709, 0.1108, 0.1527, 0.0776, 0.0560, 0.0619], device='cuda:1'), in_proj_covar=tensor([0.0772, 0.0457, 0.0536, 0.0593, 0.0789, 0.0544, 0.0434, 0.0533], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 10:02:02,590 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53315.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 10:02:20,570 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.404e+02 2.202e+02 2.545e+02 3.206e+02 8.434e+02, threshold=5.090e+02, percent-clipped=3.0 2023-03-08 10:02:41,148 INFO [train2.py:809] (1/4) Epoch 14, batch 1550, loss[ctc_loss=0.1171, att_loss=0.2665, loss=0.2367, over 17310.00 frames. utt_duration=1260 frames, utt_pad_proportion=0.009391, over 55.00 utterances.], tot_loss[ctc_loss=0.08988, att_loss=0.2435, loss=0.2128, over 3261242.27 frames. utt_duration=1251 frames, utt_pad_proportion=0.05755, over 10440.24 utterances.], batch size: 55, lr: 7.79e-03, grad_scale: 8.0 2023-03-08 10:03:21,734 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53365.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 10:03:24,965 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53367.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 10:03:38,803 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53376.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 10:03:56,270 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=2.06 vs. limit=2.0 2023-03-08 10:03:59,969 INFO [train2.py:809] (1/4) Epoch 14, batch 1600, loss[ctc_loss=0.07581, att_loss=0.2394, loss=0.2067, over 16389.00 frames. utt_duration=1491 frames, utt_pad_proportion=0.007021, over 44.00 utterances.], tot_loss[ctc_loss=0.09083, att_loss=0.2444, loss=0.2137, over 3263530.18 frames. utt_duration=1243 frames, utt_pad_proportion=0.05875, over 10511.37 utterances.], batch size: 44, lr: 7.78e-03, grad_scale: 8.0 2023-03-08 10:04:03,178 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.1704, 5.3934, 5.7310, 5.5083, 5.6162, 6.0628, 5.2776, 6.2016], device='cuda:1'), covar=tensor([0.0611, 0.0652, 0.0668, 0.1273, 0.1730, 0.0866, 0.0662, 0.0544], device='cuda:1'), in_proj_covar=tensor([0.0781, 0.0463, 0.0544, 0.0604, 0.0799, 0.0552, 0.0440, 0.0536], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 10:04:29,524 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53408.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 10:04:36,032 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53412.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 10:04:54,382 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5378, 4.7851, 4.7663, 4.7213, 4.8508, 4.8136, 4.4691, 4.3417], device='cuda:1'), covar=tensor([0.1067, 0.0536, 0.0313, 0.0474, 0.0314, 0.0324, 0.0390, 0.0392], device='cuda:1'), in_proj_covar=tensor([0.0476, 0.0314, 0.0282, 0.0311, 0.0364, 0.0378, 0.0310, 0.0351], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 10:04:58,719 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.459e+02 2.287e+02 2.681e+02 3.372e+02 7.467e+02, threshold=5.361e+02, percent-clipped=4.0 2023-03-08 10:05:00,594 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53428.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 10:05:19,367 INFO [train2.py:809] (1/4) Epoch 14, batch 1650, loss[ctc_loss=0.08066, att_loss=0.2379, loss=0.2065, over 16337.00 frames. utt_duration=1454 frames, utt_pad_proportion=0.005761, over 45.00 utterances.], tot_loss[ctc_loss=0.0915, att_loss=0.2453, loss=0.2145, over 3272651.21 frames. utt_duration=1235 frames, utt_pad_proportion=0.05831, over 10615.75 utterances.], batch size: 45, lr: 7.78e-03, grad_scale: 8.0 2023-03-08 10:05:44,647 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=53456.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 10:06:38,741 INFO [train2.py:809] (1/4) Epoch 14, batch 1700, loss[ctc_loss=0.06293, att_loss=0.2058, loss=0.1772, over 15623.00 frames. utt_duration=1691 frames, utt_pad_proportion=0.01006, over 37.00 utterances.], tot_loss[ctc_loss=0.09143, att_loss=0.2452, loss=0.2144, over 3280979.04 frames. utt_duration=1246 frames, utt_pad_proportion=0.05363, over 10541.51 utterances.], batch size: 37, lr: 7.78e-03, grad_scale: 8.0 2023-03-08 10:06:48,313 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4173, 2.9851, 3.5074, 2.9770, 3.3818, 4.4665, 4.3876, 3.1924], device='cuda:1'), covar=tensor([0.0361, 0.1593, 0.1236, 0.1216, 0.1070, 0.0861, 0.0512, 0.1189], device='cuda:1'), in_proj_covar=tensor([0.0230, 0.0233, 0.0255, 0.0206, 0.0246, 0.0328, 0.0233, 0.0223], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 10:07:37,131 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.475e+02 2.130e+02 2.657e+02 3.171e+02 6.492e+02, threshold=5.314e+02, percent-clipped=1.0 2023-03-08 10:07:57,988 INFO [train2.py:809] (1/4) Epoch 14, batch 1750, loss[ctc_loss=0.08607, att_loss=0.2265, loss=0.1984, over 15877.00 frames. utt_duration=1630 frames, utt_pad_proportion=0.007944, over 39.00 utterances.], tot_loss[ctc_loss=0.09085, att_loss=0.2442, loss=0.2135, over 3270922.76 frames. utt_duration=1238 frames, utt_pad_proportion=0.05852, over 10577.31 utterances.], batch size: 39, lr: 7.77e-03, grad_scale: 8.0 2023-03-08 10:08:05,954 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0594, 5.1497, 4.8426, 2.9010, 4.7640, 4.7219, 3.8648, 2.3992], device='cuda:1'), covar=tensor([0.0114, 0.0082, 0.0292, 0.1024, 0.0108, 0.0168, 0.0461, 0.1561], device='cuda:1'), in_proj_covar=tensor([0.0064, 0.0089, 0.0085, 0.0105, 0.0075, 0.0099, 0.0095, 0.0100], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 10:09:17,127 INFO [train2.py:809] (1/4) Epoch 14, batch 1800, loss[ctc_loss=0.06744, att_loss=0.2088, loss=0.1805, over 15354.00 frames. utt_duration=1756 frames, utt_pad_proportion=0.01234, over 35.00 utterances.], tot_loss[ctc_loss=0.09071, att_loss=0.2441, loss=0.2135, over 3268126.49 frames. utt_duration=1232 frames, utt_pad_proportion=0.06137, over 10627.99 utterances.], batch size: 35, lr: 7.77e-03, grad_scale: 8.0 2023-03-08 10:09:28,916 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53597.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 10:10:17,821 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.277e+02 2.192e+02 2.789e+02 3.495e+02 6.708e+02, threshold=5.578e+02, percent-clipped=5.0 2023-03-08 10:10:27,086 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6161, 4.9634, 4.8991, 4.9026, 4.9609, 4.8300, 3.4948, 4.8814], device='cuda:1'), covar=tensor([0.0099, 0.0111, 0.0108, 0.0085, 0.0105, 0.0101, 0.0729, 0.0184], device='cuda:1'), in_proj_covar=tensor([0.0079, 0.0076, 0.0095, 0.0058, 0.0065, 0.0075, 0.0095, 0.0096], 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') 2023-03-08 10:10:39,136 INFO [train2.py:809] (1/4) Epoch 14, batch 1850, loss[ctc_loss=0.06509, att_loss=0.2344, loss=0.2005, over 16542.00 frames. utt_duration=1472 frames, utt_pad_proportion=0.006167, over 45.00 utterances.], tot_loss[ctc_loss=0.09064, att_loss=0.2441, loss=0.2134, over 3256584.66 frames. utt_duration=1203 frames, utt_pad_proportion=0.07062, over 10846.01 utterances.], batch size: 45, lr: 7.77e-03, grad_scale: 8.0 2023-03-08 10:10:41,035 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.5146, 2.9572, 3.4360, 4.4727, 4.0590, 4.0630, 2.9686, 2.4225], device='cuda:1'), covar=tensor([0.0603, 0.2068, 0.0954, 0.0478, 0.0713, 0.0408, 0.1386, 0.2060], device='cuda:1'), in_proj_covar=tensor([0.0168, 0.0208, 0.0188, 0.0198, 0.0201, 0.0160, 0.0195, 0.0179], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 10:11:12,790 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53660.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 10:11:29,357 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53671.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 10:11:59,261 INFO [train2.py:809] (1/4) Epoch 14, batch 1900, loss[ctc_loss=0.06955, att_loss=0.2288, loss=0.197, over 16123.00 frames. utt_duration=1537 frames, utt_pad_proportion=0.006572, over 42.00 utterances.], tot_loss[ctc_loss=0.08975, att_loss=0.2436, loss=0.2129, over 3264463.88 frames. utt_duration=1230 frames, utt_pad_proportion=0.0612, over 10625.68 utterances.], batch size: 42, lr: 7.76e-03, grad_scale: 8.0 2023-03-08 10:12:35,610 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53712.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 10:12:52,333 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53723.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 10:12:58,261 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.175e+02 2.093e+02 2.485e+02 3.000e+02 6.734e+02, threshold=4.970e+02, percent-clipped=1.0 2023-03-08 10:13:18,791 INFO [train2.py:809] (1/4) Epoch 14, batch 1950, loss[ctc_loss=0.08362, att_loss=0.2209, loss=0.1935, over 15955.00 frames. utt_duration=1558 frames, utt_pad_proportion=0.006331, over 41.00 utterances.], tot_loss[ctc_loss=0.08938, att_loss=0.2434, loss=0.2126, over 3263465.22 frames. utt_duration=1239 frames, utt_pad_proportion=0.05802, over 10545.58 utterances.], batch size: 41, lr: 7.76e-03, grad_scale: 8.0 2023-03-08 10:13:50,782 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=53760.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 10:14:11,974 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-03-08 10:14:38,395 INFO [train2.py:809] (1/4) Epoch 14, batch 2000, loss[ctc_loss=0.08844, att_loss=0.2461, loss=0.2146, over 16756.00 frames. utt_duration=1398 frames, utt_pad_proportion=0.007206, over 48.00 utterances.], tot_loss[ctc_loss=0.0897, att_loss=0.2443, loss=0.2134, over 3264745.48 frames. utt_duration=1223 frames, utt_pad_proportion=0.06052, over 10694.75 utterances.], batch size: 48, lr: 7.76e-03, grad_scale: 8.0 2023-03-08 10:15:36,776 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.520e+02 2.201e+02 2.632e+02 3.371e+02 1.571e+03, threshold=5.264e+02, percent-clipped=3.0 2023-03-08 10:15:57,725 INFO [train2.py:809] (1/4) Epoch 14, batch 2050, loss[ctc_loss=0.07639, att_loss=0.2224, loss=0.1932, over 16189.00 frames. utt_duration=1581 frames, utt_pad_proportion=0.005752, over 41.00 utterances.], tot_loss[ctc_loss=0.08988, att_loss=0.2444, loss=0.2135, over 3263948.00 frames. utt_duration=1212 frames, utt_pad_proportion=0.06502, over 10786.93 utterances.], batch size: 41, lr: 7.75e-03, grad_scale: 8.0 2023-03-08 10:16:56,643 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.5581, 1.6160, 1.7687, 2.5164, 2.3603, 1.8972, 1.8362, 2.6551], device='cuda:1'), covar=tensor([0.3046, 0.8101, 0.5494, 0.2732, 0.4017, 0.3681, 0.6309, 0.1906], device='cuda:1'), in_proj_covar=tensor([0.0088, 0.0097, 0.0098, 0.0082, 0.0087, 0.0079, 0.0102, 0.0070], device='cuda:1'), out_proj_covar=tensor([6.2138e-05, 7.0262e-05, 7.2580e-05, 6.0582e-05, 6.1601e-05, 6.0273e-05, 7.1842e-05, 5.4096e-05], device='cuda:1') 2023-03-08 10:17:18,528 INFO [train2.py:809] (1/4) Epoch 14, batch 2100, loss[ctc_loss=0.07753, att_loss=0.2214, loss=0.1926, over 11451.00 frames. utt_duration=1834 frames, utt_pad_proportion=0.1852, over 25.00 utterances.], tot_loss[ctc_loss=0.0911, att_loss=0.2453, loss=0.2144, over 3263768.12 frames. utt_duration=1197 frames, utt_pad_proportion=0.06787, over 10922.82 utterances.], batch size: 25, lr: 7.75e-03, grad_scale: 8.0 2023-03-08 10:17:30,133 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53897.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 10:17:43,255 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8628, 5.1731, 5.0119, 5.0975, 5.2311, 5.2135, 4.8654, 4.7013], device='cuda:1'), covar=tensor([0.1171, 0.0563, 0.0410, 0.0511, 0.0331, 0.0308, 0.0420, 0.0368], device='cuda:1'), in_proj_covar=tensor([0.0476, 0.0313, 0.0281, 0.0311, 0.0365, 0.0379, 0.0312, 0.0350], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 10:18:17,164 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.409e+02 2.088e+02 2.471e+02 3.126e+02 7.099e+02, threshold=4.942e+02, percent-clipped=2.0 2023-03-08 10:18:37,841 INFO [train2.py:809] (1/4) Epoch 14, batch 2150, loss[ctc_loss=0.08154, att_loss=0.2418, loss=0.2097, over 16135.00 frames. utt_duration=1538 frames, utt_pad_proportion=0.005128, over 42.00 utterances.], tot_loss[ctc_loss=0.0901, att_loss=0.2442, loss=0.2134, over 3263016.17 frames. utt_duration=1241 frames, utt_pad_proportion=0.05785, over 10528.70 utterances.], batch size: 42, lr: 7.75e-03, grad_scale: 8.0 2023-03-08 10:18:45,614 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=53945.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 10:19:10,073 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53960.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 10:19:27,841 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53971.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 10:19:54,378 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1842, 2.5739, 4.5639, 3.6016, 2.9161, 3.9817, 4.0583, 4.2862], device='cuda:1'), covar=tensor([0.0230, 0.1528, 0.0110, 0.1076, 0.1707, 0.0269, 0.0205, 0.0270], device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0243, 0.0149, 0.0302, 0.0265, 0.0189, 0.0135, 0.0164], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 10:19:57,017 INFO [train2.py:809] (1/4) Epoch 14, batch 2200, loss[ctc_loss=0.07535, att_loss=0.2167, loss=0.1884, over 15863.00 frames. utt_duration=1629 frames, utt_pad_proportion=0.0105, over 39.00 utterances.], tot_loss[ctc_loss=0.09019, att_loss=0.2439, loss=0.2132, over 3260535.47 frames. utt_duration=1237 frames, utt_pad_proportion=0.05937, over 10556.90 utterances.], batch size: 39, lr: 7.74e-03, grad_scale: 8.0 2023-03-08 10:20:30,046 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=54008.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 10:20:47,495 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=54019.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 10:20:53,795 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=54023.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 10:20:59,197 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.392e+02 2.188e+02 2.816e+02 3.562e+02 5.650e+02, threshold=5.631e+02, percent-clipped=6.0 2023-03-08 10:21:20,205 INFO [train2.py:809] (1/4) Epoch 14, batch 2250, loss[ctc_loss=0.08777, att_loss=0.2362, loss=0.2065, over 15953.00 frames. utt_duration=1558 frames, utt_pad_proportion=0.006533, over 41.00 utterances.], tot_loss[ctc_loss=0.09056, att_loss=0.2439, loss=0.2132, over 3260850.97 frames. utt_duration=1237 frames, utt_pad_proportion=0.06084, over 10559.25 utterances.], batch size: 41, lr: 7.74e-03, grad_scale: 8.0 2023-03-08 10:21:25,112 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4574, 2.9335, 3.5914, 2.6644, 3.3454, 4.6292, 4.4125, 2.9820], device='cuda:1'), covar=tensor([0.0486, 0.1921, 0.1150, 0.1724, 0.1270, 0.0775, 0.0594, 0.1657], device='cuda:1'), in_proj_covar=tensor([0.0233, 0.0233, 0.0255, 0.0207, 0.0247, 0.0329, 0.0234, 0.0226], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 10:21:36,581 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4867, 2.4595, 4.9446, 3.8238, 2.9305, 4.1989, 4.4622, 4.5738], device='cuda:1'), covar=tensor([0.0194, 0.1693, 0.0093, 0.0923, 0.1715, 0.0258, 0.0159, 0.0234], device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0245, 0.0150, 0.0306, 0.0267, 0.0192, 0.0137, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 10:21:54,051 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8254, 5.0458, 4.4097, 5.3361, 4.6430, 4.9286, 5.1901, 4.9846], device='cuda:1'), covar=tensor([0.0640, 0.0344, 0.1202, 0.0245, 0.0420, 0.0296, 0.0352, 0.0232], device='cuda:1'), in_proj_covar=tensor([0.0358, 0.0280, 0.0334, 0.0288, 0.0286, 0.0218, 0.0267, 0.0250], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0005, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 10:22:09,525 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=54071.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 10:22:39,618 INFO [train2.py:809] (1/4) Epoch 14, batch 2300, loss[ctc_loss=0.13, att_loss=0.2646, loss=0.2377, over 13952.00 frames. utt_duration=383.8 frames, utt_pad_proportion=0.3302, over 146.00 utterances.], tot_loss[ctc_loss=0.09018, att_loss=0.2434, loss=0.2127, over 3262358.13 frames. utt_duration=1237 frames, utt_pad_proportion=0.06051, over 10561.77 utterances.], batch size: 146, lr: 7.73e-03, grad_scale: 8.0 2023-03-08 10:23:28,210 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0682, 4.5980, 4.7225, 4.6724, 2.9687, 4.7667, 2.8419, 1.8436], device='cuda:1'), covar=tensor([0.0415, 0.0189, 0.0527, 0.0148, 0.1542, 0.0127, 0.1417, 0.1641], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0130, 0.0251, 0.0120, 0.0220, 0.0115, 0.0223, 0.0196], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 10:23:36,762 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.455e+02 2.251e+02 2.657e+02 3.358e+02 7.808e+02, threshold=5.315e+02, percent-clipped=1.0 2023-03-08 10:23:57,483 INFO [train2.py:809] (1/4) Epoch 14, batch 2350, loss[ctc_loss=0.1049, att_loss=0.263, loss=0.2314, over 17107.00 frames. utt_duration=1223 frames, utt_pad_proportion=0.01598, over 56.00 utterances.], tot_loss[ctc_loss=0.09093, att_loss=0.2437, loss=0.2131, over 3256532.53 frames. utt_duration=1227 frames, utt_pad_proportion=0.06543, over 10631.76 utterances.], batch size: 56, lr: 7.73e-03, grad_scale: 8.0 2023-03-08 10:25:16,439 INFO [train2.py:809] (1/4) Epoch 14, batch 2400, loss[ctc_loss=0.09849, att_loss=0.2497, loss=0.2195, over 17062.00 frames. utt_duration=1314 frames, utt_pad_proportion=0.007655, over 52.00 utterances.], tot_loss[ctc_loss=0.09018, att_loss=0.2432, loss=0.2126, over 3255608.74 frames. utt_duration=1254 frames, utt_pad_proportion=0.05993, over 10397.38 utterances.], batch size: 52, lr: 7.73e-03, grad_scale: 8.0 2023-03-08 10:25:19,784 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8975, 3.2607, 3.7948, 3.1902, 3.7715, 4.8886, 4.7240, 3.4790], device='cuda:1'), covar=tensor([0.0271, 0.1535, 0.1051, 0.1276, 0.0842, 0.0682, 0.0391, 0.1211], device='cuda:1'), in_proj_covar=tensor([0.0230, 0.0229, 0.0252, 0.0204, 0.0241, 0.0326, 0.0232, 0.0222], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 10:25:34,378 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1993, 5.4319, 4.8646, 5.2831, 5.0341, 4.7242, 4.9067, 4.6960], device='cuda:1'), covar=tensor([0.1162, 0.1006, 0.0949, 0.0812, 0.0967, 0.1594, 0.2269, 0.2204], device='cuda:1'), in_proj_covar=tensor([0.0466, 0.0537, 0.0405, 0.0401, 0.0384, 0.0436, 0.0554, 0.0484], 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') 2023-03-08 10:26:15,331 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.443e+02 2.183e+02 2.531e+02 3.194e+02 7.343e+02, threshold=5.062e+02, percent-clipped=2.0 2023-03-08 10:26:35,975 INFO [train2.py:809] (1/4) Epoch 14, batch 2450, loss[ctc_loss=0.07175, att_loss=0.2257, loss=0.1949, over 15963.00 frames. utt_duration=1559 frames, utt_pad_proportion=0.006513, over 41.00 utterances.], tot_loss[ctc_loss=0.09044, att_loss=0.2439, loss=0.2132, over 3262508.98 frames. utt_duration=1218 frames, utt_pad_proportion=0.06625, over 10731.84 utterances.], batch size: 41, lr: 7.72e-03, grad_scale: 8.0 2023-03-08 10:27:54,976 INFO [train2.py:809] (1/4) Epoch 14, batch 2500, loss[ctc_loss=0.08367, att_loss=0.2221, loss=0.1944, over 14452.00 frames. utt_duration=1808 frames, utt_pad_proportion=0.04771, over 32.00 utterances.], tot_loss[ctc_loss=0.08954, att_loss=0.2429, loss=0.2122, over 3258450.41 frames. utt_duration=1240 frames, utt_pad_proportion=0.06104, over 10525.43 utterances.], batch size: 32, lr: 7.72e-03, grad_scale: 8.0 2023-03-08 10:28:13,721 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-03-08 10:28:53,991 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.418e+02 2.435e+02 2.776e+02 3.474e+02 9.434e+02, threshold=5.553e+02, percent-clipped=7.0 2023-03-08 10:29:14,829 INFO [train2.py:809] (1/4) Epoch 14, batch 2550, loss[ctc_loss=0.09656, att_loss=0.2508, loss=0.2199, over 17350.00 frames. utt_duration=1103 frames, utt_pad_proportion=0.0365, over 63.00 utterances.], tot_loss[ctc_loss=0.09023, att_loss=0.2434, loss=0.2128, over 3259797.08 frames. utt_duration=1253 frames, utt_pad_proportion=0.05652, over 10420.27 utterances.], batch size: 63, lr: 7.72e-03, grad_scale: 8.0 2023-03-08 10:29:36,164 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5751, 4.8776, 4.3788, 4.9658, 4.3672, 4.6302, 5.0403, 4.8351], device='cuda:1'), covar=tensor([0.0649, 0.0331, 0.0865, 0.0312, 0.0457, 0.0343, 0.0256, 0.0206], device='cuda:1'), in_proj_covar=tensor([0.0361, 0.0281, 0.0338, 0.0291, 0.0289, 0.0221, 0.0271, 0.0251], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0005, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 10:30:29,517 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0616, 5.3383, 5.6341, 5.3976, 5.5260, 5.9875, 5.2298, 6.1172], device='cuda:1'), covar=tensor([0.0761, 0.0679, 0.0736, 0.1184, 0.1825, 0.1024, 0.0643, 0.0680], device='cuda:1'), in_proj_covar=tensor([0.0779, 0.0457, 0.0536, 0.0600, 0.0791, 0.0545, 0.0440, 0.0530], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 10:30:37,349 INFO [train2.py:809] (1/4) Epoch 14, batch 2600, loss[ctc_loss=0.08053, att_loss=0.2403, loss=0.2084, over 16293.00 frames. utt_duration=1517 frames, utt_pad_proportion=0.005715, over 43.00 utterances.], tot_loss[ctc_loss=0.09033, att_loss=0.2438, loss=0.2131, over 3270303.24 frames. utt_duration=1263 frames, utt_pad_proportion=0.05041, over 10366.94 utterances.], batch size: 43, lr: 7.71e-03, grad_scale: 8.0 2023-03-08 10:31:18,573 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-03-08 10:31:39,669 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.402e+02 2.230e+02 2.611e+02 3.212e+02 5.306e+02, threshold=5.223e+02, percent-clipped=0.0 2023-03-08 10:31:56,773 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.5656, 5.8296, 5.2517, 5.6349, 5.4721, 5.0558, 5.2365, 5.0091], device='cuda:1'), covar=tensor([0.1304, 0.0922, 0.0927, 0.0745, 0.0813, 0.1534, 0.2504, 0.2598], device='cuda:1'), in_proj_covar=tensor([0.0470, 0.0540, 0.0410, 0.0404, 0.0386, 0.0440, 0.0555, 0.0490], 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') 2023-03-08 10:32:01,195 INFO [train2.py:809] (1/4) Epoch 14, batch 2650, loss[ctc_loss=0.1088, att_loss=0.2594, loss=0.2293, over 16873.00 frames. utt_duration=1379 frames, utt_pad_proportion=0.007346, over 49.00 utterances.], tot_loss[ctc_loss=0.09113, att_loss=0.244, loss=0.2134, over 3252251.66 frames. utt_duration=1232 frames, utt_pad_proportion=0.0628, over 10569.42 utterances.], batch size: 49, lr: 7.71e-03, grad_scale: 8.0 2023-03-08 10:33:16,259 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.6485, 5.9176, 5.4114, 5.7334, 5.5599, 5.1904, 5.3264, 5.1796], device='cuda:1'), covar=tensor([0.1316, 0.0947, 0.0866, 0.0793, 0.0880, 0.1543, 0.2434, 0.2410], device='cuda:1'), in_proj_covar=tensor([0.0475, 0.0545, 0.0413, 0.0408, 0.0389, 0.0444, 0.0560, 0.0493], 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') 2023-03-08 10:33:23,997 INFO [train2.py:809] (1/4) Epoch 14, batch 2700, loss[ctc_loss=0.09305, att_loss=0.2428, loss=0.2129, over 16021.00 frames. utt_duration=1604 frames, utt_pad_proportion=0.006351, over 40.00 utterances.], tot_loss[ctc_loss=0.09089, att_loss=0.2442, loss=0.2135, over 3261415.47 frames. utt_duration=1259 frames, utt_pad_proportion=0.05381, over 10375.49 utterances.], batch size: 40, lr: 7.71e-03, grad_scale: 8.0 2023-03-08 10:33:48,192 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.7030, 3.8779, 3.5131, 3.7394, 3.9459, 3.7340, 3.6544, 2.4613], device='cuda:1'), covar=tensor([0.0245, 0.0325, 0.0450, 0.0342, 0.0752, 0.0254, 0.0366, 0.1746], device='cuda:1'), in_proj_covar=tensor([0.0133, 0.0145, 0.0152, 0.0160, 0.0348, 0.0131, 0.0137, 0.0213], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0002, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 10:34:25,553 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.212e+02 2.188e+02 2.730e+02 3.256e+02 6.934e+02, threshold=5.460e+02, percent-clipped=5.0 2023-03-08 10:34:38,226 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-03-08 10:34:47,228 INFO [train2.py:809] (1/4) Epoch 14, batch 2750, loss[ctc_loss=0.115, att_loss=0.2684, loss=0.2377, over 17061.00 frames. utt_duration=1289 frames, utt_pad_proportion=0.009093, over 53.00 utterances.], tot_loss[ctc_loss=0.09245, att_loss=0.2454, loss=0.2148, over 3264231.85 frames. utt_duration=1214 frames, utt_pad_proportion=0.06459, over 10772.43 utterances.], batch size: 53, lr: 7.70e-03, grad_scale: 8.0 2023-03-08 10:36:10,907 INFO [train2.py:809] (1/4) Epoch 14, batch 2800, loss[ctc_loss=0.0917, att_loss=0.2308, loss=0.203, over 16272.00 frames. utt_duration=1515 frames, utt_pad_proportion=0.0076, over 43.00 utterances.], tot_loss[ctc_loss=0.09178, att_loss=0.245, loss=0.2144, over 3264776.47 frames. utt_duration=1223 frames, utt_pad_proportion=0.06258, over 10688.46 utterances.], batch size: 43, lr: 7.70e-03, grad_scale: 8.0 2023-03-08 10:36:43,942 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5595, 2.3554, 5.0394, 4.1687, 2.9375, 4.4463, 4.9644, 4.6962], device='cuda:1'), covar=tensor([0.0173, 0.1607, 0.0211, 0.0741, 0.1749, 0.0157, 0.0106, 0.0180], device='cuda:1'), in_proj_covar=tensor([0.0161, 0.0245, 0.0151, 0.0307, 0.0268, 0.0191, 0.0139, 0.0167], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 10:37:12,827 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.585e+02 2.211e+02 2.653e+02 3.453e+02 1.073e+03, threshold=5.307e+02, percent-clipped=4.0 2023-03-08 10:37:34,795 INFO [train2.py:809] (1/4) Epoch 14, batch 2850, loss[ctc_loss=0.07447, att_loss=0.2238, loss=0.194, over 14532.00 frames. utt_duration=1818 frames, utt_pad_proportion=0.03233, over 32.00 utterances.], tot_loss[ctc_loss=0.09195, att_loss=0.2448, loss=0.2143, over 3259580.87 frames. utt_duration=1207 frames, utt_pad_proportion=0.06802, over 10819.79 utterances.], batch size: 32, lr: 7.70e-03, grad_scale: 8.0 2023-03-08 10:38:09,855 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9015, 5.1771, 5.3987, 5.3119, 5.3510, 5.8866, 5.1654, 5.9627], device='cuda:1'), covar=tensor([0.0733, 0.0715, 0.0786, 0.1161, 0.1847, 0.0793, 0.0641, 0.0625], device='cuda:1'), in_proj_covar=tensor([0.0775, 0.0456, 0.0537, 0.0597, 0.0793, 0.0542, 0.0441, 0.0533], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 10:38:14,641 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=54664.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 10:38:57,461 INFO [train2.py:809] (1/4) Epoch 14, batch 2900, loss[ctc_loss=0.07683, att_loss=0.2336, loss=0.2022, over 16166.00 frames. utt_duration=1579 frames, utt_pad_proportion=0.007102, over 41.00 utterances.], tot_loss[ctc_loss=0.09195, att_loss=0.2454, loss=0.2147, over 3272434.63 frames. utt_duration=1211 frames, utt_pad_proportion=0.06316, over 10824.74 utterances.], batch size: 41, lr: 7.69e-03, grad_scale: 8.0 2023-03-08 10:39:01,170 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5589, 4.5002, 4.4303, 4.4910, 5.0107, 4.5752, 4.4841, 2.2124], device='cuda:1'), covar=tensor([0.0156, 0.0265, 0.0288, 0.0205, 0.0976, 0.0170, 0.0295, 0.2097], device='cuda:1'), in_proj_covar=tensor([0.0131, 0.0143, 0.0150, 0.0160, 0.0344, 0.0129, 0.0135, 0.0210], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0002, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 10:39:35,118 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-03-08 10:39:50,396 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 2023-03-08 10:39:55,596 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=54725.0, num_to_drop=1, layers_to_drop={3} 2023-03-08 10:39:58,433 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.349e+02 2.074e+02 2.453e+02 3.116e+02 5.349e+02, threshold=4.905e+02, percent-clipped=1.0 2023-03-08 10:40:19,748 INFO [train2.py:809] (1/4) Epoch 14, batch 2950, loss[ctc_loss=0.09036, att_loss=0.2321, loss=0.2037, over 14509.00 frames. utt_duration=1815 frames, utt_pad_proportion=0.03394, over 32.00 utterances.], tot_loss[ctc_loss=0.092, att_loss=0.2453, loss=0.2147, over 3266322.95 frames. utt_duration=1221 frames, utt_pad_proportion=0.06219, over 10717.20 utterances.], batch size: 32, lr: 7.69e-03, grad_scale: 8.0 2023-03-08 10:40:20,174 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.0872, 2.3195, 2.5862, 2.6241, 2.5485, 2.5625, 2.4365, 3.0694], device='cuda:1'), covar=tensor([0.2401, 0.3597, 0.2651, 0.1253, 0.1813, 0.1209, 0.2632, 0.0787], device='cuda:1'), in_proj_covar=tensor([0.0087, 0.0093, 0.0095, 0.0080, 0.0085, 0.0076, 0.0098, 0.0069], device='cuda:1'), out_proj_covar=tensor([6.1680e-05, 6.7896e-05, 7.0610e-05, 5.9145e-05, 6.0254e-05, 5.8788e-05, 6.9853e-05, 5.3470e-05], device='cuda:1') 2023-03-08 10:40:43,159 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2433, 4.5027, 4.8779, 4.8180, 2.9856, 5.0394, 3.3582, 1.9303], device='cuda:1'), covar=tensor([0.0349, 0.0240, 0.0556, 0.0117, 0.1651, 0.0094, 0.1124, 0.1734], device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0132, 0.0252, 0.0121, 0.0220, 0.0115, 0.0224, 0.0197], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 10:40:55,184 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.84 vs. limit=2.0 2023-03-08 10:41:12,564 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.5693, 3.5827, 3.3643, 2.9894, 3.5985, 3.4948, 3.4532, 2.4729], device='cuda:1'), covar=tensor([0.1067, 0.1515, 0.3711, 0.5196, 0.1040, 0.3921, 0.1670, 0.6700], device='cuda:1'), in_proj_covar=tensor([0.0116, 0.0139, 0.0151, 0.0220, 0.0114, 0.0206, 0.0129, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 10:41:41,651 INFO [train2.py:809] (1/4) Epoch 14, batch 3000, loss[ctc_loss=0.07741, att_loss=0.2366, loss=0.2048, over 15955.00 frames. utt_duration=1558 frames, utt_pad_proportion=0.006315, over 41.00 utterances.], tot_loss[ctc_loss=0.09175, att_loss=0.2446, loss=0.2141, over 3261105.38 frames. utt_duration=1211 frames, utt_pad_proportion=0.06589, over 10783.05 utterances.], batch size: 41, lr: 7.69e-03, grad_scale: 8.0 2023-03-08 10:41:41,652 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-08 10:41:56,407 INFO [train2.py:843] (1/4) Epoch 14, validation: ctc_loss=0.0441, att_loss=0.2368, loss=0.1983, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-08 10:41:56,408 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-08 10:42:17,203 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-03-08 10:42:55,795 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.163e+02 2.238e+02 2.720e+02 3.334e+02 7.498e+02, threshold=5.441e+02, percent-clipped=5.0 2023-03-08 10:43:16,802 INFO [train2.py:809] (1/4) Epoch 14, batch 3050, loss[ctc_loss=0.1188, att_loss=0.2688, loss=0.2388, over 17037.00 frames. utt_duration=1287 frames, utt_pad_proportion=0.01054, over 53.00 utterances.], tot_loss[ctc_loss=0.09212, att_loss=0.2456, loss=0.2149, over 3270256.29 frames. utt_duration=1204 frames, utt_pad_proportion=0.06509, over 10875.54 utterances.], batch size: 53, lr: 7.68e-03, grad_scale: 16.0 2023-03-08 10:43:52,894 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=54862.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 10:44:38,677 INFO [train2.py:809] (1/4) Epoch 14, batch 3100, loss[ctc_loss=0.06349, att_loss=0.2076, loss=0.1788, over 15492.00 frames. utt_duration=1723 frames, utt_pad_proportion=0.00936, over 36.00 utterances.], tot_loss[ctc_loss=0.09139, att_loss=0.2447, loss=0.214, over 3269916.62 frames. utt_duration=1226 frames, utt_pad_proportion=0.06022, over 10685.90 utterances.], batch size: 36, lr: 7.68e-03, grad_scale: 16.0 2023-03-08 10:45:32,213 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=54923.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 10:45:32,617 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 2023-03-08 10:45:38,026 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.392e+02 2.291e+02 2.720e+02 3.507e+02 7.867e+02, threshold=5.439e+02, percent-clipped=2.0 2023-03-08 10:45:42,145 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1912, 5.1238, 4.9183, 3.2265, 4.9711, 4.7859, 4.2625, 2.7869], device='cuda:1'), covar=tensor([0.0098, 0.0081, 0.0237, 0.0868, 0.0080, 0.0178, 0.0335, 0.1287], device='cuda:1'), in_proj_covar=tensor([0.0065, 0.0090, 0.0088, 0.0106, 0.0075, 0.0100, 0.0096, 0.0101], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 10:45:42,245 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.4178, 2.2287, 2.2214, 2.8693, 2.2451, 2.4220, 2.2446, 3.0194], device='cuda:1'), covar=tensor([0.1975, 0.3475, 0.2523, 0.1118, 0.1990, 0.1543, 0.2972, 0.0920], device='cuda:1'), in_proj_covar=tensor([0.0086, 0.0092, 0.0095, 0.0079, 0.0085, 0.0077, 0.0098, 0.0068], device='cuda:1'), out_proj_covar=tensor([6.1217e-05, 6.7260e-05, 7.0555e-05, 5.8590e-05, 6.0022e-05, 5.8926e-05, 6.9683e-05, 5.2796e-05], device='cuda:1') 2023-03-08 10:45:59,456 INFO [train2.py:809] (1/4) Epoch 14, batch 3150, loss[ctc_loss=0.09828, att_loss=0.2492, loss=0.2191, over 16406.00 frames. utt_duration=1493 frames, utt_pad_proportion=0.007359, over 44.00 utterances.], tot_loss[ctc_loss=0.0911, att_loss=0.2447, loss=0.214, over 3266133.38 frames. utt_duration=1237 frames, utt_pad_proportion=0.05744, over 10577.93 utterances.], batch size: 44, lr: 7.68e-03, grad_scale: 16.0 2023-03-08 10:46:42,504 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7642, 4.7907, 4.6358, 4.5641, 5.2749, 4.8416, 4.6904, 2.5746], device='cuda:1'), covar=tensor([0.0156, 0.0249, 0.0244, 0.0294, 0.0866, 0.0158, 0.0286, 0.1896], device='cuda:1'), in_proj_covar=tensor([0.0132, 0.0144, 0.0151, 0.0162, 0.0346, 0.0130, 0.0137, 0.0211], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0002, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 10:47:20,535 INFO [train2.py:809] (1/4) Epoch 14, batch 3200, loss[ctc_loss=0.07209, att_loss=0.2159, loss=0.1871, over 15873.00 frames. utt_duration=1630 frames, utt_pad_proportion=0.01, over 39.00 utterances.], tot_loss[ctc_loss=0.09065, att_loss=0.2446, loss=0.2138, over 3260676.11 frames. utt_duration=1247 frames, utt_pad_proportion=0.05697, over 10468.87 utterances.], batch size: 39, lr: 7.67e-03, grad_scale: 16.0 2023-03-08 10:47:59,511 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-03-08 10:48:12,455 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55020.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 10:48:23,966 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.330e+02 2.213e+02 2.608e+02 3.360e+02 8.145e+02, threshold=5.216e+02, percent-clipped=2.0 2023-03-08 10:48:46,131 INFO [train2.py:809] (1/4) Epoch 14, batch 3250, loss[ctc_loss=0.1042, att_loss=0.2495, loss=0.2205, over 17450.00 frames. utt_duration=1013 frames, utt_pad_proportion=0.04527, over 69.00 utterances.], tot_loss[ctc_loss=0.09089, att_loss=0.2447, loss=0.2139, over 3258248.15 frames. utt_duration=1241 frames, utt_pad_proportion=0.05914, over 10511.68 utterances.], batch size: 69, lr: 7.67e-03, grad_scale: 16.0 2023-03-08 10:49:00,493 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1515, 5.2300, 5.0058, 2.5308, 2.0998, 2.8143, 2.6873, 3.8814], device='cuda:1'), covar=tensor([0.0643, 0.0261, 0.0245, 0.3985, 0.5482, 0.2576, 0.2868, 0.1698], device='cuda:1'), in_proj_covar=tensor([0.0339, 0.0237, 0.0243, 0.0220, 0.0345, 0.0333, 0.0232, 0.0354], device='cuda:1'), out_proj_covar=tensor([1.4844e-04, 8.8163e-05, 1.0492e-04, 9.5869e-05, 1.4705e-04, 1.3229e-04, 9.2401e-05, 1.4706e-04], device='cuda:1') 2023-03-08 10:49:07,723 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4868, 2.7343, 5.1412, 3.8823, 3.0444, 4.2722, 4.8517, 4.6971], device='cuda:1'), covar=tensor([0.0265, 0.1694, 0.0160, 0.1013, 0.1863, 0.0284, 0.0121, 0.0228], device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0250, 0.0154, 0.0313, 0.0271, 0.0194, 0.0141, 0.0170], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0003, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 10:49:20,085 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-03-08 10:49:48,283 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6250, 4.5595, 4.5392, 4.4459, 5.0283, 4.6801, 4.4699, 2.3679], device='cuda:1'), covar=tensor([0.0161, 0.0235, 0.0242, 0.0254, 0.0826, 0.0155, 0.0302, 0.2116], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0147, 0.0154, 0.0166, 0.0353, 0.0133, 0.0141, 0.0216], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 10:50:12,432 INFO [train2.py:809] (1/4) Epoch 14, batch 3300, loss[ctc_loss=0.1149, att_loss=0.2561, loss=0.2279, over 16390.00 frames. utt_duration=1492 frames, utt_pad_proportion=0.007546, over 44.00 utterances.], tot_loss[ctc_loss=0.09122, att_loss=0.2452, loss=0.2144, over 3267710.07 frames. utt_duration=1237 frames, utt_pad_proportion=0.05669, over 10576.88 utterances.], batch size: 44, lr: 7.66e-03, grad_scale: 16.0 2023-03-08 10:50:45,153 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.3566, 1.5533, 2.0158, 2.3631, 2.0327, 2.1227, 1.6579, 2.3691], device='cuda:1'), covar=tensor([0.1271, 0.3280, 0.2126, 0.0925, 0.2159, 0.1019, 0.1400, 0.0833], device='cuda:1'), in_proj_covar=tensor([0.0088, 0.0093, 0.0096, 0.0080, 0.0085, 0.0077, 0.0099, 0.0068], device='cuda:1'), out_proj_covar=tensor([6.2066e-05, 6.8162e-05, 7.0997e-05, 5.9460e-05, 6.0447e-05, 5.9498e-05, 7.0467e-05, 5.3091e-05], device='cuda:1') 2023-03-08 10:51:15,768 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.495e+02 2.260e+02 2.683e+02 3.437e+02 9.607e+02, threshold=5.365e+02, percent-clipped=9.0 2023-03-08 10:51:37,889 INFO [train2.py:809] (1/4) Epoch 14, batch 3350, loss[ctc_loss=0.0936, att_loss=0.2522, loss=0.2204, over 17324.00 frames. utt_duration=1261 frames, utt_pad_proportion=0.01067, over 55.00 utterances.], tot_loss[ctc_loss=0.09082, att_loss=0.245, loss=0.2141, over 3275057.44 frames. utt_duration=1234 frames, utt_pad_proportion=0.05629, over 10626.61 utterances.], batch size: 55, lr: 7.66e-03, grad_scale: 16.0 2023-03-08 10:51:51,530 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55148.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 10:53:01,327 INFO [train2.py:809] (1/4) Epoch 14, batch 3400, loss[ctc_loss=0.08365, att_loss=0.2541, loss=0.22, over 16871.00 frames. utt_duration=1379 frames, utt_pad_proportion=0.007214, over 49.00 utterances.], tot_loss[ctc_loss=0.09021, att_loss=0.2441, loss=0.2133, over 3275099.41 frames. utt_duration=1259 frames, utt_pad_proportion=0.05162, over 10421.10 utterances.], batch size: 49, lr: 7.66e-03, grad_scale: 16.0 2023-03-08 10:53:34,752 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55209.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 10:53:38,113 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55211.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 10:53:50,358 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55218.0, num_to_drop=1, layers_to_drop={3} 2023-03-08 10:54:05,572 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.493e+02 2.249e+02 2.663e+02 3.402e+02 7.529e+02, threshold=5.327e+02, percent-clipped=2.0 2023-03-08 10:54:27,348 INFO [train2.py:809] (1/4) Epoch 14, batch 3450, loss[ctc_loss=0.07864, att_loss=0.2153, loss=0.188, over 14534.00 frames. utt_duration=1819 frames, utt_pad_proportion=0.03577, over 32.00 utterances.], tot_loss[ctc_loss=0.08968, att_loss=0.2438, loss=0.213, over 3277440.07 frames. utt_duration=1272 frames, utt_pad_proportion=0.04841, over 10321.81 utterances.], batch size: 32, lr: 7.65e-03, grad_scale: 16.0 2023-03-08 10:55:20,536 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.6035, 3.9045, 3.9627, 2.4298, 2.2557, 2.7294, 2.4602, 3.4273], device='cuda:1'), covar=tensor([0.0827, 0.0322, 0.0352, 0.3263, 0.4277, 0.2264, 0.2382, 0.1515], device='cuda:1'), in_proj_covar=tensor([0.0341, 0.0236, 0.0242, 0.0220, 0.0343, 0.0335, 0.0233, 0.0355], device='cuda:1'), out_proj_covar=tensor([1.4902e-04, 8.8062e-05, 1.0406e-04, 9.6076e-05, 1.4662e-04, 1.3287e-04, 9.2615e-05, 1.4734e-04], device='cuda:1') 2023-03-08 10:55:22,000 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55272.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 10:55:52,456 INFO [train2.py:809] (1/4) Epoch 14, batch 3500, loss[ctc_loss=0.08751, att_loss=0.2522, loss=0.2193, over 17517.00 frames. utt_duration=877.5 frames, utt_pad_proportion=0.07532, over 80.00 utterances.], tot_loss[ctc_loss=0.09072, att_loss=0.2445, loss=0.2138, over 3272696.39 frames. utt_duration=1234 frames, utt_pad_proportion=0.05854, over 10620.42 utterances.], batch size: 80, lr: 7.65e-03, grad_scale: 8.0 2023-03-08 10:56:43,462 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55320.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 10:56:56,799 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.444e+02 2.024e+02 2.386e+02 2.873e+02 9.037e+02, threshold=4.771e+02, percent-clipped=2.0 2023-03-08 10:57:17,384 INFO [train2.py:809] (1/4) Epoch 14, batch 3550, loss[ctc_loss=0.08922, att_loss=0.2525, loss=0.2199, over 16855.00 frames. utt_duration=1378 frames, utt_pad_proportion=0.008243, over 49.00 utterances.], tot_loss[ctc_loss=0.09104, att_loss=0.2445, loss=0.2138, over 3271640.27 frames. utt_duration=1235 frames, utt_pad_proportion=0.05879, over 10610.52 utterances.], batch size: 49, lr: 7.65e-03, grad_scale: 8.0 2023-03-08 10:58:05,156 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=55368.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 10:58:22,209 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.7084, 5.9910, 5.4290, 5.7125, 5.5931, 5.2090, 5.3361, 5.1559], device='cuda:1'), covar=tensor([0.1349, 0.0850, 0.0834, 0.0726, 0.0825, 0.1551, 0.2521, 0.2349], device='cuda:1'), in_proj_covar=tensor([0.0473, 0.0545, 0.0416, 0.0408, 0.0394, 0.0444, 0.0565, 0.0496], 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') 2023-03-08 10:58:41,901 INFO [train2.py:809] (1/4) Epoch 14, batch 3600, loss[ctc_loss=0.1215, att_loss=0.2726, loss=0.2424, over 17303.00 frames. utt_duration=877.6 frames, utt_pad_proportion=0.08104, over 79.00 utterances.], tot_loss[ctc_loss=0.09076, att_loss=0.2446, loss=0.2139, over 3273979.35 frames. utt_duration=1231 frames, utt_pad_proportion=0.05921, over 10652.95 utterances.], batch size: 79, lr: 7.64e-03, grad_scale: 8.0 2023-03-08 10:59:46,538 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.378e+02 2.040e+02 2.412e+02 2.968e+02 1.250e+03, threshold=4.823e+02, percent-clipped=3.0 2023-03-08 11:00:07,017 INFO [train2.py:809] (1/4) Epoch 14, batch 3650, loss[ctc_loss=0.08335, att_loss=0.2402, loss=0.2088, over 16275.00 frames. utt_duration=1515 frames, utt_pad_proportion=0.007026, over 43.00 utterances.], tot_loss[ctc_loss=0.09052, att_loss=0.2441, loss=0.2134, over 3257025.82 frames. utt_duration=1220 frames, utt_pad_proportion=0.06575, over 10687.84 utterances.], batch size: 43, lr: 7.64e-03, grad_scale: 8.0 2023-03-08 11:01:29,820 INFO [train2.py:809] (1/4) Epoch 14, batch 3700, loss[ctc_loss=0.07256, att_loss=0.2129, loss=0.1849, over 13179.00 frames. utt_duration=1819 frames, utt_pad_proportion=0.04155, over 29.00 utterances.], tot_loss[ctc_loss=0.08972, att_loss=0.2436, loss=0.2129, over 3263078.00 frames. utt_duration=1243 frames, utt_pad_proportion=0.05755, over 10515.64 utterances.], batch size: 29, lr: 7.64e-03, grad_scale: 8.0 2023-03-08 11:01:53,187 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55504.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:02:16,950 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55518.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 11:02:33,336 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.607e+02 2.162e+02 2.668e+02 3.280e+02 7.175e+02, threshold=5.335e+02, percent-clipped=8.0 2023-03-08 11:02:53,079 INFO [train2.py:809] (1/4) Epoch 14, batch 3750, loss[ctc_loss=0.09613, att_loss=0.2309, loss=0.2039, over 16000.00 frames. utt_duration=1602 frames, utt_pad_proportion=0.007021, over 40.00 utterances.], tot_loss[ctc_loss=0.08965, att_loss=0.2435, loss=0.2127, over 3271141.50 frames. utt_duration=1268 frames, utt_pad_proportion=0.04971, over 10330.74 utterances.], batch size: 40, lr: 7.63e-03, grad_scale: 8.0 2023-03-08 11:03:25,809 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.93 vs. limit=2.0 2023-03-08 11:03:26,889 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1056, 2.7041, 3.4702, 2.6099, 3.3945, 4.2443, 4.0799, 3.0236], device='cuda:1'), covar=tensor([0.0378, 0.1744, 0.1084, 0.1389, 0.0919, 0.0945, 0.0570, 0.1287], device='cuda:1'), in_proj_covar=tensor([0.0237, 0.0237, 0.0259, 0.0209, 0.0249, 0.0336, 0.0238, 0.0227], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 11:03:36,234 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=55566.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:03:37,989 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55567.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:04:16,306 INFO [train2.py:809] (1/4) Epoch 14, batch 3800, loss[ctc_loss=0.07213, att_loss=0.2139, loss=0.1856, over 15761.00 frames. utt_duration=1660 frames, utt_pad_proportion=0.009327, over 38.00 utterances.], tot_loss[ctc_loss=0.09012, att_loss=0.2436, loss=0.2129, over 3272271.61 frames. utt_duration=1248 frames, utt_pad_proportion=0.05554, over 10502.93 utterances.], batch size: 38, lr: 7.63e-03, grad_scale: 8.0 2023-03-08 11:05:20,072 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.413e+02 2.212e+02 2.620e+02 3.323e+02 8.677e+02, threshold=5.239e+02, percent-clipped=3.0 2023-03-08 11:05:40,105 INFO [train2.py:809] (1/4) Epoch 14, batch 3850, loss[ctc_loss=0.09466, att_loss=0.2546, loss=0.2226, over 17144.00 frames. utt_duration=1226 frames, utt_pad_proportion=0.01375, over 56.00 utterances.], tot_loss[ctc_loss=0.08955, att_loss=0.2428, loss=0.2121, over 3264723.06 frames. utt_duration=1250 frames, utt_pad_proportion=0.05627, over 10460.83 utterances.], batch size: 56, lr: 7.63e-03, grad_scale: 8.0 2023-03-08 11:07:00,901 INFO [train2.py:809] (1/4) Epoch 14, batch 3900, loss[ctc_loss=0.08496, att_loss=0.2529, loss=0.2193, over 17015.00 frames. utt_duration=1336 frames, utt_pad_proportion=0.009135, over 51.00 utterances.], tot_loss[ctc_loss=0.08986, att_loss=0.2432, loss=0.2126, over 3263885.34 frames. utt_duration=1236 frames, utt_pad_proportion=0.0604, over 10574.90 utterances.], batch size: 51, lr: 7.62e-03, grad_scale: 8.0 2023-03-08 11:07:17,830 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.3296, 5.1923, 5.0763, 3.2007, 5.1233, 4.9237, 4.6439, 2.8334], device='cuda:1'), covar=tensor([0.0098, 0.0090, 0.0256, 0.0932, 0.0080, 0.0167, 0.0267, 0.1429], device='cuda:1'), in_proj_covar=tensor([0.0065, 0.0090, 0.0086, 0.0105, 0.0075, 0.0101, 0.0095, 0.0100], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 11:08:01,633 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.280e+02 2.167e+02 2.705e+02 3.104e+02 1.128e+03, threshold=5.409e+02, percent-clipped=4.0 2023-03-08 11:08:20,153 INFO [train2.py:809] (1/4) Epoch 14, batch 3950, loss[ctc_loss=0.0936, att_loss=0.2508, loss=0.2194, over 16948.00 frames. utt_duration=1357 frames, utt_pad_proportion=0.007749, over 50.00 utterances.], tot_loss[ctc_loss=0.09068, att_loss=0.2438, loss=0.2132, over 3267552.94 frames. utt_duration=1238 frames, utt_pad_proportion=0.05941, over 10572.27 utterances.], batch size: 50, lr: 7.62e-03, grad_scale: 8.0 2023-03-08 11:08:32,826 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55748.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:09:39,025 INFO [train2.py:809] (1/4) Epoch 15, batch 0, loss[ctc_loss=0.08716, att_loss=0.235, loss=0.2054, over 16283.00 frames. utt_duration=1516 frames, utt_pad_proportion=0.007128, over 43.00 utterances.], tot_loss[ctc_loss=0.08716, att_loss=0.235, loss=0.2054, over 16283.00 frames. utt_duration=1516 frames, utt_pad_proportion=0.007128, over 43.00 utterances.], batch size: 43, lr: 7.36e-03, grad_scale: 8.0 2023-03-08 11:09:39,026 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-08 11:09:51,763 INFO [train2.py:843] (1/4) Epoch 15, validation: ctc_loss=0.04404, att_loss=0.2365, loss=0.198, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-08 11:09:51,764 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-08 11:10:43,390 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55804.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:10:44,864 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0054, 5.2879, 5.5194, 5.4353, 5.4619, 5.9713, 5.1572, 6.1159], device='cuda:1'), covar=tensor([0.0637, 0.0686, 0.0761, 0.1142, 0.1861, 0.0826, 0.0658, 0.0591], device='cuda:1'), in_proj_covar=tensor([0.0780, 0.0464, 0.0547, 0.0606, 0.0802, 0.0556, 0.0453, 0.0539], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 11:10:46,663 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1134, 4.4812, 4.6135, 4.8327, 2.8312, 4.4702, 2.6652, 1.6526], device='cuda:1'), covar=tensor([0.0317, 0.0203, 0.0614, 0.0114, 0.1621, 0.0181, 0.1509, 0.1748], device='cuda:1'), in_proj_covar=tensor([0.0161, 0.0135, 0.0254, 0.0124, 0.0221, 0.0116, 0.0226, 0.0196], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 11:10:51,508 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55809.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:11:05,752 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.8728, 1.9305, 2.0439, 1.9952, 2.4332, 2.5487, 2.0414, 2.9175], device='cuda:1'), covar=tensor([0.1777, 0.4385, 0.2822, 0.2230, 0.1943, 0.1366, 0.3913, 0.1335], device='cuda:1'), in_proj_covar=tensor([0.0088, 0.0096, 0.0098, 0.0084, 0.0087, 0.0079, 0.0102, 0.0069], device='cuda:1'), out_proj_covar=tensor([6.3082e-05, 7.0094e-05, 7.2881e-05, 6.1847e-05, 6.1910e-05, 6.1199e-05, 7.2331e-05, 5.4228e-05], device='cuda:1') 2023-03-08 11:11:13,890 INFO [train2.py:809] (1/4) Epoch 15, batch 50, loss[ctc_loss=0.09358, att_loss=0.229, loss=0.2019, over 15750.00 frames. utt_duration=1659 frames, utt_pad_proportion=0.009955, over 38.00 utterances.], tot_loss[ctc_loss=0.08842, att_loss=0.2446, loss=0.2134, over 746286.94 frames. utt_duration=1404 frames, utt_pad_proportion=0.01092, over 2128.60 utterances.], batch size: 38, lr: 7.35e-03, grad_scale: 8.0 2023-03-08 11:11:21,608 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.330e+02 2.261e+02 2.719e+02 3.409e+02 7.287e+02, threshold=5.439e+02, percent-clipped=2.0 2023-03-08 11:11:42,966 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=5.09 vs. limit=5.0 2023-03-08 11:11:55,440 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55848.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:12:01,374 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=55852.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:12:25,416 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55867.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:12:29,298 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8348, 5.1489, 4.6318, 5.2063, 4.5466, 4.9180, 5.3033, 5.0487], device='cuda:1'), covar=tensor([0.0589, 0.0258, 0.0829, 0.0258, 0.0436, 0.0213, 0.0193, 0.0178], device='cuda:1'), in_proj_covar=tensor([0.0359, 0.0280, 0.0336, 0.0296, 0.0284, 0.0219, 0.0272, 0.0250], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0005, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 11:12:35,391 INFO [train2.py:809] (1/4) Epoch 15, batch 100, loss[ctc_loss=0.116, att_loss=0.2621, loss=0.2329, over 17040.00 frames. utt_duration=1312 frames, utt_pad_proportion=0.009434, over 52.00 utterances.], tot_loss[ctc_loss=0.0905, att_loss=0.2457, loss=0.2146, over 1311001.16 frames. utt_duration=1332 frames, utt_pad_proportion=0.02662, over 3940.35 utterances.], batch size: 52, lr: 7.35e-03, grad_scale: 8.0 2023-03-08 11:13:35,193 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55909.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:13:44,373 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=55915.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:13:48,629 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 2023-03-08 11:13:57,609 INFO [train2.py:809] (1/4) Epoch 15, batch 150, loss[ctc_loss=0.08308, att_loss=0.2309, loss=0.2013, over 15752.00 frames. utt_duration=1660 frames, utt_pad_proportion=0.009735, over 38.00 utterances.], tot_loss[ctc_loss=0.08858, att_loss=0.2436, loss=0.2126, over 1745554.51 frames. utt_duration=1322 frames, utt_pad_proportion=0.03207, over 5288.83 utterances.], batch size: 38, lr: 7.35e-03, grad_scale: 8.0 2023-03-08 11:14:05,581 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.482e+02 2.292e+02 2.863e+02 3.453e+02 8.582e+02, threshold=5.726e+02, percent-clipped=4.0 2023-03-08 11:15:19,357 INFO [train2.py:809] (1/4) Epoch 15, batch 200, loss[ctc_loss=0.094, att_loss=0.2712, loss=0.2357, over 17355.00 frames. utt_duration=1104 frames, utt_pad_proportion=0.03535, over 63.00 utterances.], tot_loss[ctc_loss=0.08834, att_loss=0.2438, loss=0.2127, over 2081678.01 frames. utt_duration=1259 frames, utt_pad_proportion=0.04875, over 6619.03 utterances.], batch size: 63, lr: 7.34e-03, grad_scale: 8.0 2023-03-08 11:16:45,491 INFO [train2.py:809] (1/4) Epoch 15, batch 250, loss[ctc_loss=0.0718, att_loss=0.2174, loss=0.1883, over 14558.00 frames. utt_duration=1821 frames, utt_pad_proportion=0.03015, over 32.00 utterances.], tot_loss[ctc_loss=0.08889, att_loss=0.244, loss=0.213, over 2349803.14 frames. utt_duration=1242 frames, utt_pad_proportion=0.05094, over 7578.10 utterances.], batch size: 32, lr: 7.34e-03, grad_scale: 8.0 2023-03-08 11:16:53,154 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.371e+02 2.127e+02 2.516e+02 2.865e+02 8.492e+02, threshold=5.032e+02, percent-clipped=2.0 2023-03-08 11:17:37,469 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6916, 4.9315, 4.4331, 5.0380, 4.3367, 4.6999, 5.0787, 4.8757], device='cuda:1'), covar=tensor([0.0566, 0.0334, 0.0924, 0.0308, 0.0491, 0.0317, 0.0250, 0.0194], device='cuda:1'), in_proj_covar=tensor([0.0364, 0.0284, 0.0340, 0.0300, 0.0289, 0.0222, 0.0276, 0.0253], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0005, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 11:18:06,090 INFO [train2.py:809] (1/4) Epoch 15, batch 300, loss[ctc_loss=0.0913, att_loss=0.2284, loss=0.201, over 16171.00 frames. utt_duration=1579 frames, utt_pad_proportion=0.006673, over 41.00 utterances.], tot_loss[ctc_loss=0.0884, att_loss=0.2434, loss=0.2124, over 2555625.80 frames. utt_duration=1277 frames, utt_pad_proportion=0.04337, over 8017.36 utterances.], batch size: 41, lr: 7.34e-03, grad_scale: 8.0 2023-03-08 11:18:55,723 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56104.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:19:25,567 INFO [train2.py:809] (1/4) Epoch 15, batch 350, loss[ctc_loss=0.07455, att_loss=0.2085, loss=0.1817, over 15366.00 frames. utt_duration=1758 frames, utt_pad_proportion=0.01146, over 35.00 utterances.], tot_loss[ctc_loss=0.08908, att_loss=0.2437, loss=0.2127, over 2718430.46 frames. utt_duration=1287 frames, utt_pad_proportion=0.03962, over 8455.75 utterances.], batch size: 35, lr: 7.34e-03, grad_scale: 8.0 2023-03-08 11:19:34,070 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.479e+02 2.148e+02 2.622e+02 3.077e+02 6.565e+02, threshold=5.243e+02, percent-clipped=3.0 2023-03-08 11:20:01,857 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56145.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:20:28,660 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=2.87 vs. limit=5.0 2023-03-08 11:20:45,056 INFO [train2.py:809] (1/4) Epoch 15, batch 400, loss[ctc_loss=0.09313, att_loss=0.2437, loss=0.2136, over 16121.00 frames. utt_duration=1537 frames, utt_pad_proportion=0.005929, over 42.00 utterances.], tot_loss[ctc_loss=0.08975, att_loss=0.2442, loss=0.2133, over 2845889.39 frames. utt_duration=1294 frames, utt_pad_proportion=0.03835, over 8809.37 utterances.], batch size: 42, lr: 7.33e-03, grad_scale: 8.0 2023-03-08 11:21:02,596 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.14 vs. limit=5.0 2023-03-08 11:21:22,525 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2436, 4.6550, 4.4466, 4.8601, 2.9193, 4.7832, 2.6898, 2.2583], device='cuda:1'), covar=tensor([0.0318, 0.0219, 0.0772, 0.0148, 0.1744, 0.0141, 0.1655, 0.1721], device='cuda:1'), in_proj_covar=tensor([0.0161, 0.0134, 0.0254, 0.0124, 0.0220, 0.0114, 0.0226, 0.0195], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 11:21:34,930 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56204.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:21:35,516 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-03-08 11:21:38,234 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56206.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:22:04,494 INFO [train2.py:809] (1/4) Epoch 15, batch 450, loss[ctc_loss=0.1026, att_loss=0.254, loss=0.2237, over 17120.00 frames. utt_duration=1224 frames, utt_pad_proportion=0.01496, over 56.00 utterances.], tot_loss[ctc_loss=0.08866, att_loss=0.2432, loss=0.2123, over 2937857.57 frames. utt_duration=1297 frames, utt_pad_proportion=0.03952, over 9073.91 utterances.], batch size: 56, lr: 7.33e-03, grad_scale: 8.0 2023-03-08 11:22:12,113 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.329e+02 2.121e+02 2.729e+02 3.274e+02 6.212e+02, threshold=5.457e+02, percent-clipped=2.0 2023-03-08 11:23:23,984 INFO [train2.py:809] (1/4) Epoch 15, batch 500, loss[ctc_loss=0.1135, att_loss=0.2493, loss=0.2222, over 15940.00 frames. utt_duration=1557 frames, utt_pad_proportion=0.007326, over 41.00 utterances.], tot_loss[ctc_loss=0.08786, att_loss=0.2423, loss=0.2115, over 3001850.34 frames. utt_duration=1279 frames, utt_pad_proportion=0.04643, over 9400.95 utterances.], batch size: 41, lr: 7.33e-03, grad_scale: 8.0 2023-03-08 11:24:36,862 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-03-08 11:24:43,552 INFO [train2.py:809] (1/4) Epoch 15, batch 550, loss[ctc_loss=0.1148, att_loss=0.2685, loss=0.2377, over 17105.00 frames. utt_duration=1223 frames, utt_pad_proportion=0.01575, over 56.00 utterances.], tot_loss[ctc_loss=0.08759, att_loss=0.2425, loss=0.2116, over 3073920.80 frames. utt_duration=1285 frames, utt_pad_proportion=0.04106, over 9578.85 utterances.], batch size: 56, lr: 7.32e-03, grad_scale: 8.0 2023-03-08 11:24:51,297 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.338e+02 2.199e+02 2.714e+02 3.543e+02 1.087e+03, threshold=5.428e+02, percent-clipped=4.0 2023-03-08 11:26:03,956 INFO [train2.py:809] (1/4) Epoch 15, batch 600, loss[ctc_loss=0.0812, att_loss=0.2314, loss=0.2014, over 16003.00 frames. utt_duration=1602 frames, utt_pad_proportion=0.007698, over 40.00 utterances.], tot_loss[ctc_loss=0.08901, att_loss=0.2428, loss=0.212, over 3112069.14 frames. utt_duration=1269 frames, utt_pad_proportion=0.04797, over 9818.33 utterances.], batch size: 40, lr: 7.32e-03, grad_scale: 8.0 2023-03-08 11:26:54,541 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=56404.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:27:23,614 INFO [train2.py:809] (1/4) Epoch 15, batch 650, loss[ctc_loss=0.08404, att_loss=0.2296, loss=0.2005, over 15774.00 frames. utt_duration=1662 frames, utt_pad_proportion=0.0072, over 38.00 utterances.], tot_loss[ctc_loss=0.08954, att_loss=0.2429, loss=0.2122, over 3149929.64 frames. utt_duration=1260 frames, utt_pad_proportion=0.0503, over 10013.74 utterances.], batch size: 38, lr: 7.32e-03, grad_scale: 8.0 2023-03-08 11:27:31,803 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.524e+02 2.098e+02 2.434e+02 3.232e+02 5.405e+02, threshold=4.867e+02, percent-clipped=0.0 2023-03-08 11:28:09,880 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=56452.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:28:42,742 INFO [train2.py:809] (1/4) Epoch 15, batch 700, loss[ctc_loss=0.08835, att_loss=0.2542, loss=0.221, over 17032.00 frames. utt_duration=689.7 frames, utt_pad_proportion=0.1335, over 99.00 utterances.], tot_loss[ctc_loss=0.0906, att_loss=0.244, loss=0.2133, over 3174732.55 frames. utt_duration=1233 frames, utt_pad_proportion=0.05644, over 10313.98 utterances.], batch size: 99, lr: 7.31e-03, grad_scale: 8.0 2023-03-08 11:29:28,158 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56501.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:29:33,103 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=56504.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:30:02,717 INFO [train2.py:809] (1/4) Epoch 15, batch 750, loss[ctc_loss=0.09869, att_loss=0.2568, loss=0.2252, over 16630.00 frames. utt_duration=1417 frames, utt_pad_proportion=0.00508, over 47.00 utterances.], tot_loss[ctc_loss=0.09063, att_loss=0.2434, loss=0.2128, over 3184828.80 frames. utt_duration=1238 frames, utt_pad_proportion=0.05945, over 10304.74 utterances.], batch size: 47, lr: 7.31e-03, grad_scale: 8.0 2023-03-08 11:30:03,076 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56523.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 11:30:07,833 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7139, 4.7461, 4.6072, 4.5743, 5.2057, 4.7515, 4.6101, 2.3175], device='cuda:1'), covar=tensor([0.0209, 0.0243, 0.0272, 0.0285, 0.0857, 0.0204, 0.0284, 0.1990], device='cuda:1'), in_proj_covar=tensor([0.0133, 0.0145, 0.0152, 0.0164, 0.0347, 0.0132, 0.0138, 0.0212], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0002, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 11:30:11,024 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.244e+02 2.381e+02 2.975e+02 3.740e+02 8.505e+02, threshold=5.949e+02, percent-clipped=7.0 2023-03-08 11:30:49,606 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=56552.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:31:03,228 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-03-08 11:31:22,462 INFO [train2.py:809] (1/4) Epoch 15, batch 800, loss[ctc_loss=0.07995, att_loss=0.2213, loss=0.193, over 15749.00 frames. utt_duration=1659 frames, utt_pad_proportion=0.009285, over 38.00 utterances.], tot_loss[ctc_loss=0.09029, att_loss=0.2436, loss=0.2129, over 3213053.29 frames. utt_duration=1260 frames, utt_pad_proportion=0.0512, over 10209.53 utterances.], batch size: 38, lr: 7.31e-03, grad_scale: 8.0 2023-03-08 11:31:41,523 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56584.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 11:32:44,277 INFO [train2.py:809] (1/4) Epoch 15, batch 850, loss[ctc_loss=0.06629, att_loss=0.2165, loss=0.1865, over 14506.00 frames. utt_duration=1815 frames, utt_pad_proportion=0.04688, over 32.00 utterances.], tot_loss[ctc_loss=0.09011, att_loss=0.2438, loss=0.2131, over 3229338.33 frames. utt_duration=1248 frames, utt_pad_proportion=0.0532, over 10362.33 utterances.], batch size: 32, lr: 7.30e-03, grad_scale: 8.0 2023-03-08 11:32:52,933 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.315e+02 2.144e+02 2.509e+02 3.065e+02 7.548e+02, threshold=5.017e+02, percent-clipped=1.0 2023-03-08 11:32:57,999 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6418, 4.6527, 4.5358, 4.5437, 4.9871, 4.8050, 4.5508, 2.2358], device='cuda:1'), covar=tensor([0.0195, 0.0198, 0.0262, 0.0245, 0.0993, 0.0178, 0.0242, 0.2056], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0147, 0.0153, 0.0165, 0.0352, 0.0134, 0.0139, 0.0214], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 11:34:04,336 INFO [train2.py:809] (1/4) Epoch 15, batch 900, loss[ctc_loss=0.08485, att_loss=0.2547, loss=0.2207, over 16477.00 frames. utt_duration=1435 frames, utt_pad_proportion=0.006549, over 46.00 utterances.], tot_loss[ctc_loss=0.08976, att_loss=0.2435, loss=0.2128, over 3231310.56 frames. utt_duration=1240 frames, utt_pad_proportion=0.05711, over 10435.00 utterances.], batch size: 46, lr: 7.30e-03, grad_scale: 8.0 2023-03-08 11:35:24,403 INFO [train2.py:809] (1/4) Epoch 15, batch 950, loss[ctc_loss=0.0722, att_loss=0.2309, loss=0.1991, over 15963.00 frames. utt_duration=1558 frames, utt_pad_proportion=0.006066, over 41.00 utterances.], tot_loss[ctc_loss=0.0899, att_loss=0.2434, loss=0.2127, over 3238016.53 frames. utt_duration=1238 frames, utt_pad_proportion=0.05792, over 10472.88 utterances.], batch size: 41, lr: 7.30e-03, grad_scale: 8.0 2023-03-08 11:35:32,189 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.300e+02 2.400e+02 2.758e+02 3.423e+02 1.509e+03, threshold=5.516e+02, percent-clipped=3.0 2023-03-08 11:36:11,930 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-03-08 11:36:19,691 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.3354, 5.1224, 4.9181, 3.1138, 4.9184, 4.8070, 4.6130, 3.1159], device='cuda:1'), covar=tensor([0.0088, 0.0091, 0.0337, 0.0989, 0.0099, 0.0155, 0.0252, 0.1108], device='cuda:1'), in_proj_covar=tensor([0.0067, 0.0091, 0.0088, 0.0108, 0.0076, 0.0102, 0.0097, 0.0101], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 11:36:44,600 INFO [train2.py:809] (1/4) Epoch 15, batch 1000, loss[ctc_loss=0.07958, att_loss=0.2446, loss=0.2116, over 16550.00 frames. utt_duration=1472 frames, utt_pad_proportion=0.00509, over 45.00 utterances.], tot_loss[ctc_loss=0.09004, att_loss=0.2438, loss=0.213, over 3246987.61 frames. utt_duration=1216 frames, utt_pad_proportion=0.06257, over 10689.53 utterances.], batch size: 45, lr: 7.29e-03, grad_scale: 8.0 2023-03-08 11:37:30,243 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=56801.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:37:33,366 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56803.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:37:49,430 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.4306, 2.5429, 2.3734, 2.3110, 2.4781, 2.4496, 2.5301, 1.9452], device='cuda:1'), covar=tensor([0.1172, 0.2082, 0.2901, 0.4859, 0.1311, 0.3594, 0.1619, 0.5960], device='cuda:1'), in_proj_covar=tensor([0.0122, 0.0144, 0.0155, 0.0222, 0.0116, 0.0207, 0.0131, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 11:38:05,062 INFO [train2.py:809] (1/4) Epoch 15, batch 1050, loss[ctc_loss=0.09948, att_loss=0.2335, loss=0.2067, over 14497.00 frames. utt_duration=1813 frames, utt_pad_proportion=0.04756, over 32.00 utterances.], tot_loss[ctc_loss=0.08927, att_loss=0.244, loss=0.2131, over 3257000.44 frames. utt_duration=1234 frames, utt_pad_proportion=0.05774, over 10571.77 utterances.], batch size: 32, lr: 7.29e-03, grad_scale: 8.0 2023-03-08 11:38:12,498 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.420e+02 2.111e+02 2.438e+02 3.282e+02 6.926e+02, threshold=4.877e+02, percent-clipped=2.0 2023-03-08 11:38:19,692 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.6041, 3.0718, 3.4431, 4.5636, 4.1547, 4.0768, 2.9414, 2.0753], device='cuda:1'), covar=tensor([0.0588, 0.1928, 0.1070, 0.0461, 0.0608, 0.0435, 0.1489, 0.2430], device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0214, 0.0192, 0.0202, 0.0208, 0.0164, 0.0200, 0.0182], 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') 2023-03-08 11:38:46,233 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=56849.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:39:10,473 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56864.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:39:24,727 INFO [train2.py:809] (1/4) Epoch 15, batch 1100, loss[ctc_loss=0.08081, att_loss=0.2468, loss=0.2136, over 16964.00 frames. utt_duration=1358 frames, utt_pad_proportion=0.007699, over 50.00 utterances.], tot_loss[ctc_loss=0.08903, att_loss=0.2445, loss=0.2134, over 3271961.70 frames. utt_duration=1223 frames, utt_pad_proportion=0.05763, over 10712.56 utterances.], batch size: 50, lr: 7.29e-03, grad_scale: 8.0 2023-03-08 11:39:34,115 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56879.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 11:40:18,811 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9769, 5.2542, 5.5726, 5.3130, 5.3930, 5.9702, 5.2566, 6.0476], device='cuda:1'), covar=tensor([0.0654, 0.0719, 0.0724, 0.1136, 0.1504, 0.0759, 0.0554, 0.0551], device='cuda:1'), in_proj_covar=tensor([0.0789, 0.0463, 0.0544, 0.0604, 0.0791, 0.0551, 0.0449, 0.0538], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 11:40:19,595 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-03-08 11:40:20,535 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56908.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:40:29,816 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-03-08 11:40:44,825 INFO [train2.py:809] (1/4) Epoch 15, batch 1150, loss[ctc_loss=0.06855, att_loss=0.2168, loss=0.1871, over 16017.00 frames. utt_duration=1603 frames, utt_pad_proportion=0.007276, over 40.00 utterances.], tot_loss[ctc_loss=0.08922, att_loss=0.2445, loss=0.2134, over 3274957.25 frames. utt_duration=1227 frames, utt_pad_proportion=0.05699, over 10691.30 utterances.], batch size: 40, lr: 7.28e-03, grad_scale: 8.0 2023-03-08 11:40:52,692 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.206e+02 2.173e+02 2.768e+02 3.173e+02 4.889e+02, threshold=5.536e+02, percent-clipped=1.0 2023-03-08 11:41:58,525 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56969.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:42:04,436 INFO [train2.py:809] (1/4) Epoch 15, batch 1200, loss[ctc_loss=0.07739, att_loss=0.2322, loss=0.2012, over 15992.00 frames. utt_duration=1601 frames, utt_pad_proportion=0.008178, over 40.00 utterances.], tot_loss[ctc_loss=0.08945, att_loss=0.2446, loss=0.2136, over 3273120.83 frames. utt_duration=1230 frames, utt_pad_proportion=0.05726, over 10654.19 utterances.], batch size: 40, lr: 7.28e-03, grad_scale: 8.0 2023-03-08 11:42:36,802 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56993.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:42:39,857 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56995.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:43:23,350 INFO [train2.py:809] (1/4) Epoch 15, batch 1250, loss[ctc_loss=0.07584, att_loss=0.2331, loss=0.2017, over 16386.00 frames. utt_duration=1491 frames, utt_pad_proportion=0.008583, over 44.00 utterances.], tot_loss[ctc_loss=0.08917, att_loss=0.2442, loss=0.2132, over 3282836.52 frames. utt_duration=1250 frames, utt_pad_proportion=0.05022, over 10518.01 utterances.], batch size: 44, lr: 7.28e-03, grad_scale: 8.0 2023-03-08 11:43:31,117 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.571e+02 2.221e+02 2.683e+02 3.397e+02 1.061e+03, threshold=5.366e+02, percent-clipped=4.0 2023-03-08 11:44:13,695 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57054.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:44:17,055 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57056.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:44:43,934 INFO [train2.py:809] (1/4) Epoch 15, batch 1300, loss[ctc_loss=0.05698, att_loss=0.2218, loss=0.1888, over 15898.00 frames. utt_duration=1632 frames, utt_pad_proportion=0.00849, over 39.00 utterances.], tot_loss[ctc_loss=0.08843, att_loss=0.2436, loss=0.2126, over 3282197.88 frames. utt_duration=1265 frames, utt_pad_proportion=0.04769, over 10392.35 utterances.], batch size: 39, lr: 7.27e-03, grad_scale: 8.0 2023-03-08 11:46:03,451 INFO [train2.py:809] (1/4) Epoch 15, batch 1350, loss[ctc_loss=0.113, att_loss=0.2591, loss=0.2299, over 16331.00 frames. utt_duration=1453 frames, utt_pad_proportion=0.005943, over 45.00 utterances.], tot_loss[ctc_loss=0.08859, att_loss=0.2435, loss=0.2125, over 3273817.54 frames. utt_duration=1243 frames, utt_pad_proportion=0.05578, over 10550.46 utterances.], batch size: 45, lr: 7.27e-03, grad_scale: 8.0 2023-03-08 11:46:11,170 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.494e+02 2.073e+02 2.593e+02 3.161e+02 1.501e+03, threshold=5.185e+02, percent-clipped=3.0 2023-03-08 11:46:31,597 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-03-08 11:47:00,496 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57159.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:47:22,527 INFO [train2.py:809] (1/4) Epoch 15, batch 1400, loss[ctc_loss=0.08477, att_loss=0.2477, loss=0.2151, over 17100.00 frames. utt_duration=1223 frames, utt_pad_proportion=0.01534, over 56.00 utterances.], tot_loss[ctc_loss=0.08992, att_loss=0.2441, loss=0.2132, over 3268436.80 frames. utt_duration=1201 frames, utt_pad_proportion=0.06717, over 10901.90 utterances.], batch size: 56, lr: 7.27e-03, grad_scale: 8.0 2023-03-08 11:47:32,871 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57179.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 11:47:44,288 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57186.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:48:00,877 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57197.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:48:13,782 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.8103, 1.8407, 2.0861, 1.9854, 2.6262, 2.2948, 2.0678, 2.8306], device='cuda:1'), covar=tensor([0.1131, 0.3669, 0.2385, 0.1934, 0.1365, 0.1232, 0.3039, 0.0788], device='cuda:1'), in_proj_covar=tensor([0.0091, 0.0098, 0.0099, 0.0088, 0.0090, 0.0080, 0.0105, 0.0071], device='cuda:1'), out_proj_covar=tensor([6.5342e-05, 7.2006e-05, 7.4430e-05, 6.4856e-05, 6.4321e-05, 6.2425e-05, 7.4539e-05, 5.6246e-05], device='cuda:1') 2023-03-08 11:48:42,466 INFO [train2.py:809] (1/4) Epoch 15, batch 1450, loss[ctc_loss=0.07396, att_loss=0.2451, loss=0.2108, over 16414.00 frames. utt_duration=1494 frames, utt_pad_proportion=0.006755, over 44.00 utterances.], tot_loss[ctc_loss=0.08932, att_loss=0.2439, loss=0.213, over 3276735.49 frames. utt_duration=1232 frames, utt_pad_proportion=0.05751, over 10652.69 utterances.], batch size: 44, lr: 7.26e-03, grad_scale: 8.0 2023-03-08 11:48:48,511 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=57227.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 11:48:49,835 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.461e+02 2.071e+02 2.451e+02 3.320e+02 6.721e+02, threshold=4.903e+02, percent-clipped=4.0 2023-03-08 11:49:15,906 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57244.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:49:20,466 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57247.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:49:38,312 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57258.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 11:49:48,019 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57264.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:49:49,712 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2510, 4.7225, 4.6372, 4.7265, 4.7957, 4.3743, 3.3532, 4.6362], device='cuda:1'), covar=tensor([0.0116, 0.0107, 0.0122, 0.0076, 0.0090, 0.0113, 0.0622, 0.0178], device='cuda:1'), in_proj_covar=tensor([0.0080, 0.0078, 0.0096, 0.0059, 0.0065, 0.0077, 0.0094, 0.0097], 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') 2023-03-08 11:49:56,090 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7926, 1.8058, 2.1947, 2.0357, 2.6076, 2.4686, 2.1253, 3.0403], device='cuda:1'), covar=tensor([0.1760, 0.5225, 0.3229, 0.2760, 0.1813, 0.1692, 0.3870, 0.0916], device='cuda:1'), in_proj_covar=tensor([0.0092, 0.0099, 0.0101, 0.0089, 0.0091, 0.0081, 0.0105, 0.0072], device='cuda:1'), out_proj_covar=tensor([6.6044e-05, 7.2859e-05, 7.5594e-05, 6.5551e-05, 6.5016e-05, 6.3309e-05, 7.5170e-05, 5.6698e-05], device='cuda:1') 2023-03-08 11:50:02,011 INFO [train2.py:809] (1/4) Epoch 15, batch 1500, loss[ctc_loss=0.07391, att_loss=0.2346, loss=0.2025, over 16764.00 frames. utt_duration=1398 frames, utt_pad_proportion=0.006866, over 48.00 utterances.], tot_loss[ctc_loss=0.08929, att_loss=0.2437, loss=0.2128, over 3283756.43 frames. utt_duration=1254 frames, utt_pad_proportion=0.05047, over 10486.58 utterances.], batch size: 48, lr: 7.26e-03, grad_scale: 16.0 2023-03-08 11:50:53,395 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57305.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:51:08,937 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.3279, 2.4906, 2.9667, 4.4306, 4.0242, 3.9813, 3.0614, 2.0823], device='cuda:1'), covar=tensor([0.0763, 0.2608, 0.1450, 0.0578, 0.0723, 0.0423, 0.1456, 0.2597], device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0212, 0.0189, 0.0201, 0.0206, 0.0164, 0.0199, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 11:51:20,966 INFO [train2.py:809] (1/4) Epoch 15, batch 1550, loss[ctc_loss=0.1073, att_loss=0.2408, loss=0.2141, over 16265.00 frames. utt_duration=1515 frames, utt_pad_proportion=0.007254, over 43.00 utterances.], tot_loss[ctc_loss=0.08965, att_loss=0.2441, loss=0.2132, over 3289574.48 frames. utt_duration=1252 frames, utt_pad_proportion=0.04915, over 10518.50 utterances.], batch size: 43, lr: 7.26e-03, grad_scale: 16.0 2023-03-08 11:51:29,315 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.286e+02 2.083e+02 2.433e+02 3.048e+02 8.459e+02, threshold=4.866e+02, percent-clipped=4.0 2023-03-08 11:52:02,020 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57349.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:52:05,132 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57351.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:52:09,023 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-03-08 11:52:39,938 INFO [train2.py:809] (1/4) Epoch 15, batch 1600, loss[ctc_loss=0.07204, att_loss=0.2159, loss=0.1871, over 15389.00 frames. utt_duration=1760 frames, utt_pad_proportion=0.009551, over 35.00 utterances.], tot_loss[ctc_loss=0.09007, att_loss=0.2443, loss=0.2135, over 3283307.81 frames. utt_duration=1254 frames, utt_pad_proportion=0.0508, over 10487.70 utterances.], batch size: 35, lr: 7.26e-03, grad_scale: 16.0 2023-03-08 11:53:20,349 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57398.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:54:00,797 INFO [train2.py:809] (1/4) Epoch 15, batch 1650, loss[ctc_loss=0.08287, att_loss=0.2463, loss=0.2136, over 16959.00 frames. utt_duration=1358 frames, utt_pad_proportion=0.007816, over 50.00 utterances.], tot_loss[ctc_loss=0.08956, att_loss=0.2454, loss=0.2143, over 3294060.85 frames. utt_duration=1245 frames, utt_pad_proportion=0.0498, over 10596.96 utterances.], batch size: 50, lr: 7.25e-03, grad_scale: 16.0 2023-03-08 11:54:09,582 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.363e+02 2.043e+02 2.377e+02 2.961e+02 5.254e+02, threshold=4.753e+02, percent-clipped=2.0 2023-03-08 11:54:59,144 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57459.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:54:59,235 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57459.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:55:02,013 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9703, 6.1854, 5.5319, 5.9171, 5.8030, 5.2931, 5.6082, 5.2671], device='cuda:1'), covar=tensor([0.1308, 0.0839, 0.0910, 0.0813, 0.0775, 0.1566, 0.2247, 0.2478], device='cuda:1'), in_proj_covar=tensor([0.0467, 0.0540, 0.0410, 0.0405, 0.0389, 0.0430, 0.0555, 0.0495], 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') 2023-03-08 11:55:03,831 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57462.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:55:20,182 INFO [train2.py:809] (1/4) Epoch 15, batch 1700, loss[ctc_loss=0.09966, att_loss=0.2516, loss=0.2212, over 17054.00 frames. utt_duration=1289 frames, utt_pad_proportion=0.009557, over 53.00 utterances.], tot_loss[ctc_loss=0.08888, att_loss=0.244, loss=0.213, over 3283778.32 frames. utt_duration=1239 frames, utt_pad_proportion=0.0533, over 10611.28 utterances.], batch size: 53, lr: 7.25e-03, grad_scale: 16.0 2023-03-08 11:55:33,403 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9667, 5.2417, 5.4828, 5.3793, 5.4234, 5.8855, 5.2171, 6.0619], device='cuda:1'), covar=tensor([0.0773, 0.0699, 0.0803, 0.1229, 0.1887, 0.1040, 0.0662, 0.0663], device='cuda:1'), in_proj_covar=tensor([0.0793, 0.0464, 0.0546, 0.0607, 0.0796, 0.0560, 0.0449, 0.0535], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 11:55:54,927 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57495.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:56:14,336 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=57507.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:56:38,720 INFO [train2.py:809] (1/4) Epoch 15, batch 1750, loss[ctc_loss=0.0696, att_loss=0.2281, loss=0.1964, over 16180.00 frames. utt_duration=1580 frames, utt_pad_proportion=0.006807, over 41.00 utterances.], tot_loss[ctc_loss=0.08864, att_loss=0.2434, loss=0.2124, over 3281072.54 frames. utt_duration=1232 frames, utt_pad_proportion=0.05665, over 10666.31 utterances.], batch size: 41, lr: 7.25e-03, grad_scale: 16.0 2023-03-08 11:56:39,752 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57523.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:56:46,885 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.541e+02 2.236e+02 2.703e+02 3.413e+02 7.176e+02, threshold=5.406e+02, percent-clipped=5.0 2023-03-08 11:57:00,229 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-03-08 11:57:08,508 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57542.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:57:16,430 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0715, 4.7761, 4.8180, 2.3276, 2.0833, 2.7459, 2.6024, 3.7135], device='cuda:1'), covar=tensor([0.0645, 0.0171, 0.0173, 0.3674, 0.5362, 0.2590, 0.2655, 0.1607], device='cuda:1'), in_proj_covar=tensor([0.0341, 0.0237, 0.0243, 0.0223, 0.0342, 0.0332, 0.0231, 0.0352], device='cuda:1'), out_proj_covar=tensor([1.4873e-04, 8.9151e-05, 1.0533e-04, 9.6987e-05, 1.4533e-04, 1.3188e-04, 9.2563e-05, 1.4596e-04], device='cuda:1') 2023-03-08 11:57:26,195 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57553.0, num_to_drop=1, layers_to_drop={3} 2023-03-08 11:57:31,878 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57556.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:57:44,175 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57564.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:57:47,241 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.4384, 2.7810, 3.4319, 4.4614, 3.9532, 3.9052, 2.8963, 2.1588], device='cuda:1'), covar=tensor([0.0682, 0.2027, 0.0908, 0.0559, 0.0774, 0.0412, 0.1587, 0.2311], device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0209, 0.0188, 0.0198, 0.0205, 0.0163, 0.0197, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 11:57:58,284 INFO [train2.py:809] (1/4) Epoch 15, batch 1800, loss[ctc_loss=0.08568, att_loss=0.244, loss=0.2123, over 16769.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.006422, over 48.00 utterances.], tot_loss[ctc_loss=0.08916, att_loss=0.2437, loss=0.2128, over 3274010.94 frames. utt_duration=1189 frames, utt_pad_proportion=0.06893, over 11029.36 utterances.], batch size: 48, lr: 7.24e-03, grad_scale: 16.0 2023-03-08 11:58:29,070 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-03-08 11:58:41,635 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57600.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:58:56,298 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7704, 4.9484, 4.7293, 4.7252, 5.3964, 5.1005, 4.8795, 2.7031], device='cuda:1'), covar=tensor([0.0180, 0.0212, 0.0269, 0.0311, 0.0894, 0.0145, 0.0267, 0.1692], device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0148, 0.0153, 0.0164, 0.0350, 0.0134, 0.0140, 0.0212], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 11:59:00,596 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=57612.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 11:59:19,512 INFO [train2.py:809] (1/4) Epoch 15, batch 1850, loss[ctc_loss=0.07517, att_loss=0.2511, loss=0.2159, over 16472.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.006835, over 46.00 utterances.], tot_loss[ctc_loss=0.08878, att_loss=0.2438, loss=0.2128, over 3280530.14 frames. utt_duration=1200 frames, utt_pad_proportion=0.06405, over 10945.14 utterances.], batch size: 46, lr: 7.24e-03, grad_scale: 16.0 2023-03-08 11:59:27,209 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.618e+02 2.149e+02 2.578e+02 3.271e+02 8.979e+02, threshold=5.155e+02, percent-clipped=4.0 2023-03-08 11:59:33,905 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4910, 4.9628, 4.8556, 5.1180, 5.0766, 4.7309, 3.4520, 4.9781], device='cuda:1'), covar=tensor([0.0145, 0.0164, 0.0155, 0.0085, 0.0112, 0.0158, 0.0807, 0.0292], device='cuda:1'), in_proj_covar=tensor([0.0081, 0.0079, 0.0098, 0.0060, 0.0066, 0.0078, 0.0096, 0.0099], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:1') 2023-03-08 12:00:00,126 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57649.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:00:03,803 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57651.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:00:38,110 INFO [train2.py:809] (1/4) Epoch 15, batch 1900, loss[ctc_loss=0.06474, att_loss=0.217, loss=0.1865, over 15364.00 frames. utt_duration=1757 frames, utt_pad_proportion=0.009948, over 35.00 utterances.], tot_loss[ctc_loss=0.08873, att_loss=0.2443, loss=0.2132, over 3282500.72 frames. utt_duration=1201 frames, utt_pad_proportion=0.06439, over 10948.17 utterances.], batch size: 35, lr: 7.24e-03, grad_scale: 16.0 2023-03-08 12:01:08,090 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9890, 5.2145, 4.7815, 5.3144, 4.6991, 4.9368, 5.4138, 5.1839], device='cuda:1'), covar=tensor([0.0487, 0.0268, 0.0818, 0.0229, 0.0383, 0.0256, 0.0208, 0.0180], device='cuda:1'), in_proj_covar=tensor([0.0365, 0.0285, 0.0343, 0.0301, 0.0295, 0.0225, 0.0278, 0.0256], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0005, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 12:01:12,788 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6857, 3.1039, 3.7423, 3.2350, 3.6703, 4.7862, 4.4747, 3.5909], device='cuda:1'), covar=tensor([0.0308, 0.1638, 0.1118, 0.1291, 0.1045, 0.0700, 0.0547, 0.1135], device='cuda:1'), in_proj_covar=tensor([0.0237, 0.0236, 0.0264, 0.0209, 0.0255, 0.0335, 0.0239, 0.0228], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 12:01:15,661 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=57697.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:01:18,921 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=57699.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:01:19,261 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7352, 4.8205, 4.7044, 4.6051, 5.1871, 4.9701, 4.6681, 2.4390], device='cuda:1'), covar=tensor([0.0168, 0.0204, 0.0252, 0.0276, 0.0825, 0.0143, 0.0303, 0.1894], device='cuda:1'), in_proj_covar=tensor([0.0135, 0.0148, 0.0155, 0.0166, 0.0353, 0.0135, 0.0141, 0.0214], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 12:01:56,432 INFO [train2.py:809] (1/4) Epoch 15, batch 1950, loss[ctc_loss=0.07734, att_loss=0.234, loss=0.2027, over 16190.00 frames. utt_duration=1581 frames, utt_pad_proportion=0.005614, over 41.00 utterances.], tot_loss[ctc_loss=0.08867, att_loss=0.2443, loss=0.2132, over 3284770.29 frames. utt_duration=1227 frames, utt_pad_proportion=0.05801, over 10721.78 utterances.], batch size: 41, lr: 7.23e-03, grad_scale: 16.0 2023-03-08 12:02:05,440 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.339e+02 2.187e+02 2.629e+02 3.507e+02 9.887e+02, threshold=5.258e+02, percent-clipped=4.0 2023-03-08 12:02:13,627 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57733.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:02:47,464 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57754.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:03:17,167 INFO [train2.py:809] (1/4) Epoch 15, batch 2000, loss[ctc_loss=0.07804, att_loss=0.2425, loss=0.2096, over 15961.00 frames. utt_duration=1559 frames, utt_pad_proportion=0.006684, over 41.00 utterances.], tot_loss[ctc_loss=0.08737, att_loss=0.2431, loss=0.212, over 3276692.05 frames. utt_duration=1232 frames, utt_pad_proportion=0.05895, over 10651.20 utterances.], batch size: 41, lr: 7.23e-03, grad_scale: 16.0 2023-03-08 12:03:31,266 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.03 vs. limit=5.0 2023-03-08 12:03:43,293 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-03-08 12:03:50,214 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57794.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:04:28,950 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57818.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:04:30,794 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4777, 2.5325, 3.5783, 2.5941, 3.4307, 4.6514, 4.5196, 3.0036], device='cuda:1'), covar=tensor([0.0442, 0.2171, 0.1006, 0.1787, 0.0992, 0.0571, 0.0490, 0.1640], device='cuda:1'), in_proj_covar=tensor([0.0237, 0.0237, 0.0265, 0.0209, 0.0255, 0.0333, 0.0237, 0.0227], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 12:04:37,302 INFO [train2.py:809] (1/4) Epoch 15, batch 2050, loss[ctc_loss=0.08568, att_loss=0.2457, loss=0.2137, over 17340.00 frames. utt_duration=1103 frames, utt_pad_proportion=0.03538, over 63.00 utterances.], tot_loss[ctc_loss=0.08731, att_loss=0.243, loss=0.2119, over 3268435.39 frames. utt_duration=1222 frames, utt_pad_proportion=0.06223, over 10715.09 utterances.], batch size: 63, lr: 7.23e-03, grad_scale: 16.0 2023-03-08 12:04:45,711 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.367e+02 2.062e+02 2.536e+02 3.073e+02 8.062e+02, threshold=5.072e+02, percent-clipped=2.0 2023-03-08 12:05:07,114 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57842.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:05:17,629 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.99 vs. limit=2.0 2023-03-08 12:05:22,336 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57851.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:05:25,429 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57853.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 12:05:56,964 INFO [train2.py:809] (1/4) Epoch 15, batch 2100, loss[ctc_loss=0.05822, att_loss=0.2245, loss=0.1912, over 16404.00 frames. utt_duration=1493 frames, utt_pad_proportion=0.007404, over 44.00 utterances.], tot_loss[ctc_loss=0.08781, att_loss=0.2429, loss=0.2119, over 3255541.83 frames. utt_duration=1225 frames, utt_pad_proportion=0.06421, over 10647.19 utterances.], batch size: 44, lr: 7.22e-03, grad_scale: 16.0 2023-03-08 12:06:11,956 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-03-08 12:06:19,147 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.3428, 4.0696, 3.5179, 3.7295, 4.3230, 3.9240, 3.4085, 4.5824], device='cuda:1'), covar=tensor([0.0925, 0.0470, 0.0869, 0.0651, 0.0553, 0.0662, 0.0690, 0.0375], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0200, 0.0215, 0.0188, 0.0257, 0.0228, 0.0190, 0.0276], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-08 12:06:23,269 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=57890.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:06:28,770 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-03-08 12:06:40,419 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57900.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:06:41,918 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=57901.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:06:55,195 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-03-08 12:07:17,265 INFO [train2.py:809] (1/4) Epoch 15, batch 2150, loss[ctc_loss=0.08941, att_loss=0.2356, loss=0.2064, over 14110.00 frames. utt_duration=1822 frames, utt_pad_proportion=0.04645, over 31.00 utterances.], tot_loss[ctc_loss=0.08783, att_loss=0.2427, loss=0.2118, over 3251649.66 frames. utt_duration=1221 frames, utt_pad_proportion=0.06627, over 10665.44 utterances.], batch size: 31, lr: 7.22e-03, grad_scale: 16.0 2023-03-08 12:07:25,214 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.416e+02 2.164e+02 2.681e+02 3.169e+02 6.485e+02, threshold=5.361e+02, percent-clipped=3.0 2023-03-08 12:07:57,492 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=57948.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:08:37,642 INFO [train2.py:809] (1/4) Epoch 15, batch 2200, loss[ctc_loss=0.09039, att_loss=0.2553, loss=0.2223, over 16483.00 frames. utt_duration=1435 frames, utt_pad_proportion=0.005649, over 46.00 utterances.], tot_loss[ctc_loss=0.08752, att_loss=0.2421, loss=0.2112, over 3251173.34 frames. utt_duration=1220 frames, utt_pad_proportion=0.0672, over 10669.22 utterances.], batch size: 46, lr: 7.22e-03, grad_scale: 16.0 2023-03-08 12:09:53,100 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9460, 5.2310, 5.1212, 5.0939, 5.2833, 5.2508, 4.9203, 4.6717], device='cuda:1'), covar=tensor([0.1022, 0.0495, 0.0305, 0.0500, 0.0260, 0.0298, 0.0342, 0.0368], device='cuda:1'), in_proj_covar=tensor([0.0495, 0.0330, 0.0297, 0.0329, 0.0385, 0.0405, 0.0326, 0.0365], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0004, 0.0004, 0.0005, 0.0003, 0.0004], device='cuda:1') 2023-03-08 12:10:01,092 INFO [train2.py:809] (1/4) Epoch 15, batch 2250, loss[ctc_loss=0.1276, att_loss=0.2627, loss=0.2357, over 16605.00 frames. utt_duration=1415 frames, utt_pad_proportion=0.005786, over 47.00 utterances.], tot_loss[ctc_loss=0.08853, att_loss=0.2432, loss=0.2122, over 3263821.96 frames. utt_duration=1225 frames, utt_pad_proportion=0.06233, over 10667.88 utterances.], batch size: 47, lr: 7.22e-03, grad_scale: 16.0 2023-03-08 12:10:08,785 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.417e+02 2.180e+02 2.771e+02 3.383e+02 6.703e+02, threshold=5.542e+02, percent-clipped=1.0 2023-03-08 12:10:25,235 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-03-08 12:10:49,923 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58054.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:11:07,934 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3510, 4.3853, 4.3888, 4.4110, 4.9975, 4.7061, 4.3460, 2.1702], device='cuda:1'), covar=tensor([0.0246, 0.0312, 0.0296, 0.0286, 0.0930, 0.0183, 0.0332, 0.2138], device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0148, 0.0153, 0.0165, 0.0346, 0.0133, 0.0141, 0.0212], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 12:11:20,346 INFO [train2.py:809] (1/4) Epoch 15, batch 2300, loss[ctc_loss=0.07431, att_loss=0.2289, loss=0.198, over 16161.00 frames. utt_duration=1578 frames, utt_pad_proportion=0.007624, over 41.00 utterances.], tot_loss[ctc_loss=0.08774, att_loss=0.2426, loss=0.2116, over 3264518.62 frames. utt_duration=1232 frames, utt_pad_proportion=0.06127, over 10608.40 utterances.], batch size: 41, lr: 7.21e-03, grad_scale: 8.0 2023-03-08 12:11:42,775 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9113, 5.1546, 5.4540, 5.2846, 5.3249, 5.8240, 5.1688, 5.9665], device='cuda:1'), covar=tensor([0.0711, 0.0792, 0.0800, 0.1259, 0.1983, 0.0978, 0.0726, 0.0671], device='cuda:1'), in_proj_covar=tensor([0.0790, 0.0468, 0.0543, 0.0608, 0.0800, 0.0562, 0.0448, 0.0529], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 12:11:45,858 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=58089.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:12:06,435 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=58102.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:12:33,108 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58118.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:12:40,622 INFO [train2.py:809] (1/4) Epoch 15, batch 2350, loss[ctc_loss=0.09338, att_loss=0.2329, loss=0.205, over 15769.00 frames. utt_duration=1662 frames, utt_pad_proportion=0.008604, over 38.00 utterances.], tot_loss[ctc_loss=0.08757, att_loss=0.2434, loss=0.2122, over 3267243.23 frames. utt_duration=1218 frames, utt_pad_proportion=0.06374, over 10743.04 utterances.], batch size: 38, lr: 7.21e-03, grad_scale: 8.0 2023-03-08 12:12:49,824 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.555e+02 2.243e+02 2.592e+02 3.349e+02 5.777e+02, threshold=5.184e+02, percent-clipped=2.0 2023-03-08 12:13:06,591 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9358, 5.3922, 4.4354, 5.5147, 4.8349, 5.1139, 5.4226, 5.3010], device='cuda:1'), covar=tensor([0.0602, 0.0306, 0.1333, 0.0287, 0.0403, 0.0228, 0.0313, 0.0190], device='cuda:1'), in_proj_covar=tensor([0.0369, 0.0288, 0.0348, 0.0305, 0.0298, 0.0225, 0.0279, 0.0258], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0005, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 12:13:08,901 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58141.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:13:14,163 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58144.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:13:24,734 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58151.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:13:48,263 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=58166.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:13:58,770 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-03-08 12:13:58,919 INFO [train2.py:809] (1/4) Epoch 15, batch 2400, loss[ctc_loss=0.06686, att_loss=0.2354, loss=0.2017, over 16277.00 frames. utt_duration=1516 frames, utt_pad_proportion=0.007432, over 43.00 utterances.], tot_loss[ctc_loss=0.08687, att_loss=0.2424, loss=0.2113, over 3267894.44 frames. utt_duration=1253 frames, utt_pad_proportion=0.05513, over 10446.02 utterances.], batch size: 43, lr: 7.21e-03, grad_scale: 8.0 2023-03-08 12:14:14,987 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1241, 5.1126, 5.0330, 2.4337, 1.8974, 2.8343, 2.6193, 3.8619], device='cuda:1'), covar=tensor([0.0658, 0.0217, 0.0201, 0.3970, 0.5498, 0.2407, 0.2563, 0.1683], device='cuda:1'), in_proj_covar=tensor([0.0344, 0.0239, 0.0245, 0.0225, 0.0343, 0.0336, 0.0234, 0.0355], device='cuda:1'), out_proj_covar=tensor([1.5011e-04, 9.0163e-05, 1.0601e-04, 9.7931e-05, 1.4621e-04, 1.3341e-04, 9.3777e-05, 1.4697e-04], device='cuda:1') 2023-03-08 12:14:40,626 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=58199.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:14:45,834 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58202.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:14:50,495 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58205.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:15:18,771 INFO [train2.py:809] (1/4) Epoch 15, batch 2450, loss[ctc_loss=0.06808, att_loss=0.2371, loss=0.2033, over 16268.00 frames. utt_duration=1515 frames, utt_pad_proportion=0.008011, over 43.00 utterances.], tot_loss[ctc_loss=0.08712, att_loss=0.2427, loss=0.2116, over 3271546.62 frames. utt_duration=1250 frames, utt_pad_proportion=0.05538, over 10477.48 utterances.], batch size: 43, lr: 7.20e-03, grad_scale: 8.0 2023-03-08 12:15:27,700 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.313e+02 2.141e+02 2.652e+02 3.416e+02 6.108e+02, threshold=5.305e+02, percent-clipped=4.0 2023-03-08 12:16:09,644 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0132, 5.0367, 4.9440, 2.1571, 1.8989, 2.8446, 2.4829, 3.8289], device='cuda:1'), covar=tensor([0.0763, 0.0209, 0.0191, 0.5163, 0.6114, 0.2666, 0.3073, 0.1705], device='cuda:1'), in_proj_covar=tensor([0.0349, 0.0244, 0.0248, 0.0229, 0.0348, 0.0340, 0.0238, 0.0360], device='cuda:1'), out_proj_covar=tensor([1.5226e-04, 9.1735e-05, 1.0756e-04, 9.9729e-05, 1.4799e-04, 1.3502e-04, 9.5011e-05, 1.4916e-04], device='cuda:1') 2023-03-08 12:16:38,087 INFO [train2.py:809] (1/4) Epoch 15, batch 2500, loss[ctc_loss=0.06797, att_loss=0.225, loss=0.1936, over 15935.00 frames. utt_duration=1557 frames, utt_pad_proportion=0.007928, over 41.00 utterances.], tot_loss[ctc_loss=0.08724, att_loss=0.2426, loss=0.2115, over 3272820.77 frames. utt_duration=1241 frames, utt_pad_proportion=0.05747, over 10559.20 utterances.], batch size: 41, lr: 7.20e-03, grad_scale: 8.0 2023-03-08 12:17:57,547 INFO [train2.py:809] (1/4) Epoch 15, batch 2550, loss[ctc_loss=0.09146, att_loss=0.2462, loss=0.2153, over 17411.00 frames. utt_duration=1011 frames, utt_pad_proportion=0.0473, over 69.00 utterances.], tot_loss[ctc_loss=0.08685, att_loss=0.2428, loss=0.2116, over 3288781.80 frames. utt_duration=1254 frames, utt_pad_proportion=0.0498, over 10504.92 utterances.], batch size: 69, lr: 7.20e-03, grad_scale: 8.0 2023-03-08 12:18:06,739 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.352e+02 2.062e+02 2.631e+02 3.242e+02 5.829e+02, threshold=5.263e+02, percent-clipped=2.0 2023-03-08 12:18:15,304 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.76 vs. limit=2.0 2023-03-08 12:18:20,517 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-03-08 12:18:52,980 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-03-08 12:19:04,263 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.22 vs. limit=5.0 2023-03-08 12:19:17,182 INFO [train2.py:809] (1/4) Epoch 15, batch 2600, loss[ctc_loss=0.09978, att_loss=0.2481, loss=0.2185, over 16387.00 frames. utt_duration=1491 frames, utt_pad_proportion=0.007293, over 44.00 utterances.], tot_loss[ctc_loss=0.08561, att_loss=0.2419, loss=0.2106, over 3286578.60 frames. utt_duration=1277 frames, utt_pad_proportion=0.04579, over 10304.45 utterances.], batch size: 44, lr: 7.19e-03, grad_scale: 8.0 2023-03-08 12:19:43,515 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58389.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:20:37,379 INFO [train2.py:809] (1/4) Epoch 15, batch 2650, loss[ctc_loss=0.074, att_loss=0.2241, loss=0.1941, over 15884.00 frames. utt_duration=1631 frames, utt_pad_proportion=0.008682, over 39.00 utterances.], tot_loss[ctc_loss=0.08548, att_loss=0.2418, loss=0.2105, over 3280469.81 frames. utt_duration=1273 frames, utt_pad_proportion=0.04936, over 10319.75 utterances.], batch size: 39, lr: 7.19e-03, grad_scale: 8.0 2023-03-08 12:20:46,673 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.284e+02 2.218e+02 2.453e+02 3.355e+02 6.382e+02, threshold=4.906e+02, percent-clipped=2.0 2023-03-08 12:21:00,180 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=58437.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:21:19,811 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9767, 4.3393, 4.2368, 4.5732, 2.6501, 4.4926, 2.2812, 1.4831], device='cuda:1'), covar=tensor([0.0384, 0.0215, 0.0799, 0.0141, 0.1960, 0.0157, 0.1787, 0.1966], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0134, 0.0248, 0.0124, 0.0219, 0.0117, 0.0220, 0.0194], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 12:21:57,762 INFO [train2.py:809] (1/4) Epoch 15, batch 2700, loss[ctc_loss=0.08833, att_loss=0.2405, loss=0.21, over 16556.00 frames. utt_duration=1473 frames, utt_pad_proportion=0.005312, over 45.00 utterances.], tot_loss[ctc_loss=0.08626, att_loss=0.2424, loss=0.2112, over 3279232.18 frames. utt_duration=1237 frames, utt_pad_proportion=0.05834, over 10612.92 utterances.], batch size: 45, lr: 7.19e-03, grad_scale: 8.0 2023-03-08 12:22:36,302 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=58497.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:22:41,392 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=58500.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:23:17,026 INFO [train2.py:809] (1/4) Epoch 15, batch 2750, loss[ctc_loss=0.08518, att_loss=0.2523, loss=0.2189, over 17420.00 frames. utt_duration=1108 frames, utt_pad_proportion=0.03016, over 63.00 utterances.], tot_loss[ctc_loss=0.08736, att_loss=0.2432, loss=0.212, over 3281447.19 frames. utt_duration=1229 frames, utt_pad_proportion=0.06026, over 10690.52 utterances.], batch size: 63, lr: 7.18e-03, grad_scale: 8.0 2023-03-08 12:23:26,924 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.411e+02 2.307e+02 2.848e+02 3.393e+02 6.929e+02, threshold=5.697e+02, percent-clipped=6.0 2023-03-08 12:23:43,338 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9809, 4.9855, 4.9094, 2.1756, 1.9588, 2.6836, 2.3495, 3.7978], device='cuda:1'), covar=tensor([0.0755, 0.0203, 0.0196, 0.5026, 0.5756, 0.2831, 0.3405, 0.1657], device='cuda:1'), in_proj_covar=tensor([0.0343, 0.0239, 0.0246, 0.0226, 0.0341, 0.0335, 0.0234, 0.0354], device='cuda:1'), out_proj_covar=tensor([1.4944e-04, 9.0077e-05, 1.0672e-04, 9.8309e-05, 1.4522e-04, 1.3289e-04, 9.3478e-05, 1.4661e-04], device='cuda:1') 2023-03-08 12:24:00,563 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9748, 5.2416, 5.2191, 5.1832, 5.2653, 5.2623, 4.9741, 4.7681], device='cuda:1'), covar=tensor([0.1021, 0.0548, 0.0292, 0.0550, 0.0309, 0.0341, 0.0373, 0.0387], device='cuda:1'), in_proj_covar=tensor([0.0478, 0.0322, 0.0291, 0.0318, 0.0376, 0.0393, 0.0316, 0.0357], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 12:24:27,195 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.4710, 5.3221, 5.1738, 3.0121, 5.2022, 4.8925, 4.7192, 3.1035], device='cuda:1'), covar=tensor([0.0086, 0.0081, 0.0222, 0.1042, 0.0071, 0.0166, 0.0252, 0.1165], device='cuda:1'), in_proj_covar=tensor([0.0067, 0.0093, 0.0090, 0.0108, 0.0077, 0.0103, 0.0098, 0.0103], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 12:24:28,061 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-03-08 12:24:35,860 INFO [train2.py:809] (1/4) Epoch 15, batch 2800, loss[ctc_loss=0.08324, att_loss=0.2378, loss=0.2069, over 16880.00 frames. utt_duration=1379 frames, utt_pad_proportion=0.00762, over 49.00 utterances.], tot_loss[ctc_loss=0.08787, att_loss=0.2433, loss=0.2122, over 3284827.87 frames. utt_duration=1233 frames, utt_pad_proportion=0.05856, over 10668.65 utterances.], batch size: 49, lr: 7.18e-03, grad_scale: 8.0 2023-03-08 12:25:21,891 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.6503, 3.6320, 2.9950, 3.3122, 3.8097, 3.4452, 2.9258, 3.9660], device='cuda:1'), covar=tensor([0.1125, 0.0449, 0.1027, 0.0643, 0.0625, 0.0693, 0.0812, 0.0480], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0198, 0.0215, 0.0187, 0.0257, 0.0227, 0.0191, 0.0273], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-08 12:25:53,748 INFO [train2.py:809] (1/4) Epoch 15, batch 2850, loss[ctc_loss=0.1126, att_loss=0.2566, loss=0.2278, over 16616.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.005797, over 47.00 utterances.], tot_loss[ctc_loss=0.08786, att_loss=0.2438, loss=0.2126, over 3297320.55 frames. utt_duration=1245 frames, utt_pad_proportion=0.05118, over 10605.76 utterances.], batch size: 47, lr: 7.18e-03, grad_scale: 8.0 2023-03-08 12:26:02,954 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.515e+02 2.157e+02 2.674e+02 3.074e+02 6.924e+02, threshold=5.347e+02, percent-clipped=2.0 2023-03-08 12:26:15,027 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7860, 5.0855, 4.6201, 5.1644, 4.5755, 4.8027, 5.2376, 5.0688], device='cuda:1'), covar=tensor([0.0564, 0.0271, 0.0927, 0.0280, 0.0452, 0.0312, 0.0257, 0.0186], device='cuda:1'), in_proj_covar=tensor([0.0373, 0.0293, 0.0350, 0.0311, 0.0302, 0.0228, 0.0282, 0.0261], device='cuda:1'), out_proj_covar=tensor([0.0007, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 12:27:12,215 INFO [train2.py:809] (1/4) Epoch 15, batch 2900, loss[ctc_loss=0.1318, att_loss=0.2678, loss=0.2406, over 17025.00 frames. utt_duration=1311 frames, utt_pad_proportion=0.009688, over 52.00 utterances.], tot_loss[ctc_loss=0.0884, att_loss=0.244, loss=0.2129, over 3295081.17 frames. utt_duration=1231 frames, utt_pad_proportion=0.05483, over 10717.58 utterances.], batch size: 52, lr: 7.18e-03, grad_scale: 8.0 2023-03-08 12:27:26,486 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58681.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:27:30,993 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6947, 3.0557, 3.7107, 3.1417, 3.6706, 4.8026, 4.5479, 3.5694], device='cuda:1'), covar=tensor([0.0374, 0.1662, 0.1194, 0.1365, 0.0965, 0.0630, 0.0433, 0.1143], device='cuda:1'), in_proj_covar=tensor([0.0237, 0.0237, 0.0266, 0.0210, 0.0254, 0.0333, 0.0239, 0.0229], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 12:27:34,034 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.5329, 2.1699, 2.2325, 2.6623, 2.3809, 2.6601, 2.2874, 2.9548], device='cuda:1'), covar=tensor([0.1846, 0.3031, 0.2836, 0.1471, 0.2347, 0.1342, 0.2972, 0.0743], device='cuda:1'), in_proj_covar=tensor([0.0091, 0.0096, 0.0099, 0.0085, 0.0090, 0.0080, 0.0103, 0.0070], device='cuda:1'), out_proj_covar=tensor([6.5563e-05, 7.1549e-05, 7.4570e-05, 6.3808e-05, 6.4880e-05, 6.2511e-05, 7.4042e-05, 5.5572e-05], device='cuda:1') 2023-03-08 12:28:31,199 INFO [train2.py:809] (1/4) Epoch 15, batch 2950, loss[ctc_loss=0.08345, att_loss=0.2506, loss=0.2172, over 16620.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.005782, over 47.00 utterances.], tot_loss[ctc_loss=0.08777, att_loss=0.2433, loss=0.2122, over 3284243.42 frames. utt_duration=1245 frames, utt_pad_proportion=0.05455, over 10563.04 utterances.], batch size: 47, lr: 7.17e-03, grad_scale: 8.0 2023-03-08 12:28:41,001 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.392e+02 2.063e+02 2.403e+02 2.955e+02 5.605e+02, threshold=4.806e+02, percent-clipped=1.0 2023-03-08 12:29:01,926 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58742.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:29:31,696 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0807, 5.3441, 5.6313, 5.4709, 5.4839, 6.0199, 5.1938, 6.1296], device='cuda:1'), covar=tensor([0.0672, 0.0683, 0.0690, 0.1049, 0.1844, 0.0888, 0.0700, 0.0622], device='cuda:1'), in_proj_covar=tensor([0.0799, 0.0473, 0.0552, 0.0606, 0.0805, 0.0564, 0.0452, 0.0537], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 12:29:35,008 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4528, 2.4540, 5.0049, 3.8873, 2.8317, 4.2601, 4.8141, 4.6631], device='cuda:1'), covar=tensor([0.0242, 0.1729, 0.0161, 0.0882, 0.1849, 0.0244, 0.0106, 0.0202], device='cuda:1'), in_proj_covar=tensor([0.0164, 0.0239, 0.0155, 0.0304, 0.0261, 0.0189, 0.0139, 0.0169], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 12:29:49,922 INFO [train2.py:809] (1/4) Epoch 15, batch 3000, loss[ctc_loss=0.06825, att_loss=0.2145, loss=0.1853, over 15753.00 frames. utt_duration=1659 frames, utt_pad_proportion=0.009892, over 38.00 utterances.], tot_loss[ctc_loss=0.08734, att_loss=0.243, loss=0.2119, over 3280513.13 frames. utt_duration=1239 frames, utt_pad_proportion=0.05609, over 10603.73 utterances.], batch size: 38, lr: 7.17e-03, grad_scale: 8.0 2023-03-08 12:29:49,922 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-08 12:30:03,570 INFO [train2.py:843] (1/4) Epoch 15, validation: ctc_loss=0.04336, att_loss=0.2362, loss=0.1976, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-08 12:30:03,571 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-08 12:30:07,189 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58775.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:30:42,500 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9411, 5.2314, 4.7697, 5.3177, 4.6276, 4.9913, 5.4032, 5.1658], device='cuda:1'), covar=tensor([0.0587, 0.0283, 0.0873, 0.0252, 0.0501, 0.0217, 0.0230, 0.0192], device='cuda:1'), in_proj_covar=tensor([0.0369, 0.0289, 0.0344, 0.0306, 0.0299, 0.0225, 0.0278, 0.0259], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 12:30:42,564 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58797.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:30:47,670 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58800.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:31:24,437 INFO [train2.py:809] (1/4) Epoch 15, batch 3050, loss[ctc_loss=0.08582, att_loss=0.2369, loss=0.2067, over 16193.00 frames. utt_duration=1581 frames, utt_pad_proportion=0.006086, over 41.00 utterances.], tot_loss[ctc_loss=0.08825, att_loss=0.2435, loss=0.2124, over 3281383.74 frames. utt_duration=1223 frames, utt_pad_proportion=0.05907, over 10744.31 utterances.], batch size: 41, lr: 7.17e-03, grad_scale: 8.0 2023-03-08 12:31:24,809 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5311, 2.9114, 3.6176, 3.0993, 3.4686, 4.6251, 4.4129, 3.2854], device='cuda:1'), covar=tensor([0.0372, 0.1690, 0.1220, 0.1355, 0.1058, 0.0760, 0.0537, 0.1345], device='cuda:1'), in_proj_covar=tensor([0.0236, 0.0235, 0.0264, 0.0211, 0.0253, 0.0332, 0.0238, 0.0229], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 12:31:34,021 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.376e+02 2.164e+02 2.584e+02 3.141e+02 6.750e+02, threshold=5.168e+02, percent-clipped=6.0 2023-03-08 12:31:45,919 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58836.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:31:54,495 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-03-08 12:31:56,577 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.5440, 2.8230, 3.5012, 4.3903, 3.9437, 3.8244, 3.0317, 2.1874], device='cuda:1'), covar=tensor([0.0550, 0.1998, 0.0848, 0.0584, 0.0705, 0.0502, 0.1418, 0.2220], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0210, 0.0185, 0.0198, 0.0204, 0.0166, 0.0195, 0.0178], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 12:31:59,460 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=58845.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:32:04,252 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=58848.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:32:38,151 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-03-08 12:32:43,310 INFO [train2.py:809] (1/4) Epoch 15, batch 3100, loss[ctc_loss=0.0691, att_loss=0.2139, loss=0.1849, over 15845.00 frames. utt_duration=1627 frames, utt_pad_proportion=0.01178, over 39.00 utterances.], tot_loss[ctc_loss=0.08904, att_loss=0.2443, loss=0.2132, over 3281228.10 frames. utt_duration=1197 frames, utt_pad_proportion=0.06626, over 10979.50 utterances.], batch size: 39, lr: 7.16e-03, grad_scale: 8.0 2023-03-08 12:33:10,793 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-03-08 12:33:34,863 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58906.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:34:00,661 INFO [train2.py:809] (1/4) Epoch 15, batch 3150, loss[ctc_loss=0.0879, att_loss=0.24, loss=0.2096, over 15975.00 frames. utt_duration=1560 frames, utt_pad_proportion=0.005102, over 41.00 utterances.], tot_loss[ctc_loss=0.08872, att_loss=0.2437, loss=0.2127, over 3275457.71 frames. utt_duration=1227 frames, utt_pad_proportion=0.06009, over 10690.81 utterances.], batch size: 41, lr: 7.16e-03, grad_scale: 8.0 2023-03-08 12:34:09,777 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.384e+02 2.265e+02 2.592e+02 3.060e+02 7.077e+02, threshold=5.184e+02, percent-clipped=3.0 2023-03-08 12:35:10,531 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58967.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:35:19,416 INFO [train2.py:809] (1/4) Epoch 15, batch 3200, loss[ctc_loss=0.1131, att_loss=0.2594, loss=0.2301, over 17068.00 frames. utt_duration=1314 frames, utt_pad_proportion=0.00719, over 52.00 utterances.], tot_loss[ctc_loss=0.08859, att_loss=0.2438, loss=0.2128, over 3277604.27 frames. utt_duration=1209 frames, utt_pad_proportion=0.06324, over 10859.78 utterances.], batch size: 52, lr: 7.16e-03, grad_scale: 8.0 2023-03-08 12:35:27,322 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9804, 5.0492, 4.9845, 2.3888, 1.8431, 2.9672, 2.5107, 3.7357], device='cuda:1'), covar=tensor([0.0819, 0.0326, 0.0276, 0.4820, 0.6144, 0.2545, 0.3518, 0.1959], device='cuda:1'), in_proj_covar=tensor([0.0338, 0.0237, 0.0242, 0.0220, 0.0337, 0.0327, 0.0229, 0.0349], device='cuda:1'), out_proj_covar=tensor([1.4682e-04, 8.8809e-05, 1.0470e-04, 9.5357e-05, 1.4328e-04, 1.3008e-04, 9.1767e-05, 1.4439e-04], device='cuda:1') 2023-03-08 12:36:38,232 INFO [train2.py:809] (1/4) Epoch 15, batch 3250, loss[ctc_loss=0.07968, att_loss=0.2366, loss=0.2052, over 16387.00 frames. utt_duration=1491 frames, utt_pad_proportion=0.007954, over 44.00 utterances.], tot_loss[ctc_loss=0.08797, att_loss=0.2435, loss=0.2124, over 3278169.71 frames. utt_duration=1211 frames, utt_pad_proportion=0.06361, over 10837.64 utterances.], batch size: 44, lr: 7.15e-03, grad_scale: 8.0 2023-03-08 12:36:48,043 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.212e+02 2.216e+02 2.757e+02 3.407e+02 8.183e+02, threshold=5.513e+02, percent-clipped=5.0 2023-03-08 12:37:00,978 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59037.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:37:57,616 INFO [train2.py:809] (1/4) Epoch 15, batch 3300, loss[ctc_loss=0.06876, att_loss=0.2179, loss=0.1881, over 15624.00 frames. utt_duration=1691 frames, utt_pad_proportion=0.00903, over 37.00 utterances.], tot_loss[ctc_loss=0.08786, att_loss=0.2438, loss=0.2126, over 3278053.48 frames. utt_duration=1210 frames, utt_pad_proportion=0.06176, over 10845.83 utterances.], batch size: 37, lr: 7.15e-03, grad_scale: 8.0 2023-03-08 12:38:36,015 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.8886, 4.6056, 4.7390, 2.2874, 1.9975, 2.8152, 2.4200, 3.6786], device='cuda:1'), covar=tensor([0.0817, 0.0208, 0.0207, 0.4457, 0.5668, 0.2650, 0.2950, 0.1593], device='cuda:1'), in_proj_covar=tensor([0.0347, 0.0242, 0.0249, 0.0227, 0.0346, 0.0337, 0.0236, 0.0358], device='cuda:1'), out_proj_covar=tensor([1.5095e-04, 9.0644e-05, 1.0743e-04, 9.8867e-05, 1.4688e-04, 1.3366e-04, 9.4426e-05, 1.4806e-04], device='cuda:1') 2023-03-08 12:39:16,674 INFO [train2.py:809] (1/4) Epoch 15, batch 3350, loss[ctc_loss=0.1657, att_loss=0.2852, loss=0.2613, over 13979.00 frames. utt_duration=387.2 frames, utt_pad_proportion=0.3266, over 145.00 utterances.], tot_loss[ctc_loss=0.08754, att_loss=0.2437, loss=0.2125, over 3279172.70 frames. utt_duration=1204 frames, utt_pad_proportion=0.06352, over 10908.13 utterances.], batch size: 145, lr: 7.15e-03, grad_scale: 8.0 2023-03-08 12:39:17,553 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-03-08 12:39:27,190 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.236e+02 1.917e+02 2.406e+02 2.955e+02 7.524e+02, threshold=4.811e+02, percent-clipped=3.0 2023-03-08 12:39:30,493 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59131.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:40:22,563 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59164.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:40:36,061 INFO [train2.py:809] (1/4) Epoch 15, batch 3400, loss[ctc_loss=0.1069, att_loss=0.2519, loss=0.2229, over 16623.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.005648, over 47.00 utterances.], tot_loss[ctc_loss=0.08744, att_loss=0.2432, loss=0.212, over 3285044.54 frames. utt_duration=1226 frames, utt_pad_proportion=0.0565, over 10732.38 utterances.], batch size: 47, lr: 7.15e-03, grad_scale: 8.0 2023-03-08 12:41:56,068 INFO [train2.py:809] (1/4) Epoch 15, batch 3450, loss[ctc_loss=0.09019, att_loss=0.2537, loss=0.221, over 16621.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.005439, over 47.00 utterances.], tot_loss[ctc_loss=0.08721, att_loss=0.2428, loss=0.2117, over 3278626.79 frames. utt_duration=1236 frames, utt_pad_proportion=0.0548, over 10624.58 utterances.], batch size: 47, lr: 7.14e-03, grad_scale: 8.0 2023-03-08 12:42:00,144 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59225.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:42:05,712 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0272, 4.2466, 4.2364, 4.6171, 2.5445, 4.4935, 2.3452, 1.5397], device='cuda:1'), covar=tensor([0.0423, 0.0248, 0.0732, 0.0155, 0.1803, 0.0160, 0.1742, 0.1898], device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0137, 0.0250, 0.0127, 0.0222, 0.0120, 0.0224, 0.0197], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 12:42:06,827 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.536e+02 2.401e+02 2.877e+02 3.602e+02 6.308e+02, threshold=5.755e+02, percent-clipped=6.0 2023-03-08 12:42:55,980 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0494, 4.4318, 4.4157, 4.7636, 2.7306, 4.6747, 2.2630, 1.5675], device='cuda:1'), covar=tensor([0.0419, 0.0213, 0.0688, 0.0128, 0.1833, 0.0143, 0.1808, 0.1978], device='cuda:1'), in_proj_covar=tensor([0.0162, 0.0137, 0.0252, 0.0127, 0.0223, 0.0120, 0.0225, 0.0198], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 12:42:58,739 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59262.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:43:16,322 INFO [train2.py:809] (1/4) Epoch 15, batch 3500, loss[ctc_loss=0.1075, att_loss=0.2572, loss=0.2272, over 17377.00 frames. utt_duration=1105 frames, utt_pad_proportion=0.03338, over 63.00 utterances.], tot_loss[ctc_loss=0.08776, att_loss=0.2435, loss=0.2124, over 3279356.38 frames. utt_duration=1213 frames, utt_pad_proportion=0.06027, over 10828.33 utterances.], batch size: 63, lr: 7.14e-03, grad_scale: 8.0 2023-03-08 12:44:35,608 INFO [train2.py:809] (1/4) Epoch 15, batch 3550, loss[ctc_loss=0.08432, att_loss=0.226, loss=0.1977, over 15885.00 frames. utt_duration=1631 frames, utt_pad_proportion=0.009378, over 39.00 utterances.], tot_loss[ctc_loss=0.08797, att_loss=0.2436, loss=0.2125, over 3272043.19 frames. utt_duration=1213 frames, utt_pad_proportion=0.06264, over 10803.25 utterances.], batch size: 39, lr: 7.14e-03, grad_scale: 8.0 2023-03-08 12:44:45,171 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.189e+02 2.222e+02 2.629e+02 3.391e+02 5.578e+02, threshold=5.259e+02, percent-clipped=0.0 2023-03-08 12:44:57,796 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59337.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:45:40,626 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6418, 2.4996, 5.1122, 3.9856, 2.8233, 4.3258, 4.9856, 4.6419], device='cuda:1'), covar=tensor([0.0246, 0.1684, 0.0180, 0.1028, 0.2061, 0.0254, 0.0114, 0.0260], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0238, 0.0156, 0.0306, 0.0263, 0.0190, 0.0141, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 12:45:54,582 INFO [train2.py:809] (1/4) Epoch 15, batch 3600, loss[ctc_loss=0.08737, att_loss=0.251, loss=0.2183, over 17107.00 frames. utt_duration=1223 frames, utt_pad_proportion=0.01579, over 56.00 utterances.], tot_loss[ctc_loss=0.08791, att_loss=0.2435, loss=0.2124, over 3268281.81 frames. utt_duration=1203 frames, utt_pad_proportion=0.0669, over 10880.70 utterances.], batch size: 56, lr: 7.13e-03, grad_scale: 8.0 2023-03-08 12:46:13,882 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=59385.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:47:15,176 INFO [train2.py:809] (1/4) Epoch 15, batch 3650, loss[ctc_loss=0.09494, att_loss=0.2341, loss=0.2063, over 15938.00 frames. utt_duration=1556 frames, utt_pad_proportion=0.007435, over 41.00 utterances.], tot_loss[ctc_loss=0.08752, att_loss=0.2428, loss=0.2118, over 3267079.09 frames. utt_duration=1217 frames, utt_pad_proportion=0.06395, over 10746.96 utterances.], batch size: 41, lr: 7.13e-03, grad_scale: 8.0 2023-03-08 12:47:24,349 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.406e+02 2.155e+02 2.711e+02 3.466e+02 9.955e+02, threshold=5.422e+02, percent-clipped=4.0 2023-03-08 12:47:27,722 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59431.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:47:42,936 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0826, 5.3845, 5.6269, 5.4637, 5.6107, 6.0727, 5.2475, 6.1333], device='cuda:1'), covar=tensor([0.0707, 0.0682, 0.0719, 0.1166, 0.1777, 0.0852, 0.0603, 0.0627], device='cuda:1'), in_proj_covar=tensor([0.0802, 0.0474, 0.0553, 0.0615, 0.0808, 0.0565, 0.0459, 0.0548], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 12:48:02,754 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1023, 4.5751, 4.3210, 4.6521, 2.8654, 4.6693, 2.2553, 2.0516], device='cuda:1'), covar=tensor([0.0344, 0.0163, 0.0769, 0.0152, 0.1759, 0.0142, 0.1806, 0.1707], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0134, 0.0246, 0.0125, 0.0218, 0.0117, 0.0221, 0.0193], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 12:48:26,450 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8566, 5.2279, 4.7510, 5.2850, 4.6170, 4.9613, 5.3479, 5.1035], device='cuda:1'), covar=tensor([0.0627, 0.0268, 0.0810, 0.0281, 0.0441, 0.0214, 0.0214, 0.0184], device='cuda:1'), in_proj_covar=tensor([0.0363, 0.0287, 0.0339, 0.0304, 0.0293, 0.0223, 0.0274, 0.0258], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0005, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 12:48:34,111 INFO [train2.py:809] (1/4) Epoch 15, batch 3700, loss[ctc_loss=0.1018, att_loss=0.2528, loss=0.2226, over 17289.00 frames. utt_duration=1099 frames, utt_pad_proportion=0.03946, over 63.00 utterances.], tot_loss[ctc_loss=0.08763, att_loss=0.2431, loss=0.212, over 3268482.61 frames. utt_duration=1217 frames, utt_pad_proportion=0.06305, over 10751.63 utterances.], batch size: 63, lr: 7.13e-03, grad_scale: 8.0 2023-03-08 12:48:43,277 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=59479.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:49:20,208 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59502.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:49:44,340 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.82 vs. limit=2.0 2023-03-08 12:49:49,034 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59520.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:49:53,483 INFO [train2.py:809] (1/4) Epoch 15, batch 3750, loss[ctc_loss=0.08665, att_loss=0.2301, loss=0.2014, over 16136.00 frames. utt_duration=1538 frames, utt_pad_proportion=0.005113, over 42.00 utterances.], tot_loss[ctc_loss=0.08726, att_loss=0.2429, loss=0.2117, over 3267194.98 frames. utt_duration=1225 frames, utt_pad_proportion=0.06145, over 10677.80 utterances.], batch size: 42, lr: 7.12e-03, grad_scale: 8.0 2023-03-08 12:50:02,433 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.260e+02 2.043e+02 2.443e+02 3.044e+02 5.040e+02, threshold=4.885e+02, percent-clipped=0.0 2023-03-08 12:50:29,626 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5488, 4.4946, 4.4216, 4.4996, 5.0468, 4.6015, 4.4750, 2.2180], device='cuda:1'), covar=tensor([0.0198, 0.0274, 0.0323, 0.0237, 0.0751, 0.0176, 0.0269, 0.2042], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0148, 0.0154, 0.0168, 0.0348, 0.0134, 0.0142, 0.0212], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 12:50:54,606 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59562.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:50:54,666 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7928, 3.4961, 2.9112, 3.1714, 3.7318, 3.3767, 2.5212, 3.9620], device='cuda:1'), covar=tensor([0.1101, 0.0531, 0.1109, 0.0722, 0.0747, 0.0710, 0.1040, 0.0487], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0200, 0.0216, 0.0190, 0.0257, 0.0228, 0.0192, 0.0275], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-08 12:50:56,183 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59563.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:51:12,187 INFO [train2.py:809] (1/4) Epoch 15, batch 3800, loss[ctc_loss=0.08408, att_loss=0.2293, loss=0.2003, over 15489.00 frames. utt_duration=1723 frames, utt_pad_proportion=0.009392, over 36.00 utterances.], tot_loss[ctc_loss=0.08736, att_loss=0.243, loss=0.2119, over 3277331.57 frames. utt_duration=1252 frames, utt_pad_proportion=0.05251, over 10481.59 utterances.], batch size: 36, lr: 7.12e-03, grad_scale: 8.0 2023-03-08 12:52:09,968 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=59610.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:52:31,314 INFO [train2.py:809] (1/4) Epoch 15, batch 3850, loss[ctc_loss=0.06804, att_loss=0.2301, loss=0.1977, over 16174.00 frames. utt_duration=1579 frames, utt_pad_proportion=0.007282, over 41.00 utterances.], tot_loss[ctc_loss=0.08608, att_loss=0.2418, loss=0.2106, over 3277872.54 frames. utt_duration=1285 frames, utt_pad_proportion=0.04475, over 10211.91 utterances.], batch size: 41, lr: 7.12e-03, grad_scale: 8.0 2023-03-08 12:52:40,311 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.434e+02 2.221e+02 2.684e+02 3.228e+02 7.729e+02, threshold=5.369e+02, percent-clipped=7.0 2023-03-08 12:53:46,844 INFO [train2.py:809] (1/4) Epoch 15, batch 3900, loss[ctc_loss=0.1257, att_loss=0.268, loss=0.2396, over 14826.00 frames. utt_duration=407.7 frames, utt_pad_proportion=0.2886, over 146.00 utterances.], tot_loss[ctc_loss=0.08635, att_loss=0.2419, loss=0.2108, over 3280340.51 frames. utt_duration=1264 frames, utt_pad_proportion=0.04899, over 10397.01 utterances.], batch size: 146, lr: 7.12e-03, grad_scale: 8.0 2023-03-08 12:55:03,847 INFO [train2.py:809] (1/4) Epoch 15, batch 3950, loss[ctc_loss=0.07664, att_loss=0.2486, loss=0.2142, over 17006.00 frames. utt_duration=1336 frames, utt_pad_proportion=0.00853, over 51.00 utterances.], tot_loss[ctc_loss=0.08545, att_loss=0.2418, loss=0.2106, over 3286332.77 frames. utt_duration=1268 frames, utt_pad_proportion=0.04689, over 10379.24 utterances.], batch size: 51, lr: 7.11e-03, grad_scale: 8.0 2023-03-08 12:55:12,750 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.264e+02 1.994e+02 2.418e+02 2.996e+02 5.766e+02, threshold=4.836e+02, percent-clipped=1.0 2023-03-08 12:55:36,054 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0602, 4.3129, 4.2685, 4.5555, 2.7853, 4.4964, 2.1748, 2.1708], device='cuda:1'), covar=tensor([0.0428, 0.0279, 0.0777, 0.0167, 0.1696, 0.0176, 0.1822, 0.1543], device='cuda:1'), in_proj_covar=tensor([0.0163, 0.0137, 0.0249, 0.0127, 0.0221, 0.0120, 0.0224, 0.0195], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 12:56:20,785 INFO [train2.py:809] (1/4) Epoch 16, batch 0, loss[ctc_loss=0.1358, att_loss=0.2728, loss=0.2454, over 14227.00 frames. utt_duration=388.7 frames, utt_pad_proportion=0.3193, over 147.00 utterances.], tot_loss[ctc_loss=0.1358, att_loss=0.2728, loss=0.2454, over 14227.00 frames. utt_duration=388.7 frames, utt_pad_proportion=0.3193, over 147.00 utterances.], batch size: 147, lr: 6.88e-03, grad_scale: 8.0 2023-03-08 12:56:20,786 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-08 12:56:32,579 INFO [train2.py:843] (1/4) Epoch 16, validation: ctc_loss=0.04399, att_loss=0.2363, loss=0.1978, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-08 12:56:32,580 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-08 12:57:19,963 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.5574, 5.8707, 5.2730, 5.6644, 5.4897, 5.1350, 5.2900, 5.1934], device='cuda:1'), covar=tensor([0.1421, 0.0905, 0.0826, 0.0719, 0.0946, 0.1321, 0.2219, 0.2176], device='cuda:1'), in_proj_covar=tensor([0.0475, 0.0551, 0.0416, 0.0418, 0.0397, 0.0432, 0.0557, 0.0495], 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') 2023-03-08 12:57:40,348 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.0222, 3.9236, 3.3650, 3.6244, 4.0754, 3.7568, 3.3128, 4.4238], device='cuda:1'), covar=tensor([0.0954, 0.0448, 0.0828, 0.0584, 0.0592, 0.0590, 0.0707, 0.0437], device='cuda:1'), in_proj_covar=tensor([0.0193, 0.0199, 0.0215, 0.0187, 0.0255, 0.0226, 0.0189, 0.0271], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-08 12:57:52,432 INFO [train2.py:809] (1/4) Epoch 16, batch 50, loss[ctc_loss=0.08797, att_loss=0.2393, loss=0.209, over 16280.00 frames. utt_duration=1516 frames, utt_pad_proportion=0.00728, over 43.00 utterances.], tot_loss[ctc_loss=0.08524, att_loss=0.2396, loss=0.2087, over 731080.26 frames. utt_duration=1286 frames, utt_pad_proportion=0.05515, over 2276.71 utterances.], batch size: 43, lr: 6.88e-03, grad_scale: 8.0 2023-03-08 12:58:13,672 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59820.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:58:27,785 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.394e+02 2.112e+02 2.420e+02 3.068e+02 7.828e+02, threshold=4.840e+02, percent-clipped=6.0 2023-03-08 12:58:54,469 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.5547, 4.8736, 5.0975, 4.9805, 5.0168, 5.4580, 5.0008, 5.5593], device='cuda:1'), covar=tensor([0.0715, 0.0737, 0.0729, 0.1193, 0.1855, 0.0973, 0.0912, 0.0708], device='cuda:1'), in_proj_covar=tensor([0.0805, 0.0479, 0.0557, 0.0617, 0.0816, 0.0568, 0.0459, 0.0547], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 12:58:59,306 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59848.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:59:12,667 INFO [train2.py:809] (1/4) Epoch 16, batch 100, loss[ctc_loss=0.0785, att_loss=0.2224, loss=0.1936, over 16006.00 frames. utt_duration=1602 frames, utt_pad_proportion=0.00728, over 40.00 utterances.], tot_loss[ctc_loss=0.0865, att_loss=0.2413, loss=0.2104, over 1294673.52 frames. utt_duration=1281 frames, utt_pad_proportion=0.05219, over 4046.75 utterances.], batch size: 40, lr: 6.88e-03, grad_scale: 8.0 2023-03-08 12:59:14,421 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59858.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:59:30,342 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=59868.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 12:59:30,508 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8097, 5.1047, 4.6660, 5.2172, 4.5602, 4.8142, 5.3006, 5.0605], device='cuda:1'), covar=tensor([0.0609, 0.0324, 0.0920, 0.0304, 0.0539, 0.0255, 0.0230, 0.0193], device='cuda:1'), in_proj_covar=tensor([0.0363, 0.0289, 0.0340, 0.0301, 0.0297, 0.0223, 0.0276, 0.0258], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0005, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 13:00:03,277 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59889.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:00:31,901 INFO [train2.py:809] (1/4) Epoch 16, batch 150, loss[ctc_loss=0.07213, att_loss=0.2178, loss=0.1887, over 15627.00 frames. utt_duration=1691 frames, utt_pad_proportion=0.008761, over 37.00 utterances.], tot_loss[ctc_loss=0.08665, att_loss=0.2418, loss=0.2108, over 1738011.08 frames. utt_duration=1299 frames, utt_pad_proportion=0.04213, over 5356.55 utterances.], batch size: 37, lr: 6.87e-03, grad_scale: 8.0 2023-03-08 13:00:35,494 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59909.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:01:06,715 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.476e+02 2.211e+02 2.747e+02 3.339e+02 6.489e+02, threshold=5.494e+02, percent-clipped=4.0 2023-03-08 13:01:19,429 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1972, 4.5180, 4.5038, 4.6071, 2.7894, 4.4794, 2.7469, 1.9806], device='cuda:1'), covar=tensor([0.0350, 0.0225, 0.0674, 0.0221, 0.1672, 0.0190, 0.1430, 0.1622], device='cuda:1'), in_proj_covar=tensor([0.0160, 0.0135, 0.0246, 0.0125, 0.0217, 0.0117, 0.0221, 0.0193], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 13:01:37,593 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1982, 5.0618, 4.9433, 3.0282, 4.9085, 4.7138, 4.3453, 2.7365], device='cuda:1'), covar=tensor([0.0106, 0.0097, 0.0240, 0.1022, 0.0086, 0.0195, 0.0319, 0.1354], device='cuda:1'), in_proj_covar=tensor([0.0067, 0.0094, 0.0090, 0.0108, 0.0077, 0.0103, 0.0097, 0.0102], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 13:01:40,812 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59950.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:01:51,219 INFO [train2.py:809] (1/4) Epoch 16, batch 200, loss[ctc_loss=0.09155, att_loss=0.2467, loss=0.2157, over 17206.00 frames. utt_duration=872.5 frames, utt_pad_proportion=0.08641, over 79.00 utterances.], tot_loss[ctc_loss=0.08527, att_loss=0.2411, loss=0.2099, over 2078229.57 frames. utt_duration=1305 frames, utt_pad_proportion=0.04148, over 6379.69 utterances.], batch size: 79, lr: 6.87e-03, grad_scale: 8.0 2023-03-08 13:02:45,701 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5600, 4.9371, 4.7515, 4.9670, 5.0261, 4.6969, 3.7868, 4.9310], device='cuda:1'), covar=tensor([0.0104, 0.0086, 0.0115, 0.0075, 0.0071, 0.0086, 0.0545, 0.0149], device='cuda:1'), in_proj_covar=tensor([0.0082, 0.0079, 0.0099, 0.0061, 0.0066, 0.0078, 0.0097, 0.0100], 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') 2023-03-08 13:02:47,084 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.20 vs. limit=5.0 2023-03-08 13:03:15,171 INFO [train2.py:809] (1/4) Epoch 16, batch 250, loss[ctc_loss=0.08932, att_loss=0.2509, loss=0.2186, over 16947.00 frames. utt_duration=1357 frames, utt_pad_proportion=0.007968, over 50.00 utterances.], tot_loss[ctc_loss=0.08572, att_loss=0.2424, loss=0.2111, over 2355616.06 frames. utt_duration=1274 frames, utt_pad_proportion=0.04494, over 7404.86 utterances.], batch size: 50, lr: 6.87e-03, grad_scale: 8.0 2023-03-08 13:03:31,816 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.9288, 2.0128, 2.6751, 2.8859, 2.8433, 2.7108, 2.6512, 2.9911], device='cuda:1'), covar=tensor([0.1760, 0.4087, 0.2585, 0.1498, 0.1650, 0.1413, 0.2522, 0.0843], device='cuda:1'), in_proj_covar=tensor([0.0090, 0.0095, 0.0097, 0.0085, 0.0089, 0.0080, 0.0100, 0.0070], device='cuda:1'), out_proj_covar=tensor([6.5158e-05, 7.0772e-05, 7.3692e-05, 6.3835e-05, 6.4175e-05, 6.2730e-05, 7.2493e-05, 5.5581e-05], device='cuda:1') 2023-03-08 13:03:50,525 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.435e+02 2.138e+02 2.534e+02 3.199e+02 6.195e+02, threshold=5.067e+02, percent-clipped=2.0 2023-03-08 13:04:09,062 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60040.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:04:28,388 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60052.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:04:35,591 INFO [train2.py:809] (1/4) Epoch 16, batch 300, loss[ctc_loss=0.07893, att_loss=0.2133, loss=0.1864, over 15501.00 frames. utt_duration=1724 frames, utt_pad_proportion=0.007756, over 36.00 utterances.], tot_loss[ctc_loss=0.08575, att_loss=0.2417, loss=0.2105, over 2546901.63 frames. utt_duration=1276 frames, utt_pad_proportion=0.04676, over 7995.27 utterances.], batch size: 36, lr: 6.87e-03, grad_scale: 16.0 2023-03-08 13:04:49,143 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60065.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:05:09,383 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60078.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:05:27,662 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.7404, 5.0799, 5.2625, 5.1406, 5.1735, 5.6825, 5.0816, 5.7615], device='cuda:1'), covar=tensor([0.0699, 0.0617, 0.0738, 0.1201, 0.1692, 0.0829, 0.0818, 0.0703], device='cuda:1'), in_proj_covar=tensor([0.0789, 0.0468, 0.0548, 0.0607, 0.0808, 0.0557, 0.0452, 0.0542], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 13:05:46,221 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60101.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 13:05:55,890 INFO [train2.py:809] (1/4) Epoch 16, batch 350, loss[ctc_loss=0.09159, att_loss=0.2701, loss=0.2344, over 17144.00 frames. utt_duration=1226 frames, utt_pad_proportion=0.01378, over 56.00 utterances.], tot_loss[ctc_loss=0.08648, att_loss=0.2422, loss=0.2111, over 2705114.84 frames. utt_duration=1231 frames, utt_pad_proportion=0.05954, over 8803.51 utterances.], batch size: 56, lr: 6.86e-03, grad_scale: 16.0 2023-03-08 13:06:05,691 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60113.0, num_to_drop=1, layers_to_drop={3} 2023-03-08 13:06:26,346 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60126.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:06:30,556 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.343e+02 2.169e+02 2.724e+02 3.771e+02 9.216e+02, threshold=5.447e+02, percent-clipped=7.0 2023-03-08 13:06:47,217 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60139.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:07:14,373 INFO [train2.py:809] (1/4) Epoch 16, batch 400, loss[ctc_loss=0.08054, att_loss=0.2285, loss=0.1989, over 15634.00 frames. utt_duration=1692 frames, utt_pad_proportion=0.009384, over 37.00 utterances.], tot_loss[ctc_loss=0.08783, att_loss=0.2432, loss=0.2121, over 2838824.69 frames. utt_duration=1226 frames, utt_pad_proportion=0.05842, over 9271.53 utterances.], batch size: 37, lr: 6.86e-03, grad_scale: 8.0 2023-03-08 13:07:16,910 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60158.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:08:16,878 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5085, 2.6593, 4.9743, 3.8868, 3.0847, 4.2686, 4.9091, 4.6118], device='cuda:1'), covar=tensor([0.0233, 0.1654, 0.0191, 0.0999, 0.1818, 0.0257, 0.0118, 0.0228], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0237, 0.0157, 0.0304, 0.0259, 0.0191, 0.0139, 0.0170], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 13:08:19,871 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60198.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:08:29,149 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60204.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:08:29,294 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3367, 2.7241, 3.6049, 2.7832, 3.5035, 4.4750, 4.2294, 2.9622], device='cuda:1'), covar=tensor([0.0421, 0.2058, 0.1158, 0.1601, 0.1230, 0.0898, 0.0714, 0.1520], device='cuda:1'), in_proj_covar=tensor([0.0237, 0.0237, 0.0264, 0.0209, 0.0254, 0.0335, 0.0240, 0.0229], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 13:08:32,208 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=60206.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:08:33,629 INFO [train2.py:809] (1/4) Epoch 16, batch 450, loss[ctc_loss=0.0811, att_loss=0.2363, loss=0.2053, over 15957.00 frames. utt_duration=1558 frames, utt_pad_proportion=0.0063, over 41.00 utterances.], tot_loss[ctc_loss=0.08729, att_loss=0.242, loss=0.211, over 2924336.25 frames. utt_duration=1217 frames, utt_pad_proportion=0.06427, over 9625.83 utterances.], batch size: 41, lr: 6.86e-03, grad_scale: 8.0 2023-03-08 13:08:45,949 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0067, 5.1522, 4.9741, 2.1923, 2.0305, 2.8873, 2.7175, 3.6589], device='cuda:1'), covar=tensor([0.0763, 0.0235, 0.0229, 0.4655, 0.5694, 0.2553, 0.2762, 0.1904], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0241, 0.0246, 0.0224, 0.0340, 0.0330, 0.0231, 0.0353], device='cuda:1'), out_proj_covar=tensor([1.4803e-04, 9.0050e-05, 1.0578e-04, 9.7424e-05, 1.4427e-04, 1.3089e-04, 9.2666e-05, 1.4558e-04], device='cuda:1') 2023-03-08 13:09:10,575 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.357e+02 2.196e+02 2.695e+02 3.219e+02 5.768e+02, threshold=5.391e+02, percent-clipped=1.0 2023-03-08 13:09:34,702 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60245.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:09:54,252 INFO [train2.py:809] (1/4) Epoch 16, batch 500, loss[ctc_loss=0.07196, att_loss=0.216, loss=0.1872, over 15357.00 frames. utt_duration=1756 frames, utt_pad_proportion=0.01158, over 35.00 utterances.], tot_loss[ctc_loss=0.08675, att_loss=0.2425, loss=0.2113, over 3007443.76 frames. utt_duration=1231 frames, utt_pad_proportion=0.05882, over 9780.46 utterances.], batch size: 35, lr: 6.85e-03, grad_scale: 8.0 2023-03-08 13:09:57,632 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60259.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:10:28,548 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.9651, 3.6205, 3.6508, 3.1459, 3.7076, 3.7495, 3.6542, 2.5639], device='cuda:1'), covar=tensor([0.1024, 0.1614, 0.2728, 0.5673, 0.1234, 0.3702, 0.1178, 0.6216], device='cuda:1'), in_proj_covar=tensor([0.0130, 0.0153, 0.0163, 0.0229, 0.0127, 0.0220, 0.0139, 0.0195], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 13:11:14,955 INFO [train2.py:809] (1/4) Epoch 16, batch 550, loss[ctc_loss=0.06416, att_loss=0.2322, loss=0.1986, over 16393.00 frames. utt_duration=1492 frames, utt_pad_proportion=0.008145, over 44.00 utterances.], tot_loss[ctc_loss=0.08608, att_loss=0.2427, loss=0.2114, over 3068813.51 frames. utt_duration=1225 frames, utt_pad_proportion=0.06052, over 10035.66 utterances.], batch size: 44, lr: 6.85e-03, grad_scale: 8.0 2023-03-08 13:11:51,581 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.455e+02 1.986e+02 2.421e+02 3.099e+02 6.630e+02, threshold=4.842e+02, percent-clipped=3.0 2023-03-08 13:12:35,538 INFO [train2.py:809] (1/4) Epoch 16, batch 600, loss[ctc_loss=0.09293, att_loss=0.2519, loss=0.2201, over 17371.00 frames. utt_duration=1179 frames, utt_pad_proportion=0.01984, over 59.00 utterances.], tot_loss[ctc_loss=0.08612, att_loss=0.2426, loss=0.2113, over 3115243.07 frames. utt_duration=1240 frames, utt_pad_proportion=0.05627, over 10060.13 utterances.], batch size: 59, lr: 6.85e-03, grad_scale: 8.0 2023-03-08 13:12:55,478 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6724, 4.4798, 4.5378, 4.4526, 4.9596, 4.4968, 4.4340, 2.5335], device='cuda:1'), covar=tensor([0.0206, 0.0259, 0.0238, 0.0232, 0.1122, 0.0239, 0.0279, 0.1873], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0150, 0.0155, 0.0169, 0.0350, 0.0135, 0.0145, 0.0214], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 13:13:24,933 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6584, 3.2460, 3.9188, 3.1842, 3.7404, 4.8898, 4.6374, 3.3644], device='cuda:1'), covar=tensor([0.0407, 0.1592, 0.0970, 0.1355, 0.0971, 0.0597, 0.0555, 0.1347], device='cuda:1'), in_proj_covar=tensor([0.0235, 0.0233, 0.0262, 0.0207, 0.0250, 0.0333, 0.0239, 0.0227], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 13:13:39,229 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60396.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 13:13:57,397 INFO [train2.py:809] (1/4) Epoch 16, batch 650, loss[ctc_loss=0.1279, att_loss=0.2683, loss=0.2402, over 16938.00 frames. utt_duration=685.6 frames, utt_pad_proportion=0.1408, over 99.00 utterances.], tot_loss[ctc_loss=0.08646, att_loss=0.2429, loss=0.2116, over 3158944.50 frames. utt_duration=1236 frames, utt_pad_proportion=0.05439, over 10237.93 utterances.], batch size: 99, lr: 6.85e-03, grad_scale: 8.0 2023-03-08 13:13:59,123 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60408.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 13:14:20,215 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60421.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:14:33,815 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.518e+02 2.189e+02 2.670e+02 3.387e+02 7.027e+02, threshold=5.339e+02, percent-clipped=7.0 2023-03-08 13:14:40,853 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60434.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:14:43,908 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-03-08 13:15:17,208 INFO [train2.py:809] (1/4) Epoch 16, batch 700, loss[ctc_loss=0.08045, att_loss=0.2572, loss=0.2219, over 16626.00 frames. utt_duration=1417 frames, utt_pad_proportion=0.005244, over 47.00 utterances.], tot_loss[ctc_loss=0.08693, att_loss=0.2432, loss=0.212, over 3185616.56 frames. utt_duration=1219 frames, utt_pad_proportion=0.05928, over 10464.01 utterances.], batch size: 47, lr: 6.84e-03, grad_scale: 8.0 2023-03-08 13:15:44,255 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.8347, 2.0937, 2.3128, 2.3906, 2.4739, 2.5282, 2.4142, 2.9806], device='cuda:1'), covar=tensor([0.1390, 0.3854, 0.3451, 0.1862, 0.2124, 0.1506, 0.2762, 0.0847], device='cuda:1'), in_proj_covar=tensor([0.0089, 0.0093, 0.0098, 0.0086, 0.0091, 0.0080, 0.0100, 0.0069], device='cuda:1'), out_proj_covar=tensor([6.4720e-05, 7.0389e-05, 7.4104e-05, 6.4579e-05, 6.5191e-05, 6.2778e-05, 7.3020e-05, 5.5514e-05], device='cuda:1') 2023-03-08 13:16:01,883 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.96 vs. limit=2.0 2023-03-08 13:16:18,972 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0014, 4.9909, 4.9483, 2.0591, 1.9445, 2.6786, 2.2912, 3.6815], device='cuda:1'), covar=tensor([0.0737, 0.0251, 0.0230, 0.5289, 0.6169, 0.2873, 0.3441, 0.1903], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0242, 0.0247, 0.0224, 0.0343, 0.0334, 0.0234, 0.0355], device='cuda:1'), out_proj_covar=tensor([1.4815e-04, 9.0255e-05, 1.0594e-04, 9.7036e-05, 1.4531e-04, 1.3220e-04, 9.3595e-05, 1.4616e-04], device='cuda:1') 2023-03-08 13:16:33,124 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60504.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:16:34,745 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1757, 5.0547, 4.9429, 2.9770, 4.9147, 4.6117, 4.4513, 2.8075], device='cuda:1'), covar=tensor([0.0122, 0.0105, 0.0246, 0.1015, 0.0091, 0.0210, 0.0259, 0.1312], device='cuda:1'), in_proj_covar=tensor([0.0067, 0.0093, 0.0091, 0.0108, 0.0077, 0.0104, 0.0096, 0.0102], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 13:16:37,563 INFO [train2.py:809] (1/4) Epoch 16, batch 750, loss[ctc_loss=0.0825, att_loss=0.2389, loss=0.2076, over 16338.00 frames. utt_duration=1454 frames, utt_pad_proportion=0.005761, over 45.00 utterances.], tot_loss[ctc_loss=0.08677, att_loss=0.2431, loss=0.2118, over 3205114.38 frames. utt_duration=1220 frames, utt_pad_proportion=0.05816, over 10519.60 utterances.], batch size: 45, lr: 6.84e-03, grad_scale: 8.0 2023-03-08 13:17:14,654 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.359e+02 2.152e+02 2.643e+02 3.369e+02 1.101e+03, threshold=5.285e+02, percent-clipped=4.0 2023-03-08 13:17:40,040 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60545.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:17:50,570 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=60552.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:17:54,429 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60554.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:17:58,940 INFO [train2.py:809] (1/4) Epoch 16, batch 800, loss[ctc_loss=0.09478, att_loss=0.2166, loss=0.1923, over 15636.00 frames. utt_duration=1692 frames, utt_pad_proportion=0.009336, over 37.00 utterances.], tot_loss[ctc_loss=0.08676, att_loss=0.2431, loss=0.2119, over 3225212.50 frames. utt_duration=1236 frames, utt_pad_proportion=0.05381, over 10450.27 utterances.], batch size: 37, lr: 6.84e-03, grad_scale: 8.0 2023-03-08 13:18:56,200 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=60593.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:19:00,145 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-03-08 13:19:18,473 INFO [train2.py:809] (1/4) Epoch 16, batch 850, loss[ctc_loss=0.06698, att_loss=0.2124, loss=0.1833, over 16176.00 frames. utt_duration=1580 frames, utt_pad_proportion=0.006565, over 41.00 utterances.], tot_loss[ctc_loss=0.08673, att_loss=0.2433, loss=0.212, over 3238744.07 frames. utt_duration=1225 frames, utt_pad_proportion=0.0566, over 10591.74 utterances.], batch size: 41, lr: 6.83e-03, grad_scale: 8.0 2023-03-08 13:19:23,230 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.13 vs. limit=5.0 2023-03-08 13:19:24,869 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-03-08 13:19:55,687 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.256e+02 2.159e+02 2.689e+02 3.192e+02 6.800e+02, threshold=5.378e+02, percent-clipped=2.0 2023-03-08 13:20:39,640 INFO [train2.py:809] (1/4) Epoch 16, batch 900, loss[ctc_loss=0.06341, att_loss=0.2194, loss=0.1882, over 16556.00 frames. utt_duration=1473 frames, utt_pad_proportion=0.005237, over 45.00 utterances.], tot_loss[ctc_loss=0.0859, att_loss=0.2426, loss=0.2113, over 3250087.52 frames. utt_duration=1239 frames, utt_pad_proportion=0.05352, over 10502.87 utterances.], batch size: 45, lr: 6.83e-03, grad_scale: 8.0 2023-03-08 13:21:20,733 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-03-08 13:21:39,240 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5159, 2.2861, 4.9301, 3.8525, 2.8186, 4.1364, 4.5840, 4.6460], device='cuda:1'), covar=tensor([0.0208, 0.1787, 0.0152, 0.0910, 0.1856, 0.0275, 0.0138, 0.0217], device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0237, 0.0158, 0.0304, 0.0259, 0.0193, 0.0140, 0.0171], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 13:21:42,258 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60696.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:22:00,172 INFO [train2.py:809] (1/4) Epoch 16, batch 950, loss[ctc_loss=0.1008, att_loss=0.2623, loss=0.23, over 16461.00 frames. utt_duration=1433 frames, utt_pad_proportion=0.00684, over 46.00 utterances.], tot_loss[ctc_loss=0.08545, att_loss=0.2419, loss=0.2106, over 3251775.83 frames. utt_duration=1243 frames, utt_pad_proportion=0.0542, over 10474.83 utterances.], batch size: 46, lr: 6.83e-03, grad_scale: 8.0 2023-03-08 13:22:02,088 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60708.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:22:07,240 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.47 vs. limit=5.0 2023-03-08 13:22:18,037 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60718.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:22:22,545 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60721.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:22:36,659 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.437e+02 2.156e+02 2.670e+02 3.136e+02 6.539e+02, threshold=5.340e+02, percent-clipped=2.0 2023-03-08 13:22:43,843 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60734.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:22:59,076 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=60744.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:23:11,934 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-03-08 13:23:19,077 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=60756.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:23:20,645 INFO [train2.py:809] (1/4) Epoch 16, batch 1000, loss[ctc_loss=0.1065, att_loss=0.2619, loss=0.2309, over 17099.00 frames. utt_duration=1223 frames, utt_pad_proportion=0.01631, over 56.00 utterances.], tot_loss[ctc_loss=0.08533, att_loss=0.2418, loss=0.2105, over 3247696.72 frames. utt_duration=1227 frames, utt_pad_proportion=0.06076, over 10600.29 utterances.], batch size: 56, lr: 6.83e-03, grad_scale: 8.0 2023-03-08 13:23:39,232 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=60769.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:23:55,694 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60779.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:24:00,513 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=60782.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:24:40,431 INFO [train2.py:809] (1/4) Epoch 16, batch 1050, loss[ctc_loss=0.07748, att_loss=0.2446, loss=0.2112, over 17531.00 frames. utt_duration=1018 frames, utt_pad_proportion=0.04066, over 69.00 utterances.], tot_loss[ctc_loss=0.08511, att_loss=0.2413, loss=0.2101, over 3248278.13 frames. utt_duration=1248 frames, utt_pad_proportion=0.05645, over 10424.02 utterances.], batch size: 69, lr: 6.82e-03, grad_scale: 8.0 2023-03-08 13:25:17,107 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.408e+02 2.141e+02 2.597e+02 3.263e+02 1.129e+03, threshold=5.194e+02, percent-clipped=2.0 2023-03-08 13:25:56,055 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60854.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:25:59,248 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.8595, 1.9705, 2.4583, 2.6967, 2.8160, 2.6364, 2.2691, 3.1088], device='cuda:1'), covar=tensor([0.3456, 0.4643, 0.3342, 0.1912, 0.2075, 0.2003, 0.3332, 0.1256], device='cuda:1'), in_proj_covar=tensor([0.0091, 0.0094, 0.0098, 0.0086, 0.0090, 0.0082, 0.0101, 0.0070], device='cuda:1'), out_proj_covar=tensor([6.5771e-05, 7.1292e-05, 7.4349e-05, 6.4666e-05, 6.5193e-05, 6.3640e-05, 7.3614e-05, 5.5917e-05], device='cuda:1') 2023-03-08 13:26:00,466 INFO [train2.py:809] (1/4) Epoch 16, batch 1100, loss[ctc_loss=0.08301, att_loss=0.2216, loss=0.1939, over 16291.00 frames. utt_duration=1517 frames, utt_pad_proportion=0.006594, over 43.00 utterances.], tot_loss[ctc_loss=0.08477, att_loss=0.2406, loss=0.2094, over 3253598.02 frames. utt_duration=1255 frames, utt_pad_proportion=0.05443, over 10386.28 utterances.], batch size: 43, lr: 6.82e-03, grad_scale: 8.0 2023-03-08 13:26:41,183 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2027, 2.4632, 3.1008, 4.1567, 3.7229, 3.6644, 2.7433, 2.0105], device='cuda:1'), covar=tensor([0.0780, 0.2451, 0.1065, 0.0625, 0.0827, 0.0579, 0.1589, 0.2537], device='cuda:1'), in_proj_covar=tensor([0.0176, 0.0214, 0.0191, 0.0202, 0.0211, 0.0168, 0.0197, 0.0181], 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') 2023-03-08 13:26:47,594 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5430, 2.7420, 5.0963, 4.1239, 3.1164, 4.4508, 4.9847, 4.6815], device='cuda:1'), covar=tensor([0.0290, 0.1478, 0.0223, 0.0883, 0.1630, 0.0232, 0.0120, 0.0255], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0235, 0.0158, 0.0302, 0.0257, 0.0191, 0.0139, 0.0169], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 13:27:12,923 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=60902.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:27:21,215 INFO [train2.py:809] (1/4) Epoch 16, batch 1150, loss[ctc_loss=0.0846, att_loss=0.2388, loss=0.208, over 16533.00 frames. utt_duration=1471 frames, utt_pad_proportion=0.006632, over 45.00 utterances.], tot_loss[ctc_loss=0.08519, att_loss=0.2413, loss=0.2101, over 3268578.70 frames. utt_duration=1252 frames, utt_pad_proportion=0.05157, over 10451.25 utterances.], batch size: 45, lr: 6.82e-03, grad_scale: 8.0 2023-03-08 13:27:28,089 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1879, 5.0738, 5.0562, 2.4261, 2.0316, 2.9880, 2.5667, 3.9111], device='cuda:1'), covar=tensor([0.0728, 0.0307, 0.0240, 0.5250, 0.5923, 0.2635, 0.3367, 0.1824], device='cuda:1'), in_proj_covar=tensor([0.0341, 0.0243, 0.0246, 0.0226, 0.0341, 0.0334, 0.0235, 0.0355], device='cuda:1'), out_proj_covar=tensor([1.4888e-04, 9.0526e-05, 1.0566e-04, 9.8178e-05, 1.4491e-04, 1.3221e-04, 9.4251e-05, 1.4618e-04], device='cuda:1') 2023-03-08 13:27:59,219 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.119e+02 2.021e+02 2.572e+02 3.109e+02 6.628e+02, threshold=5.143e+02, percent-clipped=1.0 2023-03-08 13:28:01,079 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9508, 5.2455, 5.4268, 5.3133, 5.3644, 5.8827, 5.1802, 6.0094], device='cuda:1'), covar=tensor([0.0678, 0.0677, 0.0804, 0.1143, 0.1653, 0.0896, 0.0686, 0.0589], device='cuda:1'), in_proj_covar=tensor([0.0798, 0.0471, 0.0556, 0.0609, 0.0811, 0.0565, 0.0453, 0.0536], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 13:28:12,337 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.6366, 3.7377, 3.9572, 2.4853, 2.3759, 2.8281, 2.3859, 3.4744], device='cuda:1'), covar=tensor([0.0726, 0.0371, 0.0310, 0.3522, 0.4110, 0.2121, 0.2531, 0.1430], device='cuda:1'), in_proj_covar=tensor([0.0345, 0.0245, 0.0249, 0.0229, 0.0346, 0.0338, 0.0238, 0.0359], device='cuda:1'), out_proj_covar=tensor([1.5053e-04, 9.1526e-05, 1.0698e-04, 9.9750e-05, 1.4678e-04, 1.3385e-04, 9.5364e-05, 1.4789e-04], device='cuda:1') 2023-03-08 13:28:42,081 INFO [train2.py:809] (1/4) Epoch 16, batch 1200, loss[ctc_loss=0.08362, att_loss=0.2442, loss=0.2121, over 16635.00 frames. utt_duration=1417 frames, utt_pad_proportion=0.004766, over 47.00 utterances.], tot_loss[ctc_loss=0.08488, att_loss=0.2411, loss=0.2098, over 3265282.96 frames. utt_duration=1245 frames, utt_pad_proportion=0.05552, over 10504.31 utterances.], batch size: 47, lr: 6.81e-03, grad_scale: 8.0 2023-03-08 13:29:34,934 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.0593, 3.5998, 3.6026, 3.1174, 3.6791, 3.7878, 3.6391, 2.8243], device='cuda:1'), covar=tensor([0.0987, 0.1782, 0.2116, 0.5169, 0.2824, 0.2705, 0.1035, 0.5115], device='cuda:1'), in_proj_covar=tensor([0.0133, 0.0154, 0.0166, 0.0231, 0.0129, 0.0223, 0.0141, 0.0198], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 13:30:02,899 INFO [train2.py:809] (1/4) Epoch 16, batch 1250, loss[ctc_loss=0.09332, att_loss=0.2417, loss=0.212, over 16274.00 frames. utt_duration=1516 frames, utt_pad_proportion=0.007523, over 43.00 utterances.], tot_loss[ctc_loss=0.0844, att_loss=0.2407, loss=0.2094, over 3264647.09 frames. utt_duration=1253 frames, utt_pad_proportion=0.0533, over 10432.17 utterances.], batch size: 43, lr: 6.81e-03, grad_scale: 8.0 2023-03-08 13:30:39,966 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.300e+02 1.974e+02 2.401e+02 2.960e+02 6.292e+02, threshold=4.803e+02, percent-clipped=4.0 2023-03-08 13:31:22,899 INFO [train2.py:809] (1/4) Epoch 16, batch 1300, loss[ctc_loss=0.1496, att_loss=0.277, loss=0.2515, over 13558.00 frames. utt_duration=372.9 frames, utt_pad_proportion=0.3503, over 146.00 utterances.], tot_loss[ctc_loss=0.08541, att_loss=0.2415, loss=0.2103, over 3264244.47 frames. utt_duration=1246 frames, utt_pad_proportion=0.05497, over 10488.91 utterances.], batch size: 146, lr: 6.81e-03, grad_scale: 8.0 2023-03-08 13:31:29,575 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61061.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:31:50,215 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=61074.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:32:41,875 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-03-08 13:32:42,648 INFO [train2.py:809] (1/4) Epoch 16, batch 1350, loss[ctc_loss=0.1051, att_loss=0.2664, loss=0.2341, over 17308.00 frames. utt_duration=1175 frames, utt_pad_proportion=0.02263, over 59.00 utterances.], tot_loss[ctc_loss=0.08581, att_loss=0.2415, loss=0.2103, over 3262004.44 frames. utt_duration=1228 frames, utt_pad_proportion=0.06069, over 10642.04 utterances.], batch size: 59, lr: 6.81e-03, grad_scale: 8.0 2023-03-08 13:33:06,276 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61122.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:33:18,943 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.675e+02 2.149e+02 2.520e+02 3.082e+02 5.998e+02, threshold=5.039e+02, percent-clipped=2.0 2023-03-08 13:34:02,072 INFO [train2.py:809] (1/4) Epoch 16, batch 1400, loss[ctc_loss=0.08493, att_loss=0.2354, loss=0.2053, over 16319.00 frames. utt_duration=1452 frames, utt_pad_proportion=0.006749, over 45.00 utterances.], tot_loss[ctc_loss=0.08511, att_loss=0.2406, loss=0.2095, over 3258946.13 frames. utt_duration=1241 frames, utt_pad_proportion=0.05793, over 10515.59 utterances.], batch size: 45, lr: 6.80e-03, grad_scale: 8.0 2023-03-08 13:34:43,272 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.1609, 3.7788, 3.1591, 3.6118, 3.9759, 3.6934, 3.1797, 4.3213], device='cuda:1'), covar=tensor([0.0917, 0.0437, 0.1043, 0.0552, 0.0588, 0.0677, 0.0741, 0.0455], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0200, 0.0216, 0.0187, 0.0259, 0.0226, 0.0192, 0.0277], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-08 13:35:21,832 INFO [train2.py:809] (1/4) Epoch 16, batch 1450, loss[ctc_loss=0.07368, att_loss=0.227, loss=0.1963, over 14157.00 frames. utt_duration=1828 frames, utt_pad_proportion=0.04724, over 31.00 utterances.], tot_loss[ctc_loss=0.08534, att_loss=0.2409, loss=0.2098, over 3262245.90 frames. utt_duration=1239 frames, utt_pad_proportion=0.05818, over 10543.17 utterances.], batch size: 31, lr: 6.80e-03, grad_scale: 8.0 2023-03-08 13:35:57,927 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.42 vs. limit=5.0 2023-03-08 13:35:58,473 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.469e+02 2.227e+02 2.591e+02 3.129e+02 8.266e+02, threshold=5.182e+02, percent-clipped=3.0 2023-03-08 13:36:38,244 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.5222, 3.8274, 3.4137, 3.7051, 4.1154, 3.8378, 3.4493, 4.5272], device='cuda:1'), covar=tensor([0.0830, 0.0534, 0.1026, 0.0642, 0.0696, 0.0675, 0.0706, 0.0420], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0200, 0.0216, 0.0188, 0.0261, 0.0228, 0.0192, 0.0277], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-08 13:36:39,617 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9629, 5.3705, 4.7548, 5.4436, 4.7780, 5.0599, 5.4572, 5.2251], device='cuda:1'), covar=tensor([0.0668, 0.0282, 0.0919, 0.0264, 0.0409, 0.0242, 0.0251, 0.0233], device='cuda:1'), in_proj_covar=tensor([0.0371, 0.0293, 0.0347, 0.0307, 0.0302, 0.0226, 0.0279, 0.0264], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 13:36:40,850 INFO [train2.py:809] (1/4) Epoch 16, batch 1500, loss[ctc_loss=0.06807, att_loss=0.2199, loss=0.1895, over 15993.00 frames. utt_duration=1601 frames, utt_pad_proportion=0.008147, over 40.00 utterances.], tot_loss[ctc_loss=0.08542, att_loss=0.2407, loss=0.2097, over 3269565.44 frames. utt_duration=1258 frames, utt_pad_proportion=0.05157, over 10408.79 utterances.], batch size: 40, lr: 6.80e-03, grad_scale: 8.0 2023-03-08 13:36:56,455 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7235, 2.9447, 3.8036, 3.2802, 3.6340, 4.7445, 4.4738, 3.3692], device='cuda:1'), covar=tensor([0.0321, 0.1846, 0.1097, 0.1196, 0.1081, 0.0841, 0.0637, 0.1264], device='cuda:1'), in_proj_covar=tensor([0.0240, 0.0238, 0.0266, 0.0211, 0.0254, 0.0339, 0.0242, 0.0230], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 13:37:34,594 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-03-08 13:37:48,542 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6082, 2.4468, 5.1124, 3.9096, 3.0131, 4.3357, 4.8560, 4.6217], device='cuda:1'), covar=tensor([0.0200, 0.1718, 0.0187, 0.1003, 0.1794, 0.0233, 0.0116, 0.0216], device='cuda:1'), in_proj_covar=tensor([0.0169, 0.0242, 0.0162, 0.0311, 0.0264, 0.0197, 0.0143, 0.0173], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 13:37:59,314 INFO [train2.py:809] (1/4) Epoch 16, batch 1550, loss[ctc_loss=0.08737, att_loss=0.2419, loss=0.211, over 16401.00 frames. utt_duration=1493 frames, utt_pad_proportion=0.006744, over 44.00 utterances.], tot_loss[ctc_loss=0.08554, att_loss=0.2411, loss=0.21, over 3276522.28 frames. utt_duration=1259 frames, utt_pad_proportion=0.04974, over 10424.30 utterances.], batch size: 44, lr: 6.80e-03, grad_scale: 8.0 2023-03-08 13:38:07,656 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0466, 4.3840, 4.3501, 4.8112, 2.6425, 4.3329, 2.6675, 1.6808], device='cuda:1'), covar=tensor([0.0474, 0.0277, 0.0767, 0.0138, 0.1797, 0.0194, 0.1548, 0.1812], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0138, 0.0255, 0.0130, 0.0220, 0.0122, 0.0228, 0.0199], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 13:38:13,900 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2423, 5.1297, 4.9787, 2.5484, 2.0747, 3.1101, 2.3858, 3.8935], device='cuda:1'), covar=tensor([0.0621, 0.0310, 0.0321, 0.4536, 0.5702, 0.2252, 0.3469, 0.1674], device='cuda:1'), in_proj_covar=tensor([0.0342, 0.0242, 0.0247, 0.0226, 0.0340, 0.0334, 0.0235, 0.0354], device='cuda:1'), out_proj_covar=tensor([1.4900e-04, 9.0426e-05, 1.0640e-04, 9.7688e-05, 1.4443e-04, 1.3220e-04, 9.4251e-05, 1.4609e-04], device='cuda:1') 2023-03-08 13:38:35,928 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.223e+02 2.094e+02 2.601e+02 3.579e+02 7.093e+02, threshold=5.201e+02, percent-clipped=2.0 2023-03-08 13:39:18,916 INFO [train2.py:809] (1/4) Epoch 16, batch 1600, loss[ctc_loss=0.07622, att_loss=0.2518, loss=0.2167, over 16969.00 frames. utt_duration=1359 frames, utt_pad_proportion=0.006433, over 50.00 utterances.], tot_loss[ctc_loss=0.0858, att_loss=0.2415, loss=0.2104, over 3274767.38 frames. utt_duration=1226 frames, utt_pad_proportion=0.05905, over 10694.38 utterances.], batch size: 50, lr: 6.79e-03, grad_scale: 8.0 2023-03-08 13:39:46,385 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=61374.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:40:38,376 INFO [train2.py:809] (1/4) Epoch 16, batch 1650, loss[ctc_loss=0.06649, att_loss=0.2201, loss=0.1894, over 15882.00 frames. utt_duration=1631 frames, utt_pad_proportion=0.0093, over 39.00 utterances.], tot_loss[ctc_loss=0.08553, att_loss=0.2417, loss=0.2105, over 3271427.70 frames. utt_duration=1225 frames, utt_pad_proportion=0.06009, over 10697.93 utterances.], batch size: 39, lr: 6.79e-03, grad_scale: 8.0 2023-03-08 13:40:54,188 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=61417.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:41:02,482 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=61422.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:41:14,484 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.484e+02 2.105e+02 2.549e+02 2.944e+02 7.207e+02, threshold=5.098e+02, percent-clipped=3.0 2023-03-08 13:41:56,931 INFO [train2.py:809] (1/4) Epoch 16, batch 1700, loss[ctc_loss=0.09466, att_loss=0.2498, loss=0.2187, over 17296.00 frames. utt_duration=1259 frames, utt_pad_proportion=0.01218, over 55.00 utterances.], tot_loss[ctc_loss=0.0856, att_loss=0.2414, loss=0.2102, over 3272312.70 frames. utt_duration=1254 frames, utt_pad_proportion=0.05397, over 10448.39 utterances.], batch size: 55, lr: 6.79e-03, grad_scale: 8.0 2023-03-08 13:42:49,260 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7446, 3.6310, 2.9953, 3.2247, 3.8114, 3.4571, 2.4890, 3.9866], device='cuda:1'), covar=tensor([0.1032, 0.0413, 0.1068, 0.0672, 0.0605, 0.0662, 0.1081, 0.0446], device='cuda:1'), in_proj_covar=tensor([0.0193, 0.0199, 0.0216, 0.0187, 0.0258, 0.0224, 0.0192, 0.0274], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-08 13:42:56,969 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5904, 3.0304, 3.8095, 3.1947, 3.6805, 4.7006, 4.4374, 3.2699], device='cuda:1'), covar=tensor([0.0360, 0.1674, 0.1011, 0.1296, 0.0989, 0.0789, 0.0646, 0.1281], device='cuda:1'), in_proj_covar=tensor([0.0240, 0.0238, 0.0266, 0.0210, 0.0255, 0.0339, 0.0242, 0.0231], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 13:43:16,828 INFO [train2.py:809] (1/4) Epoch 16, batch 1750, loss[ctc_loss=0.06604, att_loss=0.2376, loss=0.2033, over 16624.00 frames. utt_duration=1417 frames, utt_pad_proportion=0.005229, over 47.00 utterances.], tot_loss[ctc_loss=0.08534, att_loss=0.2414, loss=0.2102, over 3278212.52 frames. utt_duration=1255 frames, utt_pad_proportion=0.05165, over 10463.24 utterances.], batch size: 47, lr: 6.78e-03, grad_scale: 8.0 2023-03-08 13:43:53,593 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.239e+02 1.998e+02 2.470e+02 3.033e+02 6.251e+02, threshold=4.940e+02, percent-clipped=1.0 2023-03-08 13:44:07,601 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.7129, 5.9585, 5.3994, 5.7547, 5.5938, 5.1552, 5.3466, 5.1333], device='cuda:1'), covar=tensor([0.1223, 0.0931, 0.0868, 0.0822, 0.0918, 0.1532, 0.2230, 0.2370], device='cuda:1'), in_proj_covar=tensor([0.0487, 0.0555, 0.0423, 0.0427, 0.0405, 0.0442, 0.0574, 0.0503], 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') 2023-03-08 13:44:36,284 INFO [train2.py:809] (1/4) Epoch 16, batch 1800, loss[ctc_loss=0.0954, att_loss=0.261, loss=0.2279, over 16546.00 frames. utt_duration=1472 frames, utt_pad_proportion=0.004447, over 45.00 utterances.], tot_loss[ctc_loss=0.08516, att_loss=0.2418, loss=0.2105, over 3279606.83 frames. utt_duration=1285 frames, utt_pad_proportion=0.04412, over 10219.61 utterances.], batch size: 45, lr: 6.78e-03, grad_scale: 8.0 2023-03-08 13:44:41,707 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-03-08 13:45:55,242 INFO [train2.py:809] (1/4) Epoch 16, batch 1850, loss[ctc_loss=0.06536, att_loss=0.2124, loss=0.183, over 14504.00 frames. utt_duration=1815 frames, utt_pad_proportion=0.04841, over 32.00 utterances.], tot_loss[ctc_loss=0.08552, att_loss=0.2421, loss=0.2108, over 3280667.19 frames. utt_duration=1266 frames, utt_pad_proportion=0.04775, over 10374.26 utterances.], batch size: 32, lr: 6.78e-03, grad_scale: 8.0 2023-03-08 13:46:31,938 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.537e+02 2.233e+02 2.629e+02 3.133e+02 5.185e+02, threshold=5.258e+02, percent-clipped=1.0 2023-03-08 13:47:14,834 INFO [train2.py:809] (1/4) Epoch 16, batch 1900, loss[ctc_loss=0.08839, att_loss=0.2354, loss=0.206, over 16411.00 frames. utt_duration=1493 frames, utt_pad_proportion=0.006457, over 44.00 utterances.], tot_loss[ctc_loss=0.0849, att_loss=0.2412, loss=0.21, over 3283373.87 frames. utt_duration=1283 frames, utt_pad_proportion=0.04315, over 10247.35 utterances.], batch size: 44, lr: 6.78e-03, grad_scale: 8.0 2023-03-08 13:48:33,457 INFO [train2.py:809] (1/4) Epoch 16, batch 1950, loss[ctc_loss=0.08017, att_loss=0.2502, loss=0.2162, over 17391.00 frames. utt_duration=1106 frames, utt_pad_proportion=0.03342, over 63.00 utterances.], tot_loss[ctc_loss=0.08501, att_loss=0.2409, loss=0.2097, over 3279370.76 frames. utt_duration=1257 frames, utt_pad_proportion=0.05073, over 10450.45 utterances.], batch size: 63, lr: 6.77e-03, grad_scale: 8.0 2023-03-08 13:48:49,368 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=61717.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:49:09,854 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.327e+02 2.102e+02 2.537e+02 2.891e+02 6.099e+02, threshold=5.073e+02, percent-clipped=2.0 2023-03-08 13:49:13,259 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61732.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:49:30,476 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61743.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:49:40,476 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7016, 2.8989, 3.7296, 3.4098, 3.7016, 4.8173, 4.5923, 3.3932], device='cuda:1'), covar=tensor([0.0351, 0.1869, 0.1257, 0.1167, 0.1078, 0.0734, 0.0497, 0.1371], device='cuda:1'), in_proj_covar=tensor([0.0239, 0.0236, 0.0265, 0.0208, 0.0252, 0.0338, 0.0240, 0.0231], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 13:49:53,041 INFO [train2.py:809] (1/4) Epoch 16, batch 2000, loss[ctc_loss=0.07248, att_loss=0.2349, loss=0.2024, over 16536.00 frames. utt_duration=1471 frames, utt_pad_proportion=0.006587, over 45.00 utterances.], tot_loss[ctc_loss=0.0848, att_loss=0.2411, loss=0.2098, over 3275954.69 frames. utt_duration=1249 frames, utt_pad_proportion=0.05326, over 10501.32 utterances.], batch size: 45, lr: 6.77e-03, grad_scale: 8.0 2023-03-08 13:50:05,405 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=61765.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:50:05,768 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6076, 2.7993, 5.2221, 4.2010, 3.0181, 4.4991, 4.9565, 4.5979], device='cuda:1'), covar=tensor([0.0278, 0.1569, 0.0186, 0.0810, 0.1801, 0.0224, 0.0138, 0.0273], device='cuda:1'), in_proj_covar=tensor([0.0170, 0.0244, 0.0162, 0.0311, 0.0265, 0.0196, 0.0143, 0.0173], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 13:50:49,908 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61793.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:51:08,238 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61804.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:51:12,561 INFO [train2.py:809] (1/4) Epoch 16, batch 2050, loss[ctc_loss=0.08699, att_loss=0.2514, loss=0.2185, over 16764.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.006614, over 48.00 utterances.], tot_loss[ctc_loss=0.08509, att_loss=0.2412, loss=0.2099, over 3274272.12 frames. utt_duration=1250 frames, utt_pad_proportion=0.05378, over 10489.43 utterances.], batch size: 48, lr: 6.77e-03, grad_scale: 8.0 2023-03-08 13:51:48,692 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.602e+02 2.229e+02 2.633e+02 3.391e+02 8.027e+02, threshold=5.266e+02, percent-clipped=7.0 2023-03-08 13:52:32,059 INFO [train2.py:809] (1/4) Epoch 16, batch 2100, loss[ctc_loss=0.07957, att_loss=0.2479, loss=0.2142, over 16873.00 frames. utt_duration=1379 frames, utt_pad_proportion=0.007796, over 49.00 utterances.], tot_loss[ctc_loss=0.08548, att_loss=0.2412, loss=0.21, over 3261417.10 frames. utt_duration=1218 frames, utt_pad_proportion=0.06657, over 10726.15 utterances.], batch size: 49, lr: 6.77e-03, grad_scale: 8.0 2023-03-08 13:53:03,356 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.18 vs. limit=5.0 2023-03-08 13:53:11,653 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4312, 5.2608, 5.2854, 5.3267, 5.1699, 5.3347, 5.0911, 4.7924], device='cuda:1'), covar=tensor([0.1721, 0.0668, 0.0336, 0.0498, 0.0660, 0.0383, 0.0371, 0.0406], device='cuda:1'), in_proj_covar=tensor([0.0489, 0.0328, 0.0300, 0.0323, 0.0381, 0.0399, 0.0324, 0.0361], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 13:53:49,913 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0996, 4.4348, 4.3135, 4.3437, 4.4026, 4.1788, 3.1479, 4.4112], device='cuda:1'), covar=tensor([0.0139, 0.0118, 0.0147, 0.0103, 0.0129, 0.0127, 0.0678, 0.0221], device='cuda:1'), in_proj_covar=tensor([0.0084, 0.0080, 0.0099, 0.0063, 0.0068, 0.0080, 0.0098, 0.0101], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 13:53:51,105 INFO [train2.py:809] (1/4) Epoch 16, batch 2150, loss[ctc_loss=0.1058, att_loss=0.2531, loss=0.2237, over 16569.00 frames. utt_duration=671.1 frames, utt_pad_proportion=0.1569, over 99.00 utterances.], tot_loss[ctc_loss=0.08548, att_loss=0.2412, loss=0.21, over 3265364.71 frames. utt_duration=1225 frames, utt_pad_proportion=0.06313, over 10676.61 utterances.], batch size: 99, lr: 6.76e-03, grad_scale: 8.0 2023-03-08 13:54:03,287 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61914.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:54:27,521 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.318e+02 2.191e+02 2.731e+02 3.267e+02 7.309e+02, threshold=5.461e+02, percent-clipped=1.0 2023-03-08 13:54:41,509 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-03-08 13:55:11,211 INFO [train2.py:809] (1/4) Epoch 16, batch 2200, loss[ctc_loss=0.09916, att_loss=0.2529, loss=0.2222, over 16339.00 frames. utt_duration=1454 frames, utt_pad_proportion=0.0057, over 45.00 utterances.], tot_loss[ctc_loss=0.08502, att_loss=0.2406, loss=0.2095, over 3266248.49 frames. utt_duration=1250 frames, utt_pad_proportion=0.05768, over 10467.44 utterances.], batch size: 45, lr: 6.76e-03, grad_scale: 8.0 2023-03-08 13:55:40,164 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61975.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:56:35,307 INFO [train2.py:809] (1/4) Epoch 16, batch 2250, loss[ctc_loss=0.0973, att_loss=0.2589, loss=0.2266, over 17121.00 frames. utt_duration=1225 frames, utt_pad_proportion=0.01474, over 56.00 utterances.], tot_loss[ctc_loss=0.08473, att_loss=0.2409, loss=0.2096, over 3278864.50 frames. utt_duration=1264 frames, utt_pad_proportion=0.05065, over 10385.98 utterances.], batch size: 56, lr: 6.76e-03, grad_scale: 8.0 2023-03-08 13:57:11,172 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.510e+02 2.098e+02 2.625e+02 3.264e+02 7.065e+02, threshold=5.250e+02, percent-clipped=5.0 2023-03-08 13:57:26,555 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-03-08 13:57:54,974 INFO [train2.py:809] (1/4) Epoch 16, batch 2300, loss[ctc_loss=0.09331, att_loss=0.2347, loss=0.2064, over 15894.00 frames. utt_duration=1632 frames, utt_pad_proportion=0.008646, over 39.00 utterances.], tot_loss[ctc_loss=0.08427, att_loss=0.2409, loss=0.2096, over 3271618.50 frames. utt_duration=1282 frames, utt_pad_proportion=0.04754, over 10221.81 utterances.], batch size: 39, lr: 6.75e-03, grad_scale: 8.0 2023-03-08 13:58:43,780 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62088.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:59:02,061 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62099.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:59:09,924 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6663, 2.4602, 5.1541, 3.8937, 3.0181, 4.3536, 4.9030, 4.7018], device='cuda:1'), covar=tensor([0.0230, 0.1958, 0.0172, 0.1019, 0.1907, 0.0264, 0.0125, 0.0233], device='cuda:1'), in_proj_covar=tensor([0.0169, 0.0243, 0.0162, 0.0311, 0.0264, 0.0195, 0.0143, 0.0171], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 13:59:14,699 INFO [train2.py:809] (1/4) Epoch 16, batch 2350, loss[ctc_loss=0.1135, att_loss=0.2653, loss=0.235, over 17292.00 frames. utt_duration=1099 frames, utt_pad_proportion=0.0392, over 63.00 utterances.], tot_loss[ctc_loss=0.08466, att_loss=0.2409, loss=0.2096, over 3273466.32 frames. utt_duration=1290 frames, utt_pad_proportion=0.04413, over 10161.60 utterances.], batch size: 63, lr: 6.75e-03, grad_scale: 8.0 2023-03-08 13:59:38,205 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62122.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 13:59:49,925 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.396e+02 2.132e+02 2.401e+02 3.401e+02 9.519e+02, threshold=4.802e+02, percent-clipped=3.0 2023-03-08 14:00:06,563 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4405, 4.3786, 4.4435, 4.3713, 4.8995, 4.4294, 4.2753, 2.3641], device='cuda:1'), covar=tensor([0.0232, 0.0325, 0.0304, 0.0273, 0.0979, 0.0225, 0.0284, 0.2113], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0154, 0.0158, 0.0169, 0.0350, 0.0135, 0.0146, 0.0213], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 14:00:18,565 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-03-08 14:00:34,128 INFO [train2.py:809] (1/4) Epoch 16, batch 2400, loss[ctc_loss=0.06381, att_loss=0.2389, loss=0.2039, over 17427.00 frames. utt_duration=883.9 frames, utt_pad_proportion=0.07449, over 79.00 utterances.], tot_loss[ctc_loss=0.08403, att_loss=0.2403, loss=0.209, over 3269897.58 frames. utt_duration=1280 frames, utt_pad_proportion=0.04744, over 10227.53 utterances.], batch size: 79, lr: 6.75e-03, grad_scale: 16.0 2023-03-08 14:00:53,834 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-03-08 14:01:15,283 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62183.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:01:28,436 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.69 vs. limit=2.0 2023-03-08 14:01:53,486 INFO [train2.py:809] (1/4) Epoch 16, batch 2450, loss[ctc_loss=0.1373, att_loss=0.2743, loss=0.2469, over 14400.00 frames. utt_duration=396.1 frames, utt_pad_proportion=0.3112, over 146.00 utterances.], tot_loss[ctc_loss=0.08449, att_loss=0.241, loss=0.2097, over 3275699.23 frames. utt_duration=1270 frames, utt_pad_proportion=0.04865, over 10326.03 utterances.], batch size: 146, lr: 6.75e-03, grad_scale: 16.0 2023-03-08 14:02:29,359 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.309e+02 2.234e+02 2.691e+02 3.084e+02 9.437e+02, threshold=5.381e+02, percent-clipped=5.0 2023-03-08 14:02:35,482 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-03-08 14:03:13,139 INFO [train2.py:809] (1/4) Epoch 16, batch 2500, loss[ctc_loss=0.07587, att_loss=0.2146, loss=0.1869, over 14116.00 frames. utt_duration=1823 frames, utt_pad_proportion=0.05544, over 31.00 utterances.], tot_loss[ctc_loss=0.08326, att_loss=0.2403, loss=0.2089, over 3274517.83 frames. utt_duration=1296 frames, utt_pad_proportion=0.04278, over 10121.39 utterances.], batch size: 31, lr: 6.74e-03, grad_scale: 16.0 2023-03-08 14:03:33,440 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62270.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:03:38,174 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62273.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:04:32,512 INFO [train2.py:809] (1/4) Epoch 16, batch 2550, loss[ctc_loss=0.07757, att_loss=0.2242, loss=0.1949, over 15628.00 frames. utt_duration=1691 frames, utt_pad_proportion=0.009262, over 37.00 utterances.], tot_loss[ctc_loss=0.08323, att_loss=0.2398, loss=0.2085, over 3276067.69 frames. utt_duration=1285 frames, utt_pad_proportion=0.04459, over 10207.61 utterances.], batch size: 37, lr: 6.74e-03, grad_scale: 8.0 2023-03-08 14:04:47,707 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-03-08 14:05:09,995 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.185e+02 2.191e+02 2.595e+02 3.145e+02 7.901e+02, threshold=5.189e+02, percent-clipped=4.0 2023-03-08 14:05:15,025 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1586, 5.3996, 5.4140, 5.3229, 5.4788, 5.4500, 5.1614, 4.8915], device='cuda:1'), covar=tensor([0.0957, 0.0522, 0.0226, 0.0436, 0.0261, 0.0287, 0.0292, 0.0338], device='cuda:1'), in_proj_covar=tensor([0.0482, 0.0323, 0.0296, 0.0318, 0.0375, 0.0393, 0.0320, 0.0358], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 14:05:15,192 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62334.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:05:52,042 INFO [train2.py:809] (1/4) Epoch 16, batch 2600, loss[ctc_loss=0.07322, att_loss=0.2159, loss=0.1874, over 15505.00 frames. utt_duration=1724 frames, utt_pad_proportion=0.00853, over 36.00 utterances.], tot_loss[ctc_loss=0.08396, att_loss=0.2404, loss=0.2091, over 3270496.50 frames. utt_duration=1254 frames, utt_pad_proportion=0.05316, over 10444.43 utterances.], batch size: 36, lr: 6.74e-03, grad_scale: 8.0 2023-03-08 14:06:41,221 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62388.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:06:45,796 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62391.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:06:58,812 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62399.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:07:11,495 INFO [train2.py:809] (1/4) Epoch 16, batch 2650, loss[ctc_loss=0.08496, att_loss=0.2406, loss=0.2095, over 16531.00 frames. utt_duration=1471 frames, utt_pad_proportion=0.006156, over 45.00 utterances.], tot_loss[ctc_loss=0.08408, att_loss=0.2408, loss=0.2095, over 3279185.49 frames. utt_duration=1241 frames, utt_pad_proportion=0.05402, over 10581.67 utterances.], batch size: 45, lr: 6.74e-03, grad_scale: 8.0 2023-03-08 14:07:49,382 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.560e+02 2.130e+02 2.553e+02 3.253e+02 7.281e+02, threshold=5.105e+02, percent-clipped=6.0 2023-03-08 14:07:57,911 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=62436.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:08:15,310 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=62447.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:08:20,159 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-03-08 14:08:24,082 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62452.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:08:31,465 INFO [train2.py:809] (1/4) Epoch 16, batch 2700, loss[ctc_loss=0.1314, att_loss=0.2682, loss=0.2408, over 14213.00 frames. utt_duration=391 frames, utt_pad_proportion=0.3188, over 146.00 utterances.], tot_loss[ctc_loss=0.08481, att_loss=0.2423, loss=0.2108, over 3287019.13 frames. utt_duration=1224 frames, utt_pad_proportion=0.05721, over 10758.19 utterances.], batch size: 146, lr: 6.73e-03, grad_scale: 8.0 2023-03-08 14:09:05,212 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62478.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:09:51,787 INFO [train2.py:809] (1/4) Epoch 16, batch 2750, loss[ctc_loss=0.07203, att_loss=0.2093, loss=0.1818, over 15356.00 frames. utt_duration=1756 frames, utt_pad_proportion=0.01115, over 35.00 utterances.], tot_loss[ctc_loss=0.08565, att_loss=0.2421, loss=0.2108, over 3268646.38 frames. utt_duration=1200 frames, utt_pad_proportion=0.06722, over 10913.65 utterances.], batch size: 35, lr: 6.73e-03, grad_scale: 8.0 2023-03-08 14:10:29,272 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.352e+02 2.080e+02 2.636e+02 3.250e+02 8.951e+02, threshold=5.273e+02, percent-clipped=4.0 2023-03-08 14:10:52,874 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62545.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:11:10,991 INFO [train2.py:809] (1/4) Epoch 16, batch 2800, loss[ctc_loss=0.09531, att_loss=0.257, loss=0.2247, over 17068.00 frames. utt_duration=865.7 frames, utt_pad_proportion=0.08877, over 79.00 utterances.], tot_loss[ctc_loss=0.08627, att_loss=0.2428, loss=0.2115, over 3273085.27 frames. utt_duration=1178 frames, utt_pad_proportion=0.07134, over 11124.98 utterances.], batch size: 79, lr: 6.73e-03, grad_scale: 8.0 2023-03-08 14:11:31,247 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62570.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:11:36,428 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3172, 4.7263, 4.5479, 4.7081, 4.8549, 4.5094, 3.1580, 4.6274], device='cuda:1'), covar=tensor([0.0126, 0.0117, 0.0161, 0.0110, 0.0110, 0.0110, 0.0817, 0.0217], device='cuda:1'), in_proj_covar=tensor([0.0084, 0.0081, 0.0100, 0.0063, 0.0068, 0.0080, 0.0098, 0.0101], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 14:11:55,459 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62585.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:12:29,617 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62606.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:12:30,827 INFO [train2.py:809] (1/4) Epoch 16, batch 2850, loss[ctc_loss=0.07544, att_loss=0.2363, loss=0.2041, over 16525.00 frames. utt_duration=1471 frames, utt_pad_proportion=0.006396, over 45.00 utterances.], tot_loss[ctc_loss=0.08535, att_loss=0.2418, loss=0.2105, over 3275175.31 frames. utt_duration=1198 frames, utt_pad_proportion=0.06592, over 10946.87 utterances.], batch size: 45, lr: 6.72e-03, grad_scale: 8.0 2023-03-08 14:12:48,227 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=62618.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:13:05,995 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62629.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:13:09,005 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-03-08 14:13:09,362 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.381e+02 1.902e+02 2.514e+02 2.955e+02 4.757e+02, threshold=5.028e+02, percent-clipped=0.0 2023-03-08 14:13:34,648 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62646.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:13:51,161 INFO [train2.py:809] (1/4) Epoch 16, batch 2900, loss[ctc_loss=0.1054, att_loss=0.2559, loss=0.2258, over 17422.00 frames. utt_duration=1011 frames, utt_pad_proportion=0.04588, over 69.00 utterances.], tot_loss[ctc_loss=0.08468, att_loss=0.2413, loss=0.21, over 3269725.38 frames. utt_duration=1216 frames, utt_pad_proportion=0.06352, over 10771.07 utterances.], batch size: 69, lr: 6.72e-03, grad_scale: 8.0 2023-03-08 14:14:05,844 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62666.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:14:25,900 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6033, 2.3876, 5.0490, 3.7660, 2.9449, 4.2178, 4.8058, 4.5987], device='cuda:1'), covar=tensor([0.0221, 0.1818, 0.0146, 0.1008, 0.1872, 0.0254, 0.0110, 0.0222], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0243, 0.0161, 0.0311, 0.0265, 0.0195, 0.0143, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 14:15:03,039 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.5651, 5.8298, 5.2257, 5.6225, 5.4520, 5.0162, 5.2375, 5.0053], device='cuda:1'), covar=tensor([0.1377, 0.0945, 0.0965, 0.0919, 0.1062, 0.1699, 0.2542, 0.2493], device='cuda:1'), in_proj_covar=tensor([0.0495, 0.0562, 0.0433, 0.0435, 0.0412, 0.0449, 0.0582, 0.0511], 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') 2023-03-08 14:15:12,445 INFO [train2.py:809] (1/4) Epoch 16, batch 2950, loss[ctc_loss=0.1196, att_loss=0.2658, loss=0.2365, over 17055.00 frames. utt_duration=1313 frames, utt_pad_proportion=0.008984, over 52.00 utterances.], tot_loss[ctc_loss=0.08501, att_loss=0.2411, loss=0.2099, over 3258837.70 frames. utt_duration=1229 frames, utt_pad_proportion=0.06249, over 10618.74 utterances.], batch size: 52, lr: 6.72e-03, grad_scale: 8.0 2023-03-08 14:15:17,514 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9670, 4.9785, 4.7126, 2.8671, 4.7099, 4.6150, 4.1313, 2.7532], device='cuda:1'), covar=tensor([0.0115, 0.0099, 0.0262, 0.1009, 0.0090, 0.0202, 0.0369, 0.1325], device='cuda:1'), in_proj_covar=tensor([0.0069, 0.0096, 0.0094, 0.0109, 0.0079, 0.0107, 0.0098, 0.0103], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 14:15:44,555 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62727.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:15:50,981 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.359e+02 2.017e+02 2.387e+02 2.901e+02 6.735e+02, threshold=4.774e+02, percent-clipped=4.0 2023-03-08 14:15:59,917 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0026, 5.2254, 5.1633, 5.1674, 5.2974, 5.2725, 4.9754, 4.7423], device='cuda:1'), covar=tensor([0.0968, 0.0559, 0.0307, 0.0460, 0.0277, 0.0300, 0.0366, 0.0367], device='cuda:1'), in_proj_covar=tensor([0.0491, 0.0328, 0.0305, 0.0324, 0.0383, 0.0400, 0.0326, 0.0362], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 14:16:17,590 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62747.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:16:32,920 INFO [train2.py:809] (1/4) Epoch 16, batch 3000, loss[ctc_loss=0.07319, att_loss=0.2375, loss=0.2046, over 16174.00 frames. utt_duration=1580 frames, utt_pad_proportion=0.006489, over 41.00 utterances.], tot_loss[ctc_loss=0.08442, att_loss=0.2412, loss=0.2099, over 3269393.79 frames. utt_duration=1250 frames, utt_pad_proportion=0.05434, over 10474.45 utterances.], batch size: 41, lr: 6.72e-03, grad_scale: 8.0 2023-03-08 14:16:32,920 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-08 14:16:46,719 INFO [train2.py:843] (1/4) Epoch 16, validation: ctc_loss=0.0433, att_loss=0.235, loss=0.1967, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-08 14:16:46,720 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-08 14:17:03,933 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1326, 5.4205, 5.0338, 5.5006, 4.8957, 5.0892, 5.5917, 5.3737], device='cuda:1'), covar=tensor([0.0556, 0.0289, 0.0744, 0.0243, 0.0391, 0.0191, 0.0182, 0.0150], device='cuda:1'), in_proj_covar=tensor([0.0367, 0.0291, 0.0344, 0.0307, 0.0296, 0.0220, 0.0279, 0.0260], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0005, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 14:17:11,751 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62773.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:17:19,149 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62778.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:17:36,794 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8350, 5.2643, 5.3197, 5.2534, 5.3696, 5.3398, 5.0648, 4.8288], device='cuda:1'), covar=tensor([0.1245, 0.0688, 0.0322, 0.0506, 0.0365, 0.0384, 0.0416, 0.0421], device='cuda:1'), in_proj_covar=tensor([0.0491, 0.0328, 0.0305, 0.0324, 0.0382, 0.0399, 0.0327, 0.0363], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 14:18:06,026 INFO [train2.py:809] (1/4) Epoch 16, batch 3050, loss[ctc_loss=0.08981, att_loss=0.2246, loss=0.1976, over 15360.00 frames. utt_duration=1757 frames, utt_pad_proportion=0.01068, over 35.00 utterances.], tot_loss[ctc_loss=0.08476, att_loss=0.2411, loss=0.2098, over 3266129.51 frames. utt_duration=1251 frames, utt_pad_proportion=0.05675, over 10457.39 utterances.], batch size: 35, lr: 6.71e-03, grad_scale: 8.0 2023-03-08 14:18:36,335 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=62826.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:18:43,970 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.353e+02 2.145e+02 2.502e+02 3.037e+02 6.032e+02, threshold=5.003e+02, percent-clipped=3.0 2023-03-08 14:18:48,979 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62834.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:18:56,351 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-03-08 14:19:28,209 INFO [train2.py:809] (1/4) Epoch 16, batch 3100, loss[ctc_loss=0.08278, att_loss=0.2264, loss=0.1977, over 15944.00 frames. utt_duration=1557 frames, utt_pad_proportion=0.007202, over 41.00 utterances.], tot_loss[ctc_loss=0.08439, att_loss=0.2408, loss=0.2095, over 3263988.16 frames. utt_duration=1238 frames, utt_pad_proportion=0.06002, over 10559.78 utterances.], batch size: 41, lr: 6.71e-03, grad_scale: 8.0 2023-03-08 14:19:31,801 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.4996, 2.2626, 2.0154, 2.7339, 2.6012, 2.3311, 2.3063, 2.6463], device='cuda:1'), covar=tensor([0.1858, 0.3597, 0.2824, 0.1644, 0.2704, 0.1701, 0.2843, 0.1157], device='cuda:1'), in_proj_covar=tensor([0.0093, 0.0098, 0.0102, 0.0091, 0.0096, 0.0085, 0.0104, 0.0075], device='cuda:1'), out_proj_covar=tensor([6.7866e-05, 7.4146e-05, 7.7539e-05, 6.7867e-05, 6.9552e-05, 6.6523e-05, 7.6343e-05, 5.9450e-05], device='cuda:1') 2023-03-08 14:19:42,253 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4310, 4.6283, 4.2365, 4.6966, 4.2297, 4.3362, 4.7339, 4.5919], device='cuda:1'), covar=tensor([0.0585, 0.0310, 0.0744, 0.0305, 0.0434, 0.0393, 0.0243, 0.0186], device='cuda:1'), in_proj_covar=tensor([0.0370, 0.0293, 0.0347, 0.0311, 0.0299, 0.0222, 0.0281, 0.0262], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0005, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 14:20:25,422 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.7217, 5.9996, 5.3807, 5.7528, 5.5876, 5.1025, 5.3891, 5.1834], device='cuda:1'), covar=tensor([0.1343, 0.0848, 0.0849, 0.0799, 0.0916, 0.1421, 0.2297, 0.2232], device='cuda:1'), in_proj_covar=tensor([0.0496, 0.0565, 0.0431, 0.0435, 0.0410, 0.0452, 0.0587, 0.0509], 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') 2023-03-08 14:20:27,416 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4098, 2.1680, 4.7979, 3.6883, 2.8169, 4.1405, 4.3746, 4.4165], device='cuda:1'), covar=tensor([0.0227, 0.1967, 0.0142, 0.1075, 0.1822, 0.0269, 0.0182, 0.0271], device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0243, 0.0163, 0.0311, 0.0264, 0.0196, 0.0144, 0.0172], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 14:20:42,354 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62901.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:20:51,880 INFO [train2.py:809] (1/4) Epoch 16, batch 3150, loss[ctc_loss=0.08016, att_loss=0.2534, loss=0.2188, over 16889.00 frames. utt_duration=1380 frames, utt_pad_proportion=0.007077, over 49.00 utterances.], tot_loss[ctc_loss=0.08465, att_loss=0.2412, loss=0.2099, over 3274529.93 frames. utt_duration=1258 frames, utt_pad_proportion=0.05189, over 10421.29 utterances.], batch size: 49, lr: 6.71e-03, grad_scale: 8.0 2023-03-08 14:21:28,174 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62929.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:21:31,247 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.516e+02 2.078e+02 2.532e+02 3.325e+02 7.691e+02, threshold=5.063e+02, percent-clipped=6.0 2023-03-08 14:21:39,804 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-03-08 14:21:49,292 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62941.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:22:15,362 INFO [train2.py:809] (1/4) Epoch 16, batch 3200, loss[ctc_loss=0.07577, att_loss=0.2196, loss=0.1909, over 15498.00 frames. utt_duration=1724 frames, utt_pad_proportion=0.008881, over 36.00 utterances.], tot_loss[ctc_loss=0.08487, att_loss=0.2415, loss=0.2101, over 3273829.07 frames. utt_duration=1240 frames, utt_pad_proportion=0.0576, over 10574.26 utterances.], batch size: 36, lr: 6.71e-03, grad_scale: 8.0 2023-03-08 14:22:25,184 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-03-08 14:22:36,903 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7742, 2.3936, 2.6347, 2.6946, 2.7514, 2.6824, 2.6235, 2.7742], device='cuda:1'), covar=tensor([0.1452, 0.3411, 0.2532, 0.1617, 0.2935, 0.1313, 0.2357, 0.1090], device='cuda:1'), in_proj_covar=tensor([0.0094, 0.0100, 0.0104, 0.0092, 0.0097, 0.0086, 0.0106, 0.0076], device='cuda:1'), out_proj_covar=tensor([6.9025e-05, 7.5328e-05, 7.9110e-05, 6.8889e-05, 7.0602e-05, 6.7467e-05, 7.7406e-05, 6.0474e-05], device='cuda:1') 2023-03-08 14:22:47,673 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=62977.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:23:37,506 INFO [train2.py:809] (1/4) Epoch 16, batch 3250, loss[ctc_loss=0.08553, att_loss=0.2457, loss=0.2137, over 16344.00 frames. utt_duration=1454 frames, utt_pad_proportion=0.005305, over 45.00 utterances.], tot_loss[ctc_loss=0.08523, att_loss=0.2412, loss=0.21, over 3265556.04 frames. utt_duration=1232 frames, utt_pad_proportion=0.06282, over 10614.03 utterances.], batch size: 45, lr: 6.70e-03, grad_scale: 8.0 2023-03-08 14:23:48,341 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-03-08 14:24:03,051 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63022.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:24:17,455 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.315e+02 2.062e+02 2.466e+02 2.972e+02 8.137e+02, threshold=4.932e+02, percent-clipped=1.0 2023-03-08 14:24:45,800 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63047.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:25:01,611 INFO [train2.py:809] (1/4) Epoch 16, batch 3300, loss[ctc_loss=0.08409, att_loss=0.2542, loss=0.2202, over 16952.00 frames. utt_duration=686.5 frames, utt_pad_proportion=0.1344, over 99.00 utterances.], tot_loss[ctc_loss=0.08512, att_loss=0.2408, loss=0.2097, over 3258639.98 frames. utt_duration=1225 frames, utt_pad_proportion=0.06608, over 10654.90 utterances.], batch size: 99, lr: 6.70e-03, grad_scale: 8.0 2023-03-08 14:26:04,487 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=63095.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:26:24,099 INFO [train2.py:809] (1/4) Epoch 16, batch 3350, loss[ctc_loss=0.09859, att_loss=0.2569, loss=0.2252, over 17037.00 frames. utt_duration=1338 frames, utt_pad_proportion=0.007491, over 51.00 utterances.], tot_loss[ctc_loss=0.08414, att_loss=0.2401, loss=0.2089, over 3254969.60 frames. utt_duration=1265 frames, utt_pad_proportion=0.05662, over 10303.09 utterances.], batch size: 51, lr: 6.70e-03, grad_scale: 8.0 2023-03-08 14:26:52,250 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-03-08 14:27:00,458 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63129.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:27:03,354 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.304e+02 2.090e+02 2.534e+02 3.183e+02 5.519e+02, threshold=5.068e+02, percent-clipped=6.0 2023-03-08 14:27:47,073 INFO [train2.py:809] (1/4) Epoch 16, batch 3400, loss[ctc_loss=0.08253, att_loss=0.239, loss=0.2077, over 16405.00 frames. utt_duration=1493 frames, utt_pad_proportion=0.007601, over 44.00 utterances.], tot_loss[ctc_loss=0.08373, att_loss=0.2401, loss=0.2088, over 3263369.14 frames. utt_duration=1277 frames, utt_pad_proportion=0.05184, over 10233.45 utterances.], batch size: 44, lr: 6.70e-03, grad_scale: 8.0 2023-03-08 14:27:47,547 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7247, 4.7381, 4.6698, 4.4768, 5.3351, 4.9092, 4.6955, 2.4413], device='cuda:1'), covar=tensor([0.0176, 0.0353, 0.0288, 0.0421, 0.0757, 0.0152, 0.0300, 0.2117], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0155, 0.0161, 0.0173, 0.0351, 0.0137, 0.0146, 0.0214], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 14:29:01,693 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63201.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:29:11,062 INFO [train2.py:809] (1/4) Epoch 16, batch 3450, loss[ctc_loss=0.0838, att_loss=0.2559, loss=0.2215, over 16964.00 frames. utt_duration=1359 frames, utt_pad_proportion=0.006857, over 50.00 utterances.], tot_loss[ctc_loss=0.0837, att_loss=0.2409, loss=0.2095, over 3275720.09 frames. utt_duration=1265 frames, utt_pad_proportion=0.05057, over 10367.42 utterances.], batch size: 50, lr: 6.69e-03, grad_scale: 8.0 2023-03-08 14:29:34,441 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63221.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:29:49,778 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.272e+02 2.022e+02 2.458e+02 2.979e+02 7.741e+02, threshold=4.916e+02, percent-clipped=5.0 2023-03-08 14:30:06,811 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63241.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:30:20,165 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=63249.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:30:32,950 INFO [train2.py:809] (1/4) Epoch 16, batch 3500, loss[ctc_loss=0.08594, att_loss=0.2525, loss=0.2192, over 16760.00 frames. utt_duration=1398 frames, utt_pad_proportion=0.006851, over 48.00 utterances.], tot_loss[ctc_loss=0.08459, att_loss=0.2419, loss=0.2104, over 3275853.93 frames. utt_duration=1218 frames, utt_pad_proportion=0.06221, over 10767.47 utterances.], batch size: 48, lr: 6.69e-03, grad_scale: 8.0 2023-03-08 14:31:14,065 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63282.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:31:25,076 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=63289.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:31:55,973 INFO [train2.py:809] (1/4) Epoch 16, batch 3550, loss[ctc_loss=0.05837, att_loss=0.2132, loss=0.1822, over 15891.00 frames. utt_duration=1631 frames, utt_pad_proportion=0.008895, over 39.00 utterances.], tot_loss[ctc_loss=0.08405, att_loss=0.241, loss=0.2096, over 3268918.71 frames. utt_duration=1236 frames, utt_pad_proportion=0.06067, over 10593.73 utterances.], batch size: 39, lr: 6.69e-03, grad_scale: 8.0 2023-03-08 14:32:20,759 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63322.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:32:36,185 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.235e+02 2.051e+02 2.467e+02 3.018e+02 7.226e+02, threshold=4.933e+02, percent-clipped=6.0 2023-03-08 14:33:19,239 INFO [train2.py:809] (1/4) Epoch 16, batch 3600, loss[ctc_loss=0.06511, att_loss=0.2099, loss=0.181, over 15495.00 frames. utt_duration=1723 frames, utt_pad_proportion=0.009057, over 36.00 utterances.], tot_loss[ctc_loss=0.08424, att_loss=0.2408, loss=0.2095, over 3268751.23 frames. utt_duration=1244 frames, utt_pad_proportion=0.05874, over 10521.76 utterances.], batch size: 36, lr: 6.69e-03, grad_scale: 8.0 2023-03-08 14:33:39,692 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=63370.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:34:42,621 INFO [train2.py:809] (1/4) Epoch 16, batch 3650, loss[ctc_loss=0.09832, att_loss=0.261, loss=0.2284, over 17297.00 frames. utt_duration=877.1 frames, utt_pad_proportion=0.0796, over 79.00 utterances.], tot_loss[ctc_loss=0.08457, att_loss=0.2413, loss=0.2099, over 3273945.99 frames. utt_duration=1250 frames, utt_pad_proportion=0.05705, over 10492.90 utterances.], batch size: 79, lr: 6.68e-03, grad_scale: 8.0 2023-03-08 14:35:03,566 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63420.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:35:06,789 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63422.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:35:19,551 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63429.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:35:22,399 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.472e+02 2.142e+02 2.479e+02 3.214e+02 4.969e+02, threshold=4.958e+02, percent-clipped=1.0 2023-03-08 14:36:06,128 INFO [train2.py:809] (1/4) Epoch 16, batch 3700, loss[ctc_loss=0.08138, att_loss=0.24, loss=0.2082, over 16392.00 frames. utt_duration=1492 frames, utt_pad_proportion=0.008085, over 44.00 utterances.], tot_loss[ctc_loss=0.08457, att_loss=0.2404, loss=0.2092, over 3263913.07 frames. utt_duration=1249 frames, utt_pad_proportion=0.05873, over 10461.49 utterances.], batch size: 44, lr: 6.68e-03, grad_scale: 8.0 2023-03-08 14:36:40,544 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=63477.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:36:47,191 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63481.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:36:50,262 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63483.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:37:22,059 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2550, 3.8028, 3.7697, 3.2697, 3.9235, 3.8891, 3.8308, 3.0208], device='cuda:1'), covar=tensor([0.0676, 0.1076, 0.1791, 0.4437, 0.0871, 0.1673, 0.0721, 0.3719], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0159, 0.0171, 0.0235, 0.0131, 0.0229, 0.0146, 0.0198], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 14:37:30,416 INFO [train2.py:809] (1/4) Epoch 16, batch 3750, loss[ctc_loss=0.06579, att_loss=0.2475, loss=0.2112, over 16619.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.005767, over 47.00 utterances.], tot_loss[ctc_loss=0.0845, att_loss=0.2406, loss=0.2094, over 3270523.80 frames. utt_duration=1257 frames, utt_pad_proportion=0.05459, over 10417.56 utterances.], batch size: 47, lr: 6.68e-03, grad_scale: 8.0 2023-03-08 14:38:09,923 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.475e+02 2.179e+02 2.509e+02 3.051e+02 5.574e+02, threshold=5.018e+02, percent-clipped=2.0 2023-03-08 14:38:31,418 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.3313, 5.2466, 4.9470, 3.0334, 5.0222, 4.8406, 4.6214, 2.9367], device='cuda:1'), covar=tensor([0.0086, 0.0105, 0.0272, 0.1104, 0.0089, 0.0179, 0.0279, 0.1393], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0097, 0.0095, 0.0110, 0.0080, 0.0107, 0.0099, 0.0102], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 14:38:52,444 INFO [train2.py:809] (1/4) Epoch 16, batch 3800, loss[ctc_loss=0.07986, att_loss=0.2273, loss=0.1978, over 16165.00 frames. utt_duration=1579 frames, utt_pad_proportion=0.006539, over 41.00 utterances.], tot_loss[ctc_loss=0.08491, att_loss=0.2406, loss=0.2095, over 3263391.26 frames. utt_duration=1252 frames, utt_pad_proportion=0.05628, over 10436.53 utterances.], batch size: 41, lr: 6.67e-03, grad_scale: 8.0 2023-03-08 14:39:07,229 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2535, 4.5685, 4.5787, 4.7487, 2.7897, 4.8094, 2.7066, 1.8999], device='cuda:1'), covar=tensor([0.0345, 0.0202, 0.0596, 0.0156, 0.1718, 0.0133, 0.1467, 0.1656], device='cuda:1'), in_proj_covar=tensor([0.0168, 0.0140, 0.0255, 0.0134, 0.0220, 0.0122, 0.0228, 0.0201], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 14:39:11,858 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63569.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:39:24,682 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-03-08 14:39:25,935 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63577.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:39:41,237 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63586.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:40:15,132 INFO [train2.py:809] (1/4) Epoch 16, batch 3850, loss[ctc_loss=0.08163, att_loss=0.2306, loss=0.2008, over 15970.00 frames. utt_duration=1559 frames, utt_pad_proportion=0.006171, over 41.00 utterances.], tot_loss[ctc_loss=0.08514, att_loss=0.2413, loss=0.2101, over 3274775.26 frames. utt_duration=1240 frames, utt_pad_proportion=0.05599, over 10573.34 utterances.], batch size: 41, lr: 6.67e-03, grad_scale: 8.0 2023-03-08 14:40:52,472 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63630.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:40:53,570 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.309e+02 2.062e+02 2.555e+02 3.342e+02 5.623e+02, threshold=5.109e+02, percent-clipped=2.0 2023-03-08 14:41:19,867 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63647.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:41:35,252 INFO [train2.py:809] (1/4) Epoch 16, batch 3900, loss[ctc_loss=0.1172, att_loss=0.273, loss=0.2418, over 17289.00 frames. utt_duration=1259 frames, utt_pad_proportion=0.01176, over 55.00 utterances.], tot_loss[ctc_loss=0.08476, att_loss=0.2417, loss=0.2103, over 3279817.16 frames. utt_duration=1253 frames, utt_pad_proportion=0.05125, over 10485.83 utterances.], batch size: 55, lr: 6.67e-03, grad_scale: 8.0 2023-03-08 14:42:39,269 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1426, 4.4618, 4.0532, 4.6469, 2.6026, 4.7517, 2.4433, 1.9638], device='cuda:1'), covar=tensor([0.0434, 0.0186, 0.0983, 0.0171, 0.2090, 0.0130, 0.1943, 0.1906], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0138, 0.0253, 0.0132, 0.0217, 0.0121, 0.0226, 0.0200], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 14:42:54,346 INFO [train2.py:809] (1/4) Epoch 16, batch 3950, loss[ctc_loss=0.09348, att_loss=0.2503, loss=0.219, over 16391.00 frames. utt_duration=1492 frames, utt_pad_proportion=0.008039, over 44.00 utterances.], tot_loss[ctc_loss=0.08426, att_loss=0.2412, loss=0.2098, over 3276432.09 frames. utt_duration=1259 frames, utt_pad_proportion=0.05182, over 10419.93 utterances.], batch size: 44, lr: 6.67e-03, grad_scale: 8.0 2023-03-08 14:43:02,779 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2923, 4.2499, 4.3783, 4.1711, 4.8408, 4.3223, 4.2386, 2.3151], device='cuda:1'), covar=tensor([0.0264, 0.0359, 0.0317, 0.0322, 0.0853, 0.0234, 0.0336, 0.2093], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0154, 0.0161, 0.0172, 0.0347, 0.0135, 0.0145, 0.0211], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 14:43:31,754 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.446e+02 2.047e+02 2.429e+02 3.327e+02 6.424e+02, threshold=4.857e+02, percent-clipped=3.0 2023-03-08 14:44:08,718 INFO [train2.py:809] (1/4) Epoch 17, batch 0, loss[ctc_loss=0.08598, att_loss=0.2461, loss=0.2141, over 16458.00 frames. utt_duration=1433 frames, utt_pad_proportion=0.007783, over 46.00 utterances.], tot_loss[ctc_loss=0.08598, att_loss=0.2461, loss=0.2141, over 16458.00 frames. utt_duration=1433 frames, utt_pad_proportion=0.007783, over 46.00 utterances.], batch size: 46, lr: 6.46e-03, grad_scale: 8.0 2023-03-08 14:44:08,719 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-08 14:44:15,030 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.7269, 4.1771, 4.1775, 2.2397, 1.9090, 2.6552, 1.8949, 3.5083], device='cuda:1'), covar=tensor([0.0684, 0.0325, 0.0335, 0.4786, 0.5899, 0.2644, 0.4009, 0.1382], device='cuda:1'), in_proj_covar=tensor([0.0340, 0.0244, 0.0249, 0.0225, 0.0338, 0.0329, 0.0239, 0.0352], device='cuda:1'), out_proj_covar=tensor([1.4724e-04, 9.0898e-05, 1.0734e-04, 9.7254e-05, 1.4338e-04, 1.2993e-04, 9.5658e-05, 1.4504e-04], device='cuda:1') 2023-03-08 14:44:15,528 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.6313, 6.1869, 6.2062, 6.0181, 6.1340, 6.5919, 5.5712, 6.6626], device='cuda:1'), covar=tensor([0.0479, 0.0535, 0.0551, 0.0996, 0.1615, 0.0609, 0.0348, 0.0403], device='cuda:1'), in_proj_covar=tensor([0.0799, 0.0469, 0.0555, 0.0615, 0.0814, 0.0568, 0.0449, 0.0546], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 14:44:18,900 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.0048, 3.6571, 3.6519, 3.2575, 3.7466, 3.7099, 3.7487, 2.7595], device='cuda:1'), covar=tensor([0.0736, 0.1233, 0.1774, 0.3027, 0.0949, 0.2167, 0.0750, 0.4217], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0158, 0.0170, 0.0232, 0.0131, 0.0227, 0.0145, 0.0197], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 14:44:21,758 INFO [train2.py:843] (1/4) Epoch 17, validation: ctc_loss=0.04327, att_loss=0.2362, loss=0.1976, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-08 14:44:21,758 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-08 14:44:22,894 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-03-08 14:44:32,338 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1001, 5.3952, 4.9300, 5.4577, 4.7789, 5.0338, 5.5019, 5.3015], device='cuda:1'), covar=tensor([0.0458, 0.0248, 0.0714, 0.0227, 0.0397, 0.0259, 0.0207, 0.0154], device='cuda:1'), in_proj_covar=tensor([0.0368, 0.0294, 0.0346, 0.0310, 0.0298, 0.0222, 0.0280, 0.0261], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 14:45:20,131 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63776.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:45:20,183 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0850, 5.4338, 4.9640, 5.5445, 4.8408, 5.0537, 5.5644, 5.3407], device='cuda:1'), covar=tensor([0.0524, 0.0312, 0.0746, 0.0269, 0.0389, 0.0220, 0.0256, 0.0168], device='cuda:1'), in_proj_covar=tensor([0.0370, 0.0295, 0.0347, 0.0311, 0.0299, 0.0222, 0.0281, 0.0262], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 14:45:23,460 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63778.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:45:44,536 INFO [train2.py:809] (1/4) Epoch 17, batch 50, loss[ctc_loss=0.08713, att_loss=0.2605, loss=0.2258, over 17387.00 frames. utt_duration=1105 frames, utt_pad_proportion=0.03452, over 63.00 utterances.], tot_loss[ctc_loss=0.0846, att_loss=0.243, loss=0.2113, over 741177.97 frames. utt_duration=1176 frames, utt_pad_proportion=0.06964, over 2523.68 utterances.], batch size: 63, lr: 6.46e-03, grad_scale: 8.0 2023-03-08 14:45:59,106 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-03-08 14:46:16,972 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63810.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:46:27,068 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9237, 6.0959, 5.5479, 5.8626, 5.8237, 5.3206, 5.5821, 5.2912], device='cuda:1'), covar=tensor([0.1112, 0.0874, 0.0856, 0.0867, 0.0854, 0.1402, 0.2152, 0.2437], device='cuda:1'), in_proj_covar=tensor([0.0489, 0.0560, 0.0428, 0.0430, 0.0405, 0.0450, 0.0578, 0.0502], 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') 2023-03-08 14:46:50,925 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.085e+02 2.041e+02 2.465e+02 3.095e+02 5.834e+02, threshold=4.929e+02, percent-clipped=2.0 2023-03-08 14:47:08,065 INFO [train2.py:809] (1/4) Epoch 17, batch 100, loss[ctc_loss=0.08353, att_loss=0.2466, loss=0.214, over 16764.00 frames. utt_duration=1398 frames, utt_pad_proportion=0.006792, over 48.00 utterances.], tot_loss[ctc_loss=0.08374, att_loss=0.2413, loss=0.2098, over 1305601.44 frames. utt_duration=1238 frames, utt_pad_proportion=0.05517, over 4223.58 utterances.], batch size: 48, lr: 6.46e-03, grad_scale: 8.0 2023-03-08 14:47:16,175 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63845.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:47:58,762 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63871.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:48:08,258 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63877.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:48:31,348 INFO [train2.py:809] (1/4) Epoch 17, batch 150, loss[ctc_loss=0.1198, att_loss=0.2716, loss=0.2413, over 17064.00 frames. utt_duration=1289 frames, utt_pad_proportion=0.008963, over 53.00 utterances.], tot_loss[ctc_loss=0.08299, att_loss=0.241, loss=0.2094, over 1742100.71 frames. utt_duration=1224 frames, utt_pad_proportion=0.05791, over 5699.97 utterances.], batch size: 53, lr: 6.46e-03, grad_scale: 4.0 2023-03-08 14:48:32,404 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7554, 3.5595, 3.5068, 2.9361, 3.5101, 3.5999, 3.6059, 2.3488], device='cuda:1'), covar=tensor([0.0988, 0.1601, 0.2304, 0.4741, 0.1078, 0.4098, 0.0877, 0.6034], device='cuda:1'), in_proj_covar=tensor([0.0137, 0.0160, 0.0172, 0.0236, 0.0133, 0.0231, 0.0146, 0.0199], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 14:48:56,197 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63906.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:49:12,652 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63916.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:49:26,795 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=63925.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:49:26,902 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63925.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:49:33,475 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5315, 2.5490, 4.9413, 3.9465, 3.0653, 4.1526, 4.9378, 4.6397], device='cuda:1'), covar=tensor([0.0247, 0.1759, 0.0205, 0.0985, 0.1816, 0.0243, 0.0109, 0.0235], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0239, 0.0163, 0.0305, 0.0262, 0.0196, 0.0143, 0.0171], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 14:49:37,543 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.362e+02 1.983e+02 2.552e+02 2.961e+02 5.419e+02, threshold=5.104e+02, percent-clipped=2.0 2023-03-08 14:49:46,482 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0732, 5.0480, 4.7244, 2.7491, 4.8126, 4.6363, 4.1486, 2.5733], device='cuda:1'), covar=tensor([0.0161, 0.0112, 0.0348, 0.1331, 0.0105, 0.0216, 0.0426, 0.1762], device='cuda:1'), in_proj_covar=tensor([0.0070, 0.0097, 0.0095, 0.0110, 0.0080, 0.0107, 0.0098, 0.0103], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 14:49:52,618 INFO [train2.py:809] (1/4) Epoch 17, batch 200, loss[ctc_loss=0.05788, att_loss=0.2073, loss=0.1775, over 15636.00 frames. utt_duration=1692 frames, utt_pad_proportion=0.00874, over 37.00 utterances.], tot_loss[ctc_loss=0.08272, att_loss=0.2409, loss=0.2092, over 2085852.31 frames. utt_duration=1223 frames, utt_pad_proportion=0.05558, over 6831.62 utterances.], batch size: 37, lr: 6.45e-03, grad_scale: 4.0 2023-03-08 14:49:55,660 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63942.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:50:52,519 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63977.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:51:15,250 INFO [train2.py:809] (1/4) Epoch 17, batch 250, loss[ctc_loss=0.0802, att_loss=0.2501, loss=0.2161, over 16757.00 frames. utt_duration=1398 frames, utt_pad_proportion=0.007206, over 48.00 utterances.], tot_loss[ctc_loss=0.08236, att_loss=0.2405, loss=0.2089, over 2341217.92 frames. utt_duration=1260 frames, utt_pad_proportion=0.04963, over 7439.17 utterances.], batch size: 48, lr: 6.45e-03, grad_scale: 4.0 2023-03-08 14:52:14,231 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64023.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:52:28,262 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.355e+02 2.064e+02 2.460e+02 3.052e+02 5.985e+02, threshold=4.919e+02, percent-clipped=1.0 2023-03-08 14:52:43,728 INFO [train2.py:809] (1/4) Epoch 17, batch 300, loss[ctc_loss=0.07666, att_loss=0.2456, loss=0.2118, over 17142.00 frames. utt_duration=1226 frames, utt_pad_proportion=0.01355, over 56.00 utterances.], tot_loss[ctc_loss=0.0832, att_loss=0.2407, loss=0.2092, over 2546958.21 frames. utt_duration=1235 frames, utt_pad_proportion=0.05623, over 8261.58 utterances.], batch size: 56, lr: 6.45e-03, grad_scale: 4.0 2023-03-08 14:53:39,699 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64076.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:53:42,908 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64078.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:53:52,330 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64084.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:54:04,132 INFO [train2.py:809] (1/4) Epoch 17, batch 350, loss[ctc_loss=0.08884, att_loss=0.2553, loss=0.222, over 17304.00 frames. utt_duration=1260 frames, utt_pad_proportion=0.01201, over 55.00 utterances.], tot_loss[ctc_loss=0.08286, att_loss=0.2401, loss=0.2087, over 2705837.86 frames. utt_duration=1264 frames, utt_pad_proportion=0.04873, over 8569.98 utterances.], batch size: 55, lr: 6.45e-03, grad_scale: 4.0 2023-03-08 14:54:56,966 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=64124.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:55:00,051 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=64126.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:55:09,305 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.372e+02 2.060e+02 2.368e+02 3.330e+02 9.339e+02, threshold=4.736e+02, percent-clipped=6.0 2023-03-08 14:55:25,124 INFO [train2.py:809] (1/4) Epoch 17, batch 400, loss[ctc_loss=0.06826, att_loss=0.222, loss=0.1913, over 16280.00 frames. utt_duration=1516 frames, utt_pad_proportion=0.007356, over 43.00 utterances.], tot_loss[ctc_loss=0.0813, att_loss=0.2385, loss=0.2071, over 2824880.47 frames. utt_duration=1307 frames, utt_pad_proportion=0.04112, over 8656.87 utterances.], batch size: 43, lr: 6.44e-03, grad_scale: 8.0 2023-03-08 14:55:25,585 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0817, 4.4231, 4.1975, 4.5211, 2.6925, 4.5484, 2.6264, 1.8337], device='cuda:1'), covar=tensor([0.0412, 0.0178, 0.0748, 0.0184, 0.1669, 0.0157, 0.1545, 0.1797], device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0142, 0.0259, 0.0136, 0.0221, 0.0125, 0.0231, 0.0205], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 14:55:34,064 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0087, 6.1431, 5.5731, 5.8697, 5.7885, 5.4122, 5.6619, 5.3598], device='cuda:1'), covar=tensor([0.0829, 0.0661, 0.0798, 0.0672, 0.0742, 0.1256, 0.1828, 0.2005], device='cuda:1'), in_proj_covar=tensor([0.0489, 0.0558, 0.0428, 0.0428, 0.0410, 0.0453, 0.0583, 0.0503], 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') 2023-03-08 14:56:07,173 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64166.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:56:46,473 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.6619, 3.0141, 3.5664, 4.4322, 3.8779, 4.0088, 2.8960, 2.3050], device='cuda:1'), covar=tensor([0.0561, 0.1854, 0.0892, 0.0582, 0.0886, 0.0422, 0.1542, 0.2218], device='cuda:1'), in_proj_covar=tensor([0.0176, 0.0212, 0.0187, 0.0201, 0.0210, 0.0168, 0.0195, 0.0183], 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') 2023-03-08 14:56:47,839 INFO [train2.py:809] (1/4) Epoch 17, batch 450, loss[ctc_loss=0.06664, att_loss=0.2264, loss=0.1944, over 15953.00 frames. utt_duration=1558 frames, utt_pad_proportion=0.006346, over 41.00 utterances.], tot_loss[ctc_loss=0.08198, att_loss=0.2392, loss=0.2078, over 2918452.12 frames. utt_duration=1245 frames, utt_pad_proportion=0.05863, over 9385.62 utterances.], batch size: 41, lr: 6.44e-03, grad_scale: 8.0 2023-03-08 14:57:04,186 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64201.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:57:42,939 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64225.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:57:53,581 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.502e+02 2.140e+02 2.597e+02 3.193e+02 5.967e+02, threshold=5.195e+02, percent-clipped=4.0 2023-03-08 14:57:53,961 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64232.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:58:09,211 INFO [train2.py:809] (1/4) Epoch 17, batch 500, loss[ctc_loss=0.07904, att_loss=0.2597, loss=0.2236, over 16631.00 frames. utt_duration=1417 frames, utt_pad_proportion=0.005005, over 47.00 utterances.], tot_loss[ctc_loss=0.0824, att_loss=0.2392, loss=0.2078, over 2995740.48 frames. utt_duration=1267 frames, utt_pad_proportion=0.05174, over 9465.39 utterances.], batch size: 47, lr: 6.44e-03, grad_scale: 8.0 2023-03-08 14:58:11,140 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64242.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:58:58,187 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64272.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:58:59,748 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=64273.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:59:27,935 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=64290.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 14:59:29,433 INFO [train2.py:809] (1/4) Epoch 17, batch 550, loss[ctc_loss=0.08449, att_loss=0.253, loss=0.2193, over 16765.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.006718, over 48.00 utterances.], tot_loss[ctc_loss=0.08237, att_loss=0.2387, loss=0.2074, over 3051176.46 frames. utt_duration=1269 frames, utt_pad_proportion=0.05272, over 9629.59 utterances.], batch size: 48, lr: 6.44e-03, grad_scale: 8.0 2023-03-08 14:59:33,715 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64293.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:00:16,529 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6283, 5.1144, 4.9036, 5.1102, 5.1206, 4.8044, 3.7018, 5.0662], device='cuda:1'), covar=tensor([0.0116, 0.0099, 0.0131, 0.0072, 0.0099, 0.0105, 0.0600, 0.0164], device='cuda:1'), in_proj_covar=tensor([0.0085, 0.0082, 0.0102, 0.0064, 0.0068, 0.0080, 0.0099, 0.0103], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 15:00:26,241 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-03-08 15:00:34,620 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.348e+02 1.879e+02 2.342e+02 2.961e+02 5.931e+02, threshold=4.683e+02, percent-clipped=3.0 2023-03-08 15:00:50,372 INFO [train2.py:809] (1/4) Epoch 17, batch 600, loss[ctc_loss=0.08908, att_loss=0.2495, loss=0.2175, over 17035.00 frames. utt_duration=1287 frames, utt_pad_proportion=0.009884, over 53.00 utterances.], tot_loss[ctc_loss=0.08214, att_loss=0.2384, loss=0.2072, over 3092859.20 frames. utt_duration=1262 frames, utt_pad_proportion=0.055, over 9817.41 utterances.], batch size: 53, lr: 6.43e-03, grad_scale: 8.0 2023-03-08 15:00:57,632 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64345.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:01:51,593 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64379.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:01:57,018 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.69 vs. limit=2.0 2023-03-08 15:02:05,143 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5240, 2.9962, 3.6464, 3.0605, 3.6060, 4.6319, 4.3514, 3.2008], device='cuda:1'), covar=tensor([0.0401, 0.1807, 0.1201, 0.1367, 0.1102, 0.1020, 0.0618, 0.1457], device='cuda:1'), in_proj_covar=tensor([0.0243, 0.0242, 0.0270, 0.0213, 0.0255, 0.0347, 0.0250, 0.0231], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 15:02:05,910 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-03-08 15:02:12,268 INFO [train2.py:809] (1/4) Epoch 17, batch 650, loss[ctc_loss=0.1209, att_loss=0.2453, loss=0.2204, over 15931.00 frames. utt_duration=1556 frames, utt_pad_proportion=0.008581, over 41.00 utterances.], tot_loss[ctc_loss=0.08224, att_loss=0.2389, loss=0.2076, over 3138429.28 frames. utt_duration=1268 frames, utt_pad_proportion=0.04986, over 9910.63 utterances.], batch size: 41, lr: 6.43e-03, grad_scale: 8.0 2023-03-08 15:02:16,314 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-03-08 15:02:36,714 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64406.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:03:17,364 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.378e+02 2.002e+02 2.471e+02 3.095e+02 7.929e+02, threshold=4.943e+02, percent-clipped=8.0 2023-03-08 15:03:32,610 INFO [train2.py:809] (1/4) Epoch 17, batch 700, loss[ctc_loss=0.1213, att_loss=0.2611, loss=0.2331, over 16472.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.00688, over 46.00 utterances.], tot_loss[ctc_loss=0.08271, att_loss=0.2391, loss=0.2079, over 3167655.24 frames. utt_duration=1276 frames, utt_pad_proportion=0.049, over 9945.49 utterances.], batch size: 46, lr: 6.43e-03, grad_scale: 8.0 2023-03-08 15:03:56,489 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7902, 4.2722, 4.4869, 4.2941, 4.4093, 4.6769, 4.3885, 4.7829], device='cuda:1'), covar=tensor([0.0838, 0.0882, 0.0794, 0.1215, 0.1658, 0.0994, 0.1634, 0.0737], device='cuda:1'), in_proj_covar=tensor([0.0787, 0.0468, 0.0553, 0.0608, 0.0807, 0.0565, 0.0442, 0.0549], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 15:04:12,645 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64465.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:04:14,075 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64466.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:04:55,321 INFO [train2.py:809] (1/4) Epoch 17, batch 750, loss[ctc_loss=0.08345, att_loss=0.2357, loss=0.2053, over 16010.00 frames. utt_duration=1602 frames, utt_pad_proportion=0.00728, over 40.00 utterances.], tot_loss[ctc_loss=0.0836, att_loss=0.2403, loss=0.209, over 3190751.27 frames. utt_duration=1257 frames, utt_pad_proportion=0.05333, over 10163.85 utterances.], batch size: 40, lr: 6.43e-03, grad_scale: 8.0 2023-03-08 15:05:11,197 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64501.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:05:25,060 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-03-08 15:05:25,886 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8645, 3.3069, 3.8610, 3.3765, 4.0005, 4.9147, 4.6727, 3.8105], device='cuda:1'), covar=tensor([0.0274, 0.1506, 0.1017, 0.1135, 0.0791, 0.0841, 0.0514, 0.0897], device='cuda:1'), in_proj_covar=tensor([0.0240, 0.0240, 0.0266, 0.0211, 0.0252, 0.0343, 0.0246, 0.0229], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 15:05:31,702 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=64514.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:05:31,828 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0042, 5.2189, 5.1770, 5.2110, 5.2424, 5.2360, 4.9421, 4.6645], device='cuda:1'), covar=tensor([0.0970, 0.0514, 0.0282, 0.0400, 0.0301, 0.0319, 0.0356, 0.0356], device='cuda:1'), in_proj_covar=tensor([0.0491, 0.0331, 0.0306, 0.0323, 0.0387, 0.0402, 0.0328, 0.0365], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 15:05:51,374 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64526.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:06:01,019 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.374e+02 2.298e+02 2.805e+02 3.347e+02 6.423e+02, threshold=5.609e+02, percent-clipped=6.0 2023-03-08 15:06:16,287 INFO [train2.py:809] (1/4) Epoch 17, batch 800, loss[ctc_loss=0.09823, att_loss=0.2569, loss=0.2251, over 17355.00 frames. utt_duration=1178 frames, utt_pad_proportion=0.02165, over 59.00 utterances.], tot_loss[ctc_loss=0.08345, att_loss=0.2407, loss=0.2093, over 3214408.76 frames. utt_duration=1264 frames, utt_pad_proportion=0.05, over 10186.35 utterances.], batch size: 59, lr: 6.42e-03, grad_scale: 8.0 2023-03-08 15:06:28,760 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=64549.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:07:05,764 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64572.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:07:32,186 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64588.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:07:33,855 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64589.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:07:36,532 INFO [train2.py:809] (1/4) Epoch 17, batch 850, loss[ctc_loss=0.07639, att_loss=0.2071, loss=0.181, over 15502.00 frames. utt_duration=1724 frames, utt_pad_proportion=0.008578, over 36.00 utterances.], tot_loss[ctc_loss=0.08287, att_loss=0.2403, loss=0.2088, over 3224864.20 frames. utt_duration=1263 frames, utt_pad_proportion=0.05141, over 10228.35 utterances.], batch size: 36, lr: 6.42e-03, grad_scale: 8.0 2023-03-08 15:07:45,224 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4914, 4.5985, 4.5564, 4.5446, 5.2060, 4.6920, 4.5362, 2.3563], device='cuda:1'), covar=tensor([0.0248, 0.0279, 0.0329, 0.0270, 0.0700, 0.0195, 0.0288, 0.2141], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0158, 0.0164, 0.0178, 0.0357, 0.0138, 0.0149, 0.0215], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 15:07:45,898 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.29 vs. limit=5.0 2023-03-08 15:08:24,156 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=64620.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:08:43,533 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.557e+02 2.027e+02 2.349e+02 3.075e+02 7.936e+02, threshold=4.698e+02, percent-clipped=3.0 2023-03-08 15:08:58,454 INFO [train2.py:809] (1/4) Epoch 17, batch 900, loss[ctc_loss=0.1372, att_loss=0.2697, loss=0.2432, over 13684.00 frames. utt_duration=379 frames, utt_pad_proportion=0.3409, over 145.00 utterances.], tot_loss[ctc_loss=0.08343, att_loss=0.2401, loss=0.2087, over 3225277.76 frames. utt_duration=1259 frames, utt_pad_proportion=0.05581, over 10262.84 utterances.], batch size: 145, lr: 6.42e-03, grad_scale: 8.0 2023-03-08 15:09:10,721 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64648.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:09:14,092 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64650.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:10:01,389 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64679.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:10:21,891 INFO [train2.py:809] (1/4) Epoch 17, batch 950, loss[ctc_loss=0.06932, att_loss=0.2308, loss=0.1985, over 11125.00 frames. utt_duration=1856 frames, utt_pad_proportion=0.2067, over 24.00 utterances.], tot_loss[ctc_loss=0.08272, att_loss=0.2403, loss=0.2087, over 3236139.66 frames. utt_duration=1276 frames, utt_pad_proportion=0.04963, over 10157.79 utterances.], batch size: 24, lr: 6.42e-03, grad_scale: 8.0 2023-03-08 15:10:38,218 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64701.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:10:42,268 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64703.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:10:51,685 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64709.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:11:19,677 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=64727.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:11:28,018 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.133e+02 2.001e+02 2.338e+02 2.858e+02 5.487e+02, threshold=4.676e+02, percent-clipped=1.0 2023-03-08 15:11:36,659 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0970, 5.3714, 5.3288, 5.3464, 5.3930, 5.3537, 5.1288, 4.8308], device='cuda:1'), covar=tensor([0.0933, 0.0441, 0.0252, 0.0419, 0.0262, 0.0284, 0.0318, 0.0299], device='cuda:1'), in_proj_covar=tensor([0.0496, 0.0335, 0.0311, 0.0328, 0.0391, 0.0406, 0.0332, 0.0370], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0003, 0.0004], device='cuda:1') 2023-03-08 15:11:42,849 INFO [train2.py:809] (1/4) Epoch 17, batch 1000, loss[ctc_loss=0.09099, att_loss=0.2596, loss=0.2259, over 16853.00 frames. utt_duration=1378 frames, utt_pad_proportion=0.008213, over 49.00 utterances.], tot_loss[ctc_loss=0.08291, att_loss=0.2403, loss=0.2089, over 3249456.20 frames. utt_duration=1274 frames, utt_pad_proportion=0.04827, over 10214.19 utterances.], batch size: 49, lr: 6.41e-03, grad_scale: 8.0 2023-03-08 15:11:49,883 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64745.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:12:01,446 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64752.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:12:14,098 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2402, 2.8449, 3.0315, 4.2210, 3.8068, 3.9630, 2.7963, 2.1404], device='cuda:1'), covar=tensor([0.0772, 0.2075, 0.1151, 0.0663, 0.0825, 0.0408, 0.1530, 0.2357], device='cuda:1'), in_proj_covar=tensor([0.0178, 0.0216, 0.0188, 0.0205, 0.0212, 0.0171, 0.0197, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 15:12:20,086 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64764.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:12:42,119 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-03-08 15:12:55,513 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6566, 3.2653, 3.7281, 3.1485, 3.7392, 4.8018, 4.5417, 3.5552], device='cuda:1'), covar=tensor([0.0341, 0.1613, 0.1259, 0.1317, 0.0968, 0.0701, 0.0529, 0.1170], device='cuda:1'), in_proj_covar=tensor([0.0238, 0.0238, 0.0265, 0.0211, 0.0249, 0.0342, 0.0247, 0.0230], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 15:13:02,949 INFO [train2.py:809] (1/4) Epoch 17, batch 1050, loss[ctc_loss=0.0696, att_loss=0.2293, loss=0.1974, over 15772.00 frames. utt_duration=1662 frames, utt_pad_proportion=0.008495, over 38.00 utterances.], tot_loss[ctc_loss=0.08327, att_loss=0.2406, loss=0.2091, over 3247399.33 frames. utt_duration=1267 frames, utt_pad_proportion=0.05098, over 10264.68 utterances.], batch size: 38, lr: 6.41e-03, grad_scale: 8.0 2023-03-08 15:13:27,785 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64806.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:13:38,392 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64813.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:13:50,499 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64821.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:14:01,614 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0443, 4.4177, 4.4654, 4.7809, 2.8752, 4.5936, 2.9747, 2.0732], device='cuda:1'), covar=tensor([0.0405, 0.0246, 0.0967, 0.0144, 0.1601, 0.0158, 0.1389, 0.1697], device='cuda:1'), in_proj_covar=tensor([0.0168, 0.0141, 0.0254, 0.0134, 0.0215, 0.0122, 0.0227, 0.0202], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 15:14:08,043 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.316e+02 2.188e+02 2.644e+02 3.168e+02 6.507e+02, threshold=5.288e+02, percent-clipped=8.0 2023-03-08 15:14:08,348 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3883, 4.5450, 4.5690, 4.5753, 4.6381, 4.6206, 4.3696, 4.1830], device='cuda:1'), covar=tensor([0.0991, 0.0664, 0.0357, 0.0498, 0.0350, 0.0352, 0.0394, 0.0352], device='cuda:1'), in_proj_covar=tensor([0.0493, 0.0335, 0.0310, 0.0328, 0.0390, 0.0405, 0.0331, 0.0368], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0004, 0.0004, 0.0005, 0.0003, 0.0004], device='cuda:1') 2023-03-08 15:14:22,625 INFO [train2.py:809] (1/4) Epoch 17, batch 1100, loss[ctc_loss=0.0747, att_loss=0.2428, loss=0.2092, over 16380.00 frames. utt_duration=1491 frames, utt_pad_proportion=0.008211, over 44.00 utterances.], tot_loss[ctc_loss=0.08364, att_loss=0.2409, loss=0.2095, over 3250436.04 frames. utt_duration=1231 frames, utt_pad_proportion=0.06021, over 10572.67 utterances.], batch size: 44, lr: 6.41e-03, grad_scale: 8.0 2023-03-08 15:15:37,893 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64888.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:15:42,103 INFO [train2.py:809] (1/4) Epoch 17, batch 1150, loss[ctc_loss=0.07602, att_loss=0.2358, loss=0.2039, over 17324.00 frames. utt_duration=878.7 frames, utt_pad_proportion=0.07986, over 79.00 utterances.], tot_loss[ctc_loss=0.08423, att_loss=0.2408, loss=0.2095, over 3250322.71 frames. utt_duration=1218 frames, utt_pad_proportion=0.06496, over 10691.48 utterances.], batch size: 79, lr: 6.41e-03, grad_scale: 8.0 2023-03-08 15:16:47,809 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.183e+02 2.059e+02 2.511e+02 2.874e+02 5.628e+02, threshold=5.021e+02, percent-clipped=1.0 2023-03-08 15:16:54,646 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=64936.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:17:02,370 INFO [train2.py:809] (1/4) Epoch 17, batch 1200, loss[ctc_loss=0.06786, att_loss=0.2228, loss=0.1918, over 15946.00 frames. utt_duration=1558 frames, utt_pad_proportion=0.006024, over 41.00 utterances.], tot_loss[ctc_loss=0.08351, att_loss=0.2398, loss=0.2086, over 3255508.20 frames. utt_duration=1243 frames, utt_pad_proportion=0.05843, over 10488.42 utterances.], batch size: 41, lr: 6.40e-03, grad_scale: 8.0 2023-03-08 15:17:05,278 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-03-08 15:17:09,400 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64945.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:17:27,156 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0810, 4.4507, 4.4962, 4.7512, 2.9675, 4.6781, 2.8069, 1.9421], device='cuda:1'), covar=tensor([0.0433, 0.0217, 0.0680, 0.0152, 0.1529, 0.0148, 0.1419, 0.1791], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0143, 0.0259, 0.0137, 0.0219, 0.0124, 0.0232, 0.0207], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 15:17:58,151 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64976.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:18:22,344 INFO [train2.py:809] (1/4) Epoch 17, batch 1250, loss[ctc_loss=0.09988, att_loss=0.2564, loss=0.2251, over 16621.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.005663, over 47.00 utterances.], tot_loss[ctc_loss=0.08304, att_loss=0.2403, loss=0.2088, over 3269644.25 frames. utt_duration=1262 frames, utt_pad_proportion=0.05048, over 10372.09 utterances.], batch size: 47, lr: 6.40e-03, grad_scale: 8.0 2023-03-08 15:18:39,096 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65001.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:18:44,343 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65004.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:19:27,888 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.350e+02 2.065e+02 2.436e+02 3.098e+02 5.553e+02, threshold=4.871e+02, percent-clipped=2.0 2023-03-08 15:19:36,260 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65037.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 15:19:41,903 INFO [train2.py:809] (1/4) Epoch 17, batch 1300, loss[ctc_loss=0.08734, att_loss=0.2514, loss=0.2186, over 17045.00 frames. utt_duration=690.2 frames, utt_pad_proportion=0.134, over 99.00 utterances.], tot_loss[ctc_loss=0.08341, att_loss=0.2413, loss=0.2097, over 3275639.60 frames. utt_duration=1245 frames, utt_pad_proportion=0.05286, over 10539.16 utterances.], batch size: 99, lr: 6.40e-03, grad_scale: 8.0 2023-03-08 15:19:48,518 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-03-08 15:19:54,984 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=65049.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:20:05,123 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65055.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:20:11,138 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65059.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:21:01,328 INFO [train2.py:809] (1/4) Epoch 17, batch 1350, loss[ctc_loss=0.05708, att_loss=0.2113, loss=0.1805, over 15501.00 frames. utt_duration=1724 frames, utt_pad_proportion=0.008849, over 36.00 utterances.], tot_loss[ctc_loss=0.08207, att_loss=0.2401, loss=0.2085, over 3278323.80 frames. utt_duration=1277 frames, utt_pad_proportion=0.0454, over 10277.01 utterances.], batch size: 36, lr: 6.40e-03, grad_scale: 8.0 2023-03-08 15:21:10,120 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7879, 2.0235, 2.3852, 2.7397, 2.6215, 2.6369, 2.4892, 2.9032], device='cuda:1'), covar=tensor([0.1460, 0.3863, 0.2786, 0.1331, 0.2246, 0.1308, 0.2211, 0.1178], device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0108, 0.0110, 0.0096, 0.0102, 0.0091, 0.0110, 0.0080], device='cuda:1'), out_proj_covar=tensor([7.2524e-05, 8.1372e-05, 8.3642e-05, 7.2117e-05, 7.4556e-05, 7.1440e-05, 8.0880e-05, 6.4123e-05], device='cuda:1') 2023-03-08 15:21:18,505 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65101.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:21:23,004 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.8064, 5.9933, 5.4325, 5.7383, 5.7044, 5.1949, 5.3881, 5.1740], device='cuda:1'), covar=tensor([0.1140, 0.0907, 0.0873, 0.0749, 0.0770, 0.1399, 0.2224, 0.2394], device='cuda:1'), in_proj_covar=tensor([0.0501, 0.0574, 0.0434, 0.0434, 0.0418, 0.0459, 0.0589, 0.0515], 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') 2023-03-08 15:21:29,272 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65108.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:21:41,874 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65116.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:21:49,428 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65121.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:22:07,570 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.334e+02 2.057e+02 2.538e+02 3.006e+02 4.389e+02, threshold=5.076e+02, percent-clipped=0.0 2023-03-08 15:22:09,387 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2989, 4.6659, 4.5498, 4.6047, 4.7474, 4.3848, 3.2770, 4.6377], device='cuda:1'), covar=tensor([0.0122, 0.0132, 0.0137, 0.0104, 0.0125, 0.0140, 0.0742, 0.0225], device='cuda:1'), in_proj_covar=tensor([0.0085, 0.0083, 0.0102, 0.0065, 0.0069, 0.0081, 0.0100, 0.0104], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 15:22:22,265 INFO [train2.py:809] (1/4) Epoch 17, batch 1400, loss[ctc_loss=0.08609, att_loss=0.2452, loss=0.2133, over 16933.00 frames. utt_duration=685.6 frames, utt_pad_proportion=0.1365, over 99.00 utterances.], tot_loss[ctc_loss=0.0821, att_loss=0.2412, loss=0.2094, over 3284130.79 frames. utt_duration=1261 frames, utt_pad_proportion=0.04837, over 10432.20 utterances.], batch size: 99, lr: 6.39e-03, grad_scale: 8.0 2023-03-08 15:23:06,280 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=65169.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:23:41,175 INFO [train2.py:809] (1/4) Epoch 17, batch 1450, loss[ctc_loss=0.06603, att_loss=0.2097, loss=0.181, over 15642.00 frames. utt_duration=1693 frames, utt_pad_proportion=0.008471, over 37.00 utterances.], tot_loss[ctc_loss=0.08313, att_loss=0.2412, loss=0.2096, over 3284596.89 frames. utt_duration=1260 frames, utt_pad_proportion=0.04926, over 10442.15 utterances.], batch size: 37, lr: 6.39e-03, grad_scale: 8.0 2023-03-08 15:23:51,009 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.8941, 5.1538, 5.4270, 5.3553, 5.3426, 5.8553, 5.0989, 5.9310], device='cuda:1'), covar=tensor([0.0725, 0.0731, 0.0836, 0.1195, 0.2029, 0.0818, 0.0749, 0.0626], device='cuda:1'), in_proj_covar=tensor([0.0796, 0.0470, 0.0559, 0.0617, 0.0815, 0.0568, 0.0454, 0.0553], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 15:24:46,651 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.514e+02 2.067e+02 2.518e+02 3.214e+02 7.164e+02, threshold=5.035e+02, percent-clipped=4.0 2023-03-08 15:25:01,185 INFO [train2.py:809] (1/4) Epoch 17, batch 1500, loss[ctc_loss=0.07319, att_loss=0.2424, loss=0.2085, over 17040.00 frames. utt_duration=1338 frames, utt_pad_proportion=0.007447, over 51.00 utterances.], tot_loss[ctc_loss=0.08434, att_loss=0.2424, loss=0.2108, over 3277343.66 frames. utt_duration=1203 frames, utt_pad_proportion=0.06534, over 10909.05 utterances.], batch size: 51, lr: 6.39e-03, grad_scale: 8.0 2023-03-08 15:25:04,068 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-03-08 15:25:04,673 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4788, 4.5986, 4.6771, 4.4351, 5.2341, 4.6195, 4.5644, 2.6081], device='cuda:1'), covar=tensor([0.0222, 0.0266, 0.0248, 0.0290, 0.0554, 0.0184, 0.0280, 0.1734], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0159, 0.0165, 0.0179, 0.0360, 0.0139, 0.0150, 0.0216], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 15:25:08,250 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65245.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:26:11,699 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65285.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:26:21,180 INFO [train2.py:809] (1/4) Epoch 17, batch 1550, loss[ctc_loss=0.08356, att_loss=0.2434, loss=0.2115, over 17282.00 frames. utt_duration=1003 frames, utt_pad_proportion=0.05351, over 69.00 utterances.], tot_loss[ctc_loss=0.08406, att_loss=0.2423, loss=0.2107, over 3280644.32 frames. utt_duration=1218 frames, utt_pad_proportion=0.0605, over 10789.24 utterances.], batch size: 69, lr: 6.39e-03, grad_scale: 8.0 2023-03-08 15:26:24,402 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=65293.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:26:27,685 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.58 vs. limit=5.0 2023-03-08 15:26:42,551 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65304.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:26:55,260 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.5573, 1.9870, 2.1238, 2.7211, 2.5688, 2.6405, 2.3139, 2.7296], device='cuda:1'), covar=tensor([0.1801, 0.4342, 0.3447, 0.1567, 0.2164, 0.1588, 0.3255, 0.1279], device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0107, 0.0111, 0.0096, 0.0102, 0.0090, 0.0110, 0.0080], device='cuda:1'), out_proj_covar=tensor([7.2441e-05, 8.0937e-05, 8.4175e-05, 7.2117e-05, 7.4563e-05, 7.0928e-05, 8.0666e-05, 6.3976e-05], device='cuda:1') 2023-03-08 15:27:26,906 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.45 vs. limit=5.0 2023-03-08 15:27:27,594 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.558e+02 2.050e+02 2.369e+02 2.748e+02 8.749e+02, threshold=4.738e+02, percent-clipped=3.0 2023-03-08 15:27:27,830 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65332.0, num_to_drop=1, layers_to_drop={3} 2023-03-08 15:27:41,714 INFO [train2.py:809] (1/4) Epoch 17, batch 1600, loss[ctc_loss=0.1023, att_loss=0.2599, loss=0.2284, over 17369.00 frames. utt_duration=1008 frames, utt_pad_proportion=0.04877, over 69.00 utterances.], tot_loss[ctc_loss=0.08337, att_loss=0.2415, loss=0.2099, over 3278750.35 frames. utt_duration=1233 frames, utt_pad_proportion=0.05679, over 10651.60 utterances.], batch size: 69, lr: 6.38e-03, grad_scale: 8.0 2023-03-08 15:27:50,393 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.8044, 2.0543, 2.2928, 2.7854, 2.7681, 2.7145, 2.4526, 2.9068], device='cuda:1'), covar=tensor([0.1297, 0.3342, 0.2616, 0.1125, 0.1410, 0.1315, 0.2263, 0.0877], device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0107, 0.0110, 0.0095, 0.0102, 0.0089, 0.0109, 0.0079], device='cuda:1'), out_proj_covar=tensor([7.1823e-05, 8.0395e-05, 8.3475e-05, 7.1367e-05, 7.4061e-05, 7.0497e-05, 7.9979e-05, 6.3306e-05], device='cuda:1') 2023-03-08 15:27:50,400 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65346.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:27:54,197 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3821, 3.0059, 3.5477, 2.9493, 3.4106, 4.5429, 4.3424, 2.9165], device='cuda:1'), covar=tensor([0.0412, 0.1721, 0.1260, 0.1499, 0.1165, 0.0762, 0.0547, 0.1590], device='cuda:1'), in_proj_covar=tensor([0.0241, 0.0239, 0.0268, 0.0213, 0.0256, 0.0349, 0.0248, 0.0231], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 15:28:00,052 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=65352.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:28:10,886 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65359.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:29:02,420 INFO [train2.py:809] (1/4) Epoch 17, batch 1650, loss[ctc_loss=0.08962, att_loss=0.2541, loss=0.2212, over 17116.00 frames. utt_duration=1224 frames, utt_pad_proportion=0.01543, over 56.00 utterances.], tot_loss[ctc_loss=0.08299, att_loss=0.2413, loss=0.2096, over 3286925.12 frames. utt_duration=1260 frames, utt_pad_proportion=0.04877, over 10450.57 utterances.], batch size: 56, lr: 6.38e-03, grad_scale: 8.0 2023-03-08 15:29:20,652 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65401.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:29:29,995 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=65407.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:29:31,710 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65408.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:29:36,209 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65411.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:30:06,595 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.1066, 5.3726, 5.7012, 5.4358, 5.5165, 6.0576, 5.1986, 6.1908], device='cuda:1'), covar=tensor([0.0701, 0.0690, 0.0733, 0.1184, 0.1903, 0.0834, 0.0660, 0.0579], device='cuda:1'), in_proj_covar=tensor([0.0807, 0.0474, 0.0565, 0.0622, 0.0827, 0.0575, 0.0461, 0.0557], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 15:30:09,495 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.374e+02 2.074e+02 2.531e+02 3.179e+02 8.424e+02, threshold=5.062e+02, percent-clipped=4.0 2023-03-08 15:30:23,566 INFO [train2.py:809] (1/4) Epoch 17, batch 1700, loss[ctc_loss=0.0755, att_loss=0.2284, loss=0.1978, over 15869.00 frames. utt_duration=1629 frames, utt_pad_proportion=0.01025, over 39.00 utterances.], tot_loss[ctc_loss=0.08288, att_loss=0.241, loss=0.2094, over 3288680.43 frames. utt_duration=1276 frames, utt_pad_proportion=0.04409, over 10319.97 utterances.], batch size: 39, lr: 6.38e-03, grad_scale: 8.0 2023-03-08 15:30:37,509 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=65449.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:30:48,522 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=65456.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:31:44,386 INFO [train2.py:809] (1/4) Epoch 17, batch 1750, loss[ctc_loss=0.09542, att_loss=0.2618, loss=0.2285, over 17265.00 frames. utt_duration=1172 frames, utt_pad_proportion=0.02483, over 59.00 utterances.], tot_loss[ctc_loss=0.0834, att_loss=0.2414, loss=0.2098, over 3284180.83 frames. utt_duration=1243 frames, utt_pad_proportion=0.05229, over 10584.54 utterances.], batch size: 59, lr: 6.38e-03, grad_scale: 8.0 2023-03-08 15:32:40,963 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.9385, 2.3067, 2.7808, 2.7228, 2.9379, 2.9765, 2.5714, 2.8761], device='cuda:1'), covar=tensor([0.1305, 0.3082, 0.2104, 0.1477, 0.1289, 0.1191, 0.2120, 0.1130], device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0105, 0.0109, 0.0094, 0.0101, 0.0088, 0.0108, 0.0078], device='cuda:1'), out_proj_covar=tensor([7.1475e-05, 7.9506e-05, 8.2890e-05, 7.0911e-05, 7.3632e-05, 6.9915e-05, 7.9262e-05, 6.2688e-05], device='cuda:1') 2023-03-08 15:32:49,893 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.165e+02 2.005e+02 2.493e+02 3.105e+02 6.516e+02, threshold=4.986e+02, percent-clipped=3.0 2023-03-08 15:32:50,242 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65532.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 15:33:05,037 INFO [train2.py:809] (1/4) Epoch 17, batch 1800, loss[ctc_loss=0.1001, att_loss=0.2649, loss=0.2319, over 17052.00 frames. utt_duration=1288 frames, utt_pad_proportion=0.009862, over 53.00 utterances.], tot_loss[ctc_loss=0.08401, att_loss=0.2417, loss=0.2102, over 3275772.50 frames. utt_duration=1216 frames, utt_pad_proportion=0.06068, over 10789.82 utterances.], batch size: 53, lr: 6.37e-03, grad_scale: 8.0 2023-03-08 15:34:26,298 INFO [train2.py:809] (1/4) Epoch 17, batch 1850, loss[ctc_loss=0.06649, att_loss=0.243, loss=0.2077, over 16869.00 frames. utt_duration=1378 frames, utt_pad_proportion=0.007714, over 49.00 utterances.], tot_loss[ctc_loss=0.0823, att_loss=0.2406, loss=0.2089, over 3281046.79 frames. utt_duration=1251 frames, utt_pad_proportion=0.05196, over 10504.13 utterances.], batch size: 49, lr: 6.37e-03, grad_scale: 8.0 2023-03-08 15:34:29,963 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65593.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 15:35:31,591 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.506e+02 1.987e+02 2.376e+02 2.824e+02 5.455e+02, threshold=4.753e+02, percent-clipped=1.0 2023-03-08 15:35:31,974 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65632.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:35:46,840 INFO [train2.py:809] (1/4) Epoch 17, batch 1900, loss[ctc_loss=0.06532, att_loss=0.2284, loss=0.1958, over 16192.00 frames. utt_duration=1581 frames, utt_pad_proportion=0.005979, over 41.00 utterances.], tot_loss[ctc_loss=0.08155, att_loss=0.2397, loss=0.2081, over 3284398.91 frames. utt_duration=1272 frames, utt_pad_proportion=0.04592, over 10338.01 utterances.], batch size: 41, lr: 6.37e-03, grad_scale: 8.0 2023-03-08 15:35:47,049 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65641.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:36:48,027 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=65680.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:37:05,992 INFO [train2.py:809] (1/4) Epoch 17, batch 1950, loss[ctc_loss=0.1105, att_loss=0.2556, loss=0.2265, over 17113.00 frames. utt_duration=1224 frames, utt_pad_proportion=0.01448, over 56.00 utterances.], tot_loss[ctc_loss=0.08086, att_loss=0.2394, loss=0.2077, over 3287268.21 frames. utt_duration=1292 frames, utt_pad_proportion=0.04029, over 10189.86 utterances.], batch size: 56, lr: 6.37e-03, grad_scale: 8.0 2023-03-08 15:37:15,899 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-03-08 15:37:36,757 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65711.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:38:10,050 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.350e+02 2.136e+02 2.531e+02 3.248e+02 6.819e+02, threshold=5.061e+02, percent-clipped=5.0 2023-03-08 15:38:25,203 INFO [train2.py:809] (1/4) Epoch 17, batch 2000, loss[ctc_loss=0.1028, att_loss=0.2558, loss=0.2252, over 17306.00 frames. utt_duration=1260 frames, utt_pad_proportion=0.01105, over 55.00 utterances.], tot_loss[ctc_loss=0.08153, att_loss=0.2397, loss=0.2081, over 3285529.06 frames. utt_duration=1290 frames, utt_pad_proportion=0.04063, over 10197.57 utterances.], batch size: 55, lr: 6.37e-03, grad_scale: 8.0 2023-03-08 15:38:44,060 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0022, 5.4105, 5.1712, 5.2714, 5.3413, 5.1131, 4.1562, 5.4140], device='cuda:1'), covar=tensor([0.0087, 0.0105, 0.0115, 0.0072, 0.0083, 0.0097, 0.0524, 0.0174], device='cuda:1'), in_proj_covar=tensor([0.0085, 0.0083, 0.0102, 0.0065, 0.0069, 0.0081, 0.0100, 0.0104], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 15:38:53,013 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=65759.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:39:44,926 INFO [train2.py:809] (1/4) Epoch 17, batch 2050, loss[ctc_loss=0.07945, att_loss=0.2371, loss=0.2055, over 16551.00 frames. utt_duration=1473 frames, utt_pad_proportion=0.005732, over 45.00 utterances.], tot_loss[ctc_loss=0.08149, att_loss=0.239, loss=0.2075, over 3267667.95 frames. utt_duration=1295 frames, utt_pad_proportion=0.0443, over 10108.11 utterances.], batch size: 45, lr: 6.36e-03, grad_scale: 8.0 2023-03-08 15:40:18,789 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6978, 1.9951, 5.0954, 4.0460, 2.9425, 4.3599, 4.8153, 4.7740], device='cuda:1'), covar=tensor([0.0182, 0.1773, 0.0123, 0.0793, 0.1672, 0.0205, 0.0103, 0.0194], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0247, 0.0170, 0.0316, 0.0270, 0.0202, 0.0152, 0.0180], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 15:40:51,677 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.094e+02 1.921e+02 2.500e+02 3.054e+02 5.987e+02, threshold=5.000e+02, percent-clipped=2.0 2023-03-08 15:41:01,837 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65838.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 15:41:06,586 INFO [train2.py:809] (1/4) Epoch 17, batch 2100, loss[ctc_loss=0.1034, att_loss=0.2488, loss=0.2198, over 17463.00 frames. utt_duration=885.8 frames, utt_pad_proportion=0.07241, over 79.00 utterances.], tot_loss[ctc_loss=0.08209, att_loss=0.2396, loss=0.2081, over 3267079.66 frames. utt_duration=1277 frames, utt_pad_proportion=0.04813, over 10245.27 utterances.], batch size: 79, lr: 6.36e-03, grad_scale: 8.0 2023-03-08 15:42:14,955 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.7746, 4.2758, 4.4896, 4.5061, 4.3512, 4.4762, 4.2696, 4.0664], device='cuda:1'), covar=tensor([0.1808, 0.1073, 0.0476, 0.0584, 0.0806, 0.0531, 0.0498, 0.0497], device='cuda:1'), in_proj_covar=tensor([0.0495, 0.0333, 0.0311, 0.0328, 0.0388, 0.0404, 0.0328, 0.0365], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 15:42:21,798 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65888.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 15:42:26,208 INFO [train2.py:809] (1/4) Epoch 17, batch 2150, loss[ctc_loss=0.08851, att_loss=0.235, loss=0.2057, over 15954.00 frames. utt_duration=1558 frames, utt_pad_proportion=0.007089, over 41.00 utterances.], tot_loss[ctc_loss=0.08372, att_loss=0.2412, loss=0.2097, over 3269737.91 frames. utt_duration=1208 frames, utt_pad_proportion=0.0644, over 10836.89 utterances.], batch size: 41, lr: 6.36e-03, grad_scale: 16.0 2023-03-08 15:42:38,840 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65899.0, num_to_drop=1, layers_to_drop={3} 2023-03-08 15:43:31,156 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.335e+02 2.187e+02 2.638e+02 3.352e+02 9.129e+02, threshold=5.276e+02, percent-clipped=7.0 2023-03-08 15:43:44,655 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.1408, 3.6931, 3.2246, 3.4076, 4.0022, 3.6159, 3.0260, 4.2232], device='cuda:1'), covar=tensor([0.0972, 0.0597, 0.1049, 0.0698, 0.0649, 0.0722, 0.0844, 0.0521], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0210, 0.0222, 0.0194, 0.0268, 0.0235, 0.0197, 0.0279], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 15:43:45,840 INFO [train2.py:809] (1/4) Epoch 17, batch 2200, loss[ctc_loss=0.08776, att_loss=0.2586, loss=0.2245, over 16872.00 frames. utt_duration=1379 frames, utt_pad_proportion=0.006675, over 49.00 utterances.], tot_loss[ctc_loss=0.08325, att_loss=0.2417, loss=0.21, over 3271989.44 frames. utt_duration=1209 frames, utt_pad_proportion=0.06501, over 10835.55 utterances.], batch size: 49, lr: 6.36e-03, grad_scale: 16.0 2023-03-08 15:43:46,122 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65941.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:44:42,654 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7881, 3.6366, 3.1743, 3.3995, 3.9382, 3.5810, 2.9142, 4.1392], device='cuda:1'), covar=tensor([0.1124, 0.0557, 0.1027, 0.0667, 0.0600, 0.0677, 0.0889, 0.0471], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0210, 0.0221, 0.0193, 0.0266, 0.0234, 0.0196, 0.0278], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 15:45:01,904 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=65989.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:45:04,875 INFO [train2.py:809] (1/4) Epoch 17, batch 2250, loss[ctc_loss=0.07624, att_loss=0.2156, loss=0.1877, over 15628.00 frames. utt_duration=1691 frames, utt_pad_proportion=0.009937, over 37.00 utterances.], tot_loss[ctc_loss=0.08291, att_loss=0.2417, loss=0.21, over 3278568.83 frames. utt_duration=1230 frames, utt_pad_proportion=0.05718, over 10674.45 utterances.], batch size: 37, lr: 6.35e-03, grad_scale: 16.0 2023-03-08 15:45:33,702 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4162, 5.1303, 5.2709, 5.2675, 5.1353, 5.1841, 5.0317, 4.7318], device='cuda:1'), covar=tensor([0.1917, 0.0771, 0.0356, 0.0453, 0.0682, 0.0473, 0.0433, 0.0397], device='cuda:1'), in_proj_covar=tensor([0.0496, 0.0333, 0.0311, 0.0328, 0.0391, 0.0404, 0.0329, 0.0365], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 15:46:12,738 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.416e+02 2.093e+02 2.407e+02 3.008e+02 8.516e+02, threshold=4.813e+02, percent-clipped=4.0 2023-03-08 15:46:27,301 INFO [train2.py:809] (1/4) Epoch 17, batch 2300, loss[ctc_loss=0.0901, att_loss=0.2514, loss=0.2191, over 16878.00 frames. utt_duration=1379 frames, utt_pad_proportion=0.007576, over 49.00 utterances.], tot_loss[ctc_loss=0.08296, att_loss=0.2418, loss=0.21, over 3279133.54 frames. utt_duration=1227 frames, utt_pad_proportion=0.05792, over 10704.59 utterances.], batch size: 49, lr: 6.35e-03, grad_scale: 8.0 2023-03-08 15:47:48,109 INFO [train2.py:809] (1/4) Epoch 17, batch 2350, loss[ctc_loss=0.07461, att_loss=0.2378, loss=0.2052, over 17052.00 frames. utt_duration=690.4 frames, utt_pad_proportion=0.1349, over 99.00 utterances.], tot_loss[ctc_loss=0.08227, att_loss=0.2406, loss=0.209, over 3277617.86 frames. utt_duration=1241 frames, utt_pad_proportion=0.05467, over 10577.14 utterances.], batch size: 99, lr: 6.35e-03, grad_scale: 8.0 2023-03-08 15:48:41,639 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2679, 2.4313, 3.0112, 2.5547, 3.0532, 3.4894, 3.4031, 2.6416], device='cuda:1'), covar=tensor([0.0535, 0.1712, 0.1265, 0.1240, 0.0982, 0.1260, 0.0696, 0.1373], device='cuda:1'), in_proj_covar=tensor([0.0242, 0.0240, 0.0270, 0.0213, 0.0258, 0.0352, 0.0250, 0.0232], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 15:48:55,984 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.531e+02 2.214e+02 2.683e+02 3.352e+02 8.000e+02, threshold=5.366e+02, percent-clipped=7.0 2023-03-08 15:49:06,445 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.6779, 3.7690, 3.1922, 3.1341, 3.8610, 3.5210, 2.5408, 4.0658], device='cuda:1'), covar=tensor([0.1444, 0.0571, 0.1130, 0.1022, 0.0975, 0.0862, 0.1387, 0.0744], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0211, 0.0223, 0.0195, 0.0268, 0.0236, 0.0197, 0.0280], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 15:49:09,129 INFO [train2.py:809] (1/4) Epoch 17, batch 2400, loss[ctc_loss=0.07355, att_loss=0.2468, loss=0.2121, over 16888.00 frames. utt_duration=1380 frames, utt_pad_proportion=0.007106, over 49.00 utterances.], tot_loss[ctc_loss=0.0828, att_loss=0.2409, loss=0.2093, over 3276482.74 frames. utt_duration=1213 frames, utt_pad_proportion=0.06203, over 10816.38 utterances.], batch size: 49, lr: 6.35e-03, grad_scale: 8.0 2023-03-08 15:49:14,398 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66144.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:49:39,851 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66160.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 15:50:24,755 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66188.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 15:50:29,106 INFO [train2.py:809] (1/4) Epoch 17, batch 2450, loss[ctc_loss=0.07385, att_loss=0.2292, loss=0.1981, over 16123.00 frames. utt_duration=1537 frames, utt_pad_proportion=0.00668, over 42.00 utterances.], tot_loss[ctc_loss=0.08298, att_loss=0.2408, loss=0.2092, over 3275334.45 frames. utt_duration=1194 frames, utt_pad_proportion=0.06759, over 10983.77 utterances.], batch size: 42, lr: 6.34e-03, grad_scale: 8.0 2023-03-08 15:50:33,899 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66194.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 15:50:40,097 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.3259, 2.9952, 3.6187, 4.5813, 3.9423, 4.1082, 3.1130, 2.4892], device='cuda:1'), covar=tensor([0.0711, 0.1951, 0.0835, 0.0471, 0.0812, 0.0391, 0.1325, 0.2018], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0215, 0.0188, 0.0207, 0.0213, 0.0170, 0.0197, 0.0183], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 15:50:51,797 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66205.0, num_to_drop=1, layers_to_drop={3} 2023-03-08 15:51:16,849 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66221.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 15:51:35,450 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.379e+02 1.994e+02 2.398e+02 2.990e+02 4.880e+02, threshold=4.796e+02, percent-clipped=0.0 2023-03-08 15:51:40,272 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=66236.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 15:51:48,521 INFO [train2.py:809] (1/4) Epoch 17, batch 2500, loss[ctc_loss=0.05564, att_loss=0.2268, loss=0.1926, over 16303.00 frames. utt_duration=1518 frames, utt_pad_proportion=0.005955, over 43.00 utterances.], tot_loss[ctc_loss=0.08296, att_loss=0.2407, loss=0.2092, over 3280753.37 frames. utt_duration=1211 frames, utt_pad_proportion=0.06088, over 10848.79 utterances.], batch size: 43, lr: 6.34e-03, grad_scale: 8.0 2023-03-08 15:52:16,098 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66258.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:52:30,584 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6573, 4.5613, 4.6704, 4.5365, 5.2006, 4.4843, 4.5374, 2.4280], device='cuda:1'), covar=tensor([0.0191, 0.0304, 0.0240, 0.0266, 0.0643, 0.0228, 0.0287, 0.1965], device='cuda:1'), in_proj_covar=tensor([0.0139, 0.0156, 0.0162, 0.0177, 0.0351, 0.0137, 0.0147, 0.0213], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 15:52:48,543 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-03-08 15:53:03,917 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.2658, 5.1577, 4.7909, 3.0358, 4.8640, 4.7728, 4.2488, 2.7649], device='cuda:1'), covar=tensor([0.0095, 0.0098, 0.0304, 0.1059, 0.0106, 0.0188, 0.0378, 0.1519], device='cuda:1'), in_proj_covar=tensor([0.0071, 0.0097, 0.0095, 0.0110, 0.0081, 0.0107, 0.0097, 0.0102], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 15:53:08,203 INFO [train2.py:809] (1/4) Epoch 17, batch 2550, loss[ctc_loss=0.1048, att_loss=0.2592, loss=0.2283, over 17394.00 frames. utt_duration=1106 frames, utt_pad_proportion=0.03315, over 63.00 utterances.], tot_loss[ctc_loss=0.08239, att_loss=0.24, loss=0.2085, over 3281135.62 frames. utt_duration=1235 frames, utt_pad_proportion=0.05571, over 10642.54 utterances.], batch size: 63, lr: 6.34e-03, grad_scale: 8.0 2023-03-08 15:53:13,163 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.5318, 3.0864, 2.6794, 2.8670, 3.1671, 3.0374, 2.4862, 3.0572], device='cuda:1'), covar=tensor([0.0861, 0.0385, 0.0851, 0.0576, 0.0666, 0.0559, 0.0810, 0.0496], device='cuda:1'), in_proj_covar=tensor([0.0199, 0.0211, 0.0222, 0.0194, 0.0267, 0.0235, 0.0196, 0.0278], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 15:53:44,979 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.1591, 3.8305, 3.3511, 3.5605, 4.0891, 3.7650, 3.0882, 4.4374], device='cuda:1'), covar=tensor([0.0960, 0.0539, 0.0956, 0.0628, 0.0646, 0.0679, 0.0860, 0.0440], device='cuda:1'), in_proj_covar=tensor([0.0199, 0.0210, 0.0222, 0.0194, 0.0267, 0.0234, 0.0196, 0.0279], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 15:53:52,743 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66319.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 15:54:14,464 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.777e+02 2.284e+02 2.590e+02 3.284e+02 7.249e+02, threshold=5.180e+02, percent-clipped=6.0 2023-03-08 15:54:27,434 INFO [train2.py:809] (1/4) Epoch 17, batch 2600, loss[ctc_loss=0.07536, att_loss=0.2314, loss=0.2002, over 16175.00 frames. utt_duration=1580 frames, utt_pad_proportion=0.005925, over 41.00 utterances.], tot_loss[ctc_loss=0.08228, att_loss=0.2401, loss=0.2086, over 3279926.33 frames. utt_duration=1234 frames, utt_pad_proportion=0.05589, over 10645.39 utterances.], batch size: 41, lr: 6.34e-03, grad_scale: 8.0 2023-03-08 15:55:02,916 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3824, 4.8048, 4.6057, 4.7172, 4.8684, 4.4589, 3.5067, 4.7347], device='cuda:1'), covar=tensor([0.0129, 0.0108, 0.0141, 0.0092, 0.0092, 0.0122, 0.0640, 0.0207], device='cuda:1'), in_proj_covar=tensor([0.0086, 0.0083, 0.0104, 0.0065, 0.0069, 0.0081, 0.0099, 0.0104], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 15:55:47,242 INFO [train2.py:809] (1/4) Epoch 17, batch 2650, loss[ctc_loss=0.05967, att_loss=0.2194, loss=0.1875, over 15384.00 frames. utt_duration=1759 frames, utt_pad_proportion=0.01105, over 35.00 utterances.], tot_loss[ctc_loss=0.08141, att_loss=0.2393, loss=0.2077, over 3280384.42 frames. utt_duration=1244 frames, utt_pad_proportion=0.05312, over 10563.99 utterances.], batch size: 35, lr: 6.33e-03, grad_scale: 8.0 2023-03-08 15:56:49,121 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.5681, 1.7084, 2.3079, 2.4345, 2.1505, 2.1555, 1.8940, 2.2720], device='cuda:1'), covar=tensor([0.0974, 0.2375, 0.2086, 0.1143, 0.1632, 0.0988, 0.1375, 0.0761], device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0103, 0.0107, 0.0092, 0.0100, 0.0088, 0.0106, 0.0078], device='cuda:1'), out_proj_covar=tensor([7.0850e-05, 7.8392e-05, 8.1945e-05, 6.9695e-05, 7.3016e-05, 6.9456e-05, 7.8636e-05, 6.2608e-05], device='cuda:1') 2023-03-08 15:56:53,520 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.354e+02 2.144e+02 2.549e+02 3.292e+02 5.806e+02, threshold=5.098e+02, percent-clipped=4.0 2023-03-08 15:57:06,279 INFO [train2.py:809] (1/4) Epoch 17, batch 2700, loss[ctc_loss=0.06356, att_loss=0.2349, loss=0.2006, over 16542.00 frames. utt_duration=1472 frames, utt_pad_proportion=0.006317, over 45.00 utterances.], tot_loss[ctc_loss=0.08265, att_loss=0.2398, loss=0.2084, over 3278933.48 frames. utt_duration=1232 frames, utt_pad_proportion=0.05582, over 10662.32 utterances.], batch size: 45, lr: 6.33e-03, grad_scale: 8.0 2023-03-08 15:57:23,565 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.94 vs. limit=5.0 2023-03-08 15:58:25,423 INFO [train2.py:809] (1/4) Epoch 17, batch 2750, loss[ctc_loss=0.09142, att_loss=0.2425, loss=0.2123, over 16532.00 frames. utt_duration=1471 frames, utt_pad_proportion=0.006632, over 45.00 utterances.], tot_loss[ctc_loss=0.08293, att_loss=0.2398, loss=0.2084, over 3274018.94 frames. utt_duration=1243 frames, utt_pad_proportion=0.05515, over 10550.82 utterances.], batch size: 45, lr: 6.33e-03, grad_scale: 8.0 2023-03-08 15:58:30,325 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66494.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 15:58:40,078 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66500.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 15:59:05,224 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66516.0, num_to_drop=1, layers_to_drop={3} 2023-03-08 15:59:12,746 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66521.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 15:59:31,458 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.407e+02 2.169e+02 2.491e+02 3.348e+02 1.363e+03, threshold=4.982e+02, percent-clipped=6.0 2023-03-08 15:59:43,570 INFO [train2.py:809] (1/4) Epoch 17, batch 2800, loss[ctc_loss=0.08984, att_loss=0.2423, loss=0.2118, over 16274.00 frames. utt_duration=1516 frames, utt_pad_proportion=0.007447, over 43.00 utterances.], tot_loss[ctc_loss=0.08247, att_loss=0.2397, loss=0.2083, over 3276878.43 frames. utt_duration=1262 frames, utt_pad_proportion=0.0504, over 10397.83 utterances.], batch size: 43, lr: 6.33e-03, grad_scale: 8.0 2023-03-08 15:59:45,180 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=66542.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 16:00:00,060 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3612, 4.2969, 4.3070, 4.3799, 4.9116, 4.2280, 4.4303, 2.4845], device='cuda:1'), covar=tensor([0.0253, 0.0398, 0.0355, 0.0257, 0.0783, 0.0267, 0.0269, 0.1916], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0159, 0.0166, 0.0179, 0.0356, 0.0140, 0.0149, 0.0215], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-08 16:00:48,788 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66582.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 16:01:02,313 INFO [train2.py:809] (1/4) Epoch 17, batch 2850, loss[ctc_loss=0.1368, att_loss=0.2696, loss=0.243, over 14052.00 frames. utt_duration=386.5 frames, utt_pad_proportion=0.3278, over 146.00 utterances.], tot_loss[ctc_loss=0.08278, att_loss=0.2393, loss=0.208, over 3269300.47 frames. utt_duration=1247 frames, utt_pad_proportion=0.05567, over 10498.78 utterances.], batch size: 146, lr: 6.32e-03, grad_scale: 8.0 2023-03-08 16:01:39,561 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66614.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 16:02:09,210 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.516e+02 2.066e+02 2.507e+02 3.083e+02 5.054e+02, threshold=5.014e+02, percent-clipped=1.0 2023-03-08 16:02:21,165 INFO [train2.py:809] (1/4) Epoch 17, batch 2900, loss[ctc_loss=0.1087, att_loss=0.2709, loss=0.2385, over 17077.00 frames. utt_duration=1315 frames, utt_pad_proportion=0.007576, over 52.00 utterances.], tot_loss[ctc_loss=0.08225, att_loss=0.2394, loss=0.2079, over 3271113.36 frames. utt_duration=1252 frames, utt_pad_proportion=0.05494, over 10463.17 utterances.], batch size: 52, lr: 6.32e-03, grad_scale: 8.0 2023-03-08 16:03:21,185 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-03-08 16:03:38,263 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.22 vs. limit=5.0 2023-03-08 16:03:41,160 INFO [train2.py:809] (1/4) Epoch 17, batch 2950, loss[ctc_loss=0.09644, att_loss=0.2556, loss=0.2237, over 16483.00 frames. utt_duration=1435 frames, utt_pad_proportion=0.005484, over 46.00 utterances.], tot_loss[ctc_loss=0.08171, att_loss=0.2391, loss=0.2076, over 3274627.67 frames. utt_duration=1289 frames, utt_pad_proportion=0.04596, over 10170.43 utterances.], batch size: 46, lr: 6.32e-03, grad_scale: 8.0 2023-03-08 16:03:58,823 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.1382, 3.7536, 3.2725, 3.3850, 4.0131, 3.6570, 3.0552, 4.3435], device='cuda:1'), covar=tensor([0.0877, 0.0472, 0.0916, 0.0673, 0.0658, 0.0673, 0.0793, 0.0468], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0211, 0.0222, 0.0195, 0.0267, 0.0235, 0.0196, 0.0279], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 16:04:44,860 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66731.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 16:04:47,518 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.385e+02 2.044e+02 2.450e+02 3.054e+02 6.191e+02, threshold=4.900e+02, percent-clipped=3.0 2023-03-08 16:05:00,736 INFO [train2.py:809] (1/4) Epoch 17, batch 3000, loss[ctc_loss=0.1057, att_loss=0.2581, loss=0.2276, over 17142.00 frames. utt_duration=1226 frames, utt_pad_proportion=0.01373, over 56.00 utterances.], tot_loss[ctc_loss=0.08247, att_loss=0.24, loss=0.2085, over 3278601.12 frames. utt_duration=1250 frames, utt_pad_proportion=0.05342, over 10508.00 utterances.], batch size: 56, lr: 6.32e-03, grad_scale: 8.0 2023-03-08 16:05:00,736 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-08 16:05:14,971 INFO [train2.py:843] (1/4) Epoch 17, validation: ctc_loss=0.04199, att_loss=0.2349, loss=0.1964, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-08 16:05:14,972 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-08 16:05:45,544 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0476, 5.3290, 4.8221, 5.3812, 4.7414, 5.0873, 5.4404, 5.2302], device='cuda:1'), covar=tensor([0.0576, 0.0296, 0.0770, 0.0242, 0.0377, 0.0196, 0.0224, 0.0204], device='cuda:1'), in_proj_covar=tensor([0.0374, 0.0298, 0.0348, 0.0313, 0.0302, 0.0223, 0.0282, 0.0265], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 16:06:34,014 INFO [train2.py:809] (1/4) Epoch 17, batch 3050, loss[ctc_loss=0.07833, att_loss=0.2403, loss=0.2079, over 16113.00 frames. utt_duration=1536 frames, utt_pad_proportion=0.006834, over 42.00 utterances.], tot_loss[ctc_loss=0.08325, att_loss=0.2407, loss=0.2092, over 3285741.97 frames. utt_duration=1247 frames, utt_pad_proportion=0.05183, over 10551.03 utterances.], batch size: 42, lr: 6.32e-03, grad_scale: 4.0 2023-03-08 16:06:35,925 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66792.0, num_to_drop=1, layers_to_drop={3} 2023-03-08 16:06:49,684 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66800.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 16:07:14,775 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66816.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 16:07:27,047 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.3338, 2.7538, 3.0199, 4.4129, 3.9162, 3.9621, 2.8453, 2.2242], device='cuda:1'), covar=tensor([0.0724, 0.2165, 0.1086, 0.0590, 0.0822, 0.0447, 0.1598, 0.2307], device='cuda:1'), in_proj_covar=tensor([0.0178, 0.0215, 0.0189, 0.0209, 0.0213, 0.0172, 0.0198, 0.0185], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 16:07:42,430 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.427e+02 2.123e+02 2.532e+02 3.472e+02 1.749e+03, threshold=5.065e+02, percent-clipped=12.0 2023-03-08 16:07:53,807 INFO [train2.py:809] (1/4) Epoch 17, batch 3100, loss[ctc_loss=0.07413, att_loss=0.2185, loss=0.1896, over 15385.00 frames. utt_duration=1760 frames, utt_pad_proportion=0.0102, over 35.00 utterances.], tot_loss[ctc_loss=0.08411, att_loss=0.2409, loss=0.2095, over 3284103.59 frames. utt_duration=1228 frames, utt_pad_proportion=0.05662, over 10710.07 utterances.], batch size: 35, lr: 6.31e-03, grad_scale: 4.0 2023-03-08 16:08:05,847 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=66848.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 16:08:30,874 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=66864.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 16:08:50,683 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66877.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 16:09:12,471 INFO [train2.py:809] (1/4) Epoch 17, batch 3150, loss[ctc_loss=0.08239, att_loss=0.2517, loss=0.2179, over 16479.00 frames. utt_duration=1435 frames, utt_pad_proportion=0.006534, over 46.00 utterances.], tot_loss[ctc_loss=0.08396, att_loss=0.2409, loss=0.2095, over 3286304.58 frames. utt_duration=1224 frames, utt_pad_proportion=0.05693, over 10750.31 utterances.], batch size: 46, lr: 6.31e-03, grad_scale: 4.0 2023-03-08 16:09:17,881 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.47 vs. limit=5.0 2023-03-08 16:09:50,593 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66914.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 16:10:21,244 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.312e+02 2.012e+02 2.406e+02 2.883e+02 7.010e+02, threshold=4.812e+02, percent-clipped=2.0 2023-03-08 16:10:32,788 INFO [train2.py:809] (1/4) Epoch 17, batch 3200, loss[ctc_loss=0.06938, att_loss=0.22, loss=0.1899, over 16265.00 frames. utt_duration=1514 frames, utt_pad_proportion=0.006302, over 43.00 utterances.], tot_loss[ctc_loss=0.08323, att_loss=0.2404, loss=0.2089, over 3275934.33 frames. utt_duration=1212 frames, utt_pad_proportion=0.06419, over 10822.52 utterances.], batch size: 43, lr: 6.31e-03, grad_scale: 8.0 2023-03-08 16:10:43,131 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0681, 5.0821, 4.7810, 2.9437, 4.7596, 4.6904, 4.2964, 2.7108], device='cuda:1'), covar=tensor([0.0155, 0.0094, 0.0283, 0.1114, 0.0126, 0.0208, 0.0351, 0.1535], device='cuda:1'), in_proj_covar=tensor([0.0072, 0.0097, 0.0097, 0.0110, 0.0081, 0.0107, 0.0097, 0.0103], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 16:10:58,877 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66957.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 16:11:01,307 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-03-08 16:11:06,321 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=66962.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 16:11:51,564 INFO [train2.py:809] (1/4) Epoch 17, batch 3250, loss[ctc_loss=0.05901, att_loss=0.2108, loss=0.1804, over 15624.00 frames. utt_duration=1691 frames, utt_pad_proportion=0.00931, over 37.00 utterances.], tot_loss[ctc_loss=0.08279, att_loss=0.2401, loss=0.2086, over 3276812.35 frames. utt_duration=1238 frames, utt_pad_proportion=0.05784, over 10597.09 utterances.], batch size: 37, lr: 6.31e-03, grad_scale: 8.0 2023-03-08 16:12:30,866 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-03-08 16:12:34,782 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67018.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 16:12:58,912 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.388e+02 2.115e+02 2.445e+02 2.942e+02 7.309e+02, threshold=4.890e+02, percent-clipped=3.0 2023-03-08 16:13:11,486 INFO [train2.py:809] (1/4) Epoch 17, batch 3300, loss[ctc_loss=0.09893, att_loss=0.2493, loss=0.2192, over 15955.00 frames. utt_duration=1558 frames, utt_pad_proportion=0.006762, over 41.00 utterances.], tot_loss[ctc_loss=0.08293, att_loss=0.2402, loss=0.2087, over 3269499.62 frames. utt_duration=1229 frames, utt_pad_proportion=0.06165, over 10652.45 utterances.], batch size: 41, lr: 6.30e-03, grad_scale: 8.0 2023-03-08 16:14:17,410 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67083.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 16:14:24,019 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67087.0, num_to_drop=1, layers_to_drop={3} 2023-03-08 16:14:31,294 INFO [train2.py:809] (1/4) Epoch 17, batch 3350, loss[ctc_loss=0.07571, att_loss=0.2369, loss=0.2046, over 16955.00 frames. utt_duration=1358 frames, utt_pad_proportion=0.007325, over 50.00 utterances.], tot_loss[ctc_loss=0.08212, att_loss=0.2396, loss=0.2081, over 3267764.85 frames. utt_duration=1257 frames, utt_pad_proportion=0.05557, over 10412.14 utterances.], batch size: 50, lr: 6.30e-03, grad_scale: 8.0 2023-03-08 16:14:55,399 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-03-08 16:15:37,317 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.585e+02 2.015e+02 2.419e+02 3.041e+02 5.956e+02, threshold=4.838e+02, percent-clipped=4.0 2023-03-08 16:15:49,841 INFO [train2.py:809] (1/4) Epoch 17, batch 3400, loss[ctc_loss=0.09205, att_loss=0.246, loss=0.2152, over 16543.00 frames. utt_duration=1472 frames, utt_pad_proportion=0.005957, over 45.00 utterances.], tot_loss[ctc_loss=0.08283, att_loss=0.2402, loss=0.2087, over 3263872.42 frames. utt_duration=1255 frames, utt_pad_proportion=0.05716, over 10414.05 utterances.], batch size: 45, lr: 6.30e-03, grad_scale: 8.0 2023-03-08 16:15:54,799 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67144.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 16:16:02,687 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0182, 5.1896, 4.9915, 2.4931, 2.0676, 2.9897, 3.0239, 3.9246], device='cuda:1'), covar=tensor([0.0752, 0.0261, 0.0270, 0.4872, 0.5887, 0.2514, 0.2677, 0.1680], device='cuda:1'), in_proj_covar=tensor([0.0349, 0.0255, 0.0256, 0.0234, 0.0345, 0.0336, 0.0242, 0.0358], device='cuda:1'), out_proj_covar=tensor([1.4975e-04, 9.4023e-05, 1.1003e-04, 1.0097e-04, 1.4518e-04, 1.3244e-04, 9.6424e-05, 1.4698e-04], device='cuda:1') 2023-03-08 16:16:27,808 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-03-08 16:16:45,520 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67177.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 16:16:57,685 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-03-08 16:17:08,454 INFO [train2.py:809] (1/4) Epoch 17, batch 3450, loss[ctc_loss=0.09783, att_loss=0.2471, loss=0.2173, over 16337.00 frames. utt_duration=1454 frames, utt_pad_proportion=0.005761, over 45.00 utterances.], tot_loss[ctc_loss=0.08341, att_loss=0.2404, loss=0.209, over 3260752.75 frames. utt_duration=1227 frames, utt_pad_proportion=0.06464, over 10646.32 utterances.], batch size: 45, lr: 6.30e-03, grad_scale: 8.0 2023-03-08 16:18:01,727 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=67225.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 16:18:15,120 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.253e+02 2.178e+02 2.510e+02 3.172e+02 6.489e+02, threshold=5.020e+02, percent-clipped=3.0 2023-03-08 16:18:27,088 INFO [train2.py:809] (1/4) Epoch 17, batch 3500, loss[ctc_loss=0.09239, att_loss=0.2541, loss=0.2217, over 17289.00 frames. utt_duration=1259 frames, utt_pad_proportion=0.01255, over 55.00 utterances.], tot_loss[ctc_loss=0.08325, att_loss=0.2408, loss=0.2093, over 3269117.09 frames. utt_duration=1234 frames, utt_pad_proportion=0.06028, over 10612.92 utterances.], batch size: 55, lr: 6.29e-03, grad_scale: 8.0 2023-03-08 16:19:26,780 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0366, 4.4263, 4.4450, 4.6174, 2.8880, 4.5513, 2.8780, 2.0038], device='cuda:1'), covar=tensor([0.0484, 0.0259, 0.0732, 0.0164, 0.1746, 0.0162, 0.1529, 0.1696], device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0146, 0.0255, 0.0140, 0.0218, 0.0126, 0.0227, 0.0200], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 16:19:28,713 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.03 vs. limit=5.0 2023-03-08 16:19:33,823 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-03-08 16:19:46,695 INFO [train2.py:809] (1/4) Epoch 17, batch 3550, loss[ctc_loss=0.07311, att_loss=0.2542, loss=0.218, over 17032.00 frames. utt_duration=1337 frames, utt_pad_proportion=0.007133, over 51.00 utterances.], tot_loss[ctc_loss=0.08288, att_loss=0.2405, loss=0.209, over 3275174.04 frames. utt_duration=1248 frames, utt_pad_proportion=0.05415, over 10513.04 utterances.], batch size: 51, lr: 6.29e-03, grad_scale: 8.0 2023-03-08 16:20:21,501 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67313.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 16:20:54,016 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.300e+02 1.995e+02 2.311e+02 2.839e+02 6.877e+02, threshold=4.621e+02, percent-clipped=1.0 2023-03-08 16:21:06,790 INFO [train2.py:809] (1/4) Epoch 17, batch 3600, loss[ctc_loss=0.06659, att_loss=0.2216, loss=0.1906, over 16420.00 frames. utt_duration=1494 frames, utt_pad_proportion=0.005701, over 44.00 utterances.], tot_loss[ctc_loss=0.08221, att_loss=0.24, loss=0.2084, over 3276930.29 frames. utt_duration=1276 frames, utt_pad_proportion=0.04745, over 10287.24 utterances.], batch size: 44, lr: 6.29e-03, grad_scale: 8.0 2023-03-08 16:21:10,822 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-03-08 16:21:48,406 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-03-08 16:21:58,462 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0865, 5.4013, 4.9194, 5.4813, 4.8550, 5.0219, 5.5672, 5.3077], device='cuda:1'), covar=tensor([0.0564, 0.0294, 0.0842, 0.0311, 0.0417, 0.0225, 0.0265, 0.0196], device='cuda:1'), in_proj_covar=tensor([0.0375, 0.0299, 0.0351, 0.0317, 0.0305, 0.0227, 0.0287, 0.0269], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 16:22:19,255 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67387.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 16:22:26,318 INFO [train2.py:809] (1/4) Epoch 17, batch 3650, loss[ctc_loss=0.07991, att_loss=0.2298, loss=0.1999, over 16115.00 frames. utt_duration=1536 frames, utt_pad_proportion=0.005856, over 42.00 utterances.], tot_loss[ctc_loss=0.0815, att_loss=0.239, loss=0.2075, over 3268865.46 frames. utt_duration=1298 frames, utt_pad_proportion=0.04311, over 10086.56 utterances.], batch size: 42, lr: 6.29e-03, grad_scale: 8.0 2023-03-08 16:23:34,284 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.388e+02 2.053e+02 2.468e+02 2.850e+02 9.204e+02, threshold=4.935e+02, percent-clipped=2.0 2023-03-08 16:23:36,635 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=67435.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 16:23:44,119 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67439.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 16:23:47,245 INFO [train2.py:809] (1/4) Epoch 17, batch 3700, loss[ctc_loss=0.06926, att_loss=0.2301, loss=0.1979, over 16279.00 frames. utt_duration=1516 frames, utt_pad_proportion=0.007341, over 43.00 utterances.], tot_loss[ctc_loss=0.08215, att_loss=0.2397, loss=0.2082, over 3274480.83 frames. utt_duration=1271 frames, utt_pad_proportion=0.04762, over 10317.54 utterances.], batch size: 43, lr: 6.28e-03, grad_scale: 8.0 2023-03-08 16:24:00,478 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3651, 2.5378, 4.9498, 3.6364, 2.7803, 4.1537, 4.4876, 4.5096], device='cuda:1'), covar=tensor([0.0236, 0.1647, 0.0122, 0.1208, 0.1953, 0.0270, 0.0162, 0.0237], device='cuda:1'), in_proj_covar=tensor([0.0176, 0.0240, 0.0168, 0.0308, 0.0264, 0.0199, 0.0152, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 16:24:35,019 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2802, 4.5372, 4.3489, 4.7478, 2.9231, 4.5921, 2.9747, 2.2121], device='cuda:1'), covar=tensor([0.0393, 0.0241, 0.0731, 0.0160, 0.1627, 0.0165, 0.1397, 0.1534], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0147, 0.0256, 0.0141, 0.0220, 0.0126, 0.0227, 0.0201], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 16:25:07,047 INFO [train2.py:809] (1/4) Epoch 17, batch 3750, loss[ctc_loss=0.05273, att_loss=0.223, loss=0.189, over 16409.00 frames. utt_duration=1493 frames, utt_pad_proportion=0.006396, over 44.00 utterances.], tot_loss[ctc_loss=0.08253, att_loss=0.2401, loss=0.2086, over 3274004.73 frames. utt_duration=1273 frames, utt_pad_proportion=0.04855, over 10300.19 utterances.], batch size: 44, lr: 6.28e-03, grad_scale: 8.0 2023-03-08 16:25:10,615 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7571, 3.3806, 3.8236, 3.2412, 3.7680, 4.7713, 4.6377, 3.6417], device='cuda:1'), covar=tensor([0.0385, 0.1432, 0.1128, 0.1249, 0.0905, 0.1133, 0.0542, 0.1053], device='cuda:1'), in_proj_covar=tensor([0.0243, 0.0241, 0.0271, 0.0214, 0.0260, 0.0355, 0.0250, 0.0232], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 16:25:42,179 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67513.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 16:26:14,092 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.323e+02 2.055e+02 2.519e+02 3.069e+02 5.925e+02, threshold=5.038e+02, percent-clipped=2.0 2023-03-08 16:26:25,909 INFO [train2.py:809] (1/4) Epoch 17, batch 3800, loss[ctc_loss=0.06902, att_loss=0.2043, loss=0.1772, over 15879.00 frames. utt_duration=1630 frames, utt_pad_proportion=0.009658, over 39.00 utterances.], tot_loss[ctc_loss=0.08172, att_loss=0.2393, loss=0.2078, over 3271052.33 frames. utt_duration=1308 frames, utt_pad_proportion=0.04133, over 10014.83 utterances.], batch size: 39, lr: 6.28e-03, grad_scale: 8.0 2023-03-08 16:27:18,173 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67574.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 16:27:30,739 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5408, 3.2173, 3.7127, 2.9617, 3.6704, 4.6647, 4.4621, 3.3152], device='cuda:1'), covar=tensor([0.0403, 0.1428, 0.1119, 0.1343, 0.0887, 0.0727, 0.0497, 0.1282], device='cuda:1'), in_proj_covar=tensor([0.0243, 0.0240, 0.0270, 0.0212, 0.0257, 0.0352, 0.0250, 0.0232], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 16:27:30,873 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4655, 2.6806, 4.9391, 3.9118, 2.9342, 4.2764, 4.7437, 4.7044], device='cuda:1'), covar=tensor([0.0253, 0.1686, 0.0206, 0.0872, 0.1818, 0.0228, 0.0126, 0.0201], device='cuda:1'), in_proj_covar=tensor([0.0176, 0.0239, 0.0168, 0.0307, 0.0264, 0.0200, 0.0151, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 16:27:46,052 INFO [train2.py:809] (1/4) Epoch 17, batch 3850, loss[ctc_loss=0.06508, att_loss=0.2105, loss=0.1814, over 15372.00 frames. utt_duration=1758 frames, utt_pad_proportion=0.01056, over 35.00 utterances.], tot_loss[ctc_loss=0.08166, att_loss=0.2389, loss=0.2075, over 3269194.27 frames. utt_duration=1305 frames, utt_pad_proportion=0.04301, over 10028.55 utterances.], batch size: 35, lr: 6.28e-03, grad_scale: 8.0 2023-03-08 16:27:57,048 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5070, 2.5995, 5.0638, 3.9470, 3.0662, 4.3224, 4.7952, 4.7067], device='cuda:1'), covar=tensor([0.0254, 0.1565, 0.0153, 0.0856, 0.1709, 0.0222, 0.0119, 0.0210], device='cuda:1'), in_proj_covar=tensor([0.0176, 0.0240, 0.0168, 0.0307, 0.0264, 0.0200, 0.0151, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 16:28:04,473 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-03-08 16:28:20,439 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67613.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 16:28:53,215 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.336e+02 2.153e+02 2.695e+02 3.326e+02 5.215e+02, threshold=5.391e+02, percent-clipped=2.0 2023-03-08 16:29:04,179 INFO [train2.py:809] (1/4) Epoch 17, batch 3900, loss[ctc_loss=0.07399, att_loss=0.2202, loss=0.1909, over 15991.00 frames. utt_duration=1601 frames, utt_pad_proportion=0.008901, over 40.00 utterances.], tot_loss[ctc_loss=0.08138, att_loss=0.239, loss=0.2074, over 3275581.21 frames. utt_duration=1305 frames, utt_pad_proportion=0.04208, over 10053.52 utterances.], batch size: 40, lr: 6.28e-03, grad_scale: 8.0 2023-03-08 16:29:34,757 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=67661.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 16:29:42,714 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7332, 4.7206, 4.4878, 2.7113, 4.5625, 4.3483, 3.9861, 2.5705], device='cuda:1'), covar=tensor([0.0111, 0.0100, 0.0241, 0.1082, 0.0097, 0.0233, 0.0334, 0.1385], device='cuda:1'), in_proj_covar=tensor([0.0072, 0.0098, 0.0097, 0.0111, 0.0081, 0.0108, 0.0098, 0.0104], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 16:30:04,343 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67680.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 16:30:21,236 INFO [train2.py:809] (1/4) Epoch 17, batch 3950, loss[ctc_loss=0.1103, att_loss=0.2661, loss=0.235, over 17327.00 frames. utt_duration=1176 frames, utt_pad_proportion=0.02313, over 59.00 utterances.], tot_loss[ctc_loss=0.0816, att_loss=0.2392, loss=0.2076, over 3267918.61 frames. utt_duration=1290 frames, utt_pad_proportion=0.04588, over 10141.87 utterances.], batch size: 59, lr: 6.27e-03, grad_scale: 8.0 2023-03-08 16:30:23,193 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67692.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 16:31:39,751 INFO [train2.py:809] (1/4) Epoch 18, batch 0, loss[ctc_loss=0.09598, att_loss=0.2617, loss=0.2286, over 16766.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.006703, over 48.00 utterances.], tot_loss[ctc_loss=0.09598, att_loss=0.2617, loss=0.2286, over 16766.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.006703, over 48.00 utterances.], batch size: 48, lr: 6.09e-03, grad_scale: 8.0 2023-03-08 16:31:39,752 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-08 16:31:52,767 INFO [train2.py:843] (1/4) Epoch 18, validation: ctc_loss=0.04224, att_loss=0.2352, loss=0.1966, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-08 16:31:52,768 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-08 16:32:02,654 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-03-08 16:32:06,402 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.395e+02 2.121e+02 2.560e+02 3.250e+02 7.102e+02, threshold=5.120e+02, percent-clipped=3.0 2023-03-08 16:32:09,822 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.5959, 3.2773, 3.2349, 2.9296, 3.3254, 3.3320, 3.3009, 2.3376], device='cuda:1'), covar=tensor([0.1101, 0.1912, 0.3052, 0.4703, 0.1578, 0.3142, 0.1293, 0.5575], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0166, 0.0173, 0.0237, 0.0142, 0.0232, 0.0152, 0.0203], 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') 2023-03-08 16:32:14,417 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67739.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 16:32:17,610 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67741.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 16:32:20,455 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6044, 3.2434, 3.7527, 3.0644, 3.6031, 4.7263, 4.5752, 3.4161], device='cuda:1'), covar=tensor([0.0333, 0.1496, 0.1131, 0.1303, 0.1034, 0.0776, 0.0511, 0.1153], device='cuda:1'), in_proj_covar=tensor([0.0243, 0.0241, 0.0271, 0.0214, 0.0259, 0.0352, 0.0249, 0.0233], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 16:32:35,593 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67753.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 16:32:43,133 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9303, 6.2045, 5.7477, 5.9226, 5.8869, 5.3770, 5.6302, 5.4500], device='cuda:1'), covar=tensor([0.1289, 0.0741, 0.0890, 0.0750, 0.0824, 0.1503, 0.2089, 0.2098], device='cuda:1'), in_proj_covar=tensor([0.0498, 0.0570, 0.0432, 0.0434, 0.0412, 0.0446, 0.0584, 0.0504], 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') 2023-03-08 16:33:00,717 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-03-08 16:33:11,760 INFO [train2.py:809] (1/4) Epoch 18, batch 50, loss[ctc_loss=0.08656, att_loss=0.2279, loss=0.1996, over 16397.00 frames. utt_duration=1492 frames, utt_pad_proportion=0.007092, over 44.00 utterances.], tot_loss[ctc_loss=0.08597, att_loss=0.2412, loss=0.2102, over 735684.66 frames. utt_duration=1176 frames, utt_pad_proportion=0.07618, over 2506.41 utterances.], batch size: 44, lr: 6.09e-03, grad_scale: 8.0 2023-03-08 16:33:30,485 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=67787.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 16:34:31,362 INFO [train2.py:809] (1/4) Epoch 18, batch 100, loss[ctc_loss=0.09191, att_loss=0.2501, loss=0.2184, over 17193.00 frames. utt_duration=872 frames, utt_pad_proportion=0.08789, over 79.00 utterances.], tot_loss[ctc_loss=0.08385, att_loss=0.2401, loss=0.2089, over 1300417.99 frames. utt_duration=1224 frames, utt_pad_proportion=0.05916, over 4253.68 utterances.], batch size: 79, lr: 6.09e-03, grad_scale: 8.0 2023-03-08 16:34:45,110 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.584e+02 2.310e+02 2.796e+02 3.498e+02 6.865e+02, threshold=5.593e+02, percent-clipped=9.0 2023-03-08 16:35:39,910 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67869.0, num_to_drop=1, layers_to_drop={3} 2023-03-08 16:35:49,985 INFO [train2.py:809] (1/4) Epoch 18, batch 150, loss[ctc_loss=0.0876, att_loss=0.2506, loss=0.218, over 16967.00 frames. utt_duration=1359 frames, utt_pad_proportion=0.006696, over 50.00 utterances.], tot_loss[ctc_loss=0.08283, att_loss=0.2403, loss=0.2088, over 1745698.89 frames. utt_duration=1215 frames, utt_pad_proportion=0.05654, over 5756.39 utterances.], batch size: 50, lr: 6.09e-03, grad_scale: 8.0 2023-03-08 16:36:40,977 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.9625, 3.6203, 3.7000, 3.1826, 3.6958, 3.7654, 3.7428, 2.9359], device='cuda:1'), covar=tensor([0.0887, 0.1781, 0.2050, 0.4112, 0.2050, 0.3266, 0.0893, 0.3967], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0165, 0.0175, 0.0237, 0.0142, 0.0234, 0.0152, 0.0203], 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') 2023-03-08 16:37:08,984 INFO [train2.py:809] (1/4) Epoch 18, batch 200, loss[ctc_loss=0.06352, att_loss=0.2103, loss=0.1809, over 15647.00 frames. utt_duration=1694 frames, utt_pad_proportion=0.007901, over 37.00 utterances.], tot_loss[ctc_loss=0.08149, att_loss=0.2385, loss=0.2071, over 2074030.64 frames. utt_duration=1224 frames, utt_pad_proportion=0.06147, over 6785.21 utterances.], batch size: 37, lr: 6.08e-03, grad_scale: 8.0 2023-03-08 16:37:12,415 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4890, 2.3918, 4.9273, 3.7945, 2.9617, 4.1832, 4.7255, 4.6691], device='cuda:1'), covar=tensor([0.0251, 0.1837, 0.0162, 0.0979, 0.1734, 0.0245, 0.0111, 0.0206], device='cuda:1'), in_proj_covar=tensor([0.0176, 0.0240, 0.0169, 0.0307, 0.0263, 0.0199, 0.0151, 0.0181], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 16:37:22,690 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.418e+02 2.000e+02 2.339e+02 2.884e+02 4.176e+02, threshold=4.678e+02, percent-clipped=0.0 2023-03-08 16:37:45,898 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.4111, 1.8836, 2.0906, 2.1014, 2.4171, 2.1959, 1.7692, 2.5471], device='cuda:1'), covar=tensor([0.1269, 0.2662, 0.2238, 0.1332, 0.2029, 0.1256, 0.1502, 0.0863], device='cuda:1'), in_proj_covar=tensor([0.0100, 0.0109, 0.0113, 0.0096, 0.0104, 0.0091, 0.0112, 0.0082], device='cuda:1'), out_proj_covar=tensor([7.4216e-05, 8.2672e-05, 8.6243e-05, 7.3412e-05, 7.6518e-05, 7.2417e-05, 8.3179e-05, 6.5990e-05], device='cuda:1') 2023-03-08 16:38:27,432 INFO [train2.py:809] (1/4) Epoch 18, batch 250, loss[ctc_loss=0.06503, att_loss=0.2062, loss=0.178, over 15494.00 frames. utt_duration=1723 frames, utt_pad_proportion=0.009233, over 36.00 utterances.], tot_loss[ctc_loss=0.08202, att_loss=0.2392, loss=0.2078, over 2333241.58 frames. utt_duration=1218 frames, utt_pad_proportion=0.06382, over 7669.67 utterances.], batch size: 36, lr: 6.08e-03, grad_scale: 8.0 2023-03-08 16:38:36,834 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67981.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 16:39:31,869 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2164, 4.5841, 4.4921, 4.7582, 2.6927, 4.5552, 2.7149, 1.6947], device='cuda:1'), covar=tensor([0.0447, 0.0226, 0.0647, 0.0174, 0.1726, 0.0188, 0.1579, 0.1788], device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0146, 0.0253, 0.0140, 0.0218, 0.0126, 0.0227, 0.0201], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 16:39:50,972 INFO [train2.py:809] (1/4) Epoch 18, batch 300, loss[ctc_loss=0.08721, att_loss=0.2479, loss=0.2157, over 17430.00 frames. utt_duration=1012 frames, utt_pad_proportion=0.04535, over 69.00 utterances.], tot_loss[ctc_loss=0.08274, att_loss=0.2398, loss=0.2084, over 2535208.11 frames. utt_duration=1171 frames, utt_pad_proportion=0.07592, over 8668.51 utterances.], batch size: 69, lr: 6.08e-03, grad_scale: 8.0 2023-03-08 16:40:04,759 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.391e+02 2.007e+02 2.345e+02 2.847e+02 6.033e+02, threshold=4.689e+02, percent-clipped=3.0 2023-03-08 16:40:08,020 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68036.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 16:40:17,194 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68042.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 16:40:20,454 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.5660, 2.3829, 2.4675, 2.2772, 3.0890, 2.7089, 2.2361, 2.9318], device='cuda:1'), covar=tensor([0.2549, 0.3431, 0.3330, 0.1978, 0.1916, 0.1106, 0.2770, 0.0929], device='cuda:1'), in_proj_covar=tensor([0.0102, 0.0110, 0.0113, 0.0097, 0.0106, 0.0092, 0.0113, 0.0083], device='cuda:1'), out_proj_covar=tensor([7.5274e-05, 8.3732e-05, 8.6878e-05, 7.4247e-05, 7.7508e-05, 7.3401e-05, 8.4206e-05, 6.6685e-05], device='cuda:1') 2023-03-08 16:40:26,343 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68048.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 16:41:09,398 INFO [train2.py:809] (1/4) Epoch 18, batch 350, loss[ctc_loss=0.075, att_loss=0.2413, loss=0.2081, over 16395.00 frames. utt_duration=1492 frames, utt_pad_proportion=0.007979, over 44.00 utterances.], tot_loss[ctc_loss=0.08207, att_loss=0.2394, loss=0.208, over 2694649.46 frames. utt_duration=1173 frames, utt_pad_proportion=0.07683, over 9200.70 utterances.], batch size: 44, lr: 6.08e-03, grad_scale: 8.0 2023-03-08 16:41:27,814 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68087.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 16:42:27,983 INFO [train2.py:809] (1/4) Epoch 18, batch 400, loss[ctc_loss=0.08074, att_loss=0.2441, loss=0.2115, over 17245.00 frames. utt_duration=1097 frames, utt_pad_proportion=0.03977, over 63.00 utterances.], tot_loss[ctc_loss=0.082, att_loss=0.2395, loss=0.208, over 2831305.54 frames. utt_duration=1204 frames, utt_pad_proportion=0.06524, over 9415.39 utterances.], batch size: 63, lr: 6.07e-03, grad_scale: 8.0 2023-03-08 16:42:41,450 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.402e+02 2.125e+02 2.519e+02 3.295e+02 8.158e+02, threshold=5.039e+02, percent-clipped=7.0 2023-03-08 16:43:03,783 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68148.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 16:43:37,533 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68169.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 16:43:46,630 INFO [train2.py:809] (1/4) Epoch 18, batch 450, loss[ctc_loss=0.0674, att_loss=0.2309, loss=0.1982, over 16401.00 frames. utt_duration=1492 frames, utt_pad_proportion=0.007661, over 44.00 utterances.], tot_loss[ctc_loss=0.08181, att_loss=0.2397, loss=0.2081, over 2931563.96 frames. utt_duration=1187 frames, utt_pad_proportion=0.06854, over 9893.61 utterances.], batch size: 44, lr: 6.07e-03, grad_scale: 8.0 2023-03-08 16:43:47,623 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-03-08 16:44:28,707 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1276, 4.5386, 4.5798, 4.8015, 2.8720, 4.5865, 2.9472, 1.8901], device='cuda:1'), covar=tensor([0.0399, 0.0260, 0.0552, 0.0156, 0.1440, 0.0173, 0.1233, 0.1597], device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0146, 0.0254, 0.0140, 0.0218, 0.0126, 0.0227, 0.0200], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 16:44:45,659 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.80 vs. limit=2.0 2023-03-08 16:44:53,040 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=68217.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 16:45:05,406 INFO [train2.py:809] (1/4) Epoch 18, batch 500, loss[ctc_loss=0.07803, att_loss=0.2277, loss=0.1978, over 15867.00 frames. utt_duration=1629 frames, utt_pad_proportion=0.01019, over 39.00 utterances.], tot_loss[ctc_loss=0.08187, att_loss=0.2387, loss=0.2074, over 2998481.16 frames. utt_duration=1202 frames, utt_pad_proportion=0.0685, over 9993.94 utterances.], batch size: 39, lr: 6.07e-03, grad_scale: 8.0 2023-03-08 16:45:19,450 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.432e+02 1.924e+02 2.355e+02 3.288e+02 6.167e+02, threshold=4.710e+02, percent-clipped=3.0 2023-03-08 16:46:24,806 INFO [train2.py:809] (1/4) Epoch 18, batch 550, loss[ctc_loss=0.08685, att_loss=0.2488, loss=0.2164, over 17191.00 frames. utt_duration=872 frames, utt_pad_proportion=0.0869, over 79.00 utterances.], tot_loss[ctc_loss=0.08054, att_loss=0.2382, loss=0.2067, over 3059104.69 frames. utt_duration=1244 frames, utt_pad_proportion=0.05787, over 9851.09 utterances.], batch size: 79, lr: 6.07e-03, grad_scale: 8.0 2023-03-08 16:47:23,174 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6245, 2.5492, 4.9298, 4.2024, 2.8601, 4.5038, 4.8603, 4.7342], device='cuda:1'), covar=tensor([0.0196, 0.1594, 0.0250, 0.0737, 0.1865, 0.0174, 0.0143, 0.0201], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0240, 0.0169, 0.0307, 0.0264, 0.0200, 0.0152, 0.0182], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 16:47:44,060 INFO [train2.py:809] (1/4) Epoch 18, batch 600, loss[ctc_loss=0.06362, att_loss=0.2288, loss=0.1957, over 16114.00 frames. utt_duration=1536 frames, utt_pad_proportion=0.006453, over 42.00 utterances.], tot_loss[ctc_loss=0.08, att_loss=0.2377, loss=0.2062, over 3103731.26 frames. utt_duration=1246 frames, utt_pad_proportion=0.05786, over 9975.86 utterances.], batch size: 42, lr: 6.07e-03, grad_scale: 8.0 2023-03-08 16:47:58,121 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.378e+02 2.193e+02 2.599e+02 3.203e+02 7.782e+02, threshold=5.199e+02, percent-clipped=2.0 2023-03-08 16:48:01,685 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68336.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 16:48:03,098 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68337.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 16:48:19,835 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68348.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 16:49:03,209 INFO [train2.py:809] (1/4) Epoch 18, batch 650, loss[ctc_loss=0.0841, att_loss=0.2368, loss=0.2063, over 16519.00 frames. utt_duration=1470 frames, utt_pad_proportion=0.006862, over 45.00 utterances.], tot_loss[ctc_loss=0.07994, att_loss=0.2374, loss=0.2059, over 3137820.65 frames. utt_duration=1275 frames, utt_pad_proportion=0.05187, over 9855.66 utterances.], batch size: 45, lr: 6.06e-03, grad_scale: 8.0 2023-03-08 16:49:17,258 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=68384.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 16:49:35,684 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=68396.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 16:49:40,474 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68399.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 16:50:22,685 INFO [train2.py:809] (1/4) Epoch 18, batch 700, loss[ctc_loss=0.0994, att_loss=0.2625, loss=0.2299, over 17130.00 frames. utt_duration=1225 frames, utt_pad_proportion=0.01431, over 56.00 utterances.], tot_loss[ctc_loss=0.07993, att_loss=0.238, loss=0.2064, over 3175628.06 frames. utt_duration=1284 frames, utt_pad_proportion=0.04561, over 9905.68 utterances.], batch size: 56, lr: 6.06e-03, grad_scale: 8.0 2023-03-08 16:50:36,282 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.516e+02 2.020e+02 2.578e+02 3.082e+02 5.307e+02, threshold=5.155e+02, percent-clipped=1.0 2023-03-08 16:50:50,330 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68443.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 16:51:17,775 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68460.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 16:51:41,629 INFO [train2.py:809] (1/4) Epoch 18, batch 750, loss[ctc_loss=0.07511, att_loss=0.2464, loss=0.2122, over 17339.00 frames. utt_duration=702 frames, utt_pad_proportion=0.1159, over 99.00 utterances.], tot_loss[ctc_loss=0.07991, att_loss=0.2384, loss=0.2067, over 3199588.21 frames. utt_duration=1258 frames, utt_pad_proportion=0.05065, over 10181.89 utterances.], batch size: 99, lr: 6.06e-03, grad_scale: 8.0 2023-03-08 16:53:00,272 INFO [train2.py:809] (1/4) Epoch 18, batch 800, loss[ctc_loss=0.06927, att_loss=0.2249, loss=0.1938, over 16388.00 frames. utt_duration=1491 frames, utt_pad_proportion=0.007848, over 44.00 utterances.], tot_loss[ctc_loss=0.07986, att_loss=0.2389, loss=0.2071, over 3222638.33 frames. utt_duration=1272 frames, utt_pad_proportion=0.04655, over 10142.07 utterances.], batch size: 44, lr: 6.06e-03, grad_scale: 8.0 2023-03-08 16:53:13,829 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.459e+02 2.218e+02 2.657e+02 2.960e+02 6.407e+02, threshold=5.314e+02, percent-clipped=4.0 2023-03-08 16:53:32,831 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.7724, 5.0165, 5.3442, 5.1822, 5.2284, 5.7193, 5.1588, 5.8316], device='cuda:1'), covar=tensor([0.0789, 0.0829, 0.0842, 0.1373, 0.2088, 0.0994, 0.0691, 0.0739], device='cuda:1'), in_proj_covar=tensor([0.0842, 0.0489, 0.0578, 0.0647, 0.0850, 0.0594, 0.0474, 0.0583], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 16:53:48,927 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0131, 5.0255, 4.7721, 2.8664, 4.8760, 4.6780, 4.1592, 2.5574], device='cuda:1'), covar=tensor([0.0117, 0.0098, 0.0295, 0.1020, 0.0086, 0.0202, 0.0342, 0.1378], device='cuda:1'), in_proj_covar=tensor([0.0071, 0.0097, 0.0095, 0.0109, 0.0081, 0.0107, 0.0097, 0.0103], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 16:54:19,206 INFO [train2.py:809] (1/4) Epoch 18, batch 850, loss[ctc_loss=0.07301, att_loss=0.2264, loss=0.1957, over 16172.00 frames. utt_duration=1579 frames, utt_pad_proportion=0.006734, over 41.00 utterances.], tot_loss[ctc_loss=0.07969, att_loss=0.2396, loss=0.2076, over 3240844.31 frames. utt_duration=1262 frames, utt_pad_proportion=0.04882, over 10287.32 utterances.], batch size: 41, lr: 6.05e-03, grad_scale: 8.0 2023-03-08 16:54:24,133 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68578.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 16:55:38,614 INFO [train2.py:809] (1/4) Epoch 18, batch 900, loss[ctc_loss=0.08211, att_loss=0.2325, loss=0.2025, over 16130.00 frames. utt_duration=1538 frames, utt_pad_proportion=0.005972, over 42.00 utterances.], tot_loss[ctc_loss=0.07935, att_loss=0.239, loss=0.2071, over 3247204.55 frames. utt_duration=1277 frames, utt_pad_proportion=0.04598, over 10183.64 utterances.], batch size: 42, lr: 6.05e-03, grad_scale: 8.0 2023-03-08 16:55:52,341 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.470e+02 1.982e+02 2.432e+02 3.091e+02 4.816e+02, threshold=4.864e+02, percent-clipped=0.0 2023-03-08 16:55:57,210 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68637.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 16:56:00,281 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68639.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 16:56:57,074 INFO [train2.py:809] (1/4) Epoch 18, batch 950, loss[ctc_loss=0.07004, att_loss=0.2418, loss=0.2075, over 16966.00 frames. utt_duration=1359 frames, utt_pad_proportion=0.007626, over 50.00 utterances.], tot_loss[ctc_loss=0.07987, att_loss=0.2387, loss=0.2069, over 3249984.44 frames. utt_duration=1268 frames, utt_pad_proportion=0.0509, over 10261.69 utterances.], batch size: 50, lr: 6.05e-03, grad_scale: 8.0 2023-03-08 16:57:12,200 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=68685.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 16:57:12,455 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68685.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 16:57:41,184 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.4945, 5.7496, 5.2291, 5.5706, 5.4092, 4.9732, 5.0947, 5.0009], device='cuda:1'), covar=tensor([0.1451, 0.0976, 0.0969, 0.0832, 0.0985, 0.1556, 0.2548, 0.2309], device='cuda:1'), in_proj_covar=tensor([0.0505, 0.0579, 0.0440, 0.0440, 0.0415, 0.0447, 0.0592, 0.0511], 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') 2023-03-08 16:58:13,107 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0776, 2.7158, 3.4497, 2.6350, 3.3891, 4.2360, 4.0345, 2.9012], device='cuda:1'), covar=tensor([0.0429, 0.1699, 0.1192, 0.1390, 0.0992, 0.0846, 0.0669, 0.1379], device='cuda:1'), in_proj_covar=tensor([0.0241, 0.0241, 0.0270, 0.0212, 0.0259, 0.0350, 0.0250, 0.0232], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 16:58:15,810 INFO [train2.py:809] (1/4) Epoch 18, batch 1000, loss[ctc_loss=0.089, att_loss=0.2523, loss=0.2196, over 16865.00 frames. utt_duration=1378 frames, utt_pad_proportion=0.007714, over 49.00 utterances.], tot_loss[ctc_loss=0.08062, att_loss=0.2385, loss=0.207, over 3248244.02 frames. utt_duration=1253 frames, utt_pad_proportion=0.05585, over 10384.64 utterances.], batch size: 49, lr: 6.05e-03, grad_scale: 8.0 2023-03-08 16:58:29,417 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.279e+02 2.078e+02 2.479e+02 2.964e+02 7.208e+02, threshold=4.957e+02, percent-clipped=4.0 2023-03-08 16:58:43,593 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68743.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 16:58:48,374 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68746.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 16:59:00,917 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.0608, 2.1476, 2.2549, 2.3002, 2.8804, 2.2286, 2.0779, 2.8263], device='cuda:1'), covar=tensor([0.1725, 0.2997, 0.2474, 0.1343, 0.1581, 0.1294, 0.2536, 0.0900], device='cuda:1'), in_proj_covar=tensor([0.0104, 0.0112, 0.0114, 0.0098, 0.0108, 0.0095, 0.0116, 0.0084], device='cuda:1'), out_proj_covar=tensor([7.6678e-05, 8.4923e-05, 8.7712e-05, 7.4807e-05, 7.8675e-05, 7.5239e-05, 8.5684e-05, 6.7655e-05], device='cuda:1') 2023-03-08 16:59:02,227 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68755.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 16:59:33,518 INFO [train2.py:809] (1/4) Epoch 18, batch 1050, loss[ctc_loss=0.07538, att_loss=0.2252, loss=0.1953, over 16165.00 frames. utt_duration=1578 frames, utt_pad_proportion=0.006646, over 41.00 utterances.], tot_loss[ctc_loss=0.08062, att_loss=0.2384, loss=0.2069, over 3251746.95 frames. utt_duration=1266 frames, utt_pad_proportion=0.05375, over 10283.36 utterances.], batch size: 41, lr: 6.05e-03, grad_scale: 16.0 2023-03-08 16:59:46,394 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6986, 4.7976, 4.6047, 4.5697, 5.3364, 4.7958, 4.7344, 2.5597], device='cuda:1'), covar=tensor([0.0181, 0.0244, 0.0339, 0.0377, 0.0735, 0.0141, 0.0267, 0.1812], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0162, 0.0168, 0.0185, 0.0359, 0.0141, 0.0156, 0.0214], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 16:59:58,334 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=68791.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 17:00:42,781 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9332, 5.3233, 4.6080, 5.4253, 4.7852, 4.9475, 5.3874, 5.1134], device='cuda:1'), covar=tensor([0.0598, 0.0216, 0.1016, 0.0231, 0.0410, 0.0351, 0.0250, 0.0209], device='cuda:1'), in_proj_covar=tensor([0.0377, 0.0301, 0.0350, 0.0322, 0.0305, 0.0232, 0.0286, 0.0271], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 17:00:53,131 INFO [train2.py:809] (1/4) Epoch 18, batch 1100, loss[ctc_loss=0.08439, att_loss=0.241, loss=0.2097, over 17036.00 frames. utt_duration=689.6 frames, utt_pad_proportion=0.1347, over 99.00 utterances.], tot_loss[ctc_loss=0.08053, att_loss=0.2386, loss=0.207, over 3255483.41 frames. utt_duration=1245 frames, utt_pad_proportion=0.05833, over 10473.34 utterances.], batch size: 99, lr: 6.04e-03, grad_scale: 16.0 2023-03-08 17:01:06,990 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.271e+02 2.026e+02 2.542e+02 3.113e+02 8.116e+02, threshold=5.085e+02, percent-clipped=3.0 2023-03-08 17:01:46,049 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7259, 3.7601, 3.0346, 3.3234, 3.8940, 3.5348, 2.8017, 4.1577], device='cuda:1'), covar=tensor([0.1182, 0.0435, 0.1158, 0.0722, 0.0721, 0.0699, 0.0941, 0.0484], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0206, 0.0218, 0.0192, 0.0259, 0.0230, 0.0196, 0.0274], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-08 17:02:00,193 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2431, 3.7456, 3.8076, 3.2142, 3.9840, 3.9306, 3.8542, 2.9578], device='cuda:1'), covar=tensor([0.0663, 0.1899, 0.1632, 0.3138, 0.0524, 0.1959, 0.0625, 0.3644], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0169, 0.0178, 0.0242, 0.0144, 0.0240, 0.0157, 0.0206], 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') 2023-03-08 17:02:12,251 INFO [train2.py:809] (1/4) Epoch 18, batch 1150, loss[ctc_loss=0.07113, att_loss=0.2213, loss=0.1913, over 16003.00 frames. utt_duration=1602 frames, utt_pad_proportion=0.007683, over 40.00 utterances.], tot_loss[ctc_loss=0.07975, att_loss=0.2381, loss=0.2064, over 3249686.17 frames. utt_duration=1277 frames, utt_pad_proportion=0.05106, over 10190.22 utterances.], batch size: 40, lr: 6.04e-03, grad_scale: 16.0 2023-03-08 17:03:14,680 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4401, 4.8967, 4.6419, 4.7675, 4.9432, 4.5029, 3.4740, 4.7680], device='cuda:1'), covar=tensor([0.0130, 0.0122, 0.0150, 0.0105, 0.0112, 0.0129, 0.0702, 0.0246], device='cuda:1'), in_proj_covar=tensor([0.0087, 0.0084, 0.0105, 0.0066, 0.0070, 0.0082, 0.0100, 0.0104], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 17:03:16,703 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-03-08 17:03:31,041 INFO [train2.py:809] (1/4) Epoch 18, batch 1200, loss[ctc_loss=0.05597, att_loss=0.2058, loss=0.1759, over 15656.00 frames. utt_duration=1694 frames, utt_pad_proportion=0.008118, over 37.00 utterances.], tot_loss[ctc_loss=0.07911, att_loss=0.238, loss=0.2062, over 3260467.44 frames. utt_duration=1295 frames, utt_pad_proportion=0.04482, over 10081.38 utterances.], batch size: 37, lr: 6.04e-03, grad_scale: 16.0 2023-03-08 17:03:44,927 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.483e+02 2.054e+02 2.571e+02 3.261e+02 6.642e+02, threshold=5.143e+02, percent-clipped=4.0 2023-03-08 17:03:45,175 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68934.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:04:50,129 INFO [train2.py:809] (1/4) Epoch 18, batch 1250, loss[ctc_loss=0.1082, att_loss=0.2571, loss=0.2273, over 16893.00 frames. utt_duration=1381 frames, utt_pad_proportion=0.006754, over 49.00 utterances.], tot_loss[ctc_loss=0.07912, att_loss=0.2388, loss=0.2069, over 3274223.30 frames. utt_duration=1284 frames, utt_pad_proportion=0.04266, over 10210.93 utterances.], batch size: 49, lr: 6.04e-03, grad_scale: 16.0 2023-03-08 17:05:52,699 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.55 vs. limit=5.0 2023-03-08 17:06:08,505 INFO [train2.py:809] (1/4) Epoch 18, batch 1300, loss[ctc_loss=0.06605, att_loss=0.2092, loss=0.1806, over 15873.00 frames. utt_duration=1630 frames, utt_pad_proportion=0.009985, over 39.00 utterances.], tot_loss[ctc_loss=0.07982, att_loss=0.2391, loss=0.2072, over 3274995.87 frames. utt_duration=1268 frames, utt_pad_proportion=0.04525, over 10342.41 utterances.], batch size: 39, lr: 6.04e-03, grad_scale: 16.0 2023-03-08 17:06:22,261 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.206e+02 2.084e+02 2.462e+02 2.947e+02 6.955e+02, threshold=4.923e+02, percent-clipped=4.0 2023-03-08 17:06:33,212 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69041.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:06:33,338 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.6019, 2.9865, 3.3035, 4.6019, 4.0918, 3.9852, 2.9611, 2.1109], device='cuda:1'), covar=tensor([0.0606, 0.2087, 0.1049, 0.0506, 0.0699, 0.0461, 0.1513, 0.2448], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0216, 0.0190, 0.0209, 0.0214, 0.0171, 0.0201, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 17:06:36,474 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69043.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:06:54,783 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69055.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:07:26,323 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1050, 4.9166, 5.0646, 2.2599, 2.0860, 2.5210, 2.5904, 3.7334], device='cuda:1'), covar=tensor([0.0878, 0.0625, 0.0259, 0.4586, 0.6926, 0.3562, 0.3496, 0.2185], device='cuda:1'), in_proj_covar=tensor([0.0343, 0.0255, 0.0253, 0.0234, 0.0339, 0.0329, 0.0242, 0.0356], device='cuda:1'), out_proj_covar=tensor([1.4657e-04, 9.4268e-05, 1.0874e-04, 1.0148e-04, 1.4284e-04, 1.2954e-04, 9.6353e-05, 1.4614e-04], device='cuda:1') 2023-03-08 17:07:27,322 INFO [train2.py:809] (1/4) Epoch 18, batch 1350, loss[ctc_loss=0.09169, att_loss=0.2429, loss=0.2127, over 16337.00 frames. utt_duration=1454 frames, utt_pad_proportion=0.005609, over 45.00 utterances.], tot_loss[ctc_loss=0.08117, att_loss=0.2399, loss=0.2082, over 3264345.42 frames. utt_duration=1239 frames, utt_pad_proportion=0.05602, over 10552.28 utterances.], batch size: 45, lr: 6.03e-03, grad_scale: 16.0 2023-03-08 17:08:10,420 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=69103.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:08:12,178 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69104.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:08:29,500 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-03-08 17:08:44,935 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1982, 4.3717, 4.2940, 4.5259, 2.5437, 4.3214, 2.7021, 1.6209], device='cuda:1'), covar=tensor([0.0356, 0.0239, 0.0727, 0.0210, 0.1790, 0.0232, 0.1467, 0.1797], device='cuda:1'), in_proj_covar=tensor([0.0176, 0.0148, 0.0255, 0.0144, 0.0220, 0.0129, 0.0229, 0.0204], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 17:08:46,059 INFO [train2.py:809] (1/4) Epoch 18, batch 1400, loss[ctc_loss=0.0702, att_loss=0.2446, loss=0.2097, over 17284.00 frames. utt_duration=1259 frames, utt_pad_proportion=0.01214, over 55.00 utterances.], tot_loss[ctc_loss=0.0804, att_loss=0.2391, loss=0.2074, over 3265030.96 frames. utt_duration=1249 frames, utt_pad_proportion=0.0548, over 10466.41 utterances.], batch size: 55, lr: 6.03e-03, grad_scale: 16.0 2023-03-08 17:08:46,347 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6713, 5.0580, 4.7868, 4.9060, 5.1563, 4.7353, 3.4089, 4.9476], device='cuda:1'), covar=tensor([0.0108, 0.0099, 0.0114, 0.0088, 0.0068, 0.0104, 0.0724, 0.0182], device='cuda:1'), in_proj_covar=tensor([0.0087, 0.0084, 0.0104, 0.0065, 0.0070, 0.0081, 0.0099, 0.0103], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 17:08:59,737 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.277e+02 2.123e+02 2.542e+02 3.068e+02 7.934e+02, threshold=5.085e+02, percent-clipped=3.0 2023-03-08 17:09:44,882 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69162.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:09:54,772 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69168.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:10:05,180 INFO [train2.py:809] (1/4) Epoch 18, batch 1450, loss[ctc_loss=0.06868, att_loss=0.223, loss=0.1921, over 15373.00 frames. utt_duration=1759 frames, utt_pad_proportion=0.01083, over 35.00 utterances.], tot_loss[ctc_loss=0.08007, att_loss=0.239, loss=0.2072, over 3266868.22 frames. utt_duration=1258 frames, utt_pad_proportion=0.05209, over 10401.05 utterances.], batch size: 35, lr: 6.03e-03, grad_scale: 16.0 2023-03-08 17:11:20,782 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8568, 4.9251, 4.6890, 2.7166, 4.6594, 4.5366, 4.0645, 2.5960], device='cuda:1'), covar=tensor([0.0141, 0.0107, 0.0241, 0.1035, 0.0092, 0.0211, 0.0315, 0.1286], device='cuda:1'), in_proj_covar=tensor([0.0071, 0.0095, 0.0094, 0.0106, 0.0079, 0.0105, 0.0094, 0.0100], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 17:11:20,884 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69223.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:11:23,581 INFO [train2.py:809] (1/4) Epoch 18, batch 1500, loss[ctc_loss=0.05679, att_loss=0.2262, loss=0.1923, over 16771.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.005627, over 48.00 utterances.], tot_loss[ctc_loss=0.08003, att_loss=0.2386, loss=0.2069, over 3265786.75 frames. utt_duration=1268 frames, utt_pad_proportion=0.0515, over 10316.42 utterances.], batch size: 48, lr: 6.03e-03, grad_scale: 16.0 2023-03-08 17:11:23,961 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69225.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:11:26,836 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.3137, 4.7992, 4.9715, 4.7678, 4.8908, 5.2834, 4.8556, 5.3387], device='cuda:1'), covar=tensor([0.0799, 0.0734, 0.0830, 0.1254, 0.1822, 0.0915, 0.1051, 0.0733], device='cuda:1'), in_proj_covar=tensor([0.0835, 0.0484, 0.0570, 0.0636, 0.0836, 0.0584, 0.0472, 0.0576], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 17:11:29,981 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69229.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:11:37,070 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.337e+02 2.061e+02 2.532e+02 3.025e+02 5.458e+02, threshold=5.063e+02, percent-clipped=1.0 2023-03-08 17:11:37,462 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69234.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:11:54,408 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69245.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 17:12:39,655 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69273.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:12:42,422 INFO [train2.py:809] (1/4) Epoch 18, batch 1550, loss[ctc_loss=0.1357, att_loss=0.2682, loss=0.2417, over 13833.00 frames. utt_duration=383 frames, utt_pad_proportion=0.3339, over 145.00 utterances.], tot_loss[ctc_loss=0.08083, att_loss=0.2392, loss=0.2075, over 3264108.09 frames. utt_duration=1233 frames, utt_pad_proportion=0.06115, over 10606.20 utterances.], batch size: 145, lr: 6.02e-03, grad_scale: 16.0 2023-03-08 17:12:52,873 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=69282.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:12:59,163 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69286.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:13:31,397 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69306.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 17:13:53,260 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69320.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:14:00,353 INFO [train2.py:809] (1/4) Epoch 18, batch 1600, loss[ctc_loss=0.07776, att_loss=0.2179, loss=0.1898, over 15492.00 frames. utt_duration=1723 frames, utt_pad_proportion=0.008012, over 36.00 utterances.], tot_loss[ctc_loss=0.08042, att_loss=0.2384, loss=0.2068, over 3261122.06 frames. utt_duration=1272 frames, utt_pad_proportion=0.05242, over 10267.35 utterances.], batch size: 36, lr: 6.02e-03, grad_scale: 16.0 2023-03-08 17:14:14,478 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.214e+02 1.944e+02 2.286e+02 2.991e+02 6.797e+02, threshold=4.572e+02, percent-clipped=3.0 2023-03-08 17:14:14,893 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69334.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 17:14:25,737 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69341.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:15:10,860 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69368.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:15:16,053 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-03-08 17:15:21,712 INFO [train2.py:809] (1/4) Epoch 18, batch 1650, loss[ctc_loss=0.06525, att_loss=0.2242, loss=0.1924, over 16123.00 frames. utt_duration=1537 frames, utt_pad_proportion=0.006556, over 42.00 utterances.], tot_loss[ctc_loss=0.0807, att_loss=0.2389, loss=0.2073, over 3267209.04 frames. utt_duration=1268 frames, utt_pad_proportion=0.05252, over 10320.06 utterances.], batch size: 42, lr: 6.02e-03, grad_scale: 16.0 2023-03-08 17:15:31,640 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69381.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:15:44,491 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=69389.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:15:52,811 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69394.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:16:02,188 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69399.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:16:14,523 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0903, 4.3098, 4.2814, 4.5006, 2.5768, 4.4174, 2.4844, 1.9617], device='cuda:1'), covar=tensor([0.0370, 0.0266, 0.0745, 0.0259, 0.1765, 0.0206, 0.1714, 0.1706], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0146, 0.0251, 0.0141, 0.0215, 0.0127, 0.0225, 0.0200], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 17:16:44,771 INFO [train2.py:809] (1/4) Epoch 18, batch 1700, loss[ctc_loss=0.06223, att_loss=0.2122, loss=0.1822, over 16171.00 frames. utt_duration=1579 frames, utt_pad_proportion=0.00742, over 41.00 utterances.], tot_loss[ctc_loss=0.08049, att_loss=0.239, loss=0.2073, over 3263160.77 frames. utt_duration=1268 frames, utt_pad_proportion=0.05188, over 10310.07 utterances.], batch size: 41, lr: 6.02e-03, grad_scale: 16.0 2023-03-08 17:16:51,523 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69429.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:16:58,949 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.294e+02 2.149e+02 2.524e+02 3.202e+02 9.381e+02, threshold=5.048e+02, percent-clipped=6.0 2023-03-08 17:17:31,247 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5324, 2.5264, 5.0771, 3.9917, 3.0335, 4.2447, 4.8275, 4.6899], device='cuda:1'), covar=tensor([0.0264, 0.1749, 0.0171, 0.0906, 0.1711, 0.0233, 0.0135, 0.0248], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0240, 0.0170, 0.0309, 0.0266, 0.0203, 0.0154, 0.0184], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 17:17:34,312 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69455.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:17:54,398 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.28 vs. limit=5.0 2023-03-08 17:18:03,407 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6495, 3.1581, 3.6865, 3.2590, 3.5999, 4.7851, 4.5588, 3.6431], device='cuda:1'), covar=tensor([0.0320, 0.1517, 0.1138, 0.1242, 0.1003, 0.0570, 0.0563, 0.1027], device='cuda:1'), in_proj_covar=tensor([0.0240, 0.0238, 0.0269, 0.0212, 0.0259, 0.0349, 0.0249, 0.0229], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 17:18:06,236 INFO [train2.py:809] (1/4) Epoch 18, batch 1750, loss[ctc_loss=0.06473, att_loss=0.2209, loss=0.1896, over 15920.00 frames. utt_duration=1634 frames, utt_pad_proportion=0.007135, over 39.00 utterances.], tot_loss[ctc_loss=0.07976, att_loss=0.2384, loss=0.2067, over 3264829.71 frames. utt_duration=1288 frames, utt_pad_proportion=0.04764, over 10154.39 utterances.], batch size: 39, lr: 6.02e-03, grad_scale: 16.0 2023-03-08 17:18:45,106 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.0032, 2.4886, 2.7353, 3.7628, 3.4552, 3.4481, 2.5742, 2.1533], device='cuda:1'), covar=tensor([0.0788, 0.2269, 0.1100, 0.0672, 0.0861, 0.0527, 0.1661, 0.2254], device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0216, 0.0190, 0.0210, 0.0213, 0.0173, 0.0201, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 17:19:17,110 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69518.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:19:26,188 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69524.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:19:27,564 INFO [train2.py:809] (1/4) Epoch 18, batch 1800, loss[ctc_loss=0.06175, att_loss=0.2214, loss=0.1895, over 16390.00 frames. utt_duration=1492 frames, utt_pad_proportion=0.008281, over 44.00 utterances.], tot_loss[ctc_loss=0.08026, att_loss=0.2389, loss=0.2072, over 3264904.34 frames. utt_duration=1267 frames, utt_pad_proportion=0.05315, over 10322.17 utterances.], batch size: 44, lr: 6.01e-03, grad_scale: 16.0 2023-03-08 17:19:37,398 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.1799, 2.8329, 3.1486, 4.2179, 3.7472, 3.6674, 2.8375, 2.1826], device='cuda:1'), covar=tensor([0.0799, 0.1953, 0.0946, 0.0591, 0.0944, 0.0553, 0.1509, 0.2292], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0213, 0.0188, 0.0208, 0.0212, 0.0171, 0.0199, 0.0188], 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') 2023-03-08 17:19:41,873 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.594e+02 2.055e+02 2.270e+02 2.783e+02 4.907e+02, threshold=4.541e+02, percent-clipped=0.0 2023-03-08 17:19:55,287 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-03-08 17:20:36,812 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.0343, 3.7809, 3.6819, 3.1719, 3.7425, 3.8598, 3.7392, 2.8451], device='cuda:1'), covar=tensor([0.0869, 0.1215, 0.2619, 0.3334, 0.1338, 0.2788, 0.0892, 0.3896], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0165, 0.0174, 0.0235, 0.0141, 0.0237, 0.0155, 0.0204], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 17:20:49,338 INFO [train2.py:809] (1/4) Epoch 18, batch 1850, loss[ctc_loss=0.07301, att_loss=0.2453, loss=0.2108, over 16975.00 frames. utt_duration=1359 frames, utt_pad_proportion=0.007027, over 50.00 utterances.], tot_loss[ctc_loss=0.08002, att_loss=0.239, loss=0.2072, over 3268173.18 frames. utt_duration=1261 frames, utt_pad_proportion=0.0544, over 10381.22 utterances.], batch size: 50, lr: 6.01e-03, grad_scale: 16.0 2023-03-08 17:20:58,876 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69581.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:21:28,495 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=2.01 vs. limit=2.0 2023-03-08 17:21:31,758 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69601.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 17:22:09,405 INFO [train2.py:809] (1/4) Epoch 18, batch 1900, loss[ctc_loss=0.09538, att_loss=0.2546, loss=0.2228, over 17100.00 frames. utt_duration=1223 frames, utt_pad_proportion=0.01528, over 56.00 utterances.], tot_loss[ctc_loss=0.07953, att_loss=0.2385, loss=0.2067, over 3270731.33 frames. utt_duration=1284 frames, utt_pad_proportion=0.04788, over 10203.96 utterances.], batch size: 56, lr: 6.01e-03, grad_scale: 16.0 2023-03-08 17:22:15,623 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69629.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 17:22:23,435 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.309e+02 2.048e+02 2.360e+02 2.878e+02 5.569e+02, threshold=4.720e+02, percent-clipped=3.0 2023-03-08 17:23:28,905 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69674.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 17:23:30,117 INFO [train2.py:809] (1/4) Epoch 18, batch 1950, loss[ctc_loss=0.07976, att_loss=0.2292, loss=0.1993, over 10166.00 frames. utt_duration=1850 frames, utt_pad_proportion=0.2502, over 22.00 utterances.], tot_loss[ctc_loss=0.0799, att_loss=0.2389, loss=0.2071, over 3270238.47 frames. utt_duration=1263 frames, utt_pad_proportion=0.05162, over 10366.84 utterances.], batch size: 22, lr: 6.01e-03, grad_scale: 16.0 2023-03-08 17:23:31,801 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69676.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:24:09,275 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69699.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:24:20,268 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.4281, 3.2311, 3.6051, 4.5496, 4.0463, 4.0205, 3.1611, 2.3662], device='cuda:1'), covar=tensor([0.0679, 0.1619, 0.0767, 0.0477, 0.0643, 0.0463, 0.1195, 0.2050], device='cuda:1'), in_proj_covar=tensor([0.0175, 0.0210, 0.0185, 0.0207, 0.0210, 0.0170, 0.0196, 0.0186], 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') 2023-03-08 17:24:48,850 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69724.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:24:50,208 INFO [train2.py:809] (1/4) Epoch 18, batch 2000, loss[ctc_loss=0.06931, att_loss=0.2433, loss=0.2085, over 16972.00 frames. utt_duration=1359 frames, utt_pad_proportion=0.007027, over 50.00 utterances.], tot_loss[ctc_loss=0.08057, att_loss=0.239, loss=0.2073, over 3273445.93 frames. utt_duration=1269 frames, utt_pad_proportion=0.04927, over 10329.08 utterances.], batch size: 50, lr: 6.00e-03, grad_scale: 8.0 2023-03-08 17:25:05,734 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.492e+02 2.144e+02 2.656e+02 3.349e+02 1.249e+03, threshold=5.313e+02, percent-clipped=13.0 2023-03-08 17:25:06,190 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69735.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 17:25:09,169 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.7485, 5.1047, 5.3593, 5.2158, 5.2674, 5.7020, 5.0693, 5.8203], device='cuda:1'), covar=tensor([0.0690, 0.0751, 0.0866, 0.1204, 0.1875, 0.0907, 0.0769, 0.0669], device='cuda:1'), in_proj_covar=tensor([0.0832, 0.0484, 0.0569, 0.0636, 0.0840, 0.0586, 0.0471, 0.0582], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 17:25:25,959 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-03-08 17:25:26,864 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=69747.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:25:31,595 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69750.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:25:49,077 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0579, 5.3154, 5.6036, 5.4671, 5.5262, 5.9820, 5.1887, 6.1318], device='cuda:1'), covar=tensor([0.0632, 0.0705, 0.0834, 0.1192, 0.1608, 0.0862, 0.0671, 0.0591], device='cuda:1'), in_proj_covar=tensor([0.0831, 0.0484, 0.0570, 0.0634, 0.0839, 0.0586, 0.0470, 0.0580], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 17:26:10,681 INFO [train2.py:809] (1/4) Epoch 18, batch 2050, loss[ctc_loss=0.06962, att_loss=0.2479, loss=0.2123, over 17012.00 frames. utt_duration=1336 frames, utt_pad_proportion=0.007648, over 51.00 utterances.], tot_loss[ctc_loss=0.08087, att_loss=0.2394, loss=0.2077, over 3272350.98 frames. utt_duration=1239 frames, utt_pad_proportion=0.05797, over 10574.46 utterances.], batch size: 51, lr: 6.00e-03, grad_scale: 8.0 2023-03-08 17:26:24,414 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9879, 6.2385, 5.6472, 5.9373, 5.9076, 5.4114, 5.6621, 5.4648], device='cuda:1'), covar=tensor([0.1279, 0.0845, 0.0912, 0.0771, 0.0895, 0.1412, 0.2214, 0.2194], device='cuda:1'), in_proj_covar=tensor([0.0506, 0.0578, 0.0441, 0.0439, 0.0415, 0.0454, 0.0599, 0.0512], 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') 2023-03-08 17:27:19,503 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69818.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:27:28,550 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69824.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:27:29,780 INFO [train2.py:809] (1/4) Epoch 18, batch 2100, loss[ctc_loss=0.08272, att_loss=0.2472, loss=0.2143, over 17163.00 frames. utt_duration=870.5 frames, utt_pad_proportion=0.08463, over 79.00 utterances.], tot_loss[ctc_loss=0.08045, att_loss=0.239, loss=0.2073, over 3272829.62 frames. utt_duration=1263 frames, utt_pad_proportion=0.0514, over 10375.13 utterances.], batch size: 79, lr: 6.00e-03, grad_scale: 8.0 2023-03-08 17:27:45,296 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.227e+02 1.987e+02 2.392e+02 3.047e+02 6.775e+02, threshold=4.784e+02, percent-clipped=2.0 2023-03-08 17:28:36,239 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=69866.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:28:45,328 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=69872.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:28:49,865 INFO [train2.py:809] (1/4) Epoch 18, batch 2150, loss[ctc_loss=0.07458, att_loss=0.2187, loss=0.1899, over 15786.00 frames. utt_duration=1663 frames, utt_pad_proportion=0.007772, over 38.00 utterances.], tot_loss[ctc_loss=0.0811, att_loss=0.2393, loss=0.2077, over 3265982.11 frames. utt_duration=1241 frames, utt_pad_proportion=0.05742, over 10540.68 utterances.], batch size: 38, lr: 6.00e-03, grad_scale: 8.0 2023-03-08 17:28:59,265 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69881.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:29:31,186 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69901.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 17:30:08,953 INFO [train2.py:809] (1/4) Epoch 18, batch 2200, loss[ctc_loss=0.08924, att_loss=0.2537, loss=0.2208, over 17363.00 frames. utt_duration=1008 frames, utt_pad_proportion=0.05016, over 69.00 utterances.], tot_loss[ctc_loss=0.08159, att_loss=0.2402, loss=0.2085, over 3273548.58 frames. utt_duration=1233 frames, utt_pad_proportion=0.05736, over 10630.14 utterances.], batch size: 69, lr: 6.00e-03, grad_scale: 8.0 2023-03-08 17:30:15,136 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=69929.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:30:15,347 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69929.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 17:30:24,844 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.283e+02 2.018e+02 2.397e+02 3.065e+02 5.583e+02, threshold=4.795e+02, percent-clipped=3.0 2023-03-08 17:30:47,811 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=69949.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 17:31:28,419 INFO [train2.py:809] (1/4) Epoch 18, batch 2250, loss[ctc_loss=0.07418, att_loss=0.2319, loss=0.2004, over 16259.00 frames. utt_duration=1514 frames, utt_pad_proportion=0.008468, over 43.00 utterances.], tot_loss[ctc_loss=0.08119, att_loss=0.2397, loss=0.208, over 3277750.21 frames. utt_duration=1235 frames, utt_pad_proportion=0.05693, over 10627.81 utterances.], batch size: 43, lr: 5.99e-03, grad_scale: 8.0 2023-03-08 17:31:30,222 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69976.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:31:31,491 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=69977.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:32:51,049 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=70024.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:32:51,251 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70024.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:32:52,447 INFO [train2.py:809] (1/4) Epoch 18, batch 2300, loss[ctc_loss=0.06278, att_loss=0.2486, loss=0.2114, over 16638.00 frames. utt_duration=1417 frames, utt_pad_proportion=0.004572, over 47.00 utterances.], tot_loss[ctc_loss=0.08022, att_loss=0.2398, loss=0.2079, over 3284385.84 frames. utt_duration=1245 frames, utt_pad_proportion=0.05305, over 10564.29 utterances.], batch size: 47, lr: 5.99e-03, grad_scale: 8.0 2023-03-08 17:33:00,998 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=70030.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 17:33:09,217 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.472e+02 2.121e+02 2.733e+02 3.440e+02 9.117e+02, threshold=5.466e+02, percent-clipped=8.0 2023-03-08 17:33:21,340 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.5298, 2.2365, 2.2719, 2.2684, 2.7295, 2.7784, 2.3920, 2.9473], device='cuda:1'), covar=tensor([0.1857, 0.3213, 0.2760, 0.2106, 0.1791, 0.1194, 0.2563, 0.1140], device='cuda:1'), in_proj_covar=tensor([0.0105, 0.0113, 0.0113, 0.0099, 0.0108, 0.0096, 0.0117, 0.0086], device='cuda:1'), out_proj_covar=tensor([7.7682e-05, 8.6366e-05, 8.7562e-05, 7.5716e-05, 7.9612e-05, 7.6359e-05, 8.6346e-05, 6.8674e-05], device='cuda:1') 2023-03-08 17:33:34,380 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70050.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:34:07,957 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=70072.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:34:12,452 INFO [train2.py:809] (1/4) Epoch 18, batch 2350, loss[ctc_loss=0.07675, att_loss=0.225, loss=0.1953, over 15907.00 frames. utt_duration=1633 frames, utt_pad_proportion=0.007914, over 39.00 utterances.], tot_loss[ctc_loss=0.0804, att_loss=0.24, loss=0.2081, over 3289336.48 frames. utt_duration=1248 frames, utt_pad_proportion=0.05091, over 10558.84 utterances.], batch size: 39, lr: 5.99e-03, grad_scale: 8.0 2023-03-08 17:34:50,260 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=70098.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:35:08,728 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70110.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:35:31,918 INFO [train2.py:809] (1/4) Epoch 18, batch 2400, loss[ctc_loss=0.05722, att_loss=0.2412, loss=0.2044, over 16778.00 frames. utt_duration=1400 frames, utt_pad_proportion=0.005815, over 48.00 utterances.], tot_loss[ctc_loss=0.0812, att_loss=0.2399, loss=0.2082, over 3267365.05 frames. utt_duration=1210 frames, utt_pad_proportion=0.06362, over 10817.14 utterances.], batch size: 48, lr: 5.99e-03, grad_scale: 8.0 2023-03-08 17:35:48,285 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.485e+02 2.097e+02 2.428e+02 2.770e+02 6.273e+02, threshold=4.856e+02, percent-clipped=1.0 2023-03-08 17:36:45,626 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=70171.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:36:52,023 INFO [train2.py:809] (1/4) Epoch 18, batch 2450, loss[ctc_loss=0.05738, att_loss=0.2068, loss=0.1769, over 15634.00 frames. utt_duration=1692 frames, utt_pad_proportion=0.008412, over 37.00 utterances.], tot_loss[ctc_loss=0.08147, att_loss=0.2403, loss=0.2086, over 3267654.53 frames. utt_duration=1182 frames, utt_pad_proportion=0.07149, over 11076.30 utterances.], batch size: 37, lr: 5.99e-03, grad_scale: 8.0 2023-03-08 17:37:11,183 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.0063, 3.6940, 3.7955, 3.1778, 3.8213, 3.8249, 3.7632, 2.7221], device='cuda:1'), covar=tensor([0.0887, 0.1280, 0.1658, 0.3302, 0.1157, 0.1344, 0.0689, 0.4141], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0168, 0.0177, 0.0236, 0.0142, 0.0238, 0.0157, 0.0209], 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') 2023-03-08 17:37:48,690 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1465, 5.1674, 4.9084, 3.2188, 4.8965, 4.7520, 4.3405, 2.8217], device='cuda:1'), covar=tensor([0.0109, 0.0097, 0.0283, 0.0851, 0.0096, 0.0165, 0.0302, 0.1237], device='cuda:1'), in_proj_covar=tensor([0.0073, 0.0097, 0.0096, 0.0109, 0.0081, 0.0106, 0.0097, 0.0102], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 17:38:11,071 INFO [train2.py:809] (1/4) Epoch 18, batch 2500, loss[ctc_loss=0.08274, att_loss=0.2471, loss=0.2142, over 16963.00 frames. utt_duration=1358 frames, utt_pad_proportion=0.007772, over 50.00 utterances.], tot_loss[ctc_loss=0.07977, att_loss=0.2388, loss=0.207, over 3267663.83 frames. utt_duration=1216 frames, utt_pad_proportion=0.06415, over 10765.46 utterances.], batch size: 50, lr: 5.98e-03, grad_scale: 8.0 2023-03-08 17:38:26,724 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.323e+02 2.106e+02 2.442e+02 2.921e+02 5.704e+02, threshold=4.884e+02, percent-clipped=1.0 2023-03-08 17:38:32,189 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1627, 5.1567, 4.9224, 3.1242, 4.9022, 4.7329, 4.3830, 2.7637], device='cuda:1'), covar=tensor([0.0141, 0.0113, 0.0289, 0.0949, 0.0108, 0.0185, 0.0302, 0.1363], device='cuda:1'), in_proj_covar=tensor([0.0073, 0.0098, 0.0097, 0.0109, 0.0082, 0.0107, 0.0097, 0.0102], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 17:39:06,537 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70260.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:39:14,714 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.8547, 3.5968, 3.6987, 3.1743, 3.6762, 3.7442, 3.6915, 2.7146], device='cuda:1'), covar=tensor([0.0981, 0.1374, 0.1422, 0.3342, 0.1194, 0.1802, 0.0834, 0.4646], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0169, 0.0179, 0.0239, 0.0144, 0.0241, 0.0158, 0.0211], 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') 2023-03-08 17:39:31,633 INFO [train2.py:809] (1/4) Epoch 18, batch 2550, loss[ctc_loss=0.06161, att_loss=0.2254, loss=0.1927, over 15948.00 frames. utt_duration=1558 frames, utt_pad_proportion=0.006642, over 41.00 utterances.], tot_loss[ctc_loss=0.07949, att_loss=0.239, loss=0.2071, over 3274245.40 frames. utt_duration=1238 frames, utt_pad_proportion=0.05596, over 10590.51 utterances.], batch size: 41, lr: 5.98e-03, grad_scale: 8.0 2023-03-08 17:39:45,884 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1791, 5.4822, 4.9760, 5.5214, 4.8765, 5.0865, 5.6007, 5.3668], device='cuda:1'), covar=tensor([0.0473, 0.0302, 0.0776, 0.0322, 0.0413, 0.0217, 0.0233, 0.0196], device='cuda:1'), in_proj_covar=tensor([0.0370, 0.0303, 0.0351, 0.0319, 0.0307, 0.0231, 0.0286, 0.0271], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 17:39:48,746 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-03-08 17:40:31,403 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7042, 4.5740, 4.7040, 4.5467, 5.3175, 4.4850, 4.5924, 2.3638], device='cuda:1'), covar=tensor([0.0192, 0.0329, 0.0287, 0.0312, 0.0667, 0.0203, 0.0316, 0.1957], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0165, 0.0167, 0.0185, 0.0355, 0.0141, 0.0157, 0.0211], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 17:40:43,947 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=70321.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:40:50,213 INFO [train2.py:809] (1/4) Epoch 18, batch 2600, loss[ctc_loss=0.08823, att_loss=0.2274, loss=0.1996, over 15898.00 frames. utt_duration=1632 frames, utt_pad_proportion=0.008568, over 39.00 utterances.], tot_loss[ctc_loss=0.0798, att_loss=0.2393, loss=0.2074, over 3270061.06 frames. utt_duration=1249 frames, utt_pad_proportion=0.05358, over 10488.61 utterances.], batch size: 39, lr: 5.98e-03, grad_scale: 8.0 2023-03-08 17:40:58,248 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70330.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 17:41:05,532 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.324e+02 1.967e+02 2.510e+02 3.094e+02 6.487e+02, threshold=5.020e+02, percent-clipped=1.0 2023-03-08 17:41:38,787 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-03-08 17:41:44,468 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7558, 3.3753, 3.5193, 2.7476, 3.6045, 3.5366, 3.5754, 2.0830], device='cuda:1'), covar=tensor([0.1119, 0.1582, 0.1821, 0.7072, 0.0986, 0.2321, 0.0877, 0.8142], device='cuda:1'), in_proj_covar=tensor([0.0149, 0.0167, 0.0177, 0.0236, 0.0141, 0.0237, 0.0156, 0.0207], 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') 2023-03-08 17:42:08,699 INFO [train2.py:809] (1/4) Epoch 18, batch 2650, loss[ctc_loss=0.08749, att_loss=0.2473, loss=0.2153, over 17381.00 frames. utt_duration=881.6 frames, utt_pad_proportion=0.07591, over 79.00 utterances.], tot_loss[ctc_loss=0.07988, att_loss=0.2392, loss=0.2073, over 3271877.69 frames. utt_duration=1241 frames, utt_pad_proportion=0.05486, over 10558.07 utterances.], batch size: 79, lr: 5.98e-03, grad_scale: 8.0 2023-03-08 17:42:13,358 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=70378.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 17:42:34,418 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-03-08 17:43:27,230 INFO [train2.py:809] (1/4) Epoch 18, batch 2700, loss[ctc_loss=0.104, att_loss=0.2587, loss=0.2277, over 17013.00 frames. utt_duration=1336 frames, utt_pad_proportion=0.008195, over 51.00 utterances.], tot_loss[ctc_loss=0.07968, att_loss=0.2389, loss=0.2071, over 3274126.69 frames. utt_duration=1266 frames, utt_pad_proportion=0.04847, over 10358.96 utterances.], batch size: 51, lr: 5.98e-03, grad_scale: 8.0 2023-03-08 17:43:30,676 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.3744, 2.3936, 3.0760, 2.5859, 3.0433, 3.4922, 3.4664, 2.6122], device='cuda:1'), covar=tensor([0.0500, 0.1714, 0.1165, 0.1226, 0.1015, 0.1195, 0.0600, 0.1454], device='cuda:1'), in_proj_covar=tensor([0.0240, 0.0240, 0.0270, 0.0213, 0.0261, 0.0352, 0.0249, 0.0229], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 17:43:43,122 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.380e+02 1.957e+02 2.578e+02 3.179e+02 5.678e+02, threshold=5.155e+02, percent-clipped=6.0 2023-03-08 17:43:54,186 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9831, 4.2059, 4.0904, 4.3975, 2.4589, 4.4304, 2.7170, 1.8273], device='cuda:1'), covar=tensor([0.0504, 0.0241, 0.0871, 0.0214, 0.2232, 0.0192, 0.1813, 0.1995], device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0150, 0.0256, 0.0145, 0.0221, 0.0129, 0.0231, 0.0203], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 17:44:15,438 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.5516, 5.7992, 5.2185, 5.5668, 5.4152, 4.9380, 5.1749, 5.0134], device='cuda:1'), covar=tensor([0.1291, 0.0814, 0.0940, 0.0759, 0.1000, 0.1502, 0.2319, 0.2056], device='cuda:1'), in_proj_covar=tensor([0.0496, 0.0577, 0.0439, 0.0435, 0.0412, 0.0450, 0.0591, 0.0505], 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') 2023-03-08 17:44:31,286 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=70466.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:44:33,976 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.62 vs. limit=5.0 2023-03-08 17:44:46,107 INFO [train2.py:809] (1/4) Epoch 18, batch 2750, loss[ctc_loss=0.0753, att_loss=0.228, loss=0.1975, over 16279.00 frames. utt_duration=1515 frames, utt_pad_proportion=0.007584, over 43.00 utterances.], tot_loss[ctc_loss=0.07913, att_loss=0.2387, loss=0.2068, over 3278824.12 frames. utt_duration=1275 frames, utt_pad_proportion=0.04547, over 10302.06 utterances.], batch size: 43, lr: 5.97e-03, grad_scale: 8.0 2023-03-08 17:46:05,051 INFO [train2.py:809] (1/4) Epoch 18, batch 2800, loss[ctc_loss=0.06334, att_loss=0.2315, loss=0.1978, over 16478.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.006056, over 46.00 utterances.], tot_loss[ctc_loss=0.07918, att_loss=0.2388, loss=0.2069, over 3280936.89 frames. utt_duration=1288 frames, utt_pad_proportion=0.04272, over 10200.52 utterances.], batch size: 46, lr: 5.97e-03, grad_scale: 8.0 2023-03-08 17:46:20,627 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.291e+02 1.948e+02 2.363e+02 2.957e+02 5.460e+02, threshold=4.725e+02, percent-clipped=2.0 2023-03-08 17:47:24,019 INFO [train2.py:809] (1/4) Epoch 18, batch 2850, loss[ctc_loss=0.07896, att_loss=0.2554, loss=0.2201, over 17052.00 frames. utt_duration=1313 frames, utt_pad_proportion=0.008999, over 52.00 utterances.], tot_loss[ctc_loss=0.07885, att_loss=0.2386, loss=0.2066, over 3272684.50 frames. utt_duration=1281 frames, utt_pad_proportion=0.04626, over 10233.67 utterances.], batch size: 52, lr: 5.97e-03, grad_scale: 8.0 2023-03-08 17:48:29,691 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=70616.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:48:35,411 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-03-08 17:48:43,293 INFO [train2.py:809] (1/4) Epoch 18, batch 2900, loss[ctc_loss=0.08064, att_loss=0.2509, loss=0.2169, over 16975.00 frames. utt_duration=1360 frames, utt_pad_proportion=0.006822, over 50.00 utterances.], tot_loss[ctc_loss=0.07907, att_loss=0.2384, loss=0.2066, over 3267212.75 frames. utt_duration=1275 frames, utt_pad_proportion=0.05001, over 10262.16 utterances.], batch size: 50, lr: 5.97e-03, grad_scale: 8.0 2023-03-08 17:48:59,499 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.265e+02 1.930e+02 2.447e+02 2.876e+02 5.952e+02, threshold=4.894e+02, percent-clipped=4.0 2023-03-08 17:50:04,767 INFO [train2.py:809] (1/4) Epoch 18, batch 2950, loss[ctc_loss=0.08018, att_loss=0.2471, loss=0.2137, over 17052.00 frames. utt_duration=1289 frames, utt_pad_proportion=0.008752, over 53.00 utterances.], tot_loss[ctc_loss=0.0792, att_loss=0.239, loss=0.207, over 3271097.52 frames. utt_duration=1254 frames, utt_pad_proportion=0.05445, over 10445.82 utterances.], batch size: 53, lr: 5.96e-03, grad_scale: 8.0 2023-03-08 17:50:16,366 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.8941, 5.2289, 5.4639, 5.2731, 5.3954, 5.8216, 5.1989, 5.9117], device='cuda:1'), covar=tensor([0.0679, 0.0770, 0.0931, 0.1422, 0.1798, 0.1068, 0.0690, 0.0801], device='cuda:1'), in_proj_covar=tensor([0.0831, 0.0490, 0.0571, 0.0639, 0.0837, 0.0594, 0.0470, 0.0586], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 17:50:38,918 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3415, 4.4535, 4.6156, 4.3387, 5.0967, 4.4298, 4.4234, 2.2467], device='cuda:1'), covar=tensor([0.0267, 0.0309, 0.0266, 0.0359, 0.0822, 0.0238, 0.0346, 0.2273], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0166, 0.0169, 0.0187, 0.0357, 0.0142, 0.0158, 0.0213], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 17:51:18,072 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1938, 5.0572, 4.9039, 3.0912, 4.8874, 4.6796, 4.3079, 2.9910], device='cuda:1'), covar=tensor([0.0095, 0.0098, 0.0257, 0.0869, 0.0092, 0.0204, 0.0304, 0.1159], device='cuda:1'), in_proj_covar=tensor([0.0073, 0.0097, 0.0096, 0.0108, 0.0080, 0.0106, 0.0096, 0.0101], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 17:51:26,431 INFO [train2.py:809] (1/4) Epoch 18, batch 3000, loss[ctc_loss=0.06745, att_loss=0.2087, loss=0.1805, over 15346.00 frames. utt_duration=1755 frames, utt_pad_proportion=0.01109, over 35.00 utterances.], tot_loss[ctc_loss=0.07885, att_loss=0.2382, loss=0.2064, over 3272063.07 frames. utt_duration=1284 frames, utt_pad_proportion=0.04692, over 10202.31 utterances.], batch size: 35, lr: 5.96e-03, grad_scale: 8.0 2023-03-08 17:51:26,431 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-08 17:51:43,589 INFO [train2.py:843] (1/4) Epoch 18, validation: ctc_loss=0.04147, att_loss=0.2347, loss=0.196, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-08 17:51:43,590 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-08 17:51:59,784 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.535e+02 1.987e+02 2.490e+02 3.021e+02 7.545e+02, threshold=4.980e+02, percent-clipped=3.0 2023-03-08 17:52:49,421 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70766.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:53:03,425 INFO [train2.py:809] (1/4) Epoch 18, batch 3050, loss[ctc_loss=0.09114, att_loss=0.2477, loss=0.2164, over 16629.00 frames. utt_duration=1417 frames, utt_pad_proportion=0.00499, over 47.00 utterances.], tot_loss[ctc_loss=0.07864, att_loss=0.238, loss=0.2061, over 3278887.67 frames. utt_duration=1284 frames, utt_pad_proportion=0.04449, over 10230.27 utterances.], batch size: 47, lr: 5.96e-03, grad_scale: 8.0 2023-03-08 17:54:07,348 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=70814.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:54:26,016 INFO [train2.py:809] (1/4) Epoch 18, batch 3100, loss[ctc_loss=0.07961, att_loss=0.2417, loss=0.2093, over 17088.00 frames. utt_duration=1222 frames, utt_pad_proportion=0.01694, over 56.00 utterances.], tot_loss[ctc_loss=0.07924, att_loss=0.2383, loss=0.2065, over 3274214.23 frames. utt_duration=1271 frames, utt_pad_proportion=0.04779, over 10319.61 utterances.], batch size: 56, lr: 5.96e-03, grad_scale: 8.0 2023-03-08 17:54:41,867 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.217e+02 2.010e+02 2.337e+02 3.007e+02 9.236e+02, threshold=4.674e+02, percent-clipped=2.0 2023-03-08 17:55:35,489 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.68 vs. limit=2.0 2023-03-08 17:55:46,924 INFO [train2.py:809] (1/4) Epoch 18, batch 3150, loss[ctc_loss=0.05139, att_loss=0.2286, loss=0.1931, over 16467.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.006508, over 46.00 utterances.], tot_loss[ctc_loss=0.07969, att_loss=0.2384, loss=0.2066, over 3268097.39 frames. utt_duration=1248 frames, utt_pad_proportion=0.055, over 10489.70 utterances.], batch size: 46, lr: 5.96e-03, grad_scale: 8.0 2023-03-08 17:56:19,732 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-03-08 17:56:52,239 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70916.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:57:07,454 INFO [train2.py:809] (1/4) Epoch 18, batch 3200, loss[ctc_loss=0.08286, att_loss=0.2532, loss=0.2192, over 17391.00 frames. utt_duration=882.4 frames, utt_pad_proportion=0.07604, over 79.00 utterances.], tot_loss[ctc_loss=0.08024, att_loss=0.239, loss=0.2073, over 3268515.06 frames. utt_duration=1220 frames, utt_pad_proportion=0.06243, over 10729.69 utterances.], batch size: 79, lr: 5.95e-03, grad_scale: 8.0 2023-03-08 17:57:23,347 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.324e+02 2.043e+02 2.547e+02 2.997e+02 7.270e+02, threshold=5.093e+02, percent-clipped=4.0 2023-03-08 17:58:10,131 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=70964.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:58:28,690 INFO [train2.py:809] (1/4) Epoch 18, batch 3250, loss[ctc_loss=0.1355, att_loss=0.2714, loss=0.2442, over 14546.00 frames. utt_duration=402.7 frames, utt_pad_proportion=0.3009, over 145.00 utterances.], tot_loss[ctc_loss=0.08038, att_loss=0.2397, loss=0.2078, over 3280234.67 frames. utt_duration=1222 frames, utt_pad_proportion=0.05892, over 10751.69 utterances.], batch size: 145, lr: 5.95e-03, grad_scale: 8.0 2023-03-08 17:58:34,241 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.8653, 5.1549, 5.4427, 5.2862, 5.2882, 5.8511, 5.1234, 5.9451], device='cuda:1'), covar=tensor([0.0720, 0.0739, 0.0801, 0.1235, 0.1924, 0.0892, 0.0665, 0.0664], device='cuda:1'), in_proj_covar=tensor([0.0837, 0.0487, 0.0571, 0.0635, 0.0840, 0.0594, 0.0469, 0.0585], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 17:58:56,368 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70992.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 17:59:47,177 INFO [train2.py:809] (1/4) Epoch 18, batch 3300, loss[ctc_loss=0.1042, att_loss=0.2624, loss=0.2307, over 17400.00 frames. utt_duration=1107 frames, utt_pad_proportion=0.03193, over 63.00 utterances.], tot_loss[ctc_loss=0.08065, att_loss=0.2394, loss=0.2077, over 3281290.86 frames. utt_duration=1243 frames, utt_pad_proportion=0.05281, over 10568.39 utterances.], batch size: 63, lr: 5.95e-03, grad_scale: 8.0 2023-03-08 17:59:55,965 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=71030.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 18:00:02,926 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.258e+02 1.989e+02 2.342e+02 2.870e+02 6.957e+02, threshold=4.685e+02, percent-clipped=3.0 2023-03-08 18:00:23,637 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2015, 5.2289, 5.0112, 2.4273, 2.1590, 3.1095, 2.6738, 4.0582], device='cuda:1'), covar=tensor([0.0643, 0.0317, 0.0262, 0.4987, 0.5433, 0.2182, 0.3127, 0.1615], device='cuda:1'), in_proj_covar=tensor([0.0351, 0.0263, 0.0259, 0.0239, 0.0345, 0.0332, 0.0248, 0.0363], device='cuda:1'), out_proj_covar=tensor([1.4948e-04, 9.7775e-05, 1.1091e-04, 1.0357e-04, 1.4514e-04, 1.3064e-04, 9.8964e-05, 1.4878e-04], device='cuda:1') 2023-03-08 18:00:23,969 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-03-08 18:00:31,326 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71053.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 18:01:01,594 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-03-08 18:01:05,451 INFO [train2.py:809] (1/4) Epoch 18, batch 3350, loss[ctc_loss=0.08691, att_loss=0.25, loss=0.2174, over 17030.00 frames. utt_duration=1337 frames, utt_pad_proportion=0.008102, over 51.00 utterances.], tot_loss[ctc_loss=0.08066, att_loss=0.2396, loss=0.2078, over 3280194.34 frames. utt_duration=1243 frames, utt_pad_proportion=0.05241, over 10572.21 utterances.], batch size: 51, lr: 5.95e-03, grad_scale: 8.0 2023-03-08 18:01:31,189 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71091.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 18:02:25,947 INFO [train2.py:809] (1/4) Epoch 18, batch 3400, loss[ctc_loss=0.05402, att_loss=0.2324, loss=0.1968, over 16964.00 frames. utt_duration=1359 frames, utt_pad_proportion=0.007611, over 50.00 utterances.], tot_loss[ctc_loss=0.07949, att_loss=0.2389, loss=0.207, over 3277437.42 frames. utt_duration=1255 frames, utt_pad_proportion=0.0502, over 10457.83 utterances.], batch size: 50, lr: 5.95e-03, grad_scale: 8.0 2023-03-08 18:02:35,213 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.93 vs. limit=2.0 2023-03-08 18:02:42,143 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.526e+02 2.089e+02 2.385e+02 3.204e+02 8.386e+02, threshold=4.770e+02, percent-clipped=4.0 2023-03-08 18:03:01,768 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5476, 2.8748, 3.6385, 2.9472, 3.4272, 4.6568, 4.4421, 3.0486], device='cuda:1'), covar=tensor([0.0361, 0.1825, 0.1169, 0.1381, 0.1149, 0.0676, 0.0512, 0.1489], device='cuda:1'), in_proj_covar=tensor([0.0240, 0.0238, 0.0269, 0.0213, 0.0256, 0.0349, 0.0249, 0.0227], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 18:03:12,823 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2266, 2.2883, 2.7033, 4.2587, 3.8523, 3.8363, 2.8555, 1.8442], device='cuda:1'), covar=tensor([0.0730, 0.2571, 0.1400, 0.0592, 0.0763, 0.0438, 0.1451, 0.2710], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0215, 0.0188, 0.0212, 0.0216, 0.0173, 0.0200, 0.0190], 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') 2023-03-08 18:03:46,801 INFO [train2.py:809] (1/4) Epoch 18, batch 3450, loss[ctc_loss=0.0867, att_loss=0.2456, loss=0.2138, over 16931.00 frames. utt_duration=692.6 frames, utt_pad_proportion=0.132, over 98.00 utterances.], tot_loss[ctc_loss=0.07968, att_loss=0.2395, loss=0.2075, over 3278899.61 frames. utt_duration=1242 frames, utt_pad_proportion=0.05415, over 10573.49 utterances.], batch size: 98, lr: 5.94e-03, grad_scale: 8.0 2023-03-08 18:04:31,253 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-03-08 18:04:51,649 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-03-08 18:05:06,129 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-03-08 18:05:06,672 INFO [train2.py:809] (1/4) Epoch 18, batch 3500, loss[ctc_loss=0.06138, att_loss=0.2145, loss=0.1839, over 16126.00 frames. utt_duration=1537 frames, utt_pad_proportion=0.00631, over 42.00 utterances.], tot_loss[ctc_loss=0.07931, att_loss=0.2389, loss=0.207, over 3262837.62 frames. utt_duration=1228 frames, utt_pad_proportion=0.0636, over 10637.91 utterances.], batch size: 42, lr: 5.94e-03, grad_scale: 8.0 2023-03-08 18:05:22,701 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.257e+02 1.989e+02 2.438e+02 2.819e+02 5.372e+02, threshold=4.876e+02, percent-clipped=3.0 2023-03-08 18:05:30,020 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-03-08 18:06:03,805 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0314, 6.2048, 5.7136, 5.9445, 5.8812, 5.3252, 5.7061, 5.3635], device='cuda:1'), covar=tensor([0.1049, 0.0925, 0.0866, 0.0842, 0.0863, 0.1420, 0.2136, 0.2442], device='cuda:1'), in_proj_covar=tensor([0.0497, 0.0580, 0.0435, 0.0432, 0.0412, 0.0449, 0.0583, 0.0506], 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') 2023-03-08 18:06:26,581 INFO [train2.py:809] (1/4) Epoch 18, batch 3550, loss[ctc_loss=0.06301, att_loss=0.2163, loss=0.1857, over 15347.00 frames. utt_duration=1756 frames, utt_pad_proportion=0.01037, over 35.00 utterances.], tot_loss[ctc_loss=0.07875, att_loss=0.2382, loss=0.2063, over 3266828.76 frames. utt_duration=1255 frames, utt_pad_proportion=0.05563, over 10422.08 utterances.], batch size: 35, lr: 5.94e-03, grad_scale: 8.0 2023-03-08 18:07:46,257 INFO [train2.py:809] (1/4) Epoch 18, batch 3600, loss[ctc_loss=0.126, att_loss=0.2628, loss=0.2354, over 16982.00 frames. utt_duration=687.7 frames, utt_pad_proportion=0.1361, over 99.00 utterances.], tot_loss[ctc_loss=0.07992, att_loss=0.239, loss=0.2072, over 3270672.22 frames. utt_duration=1241 frames, utt_pad_proportion=0.05743, over 10558.63 utterances.], batch size: 99, lr: 5.94e-03, grad_scale: 8.0 2023-03-08 18:08:02,042 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.388e+02 1.997e+02 2.305e+02 2.777e+02 7.963e+02, threshold=4.609e+02, percent-clipped=3.0 2023-03-08 18:08:23,284 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=71348.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 18:09:09,354 INFO [train2.py:809] (1/4) Epoch 18, batch 3650, loss[ctc_loss=0.07434, att_loss=0.2296, loss=0.1986, over 16700.00 frames. utt_duration=1454 frames, utt_pad_proportion=0.005754, over 46.00 utterances.], tot_loss[ctc_loss=0.07986, att_loss=0.2388, loss=0.207, over 3265537.56 frames. utt_duration=1237 frames, utt_pad_proportion=0.05889, over 10568.49 utterances.], batch size: 46, lr: 5.94e-03, grad_scale: 8.0 2023-03-08 18:09:28,402 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=71386.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 18:09:31,935 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4643, 2.5831, 4.9714, 3.7385, 3.1734, 4.1865, 4.6935, 4.5545], device='cuda:1'), covar=tensor([0.0239, 0.1717, 0.0163, 0.1029, 0.1681, 0.0274, 0.0136, 0.0229], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0237, 0.0168, 0.0306, 0.0262, 0.0203, 0.0154, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 18:10:33,059 INFO [train2.py:809] (1/4) Epoch 18, batch 3700, loss[ctc_loss=0.07583, att_loss=0.2413, loss=0.2082, over 17160.00 frames. utt_duration=694.7 frames, utt_pad_proportion=0.1273, over 99.00 utterances.], tot_loss[ctc_loss=0.08014, att_loss=0.2389, loss=0.2072, over 3256610.08 frames. utt_duration=1187 frames, utt_pad_proportion=0.07468, over 10984.29 utterances.], batch size: 99, lr: 5.93e-03, grad_scale: 8.0 2023-03-08 18:10:49,963 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.274e+02 2.052e+02 2.398e+02 2.900e+02 5.315e+02, threshold=4.797e+02, percent-clipped=3.0 2023-03-08 18:11:56,329 INFO [train2.py:809] (1/4) Epoch 18, batch 3750, loss[ctc_loss=0.06862, att_loss=0.2097, loss=0.1815, over 15772.00 frames. utt_duration=1662 frames, utt_pad_proportion=0.008589, over 38.00 utterances.], tot_loss[ctc_loss=0.07979, att_loss=0.2382, loss=0.2065, over 3247592.96 frames. utt_duration=1177 frames, utt_pad_proportion=0.08071, over 11052.75 utterances.], batch size: 38, lr: 5.93e-03, grad_scale: 8.0 2023-03-08 18:12:26,286 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1235, 5.1315, 4.8759, 3.0672, 4.8698, 4.6766, 4.4523, 2.8818], device='cuda:1'), covar=tensor([0.0142, 0.0088, 0.0267, 0.0925, 0.0096, 0.0194, 0.0252, 0.1278], device='cuda:1'), in_proj_covar=tensor([0.0073, 0.0098, 0.0097, 0.0108, 0.0081, 0.0107, 0.0097, 0.0102], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 18:13:18,686 INFO [train2.py:809] (1/4) Epoch 18, batch 3800, loss[ctc_loss=0.08582, att_loss=0.2569, loss=0.2227, over 17142.00 frames. utt_duration=1226 frames, utt_pad_proportion=0.01311, over 56.00 utterances.], tot_loss[ctc_loss=0.07959, att_loss=0.2383, loss=0.2065, over 3249370.03 frames. utt_duration=1203 frames, utt_pad_proportion=0.07426, over 10814.56 utterances.], batch size: 56, lr: 5.93e-03, grad_scale: 8.0 2023-03-08 18:13:34,473 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.420e+02 2.121e+02 2.472e+02 3.058e+02 5.629e+02, threshold=4.945e+02, percent-clipped=2.0 2023-03-08 18:13:34,906 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5996, 3.2904, 5.0168, 4.1782, 3.4713, 4.5066, 4.8447, 4.6717], device='cuda:1'), covar=tensor([0.0294, 0.1227, 0.0256, 0.0968, 0.1497, 0.0240, 0.0174, 0.0285], device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0239, 0.0169, 0.0308, 0.0264, 0.0205, 0.0156, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 18:14:12,881 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-03-08 18:14:37,812 INFO [train2.py:809] (1/4) Epoch 18, batch 3850, loss[ctc_loss=0.0739, att_loss=0.2325, loss=0.2008, over 16124.00 frames. utt_duration=1537 frames, utt_pad_proportion=0.00576, over 42.00 utterances.], tot_loss[ctc_loss=0.07974, att_loss=0.2381, loss=0.2064, over 3241327.53 frames. utt_duration=1211 frames, utt_pad_proportion=0.07433, over 10721.80 utterances.], batch size: 42, lr: 5.93e-03, grad_scale: 8.0 2023-03-08 18:15:17,994 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.7748, 3.9422, 3.9310, 3.9447, 4.0143, 3.8339, 3.0629, 3.9120], device='cuda:1'), covar=tensor([0.0135, 0.0138, 0.0142, 0.0097, 0.0099, 0.0122, 0.0630, 0.0190], device='cuda:1'), in_proj_covar=tensor([0.0088, 0.0086, 0.0106, 0.0066, 0.0072, 0.0083, 0.0102, 0.0106], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 18:15:41,493 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=71616.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 18:15:54,890 INFO [train2.py:809] (1/4) Epoch 18, batch 3900, loss[ctc_loss=0.0734, att_loss=0.2315, loss=0.1999, over 16401.00 frames. utt_duration=1493 frames, utt_pad_proportion=0.006971, over 44.00 utterances.], tot_loss[ctc_loss=0.07993, att_loss=0.2383, loss=0.2066, over 3245471.87 frames. utt_duration=1204 frames, utt_pad_proportion=0.07611, over 10799.66 utterances.], batch size: 44, lr: 5.93e-03, grad_scale: 8.0 2023-03-08 18:16:03,373 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6410, 5.0493, 4.9334, 4.8400, 5.1342, 4.6841, 3.7281, 4.9918], device='cuda:1'), covar=tensor([0.0120, 0.0108, 0.0116, 0.0099, 0.0085, 0.0116, 0.0616, 0.0203], device='cuda:1'), in_proj_covar=tensor([0.0088, 0.0085, 0.0105, 0.0066, 0.0071, 0.0082, 0.0101, 0.0105], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 18:16:10,967 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.518e+02 2.131e+02 2.535e+02 3.133e+02 8.645e+02, threshold=5.071e+02, percent-clipped=4.0 2023-03-08 18:16:31,460 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=71648.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 18:17:13,024 INFO [train2.py:809] (1/4) Epoch 18, batch 3950, loss[ctc_loss=0.08468, att_loss=0.2233, loss=0.1956, over 11517.00 frames. utt_duration=1844 frames, utt_pad_proportion=0.1839, over 25.00 utterances.], tot_loss[ctc_loss=0.07948, att_loss=0.2384, loss=0.2066, over 3251384.85 frames. utt_duration=1225 frames, utt_pad_proportion=0.06819, over 10628.13 utterances.], batch size: 25, lr: 5.92e-03, grad_scale: 8.0 2023-03-08 18:17:16,264 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71677.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 18:17:27,720 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-03-08 18:17:29,984 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=71686.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 18:17:45,661 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=71696.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 18:18:26,972 INFO [train2.py:809] (1/4) Epoch 19, batch 0, loss[ctc_loss=0.06647, att_loss=0.2402, loss=0.2054, over 17358.00 frames. utt_duration=1008 frames, utt_pad_proportion=0.04928, over 69.00 utterances.], tot_loss[ctc_loss=0.06647, att_loss=0.2402, loss=0.2054, over 17358.00 frames. utt_duration=1008 frames, utt_pad_proportion=0.04928, over 69.00 utterances.], batch size: 69, lr: 5.76e-03, grad_scale: 16.0 2023-03-08 18:18:26,973 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-08 18:18:38,963 INFO [train2.py:843] (1/4) Epoch 19, validation: ctc_loss=0.04291, att_loss=0.2348, loss=0.1964, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-08 18:18:38,963 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-08 18:19:20,133 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=71734.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 18:19:21,555 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.522e+02 2.028e+02 2.615e+02 3.173e+02 7.413e+02, threshold=5.229e+02, percent-clipped=5.0 2023-03-08 18:19:34,198 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.3393, 2.1128, 2.1429, 2.2851, 2.4511, 2.6253, 2.2409, 2.7743], device='cuda:1'), covar=tensor([0.4829, 0.5697, 0.3870, 0.3755, 0.4688, 0.2348, 0.5102, 0.2428], device='cuda:1'), in_proj_covar=tensor([0.0110, 0.0117, 0.0114, 0.0102, 0.0114, 0.0100, 0.0121, 0.0089], device='cuda:1'), out_proj_covar=tensor([8.0986e-05, 8.9229e-05, 8.9289e-05, 7.8470e-05, 8.3610e-05, 7.8828e-05, 8.9469e-05, 7.1609e-05], device='cuda:1') 2023-03-08 18:19:57,603 INFO [train2.py:809] (1/4) Epoch 19, batch 50, loss[ctc_loss=0.0625, att_loss=0.2087, loss=0.1795, over 15618.00 frames. utt_duration=1690 frames, utt_pad_proportion=0.01004, over 37.00 utterances.], tot_loss[ctc_loss=0.07942, att_loss=0.2358, loss=0.2045, over 731280.25 frames. utt_duration=1248 frames, utt_pad_proportion=0.06905, over 2345.83 utterances.], batch size: 37, lr: 5.76e-03, grad_scale: 16.0 2023-03-08 18:21:17,853 INFO [train2.py:809] (1/4) Epoch 19, batch 100, loss[ctc_loss=0.08012, att_loss=0.232, loss=0.2016, over 16333.00 frames. utt_duration=1453 frames, utt_pad_proportion=0.005852, over 45.00 utterances.], tot_loss[ctc_loss=0.0777, att_loss=0.2363, loss=0.2045, over 1297419.71 frames. utt_duration=1297 frames, utt_pad_proportion=0.0493, over 4005.37 utterances.], batch size: 45, lr: 5.76e-03, grad_scale: 16.0 2023-03-08 18:22:00,749 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.482e+02 2.106e+02 2.518e+02 3.036e+02 5.583e+02, threshold=5.037e+02, percent-clipped=2.0 2023-03-08 18:22:21,247 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-03-08 18:22:36,463 INFO [train2.py:809] (1/4) Epoch 19, batch 150, loss[ctc_loss=0.06932, att_loss=0.231, loss=0.1987, over 16407.00 frames. utt_duration=1493 frames, utt_pad_proportion=0.006714, over 44.00 utterances.], tot_loss[ctc_loss=0.07869, att_loss=0.238, loss=0.2062, over 1736994.76 frames. utt_duration=1262 frames, utt_pad_proportion=0.05396, over 5513.64 utterances.], batch size: 44, lr: 5.76e-03, grad_scale: 16.0 2023-03-08 18:23:56,063 INFO [train2.py:809] (1/4) Epoch 19, batch 200, loss[ctc_loss=0.07559, att_loss=0.2346, loss=0.2028, over 16142.00 frames. utt_duration=1539 frames, utt_pad_proportion=0.005264, over 42.00 utterances.], tot_loss[ctc_loss=0.0778, att_loss=0.2372, loss=0.2053, over 2076605.52 frames. utt_duration=1268 frames, utt_pad_proportion=0.05154, over 6557.15 utterances.], batch size: 42, lr: 5.75e-03, grad_scale: 16.0 2023-03-08 18:24:10,802 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-03-08 18:24:14,386 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=71919.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 18:24:39,025 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.141e+02 1.975e+02 2.308e+02 2.986e+02 5.349e+02, threshold=4.617e+02, percent-clipped=2.0 2023-03-08 18:25:14,792 INFO [train2.py:809] (1/4) Epoch 19, batch 250, loss[ctc_loss=0.05767, att_loss=0.2295, loss=0.1952, over 16538.00 frames. utt_duration=1472 frames, utt_pad_proportion=0.006392, over 45.00 utterances.], tot_loss[ctc_loss=0.07795, att_loss=0.2367, loss=0.205, over 2336463.87 frames. utt_duration=1284 frames, utt_pad_proportion=0.04858, over 7287.67 utterances.], batch size: 45, lr: 5.75e-03, grad_scale: 16.0 2023-03-08 18:25:37,760 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=71972.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 18:25:51,039 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71980.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 18:26:39,091 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3539, 2.7893, 3.5321, 2.8165, 3.5392, 4.5035, 4.3461, 3.0759], device='cuda:1'), covar=tensor([0.0393, 0.1760, 0.1297, 0.1433, 0.0980, 0.0788, 0.0491, 0.1408], device='cuda:1'), in_proj_covar=tensor([0.0240, 0.0241, 0.0273, 0.0216, 0.0261, 0.0353, 0.0253, 0.0229], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 18:26:40,274 INFO [train2.py:809] (1/4) Epoch 19, batch 300, loss[ctc_loss=0.09492, att_loss=0.2577, loss=0.2251, over 17105.00 frames. utt_duration=1223 frames, utt_pad_proportion=0.01592, over 56.00 utterances.], tot_loss[ctc_loss=0.0771, att_loss=0.2355, loss=0.2038, over 2544664.88 frames. utt_duration=1279 frames, utt_pad_proportion=0.04938, over 7966.27 utterances.], batch size: 56, lr: 5.75e-03, grad_scale: 16.0 2023-03-08 18:27:10,081 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-03-08 18:27:12,127 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0739, 5.0991, 4.8198, 2.6575, 4.9376, 4.6082, 4.3508, 2.5209], device='cuda:1'), covar=tensor([0.0113, 0.0098, 0.0282, 0.1182, 0.0091, 0.0220, 0.0300, 0.1481], device='cuda:1'), in_proj_covar=tensor([0.0072, 0.0098, 0.0099, 0.0109, 0.0082, 0.0108, 0.0098, 0.0103], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 18:27:21,330 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.1042, 6.2714, 5.6982, 5.9640, 5.8779, 5.4377, 5.7222, 5.3407], device='cuda:1'), covar=tensor([0.0983, 0.0856, 0.0972, 0.0841, 0.0845, 0.1368, 0.2015, 0.2386], device='cuda:1'), in_proj_covar=tensor([0.0500, 0.0584, 0.0440, 0.0442, 0.0417, 0.0456, 0.0591, 0.0515], 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') 2023-03-08 18:27:22,666 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.261e+02 2.009e+02 2.404e+02 3.052e+02 6.487e+02, threshold=4.807e+02, percent-clipped=6.0 2023-03-08 18:27:23,075 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.3582, 3.0630, 2.5887, 2.8420, 3.1650, 2.9645, 2.4864, 3.0275], device='cuda:1'), covar=tensor([0.0956, 0.0402, 0.0822, 0.0581, 0.0568, 0.0607, 0.0785, 0.0450], device='cuda:1'), in_proj_covar=tensor([0.0200, 0.0217, 0.0226, 0.0196, 0.0271, 0.0237, 0.0199, 0.0285], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 18:27:55,370 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.23 vs. limit=5.0 2023-03-08 18:27:59,507 INFO [train2.py:809] (1/4) Epoch 19, batch 350, loss[ctc_loss=0.09363, att_loss=0.2625, loss=0.2287, over 16471.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.006342, over 46.00 utterances.], tot_loss[ctc_loss=0.07784, att_loss=0.2366, loss=0.2048, over 2704923.16 frames. utt_duration=1239 frames, utt_pad_proportion=0.05775, over 8741.52 utterances.], batch size: 46, lr: 5.75e-03, grad_scale: 16.0 2023-03-08 18:28:09,218 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5406, 2.4959, 4.9670, 3.9611, 2.9262, 4.1701, 4.7866, 4.6714], device='cuda:1'), covar=tensor([0.0221, 0.1702, 0.0157, 0.0855, 0.1812, 0.0250, 0.0126, 0.0212], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0244, 0.0174, 0.0315, 0.0271, 0.0207, 0.0160, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 18:28:36,897 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.7804, 5.0980, 5.3584, 5.1107, 5.1866, 5.7197, 5.1118, 5.8437], device='cuda:1'), covar=tensor([0.0760, 0.0787, 0.0831, 0.1389, 0.1933, 0.0988, 0.0721, 0.0726], device='cuda:1'), in_proj_covar=tensor([0.0823, 0.0484, 0.0567, 0.0626, 0.0837, 0.0588, 0.0464, 0.0576], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 18:28:38,718 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6306, 2.9333, 5.0679, 4.0212, 3.0797, 4.2488, 4.8896, 4.6557], device='cuda:1'), covar=tensor([0.0266, 0.1470, 0.0184, 0.0973, 0.1729, 0.0244, 0.0141, 0.0246], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0244, 0.0174, 0.0315, 0.0271, 0.0207, 0.0160, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 18:29:01,416 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-03-08 18:29:17,683 INFO [train2.py:809] (1/4) Epoch 19, batch 400, loss[ctc_loss=0.09583, att_loss=0.259, loss=0.2264, over 17020.00 frames. utt_duration=1336 frames, utt_pad_proportion=0.008626, over 51.00 utterances.], tot_loss[ctc_loss=0.07867, att_loss=0.238, loss=0.2061, over 2831285.33 frames. utt_duration=1220 frames, utt_pad_proportion=0.06316, over 9295.32 utterances.], batch size: 51, lr: 5.75e-03, grad_scale: 8.0 2023-03-08 18:29:42,558 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1470, 4.4971, 4.4416, 4.5732, 2.6577, 4.4871, 2.2930, 1.6611], device='cuda:1'), covar=tensor([0.0350, 0.0221, 0.0631, 0.0216, 0.1794, 0.0167, 0.1739, 0.1850], device='cuda:1'), in_proj_covar=tensor([0.0181, 0.0154, 0.0257, 0.0149, 0.0224, 0.0130, 0.0232, 0.0206], 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') 2023-03-08 18:29:51,759 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72129.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 18:30:02,014 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.245e+02 2.032e+02 2.343e+02 2.887e+02 1.413e+03, threshold=4.687e+02, percent-clipped=3.0 2023-03-08 18:30:30,551 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.11 vs. limit=5.0 2023-03-08 18:30:37,549 INFO [train2.py:809] (1/4) Epoch 19, batch 450, loss[ctc_loss=0.07353, att_loss=0.2061, loss=0.1796, over 15768.00 frames. utt_duration=1661 frames, utt_pad_proportion=0.008856, over 38.00 utterances.], tot_loss[ctc_loss=0.07891, att_loss=0.2386, loss=0.2067, over 2932770.65 frames. utt_duration=1211 frames, utt_pad_proportion=0.06472, over 9699.91 utterances.], batch size: 38, lr: 5.74e-03, grad_scale: 8.0 2023-03-08 18:31:23,409 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-03-08 18:31:28,134 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=72190.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 18:31:56,583 INFO [train2.py:809] (1/4) Epoch 19, batch 500, loss[ctc_loss=0.09048, att_loss=0.2616, loss=0.2273, over 17053.00 frames. utt_duration=1313 frames, utt_pad_proportion=0.008032, over 52.00 utterances.], tot_loss[ctc_loss=0.07857, att_loss=0.2386, loss=0.2066, over 3006386.00 frames. utt_duration=1232 frames, utt_pad_proportion=0.05952, over 9774.08 utterances.], batch size: 52, lr: 5.74e-03, grad_scale: 8.0 2023-03-08 18:32:39,383 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.209e+02 2.024e+02 2.496e+02 3.106e+02 1.224e+03, threshold=4.991e+02, percent-clipped=5.0 2023-03-08 18:33:14,211 INFO [train2.py:809] (1/4) Epoch 19, batch 550, loss[ctc_loss=0.07882, att_loss=0.2312, loss=0.2007, over 16709.00 frames. utt_duration=676.5 frames, utt_pad_proportion=0.1501, over 99.00 utterances.], tot_loss[ctc_loss=0.07832, att_loss=0.2378, loss=0.2059, over 3064947.10 frames. utt_duration=1247 frames, utt_pad_proportion=0.05544, over 9841.53 utterances.], batch size: 99, lr: 5.74e-03, grad_scale: 8.0 2023-03-08 18:33:37,387 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72272.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 18:33:41,796 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=72275.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 18:33:52,533 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6292, 2.8996, 5.0202, 4.1305, 3.0863, 4.2280, 4.8145, 4.6015], device='cuda:1'), covar=tensor([0.0235, 0.1482, 0.0194, 0.0815, 0.1629, 0.0250, 0.0145, 0.0256], device='cuda:1'), in_proj_covar=tensor([0.0183, 0.0243, 0.0174, 0.0313, 0.0269, 0.0206, 0.0160, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 18:34:28,114 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-03-08 18:34:32,966 INFO [train2.py:809] (1/4) Epoch 19, batch 600, loss[ctc_loss=0.1037, att_loss=0.2607, loss=0.2293, over 17117.00 frames. utt_duration=1224 frames, utt_pad_proportion=0.01506, over 56.00 utterances.], tot_loss[ctc_loss=0.07812, att_loss=0.2375, loss=0.2056, over 3114204.07 frames. utt_duration=1265 frames, utt_pad_proportion=0.04896, over 9855.58 utterances.], batch size: 56, lr: 5.74e-03, grad_scale: 8.0 2023-03-08 18:34:52,095 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=72320.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 18:34:57,579 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1492, 4.8497, 4.9278, 4.9193, 4.7387, 4.9130, 4.6179, 4.4294], device='cuda:1'), covar=tensor([0.1998, 0.0826, 0.0417, 0.0639, 0.0940, 0.0474, 0.0515, 0.0464], device='cuda:1'), in_proj_covar=tensor([0.0509, 0.0341, 0.0322, 0.0338, 0.0402, 0.0414, 0.0345, 0.0379], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0003, 0.0004], device='cuda:1') 2023-03-08 18:35:02,565 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72326.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 18:35:18,135 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.518e+02 1.966e+02 2.354e+02 3.021e+02 6.403e+02, threshold=4.709e+02, percent-clipped=3.0 2023-03-08 18:35:23,874 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.8080, 3.5011, 3.6412, 3.1370, 3.5214, 3.7198, 3.6387, 2.8460], device='cuda:1'), covar=tensor([0.1070, 0.1609, 0.1789, 0.3616, 0.2028, 0.1801, 0.0816, 0.3442], device='cuda:1'), in_proj_covar=tensor([0.0154, 0.0173, 0.0186, 0.0247, 0.0147, 0.0241, 0.0163, 0.0210], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 18:35:53,227 INFO [train2.py:809] (1/4) Epoch 19, batch 650, loss[ctc_loss=0.06211, att_loss=0.2126, loss=0.1825, over 15496.00 frames. utt_duration=1723 frames, utt_pad_proportion=0.008977, over 36.00 utterances.], tot_loss[ctc_loss=0.07774, att_loss=0.2379, loss=0.2058, over 3162149.55 frames. utt_duration=1278 frames, utt_pad_proportion=0.04252, over 9906.74 utterances.], batch size: 36, lr: 5.74e-03, grad_scale: 8.0 2023-03-08 18:36:40,901 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=72387.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 18:37:01,849 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-03-08 18:37:13,714 INFO [train2.py:809] (1/4) Epoch 19, batch 700, loss[ctc_loss=0.0736, att_loss=0.2266, loss=0.196, over 16182.00 frames. utt_duration=1580 frames, utt_pad_proportion=0.005556, over 41.00 utterances.], tot_loss[ctc_loss=0.07776, att_loss=0.2374, loss=0.2055, over 3177332.61 frames. utt_duration=1275 frames, utt_pad_proportion=0.04704, over 9979.10 utterances.], batch size: 41, lr: 5.73e-03, grad_scale: 8.0 2023-03-08 18:37:25,364 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2337, 2.4887, 3.0039, 4.0850, 3.7554, 3.8426, 2.8617, 2.1634], device='cuda:1'), covar=tensor([0.0737, 0.2290, 0.1151, 0.0824, 0.0852, 0.0497, 0.1425, 0.2338], device='cuda:1'), in_proj_covar=tensor([0.0174, 0.0208, 0.0183, 0.0208, 0.0212, 0.0167, 0.0195, 0.0184], 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') 2023-03-08 18:37:57,758 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.096e+02 2.010e+02 2.298e+02 2.722e+02 6.112e+02, threshold=4.595e+02, percent-clipped=4.0 2023-03-08 18:38:32,588 INFO [train2.py:809] (1/4) Epoch 19, batch 750, loss[ctc_loss=0.07578, att_loss=0.2447, loss=0.2109, over 16480.00 frames. utt_duration=1435 frames, utt_pad_proportion=0.00571, over 46.00 utterances.], tot_loss[ctc_loss=0.07788, att_loss=0.2374, loss=0.2055, over 3186041.58 frames. utt_duration=1262 frames, utt_pad_proportion=0.05347, over 10111.25 utterances.], batch size: 46, lr: 5.73e-03, grad_scale: 8.0 2023-03-08 18:39:16,283 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=72485.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 18:39:46,963 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-03-08 18:39:51,882 INFO [train2.py:809] (1/4) Epoch 19, batch 800, loss[ctc_loss=0.08014, att_loss=0.2422, loss=0.2098, over 16779.00 frames. utt_duration=1400 frames, utt_pad_proportion=0.005874, over 48.00 utterances.], tot_loss[ctc_loss=0.07786, att_loss=0.2374, loss=0.2055, over 3206587.25 frames. utt_duration=1256 frames, utt_pad_proportion=0.05419, over 10223.95 utterances.], batch size: 48, lr: 5.73e-03, grad_scale: 8.0 2023-03-08 18:40:37,047 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.504e+02 2.045e+02 2.501e+02 3.023e+02 6.959e+02, threshold=5.001e+02, percent-clipped=5.0 2023-03-08 18:41:12,445 INFO [train2.py:809] (1/4) Epoch 19, batch 850, loss[ctc_loss=0.06301, att_loss=0.2024, loss=0.1745, over 15346.00 frames. utt_duration=1756 frames, utt_pad_proportion=0.01265, over 35.00 utterances.], tot_loss[ctc_loss=0.07807, att_loss=0.238, loss=0.206, over 3228250.58 frames. utt_duration=1254 frames, utt_pad_proportion=0.05184, over 10312.79 utterances.], batch size: 35, lr: 5.73e-03, grad_scale: 8.0 2023-03-08 18:41:39,242 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72575.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 18:41:41,603 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-03-08 18:42:10,169 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.28 vs. limit=5.0 2023-03-08 18:42:31,644 INFO [train2.py:809] (1/4) Epoch 19, batch 900, loss[ctc_loss=0.08024, att_loss=0.2437, loss=0.211, over 16885.00 frames. utt_duration=1380 frames, utt_pad_proportion=0.006597, over 49.00 utterances.], tot_loss[ctc_loss=0.07806, att_loss=0.2383, loss=0.2063, over 3240621.36 frames. utt_duration=1233 frames, utt_pad_proportion=0.05601, over 10529.68 utterances.], batch size: 49, lr: 5.73e-03, grad_scale: 8.0 2023-03-08 18:42:51,252 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.5521, 3.1562, 3.1766, 2.7866, 3.1939, 3.1890, 3.2086, 2.1810], device='cuda:1'), covar=tensor([0.1216, 0.2191, 0.2812, 0.4484, 0.1434, 0.3144, 0.1622, 0.5223], device='cuda:1'), in_proj_covar=tensor([0.0153, 0.0172, 0.0186, 0.0247, 0.0146, 0.0241, 0.0163, 0.0210], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 18:42:55,531 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=72623.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 18:43:16,190 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.377e+02 2.045e+02 2.502e+02 3.069e+02 5.173e+02, threshold=5.005e+02, percent-clipped=1.0 2023-03-08 18:43:51,341 INFO [train2.py:809] (1/4) Epoch 19, batch 950, loss[ctc_loss=0.05277, att_loss=0.2244, loss=0.1901, over 16397.00 frames. utt_duration=1492 frames, utt_pad_proportion=0.008236, over 44.00 utterances.], tot_loss[ctc_loss=0.07751, att_loss=0.2376, loss=0.2056, over 3251296.35 frames. utt_duration=1243 frames, utt_pad_proportion=0.05194, over 10474.14 utterances.], batch size: 44, lr: 5.72e-03, grad_scale: 8.0 2023-03-08 18:44:26,379 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-03-08 18:44:30,100 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=72682.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 18:45:11,228 INFO [train2.py:809] (1/4) Epoch 19, batch 1000, loss[ctc_loss=0.05811, att_loss=0.2345, loss=0.1992, over 16767.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.005775, over 48.00 utterances.], tot_loss[ctc_loss=0.07653, att_loss=0.2366, loss=0.2046, over 3249051.67 frames. utt_duration=1272 frames, utt_pad_proportion=0.04667, over 10231.18 utterances.], batch size: 48, lr: 5.72e-03, grad_scale: 8.0 2023-03-08 18:45:47,475 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.6230, 3.7538, 2.9680, 3.0924, 3.8370, 3.4380, 2.3225, 4.0535], device='cuda:1'), covar=tensor([0.1247, 0.0500, 0.1130, 0.0848, 0.0742, 0.0743, 0.1270, 0.0510], device='cuda:1'), in_proj_covar=tensor([0.0199, 0.0215, 0.0223, 0.0193, 0.0270, 0.0234, 0.0198, 0.0281], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-08 18:45:56,179 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.301e+02 2.016e+02 2.281e+02 2.809e+02 6.026e+02, threshold=4.562e+02, percent-clipped=5.0 2023-03-08 18:46:02,426 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-03-08 18:46:31,108 INFO [train2.py:809] (1/4) Epoch 19, batch 1050, loss[ctc_loss=0.07514, att_loss=0.2265, loss=0.1962, over 16015.00 frames. utt_duration=1603 frames, utt_pad_proportion=0.006939, over 40.00 utterances.], tot_loss[ctc_loss=0.07686, att_loss=0.237, loss=0.205, over 3257091.99 frames. utt_duration=1251 frames, utt_pad_proportion=0.05065, over 10424.34 utterances.], batch size: 40, lr: 5.72e-03, grad_scale: 8.0 2023-03-08 18:47:13,817 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72785.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 18:47:50,614 INFO [train2.py:809] (1/4) Epoch 19, batch 1100, loss[ctc_loss=0.06923, att_loss=0.2254, loss=0.1942, over 15941.00 frames. utt_duration=1556 frames, utt_pad_proportion=0.007975, over 41.00 utterances.], tot_loss[ctc_loss=0.07698, att_loss=0.2368, loss=0.2049, over 3253728.65 frames. utt_duration=1232 frames, utt_pad_proportion=0.05814, over 10576.17 utterances.], batch size: 41, lr: 5.72e-03, grad_scale: 8.0 2023-03-08 18:48:22,188 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-03-08 18:48:30,782 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=72833.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 18:48:35,938 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.339e+02 1.909e+02 2.229e+02 2.585e+02 5.838e+02, threshold=4.459e+02, percent-clipped=1.0 2023-03-08 18:49:10,574 INFO [train2.py:809] (1/4) Epoch 19, batch 1150, loss[ctc_loss=0.06339, att_loss=0.2221, loss=0.1904, over 16262.00 frames. utt_duration=1515 frames, utt_pad_proportion=0.008148, over 43.00 utterances.], tot_loss[ctc_loss=0.07683, att_loss=0.237, loss=0.205, over 3268157.77 frames. utt_duration=1250 frames, utt_pad_proportion=0.05108, over 10471.86 utterances.], batch size: 43, lr: 5.72e-03, grad_scale: 8.0 2023-03-08 18:49:43,971 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-03-08 18:50:08,402 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72894.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 18:50:30,472 INFO [train2.py:809] (1/4) Epoch 19, batch 1200, loss[ctc_loss=0.06943, att_loss=0.2353, loss=0.2022, over 16538.00 frames. utt_duration=1471 frames, utt_pad_proportion=0.006497, over 45.00 utterances.], tot_loss[ctc_loss=0.07706, att_loss=0.2375, loss=0.2054, over 3271767.67 frames. utt_duration=1246 frames, utt_pad_proportion=0.05253, over 10514.52 utterances.], batch size: 45, lr: 5.71e-03, grad_scale: 8.0 2023-03-08 18:50:36,582 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-03-08 18:50:53,163 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-03-08 18:51:15,279 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.411e+02 1.921e+02 2.301e+02 2.976e+02 5.020e+02, threshold=4.601e+02, percent-clipped=3.0 2023-03-08 18:51:45,240 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=72955.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 18:51:50,080 INFO [train2.py:809] (1/4) Epoch 19, batch 1250, loss[ctc_loss=0.05451, att_loss=0.2249, loss=0.1908, over 15962.00 frames. utt_duration=1559 frames, utt_pad_proportion=0.006607, over 41.00 utterances.], tot_loss[ctc_loss=0.07774, att_loss=0.2381, loss=0.206, over 3270062.08 frames. utt_duration=1209 frames, utt_pad_proportion=0.06301, over 10830.13 utterances.], batch size: 41, lr: 5.71e-03, grad_scale: 8.0 2023-03-08 18:51:59,969 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72964.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 18:52:27,721 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72982.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 18:52:56,023 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-03-08 18:53:08,944 INFO [train2.py:809] (1/4) Epoch 19, batch 1300, loss[ctc_loss=0.08913, att_loss=0.237, loss=0.2074, over 16120.00 frames. utt_duration=1537 frames, utt_pad_proportion=0.00648, over 42.00 utterances.], tot_loss[ctc_loss=0.07724, att_loss=0.2373, loss=0.2053, over 3273588.00 frames. utt_duration=1235 frames, utt_pad_proportion=0.05578, over 10611.79 utterances.], batch size: 42, lr: 5.71e-03, grad_scale: 8.0 2023-03-08 18:53:37,504 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73025.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 18:53:45,282 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=73030.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 18:53:55,192 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.296e+02 2.017e+02 2.532e+02 3.108e+02 5.795e+02, threshold=5.064e+02, percent-clipped=4.0 2023-03-08 18:54:30,469 INFO [train2.py:809] (1/4) Epoch 19, batch 1350, loss[ctc_loss=0.06345, att_loss=0.2351, loss=0.2008, over 16627.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.005289, over 47.00 utterances.], tot_loss[ctc_loss=0.07688, att_loss=0.2368, loss=0.2048, over 3270661.17 frames. utt_duration=1255 frames, utt_pad_proportion=0.05191, over 10434.23 utterances.], batch size: 47, lr: 5.71e-03, grad_scale: 8.0 2023-03-08 18:55:03,728 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-03-08 18:55:25,013 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2262, 3.7838, 3.7671, 3.2803, 3.8252, 3.9226, 3.8696, 2.9903], device='cuda:1'), covar=tensor([0.0794, 0.1500, 0.2404, 0.3445, 0.1398, 0.2974, 0.0945, 0.3514], device='cuda:1'), in_proj_covar=tensor([0.0154, 0.0170, 0.0185, 0.0245, 0.0147, 0.0243, 0.0163, 0.0208], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 18:55:31,850 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-03-08 18:55:50,789 INFO [train2.py:809] (1/4) Epoch 19, batch 1400, loss[ctc_loss=0.1403, att_loss=0.2716, loss=0.2454, over 13956.00 frames. utt_duration=383.7 frames, utt_pad_proportion=0.3304, over 146.00 utterances.], tot_loss[ctc_loss=0.07735, att_loss=0.2373, loss=0.2053, over 3266614.64 frames. utt_duration=1228 frames, utt_pad_proportion=0.0605, over 10656.48 utterances.], batch size: 146, lr: 5.71e-03, grad_scale: 8.0 2023-03-08 18:56:35,400 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.220e+02 2.120e+02 2.397e+02 2.926e+02 6.281e+02, threshold=4.794e+02, percent-clipped=3.0 2023-03-08 18:57:10,513 INFO [train2.py:809] (1/4) Epoch 19, batch 1450, loss[ctc_loss=0.06614, att_loss=0.2494, loss=0.2128, over 17386.00 frames. utt_duration=1009 frames, utt_pad_proportion=0.04789, over 69.00 utterances.], tot_loss[ctc_loss=0.07714, att_loss=0.2368, loss=0.2049, over 3265344.44 frames. utt_duration=1238 frames, utt_pad_proportion=0.05916, over 10565.87 utterances.], batch size: 69, lr: 5.70e-03, grad_scale: 8.0 2023-03-08 18:58:19,099 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-03-08 18:58:30,643 INFO [train2.py:809] (1/4) Epoch 19, batch 1500, loss[ctc_loss=0.06329, att_loss=0.2348, loss=0.2005, over 16264.00 frames. utt_duration=1514 frames, utt_pad_proportion=0.006923, over 43.00 utterances.], tot_loss[ctc_loss=0.07705, att_loss=0.237, loss=0.205, over 3269138.87 frames. utt_duration=1249 frames, utt_pad_proportion=0.0562, over 10483.66 utterances.], batch size: 43, lr: 5.70e-03, grad_scale: 8.0 2023-03-08 18:58:48,937 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.15 vs. limit=5.0 2023-03-08 18:59:15,228 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.382e+02 2.048e+02 2.505e+02 3.116e+02 1.439e+03, threshold=5.009e+02, percent-clipped=5.0 2023-03-08 18:59:36,944 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73250.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 18:59:49,905 INFO [train2.py:809] (1/4) Epoch 19, batch 1550, loss[ctc_loss=0.07504, att_loss=0.2489, loss=0.2142, over 17036.00 frames. utt_duration=1338 frames, utt_pad_proportion=0.00768, over 51.00 utterances.], tot_loss[ctc_loss=0.077, att_loss=0.2363, loss=0.2044, over 3259338.26 frames. utt_duration=1263 frames, utt_pad_proportion=0.05588, over 10334.74 utterances.], batch size: 51, lr: 5.70e-03, grad_scale: 8.0 2023-03-08 19:00:45,899 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=73293.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:00:48,920 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.0120, 3.7118, 3.6253, 3.2905, 3.7503, 3.8770, 3.8267, 3.0060], device='cuda:1'), covar=tensor([0.0945, 0.1483, 0.2055, 0.3275, 0.1054, 0.1461, 0.0880, 0.3435], device='cuda:1'), in_proj_covar=tensor([0.0152, 0.0168, 0.0183, 0.0241, 0.0145, 0.0240, 0.0161, 0.0206], 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') 2023-03-08 19:01:03,269 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4379, 4.5861, 4.5761, 4.5652, 4.6667, 4.6291, 4.3738, 4.1995], device='cuda:1'), covar=tensor([0.0965, 0.0679, 0.0487, 0.0526, 0.0319, 0.0342, 0.0410, 0.0360], device='cuda:1'), in_proj_covar=tensor([0.0513, 0.0346, 0.0330, 0.0346, 0.0404, 0.0416, 0.0345, 0.0381], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0003, 0.0004], device='cuda:1') 2023-03-08 19:01:09,924 INFO [train2.py:809] (1/4) Epoch 19, batch 1600, loss[ctc_loss=0.08076, att_loss=0.2505, loss=0.2165, over 16862.00 frames. utt_duration=1378 frames, utt_pad_proportion=0.008545, over 49.00 utterances.], tot_loss[ctc_loss=0.07742, att_loss=0.2368, loss=0.2049, over 3265422.42 frames. utt_duration=1253 frames, utt_pad_proportion=0.0566, over 10434.63 utterances.], batch size: 49, lr: 5.70e-03, grad_scale: 8.0 2023-03-08 19:01:28,648 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73320.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 19:01:54,274 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.313e+02 2.054e+02 2.356e+02 2.874e+02 5.257e+02, threshold=4.713e+02, percent-clipped=1.0 2023-03-08 19:02:23,066 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73354.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:02:29,413 INFO [train2.py:809] (1/4) Epoch 19, batch 1650, loss[ctc_loss=0.07478, att_loss=0.2288, loss=0.198, over 16177.00 frames. utt_duration=1580 frames, utt_pad_proportion=0.007129, over 41.00 utterances.], tot_loss[ctc_loss=0.07686, att_loss=0.2364, loss=0.2045, over 3259027.27 frames. utt_duration=1253 frames, utt_pad_proportion=0.05747, over 10415.65 utterances.], batch size: 41, lr: 5.70e-03, grad_scale: 8.0 2023-03-08 19:02:58,061 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0547, 5.0906, 4.9573, 2.2596, 1.8825, 3.0705, 2.5710, 3.7986], device='cuda:1'), covar=tensor([0.0711, 0.0282, 0.0263, 0.4994, 0.6208, 0.2331, 0.3440, 0.1751], device='cuda:1'), in_proj_covar=tensor([0.0356, 0.0269, 0.0267, 0.0245, 0.0349, 0.0339, 0.0254, 0.0366], device='cuda:1'), out_proj_covar=tensor([1.5140e-04, 9.9781e-05, 1.1352e-04, 1.0614e-04, 1.4679e-04, 1.3308e-04, 1.0201e-04, 1.4969e-04], device='cuda:1') 2023-03-08 19:03:48,206 INFO [train2.py:809] (1/4) Epoch 19, batch 1700, loss[ctc_loss=0.06079, att_loss=0.2196, loss=0.1878, over 15898.00 frames. utt_duration=1632 frames, utt_pad_proportion=0.008537, over 39.00 utterances.], tot_loss[ctc_loss=0.07742, att_loss=0.2373, loss=0.2053, over 3265648.11 frames. utt_duration=1254 frames, utt_pad_proportion=0.05454, over 10425.39 utterances.], batch size: 39, lr: 5.69e-03, grad_scale: 8.0 2023-03-08 19:04:32,147 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.279e+02 2.058e+02 2.463e+02 2.943e+02 5.957e+02, threshold=4.927e+02, percent-clipped=3.0 2023-03-08 19:04:59,757 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-03-08 19:05:06,959 INFO [train2.py:809] (1/4) Epoch 19, batch 1750, loss[ctc_loss=0.08718, att_loss=0.2587, loss=0.2244, over 17298.00 frames. utt_duration=1174 frames, utt_pad_proportion=0.02377, over 59.00 utterances.], tot_loss[ctc_loss=0.07728, att_loss=0.2368, loss=0.2049, over 3260296.02 frames. utt_duration=1270 frames, utt_pad_proportion=0.05325, over 10280.35 utterances.], batch size: 59, lr: 5.69e-03, grad_scale: 8.0 2023-03-08 19:05:36,135 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0546, 5.1361, 4.8613, 2.7297, 4.8782, 4.7025, 4.0737, 2.6161], device='cuda:1'), covar=tensor([0.0123, 0.0078, 0.0235, 0.1032, 0.0091, 0.0179, 0.0369, 0.1455], device='cuda:1'), in_proj_covar=tensor([0.0072, 0.0099, 0.0099, 0.0108, 0.0082, 0.0109, 0.0097, 0.0103], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 19:05:46,918 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2374, 4.5889, 4.6738, 4.8168, 2.7859, 4.7127, 2.9031, 1.6604], device='cuda:1'), covar=tensor([0.0414, 0.0272, 0.0586, 0.0344, 0.1629, 0.0172, 0.1341, 0.1856], device='cuda:1'), in_proj_covar=tensor([0.0178, 0.0150, 0.0250, 0.0147, 0.0216, 0.0129, 0.0225, 0.0199], 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') 2023-03-08 19:06:22,416 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-03-08 19:06:25,916 INFO [train2.py:809] (1/4) Epoch 19, batch 1800, loss[ctc_loss=0.06446, att_loss=0.2358, loss=0.2015, over 16544.00 frames. utt_duration=1472 frames, utt_pad_proportion=0.005405, over 45.00 utterances.], tot_loss[ctc_loss=0.07803, att_loss=0.2381, loss=0.2061, over 3274289.41 frames. utt_duration=1249 frames, utt_pad_proportion=0.05432, over 10502.27 utterances.], batch size: 45, lr: 5.69e-03, grad_scale: 8.0 2023-03-08 19:07:10,632 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.143e+02 1.911e+02 2.242e+02 2.883e+02 9.834e+02, threshold=4.485e+02, percent-clipped=3.0 2023-03-08 19:07:33,098 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=73550.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:07:45,877 INFO [train2.py:809] (1/4) Epoch 19, batch 1850, loss[ctc_loss=0.07116, att_loss=0.2371, loss=0.2039, over 16951.00 frames. utt_duration=1358 frames, utt_pad_proportion=0.007544, over 50.00 utterances.], tot_loss[ctc_loss=0.07796, att_loss=0.2384, loss=0.2063, over 3276255.72 frames. utt_duration=1242 frames, utt_pad_proportion=0.05571, over 10561.75 utterances.], batch size: 50, lr: 5.69e-03, grad_scale: 8.0 2023-03-08 19:08:36,607 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=73590.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:08:49,384 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=73598.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:08:59,056 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5596, 4.9860, 4.9344, 4.9453, 5.0035, 4.6569, 3.2648, 4.8542], device='cuda:1'), covar=tensor([0.0121, 0.0130, 0.0124, 0.0094, 0.0113, 0.0136, 0.0886, 0.0256], device='cuda:1'), in_proj_covar=tensor([0.0089, 0.0086, 0.0108, 0.0066, 0.0072, 0.0084, 0.0103, 0.0107], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 19:09:04,980 INFO [train2.py:809] (1/4) Epoch 19, batch 1900, loss[ctc_loss=0.07594, att_loss=0.2469, loss=0.2127, over 16466.00 frames. utt_duration=1433 frames, utt_pad_proportion=0.006644, over 46.00 utterances.], tot_loss[ctc_loss=0.07835, att_loss=0.2392, loss=0.207, over 3268106.90 frames. utt_duration=1222 frames, utt_pad_proportion=0.06307, over 10711.02 utterances.], batch size: 46, lr: 5.69e-03, grad_scale: 8.0 2023-03-08 19:09:17,809 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7531, 4.7925, 4.4952, 2.3387, 4.5921, 4.4644, 3.9076, 2.3644], device='cuda:1'), covar=tensor([0.0140, 0.0128, 0.0305, 0.1296, 0.0111, 0.0261, 0.0408, 0.1700], device='cuda:1'), in_proj_covar=tensor([0.0073, 0.0099, 0.0099, 0.0109, 0.0082, 0.0109, 0.0097, 0.0103], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 19:09:23,765 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=73620.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 19:09:48,775 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.371e+02 1.956e+02 2.391e+02 2.954e+02 5.478e+02, threshold=4.781e+02, percent-clipped=4.0 2023-03-08 19:10:10,625 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73649.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:10:13,957 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73651.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:10:24,370 INFO [train2.py:809] (1/4) Epoch 19, batch 1950, loss[ctc_loss=0.07517, att_loss=0.2404, loss=0.2073, over 16883.00 frames. utt_duration=1379 frames, utt_pad_proportion=0.006876, over 49.00 utterances.], tot_loss[ctc_loss=0.07818, att_loss=0.2392, loss=0.207, over 3271966.76 frames. utt_duration=1209 frames, utt_pad_proportion=0.06504, over 10837.16 utterances.], batch size: 49, lr: 5.68e-03, grad_scale: 8.0 2023-03-08 19:10:40,467 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=73668.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 19:10:41,665 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-03-08 19:11:18,273 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9787, 5.2673, 5.4924, 5.3108, 5.3953, 5.9186, 5.2002, 6.0316], device='cuda:1'), covar=tensor([0.0711, 0.0784, 0.0851, 0.1331, 0.1857, 0.0866, 0.0635, 0.0681], device='cuda:1'), in_proj_covar=tensor([0.0850, 0.0494, 0.0580, 0.0640, 0.0855, 0.0604, 0.0479, 0.0590], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 19:11:25,316 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.80 vs. limit=2.0 2023-03-08 19:11:33,176 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.5087, 2.9958, 3.4895, 4.5459, 4.0387, 4.0178, 3.0247, 2.1886], device='cuda:1'), covar=tensor([0.0673, 0.1998, 0.0897, 0.0585, 0.0792, 0.0445, 0.1509, 0.2360], device='cuda:1'), in_proj_covar=tensor([0.0178, 0.0215, 0.0188, 0.0214, 0.0217, 0.0174, 0.0200, 0.0187], 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') 2023-03-08 19:11:43,432 INFO [train2.py:809] (1/4) Epoch 19, batch 2000, loss[ctc_loss=0.075, att_loss=0.2149, loss=0.1869, over 13163.00 frames. utt_duration=1817 frames, utt_pad_proportion=0.03964, over 29.00 utterances.], tot_loss[ctc_loss=0.07826, att_loss=0.2389, loss=0.2068, over 3270887.83 frames. utt_duration=1223 frames, utt_pad_proportion=0.06112, over 10714.18 utterances.], batch size: 29, lr: 5.68e-03, grad_scale: 8.0 2023-03-08 19:12:22,476 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=73732.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:12:28,297 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.224e+02 2.150e+02 2.565e+02 3.312e+02 1.867e+03, threshold=5.130e+02, percent-clipped=8.0 2023-03-08 19:13:03,765 INFO [train2.py:809] (1/4) Epoch 19, batch 2050, loss[ctc_loss=0.05274, att_loss=0.2319, loss=0.1961, over 16466.00 frames. utt_duration=1433 frames, utt_pad_proportion=0.006614, over 46.00 utterances.], tot_loss[ctc_loss=0.07759, att_loss=0.2381, loss=0.206, over 3272091.46 frames. utt_duration=1240 frames, utt_pad_proportion=0.05655, over 10570.65 utterances.], batch size: 46, lr: 5.68e-03, grad_scale: 8.0 2023-03-08 19:13:05,613 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0095, 4.9719, 4.5742, 2.4996, 4.7397, 4.5404, 4.1069, 2.2606], device='cuda:1'), covar=tensor([0.0133, 0.0114, 0.0386, 0.1347, 0.0109, 0.0264, 0.0399, 0.1802], device='cuda:1'), in_proj_covar=tensor([0.0073, 0.0099, 0.0100, 0.0110, 0.0083, 0.0110, 0.0098, 0.0104], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 19:13:59,673 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73793.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:14:24,524 INFO [train2.py:809] (1/4) Epoch 19, batch 2100, loss[ctc_loss=0.04613, att_loss=0.2182, loss=0.1838, over 15967.00 frames. utt_duration=1559 frames, utt_pad_proportion=0.00642, over 41.00 utterances.], tot_loss[ctc_loss=0.07772, att_loss=0.2376, loss=0.2056, over 3264973.47 frames. utt_duration=1251 frames, utt_pad_proportion=0.0558, over 10451.74 utterances.], batch size: 41, lr: 5.68e-03, grad_scale: 8.0 2023-03-08 19:14:52,252 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7397, 3.4274, 3.5537, 2.8382, 3.5514, 3.5944, 3.6184, 2.1650], device='cuda:1'), covar=tensor([0.1234, 0.1756, 0.2671, 0.6086, 0.1417, 0.2631, 0.0993, 0.7636], device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0175, 0.0186, 0.0249, 0.0150, 0.0247, 0.0164, 0.0211], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 19:15:08,599 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.215e+02 2.020e+02 2.356e+02 2.940e+02 5.527e+02, threshold=4.712e+02, percent-clipped=4.0 2023-03-08 19:15:13,354 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.8539, 5.1067, 5.4156, 5.2698, 5.3293, 5.7713, 5.1239, 5.8934], device='cuda:1'), covar=tensor([0.0666, 0.0726, 0.0716, 0.1121, 0.1615, 0.0859, 0.0751, 0.0679], device='cuda:1'), in_proj_covar=tensor([0.0843, 0.0495, 0.0577, 0.0640, 0.0852, 0.0600, 0.0477, 0.0586], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 19:15:16,365 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2859, 2.7044, 2.9477, 4.2897, 3.8991, 3.9072, 2.8383, 1.9474], device='cuda:1'), covar=tensor([0.0749, 0.2106, 0.1179, 0.0547, 0.0864, 0.0416, 0.1470, 0.2436], device='cuda:1'), in_proj_covar=tensor([0.0176, 0.0213, 0.0187, 0.0212, 0.0216, 0.0174, 0.0198, 0.0185], 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') 2023-03-08 19:15:43,847 INFO [train2.py:809] (1/4) Epoch 19, batch 2150, loss[ctc_loss=0.04883, att_loss=0.2321, loss=0.1954, over 16899.00 frames. utt_duration=1381 frames, utt_pad_proportion=0.005789, over 49.00 utterances.], tot_loss[ctc_loss=0.07775, att_loss=0.2379, loss=0.2059, over 3271490.35 frames. utt_duration=1238 frames, utt_pad_proportion=0.05719, over 10580.42 utterances.], batch size: 49, lr: 5.68e-03, grad_scale: 8.0 2023-03-08 19:15:54,605 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2023-03-08 19:16:14,947 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.06 vs. limit=5.0 2023-03-08 19:17:03,555 INFO [train2.py:809] (1/4) Epoch 19, batch 2200, loss[ctc_loss=0.05828, att_loss=0.2053, loss=0.1759, over 15496.00 frames. utt_duration=1723 frames, utt_pad_proportion=0.009217, over 36.00 utterances.], tot_loss[ctc_loss=0.07724, att_loss=0.237, loss=0.2051, over 3268364.19 frames. utt_duration=1277 frames, utt_pad_proportion=0.04895, over 10249.11 utterances.], batch size: 36, lr: 5.68e-03, grad_scale: 8.0 2023-03-08 19:17:15,620 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-03-08 19:17:47,494 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.450e+02 2.057e+02 2.535e+02 3.122e+02 7.194e+02, threshold=5.070e+02, percent-clipped=8.0 2023-03-08 19:18:04,657 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73946.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:18:09,413 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=73949.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:18:23,575 INFO [train2.py:809] (1/4) Epoch 19, batch 2250, loss[ctc_loss=0.07924, att_loss=0.2418, loss=0.2093, over 16678.00 frames. utt_duration=1452 frames, utt_pad_proportion=0.006398, over 46.00 utterances.], tot_loss[ctc_loss=0.07713, att_loss=0.2377, loss=0.2056, over 3282949.27 frames. utt_duration=1275 frames, utt_pad_proportion=0.04611, over 10310.42 utterances.], batch size: 46, lr: 5.67e-03, grad_scale: 8.0 2023-03-08 19:18:25,995 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.70 vs. limit=2.0 2023-03-08 19:19:04,332 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0371, 5.0779, 4.5857, 2.6772, 4.8117, 4.5954, 4.2094, 2.3743], device='cuda:1'), covar=tensor([0.0204, 0.0125, 0.0390, 0.1446, 0.0122, 0.0265, 0.0467, 0.2451], device='cuda:1'), in_proj_covar=tensor([0.0074, 0.0100, 0.0101, 0.0111, 0.0083, 0.0111, 0.0099, 0.0104], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 19:19:25,457 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=73997.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:19:47,871 INFO [train2.py:809] (1/4) Epoch 19, batch 2300, loss[ctc_loss=0.072, att_loss=0.2459, loss=0.2112, over 16537.00 frames. utt_duration=1472 frames, utt_pad_proportion=0.006317, over 45.00 utterances.], tot_loss[ctc_loss=0.07734, att_loss=0.2384, loss=0.2062, over 3284947.45 frames. utt_duration=1274 frames, utt_pad_proportion=0.04469, over 10322.93 utterances.], batch size: 45, lr: 5.67e-03, grad_scale: 8.0 2023-03-08 19:20:30,630 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.391e+02 2.198e+02 2.569e+02 3.135e+02 6.426e+02, threshold=5.138e+02, percent-clipped=5.0 2023-03-08 19:21:05,902 INFO [train2.py:809] (1/4) Epoch 19, batch 2350, loss[ctc_loss=0.1107, att_loss=0.2614, loss=0.2313, over 16721.00 frames. utt_duration=677.3 frames, utt_pad_proportion=0.147, over 99.00 utterances.], tot_loss[ctc_loss=0.07841, att_loss=0.2386, loss=0.2066, over 3284194.65 frames. utt_duration=1267 frames, utt_pad_proportion=0.04739, over 10382.78 utterances.], batch size: 99, lr: 5.67e-03, grad_scale: 8.0 2023-03-08 19:21:15,256 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74064.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:21:52,207 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74088.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:22:02,554 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.7110, 4.0754, 3.2029, 4.2703, 3.8967, 3.7784, 3.9358, 3.9534], device='cuda:1'), covar=tensor([0.0752, 0.0450, 0.1608, 0.0378, 0.0393, 0.1014, 0.0679, 0.0428], device='cuda:1'), in_proj_covar=tensor([0.0371, 0.0304, 0.0353, 0.0326, 0.0308, 0.0229, 0.0292, 0.0274], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 19:22:24,341 INFO [train2.py:809] (1/4) Epoch 19, batch 2400, loss[ctc_loss=0.06364, att_loss=0.2298, loss=0.1966, over 16523.00 frames. utt_duration=1470 frames, utt_pad_proportion=0.006667, over 45.00 utterances.], tot_loss[ctc_loss=0.07878, att_loss=0.2385, loss=0.2065, over 3273926.99 frames. utt_duration=1267 frames, utt_pad_proportion=0.04922, over 10344.65 utterances.], batch size: 45, lr: 5.67e-03, grad_scale: 16.0 2023-03-08 19:22:51,342 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74125.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:22:56,482 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=2.92 vs. limit=5.0 2023-03-08 19:23:07,414 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.383e+02 2.133e+02 2.520e+02 3.216e+02 6.079e+02, threshold=5.039e+02, percent-clipped=1.0 2023-03-08 19:23:43,618 INFO [train2.py:809] (1/4) Epoch 19, batch 2450, loss[ctc_loss=0.08968, att_loss=0.2536, loss=0.2208, over 17325.00 frames. utt_duration=1176 frames, utt_pad_proportion=0.0224, over 59.00 utterances.], tot_loss[ctc_loss=0.07849, att_loss=0.2382, loss=0.2062, over 3275962.03 frames. utt_duration=1275 frames, utt_pad_proportion=0.04761, over 10293.02 utterances.], batch size: 59, lr: 5.67e-03, grad_scale: 16.0 2023-03-08 19:25:02,844 INFO [train2.py:809] (1/4) Epoch 19, batch 2500, loss[ctc_loss=0.08178, att_loss=0.2527, loss=0.2185, over 16877.00 frames. utt_duration=1379 frames, utt_pad_proportion=0.006994, over 49.00 utterances.], tot_loss[ctc_loss=0.07731, att_loss=0.2371, loss=0.2051, over 3275089.35 frames. utt_duration=1294 frames, utt_pad_proportion=0.04385, over 10139.09 utterances.], batch size: 49, lr: 5.66e-03, grad_scale: 16.0 2023-03-08 19:25:47,410 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.253e+02 2.096e+02 2.486e+02 2.922e+02 5.921e+02, threshold=4.971e+02, percent-clipped=2.0 2023-03-08 19:25:54,834 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74240.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:26:04,002 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74246.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:26:14,136 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1738, 4.5157, 4.3958, 4.7241, 2.7995, 4.5384, 2.5026, 1.8589], device='cuda:1'), covar=tensor([0.0345, 0.0220, 0.0693, 0.0182, 0.1550, 0.0162, 0.1643, 0.1661], device='cuda:1'), in_proj_covar=tensor([0.0180, 0.0151, 0.0254, 0.0148, 0.0217, 0.0131, 0.0228, 0.0201], 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') 2023-03-08 19:26:23,588 INFO [train2.py:809] (1/4) Epoch 19, batch 2550, loss[ctc_loss=0.07709, att_loss=0.2219, loss=0.1929, over 15377.00 frames. utt_duration=1759 frames, utt_pad_proportion=0.01137, over 35.00 utterances.], tot_loss[ctc_loss=0.0775, att_loss=0.2374, loss=0.2055, over 3274271.72 frames. utt_duration=1265 frames, utt_pad_proportion=0.0509, over 10363.00 utterances.], batch size: 35, lr: 5.66e-03, grad_scale: 16.0 2023-03-08 19:26:49,331 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-03-08 19:26:51,392 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0422, 5.2656, 5.5747, 5.4213, 5.5043, 5.9909, 5.2724, 6.1082], device='cuda:1'), covar=tensor([0.0678, 0.0714, 0.0805, 0.1204, 0.1812, 0.0920, 0.0594, 0.0654], device='cuda:1'), in_proj_covar=tensor([0.0834, 0.0488, 0.0575, 0.0637, 0.0846, 0.0598, 0.0473, 0.0583], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 19:27:05,461 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-03-08 19:27:20,418 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=74294.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:27:32,405 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74301.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:27:43,385 INFO [train2.py:809] (1/4) Epoch 19, batch 2600, loss[ctc_loss=0.0639, att_loss=0.2189, loss=0.1879, over 16026.00 frames. utt_duration=1604 frames, utt_pad_proportion=0.006087, over 40.00 utterances.], tot_loss[ctc_loss=0.07768, att_loss=0.2375, loss=0.2055, over 3275074.24 frames. utt_duration=1250 frames, utt_pad_proportion=0.05365, over 10488.64 utterances.], batch size: 40, lr: 5.66e-03, grad_scale: 8.0 2023-03-08 19:28:29,339 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.314e+02 1.890e+02 2.336e+02 2.791e+02 4.929e+02, threshold=4.673e+02, percent-clipped=0.0 2023-03-08 19:28:38,004 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9743, 5.2822, 5.0882, 5.2041, 5.2700, 5.2337, 4.9579, 4.7359], device='cuda:1'), covar=tensor([0.1095, 0.0530, 0.0353, 0.0452, 0.0324, 0.0357, 0.0375, 0.0354], device='cuda:1'), in_proj_covar=tensor([0.0516, 0.0348, 0.0333, 0.0344, 0.0407, 0.0418, 0.0347, 0.0384], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0003, 0.0004], device='cuda:1') 2023-03-08 19:29:04,097 INFO [train2.py:809] (1/4) Epoch 19, batch 2650, loss[ctc_loss=0.04228, att_loss=0.2072, loss=0.1742, over 14537.00 frames. utt_duration=1819 frames, utt_pad_proportion=0.04203, over 32.00 utterances.], tot_loss[ctc_loss=0.07821, att_loss=0.2381, loss=0.2061, over 3274733.57 frames. utt_duration=1219 frames, utt_pad_proportion=0.06112, over 10756.16 utterances.], batch size: 32, lr: 5.66e-03, grad_scale: 8.0 2023-03-08 19:29:40,380 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1145, 4.4598, 4.6464, 4.7948, 2.7953, 4.5064, 2.6763, 1.8776], device='cuda:1'), covar=tensor([0.0363, 0.0249, 0.0563, 0.0172, 0.1494, 0.0172, 0.1487, 0.1573], device='cuda:1'), in_proj_covar=tensor([0.0180, 0.0152, 0.0254, 0.0148, 0.0217, 0.0132, 0.0229, 0.0200], 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') 2023-03-08 19:29:52,051 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74388.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:30:23,754 INFO [train2.py:809] (1/4) Epoch 19, batch 2700, loss[ctc_loss=0.08469, att_loss=0.2522, loss=0.2187, over 17259.00 frames. utt_duration=1171 frames, utt_pad_proportion=0.02707, over 59.00 utterances.], tot_loss[ctc_loss=0.07818, att_loss=0.238, loss=0.206, over 3271284.58 frames. utt_duration=1223 frames, utt_pad_proportion=0.05947, over 10713.21 utterances.], batch size: 59, lr: 5.66e-03, grad_scale: 8.0 2023-03-08 19:30:42,771 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74420.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:31:02,287 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1078, 4.5385, 4.6200, 4.7508, 2.7942, 4.4876, 2.5437, 1.9784], device='cuda:1'), covar=tensor([0.0377, 0.0257, 0.0580, 0.0397, 0.1528, 0.0168, 0.1547, 0.1606], device='cuda:1'), in_proj_covar=tensor([0.0180, 0.0151, 0.0252, 0.0147, 0.0215, 0.0132, 0.0228, 0.0199], 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') 2023-03-08 19:31:03,573 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0849, 6.2829, 5.6989, 5.9837, 5.9778, 5.4601, 5.7176, 5.3937], device='cuda:1'), covar=tensor([0.1205, 0.0858, 0.0872, 0.0813, 0.0927, 0.1430, 0.2213, 0.2573], device='cuda:1'), in_proj_covar=tensor([0.0508, 0.0593, 0.0448, 0.0440, 0.0418, 0.0457, 0.0604, 0.0515], 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') 2023-03-08 19:31:09,478 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=74436.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:31:10,817 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.294e+02 2.123e+02 2.465e+02 2.927e+02 6.214e+02, threshold=4.930e+02, percent-clipped=3.0 2023-03-08 19:31:46,099 INFO [train2.py:809] (1/4) Epoch 19, batch 2750, loss[ctc_loss=0.06804, att_loss=0.2256, loss=0.1941, over 16116.00 frames. utt_duration=1537 frames, utt_pad_proportion=0.006757, over 42.00 utterances.], tot_loss[ctc_loss=0.07817, att_loss=0.238, loss=0.2061, over 3275250.69 frames. utt_duration=1220 frames, utt_pad_proportion=0.0594, over 10749.88 utterances.], batch size: 42, lr: 5.65e-03, grad_scale: 8.0 2023-03-08 19:32:34,540 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6483, 4.8491, 4.3060, 4.7329, 4.5070, 4.1143, 4.3725, 4.1713], device='cuda:1'), covar=tensor([0.1294, 0.1289, 0.1125, 0.0928, 0.1107, 0.1688, 0.2476, 0.2562], device='cuda:1'), in_proj_covar=tensor([0.0504, 0.0588, 0.0444, 0.0436, 0.0415, 0.0453, 0.0600, 0.0511], 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') 2023-03-08 19:33:07,059 INFO [train2.py:809] (1/4) Epoch 19, batch 2800, loss[ctc_loss=0.0588, att_loss=0.2317, loss=0.1972, over 16858.00 frames. utt_duration=682.7 frames, utt_pad_proportion=0.1423, over 99.00 utterances.], tot_loss[ctc_loss=0.07765, att_loss=0.2374, loss=0.2055, over 3270123.67 frames. utt_duration=1232 frames, utt_pad_proportion=0.05861, over 10632.69 utterances.], batch size: 99, lr: 5.65e-03, grad_scale: 8.0 2023-03-08 19:33:53,762 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.261e+02 2.053e+02 2.431e+02 2.775e+02 7.062e+02, threshold=4.862e+02, percent-clipped=1.0 2023-03-08 19:34:27,617 INFO [train2.py:809] (1/4) Epoch 19, batch 2850, loss[ctc_loss=0.1306, att_loss=0.2676, loss=0.2402, over 14365.00 frames. utt_duration=397.9 frames, utt_pad_proportion=0.308, over 145.00 utterances.], tot_loss[ctc_loss=0.07848, att_loss=0.2374, loss=0.2056, over 3263013.51 frames. utt_duration=1228 frames, utt_pad_proportion=0.06196, over 10638.91 utterances.], batch size: 145, lr: 5.65e-03, grad_scale: 8.0 2023-03-08 19:35:27,830 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74595.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:35:29,192 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74596.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:35:48,360 INFO [train2.py:809] (1/4) Epoch 19, batch 2900, loss[ctc_loss=0.09001, att_loss=0.2561, loss=0.2229, over 16762.00 frames. utt_duration=1398 frames, utt_pad_proportion=0.006762, over 48.00 utterances.], tot_loss[ctc_loss=0.07799, att_loss=0.2376, loss=0.2057, over 3267739.10 frames. utt_duration=1240 frames, utt_pad_proportion=0.0595, over 10556.17 utterances.], batch size: 48, lr: 5.65e-03, grad_scale: 8.0 2023-03-08 19:36:20,453 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-03-08 19:36:34,196 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.339e+02 1.992e+02 2.354e+02 2.922e+02 1.423e+03, threshold=4.708e+02, percent-clipped=1.0 2023-03-08 19:37:06,216 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74656.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:37:08,929 INFO [train2.py:809] (1/4) Epoch 19, batch 2950, loss[ctc_loss=0.1439, att_loss=0.2785, loss=0.2516, over 13987.00 frames. utt_duration=384.6 frames, utt_pad_proportion=0.3276, over 146.00 utterances.], tot_loss[ctc_loss=0.07793, att_loss=0.2379, loss=0.2059, over 3261737.26 frames. utt_duration=1221 frames, utt_pad_proportion=0.06539, over 10695.98 utterances.], batch size: 146, lr: 5.65e-03, grad_scale: 8.0 2023-03-08 19:37:38,523 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.9401, 2.5013, 2.8397, 3.9735, 3.5872, 3.6478, 2.7648, 1.9438], device='cuda:1'), covar=tensor([0.0824, 0.2132, 0.1190, 0.0566, 0.0832, 0.0485, 0.1360, 0.2374], device='cuda:1'), in_proj_covar=tensor([0.0178, 0.0213, 0.0188, 0.0214, 0.0220, 0.0175, 0.0199, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 19:37:52,006 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7286, 3.5869, 3.6010, 3.1038, 3.6541, 3.5230, 3.5464, 2.5174], device='cuda:1'), covar=tensor([0.1053, 0.1126, 0.1948, 0.3290, 0.0939, 0.3456, 0.0942, 0.4212], device='cuda:1'), in_proj_covar=tensor([0.0157, 0.0172, 0.0187, 0.0246, 0.0149, 0.0246, 0.0166, 0.0209], device='cuda:1'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 19:38:17,649 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-03-08 19:38:28,662 INFO [train2.py:809] (1/4) Epoch 19, batch 3000, loss[ctc_loss=0.06612, att_loss=0.2285, loss=0.196, over 16123.00 frames. utt_duration=1537 frames, utt_pad_proportion=0.006264, over 42.00 utterances.], tot_loss[ctc_loss=0.07902, att_loss=0.2391, loss=0.2071, over 3268386.42 frames. utt_duration=1197 frames, utt_pad_proportion=0.07029, over 10934.90 utterances.], batch size: 42, lr: 5.64e-03, grad_scale: 8.0 2023-03-08 19:38:28,662 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-08 19:38:42,984 INFO [train2.py:843] (1/4) Epoch 19, validation: ctc_loss=0.04253, att_loss=0.235, loss=0.1965, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-08 19:38:42,984 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-08 19:38:56,085 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4463, 4.4677, 4.5407, 4.4428, 5.1074, 4.4182, 4.5066, 2.3948], device='cuda:1'), covar=tensor([0.0240, 0.0314, 0.0281, 0.0270, 0.1107, 0.0247, 0.0282, 0.2029], device='cuda:1'), in_proj_covar=tensor([0.0153, 0.0175, 0.0179, 0.0195, 0.0366, 0.0150, 0.0166, 0.0218], 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') 2023-03-08 19:39:02,193 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74720.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:39:29,768 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.374e+02 2.054e+02 2.397e+02 2.986e+02 9.234e+02, threshold=4.795e+02, percent-clipped=4.0 2023-03-08 19:39:58,921 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2637, 4.4880, 4.5338, 4.7995, 2.6704, 4.4207, 2.6910, 2.0041], device='cuda:1'), covar=tensor([0.0401, 0.0257, 0.0714, 0.0179, 0.1887, 0.0192, 0.1730, 0.1781], device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0152, 0.0251, 0.0147, 0.0214, 0.0131, 0.0225, 0.0197], 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') 2023-03-08 19:40:05,295 INFO [train2.py:809] (1/4) Epoch 19, batch 3050, loss[ctc_loss=0.06731, att_loss=0.2356, loss=0.2019, over 16477.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.006654, over 46.00 utterances.], tot_loss[ctc_loss=0.07804, att_loss=0.2385, loss=0.2064, over 3265736.42 frames. utt_duration=1207 frames, utt_pad_proportion=0.06931, over 10835.94 utterances.], batch size: 46, lr: 5.64e-03, grad_scale: 8.0 2023-03-08 19:40:20,709 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=74768.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:41:26,177 INFO [train2.py:809] (1/4) Epoch 19, batch 3100, loss[ctc_loss=0.09169, att_loss=0.2585, loss=0.2251, over 16938.00 frames. utt_duration=1357 frames, utt_pad_proportion=0.009087, over 50.00 utterances.], tot_loss[ctc_loss=0.07848, att_loss=0.2387, loss=0.2067, over 3271622.87 frames. utt_duration=1224 frames, utt_pad_proportion=0.06447, over 10706.46 utterances.], batch size: 50, lr: 5.64e-03, grad_scale: 8.0 2023-03-08 19:41:53,353 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7433, 3.2067, 3.7477, 3.0121, 3.6756, 4.7922, 4.5026, 3.2858], device='cuda:1'), covar=tensor([0.0271, 0.1575, 0.1197, 0.1347, 0.1000, 0.0688, 0.0539, 0.1323], device='cuda:1'), in_proj_covar=tensor([0.0243, 0.0243, 0.0276, 0.0215, 0.0261, 0.0360, 0.0256, 0.0230], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 19:42:11,536 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.275e+02 1.971e+02 2.423e+02 3.077e+02 6.495e+02, threshold=4.846e+02, percent-clipped=7.0 2023-03-08 19:42:46,968 INFO [train2.py:809] (1/4) Epoch 19, batch 3150, loss[ctc_loss=0.06819, att_loss=0.2206, loss=0.1901, over 15868.00 frames. utt_duration=1629 frames, utt_pad_proportion=0.01039, over 39.00 utterances.], tot_loss[ctc_loss=0.07808, att_loss=0.2382, loss=0.2062, over 3268241.40 frames. utt_duration=1235 frames, utt_pad_proportion=0.0586, over 10598.80 utterances.], batch size: 39, lr: 5.64e-03, grad_scale: 8.0 2023-03-08 19:42:55,116 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.5900, 2.1159, 2.2847, 2.6164, 2.9471, 2.1865, 2.2354, 2.5664], device='cuda:1'), covar=tensor([0.1375, 0.3067, 0.2665, 0.1266, 0.1507, 0.1535, 0.2504, 0.1215], device='cuda:1'), in_proj_covar=tensor([0.0108, 0.0116, 0.0114, 0.0102, 0.0115, 0.0101, 0.0122, 0.0091], device='cuda:1'), out_proj_covar=tensor([8.0876e-05, 8.9120e-05, 8.9486e-05, 7.8912e-05, 8.4188e-05, 7.9772e-05, 9.0307e-05, 7.2850e-05], device='cuda:1') 2023-03-08 19:43:47,034 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74896.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:44:03,087 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-03-08 19:44:06,756 INFO [train2.py:809] (1/4) Epoch 19, batch 3200, loss[ctc_loss=0.06925, att_loss=0.2466, loss=0.2111, over 16877.00 frames. utt_duration=1379 frames, utt_pad_proportion=0.007752, over 49.00 utterances.], tot_loss[ctc_loss=0.0784, att_loss=0.2384, loss=0.2064, over 3273945.89 frames. utt_duration=1223 frames, utt_pad_proportion=0.06043, over 10721.39 utterances.], batch size: 49, lr: 5.64e-03, grad_scale: 8.0 2023-03-08 19:44:08,708 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74909.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:44:51,312 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.517e+02 1.907e+02 2.285e+02 3.012e+02 5.502e+02, threshold=4.570e+02, percent-clipped=3.0 2023-03-08 19:45:03,893 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=74944.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:45:15,409 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74951.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:45:15,573 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7832, 3.6993, 3.0888, 3.3522, 3.8535, 3.5430, 2.9782, 4.1142], device='cuda:1'), covar=tensor([0.1115, 0.0514, 0.1074, 0.0674, 0.0641, 0.0718, 0.0844, 0.0414], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0209, 0.0221, 0.0192, 0.0265, 0.0232, 0.0196, 0.0279], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-08 19:45:26,718 INFO [train2.py:809] (1/4) Epoch 19, batch 3250, loss[ctc_loss=0.05317, att_loss=0.2171, loss=0.1843, over 16180.00 frames. utt_duration=1580 frames, utt_pad_proportion=0.006397, over 41.00 utterances.], tot_loss[ctc_loss=0.07868, att_loss=0.2388, loss=0.2068, over 3281655.84 frames. utt_duration=1211 frames, utt_pad_proportion=0.06121, over 10848.94 utterances.], batch size: 41, lr: 5.64e-03, grad_scale: 8.0 2023-03-08 19:45:36,405 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74964.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:45:45,848 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74970.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:46:46,973 INFO [train2.py:809] (1/4) Epoch 19, batch 3300, loss[ctc_loss=0.09613, att_loss=0.2299, loss=0.2031, over 15764.00 frames. utt_duration=1661 frames, utt_pad_proportion=0.007267, over 38.00 utterances.], tot_loss[ctc_loss=0.07903, att_loss=0.2392, loss=0.2072, over 3284870.61 frames. utt_duration=1197 frames, utt_pad_proportion=0.06345, over 10994.05 utterances.], batch size: 38, lr: 5.63e-03, grad_scale: 8.0 2023-03-08 19:47:09,519 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-03-08 19:47:13,511 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=75025.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:47:30,792 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0754, 5.3476, 4.9946, 5.4456, 4.8194, 4.9933, 5.5061, 5.3097], device='cuda:1'), covar=tensor([0.0513, 0.0310, 0.0650, 0.0242, 0.0394, 0.0238, 0.0208, 0.0171], device='cuda:1'), in_proj_covar=tensor([0.0378, 0.0311, 0.0358, 0.0333, 0.0312, 0.0233, 0.0294, 0.0277], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 19:47:31,948 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.356e+02 2.051e+02 2.410e+02 2.967e+02 4.688e+02, threshold=4.820e+02, percent-clipped=1.0 2023-03-08 19:48:06,391 INFO [train2.py:809] (1/4) Epoch 19, batch 3350, loss[ctc_loss=0.0707, att_loss=0.2436, loss=0.209, over 17293.00 frames. utt_duration=1174 frames, utt_pad_proportion=0.02432, over 59.00 utterances.], tot_loss[ctc_loss=0.07858, att_loss=0.2386, loss=0.2066, over 3275573.34 frames. utt_duration=1212 frames, utt_pad_proportion=0.06062, over 10821.98 utterances.], batch size: 59, lr: 5.63e-03, grad_scale: 8.0 2023-03-08 19:49:26,190 INFO [train2.py:809] (1/4) Epoch 19, batch 3400, loss[ctc_loss=0.07374, att_loss=0.234, loss=0.2019, over 16280.00 frames. utt_duration=1516 frames, utt_pad_proportion=0.006523, over 43.00 utterances.], tot_loss[ctc_loss=0.07875, att_loss=0.239, loss=0.2069, over 3276024.91 frames. utt_duration=1193 frames, utt_pad_proportion=0.06653, over 10999.86 utterances.], batch size: 43, lr: 5.63e-03, grad_scale: 8.0 2023-03-08 19:50:11,775 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.156e+02 2.118e+02 2.457e+02 2.938e+02 6.241e+02, threshold=4.915e+02, percent-clipped=4.0 2023-03-08 19:50:32,198 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-03-08 19:50:46,055 INFO [train2.py:809] (1/4) Epoch 19, batch 3450, loss[ctc_loss=0.08616, att_loss=0.224, loss=0.1964, over 15993.00 frames. utt_duration=1601 frames, utt_pad_proportion=0.008116, over 40.00 utterances.], tot_loss[ctc_loss=0.0787, att_loss=0.2392, loss=0.2071, over 3288300.15 frames. utt_duration=1218 frames, utt_pad_proportion=0.05755, over 10816.70 utterances.], batch size: 40, lr: 5.63e-03, grad_scale: 8.0 2023-03-08 19:52:06,158 INFO [train2.py:809] (1/4) Epoch 19, batch 3500, loss[ctc_loss=0.09894, att_loss=0.2581, loss=0.2263, over 16333.00 frames. utt_duration=1453 frames, utt_pad_proportion=0.005913, over 45.00 utterances.], tot_loss[ctc_loss=0.07833, att_loss=0.239, loss=0.2068, over 3279580.47 frames. utt_duration=1218 frames, utt_pad_proportion=0.059, over 10779.33 utterances.], batch size: 45, lr: 5.63e-03, grad_scale: 8.0 2023-03-08 19:52:38,248 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=75228.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:52:52,420 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.557e+02 2.026e+02 2.447e+02 3.089e+02 8.845e+02, threshold=4.893e+02, percent-clipped=4.0 2023-03-08 19:53:15,774 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75251.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:53:26,628 INFO [train2.py:809] (1/4) Epoch 19, batch 3550, loss[ctc_loss=0.05361, att_loss=0.2144, loss=0.1822, over 16184.00 frames. utt_duration=1581 frames, utt_pad_proportion=0.006561, over 41.00 utterances.], tot_loss[ctc_loss=0.07772, att_loss=0.2385, loss=0.2063, over 3274813.02 frames. utt_duration=1236 frames, utt_pad_proportion=0.05586, over 10612.81 utterances.], batch size: 41, lr: 5.62e-03, grad_scale: 8.0 2023-03-08 19:53:37,475 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75265.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:53:45,175 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=75270.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:54:15,819 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=75289.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:54:32,158 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=75299.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:54:46,285 INFO [train2.py:809] (1/4) Epoch 19, batch 3600, loss[ctc_loss=0.1068, att_loss=0.2511, loss=0.2222, over 16491.00 frames. utt_duration=667.7 frames, utt_pad_proportion=0.1591, over 99.00 utterances.], tot_loss[ctc_loss=0.07851, att_loss=0.2384, loss=0.2064, over 3271174.62 frames. utt_duration=1223 frames, utt_pad_proportion=0.06128, over 10707.62 utterances.], batch size: 99, lr: 5.62e-03, grad_scale: 8.0 2023-03-08 19:55:04,935 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75320.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 19:55:23,541 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=75331.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 19:55:32,227 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.166e+02 1.965e+02 2.324e+02 2.872e+02 5.380e+02, threshold=4.648e+02, percent-clipped=3.0 2023-03-08 19:55:39,544 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3318, 2.5761, 4.7395, 3.8507, 2.8480, 4.0630, 4.4381, 4.4528], device='cuda:1'), covar=tensor([0.0215, 0.1650, 0.0188, 0.0911, 0.1778, 0.0276, 0.0210, 0.0233], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0240, 0.0178, 0.0307, 0.0263, 0.0205, 0.0162, 0.0193], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:1') 2023-03-08 19:56:05,135 INFO [train2.py:809] (1/4) Epoch 19, batch 3650, loss[ctc_loss=0.05837, att_loss=0.2152, loss=0.1838, over 15865.00 frames. utt_duration=1629 frames, utt_pad_proportion=0.009062, over 39.00 utterances.], tot_loss[ctc_loss=0.07884, att_loss=0.2384, loss=0.2065, over 3271798.22 frames. utt_duration=1216 frames, utt_pad_proportion=0.06172, over 10773.76 utterances.], batch size: 39, lr: 5.62e-03, grad_scale: 8.0 2023-03-08 19:56:41,872 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9706, 5.0152, 4.7204, 2.7847, 4.6980, 4.6094, 4.1403, 2.7123], device='cuda:1'), covar=tensor([0.0088, 0.0103, 0.0274, 0.1065, 0.0116, 0.0197, 0.0346, 0.1382], device='cuda:1'), in_proj_covar=tensor([0.0073, 0.0101, 0.0101, 0.0111, 0.0084, 0.0111, 0.0099, 0.0105], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 19:57:07,512 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-03-08 19:57:25,713 INFO [train2.py:809] (1/4) Epoch 19, batch 3700, loss[ctc_loss=0.07203, att_loss=0.2383, loss=0.205, over 16868.00 frames. utt_duration=1379 frames, utt_pad_proportion=0.007523, over 49.00 utterances.], tot_loss[ctc_loss=0.07839, att_loss=0.2378, loss=0.2059, over 3279398.07 frames. utt_duration=1244 frames, utt_pad_proportion=0.05327, over 10553.56 utterances.], batch size: 49, lr: 5.62e-03, grad_scale: 8.0 2023-03-08 19:58:11,213 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.172e+02 2.064e+02 2.535e+02 3.260e+02 9.270e+02, threshold=5.071e+02, percent-clipped=6.0 2023-03-08 19:58:35,419 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4869, 4.9960, 4.8831, 5.0089, 5.0570, 4.6602, 3.4097, 4.9023], device='cuda:1'), covar=tensor([0.0131, 0.0134, 0.0132, 0.0090, 0.0088, 0.0121, 0.0791, 0.0217], device='cuda:1'), in_proj_covar=tensor([0.0088, 0.0084, 0.0107, 0.0066, 0.0071, 0.0082, 0.0101, 0.0105], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 19:58:44,489 INFO [train2.py:809] (1/4) Epoch 19, batch 3750, loss[ctc_loss=0.06589, att_loss=0.2481, loss=0.2116, over 16769.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.005715, over 48.00 utterances.], tot_loss[ctc_loss=0.07777, att_loss=0.2373, loss=0.2054, over 3276349.36 frames. utt_duration=1268 frames, utt_pad_proportion=0.04789, over 10349.39 utterances.], batch size: 48, lr: 5.62e-03, grad_scale: 8.0 2023-03-08 20:00:03,992 INFO [train2.py:809] (1/4) Epoch 19, batch 3800, loss[ctc_loss=0.1043, att_loss=0.2583, loss=0.2275, over 16474.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.006026, over 46.00 utterances.], tot_loss[ctc_loss=0.07874, att_loss=0.2379, loss=0.2061, over 3273792.24 frames. utt_duration=1244 frames, utt_pad_proportion=0.05593, over 10539.41 utterances.], batch size: 46, lr: 5.62e-03, grad_scale: 8.0 2023-03-08 20:00:50,191 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.424e+02 2.066e+02 2.537e+02 3.345e+02 7.460e+02, threshold=5.074e+02, percent-clipped=4.0 2023-03-08 20:01:23,606 INFO [train2.py:809] (1/4) Epoch 19, batch 3850, loss[ctc_loss=0.05088, att_loss=0.2203, loss=0.1864, over 16136.00 frames. utt_duration=1538 frames, utt_pad_proportion=0.005725, over 42.00 utterances.], tot_loss[ctc_loss=0.07867, att_loss=0.2378, loss=0.206, over 3267047.10 frames. utt_duration=1231 frames, utt_pad_proportion=0.06176, over 10629.66 utterances.], batch size: 42, lr: 5.61e-03, grad_scale: 8.0 2023-03-08 20:01:34,612 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75565.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 20:01:38,661 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-03-08 20:01:45,825 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.02 vs. limit=5.0 2023-03-08 20:01:48,081 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6592, 5.0398, 4.8327, 4.9772, 5.0920, 4.6782, 3.7035, 4.9943], device='cuda:1'), covar=tensor([0.0118, 0.0107, 0.0141, 0.0087, 0.0090, 0.0121, 0.0598, 0.0210], device='cuda:1'), in_proj_covar=tensor([0.0090, 0.0085, 0.0108, 0.0067, 0.0072, 0.0084, 0.0103, 0.0107], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 20:02:03,509 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75584.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 20:02:40,370 INFO [train2.py:809] (1/4) Epoch 19, batch 3900, loss[ctc_loss=0.05464, att_loss=0.2103, loss=0.1791, over 16004.00 frames. utt_duration=1602 frames, utt_pad_proportion=0.008034, over 40.00 utterances.], tot_loss[ctc_loss=0.07862, att_loss=0.2382, loss=0.2063, over 3265865.62 frames. utt_duration=1212 frames, utt_pad_proportion=0.06728, over 10787.54 utterances.], batch size: 40, lr: 5.61e-03, grad_scale: 8.0 2023-03-08 20:02:47,812 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=75613.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 20:02:58,550 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75620.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 20:03:07,695 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75626.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 20:03:24,259 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.284e+02 1.998e+02 2.333e+02 2.845e+02 5.732e+02, threshold=4.667e+02, percent-clipped=1.0 2023-03-08 20:03:57,010 INFO [train2.py:809] (1/4) Epoch 19, batch 3950, loss[ctc_loss=0.077, att_loss=0.2231, loss=0.1939, over 15963.00 frames. utt_duration=1559 frames, utt_pad_proportion=0.006467, over 41.00 utterances.], tot_loss[ctc_loss=0.07862, att_loss=0.238, loss=0.2061, over 3268557.62 frames. utt_duration=1221 frames, utt_pad_proportion=0.06518, over 10723.66 utterances.], batch size: 41, lr: 5.61e-03, grad_scale: 8.0 2023-03-08 20:04:00,447 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2538, 2.7690, 3.0716, 4.2001, 3.8848, 3.9255, 2.8872, 2.0454], device='cuda:1'), covar=tensor([0.0795, 0.2029, 0.1048, 0.0635, 0.0780, 0.0443, 0.1544, 0.2442], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0213, 0.0187, 0.0214, 0.0217, 0.0174, 0.0202, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 20:04:12,252 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=75668.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 20:04:41,178 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6737, 5.1076, 4.9168, 5.0540, 5.1466, 4.7421, 3.6688, 5.1154], device='cuda:1'), covar=tensor([0.0114, 0.0107, 0.0110, 0.0076, 0.0075, 0.0094, 0.0616, 0.0167], device='cuda:1'), in_proj_covar=tensor([0.0090, 0.0086, 0.0108, 0.0067, 0.0072, 0.0084, 0.0103, 0.0107], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 20:05:12,143 INFO [train2.py:809] (1/4) Epoch 20, batch 0, loss[ctc_loss=0.08456, att_loss=0.2566, loss=0.2222, over 16995.00 frames. utt_duration=688.2 frames, utt_pad_proportion=0.1375, over 99.00 utterances.], tot_loss[ctc_loss=0.08456, att_loss=0.2566, loss=0.2222, over 16995.00 frames. utt_duration=688.2 frames, utt_pad_proportion=0.1375, over 99.00 utterances.], batch size: 99, lr: 5.46e-03, grad_scale: 8.0 2023-03-08 20:05:12,143 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-08 20:05:24,208 INFO [train2.py:843] (1/4) Epoch 20, validation: ctc_loss=0.04136, att_loss=0.235, loss=0.1963, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-08 20:05:24,208 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-08 20:06:35,314 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.268e+02 2.109e+02 2.667e+02 3.273e+02 6.239e+02, threshold=5.334e+02, percent-clipped=6.0 2023-03-08 20:06:37,670 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-03-08 20:06:43,084 INFO [train2.py:809] (1/4) Epoch 20, batch 50, loss[ctc_loss=0.06334, att_loss=0.2146, loss=0.1843, over 15638.00 frames. utt_duration=1692 frames, utt_pad_proportion=0.009099, over 37.00 utterances.], tot_loss[ctc_loss=0.07361, att_loss=0.237, loss=0.2043, over 744130.83 frames. utt_duration=1244 frames, utt_pad_proportion=0.04403, over 2395.00 utterances.], batch size: 37, lr: 5.46e-03, grad_scale: 8.0 2023-03-08 20:07:27,234 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-03-08 20:08:03,726 INFO [train2.py:809] (1/4) Epoch 20, batch 100, loss[ctc_loss=0.05671, att_loss=0.2094, loss=0.1788, over 15530.00 frames. utt_duration=1727 frames, utt_pad_proportion=0.006869, over 36.00 utterances.], tot_loss[ctc_loss=0.07325, att_loss=0.2378, loss=0.2049, over 1311959.71 frames. utt_duration=1249 frames, utt_pad_proportion=0.04505, over 4206.75 utterances.], batch size: 36, lr: 5.46e-03, grad_scale: 8.0 2023-03-08 20:08:16,366 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4819, 4.8466, 4.6678, 4.7792, 4.8415, 4.5192, 3.2661, 4.7485], device='cuda:1'), covar=tensor([0.0114, 0.0115, 0.0142, 0.0091, 0.0101, 0.0121, 0.0710, 0.0213], device='cuda:1'), in_proj_covar=tensor([0.0089, 0.0085, 0.0107, 0.0066, 0.0072, 0.0083, 0.0102, 0.0106], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 20:08:56,147 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-03-08 20:09:16,372 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.261e+02 1.985e+02 2.361e+02 2.843e+02 6.853e+02, threshold=4.722e+02, percent-clipped=2.0 2023-03-08 20:09:24,379 INFO [train2.py:809] (1/4) Epoch 20, batch 150, loss[ctc_loss=0.1065, att_loss=0.2582, loss=0.2278, over 17337.00 frames. utt_duration=1177 frames, utt_pad_proportion=0.02163, over 59.00 utterances.], tot_loss[ctc_loss=0.07453, att_loss=0.2375, loss=0.2049, over 1753391.19 frames. utt_duration=1285 frames, utt_pad_proportion=0.03642, over 5465.08 utterances.], batch size: 59, lr: 5.46e-03, grad_scale: 8.0 2023-03-08 20:10:33,727 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75884.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 20:10:44,799 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6391, 3.1394, 3.8235, 3.1172, 3.7381, 4.7903, 4.6164, 3.3688], device='cuda:1'), covar=tensor([0.0378, 0.1525, 0.1108, 0.1272, 0.0959, 0.0678, 0.0475, 0.1249], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0241, 0.0275, 0.0216, 0.0263, 0.0355, 0.0256, 0.0229], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 20:10:46,196 INFO [train2.py:809] (1/4) Epoch 20, batch 200, loss[ctc_loss=0.06932, att_loss=0.2151, loss=0.186, over 15857.00 frames. utt_duration=1628 frames, utt_pad_proportion=0.01049, over 39.00 utterances.], tot_loss[ctc_loss=0.07497, att_loss=0.2373, loss=0.2048, over 2098265.04 frames. utt_duration=1304 frames, utt_pad_proportion=0.03191, over 6445.41 utterances.], batch size: 39, lr: 5.46e-03, grad_scale: 8.0 2023-03-08 20:11:42,085 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75926.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 20:11:51,161 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=75932.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 20:11:59,002 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.248e+02 2.084e+02 2.421e+02 2.926e+02 5.998e+02, threshold=4.842e+02, percent-clipped=2.0 2023-03-08 20:12:06,831 INFO [train2.py:809] (1/4) Epoch 20, batch 250, loss[ctc_loss=0.07814, att_loss=0.2553, loss=0.2199, over 17370.00 frames. utt_duration=1104 frames, utt_pad_proportion=0.03483, over 63.00 utterances.], tot_loss[ctc_loss=0.07455, att_loss=0.2367, loss=0.2043, over 2363235.93 frames. utt_duration=1317 frames, utt_pad_proportion=0.02956, over 7184.44 utterances.], batch size: 63, lr: 5.45e-03, grad_scale: 8.0 2023-03-08 20:12:59,213 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=75974.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 20:13:08,687 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0187, 4.0829, 4.0115, 4.1064, 4.4851, 4.0046, 3.8710, 2.4242], device='cuda:1'), covar=tensor([0.0277, 0.0410, 0.0420, 0.0330, 0.0838, 0.0263, 0.0423, 0.1765], device='cuda:1'), in_proj_covar=tensor([0.0151, 0.0174, 0.0176, 0.0192, 0.0356, 0.0147, 0.0165, 0.0211], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 20:13:09,182 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-03-08 20:13:10,205 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1346, 4.4877, 4.6346, 4.7929, 2.9073, 4.5843, 2.9709, 1.4943], device='cuda:1'), covar=tensor([0.0489, 0.0226, 0.0584, 0.0247, 0.1543, 0.0148, 0.1329, 0.1867], device='cuda:1'), in_proj_covar=tensor([0.0182, 0.0153, 0.0253, 0.0150, 0.0215, 0.0133, 0.0227, 0.0198], 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') 2023-03-08 20:13:25,319 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1640, 5.5132, 4.8550, 5.6214, 4.9810, 5.1316, 5.6053, 5.3791], device='cuda:1'), covar=tensor([0.0478, 0.0229, 0.0825, 0.0269, 0.0331, 0.0192, 0.0231, 0.0213], device='cuda:1'), in_proj_covar=tensor([0.0378, 0.0311, 0.0360, 0.0336, 0.0313, 0.0233, 0.0295, 0.0280], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 20:13:26,595 INFO [train2.py:809] (1/4) Epoch 20, batch 300, loss[ctc_loss=0.06536, att_loss=0.2095, loss=0.1807, over 14063.00 frames. utt_duration=1816 frames, utt_pad_proportion=0.05107, over 31.00 utterances.], tot_loss[ctc_loss=0.07544, att_loss=0.2368, loss=0.2045, over 2561433.34 frames. utt_duration=1311 frames, utt_pad_proportion=0.03608, over 7825.75 utterances.], batch size: 31, lr: 5.45e-03, grad_scale: 8.0 2023-03-08 20:14:42,640 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.429e+02 2.144e+02 2.516e+02 3.059e+02 1.142e+03, threshold=5.032e+02, percent-clipped=7.0 2023-03-08 20:14:50,252 INFO [train2.py:809] (1/4) Epoch 20, batch 350, loss[ctc_loss=0.1091, att_loss=0.2573, loss=0.2277, over 14664.00 frames. utt_duration=403.2 frames, utt_pad_proportion=0.2975, over 146.00 utterances.], tot_loss[ctc_loss=0.07617, att_loss=0.2374, loss=0.2052, over 2725582.85 frames. utt_duration=1284 frames, utt_pad_proportion=0.04192, over 8501.63 utterances.], batch size: 146, lr: 5.45e-03, grad_scale: 8.0 2023-03-08 20:16:10,637 INFO [train2.py:809] (1/4) Epoch 20, batch 400, loss[ctc_loss=0.104, att_loss=0.2563, loss=0.2258, over 17293.00 frames. utt_duration=1004 frames, utt_pad_proportion=0.05287, over 69.00 utterances.], tot_loss[ctc_loss=0.07636, att_loss=0.2377, loss=0.2054, over 2849050.78 frames. utt_duration=1272 frames, utt_pad_proportion=0.04544, over 8972.78 utterances.], batch size: 69, lr: 5.45e-03, grad_scale: 8.0 2023-03-08 20:17:05,469 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-03-08 20:17:22,265 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.475e+02 1.983e+02 2.448e+02 2.910e+02 4.561e+02, threshold=4.897e+02, percent-clipped=0.0 2023-03-08 20:17:29,976 INFO [train2.py:809] (1/4) Epoch 20, batch 450, loss[ctc_loss=0.06278, att_loss=0.241, loss=0.2054, over 16765.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.006599, over 48.00 utterances.], tot_loss[ctc_loss=0.07593, att_loss=0.2368, loss=0.2046, over 2942123.31 frames. utt_duration=1312 frames, utt_pad_proportion=0.03789, over 8980.79 utterances.], batch size: 48, lr: 5.45e-03, grad_scale: 8.0 2023-03-08 20:18:49,889 INFO [train2.py:809] (1/4) Epoch 20, batch 500, loss[ctc_loss=0.05425, att_loss=0.2062, loss=0.1758, over 15663.00 frames. utt_duration=1695 frames, utt_pad_proportion=0.007611, over 37.00 utterances.], tot_loss[ctc_loss=0.07533, att_loss=0.236, loss=0.2039, over 3012632.29 frames. utt_duration=1317 frames, utt_pad_proportion=0.038, over 9159.79 utterances.], batch size: 37, lr: 5.45e-03, grad_scale: 8.0 2023-03-08 20:20:01,813 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.260e+02 1.846e+02 2.334e+02 2.812e+02 5.634e+02, threshold=4.669e+02, percent-clipped=1.0 2023-03-08 20:20:09,801 INFO [train2.py:809] (1/4) Epoch 20, batch 550, loss[ctc_loss=0.05879, att_loss=0.2062, loss=0.1768, over 15646.00 frames. utt_duration=1693 frames, utt_pad_proportion=0.008782, over 37.00 utterances.], tot_loss[ctc_loss=0.07551, att_loss=0.2361, loss=0.204, over 3076096.22 frames. utt_duration=1320 frames, utt_pad_proportion=0.03489, over 9329.52 utterances.], batch size: 37, lr: 5.44e-03, grad_scale: 8.0 2023-03-08 20:20:54,243 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76269.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 20:20:55,932 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0712, 5.0510, 4.8581, 2.3239, 1.8533, 2.9794, 2.2226, 3.8011], device='cuda:1'), covar=tensor([0.0726, 0.0285, 0.0276, 0.5273, 0.6223, 0.2315, 0.3739, 0.1751], device='cuda:1'), in_proj_covar=tensor([0.0356, 0.0268, 0.0265, 0.0243, 0.0346, 0.0336, 0.0252, 0.0366], device='cuda:1'), out_proj_covar=tensor([1.5187e-04, 9.9401e-05, 1.1314e-04, 1.0509e-04, 1.4534e-04, 1.3169e-04, 1.0068e-04, 1.4920e-04], device='cuda:1') 2023-03-08 20:21:29,926 INFO [train2.py:809] (1/4) Epoch 20, batch 600, loss[ctc_loss=0.07224, att_loss=0.2178, loss=0.1887, over 15636.00 frames. utt_duration=1692 frames, utt_pad_proportion=0.008724, over 37.00 utterances.], tot_loss[ctc_loss=0.07518, att_loss=0.2363, loss=0.2041, over 3129341.53 frames. utt_duration=1323 frames, utt_pad_proportion=0.03173, over 9470.31 utterances.], batch size: 37, lr: 5.44e-03, grad_scale: 16.0 2023-03-08 20:21:33,218 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76294.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 20:22:13,521 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.33 vs. limit=5.0 2023-03-08 20:22:14,832 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.89 vs. limit=2.0 2023-03-08 20:22:20,720 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76323.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 20:22:28,153 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9943, 5.2378, 5.1289, 5.1180, 5.2488, 5.1906, 4.9231, 4.7229], device='cuda:1'), covar=tensor([0.0856, 0.0488, 0.0297, 0.0469, 0.0283, 0.0305, 0.0332, 0.0334], device='cuda:1'), in_proj_covar=tensor([0.0520, 0.0353, 0.0333, 0.0348, 0.0410, 0.0422, 0.0344, 0.0387], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0003, 0.0004], device='cuda:1') 2023-03-08 20:22:31,334 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76330.0, num_to_drop=1, layers_to_drop={3} 2023-03-08 20:22:41,634 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.413e+02 1.945e+02 2.306e+02 2.690e+02 7.016e+02, threshold=4.612e+02, percent-clipped=2.0 2023-03-08 20:22:49,616 INFO [train2.py:809] (1/4) Epoch 20, batch 650, loss[ctc_loss=0.0724, att_loss=0.2442, loss=0.2099, over 16876.00 frames. utt_duration=1379 frames, utt_pad_proportion=0.007605, over 49.00 utterances.], tot_loss[ctc_loss=0.07571, att_loss=0.237, loss=0.2047, over 3165674.19 frames. utt_duration=1305 frames, utt_pad_proportion=0.03735, over 9716.69 utterances.], batch size: 49, lr: 5.44e-03, grad_scale: 16.0 2023-03-08 20:23:02,187 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-03-08 20:23:11,372 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76355.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 20:23:47,836 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.7642, 3.9824, 3.9433, 3.9951, 4.0326, 3.8512, 3.1240, 3.9340], device='cuda:1'), covar=tensor([0.0130, 0.0133, 0.0134, 0.0097, 0.0103, 0.0129, 0.0588, 0.0233], device='cuda:1'), in_proj_covar=tensor([0.0091, 0.0087, 0.0110, 0.0068, 0.0074, 0.0085, 0.0104, 0.0110], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 20:23:56,999 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76384.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 20:24:09,248 INFO [train2.py:809] (1/4) Epoch 20, batch 700, loss[ctc_loss=0.07771, att_loss=0.2323, loss=0.2013, over 16120.00 frames. utt_duration=1537 frames, utt_pad_proportion=0.006526, over 42.00 utterances.], tot_loss[ctc_loss=0.0761, att_loss=0.2372, loss=0.205, over 3187440.14 frames. utt_duration=1286 frames, utt_pad_proportion=0.04265, over 9923.51 utterances.], batch size: 42, lr: 5.44e-03, grad_scale: 16.0 2023-03-08 20:25:20,470 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.218e+02 1.882e+02 2.322e+02 2.702e+02 6.742e+02, threshold=4.645e+02, percent-clipped=2.0 2023-03-08 20:25:28,805 INFO [train2.py:809] (1/4) Epoch 20, batch 750, loss[ctc_loss=0.06285, att_loss=0.2282, loss=0.1951, over 16397.00 frames. utt_duration=1492 frames, utt_pad_proportion=0.006673, over 44.00 utterances.], tot_loss[ctc_loss=0.07602, att_loss=0.2373, loss=0.205, over 3212328.16 frames. utt_duration=1277 frames, utt_pad_proportion=0.04322, over 10074.95 utterances.], batch size: 44, lr: 5.44e-03, grad_scale: 16.0 2023-03-08 20:26:20,847 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-03-08 20:26:48,173 INFO [train2.py:809] (1/4) Epoch 20, batch 800, loss[ctc_loss=0.08364, att_loss=0.247, loss=0.2143, over 17269.00 frames. utt_duration=1258 frames, utt_pad_proportion=0.0129, over 55.00 utterances.], tot_loss[ctc_loss=0.07747, att_loss=0.2386, loss=0.2063, over 3229425.47 frames. utt_duration=1258 frames, utt_pad_proportion=0.04889, over 10278.56 utterances.], batch size: 55, lr: 5.44e-03, grad_scale: 16.0 2023-03-08 20:28:00,598 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.342e+02 1.827e+02 2.351e+02 2.962e+02 7.560e+02, threshold=4.702e+02, percent-clipped=5.0 2023-03-08 20:28:07,272 INFO [train2.py:809] (1/4) Epoch 20, batch 850, loss[ctc_loss=0.07201, att_loss=0.2212, loss=0.1913, over 15766.00 frames. utt_duration=1661 frames, utt_pad_proportion=0.00722, over 38.00 utterances.], tot_loss[ctc_loss=0.07687, att_loss=0.238, loss=0.2058, over 3243021.68 frames. utt_duration=1241 frames, utt_pad_proportion=0.05257, over 10469.72 utterances.], batch size: 38, lr: 5.43e-03, grad_scale: 8.0 2023-03-08 20:28:45,936 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6313, 4.5693, 4.6369, 4.6938, 5.1224, 4.6835, 4.5837, 2.5682], device='cuda:1'), covar=tensor([0.0190, 0.0332, 0.0318, 0.0313, 0.1067, 0.0174, 0.0319, 0.1893], device='cuda:1'), in_proj_covar=tensor([0.0153, 0.0175, 0.0178, 0.0193, 0.0360, 0.0149, 0.0166, 0.0212], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 20:28:55,446 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-03-08 20:29:26,685 INFO [train2.py:809] (1/4) Epoch 20, batch 900, loss[ctc_loss=0.1033, att_loss=0.2484, loss=0.2193, over 15968.00 frames. utt_duration=1559 frames, utt_pad_proportion=0.006234, over 41.00 utterances.], tot_loss[ctc_loss=0.07676, att_loss=0.2377, loss=0.2055, over 3244948.05 frames. utt_duration=1223 frames, utt_pad_proportion=0.05828, over 10623.67 utterances.], batch size: 41, lr: 5.43e-03, grad_scale: 8.0 2023-03-08 20:30:21,608 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=76625.0, num_to_drop=1, layers_to_drop={3} 2023-03-08 20:30:41,671 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.428e+02 2.021e+02 2.452e+02 3.045e+02 6.796e+02, threshold=4.904e+02, percent-clipped=5.0 2023-03-08 20:30:47,771 INFO [train2.py:809] (1/4) Epoch 20, batch 950, loss[ctc_loss=0.09443, att_loss=0.2459, loss=0.2156, over 17067.00 frames. utt_duration=691.3 frames, utt_pad_proportion=0.1327, over 99.00 utterances.], tot_loss[ctc_loss=0.0766, att_loss=0.2371, loss=0.205, over 3243446.90 frames. utt_duration=1202 frames, utt_pad_proportion=0.06633, over 10802.76 utterances.], batch size: 99, lr: 5.43e-03, grad_scale: 8.0 2023-03-08 20:31:02,303 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=76650.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 20:31:47,599 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=76679.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 20:32:08,851 INFO [train2.py:809] (1/4) Epoch 20, batch 1000, loss[ctc_loss=0.08118, att_loss=0.2495, loss=0.2158, over 17045.00 frames. utt_duration=1288 frames, utt_pad_proportion=0.009332, over 53.00 utterances.], tot_loss[ctc_loss=0.0768, att_loss=0.2377, loss=0.2055, over 3247560.00 frames. utt_duration=1188 frames, utt_pad_proportion=0.07094, over 10950.83 utterances.], batch size: 53, lr: 5.43e-03, grad_scale: 8.0 2023-03-08 20:32:41,971 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5258, 5.2494, 5.3439, 5.3136, 5.2037, 5.2549, 5.1451, 4.8377], device='cuda:1'), covar=tensor([0.1648, 0.0720, 0.0348, 0.0513, 0.0639, 0.0459, 0.0341, 0.0384], device='cuda:1'), in_proj_covar=tensor([0.0517, 0.0353, 0.0334, 0.0351, 0.0410, 0.0424, 0.0346, 0.0387], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0003, 0.0004], device='cuda:1') 2023-03-08 20:33:21,296 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.239e+02 1.956e+02 2.404e+02 2.887e+02 6.438e+02, threshold=4.808e+02, percent-clipped=2.0 2023-03-08 20:33:28,411 INFO [train2.py:809] (1/4) Epoch 20, batch 1050, loss[ctc_loss=0.118, att_loss=0.2637, loss=0.2345, over 13300.00 frames. utt_duration=368.5 frames, utt_pad_proportion=0.3603, over 145.00 utterances.], tot_loss[ctc_loss=0.07692, att_loss=0.2376, loss=0.2054, over 3255879.97 frames. utt_duration=1185 frames, utt_pad_proportion=0.07005, over 11004.05 utterances.], batch size: 145, lr: 5.43e-03, grad_scale: 8.0 2023-03-08 20:34:48,265 INFO [train2.py:809] (1/4) Epoch 20, batch 1100, loss[ctc_loss=0.08504, att_loss=0.232, loss=0.2026, over 15996.00 frames. utt_duration=1601 frames, utt_pad_proportion=0.008514, over 40.00 utterances.], tot_loss[ctc_loss=0.07688, att_loss=0.2374, loss=0.2053, over 3253738.30 frames. utt_duration=1200 frames, utt_pad_proportion=0.06758, over 10855.37 utterances.], batch size: 40, lr: 5.42e-03, grad_scale: 8.0 2023-03-08 20:36:02,083 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.131e+02 2.008e+02 2.417e+02 3.031e+02 6.018e+02, threshold=4.833e+02, percent-clipped=3.0 2023-03-08 20:36:03,932 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76839.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 20:36:08,274 INFO [train2.py:809] (1/4) Epoch 20, batch 1150, loss[ctc_loss=0.06655, att_loss=0.2233, loss=0.1919, over 14475.00 frames. utt_duration=1811 frames, utt_pad_proportion=0.04631, over 32.00 utterances.], tot_loss[ctc_loss=0.07682, att_loss=0.237, loss=0.205, over 3249651.49 frames. utt_duration=1206 frames, utt_pad_proportion=0.06657, over 10794.00 utterances.], batch size: 32, lr: 5.42e-03, grad_scale: 8.0 2023-03-08 20:36:11,267 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.32 vs. limit=5.0 2023-03-08 20:37:06,860 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-03-08 20:37:28,708 INFO [train2.py:809] (1/4) Epoch 20, batch 1200, loss[ctc_loss=0.09384, att_loss=0.255, loss=0.2227, over 17222.00 frames. utt_duration=873.6 frames, utt_pad_proportion=0.08327, over 79.00 utterances.], tot_loss[ctc_loss=0.07623, att_loss=0.2362, loss=0.2042, over 3255629.93 frames. utt_duration=1217 frames, utt_pad_proportion=0.06294, over 10713.87 utterances.], batch size: 79, lr: 5.42e-03, grad_scale: 8.0 2023-03-08 20:37:41,994 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76900.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 20:38:01,668 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-03-08 20:38:21,828 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=76925.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 20:38:42,103 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.259e+02 1.866e+02 2.347e+02 2.914e+02 5.639e+02, threshold=4.693e+02, percent-clipped=1.0 2023-03-08 20:38:49,696 INFO [train2.py:809] (1/4) Epoch 20, batch 1250, loss[ctc_loss=0.06165, att_loss=0.2258, loss=0.193, over 16123.00 frames. utt_duration=1537 frames, utt_pad_proportion=0.006341, over 42.00 utterances.], tot_loss[ctc_loss=0.0764, att_loss=0.2365, loss=0.2045, over 3266528.67 frames. utt_duration=1195 frames, utt_pad_proportion=0.0634, over 10948.84 utterances.], batch size: 42, lr: 5.42e-03, grad_scale: 8.0 2023-03-08 20:39:02,625 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=76950.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 20:39:38,061 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=76973.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 20:39:48,071 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=76979.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 20:40:10,280 INFO [train2.py:809] (1/4) Epoch 20, batch 1300, loss[ctc_loss=0.06669, att_loss=0.2236, loss=0.1922, over 16170.00 frames. utt_duration=1579 frames, utt_pad_proportion=0.007102, over 41.00 utterances.], tot_loss[ctc_loss=0.07665, att_loss=0.2366, loss=0.2046, over 3264559.51 frames. utt_duration=1183 frames, utt_pad_proportion=0.06864, over 11051.18 utterances.], batch size: 41, lr: 5.42e-03, grad_scale: 8.0 2023-03-08 20:40:19,513 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=76998.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 20:40:26,686 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-03-08 20:41:04,683 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=77027.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 20:41:22,762 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.337e+02 2.046e+02 2.509e+02 3.001e+02 6.074e+02, threshold=5.017e+02, percent-clipped=4.0 2023-03-08 20:41:29,843 INFO [train2.py:809] (1/4) Epoch 20, batch 1350, loss[ctc_loss=0.08794, att_loss=0.2403, loss=0.2098, over 17035.00 frames. utt_duration=689.9 frames, utt_pad_proportion=0.1355, over 99.00 utterances.], tot_loss[ctc_loss=0.07683, att_loss=0.237, loss=0.205, over 3269878.18 frames. utt_duration=1179 frames, utt_pad_proportion=0.06937, over 11105.82 utterances.], batch size: 99, lr: 5.42e-03, grad_scale: 8.0 2023-03-08 20:42:48,972 INFO [train2.py:809] (1/4) Epoch 20, batch 1400, loss[ctc_loss=0.0814, att_loss=0.2479, loss=0.2146, over 16974.00 frames. utt_duration=1359 frames, utt_pad_proportion=0.007012, over 50.00 utterances.], tot_loss[ctc_loss=0.0776, att_loss=0.2376, loss=0.2056, over 3268727.62 frames. utt_duration=1164 frames, utt_pad_proportion=0.07442, over 11243.43 utterances.], batch size: 50, lr: 5.41e-03, grad_scale: 8.0 2023-03-08 20:42:54,058 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77095.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 20:44:00,818 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0575, 4.5143, 4.5718, 4.6188, 2.8710, 4.4335, 2.8382, 1.8949], device='cuda:1'), covar=tensor([0.0376, 0.0206, 0.0684, 0.0204, 0.1697, 0.0174, 0.1511, 0.1819], device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0156, 0.0256, 0.0152, 0.0218, 0.0137, 0.0230, 0.0202], 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') 2023-03-08 20:44:01,977 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.377e+02 1.974e+02 2.358e+02 2.926e+02 6.381e+02, threshold=4.717e+02, percent-clipped=2.0 2023-03-08 20:44:08,006 INFO [train2.py:809] (1/4) Epoch 20, batch 1450, loss[ctc_loss=0.07454, att_loss=0.2438, loss=0.2099, over 16630.00 frames. utt_duration=1417 frames, utt_pad_proportion=0.00505, over 47.00 utterances.], tot_loss[ctc_loss=0.07808, att_loss=0.2386, loss=0.2065, over 3275465.52 frames. utt_duration=1179 frames, utt_pad_proportion=0.06956, over 11130.89 utterances.], batch size: 47, lr: 5.41e-03, grad_scale: 8.0 2023-03-08 20:44:21,056 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.93 vs. limit=2.0 2023-03-08 20:44:29,621 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77156.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 20:45:19,703 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4882, 2.4145, 4.9137, 3.7786, 2.8740, 4.1657, 4.6675, 4.4699], device='cuda:1'), covar=tensor([0.0223, 0.1776, 0.0143, 0.0953, 0.1875, 0.0259, 0.0152, 0.0262], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0239, 0.0177, 0.0306, 0.0264, 0.0206, 0.0164, 0.0195], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 20:45:27,439 INFO [train2.py:809] (1/4) Epoch 20, batch 1500, loss[ctc_loss=0.07314, att_loss=0.2382, loss=0.2052, over 16969.00 frames. utt_duration=1359 frames, utt_pad_proportion=0.006637, over 50.00 utterances.], tot_loss[ctc_loss=0.07735, att_loss=0.2378, loss=0.2057, over 3280405.55 frames. utt_duration=1204 frames, utt_pad_proportion=0.06161, over 10908.74 utterances.], batch size: 50, lr: 5.41e-03, grad_scale: 8.0 2023-03-08 20:45:31,983 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77195.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 20:45:42,321 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 2023-03-08 20:45:57,151 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77211.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 20:46:40,578 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.285e+02 1.945e+02 2.357e+02 2.826e+02 4.354e+02, threshold=4.714e+02, percent-clipped=0.0 2023-03-08 20:46:46,772 INFO [train2.py:809] (1/4) Epoch 20, batch 1550, loss[ctc_loss=0.07084, att_loss=0.2408, loss=0.2068, over 16976.00 frames. utt_duration=1359 frames, utt_pad_proportion=0.006998, over 50.00 utterances.], tot_loss[ctc_loss=0.07626, att_loss=0.2368, loss=0.2047, over 3279314.15 frames. utt_duration=1249 frames, utt_pad_proportion=0.05227, over 10513.97 utterances.], batch size: 50, lr: 5.41e-03, grad_scale: 8.0 2023-03-08 20:46:59,887 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-03-08 20:47:34,345 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77272.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 20:47:35,811 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77273.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 20:47:42,422 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77277.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 20:48:05,778 INFO [train2.py:809] (1/4) Epoch 20, batch 1600, loss[ctc_loss=0.06508, att_loss=0.2135, loss=0.1838, over 15353.00 frames. utt_duration=1756 frames, utt_pad_proportion=0.01212, over 35.00 utterances.], tot_loss[ctc_loss=0.07607, att_loss=0.2371, loss=0.2049, over 3284399.22 frames. utt_duration=1236 frames, utt_pad_proportion=0.05391, over 10639.87 utterances.], batch size: 35, lr: 5.41e-03, grad_scale: 8.0 2023-03-08 20:48:12,242 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.2548, 5.3150, 5.1076, 3.3331, 5.0478, 4.9766, 4.7046, 3.4449], device='cuda:1'), covar=tensor([0.0138, 0.0076, 0.0230, 0.0850, 0.0094, 0.0159, 0.0242, 0.1041], device='cuda:1'), in_proj_covar=tensor([0.0074, 0.0101, 0.0103, 0.0112, 0.0084, 0.0111, 0.0099, 0.0105], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 20:49:04,242 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0024, 4.3200, 4.1951, 4.3961, 2.8016, 4.3202, 2.5582, 1.6178], device='cuda:1'), covar=tensor([0.0446, 0.0200, 0.0701, 0.0241, 0.1576, 0.0181, 0.1547, 0.1795], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0156, 0.0256, 0.0153, 0.0219, 0.0136, 0.0229, 0.0203], 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') 2023-03-08 20:49:11,742 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77334.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 20:49:18,256 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.314e+02 1.917e+02 2.281e+02 2.951e+02 6.682e+02, threshold=4.563e+02, percent-clipped=3.0 2023-03-08 20:49:18,721 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77338.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 20:49:25,152 INFO [train2.py:809] (1/4) Epoch 20, batch 1650, loss[ctc_loss=0.06386, att_loss=0.213, loss=0.1832, over 16164.00 frames. utt_duration=1578 frames, utt_pad_proportion=0.007302, over 41.00 utterances.], tot_loss[ctc_loss=0.07629, att_loss=0.2373, loss=0.2051, over 3286837.90 frames. utt_duration=1220 frames, utt_pad_proportion=0.05553, over 10787.19 utterances.], batch size: 41, lr: 5.41e-03, grad_scale: 8.0 2023-03-08 20:50:09,223 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.6118, 2.3342, 2.3901, 2.3410, 2.9396, 2.4807, 2.5738, 3.1188], device='cuda:1'), covar=tensor([0.2116, 0.3248, 0.2393, 0.1758, 0.1623, 0.1508, 0.2409, 0.1731], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0124, 0.0118, 0.0108, 0.0119, 0.0104, 0.0127, 0.0096], device='cuda:1'), out_proj_covar=tensor([8.7155e-05, 9.4782e-05, 9.3330e-05, 8.3501e-05, 8.7820e-05, 8.3222e-05, 9.4927e-05, 7.6675e-05], device='cuda:1') 2023-03-08 20:50:44,527 INFO [train2.py:809] (1/4) Epoch 20, batch 1700, loss[ctc_loss=0.08256, att_loss=0.2459, loss=0.2133, over 16266.00 frames. utt_duration=1515 frames, utt_pad_proportion=0.007163, over 43.00 utterances.], tot_loss[ctc_loss=0.0766, att_loss=0.2378, loss=0.2056, over 3293046.84 frames. utt_duration=1233 frames, utt_pad_proportion=0.05177, over 10694.64 utterances.], batch size: 43, lr: 5.40e-03, grad_scale: 8.0 2023-03-08 20:51:57,825 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.333e+02 1.985e+02 2.370e+02 2.830e+02 5.351e+02, threshold=4.740e+02, percent-clipped=1.0 2023-03-08 20:52:04,405 INFO [train2.py:809] (1/4) Epoch 20, batch 1750, loss[ctc_loss=0.07074, att_loss=0.2313, loss=0.1992, over 16293.00 frames. utt_duration=1517 frames, utt_pad_proportion=0.005913, over 43.00 utterances.], tot_loss[ctc_loss=0.07625, att_loss=0.2371, loss=0.2049, over 3291977.67 frames. utt_duration=1259 frames, utt_pad_proportion=0.04592, over 10474.60 utterances.], batch size: 43, lr: 5.40e-03, grad_scale: 8.0 2023-03-08 20:52:18,567 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77451.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 20:52:26,278 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0361, 6.3029, 5.7015, 6.0165, 5.9461, 5.4103, 5.7692, 5.4885], device='cuda:1'), covar=tensor([0.1410, 0.0887, 0.0982, 0.0746, 0.0978, 0.1563, 0.2099, 0.2398], device='cuda:1'), in_proj_covar=tensor([0.0521, 0.0609, 0.0461, 0.0450, 0.0433, 0.0471, 0.0613, 0.0529], 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') 2023-03-08 20:53:24,308 INFO [train2.py:809] (1/4) Epoch 20, batch 1800, loss[ctc_loss=0.07571, att_loss=0.2378, loss=0.2054, over 16972.00 frames. utt_duration=1359 frames, utt_pad_proportion=0.006491, over 50.00 utterances.], tot_loss[ctc_loss=0.07646, att_loss=0.2375, loss=0.2053, over 3295133.06 frames. utt_duration=1264 frames, utt_pad_proportion=0.04429, over 10440.14 utterances.], batch size: 50, lr: 5.40e-03, grad_scale: 8.0 2023-03-08 20:53:29,183 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77495.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 20:54:08,878 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0856, 4.4116, 4.6464, 4.5767, 2.7482, 4.5601, 2.5522, 1.8245], device='cuda:1'), covar=tensor([0.0410, 0.0220, 0.0570, 0.0209, 0.1646, 0.0156, 0.1555, 0.1712], device='cuda:1'), in_proj_covar=tensor([0.0187, 0.0159, 0.0261, 0.0155, 0.0223, 0.0139, 0.0233, 0.0206], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 20:54:13,453 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5460, 3.0058, 3.7121, 2.9342, 3.6687, 4.6491, 4.5242, 3.2389], device='cuda:1'), covar=tensor([0.0350, 0.1531, 0.1090, 0.1357, 0.0930, 0.0810, 0.0457, 0.1219], device='cuda:1'), in_proj_covar=tensor([0.0242, 0.0243, 0.0274, 0.0214, 0.0263, 0.0358, 0.0257, 0.0229], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 20:54:37,625 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.274e+02 1.944e+02 2.387e+02 2.834e+02 5.542e+02, threshold=4.773e+02, percent-clipped=3.0 2023-03-08 20:54:44,322 INFO [train2.py:809] (1/4) Epoch 20, batch 1850, loss[ctc_loss=0.0795, att_loss=0.2514, loss=0.217, over 17127.00 frames. utt_duration=1225 frames, utt_pad_proportion=0.01478, over 56.00 utterances.], tot_loss[ctc_loss=0.07579, att_loss=0.237, loss=0.2048, over 3291146.34 frames. utt_duration=1267 frames, utt_pad_proportion=0.0462, over 10401.64 utterances.], batch size: 56, lr: 5.40e-03, grad_scale: 8.0 2023-03-08 20:54:45,978 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=77543.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 20:55:23,577 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77567.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 20:56:03,932 INFO [train2.py:809] (1/4) Epoch 20, batch 1900, loss[ctc_loss=0.06353, att_loss=0.2211, loss=0.1896, over 15771.00 frames. utt_duration=1662 frames, utt_pad_proportion=0.008479, over 38.00 utterances.], tot_loss[ctc_loss=0.07576, att_loss=0.2374, loss=0.205, over 3292964.30 frames. utt_duration=1277 frames, utt_pad_proportion=0.04419, over 10330.79 utterances.], batch size: 38, lr: 5.40e-03, grad_scale: 8.0 2023-03-08 20:56:42,723 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77616.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 20:56:46,435 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-03-08 20:57:03,873 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77629.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 20:57:10,172 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77633.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 20:57:15,023 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6480, 4.7997, 4.8184, 4.7673, 5.3755, 4.6448, 4.7987, 2.7217], device='cuda:1'), covar=tensor([0.0232, 0.0279, 0.0251, 0.0299, 0.0700, 0.0208, 0.0266, 0.1813], device='cuda:1'), in_proj_covar=tensor([0.0155, 0.0178, 0.0179, 0.0195, 0.0362, 0.0149, 0.0167, 0.0212], 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') 2023-03-08 20:57:17,665 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.149e+02 1.953e+02 2.465e+02 3.107e+02 2.003e+03, threshold=4.931e+02, percent-clipped=5.0 2023-03-08 20:57:24,505 INFO [train2.py:809] (1/4) Epoch 20, batch 1950, loss[ctc_loss=0.04504, att_loss=0.1999, loss=0.1689, over 15627.00 frames. utt_duration=1691 frames, utt_pad_proportion=0.009405, over 37.00 utterances.], tot_loss[ctc_loss=0.07509, att_loss=0.2366, loss=0.2043, over 3278559.71 frames. utt_duration=1262 frames, utt_pad_proportion=0.05078, over 10405.70 utterances.], batch size: 37, lr: 5.39e-03, grad_scale: 8.0 2023-03-08 20:58:19,908 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77677.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 20:58:19,938 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77677.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 20:58:44,129 INFO [train2.py:809] (1/4) Epoch 20, batch 2000, loss[ctc_loss=0.06848, att_loss=0.2194, loss=0.1892, over 15506.00 frames. utt_duration=1724 frames, utt_pad_proportion=0.007784, over 36.00 utterances.], tot_loss[ctc_loss=0.07553, att_loss=0.2371, loss=0.2048, over 3275440.33 frames. utt_duration=1250 frames, utt_pad_proportion=0.0518, over 10492.29 utterances.], batch size: 36, lr: 5.39e-03, grad_scale: 8.0 2023-03-08 20:59:03,213 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.3831, 2.8858, 3.4205, 4.3401, 3.7891, 3.7894, 2.8098, 2.1757], device='cuda:1'), covar=tensor([0.0707, 0.2089, 0.0884, 0.0605, 0.0881, 0.0515, 0.1667, 0.2469], device='cuda:1'), in_proj_covar=tensor([0.0180, 0.0219, 0.0190, 0.0218, 0.0221, 0.0176, 0.0205, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 20:59:57,585 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.253e+02 2.073e+02 2.649e+02 3.244e+02 1.218e+03, threshold=5.298e+02, percent-clipped=5.0 2023-03-08 20:59:57,950 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77738.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:00:03,843 INFO [train2.py:809] (1/4) Epoch 20, batch 2050, loss[ctc_loss=0.07381, att_loss=0.2388, loss=0.2058, over 17321.00 frames. utt_duration=1175 frames, utt_pad_proportion=0.02295, over 59.00 utterances.], tot_loss[ctc_loss=0.07618, att_loss=0.2377, loss=0.2054, over 3277256.11 frames. utt_duration=1251 frames, utt_pad_proportion=0.05174, over 10489.24 utterances.], batch size: 59, lr: 5.39e-03, grad_scale: 8.0 2023-03-08 21:00:17,890 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77751.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 21:00:25,700 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2362, 5.1660, 5.0681, 2.5272, 2.0718, 2.9237, 2.7293, 4.0121], device='cuda:1'), covar=tensor([0.0679, 0.0316, 0.0248, 0.4341, 0.5328, 0.2391, 0.3042, 0.1644], device='cuda:1'), in_proj_covar=tensor([0.0352, 0.0269, 0.0264, 0.0243, 0.0344, 0.0335, 0.0251, 0.0367], device='cuda:1'), out_proj_covar=tensor([1.4991e-04, 9.9484e-05, 1.1213e-04, 1.0486e-04, 1.4437e-04, 1.3116e-04, 1.0051e-04, 1.4885e-04], device='cuda:1') 2023-03-08 21:01:23,889 INFO [train2.py:809] (1/4) Epoch 20, batch 2100, loss[ctc_loss=0.07425, att_loss=0.2297, loss=0.1986, over 15642.00 frames. utt_duration=1693 frames, utt_pad_proportion=0.009035, over 37.00 utterances.], tot_loss[ctc_loss=0.07531, att_loss=0.2363, loss=0.2041, over 3267596.40 frames. utt_duration=1281 frames, utt_pad_proportion=0.04791, over 10215.54 utterances.], batch size: 37, lr: 5.39e-03, grad_scale: 8.0 2023-03-08 21:01:34,647 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=77799.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:02:36,993 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.374e+02 1.913e+02 2.227e+02 2.675e+02 5.227e+02, threshold=4.453e+02, percent-clipped=0.0 2023-03-08 21:02:43,922 INFO [train2.py:809] (1/4) Epoch 20, batch 2150, loss[ctc_loss=0.09294, att_loss=0.2485, loss=0.2174, over 16972.00 frames. utt_duration=687 frames, utt_pad_proportion=0.1391, over 99.00 utterances.], tot_loss[ctc_loss=0.0755, att_loss=0.2364, loss=0.2042, over 3264718.42 frames. utt_duration=1272 frames, utt_pad_proportion=0.05018, over 10280.39 utterances.], batch size: 99, lr: 5.39e-03, grad_scale: 8.0 2023-03-08 21:02:47,147 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9871, 5.3160, 4.8524, 5.3324, 4.7346, 4.9722, 5.4233, 5.2080], device='cuda:1'), covar=tensor([0.0544, 0.0288, 0.0791, 0.0347, 0.0400, 0.0224, 0.0277, 0.0214], device='cuda:1'), in_proj_covar=tensor([0.0383, 0.0315, 0.0362, 0.0338, 0.0313, 0.0235, 0.0296, 0.0280], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 21:03:24,192 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77867.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:03:54,831 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.8617, 5.2245, 5.4551, 5.3168, 5.4070, 5.8347, 5.1662, 5.9396], device='cuda:1'), covar=tensor([0.0704, 0.0714, 0.0784, 0.1230, 0.1809, 0.0954, 0.0784, 0.0716], device='cuda:1'), in_proj_covar=tensor([0.0875, 0.0511, 0.0590, 0.0658, 0.0868, 0.0616, 0.0486, 0.0601], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 21:04:04,229 INFO [train2.py:809] (1/4) Epoch 20, batch 2200, loss[ctc_loss=0.07838, att_loss=0.2454, loss=0.212, over 16319.00 frames. utt_duration=1452 frames, utt_pad_proportion=0.006901, over 45.00 utterances.], tot_loss[ctc_loss=0.07561, att_loss=0.2364, loss=0.2043, over 3262361.60 frames. utt_duration=1267 frames, utt_pad_proportion=0.05079, over 10314.64 utterances.], batch size: 45, lr: 5.39e-03, grad_scale: 8.0 2023-03-08 21:04:41,057 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=77915.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:05:03,626 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77929.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:05:10,166 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77933.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:05:17,370 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.298e+02 1.939e+02 2.291e+02 2.587e+02 5.643e+02, threshold=4.582e+02, percent-clipped=3.0 2023-03-08 21:05:23,449 INFO [train2.py:809] (1/4) Epoch 20, batch 2250, loss[ctc_loss=0.07659, att_loss=0.2496, loss=0.215, over 16874.00 frames. utt_duration=1379 frames, utt_pad_proportion=0.007258, over 49.00 utterances.], tot_loss[ctc_loss=0.07616, att_loss=0.2374, loss=0.2052, over 3277252.49 frames. utt_duration=1261 frames, utt_pad_proportion=0.04888, over 10409.74 utterances.], batch size: 49, lr: 5.38e-03, grad_scale: 8.0 2023-03-08 21:05:59,736 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-03-08 21:06:11,706 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77972.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:06:19,581 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=77977.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:06:26,479 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=77981.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:06:43,281 INFO [train2.py:809] (1/4) Epoch 20, batch 2300, loss[ctc_loss=0.09428, att_loss=0.2498, loss=0.2187, over 17379.00 frames. utt_duration=1009 frames, utt_pad_proportion=0.04796, over 69.00 utterances.], tot_loss[ctc_loss=0.07555, att_loss=0.2373, loss=0.205, over 3274664.86 frames. utt_duration=1275 frames, utt_pad_proportion=0.04643, over 10283.26 utterances.], batch size: 69, lr: 5.38e-03, grad_scale: 8.0 2023-03-08 21:07:51,971 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0150, 5.2938, 5.2288, 5.1907, 5.3335, 5.2355, 5.0033, 4.7894], device='cuda:1'), covar=tensor([0.1046, 0.0545, 0.0343, 0.0622, 0.0314, 0.0334, 0.0365, 0.0353], device='cuda:1'), in_proj_covar=tensor([0.0513, 0.0350, 0.0333, 0.0343, 0.0402, 0.0421, 0.0346, 0.0385], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0003, 0.0004], device='cuda:1') 2023-03-08 21:07:53,484 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78033.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:08:01,054 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.567e+02 2.189e+02 2.492e+02 2.915e+02 6.479e+02, threshold=4.984e+02, percent-clipped=5.0 2023-03-08 21:08:07,071 INFO [train2.py:809] (1/4) Epoch 20, batch 2350, loss[ctc_loss=0.04751, att_loss=0.2121, loss=0.1792, over 15359.00 frames. utt_duration=1757 frames, utt_pad_proportion=0.01199, over 35.00 utterances.], tot_loss[ctc_loss=0.07549, att_loss=0.2371, loss=0.2048, over 3276971.33 frames. utt_duration=1257 frames, utt_pad_proportion=0.04933, over 10442.05 utterances.], batch size: 35, lr: 5.38e-03, grad_scale: 8.0 2023-03-08 21:08:21,600 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4105, 4.7347, 4.6494, 4.6832, 4.7991, 4.4960, 3.2368, 4.7038], device='cuda:1'), covar=tensor([0.0129, 0.0131, 0.0132, 0.0105, 0.0108, 0.0115, 0.0773, 0.0234], device='cuda:1'), in_proj_covar=tensor([0.0089, 0.0084, 0.0106, 0.0066, 0.0072, 0.0082, 0.0101, 0.0105], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 21:09:26,277 INFO [train2.py:809] (1/4) Epoch 20, batch 2400, loss[ctc_loss=0.1262, att_loss=0.2584, loss=0.232, over 13067.00 frames. utt_duration=362.1 frames, utt_pad_proportion=0.3713, over 145.00 utterances.], tot_loss[ctc_loss=0.0763, att_loss=0.2378, loss=0.2055, over 3274725.10 frames. utt_duration=1216 frames, utt_pad_proportion=0.0615, over 10787.62 utterances.], batch size: 145, lr: 5.38e-03, grad_scale: 8.0 2023-03-08 21:10:10,175 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0857, 5.0102, 4.7865, 3.1743, 4.7334, 4.6497, 4.4297, 3.0428], device='cuda:1'), covar=tensor([0.0121, 0.0107, 0.0256, 0.0917, 0.0117, 0.0204, 0.0282, 0.1255], device='cuda:1'), in_proj_covar=tensor([0.0073, 0.0101, 0.0102, 0.0110, 0.0084, 0.0110, 0.0098, 0.0104], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 21:10:39,633 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.436e+02 2.004e+02 2.481e+02 3.189e+02 5.932e+02, threshold=4.963e+02, percent-clipped=2.0 2023-03-08 21:10:46,378 INFO [train2.py:809] (1/4) Epoch 20, batch 2450, loss[ctc_loss=0.06988, att_loss=0.2139, loss=0.1851, over 15510.00 frames. utt_duration=1725 frames, utt_pad_proportion=0.008162, over 36.00 utterances.], tot_loss[ctc_loss=0.07607, att_loss=0.2377, loss=0.2054, over 3272258.52 frames. utt_duration=1237 frames, utt_pad_proportion=0.0571, over 10591.21 utterances.], batch size: 36, lr: 5.38e-03, grad_scale: 8.0 2023-03-08 21:12:06,744 INFO [train2.py:809] (1/4) Epoch 20, batch 2500, loss[ctc_loss=0.07483, att_loss=0.2387, loss=0.2059, over 16629.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.005349, over 47.00 utterances.], tot_loss[ctc_loss=0.07639, att_loss=0.2378, loss=0.2055, over 3279850.73 frames. utt_duration=1244 frames, utt_pad_proportion=0.05351, over 10557.28 utterances.], batch size: 47, lr: 5.38e-03, grad_scale: 8.0 2023-03-08 21:12:26,873 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5987, 3.1066, 3.6003, 3.0983, 3.5468, 4.6478, 4.4186, 3.3560], device='cuda:1'), covar=tensor([0.0311, 0.1587, 0.1287, 0.1280, 0.1081, 0.0768, 0.0590, 0.1207], device='cuda:1'), in_proj_covar=tensor([0.0245, 0.0242, 0.0279, 0.0216, 0.0265, 0.0359, 0.0259, 0.0230], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 21:13:12,136 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78232.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:13:21,003 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.184e+02 1.956e+02 2.237e+02 2.783e+02 5.149e+02, threshold=4.474e+02, percent-clipped=1.0 2023-03-08 21:13:27,757 INFO [train2.py:809] (1/4) Epoch 20, batch 2550, loss[ctc_loss=0.04952, att_loss=0.2039, loss=0.173, over 15750.00 frames. utt_duration=1659 frames, utt_pad_proportion=0.008195, over 38.00 utterances.], tot_loss[ctc_loss=0.0751, att_loss=0.2363, loss=0.204, over 3270457.99 frames. utt_duration=1270 frames, utt_pad_proportion=0.04984, over 10314.64 utterances.], batch size: 38, lr: 5.37e-03, grad_scale: 8.0 2023-03-08 21:14:15,812 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78272.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:14:22,664 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6736, 4.5284, 4.5969, 4.6793, 5.2307, 4.3765, 4.5126, 2.5109], device='cuda:1'), covar=tensor([0.0234, 0.0369, 0.0313, 0.0288, 0.0806, 0.0296, 0.0417, 0.1833], device='cuda:1'), in_proj_covar=tensor([0.0158, 0.0181, 0.0182, 0.0199, 0.0366, 0.0152, 0.0171, 0.0216], 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') 2023-03-08 21:14:47,851 INFO [train2.py:809] (1/4) Epoch 20, batch 2600, loss[ctc_loss=0.0765, att_loss=0.2245, loss=0.1949, over 16178.00 frames. utt_duration=1580 frames, utt_pad_proportion=0.006519, over 41.00 utterances.], tot_loss[ctc_loss=0.07501, att_loss=0.236, loss=0.2038, over 3269137.36 frames. utt_duration=1285 frames, utt_pad_proportion=0.04714, over 10184.95 utterances.], batch size: 41, lr: 5.37e-03, grad_scale: 8.0 2023-03-08 21:14:49,782 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1357, 5.0602, 4.8771, 3.0806, 4.8022, 4.7267, 4.4238, 2.7474], device='cuda:1'), covar=tensor([0.0109, 0.0108, 0.0251, 0.1009, 0.0112, 0.0185, 0.0309, 0.1461], device='cuda:1'), in_proj_covar=tensor([0.0073, 0.0101, 0.0101, 0.0110, 0.0084, 0.0110, 0.0098, 0.0104], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 21:14:49,883 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78293.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 21:15:32,434 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=78320.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:15:53,574 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78333.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:16:00,907 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.091e+02 1.986e+02 2.327e+02 2.880e+02 6.602e+02, threshold=4.653e+02, percent-clipped=4.0 2023-03-08 21:16:04,262 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-03-08 21:16:07,836 INFO [train2.py:809] (1/4) Epoch 20, batch 2650, loss[ctc_loss=0.04866, att_loss=0.2145, loss=0.1813, over 16000.00 frames. utt_duration=1602 frames, utt_pad_proportion=0.007667, over 40.00 utterances.], tot_loss[ctc_loss=0.07534, att_loss=0.236, loss=0.2039, over 3264679.03 frames. utt_duration=1266 frames, utt_pad_proportion=0.05132, over 10330.67 utterances.], batch size: 40, lr: 5.37e-03, grad_scale: 8.0 2023-03-08 21:16:54,885 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0388, 4.5039, 4.5666, 4.7551, 2.7315, 4.7022, 2.7369, 1.6233], device='cuda:1'), covar=tensor([0.0476, 0.0215, 0.0566, 0.0190, 0.1610, 0.0174, 0.1402, 0.1766], device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0159, 0.0261, 0.0154, 0.0220, 0.0139, 0.0230, 0.0204], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 21:17:10,072 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=78381.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:17:23,911 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-03-08 21:17:27,557 INFO [train2.py:809] (1/4) Epoch 20, batch 2700, loss[ctc_loss=0.08678, att_loss=0.2513, loss=0.2184, over 16960.00 frames. utt_duration=1358 frames, utt_pad_proportion=0.007295, over 50.00 utterances.], tot_loss[ctc_loss=0.07572, att_loss=0.2363, loss=0.2042, over 3265527.10 frames. utt_duration=1261 frames, utt_pad_proportion=0.05349, over 10373.95 utterances.], batch size: 50, lr: 5.37e-03, grad_scale: 8.0 2023-03-08 21:17:30,034 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.86 vs. limit=2.0 2023-03-08 21:18:20,072 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.28 vs. limit=5.0 2023-03-08 21:18:41,453 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.385e+02 1.853e+02 2.270e+02 2.670e+02 9.199e+02, threshold=4.540e+02, percent-clipped=5.0 2023-03-08 21:18:47,883 INFO [train2.py:809] (1/4) Epoch 20, batch 2750, loss[ctc_loss=0.06035, att_loss=0.2338, loss=0.1991, over 16528.00 frames. utt_duration=1471 frames, utt_pad_proportion=0.005514, over 45.00 utterances.], tot_loss[ctc_loss=0.07546, att_loss=0.2361, loss=0.204, over 3266769.99 frames. utt_duration=1268 frames, utt_pad_proportion=0.05154, over 10321.41 utterances.], batch size: 45, lr: 5.37e-03, grad_scale: 8.0 2023-03-08 21:18:55,095 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=2.88 vs. limit=5.0 2023-03-08 21:19:17,762 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78460.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:20:08,383 INFO [train2.py:809] (1/4) Epoch 20, batch 2800, loss[ctc_loss=0.06441, att_loss=0.2197, loss=0.1886, over 12716.00 frames. utt_duration=1818 frames, utt_pad_proportion=0.1142, over 28.00 utterances.], tot_loss[ctc_loss=0.0756, att_loss=0.237, loss=0.2047, over 3273203.05 frames. utt_duration=1272 frames, utt_pad_proportion=0.04841, over 10307.69 utterances.], batch size: 28, lr: 5.37e-03, grad_scale: 8.0 2023-03-08 21:20:16,251 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-03-08 21:20:41,133 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.0809, 3.7570, 3.7511, 3.3149, 3.7874, 3.7970, 3.7502, 2.9626], device='cuda:1'), covar=tensor([0.0999, 0.1903, 0.2429, 0.3152, 0.0929, 0.3887, 0.1202, 0.3574], device='cuda:1'), in_proj_covar=tensor([0.0168, 0.0184, 0.0195, 0.0251, 0.0155, 0.0256, 0.0173, 0.0218], 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') 2023-03-08 21:20:55,811 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78521.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:21:21,798 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.455e+02 2.103e+02 2.542e+02 3.045e+02 6.490e+02, threshold=5.083e+02, percent-clipped=4.0 2023-03-08 21:21:28,205 INFO [train2.py:809] (1/4) Epoch 20, batch 2850, loss[ctc_loss=0.09276, att_loss=0.2618, loss=0.228, over 16812.00 frames. utt_duration=680.8 frames, utt_pad_proportion=0.1405, over 99.00 utterances.], tot_loss[ctc_loss=0.07722, att_loss=0.2384, loss=0.2062, over 3266777.83 frames. utt_duration=1217 frames, utt_pad_proportion=0.06371, over 10753.97 utterances.], batch size: 99, lr: 5.36e-03, grad_scale: 16.0 2023-03-08 21:21:51,214 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78556.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:22:09,557 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-03-08 21:22:42,487 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78588.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 21:22:48,504 INFO [train2.py:809] (1/4) Epoch 20, batch 2900, loss[ctc_loss=0.0846, att_loss=0.225, loss=0.1969, over 15957.00 frames. utt_duration=1558 frames, utt_pad_proportion=0.00616, over 41.00 utterances.], tot_loss[ctc_loss=0.07691, att_loss=0.2375, loss=0.2054, over 3260840.47 frames. utt_duration=1237 frames, utt_pad_proportion=0.06026, over 10560.07 utterances.], batch size: 41, lr: 5.36e-03, grad_scale: 16.0 2023-03-08 21:22:53,143 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.75 vs. limit=5.0 2023-03-08 21:23:26,673 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.88 vs. limit=5.0 2023-03-08 21:23:30,344 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78617.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:24:02,855 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.232e+02 1.965e+02 2.501e+02 3.006e+02 5.766e+02, threshold=5.002e+02, percent-clipped=1.0 2023-03-08 21:24:08,864 INFO [train2.py:809] (1/4) Epoch 20, batch 2950, loss[ctc_loss=0.07463, att_loss=0.2316, loss=0.2002, over 16396.00 frames. utt_duration=1492 frames, utt_pad_proportion=0.007167, over 44.00 utterances.], tot_loss[ctc_loss=0.07573, att_loss=0.2366, loss=0.2044, over 3262474.53 frames. utt_duration=1264 frames, utt_pad_proportion=0.05264, over 10339.33 utterances.], batch size: 44, lr: 5.36e-03, grad_scale: 16.0 2023-03-08 21:24:17,226 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.1722, 5.4511, 5.7278, 5.5745, 5.6581, 6.1066, 5.2881, 6.1698], device='cuda:1'), covar=tensor([0.0798, 0.0720, 0.0778, 0.1156, 0.1846, 0.0909, 0.0608, 0.0757], device='cuda:1'), in_proj_covar=tensor([0.0869, 0.0508, 0.0592, 0.0656, 0.0866, 0.0610, 0.0481, 0.0601], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 21:24:45,060 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1044, 5.1414, 4.9145, 2.5432, 2.0047, 2.8511, 2.6251, 3.8281], device='cuda:1'), covar=tensor([0.0650, 0.0291, 0.0264, 0.4128, 0.5393, 0.2411, 0.3188, 0.1718], device='cuda:1'), in_proj_covar=tensor([0.0350, 0.0268, 0.0263, 0.0241, 0.0341, 0.0332, 0.0250, 0.0363], device='cuda:1'), out_proj_covar=tensor([1.4863e-04, 9.9626e-05, 1.1182e-04, 1.0349e-04, 1.4295e-04, 1.2996e-04, 1.0049e-04, 1.4743e-04], device='cuda:1') 2023-03-08 21:25:28,470 INFO [train2.py:809] (1/4) Epoch 20, batch 3000, loss[ctc_loss=0.06272, att_loss=0.2342, loss=0.1999, over 16885.00 frames. utt_duration=1380 frames, utt_pad_proportion=0.007444, over 49.00 utterances.], tot_loss[ctc_loss=0.07569, att_loss=0.2367, loss=0.2045, over 3270207.46 frames. utt_duration=1267 frames, utt_pad_proportion=0.05001, over 10334.80 utterances.], batch size: 49, lr: 5.36e-03, grad_scale: 16.0 2023-03-08 21:25:28,471 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-08 21:25:42,049 INFO [train2.py:843] (1/4) Epoch 20, validation: ctc_loss=0.04025, att_loss=0.2341, loss=0.1953, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-08 21:25:42,050 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-08 21:26:41,151 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-03-08 21:26:46,744 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-03-08 21:26:55,739 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.291e+02 1.875e+02 2.186e+02 2.641e+02 4.823e+02, threshold=4.373e+02, percent-clipped=0.0 2023-03-08 21:27:02,056 INFO [train2.py:809] (1/4) Epoch 20, batch 3050, loss[ctc_loss=0.09924, att_loss=0.2604, loss=0.2282, over 17069.00 frames. utt_duration=1314 frames, utt_pad_proportion=0.007925, over 52.00 utterances.], tot_loss[ctc_loss=0.07576, att_loss=0.2371, loss=0.2048, over 3276698.39 frames. utt_duration=1262 frames, utt_pad_proportion=0.05045, over 10399.09 utterances.], batch size: 52, lr: 5.36e-03, grad_scale: 16.0 2023-03-08 21:27:33,411 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.1409, 5.3385, 5.6415, 5.4907, 5.6520, 6.0374, 5.2876, 6.1745], device='cuda:1'), covar=tensor([0.0660, 0.0759, 0.0806, 0.1225, 0.1646, 0.0944, 0.0672, 0.0761], device='cuda:1'), in_proj_covar=tensor([0.0868, 0.0507, 0.0590, 0.0657, 0.0869, 0.0610, 0.0484, 0.0603], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 21:27:51,241 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6541, 2.0839, 4.9833, 3.9704, 2.9959, 4.3092, 4.7196, 4.7339], device='cuda:1'), covar=tensor([0.0168, 0.1734, 0.0125, 0.0821, 0.1606, 0.0194, 0.0113, 0.0179], device='cuda:1'), in_proj_covar=tensor([0.0187, 0.0241, 0.0179, 0.0309, 0.0263, 0.0211, 0.0169, 0.0197], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 21:28:22,269 INFO [train2.py:809] (1/4) Epoch 20, batch 3100, loss[ctc_loss=0.05549, att_loss=0.2119, loss=0.1806, over 16109.00 frames. utt_duration=1536 frames, utt_pad_proportion=0.007264, over 42.00 utterances.], tot_loss[ctc_loss=0.07531, att_loss=0.2368, loss=0.2045, over 3268729.88 frames. utt_duration=1255 frames, utt_pad_proportion=0.0531, over 10434.74 utterances.], batch size: 42, lr: 5.36e-03, grad_scale: 16.0 2023-03-08 21:28:22,631 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6732, 3.0108, 3.7132, 3.2510, 3.6958, 4.7320, 4.5353, 3.4292], device='cuda:1'), covar=tensor([0.0306, 0.1580, 0.1169, 0.1141, 0.0994, 0.0818, 0.0553, 0.1073], device='cuda:1'), in_proj_covar=tensor([0.0241, 0.0241, 0.0277, 0.0214, 0.0262, 0.0357, 0.0256, 0.0228], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 21:29:01,282 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78816.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:29:15,667 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.7594, 6.0058, 5.4391, 5.7613, 5.7149, 5.1030, 5.4451, 5.1903], device='cuda:1'), covar=tensor([0.1187, 0.0844, 0.0881, 0.0772, 0.0984, 0.1403, 0.2034, 0.2294], device='cuda:1'), in_proj_covar=tensor([0.0512, 0.0597, 0.0451, 0.0450, 0.0423, 0.0455, 0.0607, 0.0520], 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') 2023-03-08 21:29:36,348 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.264e+02 1.977e+02 2.476e+02 2.944e+02 6.319e+02, threshold=4.952e+02, percent-clipped=6.0 2023-03-08 21:29:42,606 INFO [train2.py:809] (1/4) Epoch 20, batch 3150, loss[ctc_loss=0.05816, att_loss=0.2219, loss=0.1891, over 16383.00 frames. utt_duration=1491 frames, utt_pad_proportion=0.007924, over 44.00 utterances.], tot_loss[ctc_loss=0.07559, att_loss=0.2372, loss=0.2049, over 3273603.62 frames. utt_duration=1240 frames, utt_pad_proportion=0.05514, over 10576.81 utterances.], batch size: 44, lr: 5.35e-03, grad_scale: 16.0 2023-03-08 21:29:59,081 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.9439, 3.5509, 3.6683, 2.9460, 3.6188, 3.7030, 3.6589, 2.3331], device='cuda:1'), covar=tensor([0.1114, 0.1657, 0.1896, 0.6586, 0.1427, 0.2132, 0.1008, 0.7993], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0182, 0.0193, 0.0248, 0.0154, 0.0254, 0.0172, 0.0215], 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') 2023-03-08 21:30:55,901 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78888.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:31:01,819 INFO [train2.py:809] (1/4) Epoch 20, batch 3200, loss[ctc_loss=0.07163, att_loss=0.2452, loss=0.2105, over 17042.00 frames. utt_duration=1312 frames, utt_pad_proportion=0.00955, over 52.00 utterances.], tot_loss[ctc_loss=0.07517, att_loss=0.237, loss=0.2047, over 3278762.51 frames. utt_duration=1250 frames, utt_pad_proportion=0.05048, over 10505.79 utterances.], batch size: 52, lr: 5.35e-03, grad_scale: 16.0 2023-03-08 21:31:34,108 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0259, 5.3384, 5.6059, 5.4641, 5.4550, 5.9641, 5.2466, 6.0380], device='cuda:1'), covar=tensor([0.0716, 0.0681, 0.0753, 0.1071, 0.1753, 0.0932, 0.0710, 0.0708], device='cuda:1'), in_proj_covar=tensor([0.0860, 0.0504, 0.0588, 0.0654, 0.0861, 0.0608, 0.0481, 0.0599], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 21:31:34,122 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78912.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:32:12,200 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=78936.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:32:15,122 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.285e+02 2.045e+02 2.242e+02 2.704e+02 5.175e+02, threshold=4.484e+02, percent-clipped=1.0 2023-03-08 21:32:17,045 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7586, 3.0666, 3.7462, 3.1778, 3.7393, 4.8188, 4.6276, 3.3606], device='cuda:1'), covar=tensor([0.0331, 0.1703, 0.1148, 0.1374, 0.1075, 0.0735, 0.0522, 0.1335], device='cuda:1'), in_proj_covar=tensor([0.0245, 0.0245, 0.0281, 0.0219, 0.0267, 0.0363, 0.0260, 0.0232], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 21:32:21,327 INFO [train2.py:809] (1/4) Epoch 20, batch 3250, loss[ctc_loss=0.06002, att_loss=0.2108, loss=0.1807, over 15349.00 frames. utt_duration=1755 frames, utt_pad_proportion=0.01212, over 35.00 utterances.], tot_loss[ctc_loss=0.07543, att_loss=0.2369, loss=0.2046, over 3273025.45 frames. utt_duration=1250 frames, utt_pad_proportion=0.0529, over 10488.90 utterances.], batch size: 35, lr: 5.35e-03, grad_scale: 16.0 2023-03-08 21:32:58,073 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78965.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:33:40,329 INFO [train2.py:809] (1/4) Epoch 20, batch 3300, loss[ctc_loss=0.06502, att_loss=0.2264, loss=0.1941, over 16277.00 frames. utt_duration=1516 frames, utt_pad_proportion=0.007371, over 43.00 utterances.], tot_loss[ctc_loss=0.07543, att_loss=0.2365, loss=0.2043, over 3270136.17 frames. utt_duration=1227 frames, utt_pad_proportion=0.05946, over 10673.86 utterances.], batch size: 43, lr: 5.35e-03, grad_scale: 16.0 2023-03-08 21:34:35,286 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79026.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:34:54,086 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.427e+02 1.921e+02 2.242e+02 2.810e+02 6.510e+02, threshold=4.483e+02, percent-clipped=2.0 2023-03-08 21:35:00,359 INFO [train2.py:809] (1/4) Epoch 20, batch 3350, loss[ctc_loss=0.05806, att_loss=0.2108, loss=0.1802, over 15658.00 frames. utt_duration=1694 frames, utt_pad_proportion=0.008149, over 37.00 utterances.], tot_loss[ctc_loss=0.07611, att_loss=0.2368, loss=0.2046, over 3263053.22 frames. utt_duration=1218 frames, utt_pad_proportion=0.06457, over 10729.76 utterances.], batch size: 37, lr: 5.35e-03, grad_scale: 16.0 2023-03-08 21:35:03,015 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.01 vs. limit=5.0 2023-03-08 21:36:00,697 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1611, 4.5019, 4.4770, 4.6339, 3.0630, 4.5778, 2.8906, 1.7049], device='cuda:1'), covar=tensor([0.0403, 0.0258, 0.0603, 0.0239, 0.1358, 0.0185, 0.1375, 0.1790], device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0162, 0.0261, 0.0156, 0.0221, 0.0141, 0.0231, 0.0204], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:1') 2023-03-08 21:36:21,445 INFO [train2.py:809] (1/4) Epoch 20, batch 3400, loss[ctc_loss=0.06198, att_loss=0.2255, loss=0.1928, over 16278.00 frames. utt_duration=1515 frames, utt_pad_proportion=0.007645, over 43.00 utterances.], tot_loss[ctc_loss=0.07658, att_loss=0.2376, loss=0.2054, over 3265330.05 frames. utt_duration=1193 frames, utt_pad_proportion=0.06927, over 10959.82 utterances.], batch size: 43, lr: 5.35e-03, grad_scale: 16.0 2023-03-08 21:36:44,152 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.8975, 2.3917, 2.5857, 2.7345, 2.7689, 2.7998, 2.6085, 3.2838], device='cuda:1'), covar=tensor([0.1732, 0.2997, 0.2180, 0.1666, 0.2162, 0.1505, 0.2082, 0.1117], device='cuda:1'), in_proj_covar=tensor([0.0114, 0.0122, 0.0116, 0.0106, 0.0120, 0.0103, 0.0124, 0.0096], device='cuda:1'), out_proj_covar=tensor([8.5841e-05, 9.3953e-05, 9.2470e-05, 8.2572e-05, 8.8554e-05, 8.2454e-05, 9.3530e-05, 7.6745e-05], device='cuda:1') 2023-03-08 21:36:59,688 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79116.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:37:34,909 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.410e+02 1.976e+02 2.387e+02 2.813e+02 1.211e+03, threshold=4.774e+02, percent-clipped=8.0 2023-03-08 21:37:41,720 INFO [train2.py:809] (1/4) Epoch 20, batch 3450, loss[ctc_loss=0.06928, att_loss=0.2269, loss=0.1954, over 15944.00 frames. utt_duration=1557 frames, utt_pad_proportion=0.006969, over 41.00 utterances.], tot_loss[ctc_loss=0.07594, att_loss=0.2371, loss=0.2049, over 3262921.70 frames. utt_duration=1189 frames, utt_pad_proportion=0.07112, over 10992.09 utterances.], batch size: 41, lr: 5.34e-03, grad_scale: 16.0 2023-03-08 21:37:53,815 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-03-08 21:38:00,997 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5016, 2.5322, 5.0181, 3.9326, 2.9005, 4.2005, 4.8219, 4.6267], device='cuda:1'), covar=tensor([0.0255, 0.1745, 0.0151, 0.0865, 0.1850, 0.0250, 0.0135, 0.0245], device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0242, 0.0180, 0.0309, 0.0264, 0.0211, 0.0170, 0.0200], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 21:38:16,624 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=79164.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:39:01,285 INFO [train2.py:809] (1/4) Epoch 20, batch 3500, loss[ctc_loss=0.06266, att_loss=0.2087, loss=0.1795, over 15864.00 frames. utt_duration=1629 frames, utt_pad_proportion=0.01056, over 39.00 utterances.], tot_loss[ctc_loss=0.07671, att_loss=0.2382, loss=0.2059, over 3267482.08 frames. utt_duration=1196 frames, utt_pad_proportion=0.06808, over 10937.62 utterances.], batch size: 39, lr: 5.34e-03, grad_scale: 16.0 2023-03-08 21:39:33,905 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79212.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:39:35,351 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5816, 4.8651, 4.4309, 4.8577, 4.3719, 4.5406, 4.9209, 4.7806], device='cuda:1'), covar=tensor([0.0537, 0.0309, 0.0721, 0.0328, 0.0413, 0.0344, 0.0241, 0.0192], device='cuda:1'), in_proj_covar=tensor([0.0383, 0.0316, 0.0361, 0.0340, 0.0315, 0.0236, 0.0298, 0.0279], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 21:39:49,348 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-03-08 21:39:52,510 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-03-08 21:39:56,161 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9462, 3.9750, 4.0200, 4.0859, 4.0921, 4.0993, 3.8903, 3.7568], device='cuda:1'), covar=tensor([0.0958, 0.1002, 0.0981, 0.0556, 0.0388, 0.0475, 0.0485, 0.0414], device='cuda:1'), in_proj_covar=tensor([0.0524, 0.0357, 0.0343, 0.0353, 0.0414, 0.0428, 0.0353, 0.0394], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0003, 0.0004], device='cuda:1') 2023-03-08 21:40:15,350 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.171e+02 1.933e+02 2.262e+02 2.788e+02 9.048e+02, threshold=4.523e+02, percent-clipped=2.0 2023-03-08 21:40:18,124 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2767, 4.3334, 4.4713, 4.4685, 4.9225, 4.4390, 4.4162, 2.5753], device='cuda:1'), covar=tensor([0.0279, 0.0385, 0.0327, 0.0358, 0.0736, 0.0224, 0.0335, 0.1700], device='cuda:1'), in_proj_covar=tensor([0.0159, 0.0183, 0.0185, 0.0201, 0.0366, 0.0153, 0.0172, 0.0216], 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') 2023-03-08 21:40:22,330 INFO [train2.py:809] (1/4) Epoch 20, batch 3550, loss[ctc_loss=0.05653, att_loss=0.2323, loss=0.1972, over 16469.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.00588, over 46.00 utterances.], tot_loss[ctc_loss=0.07644, att_loss=0.2381, loss=0.2057, over 3270751.55 frames. utt_duration=1195 frames, utt_pad_proportion=0.06845, over 10963.68 utterances.], batch size: 46, lr: 5.34e-03, grad_scale: 16.0 2023-03-08 21:40:27,628 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.84 vs. limit=2.0 2023-03-08 21:40:51,171 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=79260.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:41:31,927 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79285.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 21:41:42,775 INFO [train2.py:809] (1/4) Epoch 20, batch 3600, loss[ctc_loss=0.0677, att_loss=0.2118, loss=0.183, over 14574.00 frames. utt_duration=1823 frames, utt_pad_proportion=0.04135, over 32.00 utterances.], tot_loss[ctc_loss=0.07544, att_loss=0.237, loss=0.2047, over 3270488.13 frames. utt_duration=1228 frames, utt_pad_proportion=0.05982, over 10665.78 utterances.], batch size: 32, lr: 5.34e-03, grad_scale: 16.0 2023-03-08 21:42:02,405 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0467, 5.3868, 4.9241, 5.3940, 4.8263, 5.0164, 5.4666, 5.2844], device='cuda:1'), covar=tensor([0.0487, 0.0284, 0.0682, 0.0277, 0.0392, 0.0260, 0.0220, 0.0162], device='cuda:1'), in_proj_covar=tensor([0.0383, 0.0317, 0.0361, 0.0340, 0.0315, 0.0236, 0.0299, 0.0279], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 21:42:13,380 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.4679, 3.1636, 3.6112, 4.5677, 4.0004, 3.9588, 3.0873, 2.2565], device='cuda:1'), covar=tensor([0.0613, 0.1623, 0.0745, 0.0468, 0.0737, 0.0496, 0.1305, 0.2040], device='cuda:1'), in_proj_covar=tensor([0.0180, 0.0214, 0.0189, 0.0217, 0.0219, 0.0177, 0.0202, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 21:42:16,499 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9016, 5.2407, 4.7282, 5.2579, 4.6440, 4.9348, 5.3208, 5.1500], device='cuda:1'), covar=tensor([0.0594, 0.0297, 0.0813, 0.0326, 0.0450, 0.0278, 0.0250, 0.0184], device='cuda:1'), in_proj_covar=tensor([0.0383, 0.0317, 0.0361, 0.0340, 0.0315, 0.0236, 0.0299, 0.0279], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 21:42:29,067 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79321.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:42:48,943 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79333.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:42:56,062 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.246e+02 1.889e+02 2.390e+02 3.056e+02 6.577e+02, threshold=4.779e+02, percent-clipped=3.0 2023-03-08 21:43:03,008 INFO [train2.py:809] (1/4) Epoch 20, batch 3650, loss[ctc_loss=0.05234, att_loss=0.2125, loss=0.1805, over 15635.00 frames. utt_duration=1692 frames, utt_pad_proportion=0.00893, over 37.00 utterances.], tot_loss[ctc_loss=0.07489, att_loss=0.2365, loss=0.2042, over 3267971.62 frames. utt_duration=1251 frames, utt_pad_proportion=0.05486, over 10459.21 utterances.], batch size: 37, lr: 5.34e-03, grad_scale: 16.0 2023-03-08 21:43:09,372 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79346.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 21:43:36,966 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79364.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:43:54,661 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79375.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 21:44:11,936 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8713, 5.1685, 5.0723, 5.0692, 5.1037, 5.1251, 4.8316, 4.6228], device='cuda:1'), covar=tensor([0.0973, 0.0475, 0.0320, 0.0527, 0.0313, 0.0330, 0.0374, 0.0356], device='cuda:1'), in_proj_covar=tensor([0.0524, 0.0355, 0.0343, 0.0353, 0.0415, 0.0428, 0.0353, 0.0393], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0003, 0.0004], device='cuda:1') 2023-03-08 21:44:21,502 INFO [train2.py:809] (1/4) Epoch 20, batch 3700, loss[ctc_loss=0.06318, att_loss=0.2145, loss=0.1842, over 15991.00 frames. utt_duration=1601 frames, utt_pad_proportion=0.008287, over 40.00 utterances.], tot_loss[ctc_loss=0.07471, att_loss=0.2359, loss=0.2037, over 3267199.96 frames. utt_duration=1274 frames, utt_pad_proportion=0.04835, over 10267.91 utterances.], batch size: 40, lr: 5.34e-03, grad_scale: 16.0 2023-03-08 21:44:24,961 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79394.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:45:14,867 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79425.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:45:25,796 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6829, 4.9873, 4.8011, 4.8034, 5.0091, 4.6261, 3.4010, 4.8661], device='cuda:1'), covar=tensor([0.0111, 0.0112, 0.0132, 0.0111, 0.0101, 0.0133, 0.0734, 0.0255], device='cuda:1'), in_proj_covar=tensor([0.0091, 0.0086, 0.0109, 0.0069, 0.0074, 0.0085, 0.0103, 0.0107], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 21:45:32,687 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79436.0, num_to_drop=1, layers_to_drop={3} 2023-03-08 21:45:35,344 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.270e+02 1.959e+02 2.392e+02 2.997e+02 5.823e+02, threshold=4.784e+02, percent-clipped=2.0 2023-03-08 21:45:41,563 INFO [train2.py:809] (1/4) Epoch 20, batch 3750, loss[ctc_loss=0.06251, att_loss=0.225, loss=0.1925, over 16181.00 frames. utt_duration=1580 frames, utt_pad_proportion=0.005983, over 41.00 utterances.], tot_loss[ctc_loss=0.07604, att_loss=0.2366, loss=0.2045, over 3258611.96 frames. utt_duration=1220 frames, utt_pad_proportion=0.06548, over 10700.64 utterances.], batch size: 41, lr: 5.33e-03, grad_scale: 16.0 2023-03-08 21:46:34,962 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.15 vs. limit=5.0 2023-03-08 21:46:47,689 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1079, 5.0930, 4.8425, 3.0913, 4.8028, 4.6389, 4.4288, 2.9939], device='cuda:1'), covar=tensor([0.0080, 0.0078, 0.0245, 0.0885, 0.0098, 0.0207, 0.0253, 0.1194], device='cuda:1'), in_proj_covar=tensor([0.0073, 0.0101, 0.0101, 0.0109, 0.0083, 0.0110, 0.0098, 0.0104], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 21:47:01,110 INFO [train2.py:809] (1/4) Epoch 20, batch 3800, loss[ctc_loss=0.09085, att_loss=0.2593, loss=0.2256, over 17298.00 frames. utt_duration=1174 frames, utt_pad_proportion=0.02388, over 59.00 utterances.], tot_loss[ctc_loss=0.07573, att_loss=0.2364, loss=0.2042, over 3264765.80 frames. utt_duration=1220 frames, utt_pad_proportion=0.06272, over 10715.69 utterances.], batch size: 59, lr: 5.33e-03, grad_scale: 16.0 2023-03-08 21:47:03,045 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5905, 3.1115, 3.6993, 3.2755, 3.6682, 4.7243, 4.5299, 3.4074], device='cuda:1'), covar=tensor([0.0410, 0.1756, 0.1477, 0.1387, 0.1123, 0.0890, 0.0656, 0.1394], device='cuda:1'), in_proj_covar=tensor([0.0245, 0.0245, 0.0281, 0.0218, 0.0266, 0.0364, 0.0260, 0.0234], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 21:47:55,337 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.94 vs. limit=2.0 2023-03-08 21:48:15,598 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.371e+02 1.856e+02 2.318e+02 2.933e+02 8.070e+02, threshold=4.637e+02, percent-clipped=4.0 2023-03-08 21:48:21,947 INFO [train2.py:809] (1/4) Epoch 20, batch 3850, loss[ctc_loss=0.05454, att_loss=0.2055, loss=0.1753, over 15629.00 frames. utt_duration=1691 frames, utt_pad_proportion=0.008761, over 37.00 utterances.], tot_loss[ctc_loss=0.07516, att_loss=0.2362, loss=0.204, over 3262262.41 frames. utt_duration=1228 frames, utt_pad_proportion=0.06191, over 10640.73 utterances.], batch size: 37, lr: 5.33e-03, grad_scale: 16.0 2023-03-08 21:48:46,321 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-03-08 21:49:29,026 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79585.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:49:39,450 INFO [train2.py:809] (1/4) Epoch 20, batch 3900, loss[ctc_loss=0.06922, att_loss=0.2222, loss=0.1916, over 16411.00 frames. utt_duration=1494 frames, utt_pad_proportion=0.006185, over 44.00 utterances.], tot_loss[ctc_loss=0.07409, att_loss=0.2348, loss=0.2027, over 3259874.44 frames. utt_duration=1254 frames, utt_pad_proportion=0.0565, over 10406.99 utterances.], batch size: 44, lr: 5.33e-03, grad_scale: 16.0 2023-03-08 21:50:23,845 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79621.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:50:50,109 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.328e+02 1.965e+02 2.336e+02 2.819e+02 5.688e+02, threshold=4.673e+02, percent-clipped=3.0 2023-03-08 21:50:54,409 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-03-08 21:50:54,934 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79641.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 21:50:56,229 INFO [train2.py:809] (1/4) Epoch 20, batch 3950, loss[ctc_loss=0.1378, att_loss=0.2759, loss=0.2483, over 14031.00 frames. utt_duration=388.6 frames, utt_pad_proportion=0.3253, over 145.00 utterances.], tot_loss[ctc_loss=0.07421, att_loss=0.2346, loss=0.2025, over 3263207.44 frames. utt_duration=1271 frames, utt_pad_proportion=0.0526, over 10278.94 utterances.], batch size: 145, lr: 5.33e-03, grad_scale: 16.0 2023-03-08 21:51:02,684 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79646.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:51:16,360 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.7541, 5.9996, 5.4669, 5.7069, 5.6747, 5.1568, 5.3361, 5.1043], device='cuda:1'), covar=tensor([0.1352, 0.0939, 0.0939, 0.0874, 0.1073, 0.1517, 0.2466, 0.2681], device='cuda:1'), in_proj_covar=tensor([0.0518, 0.0602, 0.0453, 0.0446, 0.0426, 0.0458, 0.0607, 0.0524], 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') 2023-03-08 21:51:19,616 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79657.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:51:38,161 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=79669.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:52:14,305 INFO [train2.py:809] (1/4) Epoch 21, batch 0, loss[ctc_loss=0.06742, att_loss=0.2191, loss=0.1887, over 16000.00 frames. utt_duration=1601 frames, utt_pad_proportion=0.007946, over 40.00 utterances.], tot_loss[ctc_loss=0.06742, att_loss=0.2191, loss=0.1887, over 16000.00 frames. utt_duration=1601 frames, utt_pad_proportion=0.007946, over 40.00 utterances.], batch size: 40, lr: 5.20e-03, grad_scale: 16.0 2023-03-08 21:52:14,305 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-08 21:52:26,403 INFO [train2.py:843] (1/4) Epoch 21, validation: ctc_loss=0.04229, att_loss=0.2351, loss=0.1965, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-08 21:52:26,404 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-08 21:52:29,133 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-03-08 21:52:47,096 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79689.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:53:33,523 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79718.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:53:36,528 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79720.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:53:46,011 INFO [train2.py:809] (1/4) Epoch 21, batch 50, loss[ctc_loss=0.08401, att_loss=0.2471, loss=0.2145, over 16461.00 frames. utt_duration=1433 frames, utt_pad_proportion=0.007005, over 46.00 utterances.], tot_loss[ctc_loss=0.07482, att_loss=0.238, loss=0.2054, over 744196.92 frames. utt_duration=1277 frames, utt_pad_proportion=0.04224, over 2333.79 utterances.], batch size: 46, lr: 5.19e-03, grad_scale: 16.0 2023-03-08 21:53:54,356 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79731.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 21:54:05,024 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.329e+02 1.975e+02 2.497e+02 3.178e+02 8.121e+02, threshold=4.994e+02, percent-clipped=4.0 2023-03-08 21:54:25,558 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-03-08 21:54:43,433 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79762.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:55:05,926 INFO [train2.py:809] (1/4) Epoch 21, batch 100, loss[ctc_loss=0.1044, att_loss=0.2589, loss=0.228, over 17069.00 frames. utt_duration=1314 frames, utt_pad_proportion=0.007939, over 52.00 utterances.], tot_loss[ctc_loss=0.0764, att_loss=0.2379, loss=0.2056, over 1297032.66 frames. utt_duration=1231 frames, utt_pad_proportion=0.06017, over 4218.61 utterances.], batch size: 52, lr: 5.19e-03, grad_scale: 16.0 2023-03-08 21:55:50,609 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.8469, 3.5720, 3.0442, 3.2966, 3.7755, 3.4693, 2.8789, 3.9933], device='cuda:1'), covar=tensor([0.1067, 0.0511, 0.1054, 0.0667, 0.0665, 0.0772, 0.0855, 0.0457], device='cuda:1'), in_proj_covar=tensor([0.0202, 0.0214, 0.0225, 0.0197, 0.0276, 0.0237, 0.0199, 0.0286], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 21:56:22,746 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79823.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:56:26,927 INFO [train2.py:809] (1/4) Epoch 21, batch 150, loss[ctc_loss=0.05187, att_loss=0.2073, loss=0.1762, over 15633.00 frames. utt_duration=1691 frames, utt_pad_proportion=0.0097, over 37.00 utterances.], tot_loss[ctc_loss=0.07486, att_loss=0.2363, loss=0.204, over 1740280.68 frames. utt_duration=1233 frames, utt_pad_proportion=0.05573, over 5654.56 utterances.], batch size: 37, lr: 5.19e-03, grad_scale: 16.0 2023-03-08 21:56:44,958 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.5558, 2.8295, 3.5292, 4.5926, 4.0541, 3.9935, 3.1269, 2.3170], device='cuda:1'), covar=tensor([0.0614, 0.2056, 0.0827, 0.0535, 0.0836, 0.0459, 0.1374, 0.2263], device='cuda:1'), in_proj_covar=tensor([0.0178, 0.0215, 0.0190, 0.0216, 0.0221, 0.0176, 0.0201, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 21:56:46,156 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.523e+02 1.968e+02 2.419e+02 2.959e+02 7.390e+02, threshold=4.838e+02, percent-clipped=3.0 2023-03-08 21:56:51,739 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0152, 5.3038, 5.5821, 5.3913, 5.5394, 6.0018, 5.1700, 6.1110], device='cuda:1'), covar=tensor([0.0688, 0.0689, 0.0798, 0.1289, 0.1601, 0.0798, 0.0791, 0.0598], device='cuda:1'), in_proj_covar=tensor([0.0864, 0.0509, 0.0588, 0.0655, 0.0861, 0.0615, 0.0482, 0.0602], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 21:56:52,049 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1276, 5.1566, 4.8629, 2.3910, 2.0546, 2.9911, 2.5742, 3.9146], device='cuda:1'), covar=tensor([0.0655, 0.0249, 0.0284, 0.5089, 0.5558, 0.2426, 0.3502, 0.1572], device='cuda:1'), in_proj_covar=tensor([0.0348, 0.0266, 0.0262, 0.0240, 0.0340, 0.0330, 0.0249, 0.0361], device='cuda:1'), out_proj_covar=tensor([1.4775e-04, 9.8472e-05, 1.1169e-04, 1.0265e-04, 1.4227e-04, 1.2903e-04, 9.9854e-05, 1.4678e-04], device='cuda:1') 2023-03-08 21:57:47,563 INFO [train2.py:809] (1/4) Epoch 21, batch 200, loss[ctc_loss=0.08541, att_loss=0.2535, loss=0.2199, over 17304.00 frames. utt_duration=1260 frames, utt_pad_proportion=0.01158, over 55.00 utterances.], tot_loss[ctc_loss=0.07584, att_loss=0.2376, loss=0.2052, over 2080696.85 frames. utt_duration=1203 frames, utt_pad_proportion=0.06251, over 6926.94 utterances.], batch size: 55, lr: 5.19e-03, grad_scale: 16.0 2023-03-08 21:57:51,250 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.8646, 6.0942, 5.5933, 5.8683, 5.7422, 5.3395, 5.4979, 5.2467], device='cuda:1'), covar=tensor([0.1167, 0.0903, 0.0897, 0.0817, 0.0897, 0.1434, 0.2307, 0.2616], device='cuda:1'), in_proj_covar=tensor([0.0511, 0.0595, 0.0448, 0.0438, 0.0420, 0.0452, 0.0600, 0.0519], 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') 2023-03-08 21:59:08,384 INFO [train2.py:809] (1/4) Epoch 21, batch 250, loss[ctc_loss=0.05473, att_loss=0.2378, loss=0.2012, over 16614.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.005887, over 47.00 utterances.], tot_loss[ctc_loss=0.07576, att_loss=0.2374, loss=0.2051, over 2347127.23 frames. utt_duration=1189 frames, utt_pad_proportion=0.06608, over 7905.93 utterances.], batch size: 47, lr: 5.19e-03, grad_scale: 16.0 2023-03-08 21:59:28,135 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.404e+02 2.056e+02 2.417e+02 2.822e+02 7.152e+02, threshold=4.834e+02, percent-clipped=3.0 2023-03-08 21:59:33,216 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79941.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 21:59:33,330 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79941.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 22:00:29,281 INFO [train2.py:809] (1/4) Epoch 21, batch 300, loss[ctc_loss=0.119, att_loss=0.27, loss=0.2398, over 13608.00 frames. utt_duration=374.2 frames, utt_pad_proportion=0.3458, over 146.00 utterances.], tot_loss[ctc_loss=0.07509, att_loss=0.2366, loss=0.2043, over 2539140.55 frames. utt_duration=1224 frames, utt_pad_proportion=0.06309, over 8307.89 utterances.], batch size: 146, lr: 5.19e-03, grad_scale: 16.0 2023-03-08 22:00:49,776 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=79989.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 22:00:49,902 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79989.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:01:06,867 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-03-08 22:01:32,387 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80013.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:01:43,961 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80020.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:01:48,050 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-03-08 22:01:53,324 INFO [train2.py:809] (1/4) Epoch 21, batch 350, loss[ctc_loss=0.06065, att_loss=0.2198, loss=0.188, over 15953.00 frames. utt_duration=1558 frames, utt_pad_proportion=0.006486, over 41.00 utterances.], tot_loss[ctc_loss=0.07404, att_loss=0.2354, loss=0.2032, over 2703986.83 frames. utt_duration=1245 frames, utt_pad_proportion=0.05566, over 8699.63 utterances.], batch size: 41, lr: 5.18e-03, grad_scale: 16.0 2023-03-08 22:02:02,051 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80031.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 22:02:11,085 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=80037.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:02:12,388 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.209e+02 1.969e+02 2.385e+02 2.918e+02 9.167e+02, threshold=4.769e+02, percent-clipped=4.0 2023-03-08 22:03:01,915 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=80068.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:03:14,460 INFO [train2.py:809] (1/4) Epoch 21, batch 400, loss[ctc_loss=0.07131, att_loss=0.2499, loss=0.2142, over 16477.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.006714, over 46.00 utterances.], tot_loss[ctc_loss=0.07481, att_loss=0.2368, loss=0.2044, over 2845091.14 frames. utt_duration=1255 frames, utt_pad_proportion=0.04805, over 9076.43 utterances.], batch size: 46, lr: 5.18e-03, grad_scale: 16.0 2023-03-08 22:03:19,731 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=80079.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 22:03:53,042 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-03-08 22:04:22,265 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80118.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:04:34,492 INFO [train2.py:809] (1/4) Epoch 21, batch 450, loss[ctc_loss=0.1005, att_loss=0.268, loss=0.2345, over 17130.00 frames. utt_duration=1225 frames, utt_pad_proportion=0.01414, over 56.00 utterances.], tot_loss[ctc_loss=0.07539, att_loss=0.237, loss=0.2047, over 2941425.52 frames. utt_duration=1265 frames, utt_pad_proportion=0.04643, over 9314.99 utterances.], batch size: 56, lr: 5.18e-03, grad_scale: 16.0 2023-03-08 22:04:53,682 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.266e+02 2.042e+02 2.548e+02 3.452e+02 1.189e+03, threshold=5.096e+02, percent-clipped=8.0 2023-03-08 22:05:19,410 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0944, 5.1289, 4.9138, 2.3884, 2.0703, 3.1080, 2.5717, 3.8175], device='cuda:1'), covar=tensor([0.0693, 0.0299, 0.0268, 0.5234, 0.5643, 0.2209, 0.3480, 0.1774], device='cuda:1'), in_proj_covar=tensor([0.0351, 0.0269, 0.0265, 0.0242, 0.0342, 0.0332, 0.0250, 0.0364], device='cuda:1'), out_proj_covar=tensor([1.4897e-04, 9.9720e-05, 1.1257e-04, 1.0351e-04, 1.4314e-04, 1.2978e-04, 1.0007e-04, 1.4776e-04], device='cuda:1') 2023-03-08 22:05:55,601 INFO [train2.py:809] (1/4) Epoch 21, batch 500, loss[ctc_loss=0.06422, att_loss=0.2302, loss=0.197, over 16416.00 frames. utt_duration=1494 frames, utt_pad_proportion=0.006785, over 44.00 utterances.], tot_loss[ctc_loss=0.07468, att_loss=0.2365, loss=0.2041, over 3017839.36 frames. utt_duration=1279 frames, utt_pad_proportion=0.04257, over 9452.40 utterances.], batch size: 44, lr: 5.18e-03, grad_scale: 16.0 2023-03-08 22:07:16,767 INFO [train2.py:809] (1/4) Epoch 21, batch 550, loss[ctc_loss=0.07533, att_loss=0.2302, loss=0.1992, over 16009.00 frames. utt_duration=1602 frames, utt_pad_proportion=0.007218, over 40.00 utterances.], tot_loss[ctc_loss=0.07497, att_loss=0.2358, loss=0.2037, over 3061988.35 frames. utt_duration=1238 frames, utt_pad_proportion=0.05772, over 9905.66 utterances.], batch size: 40, lr: 5.18e-03, grad_scale: 16.0 2023-03-08 22:07:35,878 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 9.287e+01 1.925e+02 2.164e+02 2.477e+02 4.088e+02, threshold=4.327e+02, percent-clipped=0.0 2023-03-08 22:07:41,502 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80241.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:08:36,283 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.6652, 4.9561, 5.2296, 5.0495, 5.0999, 5.5945, 5.0704, 5.7358], device='cuda:1'), covar=tensor([0.0696, 0.0764, 0.0791, 0.1298, 0.1869, 0.0881, 0.0752, 0.0595], device='cuda:1'), in_proj_covar=tensor([0.0858, 0.0509, 0.0586, 0.0649, 0.0864, 0.0612, 0.0476, 0.0599], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 22:08:37,652 INFO [train2.py:809] (1/4) Epoch 21, batch 600, loss[ctc_loss=0.06154, att_loss=0.2419, loss=0.2058, over 17368.00 frames. utt_duration=1008 frames, utt_pad_proportion=0.04952, over 69.00 utterances.], tot_loss[ctc_loss=0.07412, att_loss=0.2352, loss=0.203, over 3109013.18 frames. utt_duration=1263 frames, utt_pad_proportion=0.05225, over 9861.22 utterances.], batch size: 69, lr: 5.18e-03, grad_scale: 16.0 2023-03-08 22:08:49,105 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80283.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:08:59,329 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=80289.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:09:13,251 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-03-08 22:09:38,452 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80313.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:09:58,599 INFO [train2.py:809] (1/4) Epoch 21, batch 650, loss[ctc_loss=0.07054, att_loss=0.2442, loss=0.2095, over 17042.00 frames. utt_duration=1313 frames, utt_pad_proportion=0.008526, over 52.00 utterances.], tot_loss[ctc_loss=0.07433, att_loss=0.2351, loss=0.2029, over 3146396.75 frames. utt_duration=1263 frames, utt_pad_proportion=0.05112, over 9977.73 utterances.], batch size: 52, lr: 5.17e-03, grad_scale: 16.0 2023-03-08 22:10:18,136 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.082e+02 1.849e+02 2.223e+02 2.876e+02 6.376e+02, threshold=4.446e+02, percent-clipped=5.0 2023-03-08 22:10:27,848 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80344.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:10:55,586 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=80361.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:10:55,643 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0177, 6.3389, 5.7858, 6.0945, 5.9310, 5.5309, 5.7480, 5.5346], device='cuda:1'), covar=tensor([0.1362, 0.0769, 0.0811, 0.0737, 0.0832, 0.1412, 0.2154, 0.2345], device='cuda:1'), in_proj_covar=tensor([0.0519, 0.0603, 0.0455, 0.0448, 0.0426, 0.0463, 0.0607, 0.0521], 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') 2023-03-08 22:11:19,503 INFO [train2.py:809] (1/4) Epoch 21, batch 700, loss[ctc_loss=0.07059, att_loss=0.2278, loss=0.1964, over 15748.00 frames. utt_duration=1659 frames, utt_pad_proportion=0.009039, over 38.00 utterances.], tot_loss[ctc_loss=0.07366, att_loss=0.235, loss=0.2027, over 3178093.33 frames. utt_duration=1281 frames, utt_pad_proportion=0.0469, over 9934.71 utterances.], batch size: 38, lr: 5.17e-03, grad_scale: 16.0 2023-03-08 22:11:46,972 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80393.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:12:28,219 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80418.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:12:40,241 INFO [train2.py:809] (1/4) Epoch 21, batch 750, loss[ctc_loss=0.05875, att_loss=0.2086, loss=0.1786, over 15645.00 frames. utt_duration=1693 frames, utt_pad_proportion=0.008656, over 37.00 utterances.], tot_loss[ctc_loss=0.07427, att_loss=0.2355, loss=0.2032, over 3189739.32 frames. utt_duration=1264 frames, utt_pad_proportion=0.05322, over 10107.04 utterances.], batch size: 37, lr: 5.17e-03, grad_scale: 16.0 2023-03-08 22:12:59,013 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.272e+02 1.906e+02 2.332e+02 2.965e+02 6.816e+02, threshold=4.664e+02, percent-clipped=5.0 2023-03-08 22:13:25,125 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80454.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:13:39,560 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80463.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:13:44,238 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=80466.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:13:59,490 INFO [train2.py:809] (1/4) Epoch 21, batch 800, loss[ctc_loss=0.1032, att_loss=0.257, loss=0.2262, over 17303.00 frames. utt_duration=1260 frames, utt_pad_proportion=0.01179, over 55.00 utterances.], tot_loss[ctc_loss=0.0754, att_loss=0.2361, loss=0.2039, over 3212737.77 frames. utt_duration=1255 frames, utt_pad_proportion=0.05287, over 10249.18 utterances.], batch size: 55, lr: 5.17e-03, grad_scale: 16.0 2023-03-08 22:14:29,054 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80494.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:14:42,575 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=2.80 vs. limit=5.0 2023-03-08 22:15:05,165 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.5754, 3.8678, 3.8413, 2.5013, 2.2900, 2.7956, 2.4432, 3.4427], device='cuda:1'), covar=tensor([0.0690, 0.0385, 0.0386, 0.3398, 0.4200, 0.2072, 0.2689, 0.1295], device='cuda:1'), in_proj_covar=tensor([0.0351, 0.0270, 0.0265, 0.0241, 0.0342, 0.0332, 0.0251, 0.0363], 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') 2023-03-08 22:15:17,437 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80524.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:15:20,316 INFO [train2.py:809] (1/4) Epoch 21, batch 850, loss[ctc_loss=0.08358, att_loss=0.249, loss=0.2159, over 17024.00 frames. utt_duration=1336 frames, utt_pad_proportion=0.007153, over 51.00 utterances.], tot_loss[ctc_loss=0.07599, att_loss=0.2364, loss=0.2043, over 3229475.23 frames. utt_duration=1242 frames, utt_pad_proportion=0.05538, over 10410.72 utterances.], batch size: 51, lr: 5.17e-03, grad_scale: 16.0 2023-03-08 22:15:40,967 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.109e+02 2.073e+02 2.361e+02 2.912e+02 1.019e+03, threshold=4.723e+02, percent-clipped=4.0 2023-03-08 22:16:07,034 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80555.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:16:25,174 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4258, 2.5435, 4.9401, 3.8164, 3.0253, 4.1905, 4.7263, 4.5855], device='cuda:1'), covar=tensor([0.0254, 0.1613, 0.0175, 0.0919, 0.1709, 0.0260, 0.0164, 0.0272], device='cuda:1'), in_proj_covar=tensor([0.0188, 0.0237, 0.0180, 0.0306, 0.0258, 0.0209, 0.0170, 0.0199], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 22:16:40,199 INFO [train2.py:809] (1/4) Epoch 21, batch 900, loss[ctc_loss=0.07865, att_loss=0.2514, loss=0.2168, over 17263.00 frames. utt_duration=1257 frames, utt_pad_proportion=0.01332, over 55.00 utterances.], tot_loss[ctc_loss=0.07541, att_loss=0.2364, loss=0.2042, over 3238008.24 frames. utt_duration=1237 frames, utt_pad_proportion=0.0565, over 10486.83 utterances.], batch size: 55, lr: 5.17e-03, grad_scale: 16.0 2023-03-08 22:17:36,067 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.31 vs. limit=5.0 2023-03-08 22:18:00,939 INFO [train2.py:809] (1/4) Epoch 21, batch 950, loss[ctc_loss=0.06902, att_loss=0.2401, loss=0.2059, over 17312.00 frames. utt_duration=1175 frames, utt_pad_proportion=0.02304, over 59.00 utterances.], tot_loss[ctc_loss=0.07534, att_loss=0.2365, loss=0.2042, over 3245786.22 frames. utt_duration=1238 frames, utt_pad_proportion=0.05523, over 10500.18 utterances.], batch size: 59, lr: 5.16e-03, grad_scale: 16.0 2023-03-08 22:18:22,508 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.307e+02 1.900e+02 2.361e+02 2.857e+02 6.538e+02, threshold=4.723e+02, percent-clipped=3.0 2023-03-08 22:18:22,787 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80639.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:19:22,691 INFO [train2.py:809] (1/4) Epoch 21, batch 1000, loss[ctc_loss=0.08376, att_loss=0.2467, loss=0.2141, over 17369.00 frames. utt_duration=1179 frames, utt_pad_proportion=0.02095, over 59.00 utterances.], tot_loss[ctc_loss=0.07478, att_loss=0.2363, loss=0.204, over 3254911.26 frames. utt_duration=1244 frames, utt_pad_proportion=0.05397, over 10482.13 utterances.], batch size: 59, lr: 5.16e-03, grad_scale: 8.0 2023-03-08 22:19:40,991 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5051, 2.4641, 4.9812, 3.6938, 2.8116, 4.1838, 4.7589, 4.5358], device='cuda:1'), covar=tensor([0.0264, 0.1568, 0.0164, 0.1022, 0.1885, 0.0248, 0.0147, 0.0281], device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0238, 0.0182, 0.0309, 0.0260, 0.0210, 0.0171, 0.0200], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 22:20:17,011 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.8872, 6.1104, 5.6206, 5.8867, 5.8325, 5.2403, 5.5060, 5.3199], device='cuda:1'), covar=tensor([0.1234, 0.0903, 0.0890, 0.0842, 0.0908, 0.1550, 0.2419, 0.2542], device='cuda:1'), in_proj_covar=tensor([0.0509, 0.0593, 0.0448, 0.0445, 0.0418, 0.0458, 0.0598, 0.0514], 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') 2023-03-08 22:20:44,017 INFO [train2.py:809] (1/4) Epoch 21, batch 1050, loss[ctc_loss=0.05319, att_loss=0.226, loss=0.1915, over 15953.00 frames. utt_duration=1558 frames, utt_pad_proportion=0.006964, over 41.00 utterances.], tot_loss[ctc_loss=0.07402, att_loss=0.2359, loss=0.2035, over 3253409.62 frames. utt_duration=1268 frames, utt_pad_proportion=0.04875, over 10272.36 utterances.], batch size: 41, lr: 5.16e-03, grad_scale: 4.0 2023-03-08 22:21:00,936 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80736.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:21:08,354 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.034e+02 1.786e+02 2.169e+02 2.535e+02 5.351e+02, threshold=4.338e+02, percent-clipped=1.0 2023-03-08 22:21:21,088 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80749.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:22:00,949 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80773.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:22:05,368 INFO [train2.py:809] (1/4) Epoch 21, batch 1100, loss[ctc_loss=0.06743, att_loss=0.2462, loss=0.2105, over 16974.00 frames. utt_duration=1359 frames, utt_pad_proportion=0.007173, over 50.00 utterances.], tot_loss[ctc_loss=0.07405, att_loss=0.2357, loss=0.2034, over 3259243.81 frames. utt_duration=1282 frames, utt_pad_proportion=0.04485, over 10177.22 utterances.], batch size: 50, lr: 5.16e-03, grad_scale: 4.0 2023-03-08 22:22:39,451 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80797.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:23:16,402 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80819.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:23:27,017 INFO [train2.py:809] (1/4) Epoch 21, batch 1150, loss[ctc_loss=0.05724, att_loss=0.2106, loss=0.1799, over 15400.00 frames. utt_duration=1761 frames, utt_pad_proportion=0.009288, over 35.00 utterances.], tot_loss[ctc_loss=0.07397, att_loss=0.2353, loss=0.2031, over 3262613.34 frames. utt_duration=1298 frames, utt_pad_proportion=0.04211, over 10065.58 utterances.], batch size: 35, lr: 5.16e-03, grad_scale: 4.0 2023-03-08 22:23:40,085 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80834.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:23:51,459 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.205e+02 2.024e+02 2.478e+02 3.117e+02 8.099e+02, threshold=4.955e+02, percent-clipped=10.0 2023-03-08 22:23:58,334 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3188, 2.6218, 4.6146, 3.4982, 2.9383, 4.0015, 4.3091, 4.2689], device='cuda:1'), covar=tensor([0.0209, 0.1481, 0.0166, 0.1244, 0.1723, 0.0297, 0.0205, 0.0284], device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0239, 0.0182, 0.0310, 0.0261, 0.0211, 0.0171, 0.0200], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 22:24:06,585 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80850.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:24:48,749 INFO [train2.py:809] (1/4) Epoch 21, batch 1200, loss[ctc_loss=0.06185, att_loss=0.2493, loss=0.2118, over 17026.00 frames. utt_duration=1337 frames, utt_pad_proportion=0.007453, over 51.00 utterances.], tot_loss[ctc_loss=0.07365, att_loss=0.2354, loss=0.2031, over 3272414.91 frames. utt_duration=1286 frames, utt_pad_proportion=0.04267, over 10189.64 utterances.], batch size: 51, lr: 5.16e-03, grad_scale: 8.0 2023-03-08 22:24:49,296 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0249, 5.0971, 4.7791, 2.1297, 1.9981, 2.8402, 2.5846, 3.8064], device='cuda:1'), covar=tensor([0.0754, 0.0250, 0.0301, 0.5534, 0.5743, 0.2619, 0.3303, 0.1669], device='cuda:1'), in_proj_covar=tensor([0.0358, 0.0274, 0.0269, 0.0247, 0.0346, 0.0337, 0.0255, 0.0368], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-03-08 22:26:09,921 INFO [train2.py:809] (1/4) Epoch 21, batch 1250, loss[ctc_loss=0.07353, att_loss=0.2443, loss=0.2102, over 16892.00 frames. utt_duration=1381 frames, utt_pad_proportion=0.006024, over 49.00 utterances.], tot_loss[ctc_loss=0.07318, att_loss=0.2345, loss=0.2023, over 3270861.14 frames. utt_duration=1306 frames, utt_pad_proportion=0.03885, over 10028.86 utterances.], batch size: 49, lr: 5.16e-03, grad_scale: 8.0 2023-03-08 22:26:14,658 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-03-08 22:26:31,979 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80939.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:26:34,821 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.421e+02 2.023e+02 2.469e+02 3.077e+02 5.945e+02, threshold=4.937e+02, percent-clipped=1.0 2023-03-08 22:27:33,043 INFO [train2.py:809] (1/4) Epoch 21, batch 1300, loss[ctc_loss=0.07796, att_loss=0.2581, loss=0.2221, over 17151.00 frames. utt_duration=1227 frames, utt_pad_proportion=0.01317, over 56.00 utterances.], tot_loss[ctc_loss=0.07306, att_loss=0.235, loss=0.2026, over 3274119.15 frames. utt_duration=1294 frames, utt_pad_proportion=0.04176, over 10133.09 utterances.], batch size: 56, lr: 5.15e-03, grad_scale: 8.0 2023-03-08 22:27:50,967 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=80987.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:28:44,921 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4743, 2.3425, 4.8954, 3.6515, 2.8076, 4.2055, 4.7477, 4.6113], device='cuda:1'), covar=tensor([0.0255, 0.1752, 0.0192, 0.1271, 0.1926, 0.0281, 0.0146, 0.0262], device='cuda:1'), in_proj_covar=tensor([0.0188, 0.0238, 0.0182, 0.0309, 0.0260, 0.0210, 0.0171, 0.0200], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 22:28:46,369 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0276, 5.1119, 4.7855, 3.0861, 4.8611, 4.6709, 4.2812, 2.8295], device='cuda:1'), covar=tensor([0.0108, 0.0087, 0.0289, 0.0920, 0.0100, 0.0207, 0.0322, 0.1304], device='cuda:1'), in_proj_covar=tensor([0.0073, 0.0101, 0.0103, 0.0110, 0.0084, 0.0110, 0.0099, 0.0104], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 22:28:55,454 INFO [train2.py:809] (1/4) Epoch 21, batch 1350, loss[ctc_loss=0.0603, att_loss=0.209, loss=0.1793, over 15505.00 frames. utt_duration=1725 frames, utt_pad_proportion=0.00821, over 36.00 utterances.], tot_loss[ctc_loss=0.07351, att_loss=0.2354, loss=0.203, over 3277150.32 frames. utt_duration=1285 frames, utt_pad_proportion=0.04383, over 10213.74 utterances.], batch size: 36, lr: 5.15e-03, grad_scale: 8.0 2023-03-08 22:29:19,941 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.247e+02 1.963e+02 2.296e+02 2.713e+02 5.032e+02, threshold=4.592e+02, percent-clipped=2.0 2023-03-08 22:29:33,234 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81049.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:30:15,541 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0777, 5.3737, 4.8652, 5.4121, 4.7327, 5.0264, 5.4890, 5.2326], device='cuda:1'), covar=tensor([0.0567, 0.0306, 0.0872, 0.0323, 0.0452, 0.0258, 0.0210, 0.0190], device='cuda:1'), in_proj_covar=tensor([0.0381, 0.0316, 0.0363, 0.0344, 0.0315, 0.0237, 0.0297, 0.0279], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 22:30:16,877 INFO [train2.py:809] (1/4) Epoch 21, batch 1400, loss[ctc_loss=0.06743, att_loss=0.2218, loss=0.1909, over 16183.00 frames. utt_duration=1580 frames, utt_pad_proportion=0.006699, over 41.00 utterances.], tot_loss[ctc_loss=0.07393, att_loss=0.2359, loss=0.2035, over 3282052.78 frames. utt_duration=1277 frames, utt_pad_proportion=0.04448, over 10294.95 utterances.], batch size: 41, lr: 5.15e-03, grad_scale: 8.0 2023-03-08 22:30:42,817 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81092.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:30:50,488 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=81097.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:31:22,071 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81116.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:31:26,391 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81119.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:31:36,836 INFO [train2.py:809] (1/4) Epoch 21, batch 1450, loss[ctc_loss=0.07937, att_loss=0.2138, loss=0.1869, over 15518.00 frames. utt_duration=1726 frames, utt_pad_proportion=0.006825, over 36.00 utterances.], tot_loss[ctc_loss=0.07406, att_loss=0.2359, loss=0.2035, over 3290007.55 frames. utt_duration=1286 frames, utt_pad_proportion=0.03998, over 10244.99 utterances.], batch size: 36, lr: 5.15e-03, grad_scale: 8.0 2023-03-08 22:31:41,560 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81129.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:32:00,573 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.077e+02 1.791e+02 2.173e+02 2.721e+02 6.345e+02, threshold=4.347e+02, percent-clipped=3.0 2023-03-08 22:32:15,020 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81150.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:32:43,706 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=81167.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:32:57,634 INFO [train2.py:809] (1/4) Epoch 21, batch 1500, loss[ctc_loss=0.0761, att_loss=0.2372, loss=0.205, over 16873.00 frames. utt_duration=683.4 frames, utt_pad_proportion=0.1425, over 99.00 utterances.], tot_loss[ctc_loss=0.07445, att_loss=0.2362, loss=0.2039, over 3284766.44 frames. utt_duration=1249 frames, utt_pad_proportion=0.05001, over 10532.70 utterances.], batch size: 99, lr: 5.15e-03, grad_scale: 8.0 2023-03-08 22:32:59,575 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81177.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:33:33,288 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=81198.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:34:10,363 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-03-08 22:34:19,135 INFO [train2.py:809] (1/4) Epoch 21, batch 1550, loss[ctc_loss=0.0692, att_loss=0.2454, loss=0.2102, over 17017.00 frames. utt_duration=1336 frames, utt_pad_proportion=0.007977, over 51.00 utterances.], tot_loss[ctc_loss=0.07476, att_loss=0.2362, loss=0.2039, over 3285581.22 frames. utt_duration=1245 frames, utt_pad_proportion=0.05007, over 10570.95 utterances.], batch size: 51, lr: 5.15e-03, grad_scale: 8.0 2023-03-08 22:34:43,494 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.361e+02 1.828e+02 2.165e+02 2.736e+02 6.423e+02, threshold=4.329e+02, percent-clipped=4.0 2023-03-08 22:35:40,192 INFO [train2.py:809] (1/4) Epoch 21, batch 1600, loss[ctc_loss=0.1229, att_loss=0.2689, loss=0.2397, over 13505.00 frames. utt_duration=371.5 frames, utt_pad_proportion=0.3506, over 146.00 utterances.], tot_loss[ctc_loss=0.07503, att_loss=0.2367, loss=0.2044, over 3283671.25 frames. utt_duration=1239 frames, utt_pad_proportion=0.05419, over 10615.73 utterances.], batch size: 146, lr: 5.14e-03, grad_scale: 8.0 2023-03-08 22:36:16,941 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81299.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:36:20,801 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81301.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:37:01,165 INFO [train2.py:809] (1/4) Epoch 21, batch 1650, loss[ctc_loss=0.06151, att_loss=0.2317, loss=0.1977, over 16380.00 frames. utt_duration=1490 frames, utt_pad_proportion=0.008362, over 44.00 utterances.], tot_loss[ctc_loss=0.0745, att_loss=0.237, loss=0.2045, over 3285161.70 frames. utt_duration=1261 frames, utt_pad_proportion=0.04963, over 10436.77 utterances.], batch size: 44, lr: 5.14e-03, grad_scale: 8.0 2023-03-08 22:37:25,609 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.222e+02 1.957e+02 2.223e+02 2.631e+02 7.416e+02, threshold=4.446e+02, percent-clipped=3.0 2023-03-08 22:37:58,138 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81360.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:38:01,310 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81362.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:38:20,239 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.9014, 3.7504, 3.1331, 3.4309, 3.8832, 3.5760, 2.8317, 4.1452], device='cuda:1'), covar=tensor([0.1118, 0.0527, 0.1101, 0.0670, 0.0756, 0.0722, 0.0943, 0.0483], device='cuda:1'), in_proj_covar=tensor([0.0200, 0.0214, 0.0224, 0.0197, 0.0276, 0.0237, 0.0197, 0.0282], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 22:38:23,005 INFO [train2.py:809] (1/4) Epoch 21, batch 1700, loss[ctc_loss=0.07258, att_loss=0.2189, loss=0.1897, over 15381.00 frames. utt_duration=1759 frames, utt_pad_proportion=0.009936, over 35.00 utterances.], tot_loss[ctc_loss=0.07438, att_loss=0.2368, loss=0.2043, over 3288122.09 frames. utt_duration=1270 frames, utt_pad_proportion=0.04471, over 10366.60 utterances.], batch size: 35, lr: 5.14e-03, grad_scale: 8.0 2023-03-08 22:38:30,304 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-03-08 22:38:49,090 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81392.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:39:44,353 INFO [train2.py:809] (1/4) Epoch 21, batch 1750, loss[ctc_loss=0.06875, att_loss=0.2389, loss=0.2049, over 16692.00 frames. utt_duration=1453 frames, utt_pad_proportion=0.00632, over 46.00 utterances.], tot_loss[ctc_loss=0.07496, att_loss=0.2367, loss=0.2044, over 3274231.41 frames. utt_duration=1243 frames, utt_pad_proportion=0.05441, over 10548.09 utterances.], batch size: 46, lr: 5.14e-03, grad_scale: 8.0 2023-03-08 22:39:49,329 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81429.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:39:49,370 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7031, 2.9933, 3.7034, 3.1069, 3.6087, 4.7082, 4.4849, 3.2712], device='cuda:1'), covar=tensor([0.0363, 0.1674, 0.1368, 0.1401, 0.1095, 0.0740, 0.0577, 0.1335], device='cuda:1'), in_proj_covar=tensor([0.0241, 0.0240, 0.0276, 0.0217, 0.0262, 0.0362, 0.0259, 0.0230], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 22:40:06,985 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=81440.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:40:08,461 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.362e+02 2.066e+02 2.468e+02 3.056e+02 6.142e+02, threshold=4.936e+02, percent-clipped=4.0 2023-03-08 22:40:25,531 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1779, 5.2181, 4.9772, 3.0121, 4.9389, 4.7973, 4.3905, 2.8112], device='cuda:1'), covar=tensor([0.0109, 0.0093, 0.0232, 0.0968, 0.0108, 0.0173, 0.0303, 0.1354], device='cuda:1'), in_proj_covar=tensor([0.0074, 0.0102, 0.0104, 0.0110, 0.0084, 0.0112, 0.0099, 0.0105], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 22:40:59,674 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81472.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:41:05,530 INFO [train2.py:809] (1/4) Epoch 21, batch 1800, loss[ctc_loss=0.07853, att_loss=0.2393, loss=0.2071, over 17300.00 frames. utt_duration=1260 frames, utt_pad_proportion=0.01205, over 55.00 utterances.], tot_loss[ctc_loss=0.07487, att_loss=0.2365, loss=0.2042, over 3276695.03 frames. utt_duration=1239 frames, utt_pad_proportion=0.05497, over 10590.92 utterances.], batch size: 55, lr: 5.14e-03, grad_scale: 8.0 2023-03-08 22:41:07,206 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=81477.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:42:25,222 INFO [train2.py:809] (1/4) Epoch 21, batch 1850, loss[ctc_loss=0.1047, att_loss=0.2604, loss=0.2293, over 17269.00 frames. utt_duration=1257 frames, utt_pad_proportion=0.0139, over 55.00 utterances.], tot_loss[ctc_loss=0.07498, att_loss=0.2361, loss=0.2039, over 3271820.58 frames. utt_duration=1238 frames, utt_pad_proportion=0.05642, over 10586.80 utterances.], batch size: 55, lr: 5.14e-03, grad_scale: 8.0 2023-03-08 22:42:49,246 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.186e+02 1.850e+02 2.304e+02 2.710e+02 5.976e+02, threshold=4.608e+02, percent-clipped=1.0 2023-03-08 22:43:45,866 INFO [train2.py:809] (1/4) Epoch 21, batch 1900, loss[ctc_loss=0.06659, att_loss=0.2314, loss=0.1984, over 16625.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.005484, over 47.00 utterances.], tot_loss[ctc_loss=0.07451, att_loss=0.2355, loss=0.2033, over 3277819.72 frames. utt_duration=1267 frames, utt_pad_proportion=0.04863, over 10364.46 utterances.], batch size: 47, lr: 5.13e-03, grad_scale: 8.0 2023-03-08 22:45:07,143 INFO [train2.py:809] (1/4) Epoch 21, batch 1950, loss[ctc_loss=0.07319, att_loss=0.2507, loss=0.2152, over 17105.00 frames. utt_duration=692.6 frames, utt_pad_proportion=0.1299, over 99.00 utterances.], tot_loss[ctc_loss=0.07382, att_loss=0.2351, loss=0.2028, over 3268647.93 frames. utt_duration=1247 frames, utt_pad_proportion=0.05579, over 10500.15 utterances.], batch size: 99, lr: 5.13e-03, grad_scale: 8.0 2023-03-08 22:45:31,394 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.479e+02 1.984e+02 2.266e+02 2.808e+02 6.592e+02, threshold=4.533e+02, percent-clipped=3.0 2023-03-08 22:45:38,502 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-03-08 22:45:54,457 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81655.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:45:58,073 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81657.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:46:18,697 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81670.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:46:20,284 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.5396, 1.6961, 2.2101, 2.1725, 2.1288, 2.2745, 1.8668, 2.6879], device='cuda:1'), covar=tensor([0.1246, 0.2800, 0.1844, 0.1615, 0.2371, 0.1200, 0.2136, 0.1209], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0126, 0.0120, 0.0110, 0.0125, 0.0109, 0.0131, 0.0100], device='cuda:1'), out_proj_covar=tensor([8.9315e-05, 9.7581e-05, 9.5848e-05, 8.6201e-05, 9.2686e-05, 8.6791e-05, 9.8303e-05, 8.0116e-05], device='cuda:1') 2023-03-08 22:46:27,654 INFO [train2.py:809] (1/4) Epoch 21, batch 2000, loss[ctc_loss=0.06002, att_loss=0.2369, loss=0.2015, over 16617.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.005842, over 47.00 utterances.], tot_loss[ctc_loss=0.07385, att_loss=0.2346, loss=0.2025, over 3255249.05 frames. utt_duration=1248 frames, utt_pad_proportion=0.05921, over 10449.74 utterances.], batch size: 47, lr: 5.13e-03, grad_scale: 8.0 2023-03-08 22:46:53,533 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.5132, 1.9193, 2.3126, 2.2190, 2.4485, 2.3375, 2.0724, 2.8089], device='cuda:1'), covar=tensor([0.1417, 0.3180, 0.2187, 0.1835, 0.1996, 0.1515, 0.2248, 0.1104], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0126, 0.0120, 0.0110, 0.0126, 0.0109, 0.0131, 0.0100], device='cuda:1'), out_proj_covar=tensor([8.9342e-05, 9.7647e-05, 9.5809e-05, 8.6242e-05, 9.2849e-05, 8.6941e-05, 9.8407e-05, 8.0098e-05], device='cuda:1') 2023-03-08 22:47:24,556 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4874, 3.0594, 3.5999, 2.8798, 3.4255, 4.6284, 4.3433, 3.0863], device='cuda:1'), covar=tensor([0.0419, 0.1666, 0.1462, 0.1449, 0.1242, 0.1020, 0.0778, 0.1469], device='cuda:1'), in_proj_covar=tensor([0.0245, 0.0243, 0.0281, 0.0220, 0.0267, 0.0366, 0.0263, 0.0232], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 22:47:47,659 INFO [train2.py:809] (1/4) Epoch 21, batch 2050, loss[ctc_loss=0.06486, att_loss=0.2378, loss=0.2032, over 16676.00 frames. utt_duration=1451 frames, utt_pad_proportion=0.006636, over 46.00 utterances.], tot_loss[ctc_loss=0.07458, att_loss=0.2352, loss=0.2031, over 3257460.07 frames. utt_duration=1232 frames, utt_pad_proportion=0.06323, over 10589.61 utterances.], batch size: 46, lr: 5.13e-03, grad_scale: 8.0 2023-03-08 22:47:55,825 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81731.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 22:48:11,614 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.378e+02 1.913e+02 2.441e+02 2.878e+02 1.219e+03, threshold=4.882e+02, percent-clipped=4.0 2023-03-08 22:48:26,714 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9412, 4.9622, 4.5964, 2.8518, 4.6638, 4.6175, 4.1960, 2.5657], device='cuda:1'), covar=tensor([0.0147, 0.0116, 0.0372, 0.1141, 0.0129, 0.0233, 0.0356, 0.1623], device='cuda:1'), in_proj_covar=tensor([0.0074, 0.0102, 0.0104, 0.0110, 0.0084, 0.0112, 0.0100, 0.0105], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-08 22:49:02,497 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81772.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:49:07,036 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9940, 5.2515, 5.5502, 5.4119, 5.5155, 5.9661, 5.2183, 6.0548], device='cuda:1'), covar=tensor([0.0783, 0.0734, 0.0831, 0.1326, 0.1925, 0.0951, 0.0744, 0.0711], device='cuda:1'), in_proj_covar=tensor([0.0874, 0.0513, 0.0598, 0.0667, 0.0878, 0.0634, 0.0488, 0.0610], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 22:49:08,393 INFO [train2.py:809] (1/4) Epoch 21, batch 2100, loss[ctc_loss=0.05305, att_loss=0.2323, loss=0.1964, over 16537.00 frames. utt_duration=1471 frames, utt_pad_proportion=0.006422, over 45.00 utterances.], tot_loss[ctc_loss=0.07445, att_loss=0.2353, loss=0.2031, over 3256427.85 frames. utt_duration=1223 frames, utt_pad_proportion=0.06697, over 10666.02 utterances.], batch size: 45, lr: 5.13e-03, grad_scale: 8.0 2023-03-08 22:50:18,427 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.80 vs. limit=2.0 2023-03-08 22:50:20,895 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=81820.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:50:30,144 INFO [train2.py:809] (1/4) Epoch 21, batch 2150, loss[ctc_loss=0.06447, att_loss=0.2227, loss=0.1911, over 15875.00 frames. utt_duration=1630 frames, utt_pad_proportion=0.009876, over 39.00 utterances.], tot_loss[ctc_loss=0.07423, att_loss=0.2357, loss=0.2034, over 3259725.13 frames. utt_duration=1229 frames, utt_pad_proportion=0.06466, over 10623.00 utterances.], batch size: 39, lr: 5.13e-03, grad_scale: 8.0 2023-03-08 22:50:54,310 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.279e+02 1.964e+02 2.354e+02 2.825e+02 8.417e+02, threshold=4.708e+02, percent-clipped=3.0 2023-03-08 22:51:50,725 INFO [train2.py:809] (1/4) Epoch 21, batch 2200, loss[ctc_loss=0.07277, att_loss=0.2484, loss=0.2132, over 17262.00 frames. utt_duration=1257 frames, utt_pad_proportion=0.01356, over 55.00 utterances.], tot_loss[ctc_loss=0.07379, att_loss=0.236, loss=0.2035, over 3268822.28 frames. utt_duration=1238 frames, utt_pad_proportion=0.05915, over 10575.02 utterances.], batch size: 55, lr: 5.13e-03, grad_scale: 8.0 2023-03-08 22:52:06,793 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-03-08 22:52:41,347 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2404, 3.0419, 3.3933, 4.4910, 3.8865, 3.9547, 3.0207, 2.2341], device='cuda:1'), covar=tensor([0.0830, 0.1862, 0.0981, 0.0551, 0.0886, 0.0450, 0.1558, 0.2461], device='cuda:1'), in_proj_covar=tensor([0.0181, 0.0215, 0.0191, 0.0218, 0.0223, 0.0177, 0.0202, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 22:53:11,516 INFO [train2.py:809] (1/4) Epoch 21, batch 2250, loss[ctc_loss=0.06803, att_loss=0.2402, loss=0.2058, over 16325.00 frames. utt_duration=1453 frames, utt_pad_proportion=0.006384, over 45.00 utterances.], tot_loss[ctc_loss=0.07449, att_loss=0.237, loss=0.2045, over 3276231.99 frames. utt_duration=1218 frames, utt_pad_proportion=0.06101, over 10769.24 utterances.], batch size: 45, lr: 5.12e-03, grad_scale: 8.0 2023-03-08 22:53:20,350 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.0875, 3.8226, 3.3096, 3.5691, 4.1046, 3.7834, 3.0126, 4.3732], device='cuda:1'), covar=tensor([0.1087, 0.0540, 0.1067, 0.0669, 0.0650, 0.0678, 0.0864, 0.0396], device='cuda:1'), in_proj_covar=tensor([0.0200, 0.0215, 0.0225, 0.0198, 0.0277, 0.0238, 0.0199, 0.0282], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 22:53:35,603 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.367e+02 2.106e+02 2.465e+02 3.018e+02 5.655e+02, threshold=4.929e+02, percent-clipped=3.0 2023-03-08 22:53:47,111 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2102, 2.8877, 3.3492, 4.4291, 3.9034, 3.9543, 2.9491, 2.2644], device='cuda:1'), covar=tensor([0.0834, 0.1941, 0.0926, 0.0555, 0.0858, 0.0430, 0.1496, 0.2250], device='cuda:1'), in_proj_covar=tensor([0.0181, 0.0215, 0.0191, 0.0218, 0.0223, 0.0177, 0.0202, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 22:53:58,783 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81955.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:54:02,672 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81957.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:54:32,970 INFO [train2.py:809] (1/4) Epoch 21, batch 2300, loss[ctc_loss=0.06985, att_loss=0.2358, loss=0.2026, over 16530.00 frames. utt_duration=1471 frames, utt_pad_proportion=0.006992, over 45.00 utterances.], tot_loss[ctc_loss=0.07437, att_loss=0.237, loss=0.2045, over 3277357.57 frames. utt_duration=1193 frames, utt_pad_proportion=0.06739, over 11006.01 utterances.], batch size: 45, lr: 5.12e-03, grad_scale: 8.0 2023-03-08 22:54:33,375 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.9893, 3.6952, 3.6398, 3.1342, 3.7138, 3.7676, 3.6570, 2.7986], device='cuda:1'), covar=tensor([0.1051, 0.1714, 0.2917, 0.3583, 0.1160, 0.3768, 0.1069, 0.3723], device='cuda:1'), in_proj_covar=tensor([0.0171, 0.0185, 0.0195, 0.0249, 0.0155, 0.0256, 0.0177, 0.0214], 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') 2023-03-08 22:54:50,569 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81987.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:55:22,030 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=82003.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:55:24,614 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-03-08 22:55:25,325 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=82005.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:55:27,213 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.5187, 3.1048, 2.6538, 2.9355, 3.1518, 3.0646, 2.5503, 3.0527], device='cuda:1'), covar=tensor([0.0985, 0.0429, 0.0913, 0.0586, 0.0723, 0.0561, 0.0783, 0.0464], device='cuda:1'), in_proj_covar=tensor([0.0201, 0.0216, 0.0226, 0.0199, 0.0278, 0.0239, 0.0200, 0.0285], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 22:55:40,031 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4388, 4.3299, 4.2965, 4.3740, 4.9695, 4.4716, 4.4408, 2.3449], device='cuda:1'), covar=tensor([0.0247, 0.0393, 0.0388, 0.0288, 0.0782, 0.0219, 0.0308, 0.1993], device='cuda:1'), in_proj_covar=tensor([0.0161, 0.0184, 0.0183, 0.0199, 0.0363, 0.0156, 0.0173, 0.0213], 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') 2023-03-08 22:55:59,320 INFO [train2.py:809] (1/4) Epoch 21, batch 2350, loss[ctc_loss=0.07488, att_loss=0.2511, loss=0.2159, over 17441.00 frames. utt_duration=1109 frames, utt_pad_proportion=0.02982, over 63.00 utterances.], tot_loss[ctc_loss=0.0741, att_loss=0.2367, loss=0.2042, over 3276461.78 frames. utt_duration=1208 frames, utt_pad_proportion=0.06446, over 10861.74 utterances.], batch size: 63, lr: 5.12e-03, grad_scale: 8.0 2023-03-08 22:55:59,513 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82026.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 22:56:14,681 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-03-08 22:56:20,096 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3031, 2.8275, 3.5588, 2.8614, 3.4445, 4.4521, 4.2278, 3.1110], device='cuda:1'), covar=tensor([0.0364, 0.1749, 0.1211, 0.1469, 0.1182, 0.0782, 0.0629, 0.1344], device='cuda:1'), in_proj_covar=tensor([0.0245, 0.0244, 0.0279, 0.0220, 0.0266, 0.0365, 0.0262, 0.0232], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 22:56:22,870 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.390e+02 1.965e+02 2.374e+02 2.963e+02 4.469e+02, threshold=4.748e+02, percent-clipped=0.0 2023-03-08 22:56:35,655 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82048.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:56:47,289 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82055.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:57:20,438 INFO [train2.py:809] (1/4) Epoch 21, batch 2400, loss[ctc_loss=0.07664, att_loss=0.224, loss=0.1945, over 15653.00 frames. utt_duration=1694 frames, utt_pad_proportion=0.007632, over 37.00 utterances.], tot_loss[ctc_loss=0.07397, att_loss=0.2365, loss=0.204, over 3273265.16 frames. utt_duration=1200 frames, utt_pad_proportion=0.06678, over 10928.14 utterances.], batch size: 37, lr: 5.12e-03, grad_scale: 8.0 2023-03-08 22:57:39,983 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-03-08 22:57:40,713 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5813, 4.9160, 4.5014, 4.9712, 4.4382, 4.5863, 4.9777, 4.7993], device='cuda:1'), covar=tensor([0.0665, 0.0308, 0.0780, 0.0343, 0.0419, 0.0360, 0.0265, 0.0227], device='cuda:1'), in_proj_covar=tensor([0.0384, 0.0317, 0.0361, 0.0346, 0.0318, 0.0237, 0.0298, 0.0283], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 22:58:25,042 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82116.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 22:58:40,651 INFO [train2.py:809] (1/4) Epoch 21, batch 2450, loss[ctc_loss=0.05698, att_loss=0.2074, loss=0.1774, over 15775.00 frames. utt_duration=1662 frames, utt_pad_proportion=0.008228, over 38.00 utterances.], tot_loss[ctc_loss=0.07363, att_loss=0.2362, loss=0.2037, over 3268625.74 frames. utt_duration=1207 frames, utt_pad_proportion=0.06614, over 10843.96 utterances.], batch size: 38, lr: 5.12e-03, grad_scale: 8.0 2023-03-08 22:59:04,489 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 9.920e+01 1.856e+02 2.212e+02 2.703e+02 4.819e+02, threshold=4.424e+02, percent-clipped=1.0 2023-03-08 23:00:00,564 INFO [train2.py:809] (1/4) Epoch 21, batch 2500, loss[ctc_loss=0.05437, att_loss=0.2348, loss=0.1987, over 17352.00 frames. utt_duration=1103 frames, utt_pad_proportion=0.03507, over 63.00 utterances.], tot_loss[ctc_loss=0.07439, att_loss=0.2366, loss=0.2042, over 3265585.22 frames. utt_duration=1197 frames, utt_pad_proportion=0.06988, over 10923.95 utterances.], batch size: 63, lr: 5.12e-03, grad_scale: 8.0 2023-03-08 23:01:15,863 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-03-08 23:01:21,096 INFO [train2.py:809] (1/4) Epoch 21, batch 2550, loss[ctc_loss=0.07444, att_loss=0.2361, loss=0.2037, over 16369.00 frames. utt_duration=1490 frames, utt_pad_proportion=0.009475, over 44.00 utterances.], tot_loss[ctc_loss=0.0741, att_loss=0.236, loss=0.2036, over 3270086.32 frames. utt_duration=1228 frames, utt_pad_proportion=0.0609, over 10668.55 utterances.], batch size: 44, lr: 5.11e-03, grad_scale: 8.0 2023-03-08 23:01:45,117 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.468e+02 1.926e+02 2.375e+02 3.006e+02 1.037e+03, threshold=4.749e+02, percent-clipped=4.0 2023-03-08 23:02:13,551 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1297, 5.4115, 4.9024, 5.5033, 4.8786, 5.0199, 5.4878, 5.2992], device='cuda:1'), covar=tensor([0.0527, 0.0236, 0.0821, 0.0277, 0.0370, 0.0262, 0.0237, 0.0192], device='cuda:1'), in_proj_covar=tensor([0.0379, 0.0316, 0.0359, 0.0342, 0.0315, 0.0236, 0.0296, 0.0280], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0005, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 23:02:42,048 INFO [train2.py:809] (1/4) Epoch 21, batch 2600, loss[ctc_loss=0.06527, att_loss=0.2391, loss=0.2043, over 16762.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.005938, over 48.00 utterances.], tot_loss[ctc_loss=0.07309, att_loss=0.2349, loss=0.2026, over 3262723.10 frames. utt_duration=1247 frames, utt_pad_proportion=0.05915, over 10480.21 utterances.], batch size: 48, lr: 5.11e-03, grad_scale: 8.0 2023-03-08 23:02:47,106 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2070, 3.7577, 3.2968, 3.4015, 4.0019, 3.6042, 3.1564, 4.3109], device='cuda:1'), covar=tensor([0.0951, 0.0547, 0.1077, 0.0749, 0.0730, 0.0762, 0.0837, 0.0476], device='cuda:1'), in_proj_covar=tensor([0.0199, 0.0214, 0.0224, 0.0198, 0.0276, 0.0239, 0.0198, 0.0283], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 23:03:32,483 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6054, 4.8562, 4.4940, 4.9239, 4.3753, 4.5195, 4.9658, 4.7619], device='cuda:1'), covar=tensor([0.0592, 0.0283, 0.0748, 0.0353, 0.0474, 0.0380, 0.0253, 0.0210], device='cuda:1'), in_proj_covar=tensor([0.0377, 0.0315, 0.0358, 0.0341, 0.0314, 0.0235, 0.0294, 0.0280], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0005, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-08 23:04:03,376 INFO [train2.py:809] (1/4) Epoch 21, batch 2650, loss[ctc_loss=0.06737, att_loss=0.218, loss=0.1879, over 15993.00 frames. utt_duration=1601 frames, utt_pad_proportion=0.008576, over 40.00 utterances.], tot_loss[ctc_loss=0.07298, att_loss=0.2355, loss=0.203, over 3271880.78 frames. utt_duration=1242 frames, utt_pad_proportion=0.05798, over 10551.26 utterances.], batch size: 40, lr: 5.11e-03, grad_scale: 8.0 2023-03-08 23:04:03,720 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82326.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:04:27,512 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.225e+02 1.898e+02 2.205e+02 2.899e+02 5.130e+02, threshold=4.409e+02, percent-clipped=1.0 2023-03-08 23:04:31,581 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82343.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:04:36,577 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82346.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:05:18,875 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.1156, 2.8356, 3.2264, 4.2697, 3.7862, 3.7971, 2.8829, 2.0694], device='cuda:1'), covar=tensor([0.0881, 0.1929, 0.0971, 0.0630, 0.0999, 0.0479, 0.1574, 0.2465], device='cuda:1'), in_proj_covar=tensor([0.0181, 0.0214, 0.0190, 0.0219, 0.0224, 0.0178, 0.0203, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 23:05:21,684 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=82374.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:05:24,743 INFO [train2.py:809] (1/4) Epoch 21, batch 2700, loss[ctc_loss=0.09608, att_loss=0.2543, loss=0.2226, over 17305.00 frames. utt_duration=1260 frames, utt_pad_proportion=0.01112, over 55.00 utterances.], tot_loss[ctc_loss=0.07288, att_loss=0.2351, loss=0.2027, over 3274183.07 frames. utt_duration=1251 frames, utt_pad_proportion=0.05503, over 10482.72 utterances.], batch size: 55, lr: 5.11e-03, grad_scale: 8.0 2023-03-08 23:06:02,743 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.5210, 3.0979, 3.5554, 4.5068, 4.0021, 3.9432, 3.0581, 2.4624], device='cuda:1'), covar=tensor([0.0681, 0.1716, 0.0859, 0.0581, 0.0851, 0.0501, 0.1440, 0.2049], device='cuda:1'), in_proj_covar=tensor([0.0180, 0.0214, 0.0190, 0.0218, 0.0223, 0.0178, 0.0203, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 23:06:16,088 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82407.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:06:22,269 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82411.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:06:46,064 INFO [train2.py:809] (1/4) Epoch 21, batch 2750, loss[ctc_loss=0.05313, att_loss=0.2082, loss=0.1772, over 15644.00 frames. utt_duration=1693 frames, utt_pad_proportion=0.008877, over 37.00 utterances.], tot_loss[ctc_loss=0.0718, att_loss=0.2341, loss=0.2017, over 3269604.11 frames. utt_duration=1268 frames, utt_pad_proportion=0.05017, over 10325.15 utterances.], batch size: 37, lr: 5.11e-03, grad_scale: 8.0 2023-03-08 23:07:07,957 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.5202, 3.1008, 2.6558, 2.8975, 3.2035, 3.0448, 2.4810, 3.0797], device='cuda:1'), covar=tensor([0.0958, 0.0437, 0.0897, 0.0618, 0.0631, 0.0596, 0.0858, 0.0545], device='cuda:1'), in_proj_covar=tensor([0.0201, 0.0216, 0.0226, 0.0200, 0.0277, 0.0241, 0.0200, 0.0286], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 23:07:10,786 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.410e+02 1.903e+02 2.360e+02 2.836e+02 7.135e+02, threshold=4.721e+02, percent-clipped=4.0 2023-03-08 23:07:46,369 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-03-08 23:08:06,987 INFO [train2.py:809] (1/4) Epoch 21, batch 2800, loss[ctc_loss=0.08077, att_loss=0.2502, loss=0.2163, over 17261.00 frames. utt_duration=875.4 frames, utt_pad_proportion=0.08336, over 79.00 utterances.], tot_loss[ctc_loss=0.07381, att_loss=0.2356, loss=0.2033, over 3267004.36 frames. utt_duration=1208 frames, utt_pad_proportion=0.06516, over 10832.32 utterances.], batch size: 79, lr: 5.11e-03, grad_scale: 8.0 2023-03-08 23:08:12,427 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.6338, 2.1621, 2.3109, 2.3798, 2.7939, 2.3631, 2.3782, 3.1279], device='cuda:1'), covar=tensor([0.1433, 0.3530, 0.2122, 0.1920, 0.1603, 0.1828, 0.2869, 0.1140], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0124, 0.0119, 0.0108, 0.0123, 0.0107, 0.0129, 0.0099], device='cuda:1'), out_proj_covar=tensor([8.8015e-05, 9.6208e-05, 9.5184e-05, 8.4507e-05, 9.1126e-05, 8.5594e-05, 9.6768e-05, 7.8965e-05], device='cuda:1') 2023-03-08 23:08:34,718 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82493.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 23:08:37,822 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5316, 2.3916, 5.0169, 3.8780, 3.0208, 4.3102, 4.7374, 4.6177], device='cuda:1'), covar=tensor([0.0258, 0.1731, 0.0157, 0.0907, 0.1729, 0.0247, 0.0161, 0.0251], device='cuda:1'), in_proj_covar=tensor([0.0193, 0.0240, 0.0184, 0.0308, 0.0264, 0.0212, 0.0174, 0.0203], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 23:09:26,649 INFO [train2.py:809] (1/4) Epoch 21, batch 2850, loss[ctc_loss=0.07759, att_loss=0.2198, loss=0.1913, over 11821.00 frames. utt_duration=1820 frames, utt_pad_proportion=0.1769, over 26.00 utterances.], tot_loss[ctc_loss=0.0737, att_loss=0.235, loss=0.2028, over 3262054.55 frames. utt_duration=1222 frames, utt_pad_proportion=0.06224, over 10687.17 utterances.], batch size: 26, lr: 5.11e-03, grad_scale: 8.0 2023-03-08 23:09:50,389 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.302e+02 1.879e+02 2.361e+02 2.971e+02 6.221e+02, threshold=4.723e+02, percent-clipped=6.0 2023-03-08 23:10:09,184 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.87 vs. limit=2.0 2023-03-08 23:10:11,731 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82554.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 23:10:35,784 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82569.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:10:46,927 INFO [train2.py:809] (1/4) Epoch 21, batch 2900, loss[ctc_loss=0.05655, att_loss=0.2204, loss=0.1876, over 16265.00 frames. utt_duration=1515 frames, utt_pad_proportion=0.008117, over 43.00 utterances.], tot_loss[ctc_loss=0.0729, att_loss=0.2341, loss=0.2018, over 3258555.90 frames. utt_duration=1247 frames, utt_pad_proportion=0.05592, over 10466.17 utterances.], batch size: 43, lr: 5.10e-03, grad_scale: 8.0 2023-03-08 23:11:54,937 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9643, 5.2371, 5.5348, 5.3130, 5.4463, 5.9049, 5.1771, 6.0136], device='cuda:1'), covar=tensor([0.0650, 0.0727, 0.0835, 0.1481, 0.1783, 0.0909, 0.0743, 0.0620], device='cuda:1'), in_proj_covar=tensor([0.0870, 0.0511, 0.0597, 0.0663, 0.0873, 0.0630, 0.0487, 0.0610], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 23:12:06,192 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-03-08 23:12:07,976 INFO [train2.py:809] (1/4) Epoch 21, batch 2950, loss[ctc_loss=0.05714, att_loss=0.205, loss=0.1754, over 15624.00 frames. utt_duration=1691 frames, utt_pad_proportion=0.00903, over 37.00 utterances.], tot_loss[ctc_loss=0.07207, att_loss=0.2339, loss=0.2015, over 3261805.18 frames. utt_duration=1263 frames, utt_pad_proportion=0.05217, over 10343.42 utterances.], batch size: 37, lr: 5.10e-03, grad_scale: 8.0 2023-03-08 23:12:11,625 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3790, 2.5152, 4.7697, 3.8431, 2.9084, 4.0927, 4.4225, 4.4024], device='cuda:1'), covar=tensor([0.0215, 0.1604, 0.0137, 0.0871, 0.1719, 0.0269, 0.0184, 0.0259], device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0236, 0.0181, 0.0303, 0.0259, 0.0209, 0.0171, 0.0200], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 23:12:14,620 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82630.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:12:26,293 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7261, 5.0928, 4.9537, 5.0745, 5.1257, 4.8484, 3.5234, 5.0812], device='cuda:1'), covar=tensor([0.0123, 0.0123, 0.0141, 0.0085, 0.0114, 0.0119, 0.0748, 0.0213], device='cuda:1'), in_proj_covar=tensor([0.0093, 0.0088, 0.0112, 0.0070, 0.0076, 0.0087, 0.0105, 0.0109], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 23:12:31,985 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.421e+02 1.787e+02 2.245e+02 2.585e+02 7.263e+02, threshold=4.491e+02, percent-clipped=3.0 2023-03-08 23:12:36,023 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82643.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:12:44,193 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.50 vs. limit=5.0 2023-03-08 23:13:27,967 INFO [train2.py:809] (1/4) Epoch 21, batch 3000, loss[ctc_loss=0.07254, att_loss=0.2217, loss=0.1919, over 15487.00 frames. utt_duration=1722 frames, utt_pad_proportion=0.008982, over 36.00 utterances.], tot_loss[ctc_loss=0.07243, att_loss=0.2344, loss=0.202, over 3269649.94 frames. utt_duration=1294 frames, utt_pad_proportion=0.04393, over 10121.72 utterances.], batch size: 36, lr: 5.10e-03, grad_scale: 8.0 2023-03-08 23:13:27,967 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-08 23:13:41,812 INFO [train2.py:843] (1/4) Epoch 21, validation: ctc_loss=0.04141, att_loss=0.2346, loss=0.196, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-08 23:13:41,813 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-08 23:14:01,655 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.8316, 4.4326, 4.5120, 2.3641, 2.1038, 2.8398, 2.1667, 3.6796], device='cuda:1'), covar=tensor([0.0734, 0.0307, 0.0253, 0.4379, 0.5150, 0.2307, 0.3858, 0.1484], device='cuda:1'), in_proj_covar=tensor([0.0357, 0.0275, 0.0270, 0.0245, 0.0344, 0.0333, 0.0252, 0.0365], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-03-08 23:14:05,949 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=82691.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:14:24,600 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82702.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:14:38,811 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82711.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:15:02,603 INFO [train2.py:809] (1/4) Epoch 21, batch 3050, loss[ctc_loss=0.06716, att_loss=0.2435, loss=0.2082, over 16883.00 frames. utt_duration=1380 frames, utt_pad_proportion=0.006759, over 49.00 utterances.], tot_loss[ctc_loss=0.07229, att_loss=0.2343, loss=0.2019, over 3272490.07 frames. utt_duration=1287 frames, utt_pad_proportion=0.04461, over 10184.11 utterances.], batch size: 49, lr: 5.10e-03, grad_scale: 16.0 2023-03-08 23:15:26,171 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 8.870e+01 1.786e+02 2.187e+02 2.749e+02 5.841e+02, threshold=4.374e+02, percent-clipped=4.0 2023-03-08 23:15:55,319 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=82759.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:16:22,506 INFO [train2.py:809] (1/4) Epoch 21, batch 3100, loss[ctc_loss=0.08008, att_loss=0.2349, loss=0.204, over 16541.00 frames. utt_duration=1472 frames, utt_pad_proportion=0.006332, over 45.00 utterances.], tot_loss[ctc_loss=0.07267, att_loss=0.2348, loss=0.2024, over 3275388.45 frames. utt_duration=1264 frames, utt_pad_proportion=0.04924, over 10377.87 utterances.], batch size: 45, lr: 5.10e-03, grad_scale: 16.0 2023-03-08 23:16:32,912 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.3966, 1.7634, 2.0740, 2.2373, 2.2035, 2.1623, 1.7527, 2.4642], device='cuda:1'), covar=tensor([0.1659, 0.2441, 0.1879, 0.1530, 0.2211, 0.1298, 0.1865, 0.1073], device='cuda:1'), in_proj_covar=tensor([0.0117, 0.0125, 0.0119, 0.0108, 0.0123, 0.0106, 0.0129, 0.0098], device='cuda:1'), out_proj_covar=tensor([8.8457e-05, 9.6475e-05, 9.5032e-05, 8.4792e-05, 9.1137e-05, 8.4991e-05, 9.6670e-05, 7.8763e-05], device='cuda:1') 2023-03-08 23:17:43,780 INFO [train2.py:809] (1/4) Epoch 21, batch 3150, loss[ctc_loss=0.1004, att_loss=0.264, loss=0.2312, over 17380.00 frames. utt_duration=1180 frames, utt_pad_proportion=0.01863, over 59.00 utterances.], tot_loss[ctc_loss=0.07362, att_loss=0.2359, loss=0.2034, over 3273851.49 frames. utt_duration=1256 frames, utt_pad_proportion=0.05263, over 10434.77 utterances.], batch size: 59, lr: 5.10e-03, grad_scale: 8.0 2023-03-08 23:18:09,383 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.159e+02 2.205e+02 2.472e+02 3.065e+02 6.005e+02, threshold=4.943e+02, percent-clipped=5.0 2023-03-08 23:18:21,029 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82849.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 23:19:03,809 INFO [train2.py:809] (1/4) Epoch 21, batch 3200, loss[ctc_loss=0.09995, att_loss=0.2515, loss=0.2212, over 16409.00 frames. utt_duration=1493 frames, utt_pad_proportion=0.007208, over 44.00 utterances.], tot_loss[ctc_loss=0.07365, att_loss=0.2356, loss=0.2032, over 3273683.67 frames. utt_duration=1283 frames, utt_pad_proportion=0.04603, over 10219.03 utterances.], batch size: 44, lr: 5.09e-03, grad_scale: 8.0 2023-03-08 23:19:54,111 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.92 vs. limit=2.0 2023-03-08 23:20:01,435 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.6179, 3.4354, 3.4246, 3.0016, 3.4105, 3.4323, 3.4640, 2.5228], device='cuda:1'), covar=tensor([0.1290, 0.1281, 0.1851, 0.2951, 0.2134, 0.2770, 0.0923, 0.4000], device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0187, 0.0198, 0.0251, 0.0156, 0.0257, 0.0178, 0.0214], 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') 2023-03-08 23:20:13,550 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.8728, 3.8192, 3.0015, 3.1997, 3.9127, 3.5940, 2.6226, 4.1138], device='cuda:1'), covar=tensor([0.1138, 0.0496, 0.1132, 0.0802, 0.0750, 0.0739, 0.1106, 0.0482], device='cuda:1'), in_proj_covar=tensor([0.0202, 0.0215, 0.0225, 0.0197, 0.0275, 0.0239, 0.0198, 0.0281], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 23:20:23,221 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82925.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:20:24,596 INFO [train2.py:809] (1/4) Epoch 21, batch 3250, loss[ctc_loss=0.08321, att_loss=0.2556, loss=0.2211, over 17320.00 frames. utt_duration=1261 frames, utt_pad_proportion=0.01097, over 55.00 utterances.], tot_loss[ctc_loss=0.07359, att_loss=0.2363, loss=0.2038, over 3288357.59 frames. utt_duration=1286 frames, utt_pad_proportion=0.04134, over 10237.29 utterances.], batch size: 55, lr: 5.09e-03, grad_scale: 8.0 2023-03-08 23:20:39,202 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.1622, 5.4212, 5.7352, 5.4208, 5.6586, 6.0781, 5.3691, 6.1648], device='cuda:1'), covar=tensor([0.0650, 0.0755, 0.0754, 0.1306, 0.1643, 0.0905, 0.0651, 0.0646], device='cuda:1'), in_proj_covar=tensor([0.0863, 0.0508, 0.0593, 0.0655, 0.0866, 0.0623, 0.0483, 0.0604], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 23:20:50,592 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.258e+02 1.807e+02 2.273e+02 2.911e+02 7.187e+02, threshold=4.546e+02, percent-clipped=3.0 2023-03-08 23:21:02,212 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.83 vs. limit=5.0 2023-03-08 23:21:24,881 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82963.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:21:45,459 INFO [train2.py:809] (1/4) Epoch 21, batch 3300, loss[ctc_loss=0.06033, att_loss=0.2209, loss=0.1888, over 14605.00 frames. utt_duration=1827 frames, utt_pad_proportion=0.03551, over 32.00 utterances.], tot_loss[ctc_loss=0.0729, att_loss=0.2359, loss=0.2033, over 3286948.18 frames. utt_duration=1279 frames, utt_pad_proportion=0.04407, over 10294.21 utterances.], batch size: 32, lr: 5.09e-03, grad_scale: 8.0 2023-03-08 23:21:54,720 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.20 vs. limit=5.0 2023-03-08 23:22:28,793 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83002.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:23:04,105 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83024.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:23:07,439 INFO [train2.py:809] (1/4) Epoch 21, batch 3350, loss[ctc_loss=0.05024, att_loss=0.2194, loss=0.1856, over 16128.00 frames. utt_duration=1538 frames, utt_pad_proportion=0.006018, over 42.00 utterances.], tot_loss[ctc_loss=0.0735, att_loss=0.2366, loss=0.204, over 3291202.60 frames. utt_duration=1251 frames, utt_pad_proportion=0.04901, over 10537.28 utterances.], batch size: 42, lr: 5.09e-03, grad_scale: 8.0 2023-03-08 23:23:33,602 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.287e+02 1.819e+02 2.175e+02 2.592e+02 7.580e+02, threshold=4.350e+02, percent-clipped=2.0 2023-03-08 23:23:46,015 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=83050.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:24:28,079 INFO [train2.py:809] (1/4) Epoch 21, batch 3400, loss[ctc_loss=0.09279, att_loss=0.2588, loss=0.2256, over 17254.00 frames. utt_duration=1257 frames, utt_pad_proportion=0.0137, over 55.00 utterances.], tot_loss[ctc_loss=0.07344, att_loss=0.2369, loss=0.2042, over 3285824.37 frames. utt_duration=1236 frames, utt_pad_proportion=0.05461, over 10648.32 utterances.], batch size: 55, lr: 5.09e-03, grad_scale: 8.0 2023-03-08 23:24:43,758 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7693, 2.4288, 2.6495, 3.2521, 3.0963, 3.2086, 2.5945, 2.2899], device='cuda:1'), covar=tensor([0.0747, 0.1799, 0.0955, 0.0865, 0.0856, 0.0512, 0.1329, 0.1728], device='cuda:1'), in_proj_covar=tensor([0.0180, 0.0213, 0.0189, 0.0217, 0.0221, 0.0177, 0.0200, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 23:25:47,389 INFO [train2.py:809] (1/4) Epoch 21, batch 3450, loss[ctc_loss=0.07543, att_loss=0.25, loss=0.2151, over 17302.00 frames. utt_duration=1175 frames, utt_pad_proportion=0.02366, over 59.00 utterances.], tot_loss[ctc_loss=0.0732, att_loss=0.2363, loss=0.2037, over 3284618.01 frames. utt_duration=1250 frames, utt_pad_proportion=0.05142, over 10521.11 utterances.], batch size: 59, lr: 5.09e-03, grad_scale: 8.0 2023-03-08 23:26:03,450 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.8725, 5.2805, 5.4432, 5.2008, 5.3603, 5.8084, 5.2400, 5.8968], device='cuda:1'), covar=tensor([0.0802, 0.0802, 0.0826, 0.1513, 0.1915, 0.1038, 0.0755, 0.0776], device='cuda:1'), in_proj_covar=tensor([0.0876, 0.0514, 0.0601, 0.0664, 0.0878, 0.0632, 0.0489, 0.0612], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 23:26:13,619 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.358e+02 1.851e+02 2.238e+02 2.723e+02 5.748e+02, threshold=4.475e+02, percent-clipped=3.0 2023-03-08 23:26:24,490 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83149.0, num_to_drop=1, layers_to_drop={1} 2023-03-08 23:27:06,900 INFO [train2.py:809] (1/4) Epoch 21, batch 3500, loss[ctc_loss=0.08692, att_loss=0.2277, loss=0.1995, over 16520.00 frames. utt_duration=1470 frames, utt_pad_proportion=0.007442, over 45.00 utterances.], tot_loss[ctc_loss=0.07307, att_loss=0.2355, loss=0.203, over 3283275.42 frames. utt_duration=1259 frames, utt_pad_proportion=0.05051, over 10442.59 utterances.], batch size: 45, lr: 5.09e-03, grad_scale: 8.0 2023-03-08 23:27:39,120 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83196.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:27:40,349 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=83197.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 23:27:45,443 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.8875, 3.6813, 3.6105, 3.1258, 3.7030, 3.7100, 3.7235, 2.7652], device='cuda:1'), covar=tensor([0.1182, 0.1299, 0.1974, 0.3154, 0.2926, 0.2038, 0.0870, 0.3533], device='cuda:1'), in_proj_covar=tensor([0.0172, 0.0185, 0.0197, 0.0249, 0.0156, 0.0256, 0.0176, 0.0214], 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') 2023-03-08 23:28:24,869 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.14 vs. limit=2.0 2023-03-08 23:28:25,797 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83225.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:28:26,902 INFO [train2.py:809] (1/4) Epoch 21, batch 3550, loss[ctc_loss=0.07813, att_loss=0.2286, loss=0.1985, over 16256.00 frames. utt_duration=1514 frames, utt_pad_proportion=0.008757, over 43.00 utterances.], tot_loss[ctc_loss=0.0728, att_loss=0.2356, loss=0.203, over 3284092.60 frames. utt_duration=1275 frames, utt_pad_proportion=0.04533, over 10315.93 utterances.], batch size: 43, lr: 5.08e-03, grad_scale: 8.0 2023-03-08 23:28:53,067 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.215e+02 1.836e+02 2.204e+02 2.637e+02 6.921e+02, threshold=4.407e+02, percent-clipped=5.0 2023-03-08 23:29:16,524 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83257.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:29:41,893 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=83273.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:29:47,129 INFO [train2.py:809] (1/4) Epoch 21, batch 3600, loss[ctc_loss=0.06753, att_loss=0.2296, loss=0.1972, over 15966.00 frames. utt_duration=1559 frames, utt_pad_proportion=0.005755, over 41.00 utterances.], tot_loss[ctc_loss=0.07253, att_loss=0.2354, loss=0.2028, over 3285498.20 frames. utt_duration=1281 frames, utt_pad_proportion=0.04371, over 10268.70 utterances.], batch size: 41, lr: 5.08e-03, grad_scale: 8.0 2023-03-08 23:30:00,118 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83284.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:30:02,701 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-03-08 23:30:56,572 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=83319.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:31:07,604 INFO [train2.py:809] (1/4) Epoch 21, batch 3650, loss[ctc_loss=0.04969, att_loss=0.2132, loss=0.1805, over 16185.00 frames. utt_duration=1581 frames, utt_pad_proportion=0.006439, over 41.00 utterances.], tot_loss[ctc_loss=0.07322, att_loss=0.2355, loss=0.203, over 3278041.98 frames. utt_duration=1257 frames, utt_pad_proportion=0.05094, over 10444.88 utterances.], batch size: 41, lr: 5.08e-03, grad_scale: 8.0 2023-03-08 23:31:33,948 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.181e+02 2.085e+02 2.475e+02 3.033e+02 6.314e+02, threshold=4.951e+02, percent-clipped=5.0 2023-03-08 23:31:39,069 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83345.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:32:27,576 INFO [train2.py:809] (1/4) Epoch 21, batch 3700, loss[ctc_loss=0.07937, att_loss=0.2533, loss=0.2185, over 16886.00 frames. utt_duration=1380 frames, utt_pad_proportion=0.007238, over 49.00 utterances.], tot_loss[ctc_loss=0.07374, att_loss=0.2359, loss=0.2034, over 3275248.00 frames. utt_duration=1230 frames, utt_pad_proportion=0.05688, over 10666.43 utterances.], batch size: 49, lr: 5.08e-03, grad_scale: 8.0 2023-03-08 23:32:29,488 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9198, 5.2547, 5.4873, 5.2967, 5.3922, 5.8878, 5.2596, 5.9425], device='cuda:1'), covar=tensor([0.0759, 0.0794, 0.0813, 0.1450, 0.1801, 0.0987, 0.0667, 0.0755], device='cuda:1'), in_proj_covar=tensor([0.0866, 0.0511, 0.0596, 0.0659, 0.0869, 0.0625, 0.0482, 0.0602], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 23:32:59,868 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.0174, 3.8916, 3.2581, 3.5178, 4.0211, 3.6864, 3.1094, 4.4317], device='cuda:1'), covar=tensor([0.1030, 0.0441, 0.1069, 0.0713, 0.0699, 0.0723, 0.0819, 0.0392], device='cuda:1'), in_proj_covar=tensor([0.0201, 0.0215, 0.0224, 0.0197, 0.0275, 0.0238, 0.0198, 0.0281], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 23:33:49,258 INFO [train2.py:809] (1/4) Epoch 21, batch 3750, loss[ctc_loss=0.05928, att_loss=0.2115, loss=0.1811, over 16258.00 frames. utt_duration=1514 frames, utt_pad_proportion=0.008772, over 43.00 utterances.], tot_loss[ctc_loss=0.07355, att_loss=0.236, loss=0.2035, over 3277670.45 frames. utt_duration=1245 frames, utt_pad_proportion=0.05263, over 10542.08 utterances.], batch size: 43, lr: 5.08e-03, grad_scale: 8.0 2023-03-08 23:34:03,559 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.43 vs. limit=5.0 2023-03-08 23:34:15,117 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.225e+02 1.925e+02 2.407e+02 3.142e+02 5.755e+02, threshold=4.814e+02, percent-clipped=4.0 2023-03-08 23:34:23,525 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-03-08 23:34:32,972 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-03-08 23:34:47,840 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.09 vs. limit=5.0 2023-03-08 23:35:09,002 INFO [train2.py:809] (1/4) Epoch 21, batch 3800, loss[ctc_loss=0.0763, att_loss=0.2353, loss=0.2035, over 16463.00 frames. utt_duration=1433 frames, utt_pad_proportion=0.007527, over 46.00 utterances.], tot_loss[ctc_loss=0.07408, att_loss=0.2365, loss=0.204, over 3279230.14 frames. utt_duration=1247 frames, utt_pad_proportion=0.05287, over 10529.08 utterances.], batch size: 46, lr: 5.08e-03, grad_scale: 8.0 2023-03-08 23:36:06,616 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.28 vs. limit=5.0 2023-03-08 23:36:29,156 INFO [train2.py:809] (1/4) Epoch 21, batch 3850, loss[ctc_loss=0.07927, att_loss=0.2364, loss=0.205, over 16278.00 frames. utt_duration=1516 frames, utt_pad_proportion=0.007356, over 43.00 utterances.], tot_loss[ctc_loss=0.07438, att_loss=0.2371, loss=0.2045, over 3289774.75 frames. utt_duration=1244 frames, utt_pad_proportion=0.04948, over 10591.68 utterances.], batch size: 43, lr: 5.07e-03, grad_scale: 8.0 2023-03-08 23:36:53,728 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.289e+02 2.030e+02 2.457e+02 3.139e+02 8.463e+02, threshold=4.915e+02, percent-clipped=7.0 2023-03-08 23:37:09,547 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=83552.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:37:46,943 INFO [train2.py:809] (1/4) Epoch 21, batch 3900, loss[ctc_loss=0.09086, att_loss=0.2445, loss=0.2138, over 17373.00 frames. utt_duration=1104 frames, utt_pad_proportion=0.0339, over 63.00 utterances.], tot_loss[ctc_loss=0.07429, att_loss=0.2373, loss=0.2047, over 3286893.96 frames. utt_duration=1234 frames, utt_pad_proportion=0.0537, over 10668.28 utterances.], batch size: 63, lr: 5.07e-03, grad_scale: 8.0 2023-03-08 23:38:15,622 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5655, 2.9883, 3.7395, 2.9661, 3.6658, 4.6858, 4.5173, 3.2710], device='cuda:1'), covar=tensor([0.0383, 0.1629, 0.1167, 0.1404, 0.0993, 0.0793, 0.0550, 0.1287], device='cuda:1'), in_proj_covar=tensor([0.0247, 0.0243, 0.0280, 0.0220, 0.0264, 0.0366, 0.0259, 0.0232], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 23:38:39,851 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5556, 4.4489, 4.6434, 4.5710, 5.1702, 4.5465, 4.5707, 2.6271], device='cuda:1'), covar=tensor([0.0283, 0.0411, 0.0361, 0.0386, 0.0907, 0.0281, 0.0372, 0.1775], device='cuda:1'), in_proj_covar=tensor([0.0167, 0.0191, 0.0189, 0.0208, 0.0373, 0.0159, 0.0179, 0.0217], 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') 2023-03-08 23:38:49,273 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.4531, 2.2485, 2.0237, 2.3743, 2.7336, 2.5824, 2.1209, 3.1052], device='cuda:1'), covar=tensor([0.1926, 0.3027, 0.2512, 0.1574, 0.1822, 0.1239, 0.2507, 0.1081], device='cuda:1'), in_proj_covar=tensor([0.0118, 0.0125, 0.0123, 0.0110, 0.0124, 0.0106, 0.0130, 0.0100], device='cuda:1'), out_proj_covar=tensor([8.9275e-05, 9.7028e-05, 9.7276e-05, 8.6032e-05, 9.2198e-05, 8.5438e-05, 9.7994e-05, 8.0134e-05], device='cuda:1') 2023-03-08 23:38:55,370 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83619.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:38:56,814 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.7162, 3.9862, 3.9376, 3.9426, 4.0514, 3.7904, 2.9506, 3.9289], device='cuda:1'), covar=tensor([0.0151, 0.0118, 0.0129, 0.0101, 0.0094, 0.0135, 0.0682, 0.0190], device='cuda:1'), in_proj_covar=tensor([0.0091, 0.0086, 0.0110, 0.0069, 0.0075, 0.0085, 0.0103, 0.0107], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 23:39:05,539 INFO [train2.py:809] (1/4) Epoch 21, batch 3950, loss[ctc_loss=0.06246, att_loss=0.2121, loss=0.1822, over 15762.00 frames. utt_duration=1660 frames, utt_pad_proportion=0.007503, over 38.00 utterances.], tot_loss[ctc_loss=0.07417, att_loss=0.2372, loss=0.2046, over 3291707.04 frames. utt_duration=1248 frames, utt_pad_proportion=0.04992, over 10560.80 utterances.], batch size: 38, lr: 5.07e-03, grad_scale: 8.0 2023-03-08 23:39:27,335 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=83640.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:39:30,215 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.299e+02 1.844e+02 2.165e+02 2.793e+02 5.808e+02, threshold=4.331e+02, percent-clipped=4.0 2023-03-08 23:40:16,718 INFO [train2.py:809] (1/4) Epoch 22, batch 0, loss[ctc_loss=0.08892, att_loss=0.2473, loss=0.2156, over 16765.00 frames. utt_duration=1398 frames, utt_pad_proportion=0.007028, over 48.00 utterances.], tot_loss[ctc_loss=0.08892, att_loss=0.2473, loss=0.2156, over 16765.00 frames. utt_duration=1398 frames, utt_pad_proportion=0.007028, over 48.00 utterances.], batch size: 48, lr: 4.95e-03, grad_scale: 8.0 2023-03-08 23:40:16,718 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-08 23:40:29,654 INFO [train2.py:843] (1/4) Epoch 22, validation: ctc_loss=0.04004, att_loss=0.2341, loss=0.1953, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-08 23:40:29,655 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-08 23:40:40,583 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0013, 5.3033, 5.2306, 5.1289, 5.2859, 5.2415, 4.9117, 4.6919], device='cuda:1'), covar=tensor([0.0974, 0.0483, 0.0280, 0.0594, 0.0315, 0.0352, 0.0445, 0.0404], device='cuda:1'), in_proj_covar=tensor([0.0516, 0.0362, 0.0342, 0.0356, 0.0418, 0.0429, 0.0353, 0.0392], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0003, 0.0004], device='cuda:1') 2023-03-08 23:40:41,981 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=83667.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:41:48,591 INFO [train2.py:809] (1/4) Epoch 22, batch 50, loss[ctc_loss=0.07297, att_loss=0.2253, loss=0.1948, over 16176.00 frames. utt_duration=1580 frames, utt_pad_proportion=0.006504, over 41.00 utterances.], tot_loss[ctc_loss=0.07439, att_loss=0.2362, loss=0.2039, over 739169.18 frames. utt_duration=1254 frames, utt_pad_proportion=0.0494, over 2360.69 utterances.], batch size: 41, lr: 4.95e-03, grad_scale: 8.0 2023-03-08 23:42:40,552 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.187e+02 1.991e+02 2.354e+02 2.967e+02 6.077e+02, threshold=4.708e+02, percent-clipped=4.0 2023-03-08 23:42:43,902 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.8647, 6.1324, 5.6559, 5.9335, 5.8324, 5.3153, 5.5412, 5.3700], device='cuda:1'), covar=tensor([0.1388, 0.0940, 0.0907, 0.0788, 0.0906, 0.1693, 0.2450, 0.2348], device='cuda:1'), in_proj_covar=tensor([0.0522, 0.0602, 0.0455, 0.0452, 0.0422, 0.0463, 0.0602, 0.0518], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:1') 2023-03-08 23:43:01,756 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1204, 4.4316, 4.3957, 4.3949, 4.4996, 4.2370, 3.0731, 4.3518], device='cuda:1'), covar=tensor([0.0156, 0.0143, 0.0130, 0.0114, 0.0112, 0.0126, 0.0775, 0.0253], device='cuda:1'), in_proj_covar=tensor([0.0091, 0.0086, 0.0109, 0.0069, 0.0075, 0.0085, 0.0103, 0.0107], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 23:43:08,482 INFO [train2.py:809] (1/4) Epoch 22, batch 100, loss[ctc_loss=0.07384, att_loss=0.2488, loss=0.2138, over 16767.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.006673, over 48.00 utterances.], tot_loss[ctc_loss=0.07475, att_loss=0.237, loss=0.2045, over 1293960.28 frames. utt_duration=1242 frames, utt_pad_proportion=0.05672, over 4170.93 utterances.], batch size: 48, lr: 4.95e-03, grad_scale: 8.0 2023-03-08 23:43:10,921 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83760.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:43:48,882 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9453, 5.1390, 5.1531, 5.0550, 5.1981, 5.1599, 4.8525, 4.6409], device='cuda:1'), covar=tensor([0.0999, 0.0556, 0.0270, 0.0590, 0.0316, 0.0326, 0.0377, 0.0386], device='cuda:1'), in_proj_covar=tensor([0.0516, 0.0361, 0.0343, 0.0355, 0.0418, 0.0429, 0.0353, 0.0391], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0003, 0.0004], device='cuda:1') 2023-03-08 23:44:29,205 INFO [train2.py:809] (1/4) Epoch 22, batch 150, loss[ctc_loss=0.05406, att_loss=0.2289, loss=0.194, over 16865.00 frames. utt_duration=1378 frames, utt_pad_proportion=0.007626, over 49.00 utterances.], tot_loss[ctc_loss=0.07469, att_loss=0.2369, loss=0.2044, over 1734569.42 frames. utt_duration=1257 frames, utt_pad_proportion=0.05227, over 5525.75 utterances.], batch size: 49, lr: 4.95e-03, grad_scale: 8.0 2023-03-08 23:44:48,389 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83821.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:45:21,311 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.349e+02 1.932e+02 2.277e+02 2.985e+02 7.139e+02, threshold=4.555e+02, percent-clipped=2.0 2023-03-08 23:45:37,560 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83852.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:45:37,618 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.1655, 3.9044, 3.3349, 3.4667, 4.0826, 3.7067, 3.0110, 4.3211], device='cuda:1'), covar=tensor([0.0948, 0.0473, 0.0909, 0.0657, 0.0615, 0.0691, 0.0875, 0.0465], device='cuda:1'), in_proj_covar=tensor([0.0203, 0.0215, 0.0227, 0.0198, 0.0277, 0.0241, 0.0199, 0.0285], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 23:45:49,304 INFO [train2.py:809] (1/4) Epoch 22, batch 200, loss[ctc_loss=0.07283, att_loss=0.2504, loss=0.2149, over 17047.00 frames. utt_duration=1288 frames, utt_pad_proportion=0.009978, over 53.00 utterances.], tot_loss[ctc_loss=0.0747, att_loss=0.2361, loss=0.2038, over 2067282.74 frames. utt_duration=1229 frames, utt_pad_proportion=0.06366, over 6736.90 utterances.], batch size: 53, lr: 4.95e-03, grad_scale: 8.0 2023-03-08 23:46:28,878 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.8089, 5.1336, 5.3868, 5.1795, 5.3043, 5.7078, 5.1250, 5.8405], device='cuda:1'), covar=tensor([0.0778, 0.0867, 0.0854, 0.1324, 0.1833, 0.1082, 0.0839, 0.0672], device='cuda:1'), in_proj_covar=tensor([0.0881, 0.0517, 0.0599, 0.0669, 0.0878, 0.0631, 0.0490, 0.0609], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 23:46:53,864 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=83900.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:46:55,864 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.9633, 3.7692, 3.1305, 3.3383, 3.9442, 3.5955, 2.8546, 4.2034], device='cuda:1'), covar=tensor([0.1088, 0.0561, 0.1118, 0.0721, 0.0729, 0.0746, 0.0951, 0.0484], device='cuda:1'), in_proj_covar=tensor([0.0204, 0.0217, 0.0229, 0.0199, 0.0279, 0.0243, 0.0200, 0.0288], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-08 23:47:08,781 INFO [train2.py:809] (1/4) Epoch 22, batch 250, loss[ctc_loss=0.09759, att_loss=0.2579, loss=0.2258, over 17060.00 frames. utt_duration=1314 frames, utt_pad_proportion=0.008345, over 52.00 utterances.], tot_loss[ctc_loss=0.07461, att_loss=0.2362, loss=0.2038, over 2339565.57 frames. utt_duration=1255 frames, utt_pad_proportion=0.05464, over 7464.59 utterances.], batch size: 52, lr: 4.94e-03, grad_scale: 8.0 2023-03-08 23:47:57,195 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83940.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:47:59,820 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.391e+02 1.880e+02 2.225e+02 2.668e+02 4.764e+02, threshold=4.450e+02, percent-clipped=1.0 2023-03-08 23:48:11,496 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83949.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:48:27,878 INFO [train2.py:809] (1/4) Epoch 22, batch 300, loss[ctc_loss=0.08678, att_loss=0.2462, loss=0.2143, over 17048.00 frames. utt_duration=1288 frames, utt_pad_proportion=0.01005, over 53.00 utterances.], tot_loss[ctc_loss=0.07478, att_loss=0.2364, loss=0.2041, over 2547646.92 frames. utt_duration=1245 frames, utt_pad_proportion=0.05768, over 8196.38 utterances.], batch size: 53, lr: 4.94e-03, grad_scale: 8.0 2023-03-08 23:48:44,312 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.02 vs. limit=5.0 2023-03-08 23:48:46,391 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-03-08 23:49:09,339 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83985.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:49:13,746 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=83988.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:49:51,795 INFO [train2.py:809] (1/4) Epoch 22, batch 350, loss[ctc_loss=0.06784, att_loss=0.2313, loss=0.1986, over 16135.00 frames. utt_duration=1538 frames, utt_pad_proportion=0.005175, over 42.00 utterances.], tot_loss[ctc_loss=0.0742, att_loss=0.2361, loss=0.2037, over 2713457.20 frames. utt_duration=1243 frames, utt_pad_proportion=0.05636, over 8739.83 utterances.], batch size: 42, lr: 4.94e-03, grad_scale: 8.0 2023-03-08 23:49:53,711 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84010.0, num_to_drop=1, layers_to_drop={0} 2023-03-08 23:49:59,912 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5347, 3.0094, 3.7368, 3.0785, 3.6341, 4.6919, 4.4933, 3.2820], device='cuda:1'), covar=tensor([0.0372, 0.1626, 0.1142, 0.1324, 0.1071, 0.0714, 0.0531, 0.1240], device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0242, 0.0281, 0.0219, 0.0264, 0.0366, 0.0259, 0.0231], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 23:50:06,547 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.85 vs. limit=2.0 2023-03-08 23:50:43,093 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.569e+02 1.981e+02 2.404e+02 2.952e+02 5.790e+02, threshold=4.808e+02, percent-clipped=6.0 2023-03-08 23:50:50,296 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84046.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:50:51,785 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6628, 3.0481, 3.7261, 3.2537, 3.6944, 4.7190, 4.5356, 3.3791], device='cuda:1'), covar=tensor([0.0281, 0.1627, 0.1324, 0.1234, 0.1022, 0.0826, 0.0491, 0.1196], device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0243, 0.0282, 0.0219, 0.0264, 0.0368, 0.0260, 0.0232], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 23:51:11,737 INFO [train2.py:809] (1/4) Epoch 22, batch 400, loss[ctc_loss=0.07318, att_loss=0.2223, loss=0.1925, over 15886.00 frames. utt_duration=1631 frames, utt_pad_proportion=0.009331, over 39.00 utterances.], tot_loss[ctc_loss=0.07361, att_loss=0.2357, loss=0.2033, over 2838811.67 frames. utt_duration=1244 frames, utt_pad_proportion=0.05497, over 9139.94 utterances.], batch size: 39, lr: 4.94e-03, grad_scale: 8.0 2023-03-08 23:51:57,486 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7113, 5.0352, 4.9095, 5.0066, 5.1863, 4.7159, 3.7981, 5.0516], device='cuda:1'), covar=tensor([0.0113, 0.0116, 0.0123, 0.0075, 0.0088, 0.0115, 0.0588, 0.0179], device='cuda:1'), in_proj_covar=tensor([0.0091, 0.0086, 0.0109, 0.0069, 0.0075, 0.0084, 0.0103, 0.0107], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 23:52:12,381 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84097.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:52:31,747 INFO [train2.py:809] (1/4) Epoch 22, batch 450, loss[ctc_loss=0.07804, att_loss=0.235, loss=0.2036, over 16271.00 frames. utt_duration=1515 frames, utt_pad_proportion=0.007315, over 43.00 utterances.], tot_loss[ctc_loss=0.07325, att_loss=0.2348, loss=0.2025, over 2932347.45 frames. utt_duration=1279 frames, utt_pad_proportion=0.04735, over 9183.14 utterances.], batch size: 43, lr: 4.94e-03, grad_scale: 8.0 2023-03-08 23:52:42,546 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84116.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:53:22,924 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.259e+02 1.943e+02 2.393e+02 2.892e+02 4.484e+02, threshold=4.787e+02, percent-clipped=0.0 2023-03-08 23:53:39,937 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-03-08 23:53:50,182 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84158.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:53:51,320 INFO [train2.py:809] (1/4) Epoch 22, batch 500, loss[ctc_loss=0.07102, att_loss=0.2359, loss=0.2029, over 16542.00 frames. utt_duration=1472 frames, utt_pad_proportion=0.006287, over 45.00 utterances.], tot_loss[ctc_loss=0.0735, att_loss=0.2353, loss=0.2029, over 3016701.66 frames. utt_duration=1292 frames, utt_pad_proportion=0.04166, over 9353.03 utterances.], batch size: 45, lr: 4.94e-03, grad_scale: 8.0 2023-03-08 23:54:55,807 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84199.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:55:12,694 INFO [train2.py:809] (1/4) Epoch 22, batch 550, loss[ctc_loss=0.05638, att_loss=0.2124, loss=0.1812, over 15891.00 frames. utt_duration=1631 frames, utt_pad_proportion=0.008833, over 39.00 utterances.], tot_loss[ctc_loss=0.07342, att_loss=0.2351, loss=0.2027, over 3067396.95 frames. utt_duration=1268 frames, utt_pad_proportion=0.04934, over 9686.13 utterances.], batch size: 39, lr: 4.94e-03, grad_scale: 8.0 2023-03-08 23:55:55,708 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84236.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:56:04,933 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.221e+02 1.935e+02 2.218e+02 2.633e+02 6.667e+02, threshold=4.436e+02, percent-clipped=1.0 2023-03-08 23:56:33,890 INFO [train2.py:809] (1/4) Epoch 22, batch 600, loss[ctc_loss=0.05051, att_loss=0.2065, loss=0.1753, over 15885.00 frames. utt_duration=1631 frames, utt_pad_proportion=0.009222, over 39.00 utterances.], tot_loss[ctc_loss=0.07286, att_loss=0.2344, loss=0.2021, over 3105883.90 frames. utt_duration=1262 frames, utt_pad_proportion=0.05411, over 9859.78 utterances.], batch size: 39, lr: 4.93e-03, grad_scale: 8.0 2023-03-08 23:56:35,667 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84260.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:56:44,702 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.8969, 5.2470, 5.4022, 5.2938, 5.4387, 5.8803, 5.2346, 5.9691], device='cuda:1'), covar=tensor([0.0809, 0.0786, 0.1005, 0.1367, 0.1898, 0.0992, 0.0661, 0.0723], device='cuda:1'), in_proj_covar=tensor([0.0873, 0.0514, 0.0600, 0.0666, 0.0874, 0.0627, 0.0487, 0.0612], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-08 23:56:51,576 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5443, 3.0037, 4.9814, 4.0441, 3.0308, 4.2289, 4.7980, 4.6022], device='cuda:1'), covar=tensor([0.0252, 0.1378, 0.0190, 0.0816, 0.1590, 0.0257, 0.0139, 0.0265], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0243, 0.0190, 0.0313, 0.0264, 0.0216, 0.0178, 0.0207], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 23:57:23,486 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4602, 2.6105, 4.9572, 3.9320, 3.0054, 4.1871, 4.6920, 4.6224], device='cuda:1'), covar=tensor([0.0269, 0.1689, 0.0191, 0.0880, 0.1645, 0.0262, 0.0177, 0.0256], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0243, 0.0190, 0.0313, 0.0264, 0.0216, 0.0178, 0.0207], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 23:57:35,150 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84297.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:57:48,998 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84305.0, num_to_drop=1, layers_to_drop={2} 2023-03-08 23:57:55,131 INFO [train2.py:809] (1/4) Epoch 22, batch 650, loss[ctc_loss=0.06561, att_loss=0.1987, loss=0.1721, over 15506.00 frames. utt_duration=1724 frames, utt_pad_proportion=0.008466, over 36.00 utterances.], tot_loss[ctc_loss=0.07178, att_loss=0.2335, loss=0.2011, over 3140759.62 frames. utt_duration=1275 frames, utt_pad_proportion=0.04986, over 9862.34 utterances.], batch size: 36, lr: 4.93e-03, grad_scale: 8.0 2023-03-08 23:58:38,277 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84336.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:58:45,773 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84341.0, num_to_drop=0, layers_to_drop=set() 2023-03-08 23:58:47,104 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.197e+02 2.036e+02 2.451e+02 3.016e+02 6.584e+02, threshold=4.901e+02, percent-clipped=3.0 2023-03-08 23:59:13,079 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9246, 2.5959, 3.4629, 2.7182, 3.3418, 4.0304, 3.8895, 3.0188], device='cuda:1'), covar=tensor([0.0410, 0.1893, 0.1213, 0.1414, 0.1095, 0.1091, 0.0714, 0.1224], device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0243, 0.0281, 0.0219, 0.0264, 0.0367, 0.0261, 0.0232], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-08 23:59:16,011 INFO [train2.py:809] (1/4) Epoch 22, batch 700, loss[ctc_loss=0.0823, att_loss=0.2461, loss=0.2133, over 17445.00 frames. utt_duration=884.7 frames, utt_pad_proportion=0.0746, over 79.00 utterances.], tot_loss[ctc_loss=0.07126, att_loss=0.2335, loss=0.201, over 3168548.82 frames. utt_duration=1295 frames, utt_pad_proportion=0.04398, over 9801.84 utterances.], batch size: 79, lr: 4.93e-03, grad_scale: 8.0 2023-03-08 23:59:41,700 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.8500, 3.7106, 3.6343, 3.0771, 3.7337, 3.7262, 3.6808, 2.7083], device='cuda:1'), covar=tensor([0.1011, 0.1058, 0.1828, 0.3635, 0.0769, 0.1883, 0.0837, 0.3763], device='cuda:1'), in_proj_covar=tensor([0.0177, 0.0190, 0.0202, 0.0257, 0.0159, 0.0262, 0.0182, 0.0219], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-08 23:59:51,977 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.15 vs. limit=5.0 2023-03-09 00:00:17,499 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84397.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:00:28,873 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4787, 2.6243, 5.0499, 4.0899, 3.1622, 4.2684, 4.8350, 4.7133], device='cuda:1'), covar=tensor([0.0325, 0.1624, 0.0173, 0.0808, 0.1552, 0.0256, 0.0139, 0.0259], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0244, 0.0191, 0.0315, 0.0267, 0.0218, 0.0178, 0.0210], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 00:00:37,601 INFO [train2.py:809] (1/4) Epoch 22, batch 750, loss[ctc_loss=0.0657, att_loss=0.2153, loss=0.1854, over 14461.00 frames. utt_duration=1809 frames, utt_pad_proportion=0.03566, over 32.00 utterances.], tot_loss[ctc_loss=0.07156, att_loss=0.2342, loss=0.2017, over 3198482.40 frames. utt_duration=1288 frames, utt_pad_proportion=0.04361, over 9947.43 utterances.], batch size: 32, lr: 4.93e-03, grad_scale: 8.0 2023-03-09 00:00:48,947 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84416.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:00:50,426 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.6298, 5.8931, 5.3780, 5.6686, 5.5832, 5.0851, 5.2963, 5.1178], device='cuda:1'), covar=tensor([0.1402, 0.1127, 0.0895, 0.0890, 0.1040, 0.1689, 0.2659, 0.2573], device='cuda:1'), in_proj_covar=tensor([0.0529, 0.0616, 0.0462, 0.0460, 0.0430, 0.0467, 0.0612, 0.0529], 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') 2023-03-09 00:01:29,371 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.218e+02 1.936e+02 2.365e+02 2.996e+02 6.300e+02, threshold=4.730e+02, percent-clipped=3.0 2023-03-09 00:01:48,710 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84453.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:01:57,862 INFO [train2.py:809] (1/4) Epoch 22, batch 800, loss[ctc_loss=0.07235, att_loss=0.207, loss=0.1801, over 15344.00 frames. utt_duration=1755 frames, utt_pad_proportion=0.01302, over 35.00 utterances.], tot_loss[ctc_loss=0.07241, att_loss=0.2348, loss=0.2023, over 3213537.19 frames. utt_duration=1268 frames, utt_pad_proportion=0.04844, over 10145.67 utterances.], batch size: 35, lr: 4.93e-03, grad_scale: 8.0 2023-03-09 00:02:06,037 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=84464.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:02:12,456 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5374, 4.8522, 4.7710, 4.8704, 4.9238, 4.5493, 3.2969, 4.7218], device='cuda:1'), covar=tensor([0.0125, 0.0148, 0.0152, 0.0093, 0.0116, 0.0139, 0.0848, 0.0286], device='cuda:1'), in_proj_covar=tensor([0.0090, 0.0086, 0.0109, 0.0068, 0.0075, 0.0084, 0.0102, 0.0106], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 00:03:17,422 INFO [train2.py:809] (1/4) Epoch 22, batch 850, loss[ctc_loss=0.0622, att_loss=0.2164, loss=0.1856, over 15904.00 frames. utt_duration=1633 frames, utt_pad_proportion=0.008069, over 39.00 utterances.], tot_loss[ctc_loss=0.07301, att_loss=0.2354, loss=0.203, over 3228769.08 frames. utt_duration=1249 frames, utt_pad_proportion=0.05222, over 10350.08 utterances.], batch size: 39, lr: 4.93e-03, grad_scale: 8.0 2023-03-09 00:03:42,547 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84525.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:04:07,926 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.340e+02 1.954e+02 2.376e+02 2.798e+02 7.444e+02, threshold=4.752e+02, percent-clipped=4.0 2023-03-09 00:04:30,054 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84555.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:04:36,028 INFO [train2.py:809] (1/4) Epoch 22, batch 900, loss[ctc_loss=0.07185, att_loss=0.2443, loss=0.2098, over 17325.00 frames. utt_duration=1262 frames, utt_pad_proportion=0.01041, over 55.00 utterances.], tot_loss[ctc_loss=0.07215, att_loss=0.2347, loss=0.2022, over 3242989.63 frames. utt_duration=1272 frames, utt_pad_proportion=0.04601, over 10208.54 utterances.], batch size: 55, lr: 4.93e-03, grad_scale: 8.0 2023-03-09 00:04:52,971 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9152, 5.2162, 4.7852, 5.3081, 4.6296, 4.9757, 5.3570, 5.1299], device='cuda:1'), covar=tensor([0.0637, 0.0311, 0.0775, 0.0307, 0.0415, 0.0282, 0.0225, 0.0181], device='cuda:1'), in_proj_covar=tensor([0.0390, 0.0322, 0.0367, 0.0350, 0.0324, 0.0238, 0.0303, 0.0285], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-09 00:05:08,771 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6713, 3.1526, 3.8193, 2.9807, 3.6982, 4.7671, 4.5955, 3.4346], device='cuda:1'), covar=tensor([0.0333, 0.1608, 0.1171, 0.1454, 0.1003, 0.0796, 0.0615, 0.1189], device='cuda:1'), in_proj_covar=tensor([0.0245, 0.0244, 0.0279, 0.0219, 0.0264, 0.0366, 0.0258, 0.0231], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-09 00:05:18,220 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84586.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:05:22,777 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.3820, 2.7872, 3.4049, 4.4767, 3.9737, 3.9540, 2.9899, 2.2075], device='cuda:1'), covar=tensor([0.0764, 0.2039, 0.0901, 0.0569, 0.0929, 0.0474, 0.1423, 0.2379], device='cuda:1'), in_proj_covar=tensor([0.0182, 0.0216, 0.0189, 0.0219, 0.0225, 0.0179, 0.0202, 0.0187], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 00:05:27,574 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84592.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:05:50,244 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84605.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:05:55,902 INFO [train2.py:809] (1/4) Epoch 22, batch 950, loss[ctc_loss=0.06361, att_loss=0.2383, loss=0.2034, over 16473.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.006237, over 46.00 utterances.], tot_loss[ctc_loss=0.07246, att_loss=0.235, loss=0.2025, over 3256092.91 frames. utt_duration=1271 frames, utt_pad_proportion=0.04521, over 10258.64 utterances.], batch size: 46, lr: 4.92e-03, grad_scale: 8.0 2023-03-09 00:06:33,715 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84633.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:06:45,431 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.89 vs. limit=5.0 2023-03-09 00:06:46,115 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84641.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:06:47,279 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.223e+02 1.899e+02 2.395e+02 3.054e+02 5.479e+02, threshold=4.790e+02, percent-clipped=1.0 2023-03-09 00:07:06,386 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=84653.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:07:15,603 INFO [train2.py:809] (1/4) Epoch 22, batch 1000, loss[ctc_loss=0.06877, att_loss=0.245, loss=0.2098, over 17022.00 frames. utt_duration=1336 frames, utt_pad_proportion=0.007933, over 51.00 utterances.], tot_loss[ctc_loss=0.07247, att_loss=0.2347, loss=0.2022, over 3257007.08 frames. utt_duration=1265 frames, utt_pad_proportion=0.0484, over 10311.34 utterances.], batch size: 51, lr: 4.92e-03, grad_scale: 8.0 2023-03-09 00:08:02,349 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=84689.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:08:07,646 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84692.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:08:10,910 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84694.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:08:34,846 INFO [train2.py:809] (1/4) Epoch 22, batch 1050, loss[ctc_loss=0.07012, att_loss=0.235, loss=0.202, over 16391.00 frames. utt_duration=1492 frames, utt_pad_proportion=0.007455, over 44.00 utterances.], tot_loss[ctc_loss=0.07189, att_loss=0.2347, loss=0.2021, over 3263046.29 frames. utt_duration=1279 frames, utt_pad_proportion=0.04545, over 10220.17 utterances.], batch size: 44, lr: 4.92e-03, grad_scale: 8.0 2023-03-09 00:09:25,764 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.058e+02 1.985e+02 2.384e+02 2.768e+02 5.690e+02, threshold=4.768e+02, percent-clipped=2.0 2023-03-09 00:09:45,213 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84753.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:09:54,289 INFO [train2.py:809] (1/4) Epoch 22, batch 1100, loss[ctc_loss=0.0984, att_loss=0.2637, loss=0.2307, over 17366.00 frames. utt_duration=1104 frames, utt_pad_proportion=0.03561, over 63.00 utterances.], tot_loss[ctc_loss=0.07224, att_loss=0.2348, loss=0.2023, over 3267099.31 frames. utt_duration=1293 frames, utt_pad_proportion=0.04112, over 10119.38 utterances.], batch size: 63, lr: 4.92e-03, grad_scale: 8.0 2023-03-09 00:10:41,298 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.2218, 5.4949, 5.4375, 5.3874, 5.5355, 5.4706, 5.2165, 4.9882], device='cuda:1'), covar=tensor([0.0849, 0.0483, 0.0293, 0.0517, 0.0255, 0.0295, 0.0338, 0.0301], device='cuda:1'), in_proj_covar=tensor([0.0518, 0.0365, 0.0347, 0.0358, 0.0420, 0.0429, 0.0359, 0.0395], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004], device='cuda:1') 2023-03-09 00:10:59,982 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84800.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:11:01,287 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=84801.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:11:13,082 INFO [train2.py:809] (1/4) Epoch 22, batch 1150, loss[ctc_loss=0.07037, att_loss=0.2404, loss=0.2064, over 17531.00 frames. utt_duration=889.1 frames, utt_pad_proportion=0.06896, over 79.00 utterances.], tot_loss[ctc_loss=0.07265, att_loss=0.2355, loss=0.2029, over 3278821.96 frames. utt_duration=1290 frames, utt_pad_proportion=0.04025, over 10182.29 utterances.], batch size: 79, lr: 4.92e-03, grad_scale: 16.0 2023-03-09 00:11:20,869 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.1103, 5.3741, 5.6495, 5.4327, 5.5948, 6.0445, 5.3310, 6.1452], device='cuda:1'), covar=tensor([0.0609, 0.0851, 0.0798, 0.1309, 0.1631, 0.0912, 0.0730, 0.0598], device='cuda:1'), in_proj_covar=tensor([0.0872, 0.0515, 0.0597, 0.0664, 0.0872, 0.0626, 0.0486, 0.0607], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 00:12:04,637 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.375e+02 1.971e+02 2.350e+02 2.955e+02 6.874e+02, threshold=4.701e+02, percent-clipped=3.0 2023-03-09 00:12:25,860 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84855.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:12:31,666 INFO [train2.py:809] (1/4) Epoch 22, batch 1200, loss[ctc_loss=0.06853, att_loss=0.2399, loss=0.2056, over 17343.00 frames. utt_duration=1177 frames, utt_pad_proportion=0.02214, over 59.00 utterances.], tot_loss[ctc_loss=0.07309, att_loss=0.2359, loss=0.2033, over 3282155.29 frames. utt_duration=1263 frames, utt_pad_proportion=0.04561, over 10407.82 utterances.], batch size: 59, lr: 4.92e-03, grad_scale: 16.0 2023-03-09 00:12:35,063 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84861.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:13:06,046 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84881.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:13:24,369 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84892.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:13:41,849 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=84903.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:13:51,124 INFO [train2.py:809] (1/4) Epoch 22, batch 1250, loss[ctc_loss=0.07216, att_loss=0.2109, loss=0.1832, over 15779.00 frames. utt_duration=1662 frames, utt_pad_proportion=0.008338, over 38.00 utterances.], tot_loss[ctc_loss=0.07324, att_loss=0.2359, loss=0.2034, over 3268629.37 frames. utt_duration=1233 frames, utt_pad_proportion=0.05586, over 10621.00 utterances.], batch size: 38, lr: 4.92e-03, grad_scale: 16.0 2023-03-09 00:14:40,829 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=84940.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:14:43,746 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.170e+02 1.899e+02 2.318e+02 2.747e+02 6.253e+02, threshold=4.636e+02, percent-clipped=3.0 2023-03-09 00:14:58,345 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84951.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:15:00,453 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.19 vs. limit=5.0 2023-03-09 00:15:09,889 INFO [train2.py:809] (1/4) Epoch 22, batch 1300, loss[ctc_loss=0.07729, att_loss=0.2273, loss=0.1973, over 16391.00 frames. utt_duration=1491 frames, utt_pad_proportion=0.008372, over 44.00 utterances.], tot_loss[ctc_loss=0.07276, att_loss=0.2355, loss=0.203, over 3267333.10 frames. utt_duration=1246 frames, utt_pad_proportion=0.05335, over 10504.84 utterances.], batch size: 44, lr: 4.91e-03, grad_scale: 16.0 2023-03-09 00:15:57,616 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84989.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:15:59,931 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.9402, 3.9209, 3.2390, 3.3186, 4.1620, 3.7479, 2.8270, 4.2681], device='cuda:1'), covar=tensor([0.1186, 0.0416, 0.1081, 0.0761, 0.0667, 0.0617, 0.1038, 0.0553], device='cuda:1'), in_proj_covar=tensor([0.0204, 0.0217, 0.0227, 0.0200, 0.0278, 0.0241, 0.0202, 0.0287], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-09 00:16:02,944 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84992.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:16:29,444 INFO [train2.py:809] (1/4) Epoch 22, batch 1350, loss[ctc_loss=0.08749, att_loss=0.2333, loss=0.2041, over 15949.00 frames. utt_duration=1558 frames, utt_pad_proportion=0.006642, over 41.00 utterances.], tot_loss[ctc_loss=0.07283, att_loss=0.2351, loss=0.2027, over 3265277.93 frames. utt_duration=1261 frames, utt_pad_proportion=0.05087, over 10373.17 utterances.], batch size: 41, lr: 4.91e-03, grad_scale: 16.0 2023-03-09 00:16:34,355 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85012.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:17:18,558 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=85040.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:17:21,482 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.237e+02 1.871e+02 2.280e+02 2.864e+02 6.552e+02, threshold=4.561e+02, percent-clipped=2.0 2023-03-09 00:17:22,840 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-03-09 00:17:48,045 INFO [train2.py:809] (1/4) Epoch 22, batch 1400, loss[ctc_loss=0.0841, att_loss=0.2412, loss=0.2098, over 16621.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.005678, over 47.00 utterances.], tot_loss[ctc_loss=0.07382, att_loss=0.2362, loss=0.2037, over 3270999.76 frames. utt_duration=1242 frames, utt_pad_proportion=0.05632, over 10549.66 utterances.], batch size: 47, lr: 4.91e-03, grad_scale: 16.0 2023-03-09 00:18:58,980 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3580, 2.7978, 4.7548, 3.6343, 2.7995, 4.0841, 4.3392, 4.3892], device='cuda:1'), covar=tensor([0.0244, 0.1528, 0.0174, 0.1080, 0.1894, 0.0283, 0.0212, 0.0315], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0242, 0.0190, 0.0313, 0.0265, 0.0218, 0.0178, 0.0210], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 00:19:03,480 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85107.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:19:06,358 INFO [train2.py:809] (1/4) Epoch 22, batch 1450, loss[ctc_loss=0.06108, att_loss=0.2342, loss=0.1995, over 16309.00 frames. utt_duration=1451 frames, utt_pad_proportion=0.007296, over 45.00 utterances.], tot_loss[ctc_loss=0.07364, att_loss=0.2356, loss=0.2032, over 3265840.46 frames. utt_duration=1226 frames, utt_pad_proportion=0.06245, over 10672.00 utterances.], batch size: 45, lr: 4.91e-03, grad_scale: 16.0 2023-03-09 00:19:16,255 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85115.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:19:50,811 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2035, 3.7457, 3.1930, 3.5728, 4.0127, 3.6587, 3.1083, 4.2812], device='cuda:1'), covar=tensor([0.0881, 0.0496, 0.1023, 0.0645, 0.0707, 0.0767, 0.0789, 0.0550], device='cuda:1'), in_proj_covar=tensor([0.0205, 0.0219, 0.0228, 0.0200, 0.0280, 0.0242, 0.0203, 0.0289], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-09 00:19:59,353 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.290e+02 1.922e+02 2.285e+02 2.782e+02 4.553e+02, threshold=4.570e+02, percent-clipped=0.0 2023-03-09 00:20:08,338 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-03-09 00:20:21,437 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85156.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:20:25,865 INFO [train2.py:809] (1/4) Epoch 22, batch 1500, loss[ctc_loss=0.0901, att_loss=0.2557, loss=0.2225, over 17287.00 frames. utt_duration=876.7 frames, utt_pad_proportion=0.08103, over 79.00 utterances.], tot_loss[ctc_loss=0.07253, att_loss=0.2355, loss=0.2029, over 3280626.02 frames. utt_duration=1245 frames, utt_pad_proportion=0.05406, over 10556.86 utterances.], batch size: 79, lr: 4.91e-03, grad_scale: 16.0 2023-03-09 00:20:40,514 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85168.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 00:20:52,798 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85176.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:21:01,415 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85181.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:21:25,560 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-03-09 00:21:46,579 INFO [train2.py:809] (1/4) Epoch 22, batch 1550, loss[ctc_loss=0.1045, att_loss=0.2632, loss=0.2315, over 14598.00 frames. utt_duration=404.2 frames, utt_pad_proportion=0.2971, over 145.00 utterances.], tot_loss[ctc_loss=0.07209, att_loss=0.2351, loss=0.2025, over 3272366.05 frames. utt_duration=1236 frames, utt_pad_proportion=0.05822, over 10600.09 utterances.], batch size: 145, lr: 4.91e-03, grad_scale: 16.0 2023-03-09 00:22:19,154 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=85229.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:22:42,141 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.384e+02 1.952e+02 2.297e+02 2.764e+02 6.264e+02, threshold=4.595e+02, percent-clipped=3.0 2023-03-09 00:23:06,931 INFO [train2.py:809] (1/4) Epoch 22, batch 1600, loss[ctc_loss=0.08437, att_loss=0.2256, loss=0.1974, over 15896.00 frames. utt_duration=1632 frames, utt_pad_proportion=0.008381, over 39.00 utterances.], tot_loss[ctc_loss=0.07239, att_loss=0.2347, loss=0.2023, over 3264857.85 frames. utt_duration=1230 frames, utt_pad_proportion=0.06181, over 10633.79 utterances.], batch size: 39, lr: 4.91e-03, grad_scale: 8.0 2023-03-09 00:23:29,031 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2023-03-09 00:23:56,447 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85289.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:24:19,434 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7525, 1.8547, 2.2195, 2.6171, 2.6680, 2.6603, 2.3882, 3.1313], device='cuda:1'), covar=tensor([0.1834, 0.3641, 0.2518, 0.1474, 0.1882, 0.1288, 0.2177, 0.1004], device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0127, 0.0125, 0.0112, 0.0127, 0.0108, 0.0132, 0.0102], device='cuda:1'), out_proj_covar=tensor([9.1115e-05, 9.8646e-05, 9.9027e-05, 8.7707e-05, 9.4555e-05, 8.7129e-05, 9.9867e-05, 8.1469e-05], device='cuda:1') 2023-03-09 00:24:23,650 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85307.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:24:27,038 INFO [train2.py:809] (1/4) Epoch 22, batch 1650, loss[ctc_loss=0.0894, att_loss=0.2419, loss=0.2114, over 16831.00 frames. utt_duration=688.6 frames, utt_pad_proportion=0.1371, over 98.00 utterances.], tot_loss[ctc_loss=0.07278, att_loss=0.2357, loss=0.2031, over 3269580.64 frames. utt_duration=1209 frames, utt_pad_proportion=0.06386, over 10832.22 utterances.], batch size: 98, lr: 4.90e-03, grad_scale: 8.0 2023-03-09 00:24:39,195 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0388, 4.9852, 4.7594, 2.7874, 4.7205, 4.6531, 4.3306, 2.8077], device='cuda:1'), covar=tensor([0.0091, 0.0092, 0.0302, 0.1028, 0.0114, 0.0200, 0.0287, 0.1234], device='cuda:1'), in_proj_covar=tensor([0.0074, 0.0101, 0.0104, 0.0110, 0.0085, 0.0113, 0.0099, 0.0103], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-09 00:25:11,311 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=85337.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:25:20,306 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.436e+02 1.903e+02 2.201e+02 2.688e+02 4.353e+02, threshold=4.403e+02, percent-clipped=0.0 2023-03-09 00:25:44,528 INFO [train2.py:809] (1/4) Epoch 22, batch 1700, loss[ctc_loss=0.07942, att_loss=0.2446, loss=0.2116, over 17335.00 frames. utt_duration=1177 frames, utt_pad_proportion=0.02276, over 59.00 utterances.], tot_loss[ctc_loss=0.07286, att_loss=0.2353, loss=0.2028, over 3271291.56 frames. utt_duration=1214 frames, utt_pad_proportion=0.06237, over 10794.20 utterances.], batch size: 59, lr: 4.90e-03, grad_scale: 8.0 2023-03-09 00:26:15,648 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0553, 5.0968, 4.9474, 2.0757, 2.0248, 3.0437, 2.3653, 3.7954], device='cuda:1'), covar=tensor([0.0781, 0.0285, 0.0232, 0.5873, 0.6031, 0.2442, 0.3962, 0.1857], device='cuda:1'), in_proj_covar=tensor([0.0356, 0.0280, 0.0268, 0.0247, 0.0342, 0.0333, 0.0257, 0.0365], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-03-09 00:27:03,857 INFO [train2.py:809] (1/4) Epoch 22, batch 1750, loss[ctc_loss=0.05845, att_loss=0.2121, loss=0.1813, over 15362.00 frames. utt_duration=1757 frames, utt_pad_proportion=0.01116, over 35.00 utterances.], tot_loss[ctc_loss=0.07292, att_loss=0.2354, loss=0.2029, over 3263162.03 frames. utt_duration=1212 frames, utt_pad_proportion=0.06646, over 10778.73 utterances.], batch size: 35, lr: 4.90e-03, grad_scale: 8.0 2023-03-09 00:27:35,525 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85428.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:27:45,599 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.6592, 3.7434, 3.9510, 2.3827, 2.3521, 2.9092, 2.3527, 3.5141], device='cuda:1'), covar=tensor([0.0715, 0.0496, 0.0360, 0.3970, 0.3881, 0.2043, 0.3059, 0.1310], device='cuda:1'), in_proj_covar=tensor([0.0353, 0.0277, 0.0265, 0.0244, 0.0338, 0.0330, 0.0254, 0.0362], 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') 2023-03-09 00:27:58,932 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.196e+02 1.824e+02 2.227e+02 2.751e+02 5.313e+02, threshold=4.454e+02, percent-clipped=1.0 2023-03-09 00:28:06,858 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9796, 5.0145, 4.7781, 2.6997, 4.7081, 4.6283, 4.1070, 2.7658], device='cuda:1'), covar=tensor([0.0116, 0.0101, 0.0304, 0.1181, 0.0134, 0.0223, 0.0386, 0.1385], device='cuda:1'), in_proj_covar=tensor([0.0074, 0.0101, 0.0104, 0.0111, 0.0085, 0.0114, 0.0099, 0.0103], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-09 00:28:19,234 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85456.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:28:24,087 INFO [train2.py:809] (1/4) Epoch 22, batch 1800, loss[ctc_loss=0.07668, att_loss=0.2526, loss=0.2174, over 17339.00 frames. utt_duration=879.2 frames, utt_pad_proportion=0.07645, over 79.00 utterances.], tot_loss[ctc_loss=0.07269, att_loss=0.2354, loss=0.2028, over 3260093.99 frames. utt_duration=1210 frames, utt_pad_proportion=0.06782, over 10792.64 utterances.], batch size: 79, lr: 4.90e-03, grad_scale: 8.0 2023-03-09 00:28:30,482 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85463.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 00:28:44,313 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85471.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:29:04,503 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85483.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:29:13,783 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85489.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:29:16,930 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7143, 2.0925, 2.4382, 2.7290, 2.8388, 2.7033, 2.3013, 3.2901], device='cuda:1'), covar=tensor([0.1928, 0.3078, 0.2090, 0.1365, 0.1507, 0.1606, 0.2534, 0.0850], device='cuda:1'), in_proj_covar=tensor([0.0121, 0.0128, 0.0125, 0.0113, 0.0127, 0.0109, 0.0133, 0.0102], device='cuda:1'), out_proj_covar=tensor([9.1715e-05, 9.9021e-05, 9.9244e-05, 8.8093e-05, 9.4906e-05, 8.7841e-05, 1.0056e-04, 8.1926e-05], device='cuda:1') 2023-03-09 00:29:37,443 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=85504.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:29:45,466 INFO [train2.py:809] (1/4) Epoch 22, batch 1850, loss[ctc_loss=0.05396, att_loss=0.2264, loss=0.1919, over 16306.00 frames. utt_duration=1518 frames, utt_pad_proportion=0.005757, over 43.00 utterances.], tot_loss[ctc_loss=0.07307, att_loss=0.2357, loss=0.2032, over 3268643.44 frames. utt_duration=1227 frames, utt_pad_proportion=0.05975, over 10667.82 utterances.], batch size: 43, lr: 4.90e-03, grad_scale: 8.0 2023-03-09 00:30:19,822 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85530.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:30:39,483 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.220e+02 1.961e+02 2.288e+02 2.776e+02 5.536e+02, threshold=4.576e+02, percent-clipped=1.0 2023-03-09 00:30:41,535 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85544.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:31:04,497 INFO [train2.py:809] (1/4) Epoch 22, batch 1900, loss[ctc_loss=0.1003, att_loss=0.2563, loss=0.2251, over 17304.00 frames. utt_duration=1260 frames, utt_pad_proportion=0.01182, over 55.00 utterances.], tot_loss[ctc_loss=0.07364, att_loss=0.2358, loss=0.2034, over 3271208.78 frames. utt_duration=1217 frames, utt_pad_proportion=0.06239, over 10763.37 utterances.], batch size: 55, lr: 4.90e-03, grad_scale: 8.0 2023-03-09 00:31:25,966 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0111, 4.9409, 4.7521, 2.0853, 2.0497, 2.8811, 2.3430, 3.9416], device='cuda:1'), covar=tensor([0.0788, 0.0262, 0.0272, 0.5172, 0.5416, 0.2473, 0.3661, 0.1475], device='cuda:1'), in_proj_covar=tensor([0.0355, 0.0278, 0.0267, 0.0246, 0.0340, 0.0332, 0.0255, 0.0364], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-03-09 00:31:55,956 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85591.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:32:21,686 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85607.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:32:24,415 INFO [train2.py:809] (1/4) Epoch 22, batch 1950, loss[ctc_loss=0.07613, att_loss=0.2401, loss=0.2073, over 16331.00 frames. utt_duration=1453 frames, utt_pad_proportion=0.005958, over 45.00 utterances.], tot_loss[ctc_loss=0.073, att_loss=0.2354, loss=0.2029, over 3276954.67 frames. utt_duration=1244 frames, utt_pad_proportion=0.05411, over 10545.53 utterances.], batch size: 45, lr: 4.90e-03, grad_scale: 8.0 2023-03-09 00:32:51,862 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85625.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:33:19,619 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.244e+02 1.791e+02 2.130e+02 2.666e+02 5.391e+02, threshold=4.261e+02, percent-clipped=1.0 2023-03-09 00:33:39,320 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=85655.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:33:44,683 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.08 vs. limit=5.0 2023-03-09 00:33:45,248 INFO [train2.py:809] (1/4) Epoch 22, batch 2000, loss[ctc_loss=0.06704, att_loss=0.2176, loss=0.1875, over 14539.00 frames. utt_duration=1819 frames, utt_pad_proportion=0.03894, over 32.00 utterances.], tot_loss[ctc_loss=0.07177, att_loss=0.2349, loss=0.2022, over 3279978.34 frames. utt_duration=1265 frames, utt_pad_proportion=0.04823, over 10383.62 utterances.], batch size: 32, lr: 4.89e-03, grad_scale: 8.0 2023-03-09 00:34:29,126 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85686.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 00:34:49,361 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1085, 4.2835, 4.4218, 4.6891, 2.8860, 4.3114, 2.6778, 2.0597], device='cuda:1'), covar=tensor([0.0461, 0.0271, 0.0629, 0.0223, 0.1499, 0.0244, 0.1430, 0.1507], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0166, 0.0257, 0.0158, 0.0219, 0.0147, 0.0229, 0.0202], 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') 2023-03-09 00:34:50,636 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0147, 5.4435, 4.5039, 5.5932, 4.9244, 5.2327, 5.4564, 5.3104], device='cuda:1'), covar=tensor([0.0698, 0.0343, 0.1228, 0.0396, 0.0356, 0.0206, 0.0372, 0.0230], device='cuda:1'), in_proj_covar=tensor([0.0393, 0.0324, 0.0369, 0.0356, 0.0325, 0.0240, 0.0306, 0.0288], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-09 00:35:05,236 INFO [train2.py:809] (1/4) Epoch 22, batch 2050, loss[ctc_loss=0.07119, att_loss=0.244, loss=0.2094, over 17049.00 frames. utt_duration=1313 frames, utt_pad_proportion=0.008294, over 52.00 utterances.], tot_loss[ctc_loss=0.0718, att_loss=0.2347, loss=0.2021, over 3267744.05 frames. utt_duration=1278 frames, utt_pad_proportion=0.04719, over 10240.28 utterances.], batch size: 52, lr: 4.89e-03, grad_scale: 8.0 2023-03-09 00:35:13,317 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4915, 2.6382, 4.9533, 3.9107, 2.9133, 4.2445, 4.7136, 4.6234], device='cuda:1'), covar=tensor([0.0243, 0.1560, 0.0169, 0.0856, 0.1768, 0.0247, 0.0143, 0.0255], device='cuda:1'), in_proj_covar=tensor([0.0199, 0.0242, 0.0193, 0.0315, 0.0266, 0.0219, 0.0180, 0.0211], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 00:36:01,006 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.043e+02 1.910e+02 2.299e+02 2.633e+02 5.710e+02, threshold=4.599e+02, percent-clipped=8.0 2023-03-09 00:36:14,140 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7859, 5.1149, 4.9051, 5.0185, 5.1518, 4.8311, 3.7953, 5.1563], device='cuda:1'), covar=tensor([0.0104, 0.0110, 0.0124, 0.0090, 0.0083, 0.0103, 0.0610, 0.0143], device='cuda:1'), in_proj_covar=tensor([0.0092, 0.0089, 0.0112, 0.0070, 0.0077, 0.0087, 0.0105, 0.0109], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 00:36:15,918 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3286, 4.5024, 4.5743, 4.8697, 2.9555, 4.4757, 2.8305, 2.1497], device='cuda:1'), covar=tensor([0.0392, 0.0267, 0.0656, 0.0186, 0.1511, 0.0234, 0.1416, 0.1516], device='cuda:1'), in_proj_covar=tensor([0.0195, 0.0167, 0.0258, 0.0159, 0.0219, 0.0148, 0.0230, 0.0204], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 00:36:27,087 INFO [train2.py:809] (1/4) Epoch 22, batch 2100, loss[ctc_loss=0.05396, att_loss=0.2048, loss=0.1746, over 15342.00 frames. utt_duration=1755 frames, utt_pad_proportion=0.01058, over 35.00 utterances.], tot_loss[ctc_loss=0.0722, att_loss=0.2343, loss=0.2019, over 3262130.80 frames. utt_duration=1281 frames, utt_pad_proportion=0.04889, over 10197.23 utterances.], batch size: 35, lr: 4.89e-03, grad_scale: 8.0 2023-03-09 00:36:33,867 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85763.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:36:47,314 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85771.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:37:08,383 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85784.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:37:08,436 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9188, 5.2232, 4.8070, 5.3157, 4.6607, 4.9968, 5.3610, 5.1116], device='cuda:1'), covar=tensor([0.0611, 0.0311, 0.0835, 0.0330, 0.0450, 0.0261, 0.0264, 0.0214], device='cuda:1'), in_proj_covar=tensor([0.0389, 0.0321, 0.0366, 0.0352, 0.0323, 0.0238, 0.0303, 0.0285], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-09 00:37:11,949 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.4603, 2.1670, 2.0832, 2.7260, 2.8844, 2.5607, 2.3064, 2.9894], device='cuda:1'), covar=tensor([0.1785, 0.2701, 0.2160, 0.1300, 0.1475, 0.1261, 0.2347, 0.1105], device='cuda:1'), in_proj_covar=tensor([0.0119, 0.0125, 0.0123, 0.0111, 0.0125, 0.0107, 0.0131, 0.0101], device='cuda:1'), out_proj_covar=tensor([9.0327e-05, 9.7437e-05, 9.7627e-05, 8.6711e-05, 9.3292e-05, 8.6023e-05, 9.9128e-05, 8.1287e-05], device='cuda:1') 2023-03-09 00:37:47,848 INFO [train2.py:809] (1/4) Epoch 22, batch 2150, loss[ctc_loss=0.05709, att_loss=0.243, loss=0.2058, over 17024.00 frames. utt_duration=1337 frames, utt_pad_proportion=0.007307, over 51.00 utterances.], tot_loss[ctc_loss=0.07243, att_loss=0.2347, loss=0.2023, over 3273438.10 frames. utt_duration=1287 frames, utt_pad_proportion=0.04422, over 10186.52 utterances.], batch size: 51, lr: 4.89e-03, grad_scale: 8.0 2023-03-09 00:37:51,081 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=85811.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:38:04,881 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=85819.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:38:36,678 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85839.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:38:42,697 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.253e+02 1.885e+02 2.205e+02 3.046e+02 7.448e+02, threshold=4.410e+02, percent-clipped=6.0 2023-03-09 00:39:07,842 INFO [train2.py:809] (1/4) Epoch 22, batch 2200, loss[ctc_loss=0.06147, att_loss=0.2358, loss=0.2009, over 17004.00 frames. utt_duration=1335 frames, utt_pad_proportion=0.00888, over 51.00 utterances.], tot_loss[ctc_loss=0.07242, att_loss=0.2351, loss=0.2025, over 3270681.34 frames. utt_duration=1238 frames, utt_pad_proportion=0.05866, over 10581.09 utterances.], batch size: 51, lr: 4.89e-03, grad_scale: 8.0 2023-03-09 00:39:16,941 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-03-09 00:39:35,113 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85875.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:39:52,061 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85886.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:40:18,373 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0449, 4.9666, 4.8053, 3.0998, 4.7585, 4.7091, 4.3169, 2.6853], device='cuda:1'), covar=tensor([0.0106, 0.0085, 0.0253, 0.0891, 0.0095, 0.0180, 0.0296, 0.1338], device='cuda:1'), in_proj_covar=tensor([0.0076, 0.0102, 0.0106, 0.0112, 0.0086, 0.0116, 0.0100, 0.0104], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-09 00:40:28,363 INFO [train2.py:809] (1/4) Epoch 22, batch 2250, loss[ctc_loss=0.07504, att_loss=0.2431, loss=0.2095, over 16957.00 frames. utt_duration=1358 frames, utt_pad_proportion=0.008035, over 50.00 utterances.], tot_loss[ctc_loss=0.07213, att_loss=0.2353, loss=0.2027, over 3276965.05 frames. utt_duration=1249 frames, utt_pad_proportion=0.05292, over 10508.05 utterances.], batch size: 50, lr: 4.89e-03, grad_scale: 8.0 2023-03-09 00:41:02,661 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85930.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 00:41:11,767 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85936.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:41:22,180 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.269e+02 1.895e+02 2.304e+02 2.747e+02 7.782e+02, threshold=4.607e+02, percent-clipped=2.0 2023-03-09 00:41:44,938 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85957.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:41:45,679 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-03-09 00:41:47,670 INFO [train2.py:809] (1/4) Epoch 22, batch 2300, loss[ctc_loss=0.04848, att_loss=0.2078, loss=0.1759, over 15364.00 frames. utt_duration=1757 frames, utt_pad_proportion=0.01113, over 35.00 utterances.], tot_loss[ctc_loss=0.0732, att_loss=0.2362, loss=0.2036, over 3275886.80 frames. utt_duration=1198 frames, utt_pad_proportion=0.06664, over 10953.28 utterances.], batch size: 35, lr: 4.89e-03, grad_scale: 8.0 2023-03-09 00:42:24,319 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85981.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 00:42:40,413 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85991.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 00:43:02,479 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86002.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:43:07,380 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1065, 4.2734, 4.0177, 4.5933, 2.7277, 4.3479, 2.5470, 1.7173], device='cuda:1'), covar=tensor([0.0424, 0.0226, 0.0860, 0.0203, 0.1797, 0.0219, 0.1684, 0.1881], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0166, 0.0257, 0.0157, 0.0219, 0.0147, 0.0229, 0.0202], 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') 2023-03-09 00:43:13,254 INFO [train2.py:809] (1/4) Epoch 22, batch 2350, loss[ctc_loss=0.07507, att_loss=0.2435, loss=0.2098, over 17273.00 frames. utt_duration=876.2 frames, utt_pad_proportion=0.08158, over 79.00 utterances.], tot_loss[ctc_loss=0.07253, att_loss=0.2356, loss=0.203, over 3273549.47 frames. utt_duration=1210 frames, utt_pad_proportion=0.06372, over 10835.06 utterances.], batch size: 79, lr: 4.88e-03, grad_scale: 8.0 2023-03-09 00:43:28,513 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86018.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:43:46,303 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.6066, 3.2197, 3.2244, 2.7894, 3.1970, 3.2364, 3.2532, 2.1815], device='cuda:1'), covar=tensor([0.1138, 0.1457, 0.1705, 0.4192, 0.1554, 0.2524, 0.1258, 0.4414], device='cuda:1'), in_proj_covar=tensor([0.0178, 0.0191, 0.0202, 0.0255, 0.0160, 0.0262, 0.0185, 0.0217], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 00:43:59,584 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.2782, 5.5607, 5.2190, 5.6070, 5.1002, 5.1727, 5.7198, 5.4841], device='cuda:1'), covar=tensor([0.0560, 0.0228, 0.0648, 0.0288, 0.0362, 0.0231, 0.0210, 0.0180], device='cuda:1'), in_proj_covar=tensor([0.0388, 0.0321, 0.0367, 0.0351, 0.0323, 0.0237, 0.0303, 0.0286], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-09 00:44:07,000 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.206e+02 1.776e+02 2.270e+02 2.670e+02 5.895e+02, threshold=4.541e+02, percent-clipped=3.0 2023-03-09 00:44:21,751 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86052.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:44:32,413 INFO [train2.py:809] (1/4) Epoch 22, batch 2400, loss[ctc_loss=0.08694, att_loss=0.2571, loss=0.2231, over 17050.00 frames. utt_duration=1288 frames, utt_pad_proportion=0.009775, over 53.00 utterances.], tot_loss[ctc_loss=0.07224, att_loss=0.2348, loss=0.2023, over 3270837.10 frames. utt_duration=1232 frames, utt_pad_proportion=0.05888, over 10634.46 utterances.], batch size: 53, lr: 4.88e-03, grad_scale: 8.0 2023-03-09 00:44:39,833 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86063.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:44:49,086 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86069.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:45:12,640 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86084.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:45:50,869 INFO [train2.py:809] (1/4) Epoch 22, batch 2450, loss[ctc_loss=0.06417, att_loss=0.223, loss=0.1912, over 12396.00 frames. utt_duration=1838 frames, utt_pad_proportion=0.1308, over 27.00 utterances.], tot_loss[ctc_loss=0.07319, att_loss=0.2354, loss=0.2029, over 3258546.99 frames. utt_duration=1204 frames, utt_pad_proportion=0.06923, over 10843.19 utterances.], batch size: 27, lr: 4.88e-03, grad_scale: 8.0 2023-03-09 00:45:57,240 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7630, 3.6647, 3.5948, 3.0498, 3.6488, 3.6632, 3.6614, 2.5306], device='cuda:1'), covar=tensor([0.1070, 0.0978, 0.1798, 0.3194, 0.1166, 0.3467, 0.0762, 0.3948], device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0192, 0.0203, 0.0256, 0.0161, 0.0263, 0.0186, 0.0217], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 00:45:58,780 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86113.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:46:26,055 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86130.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:46:27,622 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7169, 3.5501, 3.7911, 3.2922, 3.7441, 4.7085, 4.5395, 3.2703], device='cuda:1'), covar=tensor([0.0347, 0.1214, 0.1099, 0.1254, 0.0872, 0.0977, 0.0545, 0.1275], device='cuda:1'), in_proj_covar=tensor([0.0247, 0.0243, 0.0284, 0.0220, 0.0266, 0.0370, 0.0261, 0.0233], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-09 00:46:28,918 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86132.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:46:39,882 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86139.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:46:45,845 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.355e+02 2.126e+02 2.601e+02 3.056e+02 8.215e+02, threshold=5.203e+02, percent-clipped=8.0 2023-03-09 00:47:11,858 INFO [train2.py:809] (1/4) Epoch 22, batch 2500, loss[ctc_loss=0.05515, att_loss=0.2251, loss=0.1911, over 15968.00 frames. utt_duration=1559 frames, utt_pad_proportion=0.00642, over 41.00 utterances.], tot_loss[ctc_loss=0.07234, att_loss=0.2349, loss=0.2024, over 3254878.58 frames. utt_duration=1241 frames, utt_pad_proportion=0.05996, over 10506.78 utterances.], batch size: 41, lr: 4.88e-03, grad_scale: 8.0 2023-03-09 00:47:38,517 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1570, 4.4370, 4.5595, 4.7970, 2.7669, 4.3081, 2.7417, 1.5724], device='cuda:1'), covar=tensor([0.0396, 0.0254, 0.0547, 0.0165, 0.1586, 0.0224, 0.1425, 0.1795], device='cuda:1'), in_proj_covar=tensor([0.0193, 0.0165, 0.0256, 0.0157, 0.0218, 0.0147, 0.0228, 0.0201], 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') 2023-03-09 00:47:57,543 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86186.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:47:58,892 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86187.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:48:33,442 INFO [train2.py:809] (1/4) Epoch 22, batch 2550, loss[ctc_loss=0.05872, att_loss=0.2336, loss=0.1987, over 16479.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.006654, over 46.00 utterances.], tot_loss[ctc_loss=0.07284, att_loss=0.2354, loss=0.2029, over 3266762.83 frames. utt_duration=1229 frames, utt_pad_proportion=0.05982, over 10647.24 utterances.], batch size: 46, lr: 4.88e-03, grad_scale: 8.0 2023-03-09 00:49:10,439 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86231.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:49:15,111 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86234.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:49:25,279 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.81 vs. limit=2.0 2023-03-09 00:49:29,164 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.366e+02 1.974e+02 2.356e+02 2.687e+02 6.031e+02, threshold=4.712e+02, percent-clipped=2.0 2023-03-09 00:49:34,352 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86246.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:49:54,972 INFO [train2.py:809] (1/4) Epoch 22, batch 2600, loss[ctc_loss=0.0885, att_loss=0.2211, loss=0.1946, over 15763.00 frames. utt_duration=1661 frames, utt_pad_proportion=0.007892, over 38.00 utterances.], tot_loss[ctc_loss=0.07268, att_loss=0.2355, loss=0.2029, over 3275485.70 frames. utt_duration=1240 frames, utt_pad_proportion=0.05547, over 10582.41 utterances.], batch size: 38, lr: 4.88e-03, grad_scale: 8.0 2023-03-09 00:50:31,786 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86281.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:50:39,195 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86286.0, num_to_drop=1, layers_to_drop={3} 2023-03-09 00:51:13,059 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86307.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:51:15,795 INFO [train2.py:809] (1/4) Epoch 22, batch 2650, loss[ctc_loss=0.06925, att_loss=0.2423, loss=0.2076, over 17127.00 frames. utt_duration=1225 frames, utt_pad_proportion=0.01468, over 56.00 utterances.], tot_loss[ctc_loss=0.07219, att_loss=0.236, loss=0.2033, over 3284040.41 frames. utt_duration=1237 frames, utt_pad_proportion=0.05379, over 10630.47 utterances.], batch size: 56, lr: 4.88e-03, grad_scale: 8.0 2023-03-09 00:51:23,145 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86313.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:51:49,594 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86329.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:52:11,920 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.265e+02 1.827e+02 2.241e+02 2.634e+02 5.171e+02, threshold=4.483e+02, percent-clipped=1.0 2023-03-09 00:52:36,640 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86358.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:52:38,759 INFO [train2.py:809] (1/4) Epoch 22, batch 2700, loss[ctc_loss=0.06573, att_loss=0.2392, loss=0.2045, over 17304.00 frames. utt_duration=1175 frames, utt_pad_proportion=0.0242, over 59.00 utterances.], tot_loss[ctc_loss=0.07121, att_loss=0.2355, loss=0.2026, over 3289471.99 frames. utt_duration=1242 frames, utt_pad_proportion=0.05008, over 10602.91 utterances.], batch size: 59, lr: 4.87e-03, grad_scale: 8.0 2023-03-09 00:52:54,788 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5916, 2.5587, 4.9122, 3.7657, 3.2338, 4.3682, 4.7173, 4.7904], device='cuda:1'), covar=tensor([0.0154, 0.1447, 0.0129, 0.1044, 0.1452, 0.0180, 0.0122, 0.0151], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0239, 0.0191, 0.0313, 0.0262, 0.0215, 0.0178, 0.0209], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 00:53:58,802 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86408.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:54:00,133 INFO [train2.py:809] (1/4) Epoch 22, batch 2750, loss[ctc_loss=0.07795, att_loss=0.2318, loss=0.201, over 16544.00 frames. utt_duration=1472 frames, utt_pad_proportion=0.005957, over 45.00 utterances.], tot_loss[ctc_loss=0.07145, att_loss=0.2356, loss=0.2027, over 3284444.00 frames. utt_duration=1249 frames, utt_pad_proportion=0.05119, over 10529.85 utterances.], batch size: 45, lr: 4.87e-03, grad_scale: 8.0 2023-03-09 00:54:19,869 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6004, 4.6301, 4.8142, 4.7168, 5.1128, 4.7054, 4.6465, 2.4181], device='cuda:1'), covar=tensor([0.0179, 0.0184, 0.0168, 0.0162, 0.0567, 0.0159, 0.0209, 0.1855], device='cuda:1'), in_proj_covar=tensor([0.0165, 0.0190, 0.0189, 0.0206, 0.0364, 0.0160, 0.0180, 0.0215], 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') 2023-03-09 00:54:25,786 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86425.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:54:44,579 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.3336, 2.8957, 3.2234, 4.4866, 4.0324, 3.8563, 2.9805, 2.3671], device='cuda:1'), covar=tensor([0.0728, 0.1832, 0.1017, 0.0488, 0.0734, 0.0505, 0.1405, 0.2135], device='cuda:1'), in_proj_covar=tensor([0.0180, 0.0213, 0.0190, 0.0219, 0.0224, 0.0181, 0.0201, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 00:54:49,470 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86440.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:54:53,696 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.448e+02 1.998e+02 2.389e+02 2.932e+02 8.437e+02, threshold=4.778e+02, percent-clipped=2.0 2023-03-09 00:54:57,216 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86445.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:55:20,793 INFO [train2.py:809] (1/4) Epoch 22, batch 2800, loss[ctc_loss=0.07966, att_loss=0.2489, loss=0.2151, over 16702.00 frames. utt_duration=1454 frames, utt_pad_proportion=0.005561, over 46.00 utterances.], tot_loss[ctc_loss=0.07209, att_loss=0.2355, loss=0.2028, over 3278764.67 frames. utt_duration=1226 frames, utt_pad_proportion=0.05835, over 10710.57 utterances.], batch size: 46, lr: 4.87e-03, grad_scale: 8.0 2023-03-09 00:55:52,960 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9524, 5.2683, 5.5064, 5.3349, 5.4479, 5.9021, 5.2009, 6.0396], device='cuda:1'), covar=tensor([0.0689, 0.0685, 0.0813, 0.1426, 0.1841, 0.0967, 0.0775, 0.0624], device='cuda:1'), in_proj_covar=tensor([0.0863, 0.0509, 0.0595, 0.0660, 0.0864, 0.0628, 0.0485, 0.0605], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 00:55:54,603 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.3215, 5.3000, 5.1304, 3.0497, 5.0664, 4.9448, 4.7952, 3.2502], device='cuda:1'), covar=tensor([0.0100, 0.0079, 0.0219, 0.0951, 0.0089, 0.0172, 0.0237, 0.1068], device='cuda:1'), in_proj_covar=tensor([0.0075, 0.0102, 0.0105, 0.0111, 0.0085, 0.0115, 0.0099, 0.0103], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-09 00:56:28,117 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86501.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:56:29,623 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.5898, 2.2605, 2.2164, 2.5136, 2.8234, 2.7552, 2.5482, 2.8735], device='cuda:1'), covar=tensor([0.1738, 0.2952, 0.2250, 0.1614, 0.1918, 0.1166, 0.2401, 0.1544], device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0126, 0.0123, 0.0111, 0.0126, 0.0107, 0.0130, 0.0102], device='cuda:1'), out_proj_covar=tensor([9.0692e-05, 9.7893e-05, 9.8182e-05, 8.7179e-05, 9.3983e-05, 8.6511e-05, 9.8888e-05, 8.1750e-05], device='cuda:1') 2023-03-09 00:56:36,589 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86506.0, num_to_drop=1, layers_to_drop={3} 2023-03-09 00:56:40,833 INFO [train2.py:809] (1/4) Epoch 22, batch 2850, loss[ctc_loss=0.05303, att_loss=0.2053, loss=0.1748, over 15868.00 frames. utt_duration=1629 frames, utt_pad_proportion=0.01039, over 39.00 utterances.], tot_loss[ctc_loss=0.07171, att_loss=0.235, loss=0.2023, over 3272211.54 frames. utt_duration=1223 frames, utt_pad_proportion=0.06143, over 10712.85 utterances.], batch size: 39, lr: 4.87e-03, grad_scale: 8.0 2023-03-09 00:57:16,135 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86531.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:57:34,136 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 9.710e+01 1.849e+02 2.202e+02 2.707e+02 3.690e+02, threshold=4.403e+02, percent-clipped=0.0 2023-03-09 00:58:01,275 INFO [train2.py:809] (1/4) Epoch 22, batch 2900, loss[ctc_loss=0.0588, att_loss=0.229, loss=0.1949, over 16474.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.006237, over 46.00 utterances.], tot_loss[ctc_loss=0.07126, att_loss=0.2347, loss=0.202, over 3271221.39 frames. utt_duration=1256 frames, utt_pad_proportion=0.05319, over 10430.68 utterances.], batch size: 46, lr: 4.87e-03, grad_scale: 8.0 2023-03-09 00:58:33,084 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86579.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:58:34,890 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86580.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:58:44,196 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86586.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 00:59:09,337 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-03-09 00:59:09,711 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86602.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:59:21,699 INFO [train2.py:809] (1/4) Epoch 22, batch 2950, loss[ctc_loss=0.084, att_loss=0.2581, loss=0.2233, over 17037.00 frames. utt_duration=1287 frames, utt_pad_proportion=0.009826, over 53.00 utterances.], tot_loss[ctc_loss=0.07163, att_loss=0.2355, loss=0.2027, over 3280619.78 frames. utt_duration=1259 frames, utt_pad_proportion=0.04889, over 10438.25 utterances.], batch size: 53, lr: 4.87e-03, grad_scale: 8.0 2023-03-09 00:59:23,653 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.8835, 3.6361, 3.0347, 3.2799, 3.8006, 3.5396, 2.9016, 4.0145], device='cuda:1'), covar=tensor([0.1075, 0.0505, 0.1086, 0.0737, 0.0720, 0.0698, 0.0902, 0.0486], device='cuda:1'), in_proj_covar=tensor([0.0203, 0.0217, 0.0228, 0.0201, 0.0281, 0.0242, 0.0200, 0.0289], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-09 00:59:28,864 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86613.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 00:59:49,126 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.0711, 2.7528, 3.0807, 4.0279, 3.6129, 3.5791, 2.7595, 2.1863], device='cuda:1'), covar=tensor([0.0814, 0.1848, 0.0955, 0.0574, 0.0879, 0.0566, 0.1419, 0.2065], device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0211, 0.0188, 0.0217, 0.0223, 0.0180, 0.0198, 0.0186], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 01:00:01,851 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86634.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 01:00:11,407 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4818, 2.7555, 4.8615, 3.6551, 3.0151, 4.1038, 4.6689, 4.5597], device='cuda:1'), covar=tensor([0.0241, 0.1525, 0.0203, 0.1199, 0.1767, 0.0280, 0.0156, 0.0272], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0242, 0.0193, 0.0317, 0.0265, 0.0218, 0.0181, 0.0211], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 01:00:12,897 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86641.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:00:15,586 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.364e+02 1.951e+02 2.348e+02 3.080e+02 7.280e+02, threshold=4.697e+02, percent-clipped=5.0 2023-03-09 01:00:37,276 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.27 vs. limit=5.0 2023-03-09 01:00:41,172 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86658.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:00:42,432 INFO [train2.py:809] (1/4) Epoch 22, batch 3000, loss[ctc_loss=0.04773, att_loss=0.1993, loss=0.169, over 15749.00 frames. utt_duration=1660 frames, utt_pad_proportion=0.009798, over 38.00 utterances.], tot_loss[ctc_loss=0.07135, att_loss=0.2351, loss=0.2024, over 3279347.57 frames. utt_duration=1255 frames, utt_pad_proportion=0.04833, over 10461.72 utterances.], batch size: 38, lr: 4.87e-03, grad_scale: 8.0 2023-03-09 01:00:42,432 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-09 01:00:55,507 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2615, 4.1901, 4.4303, 4.2536, 4.9017, 4.4279, 4.3436, 2.6020], device='cuda:1'), covar=tensor([0.0271, 0.0396, 0.0357, 0.0392, 0.0562, 0.0229, 0.0345, 0.1697], device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0193, 0.0191, 0.0208, 0.0366, 0.0161, 0.0180, 0.0216], 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') 2023-03-09 01:00:55,701 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9345, 4.0266, 3.9256, 3.7289, 4.3198, 4.0634, 4.0139, 2.3007], device='cuda:1'), covar=tensor([0.0369, 0.0501, 0.0500, 0.0580, 0.1019, 0.0299, 0.0379, 0.1960], device='cuda:1'), in_proj_covar=tensor([0.0166, 0.0193, 0.0191, 0.0208, 0.0366, 0.0161, 0.0180, 0.0216], 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') 2023-03-09 01:00:57,104 INFO [train2.py:843] (1/4) Epoch 22, validation: ctc_loss=0.03993, att_loss=0.2347, loss=0.1957, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-09 01:00:57,104 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-09 01:01:01,031 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86661.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:02:11,054 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86706.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:02:14,491 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86708.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:02:16,483 INFO [train2.py:809] (1/4) Epoch 22, batch 3050, loss[ctc_loss=0.04942, att_loss=0.2231, loss=0.1884, over 16271.00 frames. utt_duration=1515 frames, utt_pad_proportion=0.007889, over 43.00 utterances.], tot_loss[ctc_loss=0.0709, att_loss=0.234, loss=0.2014, over 3267538.24 frames. utt_duration=1298 frames, utt_pad_proportion=0.04174, over 10083.24 utterances.], batch size: 43, lr: 4.86e-03, grad_scale: 8.0 2023-03-09 01:02:43,489 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86725.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:02:48,347 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2601, 4.9196, 5.1932, 2.5902, 2.0277, 2.7997, 2.7381, 3.8615], device='cuda:1'), covar=tensor([0.0877, 0.0642, 0.0250, 0.3824, 0.6323, 0.2922, 0.3290, 0.1824], device='cuda:1'), in_proj_covar=tensor([0.0361, 0.0280, 0.0270, 0.0248, 0.0346, 0.0337, 0.0259, 0.0371], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-09 01:03:11,087 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.329e+02 1.829e+02 2.203e+02 2.781e+02 6.356e+02, threshold=4.407e+02, percent-clipped=2.0 2023-03-09 01:03:31,279 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86756.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:03:36,297 INFO [train2.py:809] (1/4) Epoch 22, batch 3100, loss[ctc_loss=0.05073, att_loss=0.2115, loss=0.1793, over 15856.00 frames. utt_duration=1628 frames, utt_pad_proportion=0.01098, over 39.00 utterances.], tot_loss[ctc_loss=0.07086, att_loss=0.2335, loss=0.201, over 3268306.46 frames. utt_duration=1319 frames, utt_pad_proportion=0.03721, over 9921.68 utterances.], batch size: 39, lr: 4.86e-03, grad_scale: 8.0 2023-03-09 01:03:51,120 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.5645, 2.2851, 2.0654, 2.4376, 2.6790, 2.5059, 2.4319, 2.8929], device='cuda:1'), covar=tensor([0.1719, 0.2920, 0.2370, 0.1550, 0.2359, 0.1286, 0.2211, 0.1018], device='cuda:1'), in_proj_covar=tensor([0.0120, 0.0127, 0.0123, 0.0111, 0.0127, 0.0107, 0.0129, 0.0102], device='cuda:1'), out_proj_covar=tensor([9.0889e-05, 9.8162e-05, 9.7958e-05, 8.6993e-05, 9.4429e-05, 8.6387e-05, 9.8634e-05, 8.1431e-05], device='cuda:1') 2023-03-09 01:03:59,916 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86773.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:04:35,405 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86796.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:04:43,622 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86801.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 01:04:57,204 INFO [train2.py:809] (1/4) Epoch 22, batch 3150, loss[ctc_loss=0.06006, att_loss=0.2282, loss=0.1946, over 16566.00 frames. utt_duration=1474 frames, utt_pad_proportion=0.004982, over 45.00 utterances.], tot_loss[ctc_loss=0.07071, att_loss=0.2334, loss=0.2009, over 3267959.03 frames. utt_duration=1323 frames, utt_pad_proportion=0.03698, over 9891.35 utterances.], batch size: 45, lr: 4.86e-03, grad_scale: 8.0 2023-03-09 01:05:06,057 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86814.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:05:51,841 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.403e+02 1.971e+02 2.330e+02 2.834e+02 6.258e+02, threshold=4.661e+02, percent-clipped=4.0 2023-03-09 01:06:18,415 INFO [train2.py:809] (1/4) Epoch 22, batch 3200, loss[ctc_loss=0.07476, att_loss=0.2499, loss=0.2149, over 16464.00 frames. utt_duration=1433 frames, utt_pad_proportion=0.007497, over 46.00 utterances.], tot_loss[ctc_loss=0.07135, att_loss=0.2344, loss=0.2018, over 3274767.55 frames. utt_duration=1280 frames, utt_pad_proportion=0.04469, over 10245.71 utterances.], batch size: 46, lr: 4.86e-03, grad_scale: 8.0 2023-03-09 01:06:45,833 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86875.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:06:50,305 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86878.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:06:50,411 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5834, 2.7528, 5.0602, 4.0790, 3.0251, 4.3279, 4.9310, 4.7676], device='cuda:1'), covar=tensor([0.0265, 0.1390, 0.0230, 0.0783, 0.1712, 0.0258, 0.0138, 0.0247], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0239, 0.0192, 0.0313, 0.0261, 0.0216, 0.0180, 0.0209], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 01:07:27,699 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86902.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:07:39,435 INFO [train2.py:809] (1/4) Epoch 22, batch 3250, loss[ctc_loss=0.05565, att_loss=0.2061, loss=0.176, over 11869.00 frames. utt_duration=1827 frames, utt_pad_proportion=0.1683, over 26.00 utterances.], tot_loss[ctc_loss=0.0707, att_loss=0.2341, loss=0.2014, over 3275461.66 frames. utt_duration=1283 frames, utt_pad_proportion=0.04259, over 10221.71 utterances.], batch size: 26, lr: 4.86e-03, grad_scale: 8.0 2023-03-09 01:07:39,885 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1788, 4.2736, 4.4748, 4.7523, 2.7550, 4.2070, 2.5867, 1.7193], device='cuda:1'), covar=tensor([0.0391, 0.0275, 0.0531, 0.0171, 0.1558, 0.0238, 0.1538, 0.1724], device='cuda:1'), in_proj_covar=tensor([0.0193, 0.0168, 0.0257, 0.0157, 0.0219, 0.0148, 0.0228, 0.0202], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 01:08:22,563 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86936.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:08:27,467 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86939.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:08:33,101 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.070e+02 1.904e+02 2.296e+02 2.851e+02 6.880e+02, threshold=4.593e+02, percent-clipped=1.0 2023-03-09 01:08:44,098 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=86950.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:08:59,941 INFO [train2.py:809] (1/4) Epoch 22, batch 3300, loss[ctc_loss=0.06417, att_loss=0.2168, loss=0.1862, over 15904.00 frames. utt_duration=1632 frames, utt_pad_proportion=0.007669, over 39.00 utterances.], tot_loss[ctc_loss=0.07076, att_loss=0.2339, loss=0.2013, over 3276767.52 frames. utt_duration=1280 frames, utt_pad_proportion=0.0429, over 10251.08 utterances.], batch size: 39, lr: 4.86e-03, grad_scale: 8.0 2023-03-09 01:09:52,348 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7294, 3.6306, 3.0567, 3.3374, 3.8210, 3.5968, 2.7800, 4.0612], device='cuda:1'), covar=tensor([0.1297, 0.0523, 0.1147, 0.0721, 0.0811, 0.0748, 0.0966, 0.0528], device='cuda:1'), in_proj_covar=tensor([0.0206, 0.0219, 0.0229, 0.0204, 0.0284, 0.0245, 0.0202, 0.0292], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-09 01:09:56,912 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.5866, 3.1836, 3.5853, 4.6545, 4.0489, 3.9734, 3.0636, 2.2140], device='cuda:1'), covar=tensor([0.0704, 0.1683, 0.0877, 0.0429, 0.0881, 0.0501, 0.1409, 0.2278], device='cuda:1'), in_proj_covar=tensor([0.0179, 0.0211, 0.0187, 0.0218, 0.0224, 0.0179, 0.0199, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 01:10:08,644 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87002.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 01:10:20,445 INFO [train2.py:809] (1/4) Epoch 22, batch 3350, loss[ctc_loss=0.09962, att_loss=0.2496, loss=0.2196, over 13995.00 frames. utt_duration=385 frames, utt_pad_proportion=0.328, over 146.00 utterances.], tot_loss[ctc_loss=0.07026, att_loss=0.2338, loss=0.2011, over 3276830.53 frames. utt_duration=1280 frames, utt_pad_proportion=0.04463, over 10249.75 utterances.], batch size: 146, lr: 4.86e-03, grad_scale: 8.0 2023-03-09 01:11:13,697 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.148e+02 1.863e+02 2.152e+02 2.724e+02 4.782e+02, threshold=4.303e+02, percent-clipped=1.0 2023-03-09 01:11:40,084 INFO [train2.py:809] (1/4) Epoch 22, batch 3400, loss[ctc_loss=0.05334, att_loss=0.2339, loss=0.1978, over 16908.00 frames. utt_duration=1382 frames, utt_pad_proportion=0.005348, over 49.00 utterances.], tot_loss[ctc_loss=0.07124, att_loss=0.2346, loss=0.2019, over 3276321.41 frames. utt_duration=1259 frames, utt_pad_proportion=0.05042, over 10419.12 utterances.], batch size: 49, lr: 4.85e-03, grad_scale: 8.0 2023-03-09 01:11:43,132 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.14 vs. limit=5.0 2023-03-09 01:11:47,042 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87063.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 01:11:50,033 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2973, 4.4797, 4.6360, 4.9921, 2.7084, 4.3704, 2.9044, 1.7425], device='cuda:1'), covar=tensor([0.0352, 0.0249, 0.0552, 0.0138, 0.1632, 0.0222, 0.1343, 0.1718], device='cuda:1'), in_proj_covar=tensor([0.0196, 0.0170, 0.0262, 0.0160, 0.0223, 0.0150, 0.0232, 0.0205], 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') 2023-03-09 01:12:09,703 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7814, 2.2903, 2.6058, 2.5883, 2.9318, 2.6898, 2.5838, 3.1696], device='cuda:1'), covar=tensor([0.1863, 0.2645, 0.1783, 0.1493, 0.1768, 0.1421, 0.2147, 0.1073], device='cuda:1'), in_proj_covar=tensor([0.0121, 0.0127, 0.0123, 0.0111, 0.0127, 0.0108, 0.0130, 0.0103], device='cuda:1'), out_proj_covar=tensor([9.1958e-05, 9.8534e-05, 9.8157e-05, 8.7193e-05, 9.4769e-05, 8.6862e-05, 9.9114e-05, 8.2111e-05], device='cuda:1') 2023-03-09 01:12:38,089 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87096.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:12:46,635 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87101.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 01:12:59,258 INFO [train2.py:809] (1/4) Epoch 22, batch 3450, loss[ctc_loss=0.08927, att_loss=0.2578, loss=0.2241, over 16874.00 frames. utt_duration=690.1 frames, utt_pad_proportion=0.1352, over 98.00 utterances.], tot_loss[ctc_loss=0.07302, att_loss=0.2361, loss=0.2035, over 3280852.35 frames. utt_duration=1209 frames, utt_pad_proportion=0.0615, over 10865.63 utterances.], batch size: 98, lr: 4.85e-03, grad_scale: 8.0 2023-03-09 01:13:52,180 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.163e+02 1.972e+02 2.278e+02 2.813e+02 5.019e+02, threshold=4.557e+02, percent-clipped=5.0 2023-03-09 01:13:53,984 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=87144.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:14:02,269 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=87149.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:14:16,709 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8347, 5.2835, 5.0896, 5.1871, 5.2317, 4.9621, 3.9098, 5.3194], device='cuda:1'), covar=tensor([0.0106, 0.0096, 0.0120, 0.0077, 0.0094, 0.0097, 0.0577, 0.0151], device='cuda:1'), in_proj_covar=tensor([0.0092, 0.0089, 0.0112, 0.0070, 0.0077, 0.0087, 0.0105, 0.0110], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 01:14:19,571 INFO [train2.py:809] (1/4) Epoch 22, batch 3500, loss[ctc_loss=0.05915, att_loss=0.2072, loss=0.1776, over 15491.00 frames. utt_duration=1723 frames, utt_pad_proportion=0.009472, over 36.00 utterances.], tot_loss[ctc_loss=0.07192, att_loss=0.235, loss=0.2024, over 3280304.18 frames. utt_duration=1233 frames, utt_pad_proportion=0.05501, over 10651.50 utterances.], batch size: 36, lr: 4.85e-03, grad_scale: 8.0 2023-03-09 01:14:37,057 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87170.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:15:39,622 INFO [train2.py:809] (1/4) Epoch 22, batch 3550, loss[ctc_loss=0.07892, att_loss=0.2501, loss=0.2159, over 17038.00 frames. utt_duration=1312 frames, utt_pad_proportion=0.008155, over 52.00 utterances.], tot_loss[ctc_loss=0.07181, att_loss=0.2354, loss=0.2027, over 3287227.30 frames. utt_duration=1254 frames, utt_pad_proportion=0.04947, over 10500.67 utterances.], batch size: 52, lr: 4.85e-03, grad_scale: 8.0 2023-03-09 01:15:40,751 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-03-09 01:16:12,827 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87230.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:16:18,890 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87234.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:16:22,085 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87236.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:16:32,599 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.105e+02 1.973e+02 2.339e+02 2.801e+02 5.420e+02, threshold=4.678e+02, percent-clipped=3.0 2023-03-09 01:16:59,777 INFO [train2.py:809] (1/4) Epoch 22, batch 3600, loss[ctc_loss=0.06106, att_loss=0.2349, loss=0.2001, over 16632.00 frames. utt_duration=1417 frames, utt_pad_proportion=0.00505, over 47.00 utterances.], tot_loss[ctc_loss=0.072, att_loss=0.2352, loss=0.2025, over 3282819.24 frames. utt_duration=1256 frames, utt_pad_proportion=0.04997, over 10465.44 utterances.], batch size: 47, lr: 4.85e-03, grad_scale: 16.0 2023-03-09 01:17:39,328 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=87284.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:17:50,587 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87291.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:17:55,889 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.5388, 2.1303, 2.2619, 2.6079, 2.7600, 2.5045, 2.4041, 3.0036], device='cuda:1'), covar=tensor([0.2278, 0.3285, 0.2240, 0.1362, 0.1921, 0.1392, 0.2359, 0.1121], device='cuda:1'), in_proj_covar=tensor([0.0124, 0.0130, 0.0126, 0.0114, 0.0131, 0.0111, 0.0134, 0.0105], device='cuda:1'), out_proj_covar=tensor([9.4200e-05, 1.0072e-04, 1.0061e-04, 8.9390e-05, 9.7623e-05, 8.9403e-05, 1.0166e-04, 8.3963e-05], device='cuda:1') 2023-03-09 01:18:19,834 INFO [train2.py:809] (1/4) Epoch 22, batch 3650, loss[ctc_loss=0.09058, att_loss=0.2438, loss=0.2132, over 16931.00 frames. utt_duration=692.6 frames, utt_pad_proportion=0.1321, over 98.00 utterances.], tot_loss[ctc_loss=0.07181, att_loss=0.2349, loss=0.2023, over 3281631.04 frames. utt_duration=1245 frames, utt_pad_proportion=0.05257, over 10553.26 utterances.], batch size: 98, lr: 4.85e-03, grad_scale: 16.0 2023-03-09 01:19:13,097 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.232e+02 1.934e+02 2.315e+02 2.927e+02 6.600e+02, threshold=4.631e+02, percent-clipped=3.0 2023-03-09 01:19:38,715 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87358.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 01:19:40,009 INFO [train2.py:809] (1/4) Epoch 22, batch 3700, loss[ctc_loss=0.101, att_loss=0.2575, loss=0.2262, over 17422.00 frames. utt_duration=883.6 frames, utt_pad_proportion=0.0757, over 79.00 utterances.], tot_loss[ctc_loss=0.07202, att_loss=0.235, loss=0.2024, over 3283308.95 frames. utt_duration=1245 frames, utt_pad_proportion=0.053, over 10560.80 utterances.], batch size: 79, lr: 4.85e-03, grad_scale: 16.0 2023-03-09 01:19:44,780 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87362.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:20:59,238 INFO [train2.py:809] (1/4) Epoch 22, batch 3750, loss[ctc_loss=0.07952, att_loss=0.2537, loss=0.2189, over 17139.00 frames. utt_duration=1226 frames, utt_pad_proportion=0.01311, over 56.00 utterances.], tot_loss[ctc_loss=0.07205, att_loss=0.2349, loss=0.2023, over 3288387.86 frames. utt_duration=1244 frames, utt_pad_proportion=0.05173, over 10588.83 utterances.], batch size: 56, lr: 4.84e-03, grad_scale: 16.0 2023-03-09 01:21:21,006 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-03-09 01:21:21,554 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87423.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:21:46,244 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.1278, 5.3334, 5.6836, 5.4970, 5.5638, 6.0970, 5.2835, 6.1530], device='cuda:1'), covar=tensor([0.0731, 0.0748, 0.0753, 0.1301, 0.1886, 0.0845, 0.0711, 0.0657], device='cuda:1'), in_proj_covar=tensor([0.0871, 0.0509, 0.0604, 0.0664, 0.0873, 0.0635, 0.0487, 0.0609], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 01:21:52,784 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.314e+02 1.951e+02 2.463e+02 3.155e+02 5.552e+02, threshold=4.927e+02, percent-clipped=3.0 2023-03-09 01:22:06,873 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9843, 5.2147, 5.1781, 5.1304, 5.2459, 5.2023, 4.8935, 4.7018], device='cuda:1'), covar=tensor([0.0956, 0.0591, 0.0288, 0.0489, 0.0287, 0.0307, 0.0411, 0.0363], device='cuda:1'), in_proj_covar=tensor([0.0526, 0.0369, 0.0349, 0.0362, 0.0425, 0.0436, 0.0360, 0.0398], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0003, 0.0004], device='cuda:1') 2023-03-09 01:22:19,688 INFO [train2.py:809] (1/4) Epoch 22, batch 3800, loss[ctc_loss=0.08204, att_loss=0.2486, loss=0.2153, over 17512.00 frames. utt_duration=1017 frames, utt_pad_proportion=0.04081, over 69.00 utterances.], tot_loss[ctc_loss=0.07223, att_loss=0.235, loss=0.2024, over 3290037.99 frames. utt_duration=1241 frames, utt_pad_proportion=0.05221, over 10618.17 utterances.], batch size: 69, lr: 4.84e-03, grad_scale: 16.0 2023-03-09 01:22:23,100 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87461.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:22:37,212 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87470.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:23:34,627 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87505.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:23:40,702 INFO [train2.py:809] (1/4) Epoch 22, batch 3850, loss[ctc_loss=0.05577, att_loss=0.2246, loss=0.1908, over 16274.00 frames. utt_duration=1515 frames, utt_pad_proportion=0.007645, over 43.00 utterances.], tot_loss[ctc_loss=0.07131, att_loss=0.2346, loss=0.202, over 3289272.67 frames. utt_duration=1255 frames, utt_pad_proportion=0.04767, over 10498.46 utterances.], batch size: 43, lr: 4.84e-03, grad_scale: 16.0 2023-03-09 01:23:54,828 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=87518.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:24:01,340 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87522.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:24:11,113 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-03-09 01:24:19,722 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87534.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:24:33,081 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.176e+02 1.913e+02 2.224e+02 2.689e+02 7.925e+02, threshold=4.448e+02, percent-clipped=4.0 2023-03-09 01:24:57,882 INFO [train2.py:809] (1/4) Epoch 22, batch 3900, loss[ctc_loss=0.08599, att_loss=0.2506, loss=0.2177, over 17338.00 frames. utt_duration=1177 frames, utt_pad_proportion=0.02244, over 59.00 utterances.], tot_loss[ctc_loss=0.07099, att_loss=0.2339, loss=0.2013, over 3281620.87 frames. utt_duration=1258 frames, utt_pad_proportion=0.04956, over 10450.28 utterances.], batch size: 59, lr: 4.84e-03, grad_scale: 16.0 2023-03-09 01:25:08,969 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87566.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 01:25:23,654 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.42 vs. limit=5.0 2023-03-09 01:25:33,515 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=87582.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:25:36,928 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87584.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:25:39,681 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87586.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:25:44,722 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87589.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:26:09,195 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87605.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:26:14,867 INFO [train2.py:809] (1/4) Epoch 22, batch 3950, loss[ctc_loss=0.07429, att_loss=0.2525, loss=0.2168, over 17273.00 frames. utt_duration=1258 frames, utt_pad_proportion=0.01276, over 55.00 utterances.], tot_loss[ctc_loss=0.07134, att_loss=0.2341, loss=0.2016, over 3281949.69 frames. utt_duration=1261 frames, utt_pad_proportion=0.04828, over 10424.66 utterances.], batch size: 55, lr: 4.84e-03, grad_scale: 16.0 2023-03-09 01:26:28,006 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-03-09 01:26:33,110 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5628, 5.2617, 5.3863, 5.3153, 5.1211, 5.2284, 5.0617, 4.7426], device='cuda:1'), covar=tensor([0.1611, 0.0724, 0.0342, 0.0563, 0.0822, 0.0419, 0.0451, 0.0441], device='cuda:1'), in_proj_covar=tensor([0.0530, 0.0371, 0.0351, 0.0364, 0.0429, 0.0440, 0.0362, 0.0401], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004], device='cuda:1') 2023-03-09 01:26:43,767 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0880, 5.3547, 5.5793, 5.4125, 5.5969, 6.0034, 5.2269, 6.0985], device='cuda:1'), covar=tensor([0.0627, 0.0644, 0.0860, 0.1353, 0.1649, 0.0854, 0.0738, 0.0628], device='cuda:1'), in_proj_covar=tensor([0.0879, 0.0510, 0.0611, 0.0666, 0.0877, 0.0638, 0.0491, 0.0617], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 01:27:30,756 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.383e+02 1.991e+02 2.302e+02 2.874e+02 7.527e+02, threshold=4.603e+02, percent-clipped=2.0 2023-03-09 01:27:30,800 INFO [train2.py:809] (1/4) Epoch 23, batch 0, loss[ctc_loss=0.05172, att_loss=0.237, loss=0.2, over 17310.00 frames. utt_duration=1100 frames, utt_pad_proportion=0.03901, over 63.00 utterances.], tot_loss[ctc_loss=0.05172, att_loss=0.237, loss=0.2, over 17310.00 frames. utt_duration=1100 frames, utt_pad_proportion=0.03901, over 63.00 utterances.], batch size: 63, lr: 4.73e-03, grad_scale: 16.0 2023-03-09 01:27:30,800 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-09 01:27:42,674 INFO [train2.py:843] (1/4) Epoch 23, validation: ctc_loss=0.04039, att_loss=0.2346, loss=0.1958, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-09 01:27:42,675 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-09 01:27:46,039 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87645.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 01:27:53,645 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87650.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:28:06,065 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87658.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 01:28:18,446 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87666.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 01:28:58,598 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-03-09 01:29:02,026 INFO [train2.py:809] (1/4) Epoch 23, batch 50, loss[ctc_loss=0.08657, att_loss=0.251, loss=0.2181, over 17367.00 frames. utt_duration=880.9 frames, utt_pad_proportion=0.07761, over 79.00 utterances.], tot_loss[ctc_loss=0.07039, att_loss=0.2366, loss=0.2034, over 747939.98 frames. utt_duration=1158 frames, utt_pad_proportion=0.0577, over 2586.81 utterances.], batch size: 79, lr: 4.73e-03, grad_scale: 16.0 2023-03-09 01:29:22,819 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=87706.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 01:29:41,129 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87718.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:29:54,233 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2725, 4.5057, 4.7256, 4.8138, 2.6202, 4.6749, 2.7530, 1.9941], device='cuda:1'), covar=tensor([0.0380, 0.0268, 0.0585, 0.0184, 0.1651, 0.0184, 0.1467, 0.1615], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0172, 0.0264, 0.0161, 0.0224, 0.0153, 0.0234, 0.0206], 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') 2023-03-09 01:30:21,899 INFO [train2.py:809] (1/4) Epoch 23, batch 100, loss[ctc_loss=0.06409, att_loss=0.2387, loss=0.2038, over 17021.00 frames. utt_duration=1337 frames, utt_pad_proportion=0.007715, over 51.00 utterances.], tot_loss[ctc_loss=0.07212, att_loss=0.2357, loss=0.2029, over 1302036.84 frames. utt_duration=1179 frames, utt_pad_proportion=0.06943, over 4423.23 utterances.], batch size: 51, lr: 4.73e-03, grad_scale: 8.0 2023-03-09 01:30:23,392 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.349e+02 2.012e+02 2.393e+02 2.818e+02 6.680e+02, threshold=4.785e+02, percent-clipped=2.0 2023-03-09 01:30:42,577 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2886, 2.9603, 3.2562, 4.3635, 3.9641, 3.8949, 2.9547, 2.4532], device='cuda:1'), covar=tensor([0.0830, 0.1899, 0.1011, 0.0604, 0.0798, 0.0449, 0.1509, 0.2039], device='cuda:1'), in_proj_covar=tensor([0.0183, 0.0217, 0.0192, 0.0224, 0.0229, 0.0183, 0.0205, 0.0192], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 01:31:41,403 INFO [train2.py:809] (1/4) Epoch 23, batch 150, loss[ctc_loss=0.05177, att_loss=0.2188, loss=0.1854, over 16395.00 frames. utt_duration=1492 frames, utt_pad_proportion=0.007455, over 44.00 utterances.], tot_loss[ctc_loss=0.07104, att_loss=0.2358, loss=0.2029, over 1749413.04 frames. utt_duration=1246 frames, utt_pad_proportion=0.04969, over 5622.91 utterances.], batch size: 44, lr: 4.73e-03, grad_scale: 8.0 2023-03-09 01:32:20,519 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87817.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:32:55,594 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.79 vs. limit=5.0 2023-03-09 01:33:02,589 INFO [train2.py:809] (1/4) Epoch 23, batch 200, loss[ctc_loss=0.07533, att_loss=0.2202, loss=0.1912, over 15352.00 frames. utt_duration=1756 frames, utt_pad_proportion=0.01196, over 35.00 utterances.], tot_loss[ctc_loss=0.07143, att_loss=0.2358, loss=0.2029, over 2089882.85 frames. utt_duration=1238 frames, utt_pad_proportion=0.05295, over 6762.03 utterances.], batch size: 35, lr: 4.72e-03, grad_scale: 8.0 2023-03-09 01:33:04,010 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.173e+02 1.980e+02 2.350e+02 2.635e+02 6.250e+02, threshold=4.701e+02, percent-clipped=5.0 2023-03-09 01:33:16,327 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-03-09 01:33:31,409 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87861.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 01:34:11,175 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87886.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:34:22,277 INFO [train2.py:809] (1/4) Epoch 23, batch 250, loss[ctc_loss=0.05382, att_loss=0.197, loss=0.1683, over 15494.00 frames. utt_duration=1723 frames, utt_pad_proportion=0.008583, over 36.00 utterances.], tot_loss[ctc_loss=0.07145, att_loss=0.2356, loss=0.2028, over 2351036.47 frames. utt_duration=1228 frames, utt_pad_proportion=0.05748, over 7668.87 utterances.], batch size: 36, lr: 4.72e-03, grad_scale: 8.0 2023-03-09 01:35:27,224 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=87934.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:35:37,866 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87940.0, num_to_drop=1, layers_to_drop={3} 2023-03-09 01:35:42,333 INFO [train2.py:809] (1/4) Epoch 23, batch 300, loss[ctc_loss=0.06121, att_loss=0.2077, loss=0.1784, over 15663.00 frames. utt_duration=1695 frames, utt_pad_proportion=0.007283, over 37.00 utterances.], tot_loss[ctc_loss=0.0706, att_loss=0.2343, loss=0.2016, over 2554712.16 frames. utt_duration=1257 frames, utt_pad_proportion=0.04979, over 8138.40 utterances.], batch size: 37, lr: 4.72e-03, grad_scale: 8.0 2023-03-09 01:35:43,860 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.187e+02 1.810e+02 2.179e+02 2.816e+02 7.106e+02, threshold=4.359e+02, percent-clipped=3.0 2023-03-09 01:35:45,671 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87945.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:36:10,771 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87961.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 01:36:15,616 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.4910, 3.1038, 3.5325, 4.4707, 3.9977, 3.9883, 2.9825, 2.3311], device='cuda:1'), covar=tensor([0.0773, 0.1757, 0.0921, 0.0627, 0.0868, 0.0504, 0.1633, 0.2208], device='cuda:1'), in_proj_covar=tensor([0.0181, 0.0215, 0.0191, 0.0223, 0.0227, 0.0182, 0.0204, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 01:36:55,333 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.64 vs. limit=5.0 2023-03-09 01:37:02,476 INFO [train2.py:809] (1/4) Epoch 23, batch 350, loss[ctc_loss=0.08519, att_loss=0.2535, loss=0.2198, over 16642.00 frames. utt_duration=680.7 frames, utt_pad_proportion=0.1469, over 98.00 utterances.], tot_loss[ctc_loss=0.07148, att_loss=0.2352, loss=0.2025, over 2713861.65 frames. utt_duration=1230 frames, utt_pad_proportion=0.05654, over 8833.11 utterances.], batch size: 98, lr: 4.72e-03, grad_scale: 8.0 2023-03-09 01:37:34,097 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7541, 4.7618, 4.3256, 2.5936, 4.4605, 4.4891, 3.8872, 2.4528], device='cuda:1'), covar=tensor([0.0193, 0.0141, 0.0432, 0.1313, 0.0146, 0.0262, 0.0470, 0.1786], device='cuda:1'), in_proj_covar=tensor([0.0076, 0.0103, 0.0107, 0.0112, 0.0086, 0.0116, 0.0101, 0.0104], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-09 01:37:46,411 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88018.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:38:27,598 INFO [train2.py:809] (1/4) Epoch 23, batch 400, loss[ctc_loss=0.07963, att_loss=0.2485, loss=0.2147, over 16460.00 frames. utt_duration=1433 frames, utt_pad_proportion=0.007156, over 46.00 utterances.], tot_loss[ctc_loss=0.07186, att_loss=0.2356, loss=0.2029, over 2840301.03 frames. utt_duration=1226 frames, utt_pad_proportion=0.05738, over 9277.03 utterances.], batch size: 46, lr: 4.72e-03, grad_scale: 8.0 2023-03-09 01:38:29,098 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.200e+02 1.944e+02 2.395e+02 3.108e+02 5.272e+02, threshold=4.790e+02, percent-clipped=4.0 2023-03-09 01:38:42,590 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.6371, 2.3061, 2.2200, 2.6933, 2.9840, 2.5877, 2.4325, 2.9876], device='cuda:1'), covar=tensor([0.1217, 0.3016, 0.1871, 0.1065, 0.1288, 0.1194, 0.2268, 0.0998], device='cuda:1'), in_proj_covar=tensor([0.0123, 0.0128, 0.0124, 0.0114, 0.0129, 0.0110, 0.0132, 0.0105], device='cuda:1'), out_proj_covar=tensor([9.3339e-05, 9.9620e-05, 9.9240e-05, 8.9184e-05, 9.5920e-05, 8.8480e-05, 1.0075e-04, 8.3709e-05], device='cuda:1') 2023-03-09 01:39:04,772 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=88066.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:39:48,103 INFO [train2.py:809] (1/4) Epoch 23, batch 450, loss[ctc_loss=0.09115, att_loss=0.2517, loss=0.2196, over 17320.00 frames. utt_duration=1101 frames, utt_pad_proportion=0.03732, over 63.00 utterances.], tot_loss[ctc_loss=0.07173, att_loss=0.2346, loss=0.2021, over 2933400.17 frames. utt_duration=1240 frames, utt_pad_proportion=0.05674, over 9476.13 utterances.], batch size: 63, lr: 4.72e-03, grad_scale: 8.0 2023-03-09 01:40:00,132 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=88100.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:40:10,125 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-03-09 01:40:26,305 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88117.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:40:56,993 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-03-09 01:41:09,082 INFO [train2.py:809] (1/4) Epoch 23, batch 500, loss[ctc_loss=0.07008, att_loss=0.2174, loss=0.188, over 15881.00 frames. utt_duration=1630 frames, utt_pad_proportion=0.008854, over 39.00 utterances.], tot_loss[ctc_loss=0.07144, att_loss=0.2347, loss=0.2021, over 3007176.00 frames. utt_duration=1232 frames, utt_pad_proportion=0.05784, over 9772.33 utterances.], batch size: 39, lr: 4.72e-03, grad_scale: 8.0 2023-03-09 01:41:10,581 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.266e+02 1.977e+02 2.397e+02 3.059e+02 5.388e+02, threshold=4.794e+02, percent-clipped=2.0 2023-03-09 01:41:37,470 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88161.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 01:41:37,525 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.9330, 3.5986, 3.6474, 2.8647, 3.6397, 3.7220, 3.7200, 2.4319], device='cuda:1'), covar=tensor([0.1335, 0.1548, 0.1887, 0.5989, 0.1721, 0.2018, 0.0971, 0.6078], device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0193, 0.0206, 0.0259, 0.0165, 0.0267, 0.0189, 0.0219], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 01:41:37,545 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=88161.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:41:44,233 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=88165.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:41:50,489 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0213, 5.3636, 4.9639, 5.4258, 4.8259, 5.0571, 5.4957, 5.2654], device='cuda:1'), covar=tensor([0.0620, 0.0272, 0.0708, 0.0333, 0.0384, 0.0274, 0.0224, 0.0193], device='cuda:1'), in_proj_covar=tensor([0.0384, 0.0320, 0.0362, 0.0351, 0.0321, 0.0238, 0.0303, 0.0288], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0005, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-09 01:42:29,383 INFO [train2.py:809] (1/4) Epoch 23, batch 550, loss[ctc_loss=0.04951, att_loss=0.2362, loss=0.1989, over 16888.00 frames. utt_duration=1380 frames, utt_pad_proportion=0.006945, over 49.00 utterances.], tot_loss[ctc_loss=0.07129, att_loss=0.2347, loss=0.202, over 3070825.55 frames. utt_duration=1233 frames, utt_pad_proportion=0.05616, over 9975.96 utterances.], batch size: 49, lr: 4.72e-03, grad_scale: 8.0 2023-03-09 01:42:41,105 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0689, 4.3394, 4.4952, 4.5535, 2.6219, 4.4486, 2.8754, 1.8608], device='cuda:1'), covar=tensor([0.0457, 0.0296, 0.0537, 0.0232, 0.1577, 0.0205, 0.1292, 0.1573], device='cuda:1'), in_proj_covar=tensor([0.0200, 0.0172, 0.0265, 0.0164, 0.0226, 0.0154, 0.0235, 0.0208], 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') 2023-03-09 01:42:54,657 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=88209.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:43:46,359 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88240.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:43:51,891 INFO [train2.py:809] (1/4) Epoch 23, batch 600, loss[ctc_loss=0.05659, att_loss=0.2042, loss=0.1747, over 15381.00 frames. utt_duration=1759 frames, utt_pad_proportion=0.01043, over 35.00 utterances.], tot_loss[ctc_loss=0.0712, att_loss=0.235, loss=0.2022, over 3128042.38 frames. utt_duration=1240 frames, utt_pad_proportion=0.05031, over 10104.82 utterances.], batch size: 35, lr: 4.71e-03, grad_scale: 8.0 2023-03-09 01:43:53,455 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.322e+02 1.886e+02 2.333e+02 2.841e+02 4.176e+02, threshold=4.666e+02, percent-clipped=0.0 2023-03-09 01:43:55,488 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88245.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:44:20,756 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88261.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 01:45:04,421 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=88288.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:45:10,130 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=88291.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 01:45:12,840 INFO [train2.py:809] (1/4) Epoch 23, batch 650, loss[ctc_loss=0.06229, att_loss=0.2281, loss=0.195, over 15965.00 frames. utt_duration=1559 frames, utt_pad_proportion=0.006482, over 41.00 utterances.], tot_loss[ctc_loss=0.07175, att_loss=0.2353, loss=0.2026, over 3153886.14 frames. utt_duration=1219 frames, utt_pad_proportion=0.05862, over 10362.95 utterances.], batch size: 41, lr: 4.71e-03, grad_scale: 8.0 2023-03-09 01:45:12,985 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=88293.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:45:38,313 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=88309.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:46:34,465 INFO [train2.py:809] (1/4) Epoch 23, batch 700, loss[ctc_loss=0.09703, att_loss=0.2584, loss=0.2261, over 17364.00 frames. utt_duration=1104 frames, utt_pad_proportion=0.03472, over 63.00 utterances.], tot_loss[ctc_loss=0.07221, att_loss=0.236, loss=0.2032, over 3190767.03 frames. utt_duration=1231 frames, utt_pad_proportion=0.05342, over 10379.63 utterances.], batch size: 63, lr: 4.71e-03, grad_scale: 8.0 2023-03-09 01:46:35,998 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.287e+02 1.968e+02 2.314e+02 2.837e+02 1.022e+03, threshold=4.627e+02, percent-clipped=4.0 2023-03-09 01:46:49,149 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=88352.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 01:47:56,429 INFO [train2.py:809] (1/4) Epoch 23, batch 750, loss[ctc_loss=0.05186, att_loss=0.2033, loss=0.173, over 15376.00 frames. utt_duration=1759 frames, utt_pad_proportion=0.01091, over 35.00 utterances.], tot_loss[ctc_loss=0.07205, att_loss=0.2357, loss=0.2029, over 3214598.52 frames. utt_duration=1243 frames, utt_pad_proportion=0.05032, over 10357.46 utterances.], batch size: 35, lr: 4.71e-03, grad_scale: 8.0 2023-03-09 01:48:17,815 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.0801, 3.8165, 3.2510, 3.5551, 4.0679, 3.7133, 3.1138, 4.3486], device='cuda:1'), covar=tensor([0.0998, 0.0540, 0.1045, 0.0687, 0.0682, 0.0752, 0.0816, 0.0478], device='cuda:1'), in_proj_covar=tensor([0.0203, 0.0217, 0.0226, 0.0202, 0.0279, 0.0242, 0.0199, 0.0289], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-09 01:48:43,678 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.3252, 5.2158, 5.0840, 3.1505, 4.9863, 4.8588, 4.6069, 2.9600], device='cuda:1'), covar=tensor([0.0101, 0.0093, 0.0237, 0.1009, 0.0102, 0.0184, 0.0279, 0.1227], device='cuda:1'), in_proj_covar=tensor([0.0075, 0.0102, 0.0105, 0.0111, 0.0086, 0.0115, 0.0100, 0.0104], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-09 01:49:17,981 INFO [train2.py:809] (1/4) Epoch 23, batch 800, loss[ctc_loss=0.0566, att_loss=0.2384, loss=0.2021, over 16969.00 frames. utt_duration=1359 frames, utt_pad_proportion=0.006623, over 50.00 utterances.], tot_loss[ctc_loss=0.07143, att_loss=0.2347, loss=0.202, over 3229480.66 frames. utt_duration=1261 frames, utt_pad_proportion=0.04614, over 10255.32 utterances.], batch size: 50, lr: 4.71e-03, grad_scale: 8.0 2023-03-09 01:49:19,538 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.200e+02 1.999e+02 2.331e+02 2.822e+02 6.920e+02, threshold=4.662e+02, percent-clipped=4.0 2023-03-09 01:49:39,407 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=88456.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:49:46,286 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.06 vs. limit=5.0 2023-03-09 01:49:57,321 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.3207, 3.9269, 3.4022, 3.7009, 4.1551, 3.9615, 3.3655, 4.5314], device='cuda:1'), covar=tensor([0.0883, 0.0533, 0.0996, 0.0590, 0.0729, 0.0574, 0.0733, 0.0434], device='cuda:1'), in_proj_covar=tensor([0.0204, 0.0219, 0.0228, 0.0203, 0.0281, 0.0244, 0.0201, 0.0291], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-09 01:50:32,034 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0621, 5.1364, 5.0259, 2.0429, 1.9773, 2.7311, 2.5547, 3.7753], device='cuda:1'), covar=tensor([0.0740, 0.0280, 0.0185, 0.5409, 0.5947, 0.2845, 0.3601, 0.1799], device='cuda:1'), in_proj_covar=tensor([0.0363, 0.0284, 0.0272, 0.0249, 0.0344, 0.0337, 0.0259, 0.0373], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-09 01:50:39,454 INFO [train2.py:809] (1/4) Epoch 23, batch 850, loss[ctc_loss=0.05644, att_loss=0.2304, loss=0.1956, over 16632.00 frames. utt_duration=1417 frames, utt_pad_proportion=0.00514, over 47.00 utterances.], tot_loss[ctc_loss=0.07077, att_loss=0.234, loss=0.2013, over 3239008.42 frames. utt_duration=1289 frames, utt_pad_proportion=0.04064, over 10061.51 utterances.], batch size: 47, lr: 4.71e-03, grad_scale: 8.0 2023-03-09 01:50:53,938 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-03-09 01:51:04,720 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5059, 2.7329, 4.9938, 4.0856, 3.2149, 4.3098, 4.8815, 4.7221], device='cuda:1'), covar=tensor([0.0289, 0.1320, 0.0183, 0.0783, 0.1576, 0.0243, 0.0132, 0.0246], device='cuda:1'), in_proj_covar=tensor([0.0202, 0.0241, 0.0192, 0.0315, 0.0263, 0.0216, 0.0182, 0.0211], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 01:52:01,932 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-03-09 01:52:02,086 INFO [train2.py:809] (1/4) Epoch 23, batch 900, loss[ctc_loss=0.06457, att_loss=0.2296, loss=0.1966, over 16477.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.006835, over 46.00 utterances.], tot_loss[ctc_loss=0.07091, att_loss=0.2348, loss=0.202, over 3258870.67 frames. utt_duration=1293 frames, utt_pad_proportion=0.03696, over 10096.09 utterances.], batch size: 46, lr: 4.71e-03, grad_scale: 8.0 2023-03-09 01:52:03,748 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.224e+02 1.915e+02 2.378e+02 3.006e+02 6.830e+02, threshold=4.756e+02, percent-clipped=6.0 2023-03-09 01:52:55,379 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-03-09 01:53:24,723 INFO [train2.py:809] (1/4) Epoch 23, batch 950, loss[ctc_loss=0.06589, att_loss=0.2327, loss=0.1994, over 17028.00 frames. utt_duration=1337 frames, utt_pad_proportion=0.007395, over 51.00 utterances.], tot_loss[ctc_loss=0.07035, att_loss=0.2342, loss=0.2014, over 3260320.46 frames. utt_duration=1299 frames, utt_pad_proportion=0.0374, over 10053.88 utterances.], batch size: 51, lr: 4.70e-03, grad_scale: 8.0 2023-03-09 01:54:29,029 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-03-09 01:54:46,090 INFO [train2.py:809] (1/4) Epoch 23, batch 1000, loss[ctc_loss=0.06705, att_loss=0.2425, loss=0.2074, over 17036.00 frames. utt_duration=1338 frames, utt_pad_proportion=0.006856, over 51.00 utterances.], tot_loss[ctc_loss=0.07051, att_loss=0.2343, loss=0.2015, over 3258559.28 frames. utt_duration=1303 frames, utt_pad_proportion=0.03852, over 10017.73 utterances.], batch size: 51, lr: 4.70e-03, grad_scale: 8.0 2023-03-09 01:54:48,311 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.066e+02 1.864e+02 2.157e+02 2.686e+02 4.602e+02, threshold=4.313e+02, percent-clipped=0.0 2023-03-09 01:54:50,989 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.16 vs. limit=5.0 2023-03-09 01:54:52,477 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.42 vs. limit=5.0 2023-03-09 01:54:53,439 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=88647.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 01:55:17,154 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.8957, 3.5738, 3.0373, 3.2419, 3.7894, 3.5007, 2.6927, 3.9527], device='cuda:1'), covar=tensor([0.1007, 0.0524, 0.1092, 0.0772, 0.0728, 0.0717, 0.0983, 0.0614], device='cuda:1'), in_proj_covar=tensor([0.0206, 0.0220, 0.0230, 0.0205, 0.0284, 0.0246, 0.0203, 0.0293], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-09 01:56:09,549 INFO [train2.py:809] (1/4) Epoch 23, batch 1050, loss[ctc_loss=0.07311, att_loss=0.2572, loss=0.2204, over 17291.00 frames. utt_duration=1259 frames, utt_pad_proportion=0.01253, over 55.00 utterances.], tot_loss[ctc_loss=0.07044, att_loss=0.2344, loss=0.2016, over 3269268.53 frames. utt_duration=1302 frames, utt_pad_proportion=0.0364, over 10057.45 utterances.], batch size: 55, lr: 4.70e-03, grad_scale: 8.0 2023-03-09 01:56:09,722 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9531, 6.1711, 5.7164, 5.8437, 5.8445, 5.2891, 5.5989, 5.3636], device='cuda:1'), covar=tensor([0.1218, 0.0838, 0.0920, 0.0841, 0.0854, 0.1444, 0.2137, 0.2035], device='cuda:1'), in_proj_covar=tensor([0.0533, 0.0620, 0.0467, 0.0464, 0.0441, 0.0470, 0.0620, 0.0532], 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') 2023-03-09 01:56:13,251 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2639, 2.8151, 3.2917, 4.4013, 3.8947, 3.7656, 2.8522, 2.1471], device='cuda:1'), covar=tensor([0.0780, 0.1982, 0.0948, 0.0454, 0.0776, 0.0543, 0.1571, 0.2266], device='cuda:1'), in_proj_covar=tensor([0.0181, 0.0214, 0.0188, 0.0222, 0.0227, 0.0181, 0.0202, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 01:56:42,415 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2165, 5.1621, 5.0334, 2.2936, 2.0083, 2.9130, 2.5044, 3.9860], device='cuda:1'), covar=tensor([0.0669, 0.0354, 0.0230, 0.4571, 0.5489, 0.2508, 0.3449, 0.1585], device='cuda:1'), in_proj_covar=tensor([0.0358, 0.0281, 0.0270, 0.0245, 0.0340, 0.0333, 0.0257, 0.0369], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-03-09 01:57:30,981 INFO [train2.py:809] (1/4) Epoch 23, batch 1100, loss[ctc_loss=0.06676, att_loss=0.2375, loss=0.2033, over 16474.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.006011, over 46.00 utterances.], tot_loss[ctc_loss=0.07122, att_loss=0.2348, loss=0.2021, over 3271483.03 frames. utt_duration=1265 frames, utt_pad_proportion=0.04601, over 10354.73 utterances.], batch size: 46, lr: 4.70e-03, grad_scale: 8.0 2023-03-09 01:57:32,494 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.348e+02 2.060e+02 2.487e+02 3.193e+02 8.233e+02, threshold=4.974e+02, percent-clipped=4.0 2023-03-09 01:57:52,261 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88756.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 01:58:52,972 INFO [train2.py:809] (1/4) Epoch 23, batch 1150, loss[ctc_loss=0.08598, att_loss=0.2442, loss=0.2126, over 16486.00 frames. utt_duration=1435 frames, utt_pad_proportion=0.006112, over 46.00 utterances.], tot_loss[ctc_loss=0.07123, att_loss=0.2346, loss=0.2019, over 3271061.11 frames. utt_duration=1263 frames, utt_pad_proportion=0.04803, over 10368.65 utterances.], batch size: 46, lr: 4.70e-03, grad_scale: 8.0 2023-03-09 01:59:11,744 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=88804.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:00:16,048 INFO [train2.py:809] (1/4) Epoch 23, batch 1200, loss[ctc_loss=0.09226, att_loss=0.2575, loss=0.2244, over 17340.00 frames. utt_duration=1177 frames, utt_pad_proportion=0.02071, over 59.00 utterances.], tot_loss[ctc_loss=0.071, att_loss=0.2346, loss=0.2019, over 3268181.55 frames. utt_duration=1269 frames, utt_pad_proportion=0.04716, over 10310.41 utterances.], batch size: 59, lr: 4.70e-03, grad_scale: 8.0 2023-03-09 02:00:17,529 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.119e+02 1.914e+02 2.277e+02 2.770e+02 5.405e+02, threshold=4.555e+02, percent-clipped=1.0 2023-03-09 02:01:37,925 INFO [train2.py:809] (1/4) Epoch 23, batch 1250, loss[ctc_loss=0.06086, att_loss=0.2375, loss=0.2022, over 16871.00 frames. utt_duration=1379 frames, utt_pad_proportion=0.007376, over 49.00 utterances.], tot_loss[ctc_loss=0.07085, att_loss=0.2348, loss=0.202, over 3278399.18 frames. utt_duration=1270 frames, utt_pad_proportion=0.04537, over 10340.07 utterances.], batch size: 49, lr: 4.70e-03, grad_scale: 8.0 2023-03-09 02:02:41,998 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9757, 5.2352, 5.1485, 5.0527, 5.2434, 5.1964, 4.8449, 4.6691], device='cuda:1'), covar=tensor([0.0992, 0.0516, 0.0321, 0.0516, 0.0306, 0.0342, 0.0421, 0.0373], device='cuda:1'), in_proj_covar=tensor([0.0528, 0.0373, 0.0357, 0.0368, 0.0432, 0.0444, 0.0366, 0.0404], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:1') 2023-03-09 02:03:00,073 INFO [train2.py:809] (1/4) Epoch 23, batch 1300, loss[ctc_loss=0.06967, att_loss=0.25, loss=0.2139, over 16868.00 frames. utt_duration=1379 frames, utt_pad_proportion=0.008251, over 49.00 utterances.], tot_loss[ctc_loss=0.07124, att_loss=0.2349, loss=0.2022, over 3274220.29 frames. utt_duration=1268 frames, utt_pad_proportion=0.04696, over 10337.01 utterances.], batch size: 49, lr: 4.70e-03, grad_scale: 8.0 2023-03-09 02:03:01,705 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.195e+02 1.977e+02 2.384e+02 3.004e+02 6.014e+02, threshold=4.768e+02, percent-clipped=3.0 2023-03-09 02:03:06,862 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88947.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 02:03:32,318 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-03-09 02:03:39,802 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-03-09 02:04:21,365 INFO [train2.py:809] (1/4) Epoch 23, batch 1350, loss[ctc_loss=0.07263, att_loss=0.2116, loss=0.1838, over 15650.00 frames. utt_duration=1693 frames, utt_pad_proportion=0.008576, over 37.00 utterances.], tot_loss[ctc_loss=0.07185, att_loss=0.235, loss=0.2023, over 3268771.74 frames. utt_duration=1251 frames, utt_pad_proportion=0.05222, over 10460.32 utterances.], batch size: 37, lr: 4.69e-03, grad_scale: 8.0 2023-03-09 02:04:24,662 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=88995.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 02:04:57,585 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.4567, 2.3992, 2.1483, 2.7396, 2.5155, 2.2980, 2.2768, 2.9190], device='cuda:1'), covar=tensor([0.1970, 0.2446, 0.1998, 0.1712, 0.1643, 0.1407, 0.2104, 0.1404], device='cuda:1'), in_proj_covar=tensor([0.0126, 0.0130, 0.0124, 0.0116, 0.0131, 0.0112, 0.0135, 0.0106], device='cuda:1'), out_proj_covar=tensor([9.5481e-05, 1.0076e-04, 9.9649e-05, 9.0397e-05, 9.7525e-05, 9.0070e-05, 1.0234e-04, 8.4703e-05], device='cuda:1') 2023-03-09 02:05:17,170 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89027.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:05:43,317 INFO [train2.py:809] (1/4) Epoch 23, batch 1400, loss[ctc_loss=0.05803, att_loss=0.2084, loss=0.1783, over 15478.00 frames. utt_duration=1722 frames, utt_pad_proportion=0.009462, over 36.00 utterances.], tot_loss[ctc_loss=0.07147, att_loss=0.2345, loss=0.2019, over 3270293.47 frames. utt_duration=1263 frames, utt_pad_proportion=0.04992, over 10372.23 utterances.], batch size: 36, lr: 4.69e-03, grad_scale: 8.0 2023-03-09 02:05:44,870 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.382e+02 1.935e+02 2.297e+02 2.917e+02 5.540e+02, threshold=4.593e+02, percent-clipped=1.0 2023-03-09 02:06:43,782 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7960, 4.2192, 4.5243, 4.3090, 4.3868, 4.7297, 4.3992, 4.7935], device='cuda:1'), covar=tensor([0.0821, 0.0845, 0.0791, 0.1258, 0.1612, 0.0943, 0.1861, 0.0750], device='cuda:1'), in_proj_covar=tensor([0.0883, 0.0518, 0.0619, 0.0673, 0.0883, 0.0640, 0.0497, 0.0619], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 02:06:47,062 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.3871, 3.8668, 3.3734, 3.6763, 4.0926, 3.8171, 3.3881, 4.4013], device='cuda:1'), covar=tensor([0.0836, 0.0477, 0.1028, 0.0642, 0.0684, 0.0692, 0.0706, 0.0485], device='cuda:1'), in_proj_covar=tensor([0.0205, 0.0217, 0.0229, 0.0203, 0.0282, 0.0244, 0.0200, 0.0292], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-09 02:06:56,646 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89088.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:07:04,585 INFO [train2.py:809] (1/4) Epoch 23, batch 1450, loss[ctc_loss=0.09514, att_loss=0.2624, loss=0.2289, over 17122.00 frames. utt_duration=1225 frames, utt_pad_proportion=0.01474, over 56.00 utterances.], tot_loss[ctc_loss=0.07145, att_loss=0.2344, loss=0.2018, over 3267437.31 frames. utt_duration=1262 frames, utt_pad_proportion=0.05176, over 10365.19 utterances.], batch size: 56, lr: 4.69e-03, grad_scale: 8.0 2023-03-09 02:07:20,987 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.53 vs. limit=5.0 2023-03-09 02:07:52,865 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-03-09 02:08:26,635 INFO [train2.py:809] (1/4) Epoch 23, batch 1500, loss[ctc_loss=0.05923, att_loss=0.234, loss=0.199, over 17026.00 frames. utt_duration=1337 frames, utt_pad_proportion=0.008218, over 51.00 utterances.], tot_loss[ctc_loss=0.0712, att_loss=0.234, loss=0.2015, over 3269376.90 frames. utt_duration=1255 frames, utt_pad_proportion=0.05242, over 10433.81 utterances.], batch size: 51, lr: 4.69e-03, grad_scale: 8.0 2023-03-09 02:08:28,105 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.367e+02 1.977e+02 2.384e+02 2.734e+02 5.349e+02, threshold=4.768e+02, percent-clipped=3.0 2023-03-09 02:09:08,705 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89169.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:09:48,304 INFO [train2.py:809] (1/4) Epoch 23, batch 1550, loss[ctc_loss=0.0547, att_loss=0.2101, loss=0.179, over 15761.00 frames. utt_duration=1660 frames, utt_pad_proportion=0.008065, over 38.00 utterances.], tot_loss[ctc_loss=0.07154, att_loss=0.2341, loss=0.2016, over 3262408.96 frames. utt_duration=1222 frames, utt_pad_proportion=0.06242, over 10690.38 utterances.], batch size: 38, lr: 4.69e-03, grad_scale: 8.0 2023-03-09 02:10:17,050 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-03-09 02:10:49,441 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89230.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:11:10,189 INFO [train2.py:809] (1/4) Epoch 23, batch 1600, loss[ctc_loss=0.0601, att_loss=0.2185, loss=0.1868, over 16018.00 frames. utt_duration=1603 frames, utt_pad_proportion=0.007183, over 40.00 utterances.], tot_loss[ctc_loss=0.07205, att_loss=0.2345, loss=0.202, over 3259764.96 frames. utt_duration=1215 frames, utt_pad_proportion=0.0655, over 10746.04 utterances.], batch size: 40, lr: 4.69e-03, grad_scale: 8.0 2023-03-09 02:11:11,786 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.099e+02 1.895e+02 2.282e+02 2.881e+02 5.348e+02, threshold=4.565e+02, percent-clipped=1.0 2023-03-09 02:11:57,061 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.83 vs. limit=2.0 2023-03-09 02:12:01,905 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-03-09 02:12:31,877 INFO [train2.py:809] (1/4) Epoch 23, batch 1650, loss[ctc_loss=0.06109, att_loss=0.2107, loss=0.1807, over 15490.00 frames. utt_duration=1723 frames, utt_pad_proportion=0.008886, over 36.00 utterances.], tot_loss[ctc_loss=0.07208, att_loss=0.2348, loss=0.2023, over 3255437.11 frames. utt_duration=1207 frames, utt_pad_proportion=0.0681, over 10799.50 utterances.], batch size: 36, lr: 4.69e-03, grad_scale: 8.0 2023-03-09 02:13:46,957 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.53 vs. limit=5.0 2023-03-09 02:13:54,203 INFO [train2.py:809] (1/4) Epoch 23, batch 1700, loss[ctc_loss=0.07677, att_loss=0.241, loss=0.2081, over 17104.00 frames. utt_duration=692.5 frames, utt_pad_proportion=0.129, over 99.00 utterances.], tot_loss[ctc_loss=0.07209, att_loss=0.2352, loss=0.2026, over 3267004.80 frames. utt_duration=1200 frames, utt_pad_proportion=0.06633, over 10900.57 utterances.], batch size: 99, lr: 4.68e-03, grad_scale: 8.0 2023-03-09 02:13:55,725 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.335e+02 1.978e+02 2.274e+02 2.807e+02 5.898e+02, threshold=4.548e+02, percent-clipped=3.0 2023-03-09 02:13:57,835 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9820, 5.0515, 4.8739, 2.1606, 1.9546, 2.9150, 2.2464, 3.7969], device='cuda:1'), covar=tensor([0.0785, 0.0267, 0.0252, 0.5162, 0.5665, 0.2441, 0.3925, 0.1718], device='cuda:1'), in_proj_covar=tensor([0.0364, 0.0284, 0.0273, 0.0248, 0.0344, 0.0337, 0.0259, 0.0372], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-09 02:14:45,034 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89374.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:15:00,021 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89383.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:15:15,727 INFO [train2.py:809] (1/4) Epoch 23, batch 1750, loss[ctc_loss=0.06584, att_loss=0.2165, loss=0.1864, over 15887.00 frames. utt_duration=1631 frames, utt_pad_proportion=0.009144, over 39.00 utterances.], tot_loss[ctc_loss=0.0715, att_loss=0.2348, loss=0.2022, over 3268990.31 frames. utt_duration=1218 frames, utt_pad_proportion=0.06175, over 10747.18 utterances.], batch size: 39, lr: 4.68e-03, grad_scale: 8.0 2023-03-09 02:16:09,004 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89425.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:16:25,522 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89435.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:16:31,832 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89439.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 02:16:33,530 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.5716, 3.6903, 3.9326, 2.5529, 2.3520, 2.7586, 2.4402, 3.4441], device='cuda:1'), covar=tensor([0.0940, 0.0530, 0.0402, 0.3055, 0.4927, 0.2658, 0.3213, 0.1669], device='cuda:1'), in_proj_covar=tensor([0.0362, 0.0283, 0.0272, 0.0247, 0.0342, 0.0335, 0.0258, 0.0370], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-09 02:16:37,756 INFO [train2.py:809] (1/4) Epoch 23, batch 1800, loss[ctc_loss=0.07318, att_loss=0.2593, loss=0.2221, over 16966.00 frames. utt_duration=1359 frames, utt_pad_proportion=0.007392, over 50.00 utterances.], tot_loss[ctc_loss=0.07201, att_loss=0.2352, loss=0.2025, over 3266787.33 frames. utt_duration=1209 frames, utt_pad_proportion=0.0644, over 10819.75 utterances.], batch size: 50, lr: 4.68e-03, grad_scale: 8.0 2023-03-09 02:16:39,274 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.252e+02 1.934e+02 2.267e+02 2.788e+02 6.296e+02, threshold=4.533e+02, percent-clipped=2.0 2023-03-09 02:16:48,204 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7452, 2.5197, 2.6258, 3.3149, 2.9993, 3.2078, 2.5530, 2.3847], device='cuda:1'), covar=tensor([0.0806, 0.1677, 0.1032, 0.0807, 0.0999, 0.0530, 0.1369, 0.1561], device='cuda:1'), in_proj_covar=tensor([0.0180, 0.0216, 0.0189, 0.0221, 0.0227, 0.0182, 0.0203, 0.0188], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 02:16:49,126 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-03-09 02:17:49,760 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89486.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:18:00,958 INFO [train2.py:809] (1/4) Epoch 23, batch 1850, loss[ctc_loss=0.0733, att_loss=0.2416, loss=0.2079, over 17342.00 frames. utt_duration=1102 frames, utt_pad_proportion=0.03635, over 63.00 utterances.], tot_loss[ctc_loss=0.0722, att_loss=0.2358, loss=0.2031, over 3267606.38 frames. utt_duration=1181 frames, utt_pad_proportion=0.0717, over 11081.63 utterances.], batch size: 63, lr: 4.68e-03, grad_scale: 8.0 2023-03-09 02:18:12,515 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89500.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 02:18:27,108 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89509.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:18:36,650 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0910, 5.0606, 4.8900, 2.9742, 4.8820, 4.6976, 4.0567, 2.6714], device='cuda:1'), covar=tensor([0.0129, 0.0101, 0.0229, 0.1028, 0.0105, 0.0202, 0.0378, 0.1317], device='cuda:1'), in_proj_covar=tensor([0.0075, 0.0102, 0.0104, 0.0111, 0.0085, 0.0114, 0.0100, 0.0103], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-09 02:18:53,261 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89525.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:19:23,208 INFO [train2.py:809] (1/4) Epoch 23, batch 1900, loss[ctc_loss=0.08987, att_loss=0.2509, loss=0.2187, over 17056.00 frames. utt_duration=1289 frames, utt_pad_proportion=0.009383, over 53.00 utterances.], tot_loss[ctc_loss=0.07234, att_loss=0.2361, loss=0.2033, over 3272855.33 frames. utt_duration=1165 frames, utt_pad_proportion=0.07459, over 11249.05 utterances.], batch size: 53, lr: 4.68e-03, grad_scale: 8.0 2023-03-09 02:19:24,723 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.258e+02 1.857e+02 2.214e+02 2.613e+02 6.463e+02, threshold=4.427e+02, percent-clipped=1.0 2023-03-09 02:19:34,946 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5669, 2.5348, 4.9577, 3.9428, 3.0568, 4.2900, 4.8115, 4.6583], device='cuda:1'), covar=tensor([0.0262, 0.1591, 0.0198, 0.0869, 0.1598, 0.0247, 0.0148, 0.0271], device='cuda:1'), in_proj_covar=tensor([0.0204, 0.0243, 0.0195, 0.0317, 0.0264, 0.0218, 0.0184, 0.0214], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 02:20:07,569 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89570.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:20:44,884 INFO [train2.py:809] (1/4) Epoch 23, batch 1950, loss[ctc_loss=0.09899, att_loss=0.2566, loss=0.2251, over 13465.00 frames. utt_duration=373.1 frames, utt_pad_proportion=0.3522, over 145.00 utterances.], tot_loss[ctc_loss=0.0734, att_loss=0.2364, loss=0.2038, over 3265383.75 frames. utt_duration=1154 frames, utt_pad_proportion=0.08089, over 11337.24 utterances.], batch size: 145, lr: 4.68e-03, grad_scale: 8.0 2023-03-09 02:20:46,787 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6803, 5.0681, 4.8824, 5.0067, 5.0862, 4.7418, 3.4906, 5.0614], device='cuda:1'), covar=tensor([0.0130, 0.0120, 0.0131, 0.0081, 0.0105, 0.0112, 0.0688, 0.0192], device='cuda:1'), in_proj_covar=tensor([0.0092, 0.0088, 0.0111, 0.0070, 0.0076, 0.0087, 0.0103, 0.0109], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 02:22:06,987 INFO [train2.py:809] (1/4) Epoch 23, batch 2000, loss[ctc_loss=0.08749, att_loss=0.2473, loss=0.2154, over 16762.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.006703, over 48.00 utterances.], tot_loss[ctc_loss=0.07276, att_loss=0.2361, loss=0.2034, over 3274258.63 frames. utt_duration=1192 frames, utt_pad_proportion=0.06884, over 10999.28 utterances.], batch size: 48, lr: 4.68e-03, grad_scale: 8.0 2023-03-09 02:22:08,496 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.359e+02 1.989e+02 2.307e+02 2.806e+02 7.990e+02, threshold=4.613e+02, percent-clipped=4.0 2023-03-09 02:23:07,562 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89680.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:23:12,435 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=89683.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:23:28,552 INFO [train2.py:809] (1/4) Epoch 23, batch 2050, loss[ctc_loss=0.08289, att_loss=0.2517, loss=0.2179, over 16765.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.006599, over 48.00 utterances.], tot_loss[ctc_loss=0.0723, att_loss=0.2352, loss=0.2026, over 3265102.37 frames. utt_duration=1208 frames, utt_pad_proportion=0.0681, over 10822.52 utterances.], batch size: 48, lr: 4.68e-03, grad_scale: 8.0 2023-03-09 02:24:00,720 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.8478, 6.1120, 5.5832, 5.8368, 5.7666, 5.3239, 5.5355, 5.3189], device='cuda:1'), covar=tensor([0.1324, 0.0875, 0.0984, 0.0801, 0.0962, 0.1551, 0.2299, 0.2270], device='cuda:1'), in_proj_covar=tensor([0.0539, 0.0622, 0.0473, 0.0468, 0.0441, 0.0476, 0.0625, 0.0539], 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') 2023-03-09 02:24:28,906 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89730.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:24:30,402 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=89731.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:24:47,349 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89741.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:24:50,038 INFO [train2.py:809] (1/4) Epoch 23, batch 2100, loss[ctc_loss=0.0696, att_loss=0.2309, loss=0.1987, over 16130.00 frames. utt_duration=1538 frames, utt_pad_proportion=0.004777, over 42.00 utterances.], tot_loss[ctc_loss=0.07217, att_loss=0.235, loss=0.2025, over 3274068.01 frames. utt_duration=1234 frames, utt_pad_proportion=0.05914, over 10624.06 utterances.], batch size: 42, lr: 4.67e-03, grad_scale: 16.0 2023-03-09 02:24:51,496 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.393e+02 1.974e+02 2.233e+02 2.769e+02 5.380e+02, threshold=4.466e+02, percent-clipped=5.0 2023-03-09 02:25:33,329 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-03-09 02:25:52,097 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89781.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:26:11,804 INFO [train2.py:809] (1/4) Epoch 23, batch 2150, loss[ctc_loss=0.08194, att_loss=0.2411, loss=0.2093, over 17006.00 frames. utt_duration=1336 frames, utt_pad_proportion=0.009237, over 51.00 utterances.], tot_loss[ctc_loss=0.0728, att_loss=0.2355, loss=0.203, over 3281362.30 frames. utt_duration=1251 frames, utt_pad_proportion=0.0538, over 10507.56 utterances.], batch size: 51, lr: 4.67e-03, grad_scale: 16.0 2023-03-09 02:26:15,131 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89795.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 02:27:04,695 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=89825.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:27:33,612 INFO [train2.py:809] (1/4) Epoch 23, batch 2200, loss[ctc_loss=0.09152, att_loss=0.255, loss=0.2223, over 17059.00 frames. utt_duration=1220 frames, utt_pad_proportion=0.01852, over 56.00 utterances.], tot_loss[ctc_loss=0.07319, att_loss=0.2357, loss=0.2032, over 3280763.71 frames. utt_duration=1241 frames, utt_pad_proportion=0.05484, over 10583.31 utterances.], batch size: 56, lr: 4.67e-03, grad_scale: 16.0 2023-03-09 02:27:34,943 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.310e+02 1.973e+02 2.386e+02 3.013e+02 6.689e+02, threshold=4.771e+02, percent-clipped=5.0 2023-03-09 02:28:09,092 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89865.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:28:09,257 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89865.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:28:09,978 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-03-09 02:28:23,119 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=89873.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:28:54,740 INFO [train2.py:809] (1/4) Epoch 23, batch 2250, loss[ctc_loss=0.05048, att_loss=0.2215, loss=0.1873, over 15942.00 frames. utt_duration=1557 frames, utt_pad_proportion=0.007218, over 41.00 utterances.], tot_loss[ctc_loss=0.07184, att_loss=0.2342, loss=0.2018, over 3270217.84 frames. utt_duration=1248 frames, utt_pad_proportion=0.05683, over 10493.21 utterances.], batch size: 41, lr: 4.67e-03, grad_scale: 16.0 2023-03-09 02:29:48,462 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89926.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 02:30:14,729 INFO [train2.py:809] (1/4) Epoch 23, batch 2300, loss[ctc_loss=0.0866, att_loss=0.2446, loss=0.213, over 17395.00 frames. utt_duration=1106 frames, utt_pad_proportion=0.03222, over 63.00 utterances.], tot_loss[ctc_loss=0.07205, att_loss=0.2346, loss=0.2021, over 3276146.54 frames. utt_duration=1254 frames, utt_pad_proportion=0.05211, over 10459.54 utterances.], batch size: 63, lr: 4.67e-03, grad_scale: 16.0 2023-03-09 02:30:16,395 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.289e+02 1.850e+02 2.227e+02 2.792e+02 5.012e+02, threshold=4.455e+02, percent-clipped=2.0 2023-03-09 02:31:36,741 INFO [train2.py:809] (1/4) Epoch 23, batch 2350, loss[ctc_loss=0.05515, att_loss=0.2069, loss=0.1765, over 15875.00 frames. utt_duration=1630 frames, utt_pad_proportion=0.009736, over 39.00 utterances.], tot_loss[ctc_loss=0.07096, att_loss=0.2341, loss=0.2015, over 3277324.19 frames. utt_duration=1278 frames, utt_pad_proportion=0.0475, over 10270.17 utterances.], batch size: 39, lr: 4.67e-03, grad_scale: 16.0 2023-03-09 02:32:29,973 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1047, 4.2751, 4.5062, 4.3683, 2.7996, 4.2251, 2.5805, 2.1674], device='cuda:1'), covar=tensor([0.0462, 0.0312, 0.0609, 0.0271, 0.1604, 0.0287, 0.1538, 0.1533], device='cuda:1'), in_proj_covar=tensor([0.0199, 0.0171, 0.0264, 0.0162, 0.0224, 0.0155, 0.0233, 0.0204], 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') 2023-03-09 02:32:42,515 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90030.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:32:52,445 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90036.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:33:03,245 INFO [train2.py:809] (1/4) Epoch 23, batch 2400, loss[ctc_loss=0.1228, att_loss=0.2672, loss=0.2383, over 14215.00 frames. utt_duration=393.8 frames, utt_pad_proportion=0.3163, over 145.00 utterances.], tot_loss[ctc_loss=0.07049, att_loss=0.2337, loss=0.2011, over 3264836.80 frames. utt_duration=1257 frames, utt_pad_proportion=0.05646, over 10399.77 utterances.], batch size: 145, lr: 4.67e-03, grad_scale: 16.0 2023-03-09 02:33:04,734 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.274e+02 1.874e+02 2.405e+02 2.931e+02 6.452e+02, threshold=4.810e+02, percent-clipped=5.0 2023-03-09 02:33:27,864 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2355, 3.8274, 3.2675, 3.3872, 4.0342, 3.6627, 2.9945, 4.3922], device='cuda:1'), covar=tensor([0.0908, 0.0529, 0.1121, 0.0781, 0.0656, 0.0748, 0.0906, 0.0385], device='cuda:1'), in_proj_covar=tensor([0.0205, 0.0220, 0.0229, 0.0205, 0.0282, 0.0245, 0.0201, 0.0293], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-09 02:33:59,626 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.19 vs. limit=5.0 2023-03-09 02:34:00,241 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=90078.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:34:05,185 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90081.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:34:24,270 INFO [train2.py:809] (1/4) Epoch 23, batch 2450, loss[ctc_loss=0.08014, att_loss=0.2556, loss=0.2205, over 17326.00 frames. utt_duration=1262 frames, utt_pad_proportion=0.01052, over 55.00 utterances.], tot_loss[ctc_loss=0.07143, att_loss=0.2342, loss=0.2016, over 3255370.58 frames. utt_duration=1244 frames, utt_pad_proportion=0.06232, over 10482.19 utterances.], batch size: 55, lr: 4.67e-03, grad_scale: 16.0 2023-03-09 02:34:27,931 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90095.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 02:34:50,640 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.2768, 5.5322, 5.1313, 5.6296, 5.0817, 5.2450, 5.6704, 5.3946], device='cuda:1'), covar=tensor([0.0443, 0.0256, 0.0712, 0.0272, 0.0355, 0.0191, 0.0196, 0.0193], device='cuda:1'), in_proj_covar=tensor([0.0387, 0.0324, 0.0367, 0.0353, 0.0323, 0.0238, 0.0307, 0.0290], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-09 02:35:06,963 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-03-09 02:35:23,981 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=90129.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:35:25,877 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90130.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:35:47,560 INFO [train2.py:809] (1/4) Epoch 23, batch 2500, loss[ctc_loss=0.04259, att_loss=0.2003, loss=0.1688, over 15643.00 frames. utt_duration=1693 frames, utt_pad_proportion=0.008376, over 37.00 utterances.], tot_loss[ctc_loss=0.07075, att_loss=0.234, loss=0.2014, over 3261584.69 frames. utt_duration=1240 frames, utt_pad_proportion=0.06048, over 10537.73 utterances.], batch size: 37, lr: 4.66e-03, grad_scale: 16.0 2023-03-09 02:35:47,648 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=90143.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 02:35:48,948 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.112e+02 2.045e+02 2.395e+02 2.937e+02 6.523e+02, threshold=4.790e+02, percent-clipped=3.0 2023-03-09 02:36:23,520 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90165.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:37:05,997 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90191.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:37:07,225 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-03-09 02:37:09,385 INFO [train2.py:809] (1/4) Epoch 23, batch 2550, loss[ctc_loss=0.06372, att_loss=0.2101, loss=0.1808, over 15658.00 frames. utt_duration=1694 frames, utt_pad_proportion=0.008007, over 37.00 utterances.], tot_loss[ctc_loss=0.07071, att_loss=0.2339, loss=0.2012, over 3263517.45 frames. utt_duration=1254 frames, utt_pad_proportion=0.05647, over 10425.99 utterances.], batch size: 37, lr: 4.66e-03, grad_scale: 16.0 2023-03-09 02:37:41,492 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=90213.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:37:54,835 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90221.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 02:38:12,241 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90232.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:38:24,856 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-03-09 02:38:30,843 INFO [train2.py:809] (1/4) Epoch 23, batch 2600, loss[ctc_loss=0.05719, att_loss=0.2109, loss=0.1801, over 15383.00 frames. utt_duration=1760 frames, utt_pad_proportion=0.009663, over 35.00 utterances.], tot_loss[ctc_loss=0.07124, att_loss=0.2341, loss=0.2015, over 3266336.09 frames. utt_duration=1251 frames, utt_pad_proportion=0.05655, over 10455.08 utterances.], batch size: 35, lr: 4.66e-03, grad_scale: 8.0 2023-03-09 02:38:34,182 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.085e+02 2.020e+02 2.371e+02 2.963e+02 6.277e+02, threshold=4.742e+02, percent-clipped=3.0 2023-03-09 02:39:52,427 INFO [train2.py:809] (1/4) Epoch 23, batch 2650, loss[ctc_loss=0.05108, att_loss=0.2044, loss=0.1737, over 15525.00 frames. utt_duration=1727 frames, utt_pad_proportion=0.007044, over 36.00 utterances.], tot_loss[ctc_loss=0.07075, att_loss=0.234, loss=0.2013, over 3270958.59 frames. utt_duration=1260 frames, utt_pad_proportion=0.05146, over 10395.65 utterances.], batch size: 36, lr: 4.66e-03, grad_scale: 8.0 2023-03-09 02:39:52,879 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90293.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:39:59,171 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1350, 5.0013, 4.8653, 2.9687, 4.8099, 4.6865, 4.3554, 2.8106], device='cuda:1'), covar=tensor([0.0093, 0.0096, 0.0260, 0.0982, 0.0106, 0.0185, 0.0297, 0.1204], device='cuda:1'), in_proj_covar=tensor([0.0075, 0.0102, 0.0104, 0.0110, 0.0085, 0.0113, 0.0099, 0.0102], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-09 02:40:24,886 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.0940, 2.6642, 2.8170, 4.3104, 3.9012, 3.8431, 2.7569, 2.0256], device='cuda:1'), covar=tensor([0.0920, 0.2184, 0.1240, 0.0642, 0.0724, 0.0456, 0.1682, 0.2438], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0219, 0.0191, 0.0224, 0.0228, 0.0182, 0.0205, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 02:40:37,346 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6974, 5.0241, 4.8554, 4.9783, 5.1478, 4.8112, 3.3400, 5.0495], device='cuda:1'), covar=tensor([0.0098, 0.0099, 0.0121, 0.0078, 0.0079, 0.0099, 0.0684, 0.0139], device='cuda:1'), in_proj_covar=tensor([0.0091, 0.0088, 0.0111, 0.0070, 0.0076, 0.0086, 0.0102, 0.0108], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 02:41:03,321 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90336.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:41:14,423 INFO [train2.py:809] (1/4) Epoch 23, batch 2700, loss[ctc_loss=0.06192, att_loss=0.2313, loss=0.1974, over 16322.00 frames. utt_duration=1452 frames, utt_pad_proportion=0.006551, over 45.00 utterances.], tot_loss[ctc_loss=0.07058, att_loss=0.2341, loss=0.2014, over 3277312.99 frames. utt_duration=1266 frames, utt_pad_proportion=0.04823, over 10369.37 utterances.], batch size: 45, lr: 4.66e-03, grad_scale: 8.0 2023-03-09 02:41:17,514 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.381e+02 1.963e+02 2.237e+02 2.916e+02 6.361e+02, threshold=4.475e+02, percent-clipped=3.0 2023-03-09 02:41:27,018 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90351.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:42:21,371 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=90384.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:42:35,956 INFO [train2.py:809] (1/4) Epoch 23, batch 2750, loss[ctc_loss=0.0787, att_loss=0.2186, loss=0.1906, over 15653.00 frames. utt_duration=1694 frames, utt_pad_proportion=0.008292, over 37.00 utterances.], tot_loss[ctc_loss=0.07103, att_loss=0.234, loss=0.2014, over 3274954.83 frames. utt_duration=1266 frames, utt_pad_proportion=0.04773, over 10357.72 utterances.], batch size: 37, lr: 4.66e-03, grad_scale: 8.0 2023-03-09 02:43:06,565 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90412.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:43:11,353 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-03-09 02:43:58,171 INFO [train2.py:809] (1/4) Epoch 23, batch 2800, loss[ctc_loss=0.05734, att_loss=0.2184, loss=0.1862, over 16129.00 frames. utt_duration=1538 frames, utt_pad_proportion=0.006079, over 42.00 utterances.], tot_loss[ctc_loss=0.07109, att_loss=0.234, loss=0.2015, over 3274350.56 frames. utt_duration=1270 frames, utt_pad_proportion=0.04816, over 10322.33 utterances.], batch size: 42, lr: 4.66e-03, grad_scale: 8.0 2023-03-09 02:44:01,316 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.235e+02 1.964e+02 2.366e+02 2.896e+02 7.050e+02, threshold=4.731e+02, percent-clipped=3.0 2023-03-09 02:44:23,938 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90459.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:44:48,548 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0487, 5.3121, 5.2742, 5.2203, 5.3249, 5.2658, 4.9143, 4.7416], device='cuda:1'), covar=tensor([0.0966, 0.0518, 0.0284, 0.0450, 0.0292, 0.0323, 0.0451, 0.0346], device='cuda:1'), in_proj_covar=tensor([0.0529, 0.0368, 0.0357, 0.0364, 0.0433, 0.0439, 0.0365, 0.0402], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:1') 2023-03-09 02:45:06,180 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8266, 5.1197, 4.7185, 5.2281, 4.6149, 4.8494, 5.2437, 5.0260], device='cuda:1'), covar=tensor([0.0612, 0.0306, 0.0854, 0.0303, 0.0408, 0.0266, 0.0252, 0.0234], device='cuda:1'), in_proj_covar=tensor([0.0394, 0.0329, 0.0370, 0.0358, 0.0326, 0.0241, 0.0312, 0.0294], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-09 02:45:07,715 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90486.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:45:19,165 INFO [train2.py:809] (1/4) Epoch 23, batch 2850, loss[ctc_loss=0.06254, att_loss=0.2193, loss=0.188, over 16005.00 frames. utt_duration=1602 frames, utt_pad_proportion=0.006343, over 40.00 utterances.], tot_loss[ctc_loss=0.07015, att_loss=0.2329, loss=0.2004, over 3271450.03 frames. utt_duration=1283 frames, utt_pad_proportion=0.04605, over 10209.54 utterances.], batch size: 40, lr: 4.66e-03, grad_scale: 8.0 2023-03-09 02:45:24,395 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90496.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:46:03,407 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90520.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:46:04,888 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90521.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 02:46:40,204 INFO [train2.py:809] (1/4) Epoch 23, batch 2900, loss[ctc_loss=0.05475, att_loss=0.2226, loss=0.189, over 16412.00 frames. utt_duration=1494 frames, utt_pad_proportion=0.006876, over 44.00 utterances.], tot_loss[ctc_loss=0.06999, att_loss=0.2337, loss=0.2009, over 3279994.35 frames. utt_duration=1297 frames, utt_pad_proportion=0.04035, over 10123.65 utterances.], batch size: 44, lr: 4.65e-03, grad_scale: 8.0 2023-03-09 02:46:43,384 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.137e+02 2.000e+02 2.282e+02 2.820e+02 8.053e+02, threshold=4.565e+02, percent-clipped=3.0 2023-03-09 02:46:54,619 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2359, 3.8955, 3.4329, 3.6257, 4.1151, 3.8116, 3.2408, 4.4886], device='cuda:1'), covar=tensor([0.0975, 0.0610, 0.1069, 0.0669, 0.0771, 0.0687, 0.0759, 0.0458], device='cuda:1'), in_proj_covar=tensor([0.0207, 0.0220, 0.0230, 0.0205, 0.0284, 0.0246, 0.0201, 0.0293], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-09 02:47:02,766 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90557.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:47:22,355 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=90569.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:47:39,761 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0332, 5.3272, 5.5397, 5.4762, 5.5259, 6.0042, 5.3268, 6.0741], device='cuda:1'), covar=tensor([0.0737, 0.0785, 0.0817, 0.1112, 0.1748, 0.0910, 0.0711, 0.0687], device='cuda:1'), in_proj_covar=tensor([0.0878, 0.0510, 0.0616, 0.0666, 0.0881, 0.0636, 0.0495, 0.0618], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 02:47:52,968 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90588.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:48:01,366 INFO [train2.py:809] (1/4) Epoch 23, batch 2950, loss[ctc_loss=0.09116, att_loss=0.2499, loss=0.2182, over 16946.00 frames. utt_duration=686.3 frames, utt_pad_proportion=0.1389, over 99.00 utterances.], tot_loss[ctc_loss=0.07007, att_loss=0.2339, loss=0.2011, over 3279797.89 frames. utt_duration=1287 frames, utt_pad_proportion=0.04371, over 10201.56 utterances.], batch size: 99, lr: 4.65e-03, grad_scale: 8.0 2023-03-09 02:48:13,385 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9422, 5.1752, 5.1508, 5.1174, 5.2330, 5.1394, 4.8387, 4.6603], device='cuda:1'), covar=tensor([0.1087, 0.0660, 0.0332, 0.0547, 0.0342, 0.0388, 0.0457, 0.0384], device='cuda:1'), in_proj_covar=tensor([0.0529, 0.0368, 0.0357, 0.0365, 0.0432, 0.0438, 0.0365, 0.0403], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:1') 2023-03-09 02:48:46,320 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8020, 5.1165, 4.7253, 5.1696, 4.6239, 4.8407, 5.2484, 5.0404], device='cuda:1'), covar=tensor([0.0642, 0.0328, 0.0824, 0.0383, 0.0433, 0.0303, 0.0252, 0.0226], device='cuda:1'), in_proj_covar=tensor([0.0396, 0.0331, 0.0371, 0.0359, 0.0327, 0.0241, 0.0313, 0.0293], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-09 02:48:46,452 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.1161, 3.7138, 3.1392, 3.4218, 3.8854, 3.6280, 3.0912, 4.1647], device='cuda:1'), covar=tensor([0.0903, 0.0499, 0.1146, 0.0680, 0.0699, 0.0690, 0.0767, 0.0488], device='cuda:1'), in_proj_covar=tensor([0.0206, 0.0220, 0.0229, 0.0204, 0.0283, 0.0245, 0.0201, 0.0292], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-09 02:49:23,756 INFO [train2.py:809] (1/4) Epoch 23, batch 3000, loss[ctc_loss=0.05845, att_loss=0.2204, loss=0.188, over 16119.00 frames. utt_duration=1537 frames, utt_pad_proportion=0.006587, over 42.00 utterances.], tot_loss[ctc_loss=0.0706, att_loss=0.234, loss=0.2013, over 3273012.94 frames. utt_duration=1271 frames, utt_pad_proportion=0.05026, over 10310.35 utterances.], batch size: 42, lr: 4.65e-03, grad_scale: 8.0 2023-03-09 02:49:23,757 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-09 02:49:38,036 INFO [train2.py:843] (1/4) Epoch 23, validation: ctc_loss=0.03973, att_loss=0.234, loss=0.1952, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-09 02:49:38,037 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-09 02:49:41,304 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.311e+02 1.804e+02 2.183e+02 2.911e+02 8.359e+02, threshold=4.365e+02, percent-clipped=3.0 2023-03-09 02:51:00,739 INFO [train2.py:809] (1/4) Epoch 23, batch 3050, loss[ctc_loss=0.07139, att_loss=0.2418, loss=0.2077, over 16948.00 frames. utt_duration=1357 frames, utt_pad_proportion=0.008736, over 50.00 utterances.], tot_loss[ctc_loss=0.07054, att_loss=0.2344, loss=0.2017, over 3271830.04 frames. utt_duration=1247 frames, utt_pad_proportion=0.05629, over 10510.16 utterances.], batch size: 50, lr: 4.65e-03, grad_scale: 8.0 2023-03-09 02:51:24,175 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90707.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:52:23,332 INFO [train2.py:809] (1/4) Epoch 23, batch 3100, loss[ctc_loss=0.07008, att_loss=0.2175, loss=0.188, over 16283.00 frames. utt_duration=1516 frames, utt_pad_proportion=0.006279, over 43.00 utterances.], tot_loss[ctc_loss=0.07067, att_loss=0.235, loss=0.2021, over 3272488.29 frames. utt_duration=1238 frames, utt_pad_proportion=0.05884, over 10586.15 utterances.], batch size: 43, lr: 4.65e-03, grad_scale: 4.0 2023-03-09 02:52:28,635 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.220e+02 1.776e+02 2.095e+02 2.519e+02 6.521e+02, threshold=4.191e+02, percent-clipped=1.0 2023-03-09 02:53:08,369 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-03-09 02:53:34,417 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90786.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:53:36,785 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-03-09 02:53:45,909 INFO [train2.py:809] (1/4) Epoch 23, batch 3150, loss[ctc_loss=0.06673, att_loss=0.2309, loss=0.1981, over 15949.00 frames. utt_duration=1557 frames, utt_pad_proportion=0.00686, over 41.00 utterances.], tot_loss[ctc_loss=0.07113, att_loss=0.235, loss=0.2022, over 3273328.84 frames. utt_duration=1250 frames, utt_pad_proportion=0.05597, over 10488.24 utterances.], batch size: 41, lr: 4.65e-03, grad_scale: 4.0 2023-03-09 02:54:21,921 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90815.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:54:33,603 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.2422, 5.1549, 4.9781, 3.4718, 5.0029, 4.7542, 4.6935, 3.0842], device='cuda:1'), covar=tensor([0.0114, 0.0110, 0.0281, 0.0789, 0.0098, 0.0191, 0.0240, 0.1201], device='cuda:1'), in_proj_covar=tensor([0.0076, 0.0104, 0.0106, 0.0112, 0.0086, 0.0116, 0.0100, 0.0104], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-09 02:54:52,435 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=90834.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:55:08,172 INFO [train2.py:809] (1/4) Epoch 23, batch 3200, loss[ctc_loss=0.05885, att_loss=0.2203, loss=0.188, over 16405.00 frames. utt_duration=1493 frames, utt_pad_proportion=0.007253, over 44.00 utterances.], tot_loss[ctc_loss=0.07076, att_loss=0.2348, loss=0.202, over 3271584.05 frames. utt_duration=1231 frames, utt_pad_proportion=0.06068, over 10640.65 utterances.], batch size: 44, lr: 4.65e-03, grad_scale: 8.0 2023-03-09 02:55:12,916 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.356e+02 1.815e+02 2.300e+02 2.792e+02 4.062e+02, threshold=4.600e+02, percent-clipped=0.0 2023-03-09 02:55:18,224 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90849.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:55:22,755 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90852.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:55:49,263 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-03-09 02:56:14,748 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2372, 2.8672, 3.5414, 2.9131, 3.3951, 4.3979, 4.2445, 2.9909], device='cuda:1'), covar=tensor([0.0451, 0.1779, 0.1288, 0.1410, 0.1131, 0.0890, 0.0639, 0.1460], device='cuda:1'), in_proj_covar=tensor([0.0248, 0.0245, 0.0283, 0.0222, 0.0269, 0.0373, 0.0264, 0.0235], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-09 02:56:21,718 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90888.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:56:29,988 INFO [train2.py:809] (1/4) Epoch 23, batch 3250, loss[ctc_loss=0.1044, att_loss=0.2637, loss=0.2319, over 13936.00 frames. utt_duration=386.1 frames, utt_pad_proportion=0.3286, over 145.00 utterances.], tot_loss[ctc_loss=0.07018, att_loss=0.2338, loss=0.2011, over 3266462.53 frames. utt_duration=1235 frames, utt_pad_proportion=0.06132, over 10593.42 utterances.], batch size: 145, lr: 4.64e-03, grad_scale: 8.0 2023-03-09 02:56:57,362 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90910.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:57:39,806 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=90936.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:57:51,514 INFO [train2.py:809] (1/4) Epoch 23, batch 3300, loss[ctc_loss=0.05521, att_loss=0.2039, loss=0.1741, over 15661.00 frames. utt_duration=1695 frames, utt_pad_proportion=0.007754, over 37.00 utterances.], tot_loss[ctc_loss=0.06977, att_loss=0.2334, loss=0.2007, over 3268073.50 frames. utt_duration=1252 frames, utt_pad_proportion=0.05652, over 10456.20 utterances.], batch size: 37, lr: 4.64e-03, grad_scale: 8.0 2023-03-09 02:57:56,189 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.358e+02 1.978e+02 2.311e+02 2.914e+02 8.897e+02, threshold=4.622e+02, percent-clipped=4.0 2023-03-09 02:59:13,554 INFO [train2.py:809] (1/4) Epoch 23, batch 3350, loss[ctc_loss=0.1044, att_loss=0.2597, loss=0.2287, over 17048.00 frames. utt_duration=690.3 frames, utt_pad_proportion=0.1339, over 99.00 utterances.], tot_loss[ctc_loss=0.06965, att_loss=0.2337, loss=0.2009, over 3273845.02 frames. utt_duration=1237 frames, utt_pad_proportion=0.05806, over 10600.43 utterances.], batch size: 99, lr: 4.64e-03, grad_scale: 8.0 2023-03-09 02:59:37,331 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91007.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 02:59:37,526 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0405, 5.0314, 4.7773, 2.1763, 2.0488, 2.8615, 2.3477, 3.8293], device='cuda:1'), covar=tensor([0.0736, 0.0260, 0.0272, 0.5138, 0.5581, 0.2494, 0.3768, 0.1636], device='cuda:1'), in_proj_covar=tensor([0.0358, 0.0282, 0.0271, 0.0246, 0.0343, 0.0334, 0.0256, 0.0370], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-03-09 03:00:23,316 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0831, 4.3451, 4.4575, 4.6530, 2.8174, 4.4501, 2.4745, 1.7056], device='cuda:1'), covar=tensor([0.0381, 0.0255, 0.0607, 0.0339, 0.1580, 0.0206, 0.1620, 0.1725], device='cuda:1'), in_proj_covar=tensor([0.0199, 0.0170, 0.0261, 0.0162, 0.0223, 0.0154, 0.0231, 0.0203], 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') 2023-03-09 03:00:35,856 INFO [train2.py:809] (1/4) Epoch 23, batch 3400, loss[ctc_loss=0.07072, att_loss=0.2517, loss=0.2155, over 17325.00 frames. utt_duration=1176 frames, utt_pad_proportion=0.02229, over 59.00 utterances.], tot_loss[ctc_loss=0.07002, att_loss=0.2347, loss=0.2018, over 3281668.73 frames. utt_duration=1198 frames, utt_pad_proportion=0.06477, over 10966.88 utterances.], batch size: 59, lr: 4.64e-03, grad_scale: 8.0 2023-03-09 03:00:40,394 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.261e+02 1.937e+02 2.293e+02 2.658e+02 6.821e+02, threshold=4.585e+02, percent-clipped=3.0 2023-03-09 03:00:55,176 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=91055.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:01:16,106 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-03-09 03:01:57,222 INFO [train2.py:809] (1/4) Epoch 23, batch 3450, loss[ctc_loss=0.07442, att_loss=0.2423, loss=0.2087, over 16777.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.006052, over 48.00 utterances.], tot_loss[ctc_loss=0.07067, att_loss=0.2352, loss=0.2023, over 3274320.20 frames. utt_duration=1182 frames, utt_pad_proportion=0.07089, over 11096.31 utterances.], batch size: 48, lr: 4.64e-03, grad_scale: 8.0 2023-03-09 03:02:15,025 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6423, 5.1009, 4.8306, 5.0290, 5.1237, 4.7301, 3.5132, 5.0201], device='cuda:1'), covar=tensor([0.0123, 0.0105, 0.0170, 0.0098, 0.0113, 0.0124, 0.0718, 0.0185], device='cuda:1'), in_proj_covar=tensor([0.0094, 0.0090, 0.0114, 0.0071, 0.0077, 0.0088, 0.0104, 0.0110], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 03:02:15,116 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.3205, 3.7999, 3.2358, 3.4316, 3.9988, 3.6566, 2.9240, 4.2886], device='cuda:1'), covar=tensor([0.0867, 0.0508, 0.1083, 0.0734, 0.0684, 0.0723, 0.0929, 0.0427], device='cuda:1'), in_proj_covar=tensor([0.0204, 0.0218, 0.0225, 0.0202, 0.0278, 0.0241, 0.0198, 0.0289], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-09 03:02:17,443 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.21 vs. limit=5.0 2023-03-09 03:02:33,668 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91115.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:03:19,284 INFO [train2.py:809] (1/4) Epoch 23, batch 3500, loss[ctc_loss=0.09164, att_loss=0.2508, loss=0.2189, over 14198.00 frames. utt_duration=390.6 frames, utt_pad_proportion=0.3112, over 146.00 utterances.], tot_loss[ctc_loss=0.07021, att_loss=0.2339, loss=0.2012, over 3260533.90 frames. utt_duration=1205 frames, utt_pad_proportion=0.06862, over 10835.24 utterances.], batch size: 146, lr: 4.64e-03, grad_scale: 8.0 2023-03-09 03:03:23,955 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.203e+02 1.873e+02 2.204e+02 2.861e+02 4.686e+02, threshold=4.409e+02, percent-clipped=2.0 2023-03-09 03:03:33,659 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91152.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:03:51,250 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=91163.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:04:40,871 INFO [train2.py:809] (1/4) Epoch 23, batch 3550, loss[ctc_loss=0.06052, att_loss=0.228, loss=0.1945, over 15962.00 frames. utt_duration=1559 frames, utt_pad_proportion=0.005059, over 41.00 utterances.], tot_loss[ctc_loss=0.07018, att_loss=0.2336, loss=0.2009, over 3259238.71 frames. utt_duration=1221 frames, utt_pad_proportion=0.06427, over 10694.14 utterances.], batch size: 41, lr: 4.64e-03, grad_scale: 8.0 2023-03-09 03:04:52,824 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=91200.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:05:01,574 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=91205.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:06:04,017 INFO [train2.py:809] (1/4) Epoch 23, batch 3600, loss[ctc_loss=0.06282, att_loss=0.2087, loss=0.1795, over 15626.00 frames. utt_duration=1691 frames, utt_pad_proportion=0.009953, over 37.00 utterances.], tot_loss[ctc_loss=0.07004, att_loss=0.2335, loss=0.2008, over 3259167.72 frames. utt_duration=1223 frames, utt_pad_proportion=0.06537, over 10669.22 utterances.], batch size: 37, lr: 4.64e-03, grad_scale: 8.0 2023-03-09 03:06:08,713 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.503e+02 1.921e+02 2.289e+02 2.747e+02 1.018e+03, threshold=4.578e+02, percent-clipped=6.0 2023-03-09 03:07:14,033 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91285.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:07:26,451 INFO [train2.py:809] (1/4) Epoch 23, batch 3650, loss[ctc_loss=0.06839, att_loss=0.2359, loss=0.2024, over 17125.00 frames. utt_duration=1225 frames, utt_pad_proportion=0.01467, over 56.00 utterances.], tot_loss[ctc_loss=0.07105, att_loss=0.2344, loss=0.2018, over 3262139.60 frames. utt_duration=1202 frames, utt_pad_proportion=0.06972, over 10865.17 utterances.], batch size: 56, lr: 4.63e-03, grad_scale: 8.0 2023-03-09 03:08:47,986 INFO [train2.py:809] (1/4) Epoch 23, batch 3700, loss[ctc_loss=0.0676, att_loss=0.2315, loss=0.1987, over 16692.00 frames. utt_duration=1453 frames, utt_pad_proportion=0.006305, over 46.00 utterances.], tot_loss[ctc_loss=0.07084, att_loss=0.2344, loss=0.2017, over 3265928.70 frames. utt_duration=1222 frames, utt_pad_proportion=0.06229, over 10701.00 utterances.], batch size: 46, lr: 4.63e-03, grad_scale: 4.0 2023-03-09 03:08:53,212 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=91346.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:08:54,348 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.187e+02 1.976e+02 2.254e+02 2.803e+02 7.124e+02, threshold=4.508e+02, percent-clipped=5.0 2023-03-09 03:08:56,499 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4588, 2.7565, 4.8831, 3.8036, 2.9362, 4.1652, 4.6156, 4.5727], device='cuda:1'), covar=tensor([0.0257, 0.1418, 0.0174, 0.0889, 0.1681, 0.0264, 0.0167, 0.0249], device='cuda:1'), in_proj_covar=tensor([0.0208, 0.0244, 0.0197, 0.0319, 0.0265, 0.0222, 0.0188, 0.0217], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 03:09:55,600 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6444, 5.1001, 4.8715, 4.9743, 5.1314, 4.7202, 3.6921, 4.9623], device='cuda:1'), covar=tensor([0.0108, 0.0096, 0.0127, 0.0074, 0.0087, 0.0105, 0.0621, 0.0182], device='cuda:1'), in_proj_covar=tensor([0.0094, 0.0090, 0.0114, 0.0071, 0.0077, 0.0088, 0.0104, 0.0110], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 03:10:10,206 INFO [train2.py:809] (1/4) Epoch 23, batch 3750, loss[ctc_loss=0.0685, att_loss=0.2257, loss=0.1943, over 15970.00 frames. utt_duration=1559 frames, utt_pad_proportion=0.006265, over 41.00 utterances.], tot_loss[ctc_loss=0.07142, att_loss=0.2349, loss=0.2022, over 3269100.13 frames. utt_duration=1247 frames, utt_pad_proportion=0.05462, over 10501.96 utterances.], batch size: 41, lr: 4.63e-03, grad_scale: 4.0 2023-03-09 03:10:17,628 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-03-09 03:10:20,380 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6968, 4.8651, 4.8725, 4.5045, 5.3856, 4.7489, 4.8079, 3.0256], device='cuda:1'), covar=tensor([0.0234, 0.0253, 0.0265, 0.0447, 0.0878, 0.0225, 0.0279, 0.1453], device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0200, 0.0198, 0.0214, 0.0375, 0.0167, 0.0187, 0.0217], 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') 2023-03-09 03:11:05,265 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.1287, 5.3276, 5.6569, 5.4636, 5.6459, 6.0592, 5.3071, 6.1267], device='cuda:1'), covar=tensor([0.0676, 0.0698, 0.0790, 0.1476, 0.1712, 0.0861, 0.0774, 0.0641], device='cuda:1'), in_proj_covar=tensor([0.0882, 0.0512, 0.0619, 0.0669, 0.0887, 0.0640, 0.0504, 0.0620], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 03:11:32,631 INFO [train2.py:809] (1/4) Epoch 23, batch 3800, loss[ctc_loss=0.07066, att_loss=0.241, loss=0.2069, over 16964.00 frames. utt_duration=1359 frames, utt_pad_proportion=0.007641, over 50.00 utterances.], tot_loss[ctc_loss=0.07115, att_loss=0.235, loss=0.2022, over 3276558.80 frames. utt_duration=1253 frames, utt_pad_proportion=0.05111, over 10471.36 utterances.], batch size: 50, lr: 4.63e-03, grad_scale: 4.0 2023-03-09 03:11:38,946 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.310e+02 1.926e+02 2.458e+02 3.092e+02 6.476e+02, threshold=4.915e+02, percent-clipped=5.0 2023-03-09 03:12:03,216 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0976, 5.0724, 4.8982, 2.0781, 2.0837, 2.7530, 2.2525, 3.8410], device='cuda:1'), covar=tensor([0.0676, 0.0278, 0.0248, 0.5192, 0.5195, 0.2597, 0.3843, 0.1684], device='cuda:1'), in_proj_covar=tensor([0.0354, 0.0278, 0.0267, 0.0241, 0.0336, 0.0331, 0.0254, 0.0364], 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') 2023-03-09 03:12:55,887 INFO [train2.py:809] (1/4) Epoch 23, batch 3850, loss[ctc_loss=0.07045, att_loss=0.2496, loss=0.2138, over 17361.00 frames. utt_duration=1104 frames, utt_pad_proportion=0.03512, over 63.00 utterances.], tot_loss[ctc_loss=0.07025, att_loss=0.2349, loss=0.202, over 3276352.67 frames. utt_duration=1244 frames, utt_pad_proportion=0.05294, over 10544.52 utterances.], batch size: 63, lr: 4.63e-03, grad_scale: 4.0 2023-03-09 03:13:14,993 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91505.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:13:19,672 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5046, 2.6807, 4.9165, 3.7716, 3.0931, 4.1724, 4.6523, 4.6705], device='cuda:1'), covar=tensor([0.0267, 0.1588, 0.0187, 0.0950, 0.1574, 0.0252, 0.0180, 0.0252], device='cuda:1'), in_proj_covar=tensor([0.0206, 0.0241, 0.0196, 0.0317, 0.0263, 0.0220, 0.0187, 0.0216], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 03:13:51,997 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0399, 4.1101, 4.0405, 3.9689, 4.4130, 4.0233, 3.8942, 2.5270], device='cuda:1'), covar=tensor([0.0341, 0.0470, 0.0472, 0.0419, 0.0916, 0.0342, 0.0490, 0.1782], device='cuda:1'), in_proj_covar=tensor([0.0175, 0.0202, 0.0201, 0.0217, 0.0379, 0.0170, 0.0189, 0.0219], 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') 2023-03-09 03:14:10,502 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.5478, 4.8843, 5.0843, 4.9730, 5.0706, 5.4911, 4.9844, 5.5838], device='cuda:1'), covar=tensor([0.0837, 0.0821, 0.0923, 0.1509, 0.2081, 0.0984, 0.1017, 0.0692], device='cuda:1'), in_proj_covar=tensor([0.0879, 0.0510, 0.0615, 0.0666, 0.0882, 0.0638, 0.0501, 0.0614], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 03:14:12,228 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91542.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 03:14:13,406 INFO [train2.py:809] (1/4) Epoch 23, batch 3900, loss[ctc_loss=0.07597, att_loss=0.2296, loss=0.1989, over 16012.00 frames. utt_duration=1602 frames, utt_pad_proportion=0.007141, over 40.00 utterances.], tot_loss[ctc_loss=0.07041, att_loss=0.2346, loss=0.2018, over 3268424.81 frames. utt_duration=1243 frames, utt_pad_proportion=0.05482, over 10533.47 utterances.], batch size: 40, lr: 4.63e-03, grad_scale: 4.0 2023-03-09 03:14:19,548 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.215e+02 1.985e+02 2.304e+02 2.861e+02 4.797e+02, threshold=4.607e+02, percent-clipped=0.0 2023-03-09 03:14:29,564 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=91553.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:14:36,527 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.81 vs. limit=2.0 2023-03-09 03:15:27,448 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9216, 6.1316, 5.6279, 5.9272, 5.8396, 5.3144, 5.6260, 5.3413], device='cuda:1'), covar=tensor([0.1243, 0.0850, 0.1025, 0.0713, 0.0850, 0.1596, 0.2103, 0.2144], device='cuda:1'), in_proj_covar=tensor([0.0526, 0.0615, 0.0467, 0.0460, 0.0434, 0.0473, 0.0616, 0.0533], 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') 2023-03-09 03:15:32,012 INFO [train2.py:809] (1/4) Epoch 23, batch 3950, loss[ctc_loss=0.04108, att_loss=0.2203, loss=0.1845, over 16897.00 frames. utt_duration=1381 frames, utt_pad_proportion=0.005906, over 49.00 utterances.], tot_loss[ctc_loss=0.06995, att_loss=0.2336, loss=0.2009, over 3263872.34 frames. utt_duration=1252 frames, utt_pad_proportion=0.05534, over 10437.62 utterances.], batch size: 49, lr: 4.63e-03, grad_scale: 4.0 2023-03-09 03:15:48,474 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=91603.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 03:16:21,135 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4888, 2.7695, 4.8706, 3.9078, 2.9004, 4.1524, 4.7012, 4.6781], device='cuda:1'), covar=tensor([0.0307, 0.1462, 0.0242, 0.0879, 0.1850, 0.0289, 0.0212, 0.0284], device='cuda:1'), in_proj_covar=tensor([0.0206, 0.0240, 0.0196, 0.0316, 0.0262, 0.0220, 0.0187, 0.0216], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 03:16:52,903 INFO [train2.py:809] (1/4) Epoch 24, batch 0, loss[ctc_loss=0.06151, att_loss=0.2108, loss=0.181, over 15870.00 frames. utt_duration=1629 frames, utt_pad_proportion=0.009602, over 39.00 utterances.], tot_loss[ctc_loss=0.06151, att_loss=0.2108, loss=0.181, over 15870.00 frames. utt_duration=1629 frames, utt_pad_proportion=0.009602, over 39.00 utterances.], batch size: 39, lr: 4.53e-03, grad_scale: 8.0 2023-03-09 03:16:52,904 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-09 03:17:00,616 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.8677, 4.7146, 4.6816, 2.1677, 2.0229, 2.9970, 2.3260, 3.7522], device='cuda:1'), covar=tensor([0.0723, 0.0277, 0.0263, 0.5202, 0.5725, 0.2405, 0.3969, 0.1366], device='cuda:1'), in_proj_covar=tensor([0.0359, 0.0282, 0.0271, 0.0245, 0.0341, 0.0334, 0.0257, 0.0369], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-03-09 03:17:05,994 INFO [train2.py:843] (1/4) Epoch 24, validation: ctc_loss=0.04095, att_loss=0.2349, loss=0.1961, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-09 03:17:05,995 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-09 03:17:28,519 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=91641.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:17:37,663 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.284e+02 2.024e+02 2.562e+02 3.130e+02 9.930e+02, threshold=5.124e+02, percent-clipped=6.0 2023-03-09 03:18:15,968 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-03-09 03:18:27,177 INFO [train2.py:809] (1/4) Epoch 24, batch 50, loss[ctc_loss=0.05976, att_loss=0.2082, loss=0.1785, over 15878.00 frames. utt_duration=1630 frames, utt_pad_proportion=0.009056, over 39.00 utterances.], tot_loss[ctc_loss=0.07126, att_loss=0.2373, loss=0.2041, over 749168.40 frames. utt_duration=1171 frames, utt_pad_proportion=0.05905, over 2562.74 utterances.], batch size: 39, lr: 4.53e-03, grad_scale: 8.0 2023-03-09 03:18:54,934 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1590, 4.6130, 4.8343, 4.8903, 3.0132, 4.5414, 3.0407, 1.9117], device='cuda:1'), covar=tensor([0.0458, 0.0265, 0.0490, 0.0205, 0.1426, 0.0227, 0.1243, 0.1644], device='cuda:1'), in_proj_covar=tensor([0.0200, 0.0171, 0.0262, 0.0165, 0.0224, 0.0156, 0.0233, 0.0204], 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') 2023-03-09 03:19:47,844 INFO [train2.py:809] (1/4) Epoch 24, batch 100, loss[ctc_loss=0.06749, att_loss=0.2382, loss=0.204, over 17028.00 frames. utt_duration=1337 frames, utt_pad_proportion=0.007337, over 51.00 utterances.], tot_loss[ctc_loss=0.06969, att_loss=0.2348, loss=0.2018, over 1313448.42 frames. utt_duration=1231 frames, utt_pad_proportion=0.04958, over 4274.43 utterances.], batch size: 51, lr: 4.52e-03, grad_scale: 8.0 2023-03-09 03:20:00,584 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.30 vs. limit=5.0 2023-03-09 03:20:20,175 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.176e+02 1.937e+02 2.364e+02 2.777e+02 5.363e+02, threshold=4.728e+02, percent-clipped=1.0 2023-03-09 03:21:08,901 INFO [train2.py:809] (1/4) Epoch 24, batch 150, loss[ctc_loss=0.0489, att_loss=0.2192, loss=0.1851, over 15966.00 frames. utt_duration=1559 frames, utt_pad_proportion=0.005522, over 41.00 utterances.], tot_loss[ctc_loss=0.06932, att_loss=0.234, loss=0.2011, over 1748646.16 frames. utt_duration=1271 frames, utt_pad_proportion=0.04431, over 5509.98 utterances.], batch size: 41, lr: 4.52e-03, grad_scale: 8.0 2023-03-09 03:21:29,510 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-03-09 03:22:03,491 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.65 vs. limit=5.0 2023-03-09 03:22:30,994 INFO [train2.py:809] (1/4) Epoch 24, batch 200, loss[ctc_loss=0.05896, att_loss=0.2178, loss=0.1861, over 16006.00 frames. utt_duration=1602 frames, utt_pad_proportion=0.00728, over 40.00 utterances.], tot_loss[ctc_loss=0.06871, att_loss=0.2338, loss=0.2008, over 2088840.37 frames. utt_duration=1283 frames, utt_pad_proportion=0.04232, over 6519.52 utterances.], batch size: 40, lr: 4.52e-03, grad_scale: 8.0 2023-03-09 03:22:49,084 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8289, 5.1184, 4.7027, 5.1889, 4.5971, 4.9272, 5.2451, 5.0135], device='cuda:1'), covar=tensor([0.0659, 0.0309, 0.0835, 0.0354, 0.0447, 0.0251, 0.0232, 0.0206], device='cuda:1'), in_proj_covar=tensor([0.0395, 0.0324, 0.0367, 0.0358, 0.0329, 0.0239, 0.0308, 0.0289], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-09 03:23:02,671 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.186e+02 1.924e+02 2.312e+02 2.679e+02 4.327e+02, threshold=4.624e+02, percent-clipped=0.0 2023-03-09 03:23:10,155 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.1602, 3.8222, 3.3116, 3.4303, 4.0251, 3.6984, 3.1287, 4.3405], device='cuda:1'), covar=tensor([0.0984, 0.0465, 0.1101, 0.0741, 0.0698, 0.0749, 0.0811, 0.0466], device='cuda:1'), in_proj_covar=tensor([0.0204, 0.0220, 0.0226, 0.0204, 0.0281, 0.0244, 0.0200, 0.0291], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-09 03:23:30,251 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0196, 5.3625, 4.8782, 5.4029, 4.7883, 5.1127, 5.4954, 5.2498], device='cuda:1'), covar=tensor([0.0593, 0.0272, 0.0887, 0.0313, 0.0436, 0.0237, 0.0235, 0.0192], device='cuda:1'), in_proj_covar=tensor([0.0397, 0.0325, 0.0369, 0.0360, 0.0330, 0.0240, 0.0309, 0.0291], device='cuda:1'), out_proj_covar=tensor([0.0007, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-09 03:23:51,120 INFO [train2.py:809] (1/4) Epoch 24, batch 250, loss[ctc_loss=0.06549, att_loss=0.2205, loss=0.1895, over 15764.00 frames. utt_duration=1661 frames, utt_pad_proportion=0.008981, over 38.00 utterances.], tot_loss[ctc_loss=0.06876, att_loss=0.2335, loss=0.2005, over 2353337.38 frames. utt_duration=1263 frames, utt_pad_proportion=0.04561, over 7460.68 utterances.], batch size: 38, lr: 4.52e-03, grad_scale: 8.0 2023-03-09 03:24:14,489 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91891.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:24:25,444 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=91898.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 03:25:11,834 INFO [train2.py:809] (1/4) Epoch 24, batch 300, loss[ctc_loss=0.06148, att_loss=0.2325, loss=0.1983, over 16330.00 frames. utt_duration=1453 frames, utt_pad_proportion=0.006126, over 45.00 utterances.], tot_loss[ctc_loss=0.06887, att_loss=0.2329, loss=0.2001, over 2556155.34 frames. utt_duration=1302 frames, utt_pad_proportion=0.03794, over 7860.08 utterances.], batch size: 45, lr: 4.52e-03, grad_scale: 8.0 2023-03-09 03:25:35,412 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91941.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:25:44,283 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.141e+02 1.878e+02 2.164e+02 2.654e+02 4.888e+02, threshold=4.328e+02, percent-clipped=1.0 2023-03-09 03:25:53,263 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=91952.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:25:57,831 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91955.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:26:32,684 INFO [train2.py:809] (1/4) Epoch 24, batch 350, loss[ctc_loss=0.06361, att_loss=0.2375, loss=0.2028, over 16760.00 frames. utt_duration=1398 frames, utt_pad_proportion=0.00694, over 48.00 utterances.], tot_loss[ctc_loss=0.06906, att_loss=0.2322, loss=0.1996, over 2708812.11 frames. utt_duration=1299 frames, utt_pad_proportion=0.04244, over 8347.80 utterances.], batch size: 48, lr: 4.52e-03, grad_scale: 8.0 2023-03-09 03:26:39,960 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91981.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 03:26:52,898 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=91989.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:27:40,079 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92016.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:27:58,109 INFO [train2.py:809] (1/4) Epoch 24, batch 400, loss[ctc_loss=0.08405, att_loss=0.2451, loss=0.2129, over 16884.00 frames. utt_duration=1380 frames, utt_pad_proportion=0.006641, over 49.00 utterances.], tot_loss[ctc_loss=0.07015, att_loss=0.2332, loss=0.2006, over 2830566.55 frames. utt_duration=1255 frames, utt_pad_proportion=0.05508, over 9031.43 utterances.], batch size: 49, lr: 4.52e-03, grad_scale: 8.0 2023-03-09 03:28:16,341 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4484, 4.5879, 4.6015, 4.6386, 4.6694, 4.6711, 4.3369, 4.2129], device='cuda:1'), covar=tensor([0.1070, 0.0707, 0.0499, 0.0520, 0.0348, 0.0390, 0.0469, 0.0413], device='cuda:1'), in_proj_covar=tensor([0.0527, 0.0366, 0.0354, 0.0361, 0.0428, 0.0438, 0.0362, 0.0402], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0003, 0.0004], device='cuda:1') 2023-03-09 03:28:22,912 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92042.0, num_to_drop=1, layers_to_drop={3} 2023-03-09 03:28:30,201 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.202e+02 1.943e+02 2.339e+02 2.970e+02 6.262e+02, threshold=4.678e+02, percent-clipped=8.0 2023-03-09 03:29:04,484 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.5141, 5.7711, 5.2615, 5.6023, 5.4459, 4.9997, 5.2049, 4.9529], device='cuda:1'), covar=tensor([0.1436, 0.0954, 0.0998, 0.0837, 0.1091, 0.1595, 0.2416, 0.2457], device='cuda:1'), in_proj_covar=tensor([0.0533, 0.0621, 0.0471, 0.0464, 0.0438, 0.0473, 0.0622, 0.0538], 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') 2023-03-09 03:29:18,995 INFO [train2.py:809] (1/4) Epoch 24, batch 450, loss[ctc_loss=0.07769, att_loss=0.2529, loss=0.2179, over 17479.00 frames. utt_duration=1015 frames, utt_pad_proportion=0.04346, over 69.00 utterances.], tot_loss[ctc_loss=0.07, att_loss=0.2334, loss=0.2007, over 2932553.42 frames. utt_duration=1241 frames, utt_pad_proportion=0.05676, over 9461.35 utterances.], batch size: 69, lr: 4.52e-03, grad_scale: 8.0 2023-03-09 03:30:40,866 INFO [train2.py:809] (1/4) Epoch 24, batch 500, loss[ctc_loss=0.06703, att_loss=0.2241, loss=0.1927, over 16116.00 frames. utt_duration=1536 frames, utt_pad_proportion=0.006268, over 42.00 utterances.], tot_loss[ctc_loss=0.07044, att_loss=0.2337, loss=0.201, over 3004785.54 frames. utt_duration=1219 frames, utt_pad_proportion=0.06161, over 9874.71 utterances.], batch size: 42, lr: 4.51e-03, grad_scale: 8.0 2023-03-09 03:31:13,731 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.285e+02 1.745e+02 2.110e+02 2.591e+02 7.932e+02, threshold=4.219e+02, percent-clipped=4.0 2023-03-09 03:32:01,718 INFO [train2.py:809] (1/4) Epoch 24, batch 550, loss[ctc_loss=0.06383, att_loss=0.2427, loss=0.207, over 16757.00 frames. utt_duration=1398 frames, utt_pad_proportion=0.007117, over 48.00 utterances.], tot_loss[ctc_loss=0.07013, att_loss=0.2338, loss=0.2011, over 3063112.92 frames. utt_duration=1220 frames, utt_pad_proportion=0.06278, over 10054.26 utterances.], batch size: 48, lr: 4.51e-03, grad_scale: 8.0 2023-03-09 03:32:18,889 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-03-09 03:32:35,884 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92198.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 03:33:22,318 INFO [train2.py:809] (1/4) Epoch 24, batch 600, loss[ctc_loss=0.07634, att_loss=0.2477, loss=0.2134, over 17433.00 frames. utt_duration=884.1 frames, utt_pad_proportion=0.07326, over 79.00 utterances.], tot_loss[ctc_loss=0.06998, att_loss=0.234, loss=0.2012, over 3112939.46 frames. utt_duration=1218 frames, utt_pad_proportion=0.06225, over 10239.89 utterances.], batch size: 79, lr: 4.51e-03, grad_scale: 8.0 2023-03-09 03:33:53,651 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=92246.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 03:33:54,959 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.139e+02 1.830e+02 2.241e+02 2.667e+02 5.450e+02, threshold=4.482e+02, percent-clipped=6.0 2023-03-09 03:33:55,222 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92247.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:34:38,849 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=5.00 vs. limit=5.0 2023-03-09 03:34:42,404 INFO [train2.py:809] (1/4) Epoch 24, batch 650, loss[ctc_loss=0.0748, att_loss=0.2426, loss=0.2091, over 17044.00 frames. utt_duration=1312 frames, utt_pad_proportion=0.009492, over 52.00 utterances.], tot_loss[ctc_loss=0.07033, att_loss=0.2343, loss=0.2015, over 3159104.05 frames. utt_duration=1219 frames, utt_pad_proportion=0.05827, over 10377.81 utterances.], batch size: 52, lr: 4.51e-03, grad_scale: 8.0 2023-03-09 03:35:24,925 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3752, 4.5969, 4.5600, 4.5377, 5.1767, 4.4888, 4.5197, 2.5656], device='cuda:1'), covar=tensor([0.0324, 0.0359, 0.0398, 0.0375, 0.0715, 0.0289, 0.0405, 0.1839], device='cuda:1'), in_proj_covar=tensor([0.0176, 0.0202, 0.0202, 0.0218, 0.0380, 0.0170, 0.0190, 0.0218], 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') 2023-03-09 03:35:37,901 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92311.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:36:03,497 INFO [train2.py:809] (1/4) Epoch 24, batch 700, loss[ctc_loss=0.05618, att_loss=0.2427, loss=0.2054, over 17305.00 frames. utt_duration=1260 frames, utt_pad_proportion=0.01191, over 55.00 utterances.], tot_loss[ctc_loss=0.07061, att_loss=0.2351, loss=0.2022, over 3193395.53 frames. utt_duration=1223 frames, utt_pad_proportion=0.05493, over 10457.27 utterances.], batch size: 55, lr: 4.51e-03, grad_scale: 8.0 2023-03-09 03:36:20,891 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92337.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 03:36:20,918 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5889, 4.9906, 4.1419, 5.1763, 4.5005, 4.8644, 5.0334, 4.9136], device='cuda:1'), covar=tensor([0.0753, 0.0384, 0.1311, 0.0420, 0.0443, 0.0354, 0.0398, 0.0258], device='cuda:1'), in_proj_covar=tensor([0.0389, 0.0322, 0.0364, 0.0355, 0.0326, 0.0237, 0.0306, 0.0286], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0005, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-09 03:36:36,599 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.268e+02 1.935e+02 2.344e+02 2.941e+02 5.112e+02, threshold=4.689e+02, percent-clipped=2.0 2023-03-09 03:37:25,733 INFO [train2.py:809] (1/4) Epoch 24, batch 750, loss[ctc_loss=0.0713, att_loss=0.2509, loss=0.215, over 16480.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.005921, over 46.00 utterances.], tot_loss[ctc_loss=0.07069, att_loss=0.2352, loss=0.2023, over 3214138.20 frames. utt_duration=1216 frames, utt_pad_proportion=0.05756, over 10587.79 utterances.], batch size: 46, lr: 4.51e-03, grad_scale: 8.0 2023-03-09 03:38:21,659 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-03-09 03:38:46,216 INFO [train2.py:809] (1/4) Epoch 24, batch 800, loss[ctc_loss=0.09541, att_loss=0.2556, loss=0.2236, over 16856.00 frames. utt_duration=682.5 frames, utt_pad_proportion=0.1447, over 99.00 utterances.], tot_loss[ctc_loss=0.07068, att_loss=0.2349, loss=0.2021, over 3225633.91 frames. utt_duration=1226 frames, utt_pad_proportion=0.05706, over 10537.76 utterances.], batch size: 99, lr: 4.51e-03, grad_scale: 8.0 2023-03-09 03:39:05,220 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9102, 5.3490, 5.1634, 5.1784, 5.2715, 5.0443, 3.7665, 5.3034], device='cuda:1'), covar=tensor([0.0102, 0.0091, 0.0117, 0.0080, 0.0099, 0.0101, 0.0614, 0.0134], device='cuda:1'), in_proj_covar=tensor([0.0093, 0.0089, 0.0112, 0.0071, 0.0076, 0.0087, 0.0103, 0.0109], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 03:39:14,956 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9658, 5.2536, 4.7686, 5.3365, 4.6239, 5.0954, 5.3953, 5.1519], device='cuda:1'), covar=tensor([0.0577, 0.0311, 0.0961, 0.0341, 0.0507, 0.0221, 0.0237, 0.0231], device='cuda:1'), in_proj_covar=tensor([0.0391, 0.0324, 0.0367, 0.0357, 0.0329, 0.0239, 0.0308, 0.0287], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-09 03:39:19,378 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.406e+02 1.923e+02 2.416e+02 3.239e+02 1.370e+03, threshold=4.832e+02, percent-clipped=5.0 2023-03-09 03:40:07,916 INFO [train2.py:809] (1/4) Epoch 24, batch 850, loss[ctc_loss=0.07198, att_loss=0.2379, loss=0.2047, over 16882.00 frames. utt_duration=1380 frames, utt_pad_proportion=0.006803, over 49.00 utterances.], tot_loss[ctc_loss=0.07059, att_loss=0.2351, loss=0.2022, over 3241675.16 frames. utt_duration=1230 frames, utt_pad_proportion=0.05458, over 10552.57 utterances.], batch size: 49, lr: 4.51e-03, grad_scale: 8.0 2023-03-09 03:41:29,527 INFO [train2.py:809] (1/4) Epoch 24, batch 900, loss[ctc_loss=0.06578, att_loss=0.2167, loss=0.1865, over 14483.00 frames. utt_duration=1812 frames, utt_pad_proportion=0.03615, over 32.00 utterances.], tot_loss[ctc_loss=0.07034, att_loss=0.2342, loss=0.2014, over 3247675.65 frames. utt_duration=1243 frames, utt_pad_proportion=0.0523, over 10461.98 utterances.], batch size: 32, lr: 4.50e-03, grad_scale: 8.0 2023-03-09 03:42:02,357 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.067e+02 1.834e+02 2.155e+02 2.748e+02 5.069e+02, threshold=4.309e+02, percent-clipped=1.0 2023-03-09 03:42:02,719 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92547.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:42:29,744 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-03-09 03:42:50,355 INFO [train2.py:809] (1/4) Epoch 24, batch 950, loss[ctc_loss=0.05454, att_loss=0.2018, loss=0.1724, over 15375.00 frames. utt_duration=1758 frames, utt_pad_proportion=0.0106, over 35.00 utterances.], tot_loss[ctc_loss=0.07051, att_loss=0.2341, loss=0.2014, over 3250918.29 frames. utt_duration=1230 frames, utt_pad_proportion=0.05729, over 10588.59 utterances.], batch size: 35, lr: 4.50e-03, grad_scale: 8.0 2023-03-09 03:43:20,155 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=92595.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:43:28,603 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1132, 4.4913, 4.3969, 4.6479, 2.9382, 4.5715, 2.7335, 2.1404], device='cuda:1'), covar=tensor([0.0512, 0.0241, 0.0752, 0.0224, 0.1666, 0.0197, 0.1676, 0.1687], device='cuda:1'), in_proj_covar=tensor([0.0201, 0.0170, 0.0263, 0.0166, 0.0223, 0.0156, 0.0231, 0.0202], 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') 2023-03-09 03:43:46,281 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92611.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:44:08,896 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92625.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:44:11,647 INFO [train2.py:809] (1/4) Epoch 24, batch 1000, loss[ctc_loss=0.05468, att_loss=0.2387, loss=0.2019, over 16625.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.005454, over 47.00 utterances.], tot_loss[ctc_loss=0.07004, att_loss=0.2335, loss=0.2008, over 3243578.09 frames. utt_duration=1222 frames, utt_pad_proportion=0.0615, over 10625.95 utterances.], batch size: 47, lr: 4.50e-03, grad_scale: 8.0 2023-03-09 03:44:28,684 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92637.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 03:44:44,465 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.218e+02 1.788e+02 2.232e+02 2.616e+02 5.460e+02, threshold=4.464e+02, percent-clipped=2.0 2023-03-09 03:45:04,282 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=92659.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:45:09,415 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2432, 4.3172, 4.3813, 4.3401, 4.9075, 4.3258, 4.2572, 2.4561], device='cuda:1'), covar=tensor([0.0327, 0.0376, 0.0366, 0.0328, 0.0754, 0.0305, 0.0403, 0.1703], device='cuda:1'), in_proj_covar=tensor([0.0175, 0.0201, 0.0200, 0.0217, 0.0377, 0.0170, 0.0189, 0.0216], 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') 2023-03-09 03:45:33,246 INFO [train2.py:809] (1/4) Epoch 24, batch 1050, loss[ctc_loss=0.08302, att_loss=0.2479, loss=0.215, over 17403.00 frames. utt_duration=1010 frames, utt_pad_proportion=0.0458, over 69.00 utterances.], tot_loss[ctc_loss=0.06982, att_loss=0.2341, loss=0.2013, over 3252773.93 frames. utt_duration=1201 frames, utt_pad_proportion=0.06667, over 10851.29 utterances.], batch size: 69, lr: 4.50e-03, grad_scale: 8.0 2023-03-09 03:45:47,609 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=92685.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 03:45:49,263 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92686.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:46:27,009 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1765, 4.6163, 4.4287, 4.5353, 4.6514, 4.2932, 3.0922, 4.5171], device='cuda:1'), covar=tensor([0.0153, 0.0124, 0.0161, 0.0094, 0.0107, 0.0145, 0.0783, 0.0212], device='cuda:1'), in_proj_covar=tensor([0.0093, 0.0089, 0.0112, 0.0071, 0.0076, 0.0087, 0.0103, 0.0108], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 03:46:43,770 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92720.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:46:54,333 INFO [train2.py:809] (1/4) Epoch 24, batch 1100, loss[ctc_loss=0.07311, att_loss=0.2485, loss=0.2134, over 17243.00 frames. utt_duration=1171 frames, utt_pad_proportion=0.0278, over 59.00 utterances.], tot_loss[ctc_loss=0.07024, att_loss=0.2342, loss=0.2014, over 3248638.19 frames. utt_duration=1187 frames, utt_pad_proportion=0.07318, over 10960.30 utterances.], batch size: 59, lr: 4.50e-03, grad_scale: 8.0 2023-03-09 03:47:27,345 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.412e+02 1.778e+02 2.254e+02 2.467e+02 8.926e+02, threshold=4.507e+02, percent-clipped=3.0 2023-03-09 03:48:16,145 INFO [train2.py:809] (1/4) Epoch 24, batch 1150, loss[ctc_loss=0.06179, att_loss=0.2121, loss=0.182, over 15988.00 frames. utt_duration=1600 frames, utt_pad_proportion=0.008457, over 40.00 utterances.], tot_loss[ctc_loss=0.07014, att_loss=0.2341, loss=0.2013, over 3258714.79 frames. utt_duration=1191 frames, utt_pad_proportion=0.06992, over 10958.26 utterances.], batch size: 40, lr: 4.50e-03, grad_scale: 8.0 2023-03-09 03:48:23,643 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92781.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:48:25,266 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5435, 4.7016, 4.6903, 4.6711, 5.2429, 4.6297, 4.5586, 2.6753], device='cuda:1'), covar=tensor([0.0282, 0.0279, 0.0315, 0.0339, 0.0842, 0.0264, 0.0357, 0.1633], device='cuda:1'), in_proj_covar=tensor([0.0175, 0.0202, 0.0201, 0.0217, 0.0377, 0.0170, 0.0190, 0.0217], 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') 2023-03-09 03:49:37,245 INFO [train2.py:809] (1/4) Epoch 24, batch 1200, loss[ctc_loss=0.09564, att_loss=0.2307, loss=0.2037, over 15944.00 frames. utt_duration=1557 frames, utt_pad_proportion=0.006969, over 41.00 utterances.], tot_loss[ctc_loss=0.06999, att_loss=0.2339, loss=0.2012, over 3260724.26 frames. utt_duration=1207 frames, utt_pad_proportion=0.06661, over 10817.38 utterances.], batch size: 41, lr: 4.50e-03, grad_scale: 8.0 2023-03-09 03:49:45,310 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92831.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:50:10,381 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.270e+02 1.947e+02 2.306e+02 3.014e+02 1.019e+03, threshold=4.613e+02, percent-clipped=6.0 2023-03-09 03:50:27,965 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-03-09 03:50:59,082 INFO [train2.py:809] (1/4) Epoch 24, batch 1250, loss[ctc_loss=0.07994, att_loss=0.2518, loss=0.2175, over 17308.00 frames. utt_duration=1260 frames, utt_pad_proportion=0.01139, over 55.00 utterances.], tot_loss[ctc_loss=0.06989, att_loss=0.2342, loss=0.2014, over 3270022.32 frames. utt_duration=1243 frames, utt_pad_proportion=0.056, over 10536.17 utterances.], batch size: 55, lr: 4.50e-03, grad_scale: 8.0 2023-03-09 03:51:24,814 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92892.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:51:35,677 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92899.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:51:49,794 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-03-09 03:52:20,572 INFO [train2.py:809] (1/4) Epoch 24, batch 1300, loss[ctc_loss=0.05641, att_loss=0.2136, loss=0.1822, over 15997.00 frames. utt_duration=1601 frames, utt_pad_proportion=0.007993, over 40.00 utterances.], tot_loss[ctc_loss=0.06977, att_loss=0.2336, loss=0.2008, over 3268571.71 frames. utt_duration=1253 frames, utt_pad_proportion=0.05357, over 10446.40 utterances.], batch size: 40, lr: 4.50e-03, grad_scale: 8.0 2023-03-09 03:52:27,986 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92931.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:52:53,102 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.274e+02 1.940e+02 2.372e+02 2.718e+02 7.433e+02, threshold=4.745e+02, percent-clipped=1.0 2023-03-09 03:53:14,565 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92960.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:53:35,860 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0664, 5.2879, 5.3225, 5.1851, 5.3392, 5.2982, 4.9493, 4.8054], device='cuda:1'), covar=tensor([0.0934, 0.0501, 0.0285, 0.0544, 0.0269, 0.0301, 0.0348, 0.0314], device='cuda:1'), in_proj_covar=tensor([0.0527, 0.0364, 0.0350, 0.0363, 0.0427, 0.0435, 0.0361, 0.0398], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0003, 0.0004], device='cuda:1') 2023-03-09 03:53:39,312 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92975.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:53:39,415 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3795, 2.5961, 4.7515, 3.7223, 2.8585, 4.1021, 4.4662, 4.4501], device='cuda:1'), covar=tensor([0.0274, 0.1466, 0.0172, 0.0911, 0.1690, 0.0273, 0.0205, 0.0275], device='cuda:1'), in_proj_covar=tensor([0.0210, 0.0243, 0.0202, 0.0320, 0.0265, 0.0223, 0.0191, 0.0218], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 03:53:41,992 INFO [train2.py:809] (1/4) Epoch 24, batch 1350, loss[ctc_loss=0.07445, att_loss=0.2227, loss=0.1931, over 14455.00 frames. utt_duration=1808 frames, utt_pad_proportion=0.05279, over 32.00 utterances.], tot_loss[ctc_loss=0.06988, att_loss=0.2342, loss=0.2014, over 3273704.16 frames. utt_duration=1257 frames, utt_pad_proportion=0.05094, over 10427.79 utterances.], batch size: 32, lr: 4.49e-03, grad_scale: 8.0 2023-03-09 03:53:49,249 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92981.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:53:56,686 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.18 vs. limit=5.0 2023-03-09 03:54:07,133 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92992.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:54:36,437 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93010.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:55:03,583 INFO [train2.py:809] (1/4) Epoch 24, batch 1400, loss[ctc_loss=0.05724, att_loss=0.2235, loss=0.1902, over 15950.00 frames. utt_duration=1558 frames, utt_pad_proportion=0.007322, over 41.00 utterances.], tot_loss[ctc_loss=0.06937, att_loss=0.2334, loss=0.2006, over 3261581.17 frames. utt_duration=1240 frames, utt_pad_proportion=0.05959, over 10535.42 utterances.], batch size: 41, lr: 4.49e-03, grad_scale: 8.0 2023-03-09 03:55:19,521 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93036.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:55:19,551 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93036.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:55:21,533 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-03-09 03:55:36,735 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.237e+02 1.820e+02 2.183e+02 2.571e+02 5.095e+02, threshold=4.367e+02, percent-clipped=2.0 2023-03-09 03:55:50,308 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9282, 5.2308, 4.8019, 5.3101, 4.6891, 4.9744, 5.3894, 5.1562], device='cuda:1'), covar=tensor([0.0566, 0.0265, 0.0829, 0.0305, 0.0409, 0.0231, 0.0209, 0.0192], device='cuda:1'), in_proj_covar=tensor([0.0393, 0.0324, 0.0368, 0.0356, 0.0327, 0.0240, 0.0308, 0.0288], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-09 03:55:53,453 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1351, 5.3443, 5.3952, 5.3159, 5.4364, 5.3932, 5.0551, 4.8684], device='cuda:1'), covar=tensor([0.1021, 0.0567, 0.0247, 0.0487, 0.0268, 0.0316, 0.0321, 0.0319], device='cuda:1'), in_proj_covar=tensor([0.0529, 0.0368, 0.0352, 0.0365, 0.0429, 0.0438, 0.0363, 0.0400], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0003, 0.0004], device='cuda:1') 2023-03-09 03:56:16,532 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93071.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:56:24,016 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93076.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:56:25,392 INFO [train2.py:809] (1/4) Epoch 24, batch 1450, loss[ctc_loss=0.08497, att_loss=0.2451, loss=0.2131, over 16466.00 frames. utt_duration=1433 frames, utt_pad_proportion=0.006644, over 46.00 utterances.], tot_loss[ctc_loss=0.06925, att_loss=0.2334, loss=0.2005, over 3257160.31 frames. utt_duration=1240 frames, utt_pad_proportion=0.06242, over 10522.21 utterances.], batch size: 46, lr: 4.49e-03, grad_scale: 8.0 2023-03-09 03:56:35,722 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9683, 5.2225, 5.5482, 5.2543, 5.4544, 5.9810, 5.1634, 6.0534], device='cuda:1'), covar=tensor([0.0818, 0.0676, 0.0844, 0.1586, 0.1977, 0.0891, 0.0805, 0.0664], device='cuda:1'), in_proj_covar=tensor([0.0895, 0.0519, 0.0622, 0.0671, 0.0894, 0.0645, 0.0506, 0.0629], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 03:56:57,971 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93097.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 03:57:26,919 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93115.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:57:28,401 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5070, 4.7186, 4.6884, 4.6118, 5.1914, 4.6331, 4.6214, 2.5977], device='cuda:1'), covar=tensor([0.0263, 0.0293, 0.0338, 0.0350, 0.0892, 0.0243, 0.0327, 0.1860], device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0200, 0.0198, 0.0213, 0.0371, 0.0168, 0.0187, 0.0214], 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') 2023-03-09 03:57:45,164 INFO [train2.py:809] (1/4) Epoch 24, batch 1500, loss[ctc_loss=0.07657, att_loss=0.2522, loss=0.217, over 16971.00 frames. utt_duration=1359 frames, utt_pad_proportion=0.007319, over 50.00 utterances.], tot_loss[ctc_loss=0.06908, att_loss=0.233, loss=0.2002, over 3259724.29 frames. utt_duration=1262 frames, utt_pad_proportion=0.05641, over 10340.65 utterances.], batch size: 50, lr: 4.49e-03, grad_scale: 8.0 2023-03-09 03:58:17,655 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.288e+02 1.858e+02 2.215e+02 2.637e+02 6.796e+02, threshold=4.431e+02, percent-clipped=2.0 2023-03-09 03:59:04,762 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93176.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 03:59:05,931 INFO [train2.py:809] (1/4) Epoch 24, batch 1550, loss[ctc_loss=0.06801, att_loss=0.2343, loss=0.201, over 16757.00 frames. utt_duration=1398 frames, utt_pad_proportion=0.006925, over 48.00 utterances.], tot_loss[ctc_loss=0.06958, att_loss=0.233, loss=0.2004, over 3255019.48 frames. utt_duration=1267 frames, utt_pad_proportion=0.05526, over 10286.98 utterances.], batch size: 48, lr: 4.49e-03, grad_scale: 8.0 2023-03-09 03:59:23,148 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93187.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:00:26,992 INFO [train2.py:809] (1/4) Epoch 24, batch 1600, loss[ctc_loss=0.08582, att_loss=0.2535, loss=0.22, over 17053.00 frames. utt_duration=1313 frames, utt_pad_proportion=0.008955, over 52.00 utterances.], tot_loss[ctc_loss=0.06983, att_loss=0.234, loss=0.2011, over 3267954.02 frames. utt_duration=1260 frames, utt_pad_proportion=0.05412, over 10387.90 utterances.], batch size: 52, lr: 4.49e-03, grad_scale: 8.0 2023-03-09 04:00:46,071 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9617, 4.9940, 4.7694, 2.2160, 2.0454, 3.0139, 2.2868, 3.7220], device='cuda:1'), covar=tensor([0.0773, 0.0303, 0.0268, 0.4904, 0.5270, 0.2258, 0.3784, 0.1744], device='cuda:1'), in_proj_covar=tensor([0.0358, 0.0283, 0.0270, 0.0245, 0.0337, 0.0334, 0.0260, 0.0369], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-03-09 04:00:59,418 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.441e+02 2.026e+02 2.316e+02 3.152e+02 2.970e+03, threshold=4.633e+02, percent-clipped=9.0 2023-03-09 04:01:12,770 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93255.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:01:26,957 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-03-09 04:01:47,571 INFO [train2.py:809] (1/4) Epoch 24, batch 1650, loss[ctc_loss=0.04782, att_loss=0.2235, loss=0.1884, over 16468.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.007106, over 46.00 utterances.], tot_loss[ctc_loss=0.07012, att_loss=0.2345, loss=0.2016, over 3273696.14 frames. utt_duration=1271 frames, utt_pad_proportion=0.05057, over 10318.87 utterances.], batch size: 46, lr: 4.49e-03, grad_scale: 8.0 2023-03-09 04:01:54,068 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93281.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:02:05,100 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93287.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:03:09,898 INFO [train2.py:809] (1/4) Epoch 24, batch 1700, loss[ctc_loss=0.06199, att_loss=0.2368, loss=0.2019, over 17421.00 frames. utt_duration=1107 frames, utt_pad_proportion=0.03194, over 63.00 utterances.], tot_loss[ctc_loss=0.06962, att_loss=0.2344, loss=0.2015, over 3279664.60 frames. utt_duration=1270 frames, utt_pad_proportion=0.0487, over 10339.10 utterances.], batch size: 63, lr: 4.49e-03, grad_scale: 16.0 2023-03-09 04:03:13,047 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=93329.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:03:16,314 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93331.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:03:42,405 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.289e+02 1.807e+02 2.162e+02 2.444e+02 4.017e+02, threshold=4.325e+02, percent-clipped=0.0 2023-03-09 04:04:14,240 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93366.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:04:29,850 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93376.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:04:31,040 INFO [train2.py:809] (1/4) Epoch 24, batch 1750, loss[ctc_loss=0.07697, att_loss=0.2362, loss=0.2043, over 16963.00 frames. utt_duration=1358 frames, utt_pad_proportion=0.007743, over 50.00 utterances.], tot_loss[ctc_loss=0.06926, att_loss=0.2345, loss=0.2014, over 3285715.43 frames. utt_duration=1268 frames, utt_pad_proportion=0.04705, over 10374.20 utterances.], batch size: 50, lr: 4.48e-03, grad_scale: 16.0 2023-03-09 04:04:56,267 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93392.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 04:04:56,883 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-03-09 04:05:28,889 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-03-09 04:05:48,344 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=93424.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:05:52,974 INFO [train2.py:809] (1/4) Epoch 24, batch 1800, loss[ctc_loss=0.05371, att_loss=0.219, loss=0.1859, over 16178.00 frames. utt_duration=1580 frames, utt_pad_proportion=0.00696, over 41.00 utterances.], tot_loss[ctc_loss=0.06849, att_loss=0.2338, loss=0.2007, over 3283421.94 frames. utt_duration=1275 frames, utt_pad_proportion=0.04741, over 10310.27 utterances.], batch size: 41, lr: 4.48e-03, grad_scale: 16.0 2023-03-09 04:06:25,985 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.320e+02 1.826e+02 2.245e+02 2.637e+02 3.732e+02, threshold=4.489e+02, percent-clipped=0.0 2023-03-09 04:07:05,218 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93471.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:07:14,513 INFO [train2.py:809] (1/4) Epoch 24, batch 1850, loss[ctc_loss=0.05323, att_loss=0.2098, loss=0.1785, over 15945.00 frames. utt_duration=1557 frames, utt_pad_proportion=0.00757, over 41.00 utterances.], tot_loss[ctc_loss=0.06849, att_loss=0.2331, loss=0.2002, over 3269909.13 frames. utt_duration=1252 frames, utt_pad_proportion=0.05658, over 10456.46 utterances.], batch size: 41, lr: 4.48e-03, grad_scale: 16.0 2023-03-09 04:07:31,196 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93487.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:08:29,227 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6631, 5.0746, 4.8404, 5.0699, 5.0705, 4.8074, 3.5015, 5.1183], device='cuda:1'), covar=tensor([0.0123, 0.0114, 0.0134, 0.0081, 0.0090, 0.0116, 0.0708, 0.0162], device='cuda:1'), in_proj_covar=tensor([0.0094, 0.0090, 0.0112, 0.0071, 0.0076, 0.0087, 0.0104, 0.0109], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 04:08:35,371 INFO [train2.py:809] (1/4) Epoch 24, batch 1900, loss[ctc_loss=0.05317, att_loss=0.2096, loss=0.1783, over 14493.00 frames. utt_duration=1813 frames, utt_pad_proportion=0.03717, over 32.00 utterances.], tot_loss[ctc_loss=0.06846, att_loss=0.233, loss=0.2001, over 3266215.02 frames. utt_duration=1249 frames, utt_pad_proportion=0.05763, over 10470.79 utterances.], batch size: 32, lr: 4.48e-03, grad_scale: 16.0 2023-03-09 04:08:48,850 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=93535.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:09:07,453 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.393e+02 1.868e+02 2.241e+02 2.979e+02 6.717e+02, threshold=4.482e+02, percent-clipped=6.0 2023-03-09 04:09:21,293 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93555.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:09:55,682 INFO [train2.py:809] (1/4) Epoch 24, batch 1950, loss[ctc_loss=0.06125, att_loss=0.2309, loss=0.1969, over 16346.00 frames. utt_duration=1454 frames, utt_pad_proportion=0.005229, over 45.00 utterances.], tot_loss[ctc_loss=0.06898, att_loss=0.2334, loss=0.2005, over 3271651.41 frames. utt_duration=1284 frames, utt_pad_proportion=0.04826, over 10205.30 utterances.], batch size: 45, lr: 4.48e-03, grad_scale: 16.0 2023-03-09 04:10:12,656 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93587.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:10:28,360 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4788, 5.0707, 5.2457, 5.2638, 5.0524, 5.2139, 4.9155, 4.7040], device='cuda:1'), covar=tensor([0.1985, 0.1025, 0.0383, 0.0599, 0.0885, 0.0481, 0.0456, 0.0467], device='cuda:1'), in_proj_covar=tensor([0.0538, 0.0373, 0.0361, 0.0372, 0.0437, 0.0444, 0.0370, 0.0405], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:1') 2023-03-09 04:10:38,951 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=93603.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:11:17,170 INFO [train2.py:809] (1/4) Epoch 24, batch 2000, loss[ctc_loss=0.1058, att_loss=0.2658, loss=0.2338, over 17069.00 frames. utt_duration=1290 frames, utt_pad_proportion=0.008528, over 53.00 utterances.], tot_loss[ctc_loss=0.07023, att_loss=0.2344, loss=0.2016, over 3277021.94 frames. utt_duration=1252 frames, utt_pad_proportion=0.0537, over 10485.78 utterances.], batch size: 53, lr: 4.48e-03, grad_scale: 16.0 2023-03-09 04:11:24,574 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93631.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:11:31,241 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=93635.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:11:50,190 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.339e+02 1.872e+02 2.240e+02 2.588e+02 7.651e+02, threshold=4.480e+02, percent-clipped=4.0 2023-03-09 04:12:09,869 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4148, 4.5558, 4.6329, 4.5733, 5.2410, 4.6374, 4.5097, 2.7699], device='cuda:1'), covar=tensor([0.0323, 0.0376, 0.0332, 0.0342, 0.0769, 0.0263, 0.0374, 0.1621], device='cuda:1'), in_proj_covar=tensor([0.0173, 0.0200, 0.0197, 0.0214, 0.0371, 0.0169, 0.0186, 0.0214], 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') 2023-03-09 04:12:22,043 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93666.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:12:38,935 INFO [train2.py:809] (1/4) Epoch 24, batch 2050, loss[ctc_loss=0.05167, att_loss=0.2343, loss=0.1978, over 16328.00 frames. utt_duration=1453 frames, utt_pad_proportion=0.006232, over 45.00 utterances.], tot_loss[ctc_loss=0.06981, att_loss=0.2334, loss=0.2007, over 3260131.36 frames. utt_duration=1226 frames, utt_pad_proportion=0.06479, over 10648.97 utterances.], batch size: 45, lr: 4.48e-03, grad_scale: 16.0 2023-03-09 04:12:42,978 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=93679.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:13:03,654 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93692.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:13:39,469 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=93714.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:14:00,051 INFO [train2.py:809] (1/4) Epoch 24, batch 2100, loss[ctc_loss=0.06215, att_loss=0.2277, loss=0.1946, over 15940.00 frames. utt_duration=1556 frames, utt_pad_proportion=0.007435, over 41.00 utterances.], tot_loss[ctc_loss=0.06938, att_loss=0.2333, loss=0.2005, over 3264755.88 frames. utt_duration=1255 frames, utt_pad_proportion=0.05596, over 10416.37 utterances.], batch size: 41, lr: 4.48e-03, grad_scale: 8.0 2023-03-09 04:14:21,494 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=93740.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:14:34,153 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.365e+02 1.856e+02 2.273e+02 2.574e+02 6.013e+02, threshold=4.546e+02, percent-clipped=4.0 2023-03-09 04:15:01,025 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 2023-03-09 04:15:09,532 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93770.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:15:11,031 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93771.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:15:19,886 INFO [train2.py:809] (1/4) Epoch 24, batch 2150, loss[ctc_loss=0.07141, att_loss=0.2419, loss=0.2078, over 17060.00 frames. utt_duration=1289 frames, utt_pad_proportion=0.008418, over 53.00 utterances.], tot_loss[ctc_loss=0.06973, att_loss=0.2339, loss=0.2011, over 3269507.85 frames. utt_duration=1231 frames, utt_pad_proportion=0.05938, over 10639.76 utterances.], batch size: 53, lr: 4.47e-03, grad_scale: 8.0 2023-03-09 04:16:28,699 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=93819.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:16:41,182 INFO [train2.py:809] (1/4) Epoch 24, batch 2200, loss[ctc_loss=0.0489, att_loss=0.2107, loss=0.1783, over 12692.00 frames. utt_duration=1815 frames, utt_pad_proportion=0.0406, over 28.00 utterances.], tot_loss[ctc_loss=0.06949, att_loss=0.2336, loss=0.2008, over 3271309.99 frames. utt_duration=1234 frames, utt_pad_proportion=0.05676, over 10618.60 utterances.], batch size: 28, lr: 4.47e-03, grad_scale: 8.0 2023-03-09 04:16:49,230 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93831.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:17:07,583 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.88 vs. limit=5.0 2023-03-09 04:17:16,341 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.265e+02 1.790e+02 2.187e+02 2.751e+02 8.387e+02, threshold=4.374e+02, percent-clipped=4.0 2023-03-09 04:18:02,864 INFO [train2.py:809] (1/4) Epoch 24, batch 2250, loss[ctc_loss=0.06787, att_loss=0.2308, loss=0.1982, over 16003.00 frames. utt_duration=1602 frames, utt_pad_proportion=0.007037, over 40.00 utterances.], tot_loss[ctc_loss=0.06946, att_loss=0.2338, loss=0.201, over 3269072.49 frames. utt_duration=1220 frames, utt_pad_proportion=0.06072, over 10729.24 utterances.], batch size: 40, lr: 4.47e-03, grad_scale: 8.0 2023-03-09 04:18:26,714 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6562, 3.0815, 3.8100, 3.4495, 3.6660, 4.6794, 4.4978, 3.4839], device='cuda:1'), covar=tensor([0.0295, 0.1628, 0.1074, 0.1027, 0.0975, 0.0932, 0.0530, 0.1040], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0246, 0.0283, 0.0217, 0.0266, 0.0368, 0.0265, 0.0232], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-09 04:19:24,949 INFO [train2.py:809] (1/4) Epoch 24, batch 2300, loss[ctc_loss=0.06403, att_loss=0.2391, loss=0.2041, over 16886.00 frames. utt_duration=1380 frames, utt_pad_proportion=0.006553, over 49.00 utterances.], tot_loss[ctc_loss=0.06931, att_loss=0.2333, loss=0.2005, over 3265424.40 frames. utt_duration=1223 frames, utt_pad_proportion=0.06245, over 10695.88 utterances.], batch size: 49, lr: 4.47e-03, grad_scale: 8.0 2023-03-09 04:19:50,410 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0545, 4.2946, 4.5697, 4.6657, 2.9432, 4.5319, 2.9638, 1.9518], device='cuda:1'), covar=tensor([0.0455, 0.0323, 0.0622, 0.0241, 0.1421, 0.0215, 0.1234, 0.1527], device='cuda:1'), in_proj_covar=tensor([0.0201, 0.0171, 0.0260, 0.0165, 0.0220, 0.0157, 0.0229, 0.0201], 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') 2023-03-09 04:20:00,417 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.133e+02 1.894e+02 2.401e+02 2.825e+02 6.941e+02, threshold=4.802e+02, percent-clipped=5.0 2023-03-09 04:20:47,685 INFO [train2.py:809] (1/4) Epoch 24, batch 2350, loss[ctc_loss=0.05226, att_loss=0.2206, loss=0.1869, over 16420.00 frames. utt_duration=1494 frames, utt_pad_proportion=0.005958, over 44.00 utterances.], tot_loss[ctc_loss=0.06943, att_loss=0.2336, loss=0.2008, over 3265976.90 frames. utt_duration=1193 frames, utt_pad_proportion=0.06977, over 10968.70 utterances.], batch size: 44, lr: 4.47e-03, grad_scale: 8.0 2023-03-09 04:20:54,645 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93981.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:22:13,388 INFO [train2.py:809] (1/4) Epoch 24, batch 2400, loss[ctc_loss=0.08907, att_loss=0.2555, loss=0.2222, over 17236.00 frames. utt_duration=874.4 frames, utt_pad_proportion=0.08152, over 79.00 utterances.], tot_loss[ctc_loss=0.07004, att_loss=0.2338, loss=0.2011, over 3264264.55 frames. utt_duration=1199 frames, utt_pad_proportion=0.06776, over 10905.17 utterances.], batch size: 79, lr: 4.47e-03, grad_scale: 8.0 2023-03-09 04:22:13,661 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9866, 5.2768, 4.8555, 5.3464, 4.7393, 5.0700, 5.4358, 5.1824], device='cuda:1'), covar=tensor([0.0590, 0.0243, 0.0758, 0.0283, 0.0409, 0.0231, 0.0231, 0.0222], device='cuda:1'), in_proj_covar=tensor([0.0386, 0.0321, 0.0362, 0.0350, 0.0322, 0.0238, 0.0304, 0.0285], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0005, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-09 04:22:38,930 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=94042.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:22:47,937 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.339e+02 1.881e+02 2.121e+02 2.677e+02 6.185e+02, threshold=4.243e+02, percent-clipped=1.0 2023-03-09 04:23:05,281 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6636, 5.1077, 4.9335, 4.9994, 5.1901, 4.7699, 3.8782, 5.1271], device='cuda:1'), covar=tensor([0.0126, 0.0106, 0.0133, 0.0083, 0.0100, 0.0133, 0.0562, 0.0180], device='cuda:1'), in_proj_covar=tensor([0.0093, 0.0089, 0.0111, 0.0070, 0.0076, 0.0087, 0.0103, 0.0108], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 04:23:34,727 INFO [train2.py:809] (1/4) Epoch 24, batch 2450, loss[ctc_loss=0.09245, att_loss=0.2664, loss=0.2316, over 17139.00 frames. utt_duration=1226 frames, utt_pad_proportion=0.01399, over 56.00 utterances.], tot_loss[ctc_loss=0.06977, att_loss=0.2338, loss=0.201, over 3261490.44 frames. utt_duration=1209 frames, utt_pad_proportion=0.06738, over 10804.99 utterances.], batch size: 56, lr: 4.47e-03, grad_scale: 8.0 2023-03-09 04:23:38,215 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.9488, 3.6632, 3.6723, 3.1949, 3.7587, 3.8327, 3.7777, 2.7110], device='cuda:1'), covar=tensor([0.0991, 0.1194, 0.2464, 0.2859, 0.0964, 0.1879, 0.0805, 0.3186], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0199, 0.0214, 0.0266, 0.0173, 0.0274, 0.0198, 0.0227], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 04:24:40,296 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.1269, 5.4171, 5.6749, 5.4079, 5.6457, 6.1104, 5.3323, 6.1325], device='cuda:1'), covar=tensor([0.0741, 0.0738, 0.0858, 0.1380, 0.1713, 0.0858, 0.0668, 0.0728], device='cuda:1'), in_proj_covar=tensor([0.0893, 0.0515, 0.0617, 0.0663, 0.0887, 0.0637, 0.0507, 0.0630], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 04:24:55,085 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94126.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:24:56,473 INFO [train2.py:809] (1/4) Epoch 24, batch 2500, loss[ctc_loss=0.06968, att_loss=0.2368, loss=0.2034, over 16474.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.007, over 46.00 utterances.], tot_loss[ctc_loss=0.06967, att_loss=0.2337, loss=0.2009, over 3264785.08 frames. utt_duration=1226 frames, utt_pad_proportion=0.06184, over 10661.46 utterances.], batch size: 46, lr: 4.47e-03, grad_scale: 8.0 2023-03-09 04:25:12,528 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7286, 3.4070, 3.9203, 3.3686, 3.6816, 4.7684, 4.5807, 3.5587], device='cuda:1'), covar=tensor([0.0350, 0.1417, 0.1096, 0.1166, 0.1106, 0.0895, 0.0636, 0.1035], device='cuda:1'), in_proj_covar=tensor([0.0245, 0.0245, 0.0285, 0.0219, 0.0267, 0.0369, 0.0266, 0.0233], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-09 04:25:31,025 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.204e+02 1.808e+02 2.198e+02 2.566e+02 4.413e+02, threshold=4.397e+02, percent-clipped=1.0 2023-03-09 04:26:18,034 INFO [train2.py:809] (1/4) Epoch 24, batch 2550, loss[ctc_loss=0.08134, att_loss=0.2503, loss=0.2165, over 17132.00 frames. utt_duration=1225 frames, utt_pad_proportion=0.01428, over 56.00 utterances.], tot_loss[ctc_loss=0.06981, att_loss=0.2332, loss=0.2005, over 3248059.15 frames. utt_duration=1209 frames, utt_pad_proportion=0.06957, over 10755.88 utterances.], batch size: 56, lr: 4.47e-03, grad_scale: 8.0 2023-03-09 04:26:45,526 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.87 vs. limit=2.0 2023-03-09 04:26:59,470 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.3673, 3.0391, 3.4068, 4.4475, 4.0022, 3.8707, 3.0115, 2.3462], device='cuda:1'), covar=tensor([0.0739, 0.1828, 0.0893, 0.0564, 0.0768, 0.0524, 0.1382, 0.2121], device='cuda:1'), in_proj_covar=tensor([0.0183, 0.0216, 0.0188, 0.0221, 0.0229, 0.0184, 0.0204, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 04:27:01,010 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7567, 5.1089, 5.0026, 5.0775, 5.1480, 4.8121, 3.6026, 5.0631], device='cuda:1'), covar=tensor([0.0108, 0.0127, 0.0121, 0.0098, 0.0131, 0.0119, 0.0680, 0.0234], device='cuda:1'), in_proj_covar=tensor([0.0094, 0.0090, 0.0112, 0.0070, 0.0076, 0.0088, 0.0104, 0.0108], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 04:27:32,997 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=94223.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 04:27:39,581 INFO [train2.py:809] (1/4) Epoch 24, batch 2600, loss[ctc_loss=0.06186, att_loss=0.231, loss=0.1972, over 16758.00 frames. utt_duration=1398 frames, utt_pad_proportion=0.007043, over 48.00 utterances.], tot_loss[ctc_loss=0.06992, att_loss=0.2333, loss=0.2006, over 3257263.23 frames. utt_duration=1220 frames, utt_pad_proportion=0.06576, over 10689.90 utterances.], batch size: 48, lr: 4.46e-03, grad_scale: 8.0 2023-03-09 04:28:14,364 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.213e+02 1.761e+02 2.111e+02 2.578e+02 4.836e+02, threshold=4.223e+02, percent-clipped=1.0 2023-03-09 04:29:01,213 INFO [train2.py:809] (1/4) Epoch 24, batch 2650, loss[ctc_loss=0.06853, att_loss=0.2286, loss=0.1966, over 15876.00 frames. utt_duration=1630 frames, utt_pad_proportion=0.009259, over 39.00 utterances.], tot_loss[ctc_loss=0.06943, att_loss=0.2335, loss=0.2007, over 3267620.80 frames. utt_duration=1257 frames, utt_pad_proportion=0.05507, over 10412.82 utterances.], batch size: 39, lr: 4.46e-03, grad_scale: 8.0 2023-03-09 04:29:14,085 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=94284.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 04:29:32,772 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0875, 6.2816, 5.7596, 6.0206, 5.9892, 5.4119, 5.7864, 5.5223], device='cuda:1'), covar=tensor([0.1173, 0.0891, 0.0827, 0.0798, 0.0780, 0.1778, 0.2343, 0.2330], device='cuda:1'), in_proj_covar=tensor([0.0537, 0.0617, 0.0472, 0.0467, 0.0435, 0.0474, 0.0627, 0.0534], 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') 2023-03-09 04:30:23,802 INFO [train2.py:809] (1/4) Epoch 24, batch 2700, loss[ctc_loss=0.0955, att_loss=0.2557, loss=0.2237, over 17370.00 frames. utt_duration=1104 frames, utt_pad_proportion=0.03453, over 63.00 utterances.], tot_loss[ctc_loss=0.06939, att_loss=0.2332, loss=0.2004, over 3265479.78 frames. utt_duration=1248 frames, utt_pad_proportion=0.05803, over 10476.28 utterances.], batch size: 63, lr: 4.46e-03, grad_scale: 8.0 2023-03-09 04:30:40,566 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94337.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:30:57,580 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.293e+02 1.818e+02 2.150e+02 2.630e+02 4.839e+02, threshold=4.300e+02, percent-clipped=3.0 2023-03-09 04:30:58,030 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=94348.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:31:45,413 INFO [train2.py:809] (1/4) Epoch 24, batch 2750, loss[ctc_loss=0.05636, att_loss=0.2204, loss=0.1876, over 16530.00 frames. utt_duration=1471 frames, utt_pad_proportion=0.006201, over 45.00 utterances.], tot_loss[ctc_loss=0.06994, att_loss=0.2334, loss=0.2007, over 3266094.01 frames. utt_duration=1234 frames, utt_pad_proportion=0.06145, over 10603.56 utterances.], batch size: 45, lr: 4.46e-03, grad_scale: 8.0 2023-03-09 04:32:38,161 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=94409.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 04:32:43,022 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2425, 3.8210, 3.2977, 3.6498, 4.0814, 3.7569, 3.2244, 4.3549], device='cuda:1'), covar=tensor([0.0921, 0.0582, 0.1003, 0.0615, 0.0686, 0.0692, 0.0806, 0.0443], device='cuda:1'), in_proj_covar=tensor([0.0203, 0.0221, 0.0225, 0.0204, 0.0282, 0.0245, 0.0201, 0.0290], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-09 04:33:05,980 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=94426.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:33:07,895 INFO [train2.py:809] (1/4) Epoch 24, batch 2800, loss[ctc_loss=0.05569, att_loss=0.2312, loss=0.1961, over 16948.00 frames. utt_duration=1357 frames, utt_pad_proportion=0.007792, over 50.00 utterances.], tot_loss[ctc_loss=0.06958, att_loss=0.2332, loss=0.2005, over 3258530.44 frames. utt_duration=1225 frames, utt_pad_proportion=0.06558, over 10657.06 utterances.], batch size: 50, lr: 4.46e-03, grad_scale: 8.0 2023-03-09 04:33:40,418 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.163e+02 1.912e+02 2.313e+02 2.793e+02 7.224e+02, threshold=4.627e+02, percent-clipped=4.0 2023-03-09 04:34:24,213 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=94474.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:34:28,879 INFO [train2.py:809] (1/4) Epoch 24, batch 2850, loss[ctc_loss=0.05205, att_loss=0.2093, loss=0.1778, over 15500.00 frames. utt_duration=1723 frames, utt_pad_proportion=0.008945, over 36.00 utterances.], tot_loss[ctc_loss=0.06964, att_loss=0.2334, loss=0.2007, over 3255360.90 frames. utt_duration=1214 frames, utt_pad_proportion=0.06775, over 10740.70 utterances.], batch size: 36, lr: 4.46e-03, grad_scale: 8.0 2023-03-09 04:35:11,761 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7808, 3.5169, 3.4741, 3.0520, 3.5930, 3.5565, 3.5587, 2.5319], device='cuda:1'), covar=tensor([0.1036, 0.1432, 0.1671, 0.3144, 0.1075, 0.2191, 0.0842, 0.3619], device='cuda:1'), in_proj_covar=tensor([0.0192, 0.0197, 0.0212, 0.0263, 0.0172, 0.0271, 0.0197, 0.0224], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 04:35:52,345 INFO [train2.py:809] (1/4) Epoch 24, batch 2900, loss[ctc_loss=0.06138, att_loss=0.2112, loss=0.1813, over 15894.00 frames. utt_duration=1631 frames, utt_pad_proportion=0.008864, over 39.00 utterances.], tot_loss[ctc_loss=0.06921, att_loss=0.233, loss=0.2003, over 3258068.02 frames. utt_duration=1206 frames, utt_pad_proportion=0.06918, over 10820.49 utterances.], batch size: 39, lr: 4.46e-03, grad_scale: 8.0 2023-03-09 04:36:26,073 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.225e+02 1.878e+02 2.321e+02 2.758e+02 4.740e+02, threshold=4.642e+02, percent-clipped=1.0 2023-03-09 04:37:14,352 INFO [train2.py:809] (1/4) Epoch 24, batch 2950, loss[ctc_loss=0.05894, att_loss=0.2122, loss=0.1815, over 15640.00 frames. utt_duration=1692 frames, utt_pad_proportion=0.008502, over 37.00 utterances.], tot_loss[ctc_loss=0.06935, att_loss=0.2332, loss=0.2004, over 3257187.40 frames. utt_duration=1219 frames, utt_pad_proportion=0.06643, over 10700.50 utterances.], batch size: 37, lr: 4.46e-03, grad_scale: 8.0 2023-03-09 04:37:17,800 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94579.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 04:37:29,504 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-03-09 04:38:27,842 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-03-09 04:38:36,142 INFO [train2.py:809] (1/4) Epoch 24, batch 3000, loss[ctc_loss=0.06915, att_loss=0.2149, loss=0.1858, over 15532.00 frames. utt_duration=1727 frames, utt_pad_proportion=0.006757, over 36.00 utterances.], tot_loss[ctc_loss=0.07021, att_loss=0.2331, loss=0.2005, over 3248252.85 frames. utt_duration=1184 frames, utt_pad_proportion=0.07843, over 10990.35 utterances.], batch size: 36, lr: 4.45e-03, grad_scale: 8.0 2023-03-09 04:38:36,142 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-09 04:38:50,630 INFO [train2.py:843] (1/4) Epoch 24, validation: ctc_loss=0.04165, att_loss=0.2345, loss=0.196, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-09 04:38:50,630 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-09 04:39:06,742 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=94637.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:39:16,363 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6168, 4.5779, 4.5551, 4.5871, 5.1476, 4.6422, 4.5962, 2.7966], device='cuda:1'), covar=tensor([0.0226, 0.0391, 0.0373, 0.0507, 0.0844, 0.0215, 0.0380, 0.1673], device='cuda:1'), in_proj_covar=tensor([0.0175, 0.0202, 0.0201, 0.0216, 0.0374, 0.0171, 0.0190, 0.0216], 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') 2023-03-09 04:39:23,558 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.235e+02 2.097e+02 2.380e+02 3.047e+02 5.871e+02, threshold=4.759e+02, percent-clipped=2.0 2023-03-09 04:40:11,574 INFO [train2.py:809] (1/4) Epoch 24, batch 3050, loss[ctc_loss=0.08279, att_loss=0.2443, loss=0.212, over 17336.00 frames. utt_duration=879.2 frames, utt_pad_proportion=0.07839, over 79.00 utterances.], tot_loss[ctc_loss=0.07043, att_loss=0.2338, loss=0.2012, over 3245280.63 frames. utt_duration=1160 frames, utt_pad_proportion=0.08369, over 11204.98 utterances.], batch size: 79, lr: 4.45e-03, grad_scale: 8.0 2023-03-09 04:40:24,110 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=94685.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:40:54,727 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94704.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 04:41:23,962 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.83 vs. limit=2.0 2023-03-09 04:41:25,011 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9768, 5.3094, 4.9338, 5.3762, 4.8162, 5.0716, 5.4938, 5.2332], device='cuda:1'), covar=tensor([0.0606, 0.0269, 0.0706, 0.0333, 0.0379, 0.0235, 0.0199, 0.0194], device='cuda:1'), in_proj_covar=tensor([0.0394, 0.0326, 0.0367, 0.0358, 0.0328, 0.0241, 0.0307, 0.0290], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-09 04:41:32,545 INFO [train2.py:809] (1/4) Epoch 24, batch 3100, loss[ctc_loss=0.0667, att_loss=0.2369, loss=0.2028, over 17140.00 frames. utt_duration=1226 frames, utt_pad_proportion=0.01375, over 56.00 utterances.], tot_loss[ctc_loss=0.07059, att_loss=0.2345, loss=0.2017, over 3262730.21 frames. utt_duration=1174 frames, utt_pad_proportion=0.07522, over 11130.06 utterances.], batch size: 56, lr: 4.45e-03, grad_scale: 8.0 2023-03-09 04:41:35,995 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0617, 5.3929, 5.0174, 5.4648, 4.8763, 5.1159, 5.5604, 5.3355], device='cuda:1'), covar=tensor([0.0553, 0.0309, 0.0705, 0.0332, 0.0378, 0.0206, 0.0197, 0.0185], device='cuda:1'), in_proj_covar=tensor([0.0394, 0.0326, 0.0367, 0.0358, 0.0329, 0.0241, 0.0308, 0.0290], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-09 04:42:05,406 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.429e+02 1.910e+02 2.247e+02 2.795e+02 4.738e+02, threshold=4.495e+02, percent-clipped=0.0 2023-03-09 04:42:12,990 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7944, 2.3966, 2.4933, 2.7790, 2.6060, 2.6103, 2.4780, 2.8575], device='cuda:1'), covar=tensor([0.1543, 0.2501, 0.1801, 0.1376, 0.1452, 0.1299, 0.2205, 0.1347], device='cuda:1'), in_proj_covar=tensor([0.0130, 0.0133, 0.0129, 0.0121, 0.0136, 0.0118, 0.0142, 0.0116], device='cuda:1'), out_proj_covar=tensor([9.9459e-05, 1.0453e-04, 1.0475e-04, 9.4678e-05, 1.0236e-04, 9.4975e-05, 1.0814e-04, 9.1759e-05], device='cuda:1') 2023-03-09 04:42:54,023 INFO [train2.py:809] (1/4) Epoch 24, batch 3150, loss[ctc_loss=0.09424, att_loss=0.2555, loss=0.2233, over 16959.00 frames. utt_duration=1358 frames, utt_pad_proportion=0.008093, over 50.00 utterances.], tot_loss[ctc_loss=0.07107, att_loss=0.2351, loss=0.2023, over 3268457.73 frames. utt_duration=1205 frames, utt_pad_proportion=0.06657, over 10860.43 utterances.], batch size: 50, lr: 4.45e-03, grad_scale: 8.0 2023-03-09 04:44:15,930 INFO [train2.py:809] (1/4) Epoch 24, batch 3200, loss[ctc_loss=0.07714, att_loss=0.2539, loss=0.2185, over 17372.00 frames. utt_duration=1105 frames, utt_pad_proportion=0.0319, over 63.00 utterances.], tot_loss[ctc_loss=0.06994, att_loss=0.2336, loss=0.2009, over 3261524.27 frames. utt_duration=1244 frames, utt_pad_proportion=0.06005, over 10500.23 utterances.], batch size: 63, lr: 4.45e-03, grad_scale: 8.0 2023-03-09 04:44:45,904 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.6233, 5.9307, 5.4249, 5.6658, 5.5640, 5.1527, 5.3173, 5.0317], device='cuda:1'), covar=tensor([0.1514, 0.1005, 0.0891, 0.0872, 0.1002, 0.1536, 0.2420, 0.2660], device='cuda:1'), in_proj_covar=tensor([0.0540, 0.0623, 0.0473, 0.0472, 0.0439, 0.0478, 0.0629, 0.0537], 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') 2023-03-09 04:44:48,826 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.290e+02 1.841e+02 2.191e+02 2.766e+02 6.449e+02, threshold=4.381e+02, percent-clipped=4.0 2023-03-09 04:45:36,259 INFO [train2.py:809] (1/4) Epoch 24, batch 3250, loss[ctc_loss=0.09115, att_loss=0.2523, loss=0.2201, over 17066.00 frames. utt_duration=1220 frames, utt_pad_proportion=0.0181, over 56.00 utterances.], tot_loss[ctc_loss=0.06963, att_loss=0.2332, loss=0.2005, over 3262720.49 frames. utt_duration=1235 frames, utt_pad_proportion=0.06207, over 10577.56 utterances.], batch size: 56, lr: 4.45e-03, grad_scale: 8.0 2023-03-09 04:45:36,583 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.3309, 5.2583, 5.1143, 3.4635, 5.0525, 4.8545, 4.5938, 3.0984], device='cuda:1'), covar=tensor([0.0096, 0.0098, 0.0234, 0.0794, 0.0096, 0.0183, 0.0281, 0.1192], device='cuda:1'), in_proj_covar=tensor([0.0077, 0.0105, 0.0108, 0.0114, 0.0088, 0.0117, 0.0101, 0.0105], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-09 04:45:39,720 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=94879.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 04:46:56,545 INFO [train2.py:809] (1/4) Epoch 24, batch 3300, loss[ctc_loss=0.07667, att_loss=0.219, loss=0.1905, over 15516.00 frames. utt_duration=1726 frames, utt_pad_proportion=0.007523, over 36.00 utterances.], tot_loss[ctc_loss=0.06954, att_loss=0.2332, loss=0.2004, over 3268159.28 frames. utt_duration=1230 frames, utt_pad_proportion=0.0615, over 10637.21 utterances.], batch size: 36, lr: 4.45e-03, grad_scale: 8.0 2023-03-09 04:46:56,643 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=94927.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 04:47:29,073 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.188e+02 1.857e+02 2.212e+02 2.661e+02 5.199e+02, threshold=4.424e+02, percent-clipped=2.0 2023-03-09 04:48:16,306 INFO [train2.py:809] (1/4) Epoch 24, batch 3350, loss[ctc_loss=0.0617, att_loss=0.2272, loss=0.1941, over 16546.00 frames. utt_duration=1472 frames, utt_pad_proportion=0.005927, over 45.00 utterances.], tot_loss[ctc_loss=0.06969, att_loss=0.2335, loss=0.2007, over 3265925.07 frames. utt_duration=1217 frames, utt_pad_proportion=0.06578, over 10743.52 utterances.], batch size: 45, lr: 4.45e-03, grad_scale: 8.0 2023-03-09 04:48:38,483 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5231, 2.5854, 4.9307, 3.9442, 3.0285, 4.2445, 4.7509, 4.7204], device='cuda:1'), covar=tensor([0.0267, 0.1489, 0.0234, 0.0800, 0.1671, 0.0278, 0.0189, 0.0249], device='cuda:1'), in_proj_covar=tensor([0.0211, 0.0241, 0.0204, 0.0317, 0.0265, 0.0223, 0.0193, 0.0218], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 04:48:59,861 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95004.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:49:36,977 INFO [train2.py:809] (1/4) Epoch 24, batch 3400, loss[ctc_loss=0.05385, att_loss=0.2166, loss=0.1841, over 15946.00 frames. utt_duration=1557 frames, utt_pad_proportion=0.006906, over 41.00 utterances.], tot_loss[ctc_loss=0.07001, att_loss=0.2335, loss=0.2008, over 3263611.65 frames. utt_duration=1205 frames, utt_pad_proportion=0.06858, over 10844.31 utterances.], batch size: 41, lr: 4.45e-03, grad_scale: 8.0 2023-03-09 04:50:06,906 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.7259, 5.9745, 5.4453, 5.7052, 5.6135, 5.2043, 5.3917, 5.1020], device='cuda:1'), covar=tensor([0.1190, 0.0891, 0.0906, 0.0820, 0.0975, 0.1398, 0.2155, 0.2396], device='cuda:1'), in_proj_covar=tensor([0.0543, 0.0625, 0.0475, 0.0475, 0.0443, 0.0481, 0.0631, 0.0541], 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') 2023-03-09 04:50:09,913 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.325e+02 1.881e+02 2.292e+02 2.864e+02 6.221e+02, threshold=4.584e+02, percent-clipped=2.0 2023-03-09 04:50:16,275 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=95052.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:50:56,264 INFO [train2.py:809] (1/4) Epoch 24, batch 3450, loss[ctc_loss=0.08611, att_loss=0.233, loss=0.2036, over 15882.00 frames. utt_duration=1630 frames, utt_pad_proportion=0.009534, over 39.00 utterances.], tot_loss[ctc_loss=0.07119, att_loss=0.2347, loss=0.202, over 3266855.06 frames. utt_duration=1204 frames, utt_pad_proportion=0.06698, over 10868.19 utterances.], batch size: 39, lr: 4.44e-03, grad_scale: 8.0 2023-03-09 04:51:04,932 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.6073, 3.5095, 3.4226, 3.8355, 2.7484, 3.7774, 2.8079, 2.1692], device='cuda:1'), covar=tensor([0.0466, 0.0380, 0.0743, 0.0292, 0.1285, 0.0274, 0.1168, 0.1404], device='cuda:1'), in_proj_covar=tensor([0.0204, 0.0175, 0.0263, 0.0168, 0.0223, 0.0161, 0.0232, 0.0206], 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') 2023-03-09 04:51:21,700 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-03-09 04:51:25,862 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4578, 2.8976, 3.6262, 2.9441, 3.4636, 4.5971, 4.4419, 3.2327], device='cuda:1'), covar=tensor([0.0371, 0.1843, 0.1193, 0.1411, 0.1149, 0.0663, 0.0482, 0.1287], device='cuda:1'), in_proj_covar=tensor([0.0249, 0.0247, 0.0286, 0.0221, 0.0269, 0.0372, 0.0267, 0.0234], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-09 04:51:41,236 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95105.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 04:51:44,213 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0935, 4.2931, 4.3763, 4.6370, 2.9239, 4.5310, 2.9474, 1.8378], device='cuda:1'), covar=tensor([0.0437, 0.0283, 0.0632, 0.0199, 0.1375, 0.0217, 0.1206, 0.1597], device='cuda:1'), in_proj_covar=tensor([0.0204, 0.0176, 0.0264, 0.0169, 0.0223, 0.0161, 0.0233, 0.0206], 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') 2023-03-09 04:52:16,795 INFO [train2.py:809] (1/4) Epoch 24, batch 3500, loss[ctc_loss=0.0795, att_loss=0.2611, loss=0.2248, over 17083.00 frames. utt_duration=1222 frames, utt_pad_proportion=0.01543, over 56.00 utterances.], tot_loss[ctc_loss=0.07105, att_loss=0.2351, loss=0.2023, over 3279065.05 frames. utt_duration=1222 frames, utt_pad_proportion=0.06027, over 10750.94 utterances.], batch size: 56, lr: 4.44e-03, grad_scale: 8.0 2023-03-09 04:52:49,702 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 9.240e+01 1.898e+02 2.336e+02 2.890e+02 6.917e+02, threshold=4.673e+02, percent-clipped=5.0 2023-03-09 04:53:18,846 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95166.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 04:53:35,722 INFO [train2.py:809] (1/4) Epoch 24, batch 3550, loss[ctc_loss=0.06515, att_loss=0.2172, loss=0.1868, over 15808.00 frames. utt_duration=1665 frames, utt_pad_proportion=0.006469, over 38.00 utterances.], tot_loss[ctc_loss=0.07068, att_loss=0.2344, loss=0.2017, over 3268703.97 frames. utt_duration=1228 frames, utt_pad_proportion=0.05893, over 10663.63 utterances.], batch size: 38, lr: 4.44e-03, grad_scale: 8.0 2023-03-09 04:54:14,461 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3798, 2.4186, 4.7723, 3.7131, 2.9305, 4.0435, 4.4680, 4.4990], device='cuda:1'), covar=tensor([0.0265, 0.1809, 0.0183, 0.1030, 0.1837, 0.0296, 0.0214, 0.0272], device='cuda:1'), in_proj_covar=tensor([0.0213, 0.0244, 0.0205, 0.0320, 0.0268, 0.0226, 0.0194, 0.0221], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 04:54:56,144 INFO [train2.py:809] (1/4) Epoch 24, batch 3600, loss[ctc_loss=0.07439, att_loss=0.239, loss=0.2061, over 16946.00 frames. utt_duration=1357 frames, utt_pad_proportion=0.007301, over 50.00 utterances.], tot_loss[ctc_loss=0.07089, att_loss=0.2346, loss=0.2018, over 3270161.92 frames. utt_duration=1222 frames, utt_pad_proportion=0.06006, over 10719.32 utterances.], batch size: 50, lr: 4.44e-03, grad_scale: 8.0 2023-03-09 04:55:10,297 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5105, 2.9614, 3.7057, 2.9280, 3.5329, 4.5983, 4.4491, 3.3685], device='cuda:1'), covar=tensor([0.0365, 0.1778, 0.1222, 0.1467, 0.1120, 0.0811, 0.0500, 0.1211], device='cuda:1'), in_proj_covar=tensor([0.0249, 0.0248, 0.0286, 0.0221, 0.0269, 0.0375, 0.0267, 0.0235], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-09 04:55:29,598 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.313e+02 1.952e+02 2.410e+02 2.837e+02 7.042e+02, threshold=4.821e+02, percent-clipped=2.0 2023-03-09 04:55:50,106 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95260.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:56:02,690 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95268.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:56:16,594 INFO [train2.py:809] (1/4) Epoch 24, batch 3650, loss[ctc_loss=0.07914, att_loss=0.2486, loss=0.2147, over 17398.00 frames. utt_duration=1106 frames, utt_pad_proportion=0.03297, over 63.00 utterances.], tot_loss[ctc_loss=0.07064, att_loss=0.2345, loss=0.2017, over 3272243.82 frames. utt_duration=1231 frames, utt_pad_proportion=0.05856, over 10642.39 utterances.], batch size: 63, lr: 4.44e-03, grad_scale: 8.0 2023-03-09 04:56:52,923 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-03-09 04:57:28,675 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95321.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:57:37,492 INFO [train2.py:809] (1/4) Epoch 24, batch 3700, loss[ctc_loss=0.05992, att_loss=0.2467, loss=0.2093, over 17010.00 frames. utt_duration=1336 frames, utt_pad_proportion=0.008472, over 51.00 utterances.], tot_loss[ctc_loss=0.07005, att_loss=0.2343, loss=0.2014, over 3271877.08 frames. utt_duration=1241 frames, utt_pad_proportion=0.05548, over 10562.92 utterances.], batch size: 51, lr: 4.44e-03, grad_scale: 8.0 2023-03-09 04:57:40,862 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95329.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:57:56,722 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95339.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 04:58:11,161 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.183e+02 1.823e+02 2.114e+02 2.643e+02 5.105e+02, threshold=4.228e+02, percent-clipped=2.0 2023-03-09 04:58:21,039 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.7635, 3.7049, 3.9456, 4.7437, 4.3277, 4.2357, 3.4235, 2.9375], device='cuda:1'), covar=tensor([0.0672, 0.1469, 0.0645, 0.0500, 0.0724, 0.0438, 0.1323, 0.1784], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0220, 0.0188, 0.0222, 0.0230, 0.0184, 0.0206, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 04:58:58,649 INFO [train2.py:809] (1/4) Epoch 24, batch 3750, loss[ctc_loss=0.05737, att_loss=0.2157, loss=0.1841, over 15906.00 frames. utt_duration=1633 frames, utt_pad_proportion=0.007851, over 39.00 utterances.], tot_loss[ctc_loss=0.06929, att_loss=0.2325, loss=0.1999, over 3252645.52 frames. utt_duration=1258 frames, utt_pad_proportion=0.05598, over 10355.07 utterances.], batch size: 39, lr: 4.44e-03, grad_scale: 8.0 2023-03-09 04:59:04,844 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5190, 4.8529, 4.8136, 4.9266, 4.9724, 4.6969, 3.7159, 4.8673], device='cuda:1'), covar=tensor([0.0132, 0.0120, 0.0131, 0.0091, 0.0093, 0.0111, 0.0589, 0.0182], device='cuda:1'), in_proj_covar=tensor([0.0095, 0.0092, 0.0115, 0.0071, 0.0078, 0.0089, 0.0105, 0.0110], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 04:59:04,900 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.1840, 2.9568, 3.2466, 4.3114, 3.8828, 3.8042, 2.8703, 2.0245], device='cuda:1'), covar=tensor([0.0877, 0.1874, 0.0901, 0.0613, 0.0898, 0.0520, 0.1654, 0.2411], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0221, 0.0188, 0.0222, 0.0230, 0.0184, 0.0207, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 04:59:27,724 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.86 vs. limit=2.0 2023-03-09 04:59:35,890 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95400.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:00:08,391 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7591, 2.5570, 2.2297, 2.4507, 2.7536, 2.4753, 2.4913, 2.9254], device='cuda:1'), covar=tensor([0.1470, 0.2167, 0.1788, 0.1443, 0.1435, 0.1211, 0.1920, 0.1063], device='cuda:1'), in_proj_covar=tensor([0.0130, 0.0133, 0.0130, 0.0121, 0.0137, 0.0118, 0.0142, 0.0116], device='cuda:1'), out_proj_covar=tensor([9.9293e-05, 1.0456e-04, 1.0484e-04, 9.5001e-05, 1.0286e-04, 9.5279e-05, 1.0778e-04, 9.2274e-05], device='cuda:1') 2023-03-09 05:00:19,271 INFO [train2.py:809] (1/4) Epoch 24, batch 3800, loss[ctc_loss=0.05917, att_loss=0.2245, loss=0.1914, over 15941.00 frames. utt_duration=1557 frames, utt_pad_proportion=0.007264, over 41.00 utterances.], tot_loss[ctc_loss=0.06925, att_loss=0.2329, loss=0.2002, over 3262646.39 frames. utt_duration=1261 frames, utt_pad_proportion=0.05306, over 10365.33 utterances.], batch size: 41, lr: 4.44e-03, grad_scale: 8.0 2023-03-09 05:00:21,714 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-03-09 05:00:52,296 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.438e+02 1.978e+02 2.351e+02 3.015e+02 8.965e+02, threshold=4.702e+02, percent-clipped=6.0 2023-03-09 05:01:13,136 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95461.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 05:01:33,310 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.22 vs. limit=5.0 2023-03-09 05:01:38,905 INFO [train2.py:809] (1/4) Epoch 24, batch 3850, loss[ctc_loss=0.08156, att_loss=0.2192, loss=0.1916, over 11837.00 frames. utt_duration=1823 frames, utt_pad_proportion=0.1649, over 26.00 utterances.], tot_loss[ctc_loss=0.06897, att_loss=0.2326, loss=0.1999, over 3264762.66 frames. utt_duration=1279 frames, utt_pad_proportion=0.04769, over 10219.38 utterances.], batch size: 26, lr: 4.43e-03, grad_scale: 8.0 2023-03-09 05:01:39,760 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-03-09 05:01:57,656 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-03-09 05:02:54,838 INFO [train2.py:809] (1/4) Epoch 24, batch 3900, loss[ctc_loss=0.07344, att_loss=0.2268, loss=0.1961, over 16113.00 frames. utt_duration=1536 frames, utt_pad_proportion=0.007003, over 42.00 utterances.], tot_loss[ctc_loss=0.06885, att_loss=0.2329, loss=0.2001, over 3272664.71 frames. utt_duration=1300 frames, utt_pad_proportion=0.04157, over 10080.28 utterances.], batch size: 42, lr: 4.43e-03, grad_scale: 8.0 2023-03-09 05:03:26,621 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.085e+02 1.865e+02 2.329e+02 2.791e+02 6.016e+02, threshold=4.658e+02, percent-clipped=3.0 2023-03-09 05:04:10,398 INFO [train2.py:809] (1/4) Epoch 24, batch 3950, loss[ctc_loss=0.05968, att_loss=0.2061, loss=0.1768, over 15367.00 frames. utt_duration=1758 frames, utt_pad_proportion=0.01136, over 35.00 utterances.], tot_loss[ctc_loss=0.06884, att_loss=0.2329, loss=0.2001, over 3277011.57 frames. utt_duration=1314 frames, utt_pad_proportion=0.03756, over 9987.47 utterances.], batch size: 35, lr: 4.43e-03, grad_scale: 8.0 2023-03-09 05:05:25,991 INFO [train2.py:809] (1/4) Epoch 25, batch 0, loss[ctc_loss=0.09573, att_loss=0.2511, loss=0.2201, over 16885.00 frames. utt_duration=1380 frames, utt_pad_proportion=0.007209, over 49.00 utterances.], tot_loss[ctc_loss=0.09573, att_loss=0.2511, loss=0.2201, over 16885.00 frames. utt_duration=1380 frames, utt_pad_proportion=0.007209, over 49.00 utterances.], batch size: 49, lr: 4.34e-03, grad_scale: 8.0 2023-03-09 05:05:25,992 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-09 05:05:38,247 INFO [train2.py:843] (1/4) Epoch 25, validation: ctc_loss=0.04004, att_loss=0.2344, loss=0.1955, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-09 05:05:38,248 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-09 05:05:46,417 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95616.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:05:48,108 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95617.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:05:58,887 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95624.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:06:36,946 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.264e+02 1.872e+02 2.311e+02 2.775e+02 5.959e+02, threshold=4.621e+02, percent-clipped=4.0 2023-03-09 05:06:54,909 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7893, 3.8188, 3.1825, 3.3131, 3.9993, 3.6955, 2.7053, 4.2843], device='cuda:1'), covar=tensor([0.1367, 0.0571, 0.1184, 0.0861, 0.0832, 0.0699, 0.1172, 0.0563], device='cuda:1'), in_proj_covar=tensor([0.0207, 0.0225, 0.0228, 0.0207, 0.0286, 0.0245, 0.0204, 0.0295], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-09 05:06:57,663 INFO [train2.py:809] (1/4) Epoch 25, batch 50, loss[ctc_loss=0.1162, att_loss=0.2605, loss=0.2316, over 14412.00 frames. utt_duration=396.4 frames, utt_pad_proportion=0.3106, over 146.00 utterances.], tot_loss[ctc_loss=0.0718, att_loss=0.2337, loss=0.2014, over 738358.59 frames. utt_duration=1267 frames, utt_pad_proportion=0.05244, over 2332.95 utterances.], batch size: 146, lr: 4.34e-03, grad_scale: 8.0 2023-03-09 05:07:25,603 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95678.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:07:51,591 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-03-09 05:07:52,061 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95695.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:08:17,157 INFO [train2.py:809] (1/4) Epoch 25, batch 100, loss[ctc_loss=0.0772, att_loss=0.2539, loss=0.2186, over 17117.00 frames. utt_duration=1224 frames, utt_pad_proportion=0.01413, over 56.00 utterances.], tot_loss[ctc_loss=0.0713, att_loss=0.2334, loss=0.201, over 1302435.95 frames. utt_duration=1265 frames, utt_pad_proportion=0.04774, over 4124.51 utterances.], batch size: 56, lr: 4.34e-03, grad_scale: 8.0 2023-03-09 05:09:15,972 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.395e+02 1.992e+02 2.344e+02 2.996e+02 8.490e+02, threshold=4.688e+02, percent-clipped=4.0 2023-03-09 05:09:36,921 INFO [train2.py:809] (1/4) Epoch 25, batch 150, loss[ctc_loss=0.08263, att_loss=0.2254, loss=0.1969, over 15769.00 frames. utt_duration=1661 frames, utt_pad_proportion=0.008809, over 38.00 utterances.], tot_loss[ctc_loss=0.07079, att_loss=0.2336, loss=0.201, over 1747584.82 frames. utt_duration=1291 frames, utt_pad_proportion=0.03782, over 5420.74 utterances.], batch size: 38, lr: 4.34e-03, grad_scale: 16.0 2023-03-09 05:09:37,267 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95761.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 05:10:02,056 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.2944, 5.4964, 5.4459, 5.4571, 5.5762, 5.5438, 5.2207, 5.0276], device='cuda:1'), covar=tensor([0.0990, 0.0530, 0.0279, 0.0476, 0.0254, 0.0290, 0.0360, 0.0308], device='cuda:1'), in_proj_covar=tensor([0.0529, 0.0371, 0.0359, 0.0371, 0.0432, 0.0438, 0.0367, 0.0400], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004], device='cuda:1') 2023-03-09 05:10:55,139 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=95809.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 05:10:58,462 INFO [train2.py:809] (1/4) Epoch 25, batch 200, loss[ctc_loss=0.068, att_loss=0.2127, loss=0.1837, over 15350.00 frames. utt_duration=1756 frames, utt_pad_proportion=0.01244, over 35.00 utterances.], tot_loss[ctc_loss=0.06969, att_loss=0.2339, loss=0.2011, over 2087608.99 frames. utt_duration=1283 frames, utt_pad_proportion=0.04294, over 6514.58 utterances.], batch size: 35, lr: 4.34e-03, grad_scale: 16.0 2023-03-09 05:11:00,457 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95812.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 05:11:55,507 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.82 vs. limit=2.0 2023-03-09 05:11:56,676 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95847.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:11:57,810 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.266e+02 1.872e+02 2.363e+02 3.018e+02 4.508e+02, threshold=4.726e+02, percent-clipped=0.0 2023-03-09 05:12:18,651 INFO [train2.py:809] (1/4) Epoch 25, batch 250, loss[ctc_loss=0.05006, att_loss=0.2094, loss=0.1775, over 15362.00 frames. utt_duration=1757 frames, utt_pad_proportion=0.0117, over 35.00 utterances.], tot_loss[ctc_loss=0.06938, att_loss=0.2336, loss=0.2007, over 2352133.21 frames. utt_duration=1265 frames, utt_pad_proportion=0.04869, over 7443.60 utterances.], batch size: 35, lr: 4.33e-03, grad_scale: 16.0 2023-03-09 05:12:38,740 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95873.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 05:13:34,971 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95908.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 05:13:39,390 INFO [train2.py:809] (1/4) Epoch 25, batch 300, loss[ctc_loss=0.06856, att_loss=0.2231, loss=0.1922, over 16538.00 frames. utt_duration=1471 frames, utt_pad_proportion=0.006482, over 45.00 utterances.], tot_loss[ctc_loss=0.06946, att_loss=0.2335, loss=0.2007, over 2555987.82 frames. utt_duration=1269 frames, utt_pad_proportion=0.04938, over 8066.46 utterances.], batch size: 45, lr: 4.33e-03, grad_scale: 16.0 2023-03-09 05:13:48,168 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95916.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:14:00,736 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95924.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:14:38,789 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.143e+02 1.855e+02 2.204e+02 2.784e+02 5.954e+02, threshold=4.408e+02, percent-clipped=3.0 2023-03-09 05:14:55,236 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0968, 5.3630, 5.3056, 5.2283, 5.4081, 5.3674, 5.0269, 4.8608], device='cuda:1'), covar=tensor([0.1049, 0.0518, 0.0294, 0.0536, 0.0275, 0.0343, 0.0424, 0.0319], device='cuda:1'), in_proj_covar=tensor([0.0532, 0.0373, 0.0359, 0.0372, 0.0433, 0.0440, 0.0368, 0.0401], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004], device='cuda:1') 2023-03-09 05:14:56,264 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-03-09 05:14:59,615 INFO [train2.py:809] (1/4) Epoch 25, batch 350, loss[ctc_loss=0.1281, att_loss=0.2668, loss=0.2391, over 13711.00 frames. utt_duration=379.8 frames, utt_pad_proportion=0.3406, over 145.00 utterances.], tot_loss[ctc_loss=0.06814, att_loss=0.2331, loss=0.2001, over 2720223.47 frames. utt_duration=1274 frames, utt_pad_proportion=0.0473, over 8551.88 utterances.], batch size: 145, lr: 4.33e-03, grad_scale: 16.0 2023-03-09 05:15:04,977 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=95964.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:15:08,200 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3843, 2.6986, 3.7240, 2.8010, 3.5460, 4.6415, 4.4756, 3.3144], device='cuda:1'), covar=tensor([0.0463, 0.2151, 0.1220, 0.1690, 0.1149, 0.0825, 0.0509, 0.1353], device='cuda:1'), in_proj_covar=tensor([0.0248, 0.0248, 0.0286, 0.0222, 0.0269, 0.0375, 0.0266, 0.0234], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-09 05:15:17,326 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=95972.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:15:19,649 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95973.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:15:30,461 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95980.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:15:54,754 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95995.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:16:08,725 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-03-09 05:16:17,998 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7001, 4.9687, 4.8020, 4.9816, 5.0489, 4.7349, 3.4887, 5.0924], device='cuda:1'), covar=tensor([0.0126, 0.0127, 0.0154, 0.0077, 0.0099, 0.0124, 0.0746, 0.0179], device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0092, 0.0115, 0.0072, 0.0079, 0.0089, 0.0106, 0.0111], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 05:16:24,358 INFO [train2.py:809] (1/4) Epoch 25, batch 400, loss[ctc_loss=0.09945, att_loss=0.2605, loss=0.2283, over 17043.00 frames. utt_duration=1288 frames, utt_pad_proportion=0.01018, over 53.00 utterances.], tot_loss[ctc_loss=0.06842, att_loss=0.2335, loss=0.2005, over 2845256.19 frames. utt_duration=1251 frames, utt_pad_proportion=0.05347, over 9111.25 utterances.], batch size: 53, lr: 4.33e-03, grad_scale: 16.0 2023-03-09 05:16:33,745 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.8564, 5.1541, 5.3897, 5.1588, 5.3025, 5.8039, 5.1725, 5.8972], device='cuda:1'), covar=tensor([0.0652, 0.0745, 0.0863, 0.1409, 0.1793, 0.0900, 0.0820, 0.0705], device='cuda:1'), in_proj_covar=tensor([0.0900, 0.0519, 0.0627, 0.0672, 0.0899, 0.0650, 0.0511, 0.0636], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 05:17:07,600 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9855, 5.0350, 4.8191, 2.8553, 4.8614, 4.7678, 4.2077, 2.7300], device='cuda:1'), covar=tensor([0.0114, 0.0097, 0.0254, 0.1075, 0.0098, 0.0180, 0.0336, 0.1356], device='cuda:1'), in_proj_covar=tensor([0.0077, 0.0105, 0.0107, 0.0113, 0.0088, 0.0116, 0.0101, 0.0104], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-09 05:17:12,324 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96041.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:17:15,164 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=96043.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:17:22,752 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.291e+02 1.820e+02 2.178e+02 2.552e+02 4.436e+02, threshold=4.357e+02, percent-clipped=1.0 2023-03-09 05:17:43,677 INFO [train2.py:809] (1/4) Epoch 25, batch 450, loss[ctc_loss=0.08786, att_loss=0.2558, loss=0.2222, over 16770.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.00573, over 48.00 utterances.], tot_loss[ctc_loss=0.06823, att_loss=0.2334, loss=0.2004, over 2945343.62 frames. utt_duration=1281 frames, utt_pad_proportion=0.04505, over 9210.15 utterances.], batch size: 48, lr: 4.33e-03, grad_scale: 16.0 2023-03-09 05:18:14,097 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96080.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:19:03,651 INFO [train2.py:809] (1/4) Epoch 25, batch 500, loss[ctc_loss=0.0774, att_loss=0.2391, loss=0.2067, over 17044.00 frames. utt_duration=1313 frames, utt_pad_proportion=0.009361, over 52.00 utterances.], tot_loss[ctc_loss=0.06839, att_loss=0.2332, loss=0.2002, over 3015303.28 frames. utt_duration=1256 frames, utt_pad_proportion=0.05104, over 9617.04 utterances.], batch size: 52, lr: 4.33e-03, grad_scale: 16.0 2023-03-09 05:19:45,969 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2043, 3.8023, 3.2845, 3.5415, 4.0580, 3.6855, 2.8766, 4.2856], device='cuda:1'), covar=tensor([0.0903, 0.0489, 0.1060, 0.0700, 0.0703, 0.0751, 0.0990, 0.0464], device='cuda:1'), in_proj_covar=tensor([0.0206, 0.0226, 0.0228, 0.0208, 0.0288, 0.0246, 0.0204, 0.0296], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-09 05:19:50,642 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96141.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:20:00,862 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.348e+02 1.791e+02 2.138e+02 2.685e+02 5.642e+02, threshold=4.276e+02, percent-clipped=2.0 2023-03-09 05:20:04,879 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-03-09 05:20:21,840 INFO [train2.py:809] (1/4) Epoch 25, batch 550, loss[ctc_loss=0.06298, att_loss=0.2196, loss=0.1883, over 15877.00 frames. utt_duration=1630 frames, utt_pad_proportion=0.009165, over 39.00 utterances.], tot_loss[ctc_loss=0.06904, att_loss=0.2335, loss=0.2006, over 3070223.90 frames. utt_duration=1231 frames, utt_pad_proportion=0.05751, over 9991.82 utterances.], batch size: 39, lr: 4.33e-03, grad_scale: 16.0 2023-03-09 05:20:28,364 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96165.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:20:32,703 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96168.0, num_to_drop=1, layers_to_drop={3} 2023-03-09 05:21:25,895 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0298, 5.3449, 4.7071, 5.4959, 4.8514, 5.0299, 5.5124, 5.1999], device='cuda:1'), covar=tensor([0.0634, 0.0333, 0.1064, 0.0309, 0.0440, 0.0299, 0.0267, 0.0234], device='cuda:1'), in_proj_covar=tensor([0.0397, 0.0330, 0.0374, 0.0362, 0.0330, 0.0244, 0.0314, 0.0293], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-09 05:21:27,358 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96203.0, num_to_drop=1, layers_to_drop={3} 2023-03-09 05:21:39,960 INFO [train2.py:809] (1/4) Epoch 25, batch 600, loss[ctc_loss=0.07156, att_loss=0.2494, loss=0.2138, over 17427.00 frames. utt_duration=1108 frames, utt_pad_proportion=0.03068, over 63.00 utterances.], tot_loss[ctc_loss=0.06909, att_loss=0.2333, loss=0.2005, over 3120157.40 frames. utt_duration=1230 frames, utt_pad_proportion=0.05658, over 10157.99 utterances.], batch size: 63, lr: 4.33e-03, grad_scale: 16.0 2023-03-09 05:22:03,767 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96226.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:22:38,438 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.174e+02 1.936e+02 2.176e+02 2.702e+02 4.909e+02, threshold=4.353e+02, percent-clipped=2.0 2023-03-09 05:22:59,842 INFO [train2.py:809] (1/4) Epoch 25, batch 650, loss[ctc_loss=0.066, att_loss=0.2485, loss=0.212, over 16882.00 frames. utt_duration=1380 frames, utt_pad_proportion=0.006803, over 49.00 utterances.], tot_loss[ctc_loss=0.06924, att_loss=0.2333, loss=0.2005, over 3150566.65 frames. utt_duration=1244 frames, utt_pad_proportion=0.05561, over 10141.43 utterances.], batch size: 49, lr: 4.33e-03, grad_scale: 16.0 2023-03-09 05:23:19,248 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96273.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:24:19,880 INFO [train2.py:809] (1/4) Epoch 25, batch 700, loss[ctc_loss=0.0648, att_loss=0.2413, loss=0.206, over 17328.00 frames. utt_duration=1102 frames, utt_pad_proportion=0.03787, over 63.00 utterances.], tot_loss[ctc_loss=0.06854, att_loss=0.2329, loss=0.2, over 3175563.92 frames. utt_duration=1248 frames, utt_pad_proportion=0.05547, over 10190.09 utterances.], batch size: 63, lr: 4.32e-03, grad_scale: 16.0 2023-03-09 05:24:36,214 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=96321.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:24:40,967 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96324.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:24:59,808 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96336.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:25:17,120 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9814, 4.2939, 4.2532, 4.6524, 2.6171, 4.4277, 2.5302, 1.7307], device='cuda:1'), covar=tensor([0.0431, 0.0254, 0.0728, 0.0216, 0.1722, 0.0223, 0.1562, 0.1747], device='cuda:1'), in_proj_covar=tensor([0.0204, 0.0176, 0.0263, 0.0169, 0.0223, 0.0160, 0.0231, 0.0204], 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') 2023-03-09 05:25:18,101 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.169e+02 1.804e+02 2.150e+02 2.706e+02 4.503e+02, threshold=4.300e+02, percent-clipped=0.0 2023-03-09 05:25:38,748 INFO [train2.py:809] (1/4) Epoch 25, batch 750, loss[ctc_loss=0.1208, att_loss=0.2723, loss=0.242, over 14044.00 frames. utt_duration=386.2 frames, utt_pad_proportion=0.3272, over 146.00 utterances.], tot_loss[ctc_loss=0.06933, att_loss=0.2343, loss=0.2013, over 3205577.52 frames. utt_duration=1242 frames, utt_pad_proportion=0.05467, over 10333.20 utterances.], batch size: 146, lr: 4.32e-03, grad_scale: 16.0 2023-03-09 05:26:17,531 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96385.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:26:18,010 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-03-09 05:26:25,361 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96390.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:26:58,845 INFO [train2.py:809] (1/4) Epoch 25, batch 800, loss[ctc_loss=0.06442, att_loss=0.2382, loss=0.2034, over 16959.00 frames. utt_duration=1358 frames, utt_pad_proportion=0.00731, over 50.00 utterances.], tot_loss[ctc_loss=0.0699, att_loss=0.2344, loss=0.2015, over 3209422.84 frames. utt_duration=1213 frames, utt_pad_proportion=0.06562, over 10595.19 utterances.], batch size: 50, lr: 4.32e-03, grad_scale: 16.0 2023-03-09 05:27:38,850 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96436.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:27:57,436 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.189e+02 1.925e+02 2.269e+02 2.805e+02 5.385e+02, threshold=4.538e+02, percent-clipped=5.0 2023-03-09 05:28:02,603 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96451.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:28:19,550 INFO [train2.py:809] (1/4) Epoch 25, batch 850, loss[ctc_loss=0.1118, att_loss=0.269, loss=0.2376, over 14535.00 frames. utt_duration=402.5 frames, utt_pad_proportion=0.3001, over 145.00 utterances.], tot_loss[ctc_loss=0.06968, att_loss=0.233, loss=0.2004, over 3213399.84 frames. utt_duration=1214 frames, utt_pad_proportion=0.06741, over 10598.63 utterances.], batch size: 145, lr: 4.32e-03, grad_scale: 16.0 2023-03-09 05:28:31,887 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96468.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 05:29:00,748 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0709, 5.0886, 4.9101, 3.0597, 4.8931, 4.6380, 4.3745, 2.9112], device='cuda:1'), covar=tensor([0.0139, 0.0108, 0.0257, 0.0915, 0.0098, 0.0215, 0.0301, 0.1230], device='cuda:1'), in_proj_covar=tensor([0.0078, 0.0105, 0.0108, 0.0114, 0.0088, 0.0117, 0.0102, 0.0105], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-09 05:29:28,036 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96503.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:29:41,365 INFO [train2.py:809] (1/4) Epoch 25, batch 900, loss[ctc_loss=0.07294, att_loss=0.2585, loss=0.2214, over 17144.00 frames. utt_duration=1226 frames, utt_pad_proportion=0.01368, over 56.00 utterances.], tot_loss[ctc_loss=0.07019, att_loss=0.2338, loss=0.2011, over 3228810.22 frames. utt_duration=1215 frames, utt_pad_proportion=0.0668, over 10644.87 utterances.], batch size: 56, lr: 4.32e-03, grad_scale: 16.0 2023-03-09 05:29:49,299 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=96516.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 05:29:57,211 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96521.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:30:40,145 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.401e+02 1.959e+02 2.352e+02 2.820e+02 9.770e+02, threshold=4.704e+02, percent-clipped=4.0 2023-03-09 05:30:45,077 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=96551.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:31:03,283 INFO [train2.py:809] (1/4) Epoch 25, batch 950, loss[ctc_loss=0.0532, att_loss=0.2363, loss=0.1997, over 16469.00 frames. utt_duration=1433 frames, utt_pad_proportion=0.006599, over 46.00 utterances.], tot_loss[ctc_loss=0.06979, att_loss=0.2336, loss=0.2008, over 3229397.26 frames. utt_duration=1241 frames, utt_pad_proportion=0.0596, over 10420.32 utterances.], batch size: 46, lr: 4.32e-03, grad_scale: 16.0 2023-03-09 05:32:25,901 INFO [train2.py:809] (1/4) Epoch 25, batch 1000, loss[ctc_loss=0.04974, att_loss=0.2335, loss=0.1967, over 16476.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.006684, over 46.00 utterances.], tot_loss[ctc_loss=0.06952, att_loss=0.2336, loss=0.2008, over 3244983.06 frames. utt_duration=1243 frames, utt_pad_proportion=0.05614, over 10457.71 utterances.], batch size: 46, lr: 4.32e-03, grad_scale: 16.0 2023-03-09 05:33:05,503 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96636.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:33:23,588 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.244e+02 1.809e+02 2.155e+02 2.476e+02 5.052e+02, threshold=4.310e+02, percent-clipped=2.0 2023-03-09 05:33:46,052 INFO [train2.py:809] (1/4) Epoch 25, batch 1050, loss[ctc_loss=0.06651, att_loss=0.2477, loss=0.2114, over 17048.00 frames. utt_duration=1313 frames, utt_pad_proportion=0.008926, over 52.00 utterances.], tot_loss[ctc_loss=0.06829, att_loss=0.2325, loss=0.1996, over 3244085.03 frames. utt_duration=1246 frames, utt_pad_proportion=0.05725, over 10426.54 utterances.], batch size: 52, lr: 4.32e-03, grad_scale: 16.0 2023-03-09 05:34:02,011 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7990, 5.1097, 4.6909, 5.1792, 4.5264, 4.8802, 5.2531, 5.0203], device='cuda:1'), covar=tensor([0.0647, 0.0343, 0.0888, 0.0324, 0.0476, 0.0256, 0.0233, 0.0216], device='cuda:1'), in_proj_covar=tensor([0.0396, 0.0330, 0.0371, 0.0362, 0.0330, 0.0242, 0.0312, 0.0291], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-09 05:34:17,076 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96680.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:34:23,048 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=96684.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:35:06,973 INFO [train2.py:809] (1/4) Epoch 25, batch 1100, loss[ctc_loss=0.05561, att_loss=0.217, loss=0.1847, over 15909.00 frames. utt_duration=1633 frames, utt_pad_proportion=0.007664, over 39.00 utterances.], tot_loss[ctc_loss=0.06808, att_loss=0.2324, loss=0.1996, over 3249740.33 frames. utt_duration=1261 frames, utt_pad_proportion=0.05389, over 10320.62 utterances.], batch size: 39, lr: 4.32e-03, grad_scale: 16.0 2023-03-09 05:35:47,551 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96736.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:36:03,168 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96746.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:36:06,427 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.271e+02 2.003e+02 2.274e+02 2.730e+02 5.677e+02, threshold=4.548e+02, percent-clipped=3.0 2023-03-09 05:36:28,283 INFO [train2.py:809] (1/4) Epoch 25, batch 1150, loss[ctc_loss=0.07407, att_loss=0.2408, loss=0.2074, over 17296.00 frames. utt_duration=1174 frames, utt_pad_proportion=0.02483, over 59.00 utterances.], tot_loss[ctc_loss=0.0686, att_loss=0.2329, loss=0.2, over 3257470.07 frames. utt_duration=1228 frames, utt_pad_proportion=0.0592, over 10620.42 utterances.], batch size: 59, lr: 4.31e-03, grad_scale: 16.0 2023-03-09 05:36:52,989 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96776.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:37:05,055 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=96784.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:37:48,671 INFO [train2.py:809] (1/4) Epoch 25, batch 1200, loss[ctc_loss=0.06927, att_loss=0.2464, loss=0.211, over 17285.00 frames. utt_duration=1259 frames, utt_pad_proportion=0.01272, over 55.00 utterances.], tot_loss[ctc_loss=0.06914, att_loss=0.2331, loss=0.2003, over 3258756.44 frames. utt_duration=1218 frames, utt_pad_proportion=0.0627, over 10711.22 utterances.], batch size: 55, lr: 4.31e-03, grad_scale: 16.0 2023-03-09 05:38:03,326 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-03-09 05:38:05,773 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96821.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:38:30,744 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96837.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:38:48,488 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.354e+02 2.088e+02 2.358e+02 2.925e+02 6.715e+02, threshold=4.715e+02, percent-clipped=3.0 2023-03-09 05:39:08,869 INFO [train2.py:809] (1/4) Epoch 25, batch 1250, loss[ctc_loss=0.07048, att_loss=0.2351, loss=0.2022, over 16384.00 frames. utt_duration=1491 frames, utt_pad_proportion=0.008719, over 44.00 utterances.], tot_loss[ctc_loss=0.06909, att_loss=0.2331, loss=0.2003, over 3264080.51 frames. utt_duration=1239 frames, utt_pad_proportion=0.05741, over 10550.82 utterances.], batch size: 44, lr: 4.31e-03, grad_scale: 16.0 2023-03-09 05:39:22,088 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=96869.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:39:43,647 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9762, 5.2781, 5.4813, 5.3316, 5.4653, 5.9207, 5.2707, 6.0137], device='cuda:1'), covar=tensor([0.0684, 0.0713, 0.0818, 0.1249, 0.1770, 0.0949, 0.0639, 0.0628], device='cuda:1'), in_proj_covar=tensor([0.0903, 0.0521, 0.0631, 0.0675, 0.0903, 0.0654, 0.0514, 0.0640], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 05:40:28,120 INFO [train2.py:809] (1/4) Epoch 25, batch 1300, loss[ctc_loss=0.05649, att_loss=0.2142, loss=0.1827, over 15785.00 frames. utt_duration=1663 frames, utt_pad_proportion=0.007678, over 38.00 utterances.], tot_loss[ctc_loss=0.06825, att_loss=0.2322, loss=0.1994, over 3258921.34 frames. utt_duration=1246 frames, utt_pad_proportion=0.05751, over 10473.73 utterances.], batch size: 38, lr: 4.31e-03, grad_scale: 16.0 2023-03-09 05:41:27,106 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.103e+02 1.796e+02 2.102e+02 2.487e+02 3.780e+02, threshold=4.203e+02, percent-clipped=0.0 2023-03-09 05:41:47,819 INFO [train2.py:809] (1/4) Epoch 25, batch 1350, loss[ctc_loss=0.05987, att_loss=0.2388, loss=0.203, over 16480.00 frames. utt_duration=1435 frames, utt_pad_proportion=0.005845, over 46.00 utterances.], tot_loss[ctc_loss=0.06794, att_loss=0.2326, loss=0.1997, over 3266159.08 frames. utt_duration=1237 frames, utt_pad_proportion=0.05648, over 10571.59 utterances.], batch size: 46, lr: 4.31e-03, grad_scale: 16.0 2023-03-09 05:42:18,041 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96980.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:42:30,053 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.1577, 3.9399, 3.3435, 3.4480, 4.0960, 3.7910, 3.1204, 4.3318], device='cuda:1'), covar=tensor([0.0998, 0.0489, 0.0970, 0.0737, 0.0721, 0.0645, 0.0901, 0.0557], device='cuda:1'), in_proj_covar=tensor([0.0206, 0.0224, 0.0227, 0.0206, 0.0287, 0.0245, 0.0202, 0.0294], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-09 05:42:51,024 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97000.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:43:07,411 INFO [train2.py:809] (1/4) Epoch 25, batch 1400, loss[ctc_loss=0.05109, att_loss=0.2202, loss=0.1864, over 15953.00 frames. utt_duration=1558 frames, utt_pad_proportion=0.00712, over 41.00 utterances.], tot_loss[ctc_loss=0.06738, att_loss=0.2319, loss=0.199, over 3267661.31 frames. utt_duration=1279 frames, utt_pad_proportion=0.04771, over 10234.35 utterances.], batch size: 41, lr: 4.31e-03, grad_scale: 16.0 2023-03-09 05:43:34,010 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=97028.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:44:02,822 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97046.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:44:05,388 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.189e+02 1.953e+02 2.309e+02 2.789e+02 6.785e+02, threshold=4.619e+02, percent-clipped=7.0 2023-03-09 05:44:26,317 INFO [train2.py:809] (1/4) Epoch 25, batch 1450, loss[ctc_loss=0.04312, att_loss=0.2053, loss=0.1729, over 15876.00 frames. utt_duration=1630 frames, utt_pad_proportion=0.009119, over 39.00 utterances.], tot_loss[ctc_loss=0.06694, att_loss=0.2317, loss=0.1987, over 3274493.51 frames. utt_duration=1287 frames, utt_pad_proportion=0.0436, over 10191.26 utterances.], batch size: 39, lr: 4.31e-03, grad_scale: 16.0 2023-03-09 05:44:26,719 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97061.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:45:09,849 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9399, 4.8920, 4.9820, 2.1892, 1.9599, 2.5264, 2.2423, 3.6284], device='cuda:1'), covar=tensor([0.0895, 0.0388, 0.0252, 0.4480, 0.6221, 0.3356, 0.4187, 0.1985], device='cuda:1'), in_proj_covar=tensor([0.0361, 0.0290, 0.0274, 0.0249, 0.0339, 0.0334, 0.0261, 0.0372], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-09 05:45:17,677 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=97094.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:45:44,475 INFO [train2.py:809] (1/4) Epoch 25, batch 1500, loss[ctc_loss=0.05716, att_loss=0.2346, loss=0.1991, over 16766.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.006673, over 48.00 utterances.], tot_loss[ctc_loss=0.06696, att_loss=0.232, loss=0.199, over 3271986.66 frames. utt_duration=1275 frames, utt_pad_proportion=0.04899, over 10280.72 utterances.], batch size: 48, lr: 4.31e-03, grad_scale: 16.0 2023-03-09 05:46:17,396 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97132.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:46:43,662 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.190e+02 2.011e+02 2.273e+02 2.604e+02 5.523e+02, threshold=4.546e+02, percent-clipped=2.0 2023-03-09 05:46:47,057 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4748, 4.4972, 4.6090, 4.5948, 5.1441, 4.4227, 4.5254, 2.5379], device='cuda:1'), covar=tensor([0.0252, 0.0361, 0.0306, 0.0309, 0.0632, 0.0268, 0.0334, 0.1856], device='cuda:1'), in_proj_covar=tensor([0.0178, 0.0205, 0.0203, 0.0221, 0.0379, 0.0177, 0.0194, 0.0220], 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') 2023-03-09 05:47:03,718 INFO [train2.py:809] (1/4) Epoch 25, batch 1550, loss[ctc_loss=0.04831, att_loss=0.2005, loss=0.17, over 14509.00 frames. utt_duration=1815 frames, utt_pad_proportion=0.04003, over 32.00 utterances.], tot_loss[ctc_loss=0.06726, att_loss=0.2317, loss=0.1988, over 3263360.12 frames. utt_duration=1260 frames, utt_pad_proportion=0.05459, over 10373.68 utterances.], batch size: 32, lr: 4.31e-03, grad_scale: 16.0 2023-03-09 05:47:44,745 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.8459, 3.4574, 3.4774, 2.8744, 3.4884, 3.5448, 3.6039, 2.4103], device='cuda:1'), covar=tensor([0.1182, 0.1369, 0.1928, 0.3870, 0.1579, 0.2253, 0.0981, 0.4155], device='cuda:1'), in_proj_covar=tensor([0.0192, 0.0200, 0.0213, 0.0267, 0.0175, 0.0275, 0.0199, 0.0225], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 05:48:24,493 INFO [train2.py:809] (1/4) Epoch 25, batch 1600, loss[ctc_loss=0.07218, att_loss=0.2425, loss=0.2084, over 16818.00 frames. utt_duration=681 frames, utt_pad_proportion=0.1445, over 99.00 utterances.], tot_loss[ctc_loss=0.06743, att_loss=0.2318, loss=0.1989, over 3266710.32 frames. utt_duration=1260 frames, utt_pad_proportion=0.05287, over 10379.13 utterances.], batch size: 99, lr: 4.30e-03, grad_scale: 16.0 2023-03-09 05:49:19,401 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0363, 5.0475, 4.8065, 3.1493, 4.7915, 4.7067, 4.3154, 2.6153], device='cuda:1'), covar=tensor([0.0129, 0.0117, 0.0296, 0.0944, 0.0115, 0.0189, 0.0328, 0.1424], device='cuda:1'), in_proj_covar=tensor([0.0076, 0.0103, 0.0106, 0.0110, 0.0086, 0.0113, 0.0099, 0.0102], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-09 05:49:22,641 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.291e+02 1.951e+02 2.275e+02 2.852e+02 6.769e+02, threshold=4.549e+02, percent-clipped=7.0 2023-03-09 05:49:41,681 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97260.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:49:43,460 INFO [train2.py:809] (1/4) Epoch 25, batch 1650, loss[ctc_loss=0.07321, att_loss=0.2383, loss=0.2053, over 16538.00 frames. utt_duration=1471 frames, utt_pad_proportion=0.005916, over 45.00 utterances.], tot_loss[ctc_loss=0.06821, att_loss=0.2326, loss=0.1997, over 3268920.31 frames. utt_duration=1265 frames, utt_pad_proportion=0.05087, over 10346.31 utterances.], batch size: 45, lr: 4.30e-03, grad_scale: 16.0 2023-03-09 05:51:02,810 INFO [train2.py:809] (1/4) Epoch 25, batch 1700, loss[ctc_loss=0.0878, att_loss=0.2633, loss=0.2282, over 17287.00 frames. utt_duration=1259 frames, utt_pad_proportion=0.01268, over 55.00 utterances.], tot_loss[ctc_loss=0.06808, att_loss=0.233, loss=0.2, over 3281557.58 frames. utt_duration=1284 frames, utt_pad_proportion=0.04368, over 10233.47 utterances.], batch size: 55, lr: 4.30e-03, grad_scale: 16.0 2023-03-09 05:51:19,393 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97321.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:51:51,943 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97342.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:52:01,029 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97347.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:52:02,097 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.222e+02 1.753e+02 2.134e+02 2.732e+02 7.784e+02, threshold=4.268e+02, percent-clipped=3.0 2023-03-09 05:52:15,772 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97356.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:52:23,284 INFO [train2.py:809] (1/4) Epoch 25, batch 1750, loss[ctc_loss=0.06047, att_loss=0.242, loss=0.2057, over 16887.00 frames. utt_duration=1380 frames, utt_pad_proportion=0.006553, over 49.00 utterances.], tot_loss[ctc_loss=0.06805, att_loss=0.2334, loss=0.2003, over 3280675.85 frames. utt_duration=1280 frames, utt_pad_proportion=0.04589, over 10268.19 utterances.], batch size: 49, lr: 4.30e-03, grad_scale: 16.0 2023-03-09 05:52:38,897 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-03-09 05:52:44,669 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97374.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:53:06,381 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5708, 4.6102, 4.7688, 4.6786, 5.2830, 4.5357, 4.7287, 2.7382], device='cuda:1'), covar=tensor([0.0270, 0.0362, 0.0298, 0.0389, 0.0663, 0.0284, 0.0306, 0.1640], device='cuda:1'), in_proj_covar=tensor([0.0180, 0.0206, 0.0204, 0.0222, 0.0381, 0.0178, 0.0195, 0.0220], 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') 2023-03-09 05:53:17,756 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-03-09 05:53:31,684 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97403.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 05:53:34,185 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-03-09 05:53:39,257 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97408.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:53:44,077 INFO [train2.py:809] (1/4) Epoch 25, batch 1800, loss[ctc_loss=0.07359, att_loss=0.2386, loss=0.2056, over 17119.00 frames. utt_duration=1224 frames, utt_pad_proportion=0.01511, over 56.00 utterances.], tot_loss[ctc_loss=0.06851, att_loss=0.234, loss=0.2009, over 3281511.30 frames. utt_duration=1250 frames, utt_pad_proportion=0.0498, over 10509.35 utterances.], batch size: 56, lr: 4.30e-03, grad_scale: 16.0 2023-03-09 05:54:03,195 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.5425, 3.0227, 3.4957, 4.4777, 4.0178, 4.0018, 3.0480, 2.6193], device='cuda:1'), covar=tensor([0.0648, 0.1831, 0.0826, 0.0577, 0.0994, 0.0494, 0.1452, 0.1877], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0220, 0.0187, 0.0221, 0.0231, 0.0185, 0.0206, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 05:54:14,432 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-03-09 05:54:16,760 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97432.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:54:21,368 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97435.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:54:43,146 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.137e+02 1.842e+02 2.140e+02 2.481e+02 6.088e+02, threshold=4.279e+02, percent-clipped=3.0 2023-03-09 05:54:56,596 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-03-09 05:55:03,731 INFO [train2.py:809] (1/4) Epoch 25, batch 1850, loss[ctc_loss=0.08109, att_loss=0.2489, loss=0.2153, over 17090.00 frames. utt_duration=1291 frames, utt_pad_proportion=0.007527, over 53.00 utterances.], tot_loss[ctc_loss=0.06742, att_loss=0.2334, loss=0.2002, over 3276010.10 frames. utt_duration=1250 frames, utt_pad_proportion=0.05052, over 10499.05 utterances.], batch size: 53, lr: 4.30e-03, grad_scale: 16.0 2023-03-09 05:55:33,361 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=97480.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:55:36,123 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-03-09 05:55:46,828 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0898, 4.3305, 4.4456, 4.6380, 2.8153, 4.5269, 2.7704, 1.6446], device='cuda:1'), covar=tensor([0.0488, 0.0271, 0.0593, 0.0265, 0.1545, 0.0196, 0.1409, 0.1742], device='cuda:1'), in_proj_covar=tensor([0.0204, 0.0175, 0.0260, 0.0168, 0.0220, 0.0160, 0.0228, 0.0202], 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') 2023-03-09 05:56:23,873 INFO [train2.py:809] (1/4) Epoch 25, batch 1900, loss[ctc_loss=0.07284, att_loss=0.2513, loss=0.2156, over 17065.00 frames. utt_duration=1289 frames, utt_pad_proportion=0.007437, over 53.00 utterances.], tot_loss[ctc_loss=0.06771, att_loss=0.2335, loss=0.2003, over 3276611.82 frames. utt_duration=1249 frames, utt_pad_proportion=0.05137, over 10506.16 utterances.], batch size: 53, lr: 4.30e-03, grad_scale: 16.0 2023-03-09 05:57:23,539 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.351e+02 1.849e+02 2.257e+02 2.657e+02 5.129e+02, threshold=4.513e+02, percent-clipped=1.0 2023-03-09 05:57:35,261 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97555.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:57:44,857 INFO [train2.py:809] (1/4) Epoch 25, batch 1950, loss[ctc_loss=0.07669, att_loss=0.2502, loss=0.2155, over 17408.00 frames. utt_duration=1011 frames, utt_pad_proportion=0.04744, over 69.00 utterances.], tot_loss[ctc_loss=0.06736, att_loss=0.2328, loss=0.1997, over 3275076.64 frames. utt_duration=1273 frames, utt_pad_proportion=0.0456, over 10306.60 utterances.], batch size: 69, lr: 4.30e-03, grad_scale: 16.0 2023-03-09 05:58:08,865 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.1191, 3.6512, 3.1582, 3.3840, 3.9608, 3.5490, 3.0116, 4.1660], device='cuda:1'), covar=tensor([0.0978, 0.0518, 0.1105, 0.0734, 0.0752, 0.0712, 0.0874, 0.0547], device='cuda:1'), in_proj_covar=tensor([0.0208, 0.0226, 0.0230, 0.0208, 0.0290, 0.0248, 0.0205, 0.0297], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-09 05:58:29,322 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-03-09 05:58:49,638 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-03-09 05:59:06,992 INFO [train2.py:809] (1/4) Epoch 25, batch 2000, loss[ctc_loss=0.06984, att_loss=0.2412, loss=0.2069, over 16622.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.005379, over 47.00 utterances.], tot_loss[ctc_loss=0.06712, att_loss=0.2328, loss=0.1997, over 3274227.68 frames. utt_duration=1286 frames, utt_pad_proportion=0.04337, over 10193.05 utterances.], batch size: 47, lr: 4.30e-03, grad_scale: 16.0 2023-03-09 05:59:15,366 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97616.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 05:59:15,563 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97616.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:00:08,128 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.275e+02 1.752e+02 2.147e+02 2.537e+02 5.924e+02, threshold=4.294e+02, percent-clipped=3.0 2023-03-09 06:00:21,287 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97656.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:00:29,815 INFO [train2.py:809] (1/4) Epoch 25, batch 2050, loss[ctc_loss=0.07321, att_loss=0.2357, loss=0.2032, over 16541.00 frames. utt_duration=1472 frames, utt_pad_proportion=0.006152, over 45.00 utterances.], tot_loss[ctc_loss=0.06695, att_loss=0.2327, loss=0.1995, over 3279792.96 frames. utt_duration=1304 frames, utt_pad_proportion=0.03797, over 10068.79 utterances.], batch size: 45, lr: 4.29e-03, grad_scale: 16.0 2023-03-09 06:00:50,822 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97674.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 06:01:10,336 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6053, 3.1198, 3.7479, 3.1058, 3.5563, 4.6874, 4.5694, 3.4169], device='cuda:1'), covar=tensor([0.0343, 0.1655, 0.1172, 0.1392, 0.1060, 0.0887, 0.0525, 0.1188], device='cuda:1'), in_proj_covar=tensor([0.0247, 0.0248, 0.0287, 0.0223, 0.0267, 0.0376, 0.0267, 0.0233], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-09 06:01:13,625 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.4777, 1.8446, 2.2861, 2.2986, 2.1759, 2.2646, 1.7953, 2.5551], device='cuda:1'), covar=tensor([0.1337, 0.2300, 0.1565, 0.1125, 0.2119, 0.1296, 0.1582, 0.1049], device='cuda:1'), in_proj_covar=tensor([0.0131, 0.0134, 0.0130, 0.0123, 0.0139, 0.0121, 0.0144, 0.0118], device='cuda:1'), out_proj_covar=tensor([1.0081e-04, 1.0598e-04, 1.0562e-04, 9.6377e-05, 1.0502e-04, 9.7642e-05, 1.0947e-04, 9.4017e-05], device='cuda:1') 2023-03-09 06:01:22,919 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1704, 5.4418, 5.4929, 5.4047, 5.5324, 5.4534, 5.1561, 4.9450], device='cuda:1'), covar=tensor([0.0995, 0.0576, 0.0238, 0.0484, 0.0231, 0.0304, 0.0381, 0.0303], device='cuda:1'), in_proj_covar=tensor([0.0543, 0.0382, 0.0370, 0.0377, 0.0439, 0.0445, 0.0376, 0.0413], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:1') 2023-03-09 06:01:30,803 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97698.0, num_to_drop=1, layers_to_drop={3} 2023-03-09 06:01:33,989 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3490, 4.4687, 4.0357, 2.3541, 4.2568, 4.2258, 3.5479, 2.2603], device='cuda:1'), covar=tensor([0.0194, 0.0162, 0.0431, 0.1652, 0.0158, 0.0360, 0.0626, 0.2492], device='cuda:1'), in_proj_covar=tensor([0.0077, 0.0105, 0.0107, 0.0111, 0.0087, 0.0115, 0.0100, 0.0103], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-09 06:01:38,591 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97703.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:01:40,006 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=97704.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:01:51,393 INFO [train2.py:809] (1/4) Epoch 25, batch 2100, loss[ctc_loss=0.03905, att_loss=0.2196, loss=0.1835, over 15944.00 frames. utt_duration=1557 frames, utt_pad_proportion=0.007633, over 41.00 utterances.], tot_loss[ctc_loss=0.06682, att_loss=0.2325, loss=0.1994, over 3282015.35 frames. utt_duration=1295 frames, utt_pad_proportion=0.04008, over 10150.22 utterances.], batch size: 41, lr: 4.29e-03, grad_scale: 16.0 2023-03-09 06:02:20,966 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97730.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:02:29,456 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97735.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 06:02:50,717 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.166e+02 1.928e+02 2.285e+02 2.765e+02 5.128e+02, threshold=4.569e+02, percent-clipped=4.0 2023-03-09 06:03:09,853 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0862, 5.1438, 4.7267, 2.7331, 4.9693, 4.8842, 4.2967, 2.4519], device='cuda:1'), covar=tensor([0.0174, 0.0124, 0.0386, 0.1423, 0.0119, 0.0216, 0.0475, 0.2204], device='cuda:1'), in_proj_covar=tensor([0.0076, 0.0104, 0.0106, 0.0111, 0.0087, 0.0115, 0.0100, 0.0103], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-09 06:03:11,116 INFO [train2.py:809] (1/4) Epoch 25, batch 2150, loss[ctc_loss=0.05259, att_loss=0.218, loss=0.1849, over 16178.00 frames. utt_duration=1580 frames, utt_pad_proportion=0.005725, over 41.00 utterances.], tot_loss[ctc_loss=0.06819, att_loss=0.2334, loss=0.2004, over 3279873.09 frames. utt_duration=1266 frames, utt_pad_proportion=0.04826, over 10373.59 utterances.], batch size: 41, lr: 4.29e-03, grad_scale: 32.0 2023-03-09 06:03:30,860 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 2023-03-09 06:03:35,210 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-03-09 06:03:59,856 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-03-09 06:04:17,329 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-03-09 06:04:32,655 INFO [train2.py:809] (1/4) Epoch 25, batch 2200, loss[ctc_loss=0.06563, att_loss=0.212, loss=0.1828, over 15472.00 frames. utt_duration=1720 frames, utt_pad_proportion=0.01012, over 36.00 utterances.], tot_loss[ctc_loss=0.06849, att_loss=0.2337, loss=0.2006, over 3270613.44 frames. utt_duration=1218 frames, utt_pad_proportion=0.0639, over 10752.56 utterances.], batch size: 36, lr: 4.29e-03, grad_scale: 32.0 2023-03-09 06:05:05,028 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0119, 4.2099, 4.3239, 4.5397, 2.8577, 4.2772, 2.8584, 1.9169], device='cuda:1'), covar=tensor([0.0520, 0.0308, 0.0659, 0.0233, 0.1511, 0.0220, 0.1328, 0.1647], device='cuda:1'), in_proj_covar=tensor([0.0204, 0.0176, 0.0260, 0.0168, 0.0221, 0.0161, 0.0229, 0.0202], 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') 2023-03-09 06:05:06,610 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7247, 3.4418, 3.3946, 3.0164, 3.4175, 3.4699, 3.5254, 2.6649], device='cuda:1'), covar=tensor([0.1158, 0.1477, 0.1673, 0.2844, 0.1333, 0.1690, 0.0796, 0.2800], device='cuda:1'), in_proj_covar=tensor([0.0191, 0.0198, 0.0211, 0.0264, 0.0174, 0.0271, 0.0196, 0.0222], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 06:05:11,502 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-03-09 06:05:34,921 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.282e+02 1.901e+02 2.356e+02 2.897e+02 7.922e+02, threshold=4.712e+02, percent-clipped=5.0 2023-03-09 06:05:53,834 INFO [train2.py:809] (1/4) Epoch 25, batch 2250, loss[ctc_loss=0.06236, att_loss=0.2194, loss=0.188, over 15637.00 frames. utt_duration=1692 frames, utt_pad_proportion=0.00764, over 37.00 utterances.], tot_loss[ctc_loss=0.06853, att_loss=0.2339, loss=0.2008, over 3273102.50 frames. utt_duration=1232 frames, utt_pad_proportion=0.06066, over 10635.93 utterances.], batch size: 37, lr: 4.29e-03, grad_scale: 16.0 2023-03-09 06:06:05,133 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8947, 4.8599, 4.7108, 3.0998, 4.6371, 4.5585, 4.2306, 2.7466], device='cuda:1'), covar=tensor([0.0124, 0.0106, 0.0261, 0.0931, 0.0117, 0.0214, 0.0316, 0.1317], device='cuda:1'), in_proj_covar=tensor([0.0077, 0.0104, 0.0107, 0.0111, 0.0087, 0.0115, 0.0100, 0.0103], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-09 06:06:29,742 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6123, 2.2424, 4.9899, 4.0823, 3.2879, 4.4485, 4.7689, 4.8458], device='cuda:1'), covar=tensor([0.0226, 0.1561, 0.0163, 0.0776, 0.1415, 0.0199, 0.0125, 0.0194], device='cuda:1'), in_proj_covar=tensor([0.0213, 0.0240, 0.0205, 0.0316, 0.0262, 0.0225, 0.0197, 0.0223], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 06:06:37,484 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97888.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:06:45,137 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.2216, 5.4233, 5.4782, 5.3885, 5.5006, 5.4426, 5.1082, 4.9203], device='cuda:1'), covar=tensor([0.0908, 0.0608, 0.0236, 0.0486, 0.0256, 0.0329, 0.0415, 0.0339], device='cuda:1'), in_proj_covar=tensor([0.0538, 0.0379, 0.0368, 0.0376, 0.0437, 0.0444, 0.0374, 0.0409], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:1') 2023-03-09 06:07:13,166 INFO [train2.py:809] (1/4) Epoch 25, batch 2300, loss[ctc_loss=0.05354, att_loss=0.2178, loss=0.1849, over 16273.00 frames. utt_duration=1515 frames, utt_pad_proportion=0.007661, over 43.00 utterances.], tot_loss[ctc_loss=0.06848, att_loss=0.2334, loss=0.2004, over 3268465.24 frames. utt_duration=1237 frames, utt_pad_proportion=0.05948, over 10581.25 utterances.], batch size: 43, lr: 4.29e-03, grad_scale: 16.0 2023-03-09 06:07:13,326 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97911.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:07:21,280 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97916.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:08:14,118 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.169e+02 1.837e+02 2.120e+02 2.605e+02 6.957e+02, threshold=4.241e+02, percent-clipped=1.0 2023-03-09 06:08:14,580 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97949.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:08:33,501 INFO [train2.py:809] (1/4) Epoch 25, batch 2350, loss[ctc_loss=0.06336, att_loss=0.2401, loss=0.2047, over 17285.00 frames. utt_duration=1259 frames, utt_pad_proportion=0.01285, over 55.00 utterances.], tot_loss[ctc_loss=0.06895, att_loss=0.2339, loss=0.2009, over 3276405.03 frames. utt_duration=1251 frames, utt_pad_proportion=0.05448, over 10491.88 utterances.], batch size: 55, lr: 4.29e-03, grad_scale: 16.0 2023-03-09 06:08:38,247 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=97964.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:08:59,924 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.20 vs. limit=5.0 2023-03-09 06:09:07,794 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1903, 5.4970, 5.0778, 5.5581, 4.9553, 5.1558, 5.6290, 5.3875], device='cuda:1'), covar=tensor([0.0582, 0.0327, 0.0721, 0.0302, 0.0379, 0.0207, 0.0224, 0.0173], device='cuda:1'), in_proj_covar=tensor([0.0399, 0.0332, 0.0374, 0.0366, 0.0333, 0.0244, 0.0316, 0.0294], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-09 06:09:34,066 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97998.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:09:45,817 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98003.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:09:57,930 INFO [train2.py:809] (1/4) Epoch 25, batch 2400, loss[ctc_loss=0.05354, att_loss=0.2111, loss=0.1796, over 15858.00 frames. utt_duration=1628 frames, utt_pad_proportion=0.00981, over 39.00 utterances.], tot_loss[ctc_loss=0.06834, att_loss=0.2336, loss=0.2006, over 3280317.90 frames. utt_duration=1243 frames, utt_pad_proportion=0.05522, over 10569.33 utterances.], batch size: 39, lr: 4.29e-03, grad_scale: 16.0 2023-03-09 06:10:00,886 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-03-09 06:10:13,279 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4817, 4.9033, 4.7131, 4.8429, 4.9451, 4.5591, 3.3791, 4.8403], device='cuda:1'), covar=tensor([0.0135, 0.0112, 0.0162, 0.0101, 0.0129, 0.0135, 0.0691, 0.0210], device='cuda:1'), in_proj_covar=tensor([0.0095, 0.0091, 0.0114, 0.0071, 0.0078, 0.0089, 0.0105, 0.0111], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 06:10:29,504 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98030.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 06:10:29,583 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98030.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:10:55,036 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=98046.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:11:00,022 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.282e+02 1.784e+02 2.263e+02 2.686e+02 5.862e+02, threshold=4.525e+02, percent-clipped=1.0 2023-03-09 06:11:03,266 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=98051.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:11:18,647 INFO [train2.py:809] (1/4) Epoch 25, batch 2450, loss[ctc_loss=0.08655, att_loss=0.2437, loss=0.2123, over 16975.00 frames. utt_duration=687.3 frames, utt_pad_proportion=0.1344, over 99.00 utterances.], tot_loss[ctc_loss=0.06867, att_loss=0.2341, loss=0.201, over 3281547.79 frames. utt_duration=1244 frames, utt_pad_proportion=0.05486, over 10566.10 utterances.], batch size: 99, lr: 4.29e-03, grad_scale: 16.0 2023-03-09 06:11:46,648 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=98078.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:12:31,665 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98106.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:12:39,097 INFO [train2.py:809] (1/4) Epoch 25, batch 2500, loss[ctc_loss=0.04578, att_loss=0.2066, loss=0.1744, over 16187.00 frames. utt_duration=1580 frames, utt_pad_proportion=0.006059, over 41.00 utterances.], tot_loss[ctc_loss=0.06852, att_loss=0.2336, loss=0.2006, over 3278664.48 frames. utt_duration=1223 frames, utt_pad_proportion=0.0613, over 10740.50 utterances.], batch size: 41, lr: 4.29e-03, grad_scale: 16.0 2023-03-09 06:13:35,882 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7253, 2.2882, 2.3097, 2.5545, 2.6898, 2.4248, 2.3283, 2.9755], device='cuda:1'), covar=tensor([0.1496, 0.2327, 0.1926, 0.1182, 0.1555, 0.1258, 0.1890, 0.1013], device='cuda:1'), in_proj_covar=tensor([0.0131, 0.0135, 0.0131, 0.0124, 0.0140, 0.0121, 0.0142, 0.0119], device='cuda:1'), out_proj_covar=tensor([1.0090e-04, 1.0656e-04, 1.0632e-04, 9.6755e-05, 1.0543e-04, 9.7745e-05, 1.0877e-04, 9.4452e-05], device='cuda:1') 2023-03-09 06:13:40,618 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.428e+02 1.837e+02 2.156e+02 2.644e+02 5.272e+02, threshold=4.313e+02, percent-clipped=2.0 2023-03-09 06:13:59,457 INFO [train2.py:809] (1/4) Epoch 25, batch 2550, loss[ctc_loss=0.06332, att_loss=0.2497, loss=0.2125, over 17038.00 frames. utt_duration=1338 frames, utt_pad_proportion=0.006798, over 51.00 utterances.], tot_loss[ctc_loss=0.06929, att_loss=0.2347, loss=0.2016, over 3273444.87 frames. utt_duration=1204 frames, utt_pad_proportion=0.06621, over 10887.61 utterances.], batch size: 51, lr: 4.28e-03, grad_scale: 16.0 2023-03-09 06:14:10,098 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98167.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:14:41,953 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0887, 5.3893, 4.9064, 5.4271, 4.8019, 5.0559, 5.4923, 5.3036], device='cuda:1'), covar=tensor([0.0552, 0.0299, 0.0774, 0.0328, 0.0404, 0.0248, 0.0231, 0.0183], device='cuda:1'), in_proj_covar=tensor([0.0395, 0.0330, 0.0372, 0.0363, 0.0331, 0.0243, 0.0313, 0.0292], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-09 06:15:20,416 INFO [train2.py:809] (1/4) Epoch 25, batch 2600, loss[ctc_loss=0.07213, att_loss=0.2499, loss=0.2143, over 16900.00 frames. utt_duration=1381 frames, utt_pad_proportion=0.006475, over 49.00 utterances.], tot_loss[ctc_loss=0.06839, att_loss=0.2334, loss=0.2004, over 3273526.04 frames. utt_duration=1250 frames, utt_pad_proportion=0.05588, over 10491.46 utterances.], batch size: 49, lr: 4.28e-03, grad_scale: 16.0 2023-03-09 06:15:20,715 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98211.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:16:14,194 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98244.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:16:21,477 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.227e+02 1.793e+02 2.177e+02 2.680e+02 7.649e+02, threshold=4.354e+02, percent-clipped=6.0 2023-03-09 06:16:37,135 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=98259.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:16:40,780 INFO [train2.py:809] (1/4) Epoch 25, batch 2650, loss[ctc_loss=0.1009, att_loss=0.2601, loss=0.2282, over 16916.00 frames. utt_duration=685 frames, utt_pad_proportion=0.1383, over 99.00 utterances.], tot_loss[ctc_loss=0.06857, att_loss=0.2337, loss=0.2006, over 3278436.76 frames. utt_duration=1245 frames, utt_pad_proportion=0.05572, over 10548.23 utterances.], batch size: 99, lr: 4.28e-03, grad_scale: 16.0 2023-03-09 06:18:00,625 INFO [train2.py:809] (1/4) Epoch 25, batch 2700, loss[ctc_loss=0.06193, att_loss=0.2212, loss=0.1893, over 16109.00 frames. utt_duration=1536 frames, utt_pad_proportion=0.007187, over 42.00 utterances.], tot_loss[ctc_loss=0.06814, att_loss=0.2332, loss=0.2002, over 3283686.22 frames. utt_duration=1242 frames, utt_pad_proportion=0.05377, over 10588.38 utterances.], batch size: 42, lr: 4.28e-03, grad_scale: 16.0 2023-03-09 06:18:07,797 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-03-09 06:18:31,193 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98330.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 06:19:00,612 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.296e+02 1.887e+02 2.414e+02 2.992e+02 7.902e+02, threshold=4.828e+02, percent-clipped=2.0 2023-03-09 06:19:19,945 INFO [train2.py:809] (1/4) Epoch 25, batch 2750, loss[ctc_loss=0.06012, att_loss=0.2403, loss=0.2043, over 16623.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.005648, over 47.00 utterances.], tot_loss[ctc_loss=0.06808, att_loss=0.233, loss=0.2, over 3276141.57 frames. utt_duration=1248 frames, utt_pad_proportion=0.05431, over 10514.41 utterances.], batch size: 47, lr: 4.28e-03, grad_scale: 16.0 2023-03-09 06:19:47,084 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=98378.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 06:19:56,236 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.6471, 5.9508, 5.3979, 5.6834, 5.6151, 5.1473, 5.3582, 5.0968], device='cuda:1'), covar=tensor([0.1223, 0.0956, 0.0971, 0.0832, 0.0875, 0.1574, 0.2112, 0.2271], device='cuda:1'), in_proj_covar=tensor([0.0543, 0.0631, 0.0478, 0.0466, 0.0435, 0.0482, 0.0628, 0.0539], 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') 2023-03-09 06:20:37,073 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98410.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:20:38,135 INFO [train2.py:809] (1/4) Epoch 25, batch 2800, loss[ctc_loss=0.0913, att_loss=0.2541, loss=0.2215, over 17297.00 frames. utt_duration=700.5 frames, utt_pad_proportion=0.1189, over 99.00 utterances.], tot_loss[ctc_loss=0.06811, att_loss=0.2331, loss=0.2001, over 3279817.64 frames. utt_duration=1266 frames, utt_pad_proportion=0.04808, over 10377.40 utterances.], batch size: 99, lr: 4.28e-03, grad_scale: 16.0 2023-03-09 06:21:09,929 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2267, 2.8107, 3.2511, 4.2468, 3.7795, 3.7236, 2.7449, 2.3281], device='cuda:1'), covar=tensor([0.0857, 0.2013, 0.0878, 0.0576, 0.0943, 0.0578, 0.1749, 0.2103], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0218, 0.0186, 0.0220, 0.0231, 0.0184, 0.0204, 0.0189], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 06:21:37,616 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.204e+02 1.960e+02 2.287e+02 2.906e+02 4.125e+02, threshold=4.574e+02, percent-clipped=0.0 2023-03-09 06:21:56,308 INFO [train2.py:809] (1/4) Epoch 25, batch 2850, loss[ctc_loss=0.07192, att_loss=0.2449, loss=0.2103, over 16771.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.006274, over 48.00 utterances.], tot_loss[ctc_loss=0.0677, att_loss=0.2326, loss=0.1996, over 3273141.26 frames. utt_duration=1253 frames, utt_pad_proportion=0.05398, over 10464.22 utterances.], batch size: 48, lr: 4.28e-03, grad_scale: 16.0 2023-03-09 06:21:57,969 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98462.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:22:04,240 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9135, 4.2737, 4.5357, 4.4882, 2.7055, 4.3306, 2.9966, 1.5049], device='cuda:1'), covar=tensor([0.0500, 0.0303, 0.0608, 0.0247, 0.1586, 0.0245, 0.1324, 0.1907], device='cuda:1'), in_proj_covar=tensor([0.0209, 0.0179, 0.0265, 0.0174, 0.0225, 0.0165, 0.0233, 0.0206], 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') 2023-03-09 06:22:05,018 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-03-09 06:22:12,362 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98471.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:23:14,287 INFO [train2.py:809] (1/4) Epoch 25, batch 2900, loss[ctc_loss=0.07262, att_loss=0.2371, loss=0.2042, over 16762.00 frames. utt_duration=1398 frames, utt_pad_proportion=0.006056, over 48.00 utterances.], tot_loss[ctc_loss=0.0678, att_loss=0.2322, loss=0.1993, over 3263344.12 frames. utt_duration=1273 frames, utt_pad_proportion=0.05096, over 10268.90 utterances.], batch size: 48, lr: 4.28e-03, grad_scale: 16.0 2023-03-09 06:23:45,947 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-03-09 06:23:52,992 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.7292, 2.7077, 3.8911, 3.5052, 3.0089, 3.7095, 3.6620, 3.7634], device='cuda:1'), covar=tensor([0.0344, 0.1226, 0.0191, 0.0781, 0.1277, 0.0324, 0.0290, 0.0358], device='cuda:1'), in_proj_covar=tensor([0.0214, 0.0243, 0.0209, 0.0320, 0.0265, 0.0229, 0.0199, 0.0226], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 06:24:06,289 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98543.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:24:07,675 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98544.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:24:14,934 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.223e+02 1.817e+02 2.098e+02 2.346e+02 4.343e+02, threshold=4.197e+02, percent-clipped=0.0 2023-03-09 06:24:16,105 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-03-09 06:24:34,079 INFO [train2.py:809] (1/4) Epoch 25, batch 2950, loss[ctc_loss=0.06289, att_loss=0.2302, loss=0.1967, over 16131.00 frames. utt_duration=1538 frames, utt_pad_proportion=0.005848, over 42.00 utterances.], tot_loss[ctc_loss=0.06814, att_loss=0.2325, loss=0.1996, over 3265017.65 frames. utt_duration=1255 frames, utt_pad_proportion=0.05573, over 10420.33 utterances.], batch size: 42, lr: 4.28e-03, grad_scale: 16.0 2023-03-09 06:25:23,552 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=98592.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:25:43,560 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98604.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:25:51,115 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.5709, 5.8523, 5.2803, 5.5859, 5.4939, 4.9733, 5.2773, 5.0855], device='cuda:1'), covar=tensor([0.1428, 0.0938, 0.1064, 0.0929, 0.0932, 0.1599, 0.2232, 0.2492], device='cuda:1'), in_proj_covar=tensor([0.0542, 0.0627, 0.0476, 0.0466, 0.0434, 0.0482, 0.0626, 0.0537], 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') 2023-03-09 06:25:54,031 INFO [train2.py:809] (1/4) Epoch 25, batch 3000, loss[ctc_loss=0.07963, att_loss=0.2474, loss=0.2138, over 17083.00 frames. utt_duration=1222 frames, utt_pad_proportion=0.01704, over 56.00 utterances.], tot_loss[ctc_loss=0.06921, att_loss=0.234, loss=0.2011, over 3278255.17 frames. utt_duration=1238 frames, utt_pad_proportion=0.05522, over 10606.17 utterances.], batch size: 56, lr: 4.27e-03, grad_scale: 16.0 2023-03-09 06:25:54,032 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-09 06:26:08,490 INFO [train2.py:843] (1/4) Epoch 25, validation: ctc_loss=0.04165, att_loss=0.235, loss=0.1963, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-09 06:26:08,491 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-09 06:27:09,504 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.233e+02 1.890e+02 2.327e+02 2.970e+02 6.667e+02, threshold=4.653e+02, percent-clipped=5.0 2023-03-09 06:27:18,683 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-03-09 06:27:28,843 INFO [train2.py:809] (1/4) Epoch 25, batch 3050, loss[ctc_loss=0.07367, att_loss=0.2288, loss=0.1978, over 15951.00 frames. utt_duration=1558 frames, utt_pad_proportion=0.007182, over 41.00 utterances.], tot_loss[ctc_loss=0.06885, att_loss=0.234, loss=0.201, over 3282370.64 frames. utt_duration=1231 frames, utt_pad_proportion=0.05568, over 10674.84 utterances.], batch size: 41, lr: 4.27e-03, grad_scale: 16.0 2023-03-09 06:27:41,917 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1590, 5.4305, 4.8389, 5.2712, 5.0670, 4.5908, 4.9281, 4.7996], device='cuda:1'), covar=tensor([0.1382, 0.1011, 0.1072, 0.0879, 0.0951, 0.1704, 0.2255, 0.2175], device='cuda:1'), in_proj_covar=tensor([0.0539, 0.0625, 0.0474, 0.0464, 0.0433, 0.0482, 0.0624, 0.0536], 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') 2023-03-09 06:28:17,434 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4514, 4.2729, 4.4433, 4.4622, 5.0680, 4.3757, 4.3016, 2.4916], device='cuda:1'), covar=tensor([0.0260, 0.0483, 0.0382, 0.0305, 0.0729, 0.0286, 0.0442, 0.1920], device='cuda:1'), in_proj_covar=tensor([0.0178, 0.0206, 0.0203, 0.0220, 0.0375, 0.0177, 0.0193, 0.0216], 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') 2023-03-09 06:28:48,631 INFO [train2.py:809] (1/4) Epoch 25, batch 3100, loss[ctc_loss=0.09633, att_loss=0.2375, loss=0.2093, over 15652.00 frames. utt_duration=1694 frames, utt_pad_proportion=0.007869, over 37.00 utterances.], tot_loss[ctc_loss=0.06786, att_loss=0.2332, loss=0.2002, over 3275698.43 frames. utt_duration=1252 frames, utt_pad_proportion=0.05212, over 10475.55 utterances.], batch size: 37, lr: 4.27e-03, grad_scale: 16.0 2023-03-09 06:29:15,432 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.12 vs. limit=5.0 2023-03-09 06:29:48,741 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.374e+02 1.858e+02 2.243e+02 2.658e+02 4.552e+02, threshold=4.486e+02, percent-clipped=0.0 2023-03-09 06:30:07,766 INFO [train2.py:809] (1/4) Epoch 25, batch 3150, loss[ctc_loss=0.04899, att_loss=0.2166, loss=0.183, over 16273.00 frames. utt_duration=1515 frames, utt_pad_proportion=0.006995, over 43.00 utterances.], tot_loss[ctc_loss=0.06769, att_loss=0.2332, loss=0.2001, over 3277528.79 frames. utt_duration=1259 frames, utt_pad_proportion=0.04981, over 10426.62 utterances.], batch size: 43, lr: 4.27e-03, grad_scale: 16.0 2023-03-09 06:30:09,797 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98762.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:30:15,619 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98766.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:31:26,675 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=98810.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:31:28,129 INFO [train2.py:809] (1/4) Epoch 25, batch 3200, loss[ctc_loss=0.06649, att_loss=0.2301, loss=0.1974, over 16279.00 frames. utt_duration=1516 frames, utt_pad_proportion=0.007051, over 43.00 utterances.], tot_loss[ctc_loss=0.06764, att_loss=0.2326, loss=0.1996, over 3270344.34 frames. utt_duration=1237 frames, utt_pad_proportion=0.05803, over 10587.86 utterances.], batch size: 43, lr: 4.27e-03, grad_scale: 8.0 2023-03-09 06:31:47,813 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7985, 2.4289, 2.5359, 2.7612, 2.8732, 2.7201, 2.5573, 3.1242], device='cuda:1'), covar=tensor([0.2041, 0.3223, 0.2336, 0.1919, 0.2461, 0.2134, 0.2742, 0.1578], device='cuda:1'), in_proj_covar=tensor([0.0131, 0.0134, 0.0131, 0.0122, 0.0139, 0.0120, 0.0141, 0.0118], device='cuda:1'), out_proj_covar=tensor([1.0060e-04, 1.0579e-04, 1.0603e-04, 9.5939e-05, 1.0499e-04, 9.7124e-05, 1.0808e-04, 9.3768e-05], device='cuda:1') 2023-03-09 06:32:28,103 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7466, 5.0416, 4.6203, 5.1109, 4.5130, 4.7687, 5.1821, 4.9395], device='cuda:1'), covar=tensor([0.0713, 0.0327, 0.0801, 0.0316, 0.0427, 0.0329, 0.0233, 0.0190], device='cuda:1'), in_proj_covar=tensor([0.0396, 0.0329, 0.0371, 0.0360, 0.0330, 0.0244, 0.0312, 0.0292], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-09 06:32:29,336 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.298e+02 1.857e+02 2.201e+02 2.576e+02 7.126e+02, threshold=4.402e+02, percent-clipped=1.0 2023-03-09 06:32:46,794 INFO [train2.py:809] (1/4) Epoch 25, batch 3250, loss[ctc_loss=0.05903, att_loss=0.2214, loss=0.189, over 16556.00 frames. utt_duration=1473 frames, utt_pad_proportion=0.005447, over 45.00 utterances.], tot_loss[ctc_loss=0.06743, att_loss=0.2322, loss=0.1992, over 3268099.37 frames. utt_duration=1244 frames, utt_pad_proportion=0.05707, over 10519.34 utterances.], batch size: 45, lr: 4.27e-03, grad_scale: 8.0 2023-03-09 06:32:54,395 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.85 vs. limit=2.0 2023-03-09 06:33:14,660 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4625, 2.6237, 4.9097, 3.9242, 3.1502, 4.3098, 4.8092, 4.6993], device='cuda:1'), covar=tensor([0.0315, 0.1583, 0.0235, 0.0969, 0.1616, 0.0261, 0.0185, 0.0273], device='cuda:1'), in_proj_covar=tensor([0.0216, 0.0246, 0.0210, 0.0322, 0.0267, 0.0231, 0.0201, 0.0229], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 06:33:33,469 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98890.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:33:41,351 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98895.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:33:47,314 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98899.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:34:05,911 INFO [train2.py:809] (1/4) Epoch 25, batch 3300, loss[ctc_loss=0.06991, att_loss=0.239, loss=0.2052, over 16973.00 frames. utt_duration=1359 frames, utt_pad_proportion=0.007202, over 50.00 utterances.], tot_loss[ctc_loss=0.06747, att_loss=0.2327, loss=0.1996, over 3273915.53 frames. utt_duration=1242 frames, utt_pad_proportion=0.05624, over 10558.16 utterances.], batch size: 50, lr: 4.27e-03, grad_scale: 8.0 2023-03-09 06:34:50,530 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.1563, 3.8238, 3.6539, 3.3344, 3.8458, 3.8808, 3.8419, 3.0581], device='cuda:1'), covar=tensor([0.0953, 0.0894, 0.1737, 0.2834, 0.0891, 0.1853, 0.1042, 0.2643], device='cuda:1'), in_proj_covar=tensor([0.0193, 0.0201, 0.0214, 0.0268, 0.0177, 0.0276, 0.0199, 0.0225], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 06:35:02,961 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.1144, 3.7784, 3.6513, 3.2901, 3.6998, 3.8669, 3.7711, 2.8965], device='cuda:1'), covar=tensor([0.0985, 0.1231, 0.1897, 0.3087, 0.1969, 0.1390, 0.1047, 0.3215], device='cuda:1'), in_proj_covar=tensor([0.0193, 0.0201, 0.0214, 0.0268, 0.0177, 0.0276, 0.0199, 0.0225], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 06:35:07,194 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.395e+02 1.882e+02 2.265e+02 2.775e+02 6.610e+02, threshold=4.530e+02, percent-clipped=3.0 2023-03-09 06:35:09,241 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98951.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:35:16,986 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98956.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:35:24,899 INFO [train2.py:809] (1/4) Epoch 25, batch 3350, loss[ctc_loss=0.05583, att_loss=0.2239, loss=0.1903, over 15961.00 frames. utt_duration=1559 frames, utt_pad_proportion=0.005926, over 41.00 utterances.], tot_loss[ctc_loss=0.0676, att_loss=0.2336, loss=0.2004, over 3284404.89 frames. utt_duration=1224 frames, utt_pad_proportion=0.05707, over 10746.88 utterances.], batch size: 41, lr: 4.27e-03, grad_scale: 8.0 2023-03-09 06:35:41,915 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98971.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:36:44,963 INFO [train2.py:809] (1/4) Epoch 25, batch 3400, loss[ctc_loss=0.08072, att_loss=0.2524, loss=0.218, over 17443.00 frames. utt_duration=1109 frames, utt_pad_proportion=0.03059, over 63.00 utterances.], tot_loss[ctc_loss=0.068, att_loss=0.2333, loss=0.2002, over 3275492.90 frames. utt_duration=1214 frames, utt_pad_proportion=0.06248, over 10802.91 utterances.], batch size: 63, lr: 4.27e-03, grad_scale: 8.0 2023-03-09 06:37:07,462 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.2951, 5.2296, 5.1224, 3.1354, 5.0564, 4.8362, 4.5711, 3.2246], device='cuda:1'), covar=tensor([0.0110, 0.0110, 0.0227, 0.0891, 0.0084, 0.0188, 0.0265, 0.1054], device='cuda:1'), in_proj_covar=tensor([0.0077, 0.0105, 0.0107, 0.0111, 0.0087, 0.0115, 0.0100, 0.0103], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-09 06:37:18,879 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99032.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:37:45,687 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.421e+02 1.932e+02 2.257e+02 2.819e+02 9.991e+02, threshold=4.513e+02, percent-clipped=2.0 2023-03-09 06:37:47,571 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99051.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:38:03,635 INFO [train2.py:809] (1/4) Epoch 25, batch 3450, loss[ctc_loss=0.06569, att_loss=0.2429, loss=0.2074, over 16770.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.005671, over 48.00 utterances.], tot_loss[ctc_loss=0.06808, att_loss=0.2336, loss=0.2005, over 3272870.77 frames. utt_duration=1230 frames, utt_pad_proportion=0.06049, over 10659.51 utterances.], batch size: 48, lr: 4.26e-03, grad_scale: 8.0 2023-03-09 06:38:12,095 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99066.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:39:22,271 INFO [train2.py:809] (1/4) Epoch 25, batch 3500, loss[ctc_loss=0.07983, att_loss=0.2541, loss=0.2192, over 17036.00 frames. utt_duration=1287 frames, utt_pad_proportion=0.01067, over 53.00 utterances.], tot_loss[ctc_loss=0.06731, att_loss=0.2325, loss=0.1995, over 3275879.85 frames. utt_duration=1254 frames, utt_pad_proportion=0.05317, over 10457.83 utterances.], batch size: 53, lr: 4.26e-03, grad_scale: 8.0 2023-03-09 06:39:24,152 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99112.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:39:27,523 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=99114.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:39:48,610 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.5076, 2.1203, 2.0523, 2.5012, 2.8244, 2.4899, 2.1401, 2.8882], device='cuda:1'), covar=tensor([0.1327, 0.2358, 0.1595, 0.1228, 0.1470, 0.1183, 0.1962, 0.1123], device='cuda:1'), in_proj_covar=tensor([0.0132, 0.0135, 0.0131, 0.0124, 0.0140, 0.0121, 0.0142, 0.0120], device='cuda:1'), out_proj_covar=tensor([1.0145e-04, 1.0651e-04, 1.0642e-04, 9.7022e-05, 1.0580e-04, 9.7354e-05, 1.0880e-04, 9.5026e-05], device='cuda:1') 2023-03-09 06:39:55,269 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=2.94 vs. limit=5.0 2023-03-09 06:40:23,774 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.048e+02 1.663e+02 2.107e+02 2.565e+02 8.689e+02, threshold=4.214e+02, percent-clipped=2.0 2023-03-09 06:40:40,990 INFO [train2.py:809] (1/4) Epoch 25, batch 3550, loss[ctc_loss=0.06305, att_loss=0.2269, loss=0.1941, over 16481.00 frames. utt_duration=1435 frames, utt_pad_proportion=0.006383, over 46.00 utterances.], tot_loss[ctc_loss=0.06789, att_loss=0.2332, loss=0.2001, over 3281619.97 frames. utt_duration=1231 frames, utt_pad_proportion=0.05733, over 10672.52 utterances.], batch size: 46, lr: 4.26e-03, grad_scale: 8.0 2023-03-09 06:41:40,265 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99199.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:41:59,398 INFO [train2.py:809] (1/4) Epoch 25, batch 3600, loss[ctc_loss=0.07333, att_loss=0.2242, loss=0.1941, over 16166.00 frames. utt_duration=1578 frames, utt_pad_proportion=0.007864, over 41.00 utterances.], tot_loss[ctc_loss=0.06805, att_loss=0.2334, loss=0.2003, over 3283570.04 frames. utt_duration=1237 frames, utt_pad_proportion=0.05472, over 10634.22 utterances.], batch size: 41, lr: 4.26e-03, grad_scale: 8.0 2023-03-09 06:42:28,470 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99229.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:42:55,191 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99246.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:42:57,251 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=99247.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:43:01,841 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.167e+02 1.837e+02 2.162e+02 2.807e+02 5.340e+02, threshold=4.325e+02, percent-clipped=4.0 2023-03-09 06:43:03,549 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99251.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:43:12,007 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9443, 4.8911, 4.7450, 2.3004, 1.8926, 2.9690, 2.2633, 3.7389], device='cuda:1'), covar=tensor([0.0757, 0.0293, 0.0306, 0.5142, 0.5705, 0.2432, 0.4103, 0.1710], device='cuda:1'), in_proj_covar=tensor([0.0361, 0.0291, 0.0275, 0.0248, 0.0340, 0.0334, 0.0262, 0.0371], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-03-09 06:43:19,234 INFO [train2.py:809] (1/4) Epoch 25, batch 3650, loss[ctc_loss=0.06977, att_loss=0.2187, loss=0.1889, over 15770.00 frames. utt_duration=1661 frames, utt_pad_proportion=0.008793, over 38.00 utterances.], tot_loss[ctc_loss=0.06838, att_loss=0.2336, loss=0.2005, over 3281811.80 frames. utt_duration=1226 frames, utt_pad_proportion=0.0572, over 10716.89 utterances.], batch size: 38, lr: 4.26e-03, grad_scale: 8.0 2023-03-09 06:43:30,890 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99268.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:43:33,867 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6569, 5.0776, 4.9161, 5.1498, 5.2397, 4.7310, 3.3558, 5.1420], device='cuda:1'), covar=tensor([0.0118, 0.0101, 0.0122, 0.0057, 0.0075, 0.0114, 0.0744, 0.0131], device='cuda:1'), in_proj_covar=tensor([0.0095, 0.0091, 0.0113, 0.0070, 0.0077, 0.0088, 0.0104, 0.0109], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 06:43:41,608 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-03-09 06:44:05,942 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99290.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:44:24,443 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0795, 4.3473, 4.7120, 4.4741, 3.0269, 4.4469, 2.5975, 1.6953], device='cuda:1'), covar=tensor([0.0618, 0.0279, 0.0497, 0.0278, 0.1364, 0.0241, 0.1483, 0.1663], device='cuda:1'), in_proj_covar=tensor([0.0208, 0.0179, 0.0262, 0.0174, 0.0223, 0.0163, 0.0232, 0.0203], 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') 2023-03-09 06:44:40,182 INFO [train2.py:809] (1/4) Epoch 25, batch 3700, loss[ctc_loss=0.06531, att_loss=0.224, loss=0.1923, over 16383.00 frames. utt_duration=1491 frames, utt_pad_proportion=0.008765, over 44.00 utterances.], tot_loss[ctc_loss=0.06847, att_loss=0.2339, loss=0.2008, over 3275383.71 frames. utt_duration=1208 frames, utt_pad_proportion=0.06244, over 10856.49 utterances.], batch size: 44, lr: 4.26e-03, grad_scale: 8.0 2023-03-09 06:45:01,489 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99324.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:45:04,519 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1887, 4.4064, 4.6434, 4.6195, 2.9165, 4.4482, 2.7759, 1.8129], device='cuda:1'), covar=tensor([0.0579, 0.0305, 0.0598, 0.0297, 0.1522, 0.0269, 0.1429, 0.1746], device='cuda:1'), in_proj_covar=tensor([0.0209, 0.0179, 0.0263, 0.0174, 0.0224, 0.0164, 0.0233, 0.0204], 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') 2023-03-09 06:45:05,819 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99327.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:45:09,012 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99329.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:45:35,364 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.8200, 2.4496, 2.4792, 3.2613, 3.0909, 3.3036, 2.5907, 2.2425], device='cuda:1'), covar=tensor([0.0717, 0.1704, 0.1038, 0.0753, 0.0897, 0.0421, 0.1318, 0.1719], device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0217, 0.0187, 0.0221, 0.0232, 0.0182, 0.0204, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 06:45:41,897 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.181e+02 1.855e+02 2.238e+02 2.669e+02 6.912e+02, threshold=4.475e+02, percent-clipped=5.0 2023-03-09 06:45:59,441 INFO [train2.py:809] (1/4) Epoch 25, batch 3750, loss[ctc_loss=0.07576, att_loss=0.241, loss=0.208, over 16459.00 frames. utt_duration=1433 frames, utt_pad_proportion=0.007005, over 46.00 utterances.], tot_loss[ctc_loss=0.0689, att_loss=0.2337, loss=0.2007, over 3267725.62 frames. utt_duration=1190 frames, utt_pad_proportion=0.07056, over 10993.57 utterances.], batch size: 46, lr: 4.26e-03, grad_scale: 8.0 2023-03-09 06:46:36,258 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6260, 3.3356, 3.7533, 3.1141, 3.5580, 4.7304, 4.5092, 3.3854], device='cuda:1'), covar=tensor([0.0429, 0.1382, 0.1204, 0.1298, 0.1201, 0.0890, 0.0618, 0.1214], device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0247, 0.0288, 0.0222, 0.0268, 0.0375, 0.0269, 0.0233], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-09 06:46:37,923 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99385.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:47:13,287 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99407.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:47:19,836 INFO [train2.py:809] (1/4) Epoch 25, batch 3800, loss[ctc_loss=0.1061, att_loss=0.2634, loss=0.2319, over 14192.00 frames. utt_duration=390.4 frames, utt_pad_proportion=0.3174, over 146.00 utterances.], tot_loss[ctc_loss=0.06902, att_loss=0.2343, loss=0.2012, over 3271543.20 frames. utt_duration=1192 frames, utt_pad_proportion=0.06974, over 10989.66 utterances.], batch size: 146, lr: 4.26e-03, grad_scale: 8.0 2023-03-09 06:47:34,703 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99420.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:48:02,840 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0522, 5.1179, 4.8990, 2.2032, 2.0183, 2.9393, 2.7824, 3.8402], device='cuda:1'), covar=tensor([0.0774, 0.0314, 0.0299, 0.5965, 0.5848, 0.2725, 0.3166, 0.1736], device='cuda:1'), in_proj_covar=tensor([0.0364, 0.0294, 0.0278, 0.0252, 0.0345, 0.0338, 0.0265, 0.0375], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-09 06:48:22,108 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.283e+02 1.840e+02 2.193e+02 2.491e+02 3.920e+02, threshold=4.386e+02, percent-clipped=0.0 2023-03-09 06:48:38,761 INFO [train2.py:809] (1/4) Epoch 25, batch 3850, loss[ctc_loss=0.08504, att_loss=0.2494, loss=0.2165, over 17073.00 frames. utt_duration=1290 frames, utt_pad_proportion=0.008455, over 53.00 utterances.], tot_loss[ctc_loss=0.06833, att_loss=0.2335, loss=0.2005, over 3267487.83 frames. utt_duration=1200 frames, utt_pad_proportion=0.06844, over 10903.47 utterances.], batch size: 53, lr: 4.26e-03, grad_scale: 8.0 2023-03-09 06:48:49,792 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99468.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 06:49:09,249 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99481.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:49:44,837 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0680, 6.2581, 5.7961, 5.9828, 5.9370, 5.4439, 5.7620, 5.4888], device='cuda:1'), covar=tensor([0.1166, 0.0915, 0.0966, 0.0924, 0.0778, 0.1601, 0.2249, 0.2447], device='cuda:1'), in_proj_covar=tensor([0.0540, 0.0626, 0.0476, 0.0467, 0.0435, 0.0486, 0.0627, 0.0540], 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') 2023-03-09 06:49:55,231 INFO [train2.py:809] (1/4) Epoch 25, batch 3900, loss[ctc_loss=0.073, att_loss=0.2444, loss=0.2101, over 17001.00 frames. utt_duration=1335 frames, utt_pad_proportion=0.009615, over 51.00 utterances.], tot_loss[ctc_loss=0.06817, att_loss=0.2336, loss=0.2005, over 3262159.64 frames. utt_duration=1212 frames, utt_pad_proportion=0.06695, over 10781.75 utterances.], batch size: 51, lr: 4.25e-03, grad_scale: 8.0 2023-03-09 06:50:23,086 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99529.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 06:50:48,864 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99546.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:50:54,663 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.126e+02 1.849e+02 2.248e+02 2.718e+02 5.190e+02, threshold=4.496e+02, percent-clipped=3.0 2023-03-09 06:50:56,451 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99551.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:51:11,367 INFO [train2.py:809] (1/4) Epoch 25, batch 3950, loss[ctc_loss=0.06609, att_loss=0.2366, loss=0.2025, over 16951.00 frames. utt_duration=1358 frames, utt_pad_proportion=0.006935, over 50.00 utterances.], tot_loss[ctc_loss=0.06781, att_loss=0.2333, loss=0.2002, over 3264859.91 frames. utt_duration=1231 frames, utt_pad_proportion=0.06188, over 10619.21 utterances.], batch size: 50, lr: 4.25e-03, grad_scale: 8.0 2023-03-09 06:51:22,253 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99568.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:51:47,852 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99585.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:51:54,100 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99589.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:52:25,627 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=99594.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:52:27,618 INFO [train2.py:809] (1/4) Epoch 26, batch 0, loss[ctc_loss=0.04805, att_loss=0.2, loss=0.1696, over 15886.00 frames. utt_duration=1630 frames, utt_pad_proportion=0.00944, over 39.00 utterances.], tot_loss[ctc_loss=0.04805, att_loss=0.2, loss=0.1696, over 15886.00 frames. utt_duration=1630 frames, utt_pad_proportion=0.00944, over 39.00 utterances.], batch size: 39, lr: 4.17e-03, grad_scale: 8.0 2023-03-09 06:52:27,618 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-09 06:52:38,167 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7136, 3.5081, 3.8405, 3.5948, 3.7776, 4.7581, 4.5626, 3.7086], device='cuda:1'), covar=tensor([0.0323, 0.1237, 0.1161, 0.1031, 0.0912, 0.0867, 0.0556, 0.0955], device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0247, 0.0287, 0.0221, 0.0265, 0.0375, 0.0269, 0.0232], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-09 06:52:40,071 INFO [train2.py:843] (1/4) Epoch 26, validation: ctc_loss=0.04131, att_loss=0.2338, loss=0.1953, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-09 06:52:40,072 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-09 06:52:46,819 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=99599.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:53:27,432 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99624.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:53:32,270 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99627.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:53:36,082 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99629.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:53:40,603 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9800, 5.2313, 5.2130, 5.1965, 5.2974, 5.2430, 4.9146, 4.6779], device='cuda:1'), covar=tensor([0.1006, 0.0561, 0.0290, 0.0535, 0.0258, 0.0305, 0.0402, 0.0367], device='cuda:1'), in_proj_covar=tensor([0.0535, 0.0379, 0.0364, 0.0373, 0.0434, 0.0445, 0.0373, 0.0408], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004], device='cuda:1') 2023-03-09 06:54:00,749 INFO [train2.py:809] (1/4) Epoch 26, batch 50, loss[ctc_loss=0.06594, att_loss=0.2346, loss=0.2009, over 16412.00 frames. utt_duration=1493 frames, utt_pad_proportion=0.007042, over 44.00 utterances.], tot_loss[ctc_loss=0.06739, att_loss=0.2331, loss=0.2, over 741235.30 frames. utt_duration=1294 frames, utt_pad_proportion=0.0382, over 2293.94 utterances.], batch size: 44, lr: 4.17e-03, grad_scale: 8.0 2023-03-09 06:54:08,989 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.180e+02 1.776e+02 2.200e+02 2.639e+02 7.025e+02, threshold=4.399e+02, percent-clipped=1.0 2023-03-09 06:54:09,456 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99650.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:54:48,108 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=99675.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:54:56,435 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99680.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:54:56,581 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99680.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:55:06,978 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0476, 5.2754, 5.2429, 5.2484, 5.3337, 5.2880, 4.9038, 4.7272], device='cuda:1'), covar=tensor([0.0964, 0.0556, 0.0296, 0.0554, 0.0289, 0.0325, 0.0479, 0.0389], device='cuda:1'), in_proj_covar=tensor([0.0534, 0.0379, 0.0364, 0.0373, 0.0435, 0.0444, 0.0373, 0.0408], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004], device='cuda:1') 2023-03-09 06:55:19,430 INFO [train2.py:809] (1/4) Epoch 26, batch 100, loss[ctc_loss=0.07394, att_loss=0.2364, loss=0.2039, over 16606.00 frames. utt_duration=1415 frames, utt_pad_proportion=0.006425, over 47.00 utterances.], tot_loss[ctc_loss=0.0685, att_loss=0.2352, loss=0.2018, over 1308765.99 frames. utt_duration=1256 frames, utt_pad_proportion=0.0414, over 4171.84 utterances.], batch size: 47, lr: 4.17e-03, grad_scale: 8.0 2023-03-09 06:55:38,612 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99707.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:56:09,382 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-03-09 06:56:12,771 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 2023-03-09 06:56:31,496 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99741.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:56:37,965 INFO [train2.py:809] (1/4) Epoch 26, batch 150, loss[ctc_loss=0.05743, att_loss=0.212, loss=0.1811, over 16114.00 frames. utt_duration=1536 frames, utt_pad_proportion=0.006345, over 42.00 utterances.], tot_loss[ctc_loss=0.06818, att_loss=0.2336, loss=0.2005, over 1741863.55 frames. utt_duration=1248 frames, utt_pad_proportion=0.04638, over 5590.81 utterances.], batch size: 42, lr: 4.17e-03, grad_scale: 8.0 2023-03-09 06:56:46,324 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.249e+02 1.892e+02 2.300e+02 2.888e+02 7.505e+02, threshold=4.601e+02, percent-clipped=4.0 2023-03-09 06:56:54,127 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=99755.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:57:07,367 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2535, 2.8912, 3.0875, 4.3706, 3.9150, 3.9087, 2.9071, 2.2047], device='cuda:1'), covar=tensor([0.0860, 0.1858, 0.1023, 0.0598, 0.0901, 0.0501, 0.1572, 0.2308], device='cuda:1'), in_proj_covar=tensor([0.0188, 0.0220, 0.0189, 0.0224, 0.0234, 0.0186, 0.0206, 0.0193], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 06:57:27,681 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99776.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 06:57:57,208 INFO [train2.py:809] (1/4) Epoch 26, batch 200, loss[ctc_loss=0.07457, att_loss=0.2288, loss=0.1979, over 16173.00 frames. utt_duration=1579 frames, utt_pad_proportion=0.006719, over 41.00 utterances.], tot_loss[ctc_loss=0.06739, att_loss=0.2335, loss=0.2003, over 2079725.04 frames. utt_duration=1247 frames, utt_pad_proportion=0.04847, over 6677.47 utterances.], batch size: 41, lr: 4.16e-03, grad_scale: 8.0 2023-03-09 06:58:41,516 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.0886, 3.7215, 3.7292, 3.2129, 3.7295, 3.7524, 3.8172, 2.8695], device='cuda:1'), covar=tensor([0.0922, 0.1264, 0.1211, 0.2802, 0.1271, 0.2023, 0.0621, 0.2754], device='cuda:1'), in_proj_covar=tensor([0.0193, 0.0200, 0.0213, 0.0267, 0.0177, 0.0275, 0.0196, 0.0226], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 06:58:44,431 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99824.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 06:59:17,386 INFO [train2.py:809] (1/4) Epoch 26, batch 250, loss[ctc_loss=0.06444, att_loss=0.2377, loss=0.2031, over 17289.00 frames. utt_duration=1259 frames, utt_pad_proportion=0.01262, over 55.00 utterances.], tot_loss[ctc_loss=0.06792, att_loss=0.2332, loss=0.2002, over 2343028.59 frames. utt_duration=1210 frames, utt_pad_proportion=0.05928, over 7755.43 utterances.], batch size: 55, lr: 4.16e-03, grad_scale: 8.0 2023-03-09 06:59:25,258 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.472e+02 1.972e+02 2.288e+02 2.962e+02 9.624e+02, threshold=4.577e+02, percent-clipped=5.0 2023-03-09 06:59:48,981 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.5244, 2.2804, 2.0512, 2.6304, 2.7227, 2.5244, 2.2288, 2.8911], device='cuda:1'), covar=tensor([0.1412, 0.2237, 0.1873, 0.0990, 0.1680, 0.1122, 0.1855, 0.0922], device='cuda:1'), in_proj_covar=tensor([0.0134, 0.0135, 0.0132, 0.0125, 0.0142, 0.0122, 0.0144, 0.0121], device='cuda:1'), out_proj_covar=tensor([1.0259e-04, 1.0682e-04, 1.0698e-04, 9.7999e-05, 1.0706e-04, 9.8241e-05, 1.1032e-04, 9.6017e-05], device='cuda:1') 2023-03-09 07:00:19,868 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99885.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 07:00:35,352 INFO [train2.py:809] (1/4) Epoch 26, batch 300, loss[ctc_loss=0.06799, att_loss=0.238, loss=0.204, over 16982.00 frames. utt_duration=1334 frames, utt_pad_proportion=0.009942, over 51.00 utterances.], tot_loss[ctc_loss=0.06809, att_loss=0.2335, loss=0.2004, over 2549612.92 frames. utt_duration=1223 frames, utt_pad_proportion=0.05902, over 8347.98 utterances.], batch size: 51, lr: 4.16e-03, grad_scale: 8.0 2023-03-09 07:01:21,957 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99924.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 07:01:22,058 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99924.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 07:01:35,966 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=99933.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 07:01:40,910 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9673, 2.3632, 3.4540, 2.6627, 3.2752, 4.2402, 4.1439, 2.5755], device='cuda:1'), covar=tensor([0.0566, 0.2428, 0.1143, 0.1797, 0.1082, 0.0977, 0.0575, 0.1916], device='cuda:1'), in_proj_covar=tensor([0.0247, 0.0248, 0.0287, 0.0222, 0.0266, 0.0377, 0.0270, 0.0232], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-09 07:01:55,232 INFO [train2.py:809] (1/4) Epoch 26, batch 350, loss[ctc_loss=0.05961, att_loss=0.2091, loss=0.1792, over 14533.00 frames. utt_duration=1818 frames, utt_pad_proportion=0.03226, over 32.00 utterances.], tot_loss[ctc_loss=0.06695, att_loss=0.2319, loss=0.1989, over 2703774.87 frames. utt_duration=1259 frames, utt_pad_proportion=0.05167, over 8603.00 utterances.], batch size: 32, lr: 4.16e-03, grad_scale: 8.0 2023-03-09 07:01:55,389 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99945.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 07:02:02,688 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.378e+02 1.873e+02 2.109e+02 2.728e+02 6.807e+02, threshold=4.219e+02, percent-clipped=3.0 2023-03-09 07:02:39,741 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=99972.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 07:02:52,054 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99980.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 07:03:15,347 INFO [train2.py:809] (1/4) Epoch 26, batch 400, loss[ctc_loss=0.06165, att_loss=0.2153, loss=0.1846, over 15621.00 frames. utt_duration=1690 frames, utt_pad_proportion=0.01029, over 37.00 utterances.], tot_loss[ctc_loss=0.06683, att_loss=0.2311, loss=0.1983, over 2822861.73 frames. utt_duration=1271 frames, utt_pad_proportion=0.05185, over 8897.85 utterances.], batch size: 37, lr: 4.16e-03, grad_scale: 8.0 2023-03-09 07:03:39,420 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100007.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 07:04:11,938 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=100028.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 07:04:24,052 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100036.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 07:04:24,243 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.8207, 2.3961, 2.5006, 2.6381, 2.8371, 2.9298, 2.4794, 2.9440], device='cuda:1'), covar=tensor([0.1509, 0.2344, 0.1731, 0.1257, 0.1408, 0.1186, 0.2096, 0.1111], device='cuda:1'), in_proj_covar=tensor([0.0133, 0.0134, 0.0130, 0.0124, 0.0142, 0.0121, 0.0144, 0.0121], device='cuda:1'), out_proj_covar=tensor([1.0212e-04, 1.0610e-04, 1.0617e-04, 9.7469e-05, 1.0663e-04, 9.7730e-05, 1.0998e-04, 9.5806e-05], device='cuda:1') 2023-03-09 07:04:38,119 INFO [train2.py:809] (1/4) Epoch 26, batch 450, loss[ctc_loss=0.04075, att_loss=0.2002, loss=0.1683, over 15872.00 frames. utt_duration=1629 frames, utt_pad_proportion=0.008937, over 39.00 utterances.], tot_loss[ctc_loss=0.06662, att_loss=0.2316, loss=0.1986, over 2928312.06 frames. utt_duration=1272 frames, utt_pad_proportion=0.04894, over 9219.41 utterances.], batch size: 39, lr: 4.16e-03, grad_scale: 8.0 2023-03-09 07:04:46,164 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.318e+02 1.836e+02 2.145e+02 2.710e+02 6.185e+02, threshold=4.291e+02, percent-clipped=5.0 2023-03-09 07:05:14,901 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100068.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 07:05:27,552 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100076.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 07:05:56,813 INFO [train2.py:809] (1/4) Epoch 26, batch 500, loss[ctc_loss=0.0703, att_loss=0.249, loss=0.2133, over 17347.00 frames. utt_duration=1177 frames, utt_pad_proportion=0.02042, over 59.00 utterances.], tot_loss[ctc_loss=0.06715, att_loss=0.2321, loss=0.1991, over 3005483.98 frames. utt_duration=1252 frames, utt_pad_proportion=0.05412, over 9616.93 utterances.], batch size: 59, lr: 4.16e-03, grad_scale: 8.0 2023-03-09 07:06:21,980 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.07 vs. limit=5.0 2023-03-09 07:06:43,578 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=100124.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 07:06:43,794 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100124.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 07:07:16,641 INFO [train2.py:809] (1/4) Epoch 26, batch 550, loss[ctc_loss=0.0738, att_loss=0.2427, loss=0.2089, over 17050.00 frames. utt_duration=1288 frames, utt_pad_proportion=0.009891, over 53.00 utterances.], tot_loss[ctc_loss=0.06757, att_loss=0.2322, loss=0.1993, over 3059574.37 frames. utt_duration=1225 frames, utt_pad_proportion=0.06239, over 10006.30 utterances.], batch size: 53, lr: 4.16e-03, grad_scale: 8.0 2023-03-09 07:07:24,150 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.301e+02 1.784e+02 2.229e+02 2.699e+02 5.667e+02, threshold=4.458e+02, percent-clipped=3.0 2023-03-09 07:08:00,400 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=100172.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 07:08:31,622 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-03-09 07:08:36,942 INFO [train2.py:809] (1/4) Epoch 26, batch 600, loss[ctc_loss=0.07399, att_loss=0.2293, loss=0.1982, over 15975.00 frames. utt_duration=1560 frames, utt_pad_proportion=0.005798, over 41.00 utterances.], tot_loss[ctc_loss=0.06843, att_loss=0.2333, loss=0.2003, over 3116829.36 frames. utt_duration=1212 frames, utt_pad_proportion=0.06193, over 10301.25 utterances.], batch size: 41, lr: 4.16e-03, grad_scale: 8.0 2023-03-09 07:09:13,959 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 2023-03-09 07:09:19,665 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.7899, 6.0549, 5.4937, 5.8030, 5.7267, 5.2175, 5.4883, 5.2632], device='cuda:1'), covar=tensor([0.1232, 0.0821, 0.1056, 0.0871, 0.0970, 0.1586, 0.2222, 0.2374], device='cuda:1'), in_proj_covar=tensor([0.0540, 0.0626, 0.0480, 0.0469, 0.0436, 0.0484, 0.0625, 0.0540], 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') 2023-03-09 07:09:23,859 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-03-09 07:09:24,646 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100224.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 07:09:47,266 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100238.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 07:09:58,240 INFO [train2.py:809] (1/4) Epoch 26, batch 650, loss[ctc_loss=0.06747, att_loss=0.2218, loss=0.1909, over 16266.00 frames. utt_duration=1515 frames, utt_pad_proportion=0.007422, over 43.00 utterances.], tot_loss[ctc_loss=0.06836, att_loss=0.233, loss=0.2001, over 3148440.40 frames. utt_duration=1212 frames, utt_pad_proportion=0.0637, over 10400.93 utterances.], batch size: 43, lr: 4.16e-03, grad_scale: 8.0 2023-03-09 07:09:58,560 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100245.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 07:10:06,971 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.335e+02 1.945e+02 2.331e+02 2.893e+02 4.719e+02, threshold=4.662e+02, percent-clipped=1.0 2023-03-09 07:10:12,291 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.0062, 2.2677, 2.3611, 2.4965, 2.7556, 2.8089, 2.4846, 3.0738], device='cuda:1'), covar=tensor([0.1378, 0.2762, 0.2065, 0.1423, 0.1875, 0.1344, 0.1942, 0.1094], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0138, 0.0134, 0.0128, 0.0145, 0.0124, 0.0147, 0.0124], device='cuda:1'), out_proj_covar=tensor([1.0446e-04, 1.0896e-04, 1.0882e-04, 1.0000e-04, 1.0934e-04, 1.0013e-04, 1.1240e-04, 9.8185e-05], device='cuda:1') 2023-03-09 07:10:43,114 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=100272.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 07:10:49,791 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1335, 4.9408, 5.1641, 2.2374, 2.0759, 2.8453, 2.5791, 3.8423], device='cuda:1'), covar=tensor([0.0884, 0.0575, 0.0268, 0.5157, 0.6368, 0.3026, 0.4015, 0.2069], device='cuda:1'), in_proj_covar=tensor([0.0362, 0.0295, 0.0278, 0.0250, 0.0343, 0.0335, 0.0262, 0.0372], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-09 07:11:16,648 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=100293.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 07:11:19,700 INFO [train2.py:809] (1/4) Epoch 26, batch 700, loss[ctc_loss=0.05461, att_loss=0.229, loss=0.1941, over 16268.00 frames. utt_duration=1515 frames, utt_pad_proportion=0.008026, over 43.00 utterances.], tot_loss[ctc_loss=0.06789, att_loss=0.2331, loss=0.2001, over 3184653.79 frames. utt_duration=1222 frames, utt_pad_proportion=0.05895, over 10440.19 utterances.], batch size: 43, lr: 4.15e-03, grad_scale: 8.0 2023-03-09 07:11:26,818 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100299.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 07:12:26,375 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100336.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 07:12:40,627 INFO [train2.py:809] (1/4) Epoch 26, batch 750, loss[ctc_loss=0.05182, att_loss=0.1975, loss=0.1683, over 15483.00 frames. utt_duration=1722 frames, utt_pad_proportion=0.009904, over 36.00 utterances.], tot_loss[ctc_loss=0.06833, att_loss=0.2329, loss=0.2, over 3198574.96 frames. utt_duration=1197 frames, utt_pad_proportion=0.06637, over 10699.54 utterances.], batch size: 36, lr: 4.15e-03, grad_scale: 8.0 2023-03-09 07:12:48,984 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.213e+02 1.813e+02 2.158e+02 2.667e+02 1.532e+03, threshold=4.316e+02, percent-clipped=4.0 2023-03-09 07:13:10,576 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100363.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 07:13:43,921 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=100384.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 07:14:01,737 INFO [train2.py:809] (1/4) Epoch 26, batch 800, loss[ctc_loss=0.0464, att_loss=0.2299, loss=0.1932, over 16550.00 frames. utt_duration=1473 frames, utt_pad_proportion=0.005582, over 45.00 utterances.], tot_loss[ctc_loss=0.06831, att_loss=0.233, loss=0.2001, over 3209533.81 frames. utt_duration=1202 frames, utt_pad_proportion=0.06615, over 10698.27 utterances.], batch size: 45, lr: 4.15e-03, grad_scale: 8.0 2023-03-09 07:14:13,180 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100402.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 07:14:23,926 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.8724, 6.1191, 5.5960, 5.8205, 5.8311, 5.2695, 5.4779, 5.3428], device='cuda:1'), covar=tensor([0.1278, 0.0934, 0.0958, 0.0804, 0.0960, 0.1701, 0.2530, 0.2271], device='cuda:1'), in_proj_covar=tensor([0.0544, 0.0629, 0.0482, 0.0471, 0.0439, 0.0487, 0.0627, 0.0545], 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') 2023-03-09 07:15:15,687 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2903, 2.3952, 3.0610, 2.5549, 3.0320, 3.4640, 3.3837, 2.6927], device='cuda:1'), covar=tensor([0.0493, 0.1715, 0.1028, 0.1235, 0.0980, 0.1122, 0.0667, 0.1116], device='cuda:1'), in_proj_covar=tensor([0.0248, 0.0250, 0.0288, 0.0223, 0.0268, 0.0378, 0.0270, 0.0234], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-09 07:15:22,221 INFO [train2.py:809] (1/4) Epoch 26, batch 850, loss[ctc_loss=0.04795, att_loss=0.2117, loss=0.1789, over 15648.00 frames. utt_duration=1693 frames, utt_pad_proportion=0.008186, over 37.00 utterances.], tot_loss[ctc_loss=0.06822, att_loss=0.2326, loss=0.1997, over 3218397.33 frames. utt_duration=1212 frames, utt_pad_proportion=0.06683, over 10636.10 utterances.], batch size: 37, lr: 4.15e-03, grad_scale: 8.0 2023-03-09 07:15:31,016 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.217e+02 1.778e+02 2.108e+02 2.517e+02 5.047e+02, threshold=4.216e+02, percent-clipped=2.0 2023-03-09 07:15:52,671 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100463.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 07:16:13,362 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2366, 2.6382, 3.0355, 4.3216, 3.9390, 3.8219, 2.8719, 2.2494], device='cuda:1'), covar=tensor([0.0823, 0.2252, 0.1116, 0.0595, 0.0968, 0.0515, 0.1572, 0.2346], device='cuda:1'), in_proj_covar=tensor([0.0187, 0.0219, 0.0188, 0.0222, 0.0233, 0.0187, 0.0205, 0.0193], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 07:16:43,810 INFO [train2.py:809] (1/4) Epoch 26, batch 900, loss[ctc_loss=0.04871, att_loss=0.2238, loss=0.1888, over 16518.00 frames. utt_duration=1470 frames, utt_pad_proportion=0.007592, over 45.00 utterances.], tot_loss[ctc_loss=0.06802, att_loss=0.2328, loss=0.1999, over 3229934.44 frames. utt_duration=1229 frames, utt_pad_proportion=0.06194, over 10524.15 utterances.], batch size: 45, lr: 4.15e-03, grad_scale: 8.0 2023-03-09 07:18:03,340 INFO [train2.py:809] (1/4) Epoch 26, batch 950, loss[ctc_loss=0.04992, att_loss=0.2085, loss=0.1768, over 15882.00 frames. utt_duration=1630 frames, utt_pad_proportion=0.008266, over 39.00 utterances.], tot_loss[ctc_loss=0.06786, att_loss=0.2325, loss=0.1995, over 3227229.28 frames. utt_duration=1218 frames, utt_pad_proportion=0.06747, over 10612.40 utterances.], batch size: 39, lr: 4.15e-03, grad_scale: 8.0 2023-03-09 07:18:11,134 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.401e+02 1.882e+02 2.194e+02 2.822e+02 5.450e+02, threshold=4.387e+02, percent-clipped=2.0 2023-03-09 07:18:24,119 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6372, 4.5631, 4.6991, 4.6579, 5.2678, 4.5727, 4.5813, 2.6645], device='cuda:1'), covar=tensor([0.0247, 0.0413, 0.0351, 0.0353, 0.0692, 0.0239, 0.0363, 0.1731], device='cuda:1'), in_proj_covar=tensor([0.0183, 0.0213, 0.0209, 0.0226, 0.0382, 0.0184, 0.0197, 0.0221], 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') 2023-03-09 07:18:43,163 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6871, 2.6435, 3.8731, 3.0782, 3.6152, 4.8678, 4.7790, 3.0525], device='cuda:1'), covar=tensor([0.0430, 0.2146, 0.0932, 0.1534, 0.0981, 0.0662, 0.0460, 0.1553], device='cuda:1'), in_proj_covar=tensor([0.0249, 0.0249, 0.0287, 0.0223, 0.0268, 0.0378, 0.0270, 0.0234], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-09 07:19:02,978 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4286, 4.3545, 4.5210, 4.4528, 5.0711, 4.3446, 4.4057, 2.4989], device='cuda:1'), covar=tensor([0.0261, 0.0451, 0.0363, 0.0390, 0.0809, 0.0282, 0.0385, 0.1823], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0213, 0.0210, 0.0227, 0.0384, 0.0185, 0.0198, 0.0222], 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') 2023-03-09 07:19:22,773 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100594.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 07:19:24,222 INFO [train2.py:809] (1/4) Epoch 26, batch 1000, loss[ctc_loss=0.07223, att_loss=0.2451, loss=0.2105, over 17374.00 frames. utt_duration=881.4 frames, utt_pad_proportion=0.07615, over 79.00 utterances.], tot_loss[ctc_loss=0.06803, att_loss=0.2324, loss=0.1995, over 3236693.73 frames. utt_duration=1223 frames, utt_pad_proportion=0.06529, over 10596.98 utterances.], batch size: 79, lr: 4.15e-03, grad_scale: 8.0 2023-03-09 07:19:57,861 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.5775, 5.8431, 5.3022, 5.5285, 5.4585, 5.0057, 5.3166, 5.0068], device='cuda:1'), covar=tensor([0.1170, 0.0915, 0.0928, 0.0884, 0.0975, 0.1650, 0.2419, 0.2450], device='cuda:1'), in_proj_covar=tensor([0.0543, 0.0629, 0.0481, 0.0473, 0.0442, 0.0486, 0.0630, 0.0545], 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') 2023-03-09 07:20:45,240 INFO [train2.py:809] (1/4) Epoch 26, batch 1050, loss[ctc_loss=0.06453, att_loss=0.2234, loss=0.1916, over 15904.00 frames. utt_duration=1633 frames, utt_pad_proportion=0.007544, over 39.00 utterances.], tot_loss[ctc_loss=0.06881, att_loss=0.2334, loss=0.2004, over 3244186.36 frames. utt_duration=1195 frames, utt_pad_proportion=0.07141, over 10876.96 utterances.], batch size: 39, lr: 4.15e-03, grad_scale: 8.0 2023-03-09 07:20:53,112 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.280e+02 1.800e+02 2.231e+02 2.686e+02 5.529e+02, threshold=4.462e+02, percent-clipped=2.0 2023-03-09 07:21:14,631 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100663.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 07:22:05,996 INFO [train2.py:809] (1/4) Epoch 26, batch 1100, loss[ctc_loss=0.06819, att_loss=0.2475, loss=0.2117, over 17296.00 frames. utt_duration=1259 frames, utt_pad_proportion=0.01149, over 55.00 utterances.], tot_loss[ctc_loss=0.06815, att_loss=0.2328, loss=0.1999, over 3239017.46 frames. utt_duration=1216 frames, utt_pad_proportion=0.06771, over 10670.38 utterances.], batch size: 55, lr: 4.15e-03, grad_scale: 8.0 2023-03-09 07:22:31,627 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=100711.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 07:22:39,725 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-03-09 07:22:44,119 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9515, 5.0062, 4.8579, 2.1856, 1.8710, 2.9444, 2.3410, 3.7455], device='cuda:1'), covar=tensor([0.0765, 0.0293, 0.0283, 0.5280, 0.5777, 0.2502, 0.3810, 0.1815], device='cuda:1'), in_proj_covar=tensor([0.0363, 0.0296, 0.0280, 0.0251, 0.0342, 0.0335, 0.0263, 0.0372], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-09 07:22:55,512 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8725, 5.1643, 4.7714, 5.2064, 4.6140, 4.8337, 5.2795, 5.0759], device='cuda:1'), covar=tensor([0.0631, 0.0311, 0.0819, 0.0375, 0.0451, 0.0332, 0.0258, 0.0197], device='cuda:1'), in_proj_covar=tensor([0.0402, 0.0336, 0.0378, 0.0370, 0.0337, 0.0246, 0.0319, 0.0301], device='cuda:1'), out_proj_covar=tensor([0.0007, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-09 07:23:01,649 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100730.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 07:23:02,660 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-03-09 07:23:25,347 INFO [train2.py:809] (1/4) Epoch 26, batch 1150, loss[ctc_loss=0.06186, att_loss=0.2236, loss=0.1913, over 16139.00 frames. utt_duration=1538 frames, utt_pad_proportion=0.005664, over 42.00 utterances.], tot_loss[ctc_loss=0.06779, att_loss=0.2325, loss=0.1996, over 3247990.66 frames. utt_duration=1233 frames, utt_pad_proportion=0.06175, over 10547.74 utterances.], batch size: 42, lr: 4.15e-03, grad_scale: 8.0 2023-03-09 07:23:32,922 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.106e+02 1.823e+02 2.156e+02 2.583e+02 8.005e+02, threshold=4.313e+02, percent-clipped=1.0 2023-03-09 07:23:40,145 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.52 vs. limit=5.0 2023-03-09 07:23:46,155 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100758.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 07:24:39,246 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100791.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 07:24:44,982 INFO [train2.py:809] (1/4) Epoch 26, batch 1200, loss[ctc_loss=0.06105, att_loss=0.2369, loss=0.2017, over 17470.00 frames. utt_duration=885.9 frames, utt_pad_proportion=0.07232, over 79.00 utterances.], tot_loss[ctc_loss=0.06739, att_loss=0.2321, loss=0.1992, over 3253092.52 frames. utt_duration=1235 frames, utt_pad_proportion=0.06108, over 10545.41 utterances.], batch size: 79, lr: 4.14e-03, grad_scale: 16.0 2023-03-09 07:25:18,269 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6968, 4.9571, 4.5680, 5.0236, 4.4144, 4.6512, 5.0537, 4.8571], device='cuda:1'), covar=tensor([0.0686, 0.0363, 0.0841, 0.0437, 0.0499, 0.0307, 0.0353, 0.0228], device='cuda:1'), in_proj_covar=tensor([0.0403, 0.0337, 0.0379, 0.0370, 0.0338, 0.0246, 0.0320, 0.0302], device='cuda:1'), out_proj_covar=tensor([0.0007, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-09 07:26:04,020 INFO [train2.py:809] (1/4) Epoch 26, batch 1250, loss[ctc_loss=0.05258, att_loss=0.2354, loss=0.1989, over 16891.00 frames. utt_duration=1380 frames, utt_pad_proportion=0.006989, over 49.00 utterances.], tot_loss[ctc_loss=0.0674, att_loss=0.2323, loss=0.1993, over 3256305.94 frames. utt_duration=1231 frames, utt_pad_proportion=0.06111, over 10595.81 utterances.], batch size: 49, lr: 4.14e-03, grad_scale: 16.0 2023-03-09 07:26:11,742 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.226e+02 1.779e+02 2.144e+02 2.683e+02 5.396e+02, threshold=4.288e+02, percent-clipped=4.0 2023-03-09 07:26:28,206 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9227, 4.7651, 4.7605, 2.2036, 2.0655, 2.7536, 2.2276, 3.9061], device='cuda:1'), covar=tensor([0.0754, 0.0276, 0.0249, 0.5117, 0.5245, 0.2580, 0.3733, 0.1374], device='cuda:1'), in_proj_covar=tensor([0.0362, 0.0296, 0.0278, 0.0250, 0.0340, 0.0334, 0.0262, 0.0370], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-03-09 07:27:22,190 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100894.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 07:27:23,459 INFO [train2.py:809] (1/4) Epoch 26, batch 1300, loss[ctc_loss=0.05331, att_loss=0.2114, loss=0.1798, over 15776.00 frames. utt_duration=1662 frames, utt_pad_proportion=0.008432, over 38.00 utterances.], tot_loss[ctc_loss=0.06756, att_loss=0.233, loss=0.1999, over 3265683.75 frames. utt_duration=1241 frames, utt_pad_proportion=0.05656, over 10541.84 utterances.], batch size: 38, lr: 4.14e-03, grad_scale: 16.0 2023-03-09 07:28:38,866 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=100942.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 07:28:43,602 INFO [train2.py:809] (1/4) Epoch 26, batch 1350, loss[ctc_loss=0.07352, att_loss=0.2459, loss=0.2115, over 17032.00 frames. utt_duration=689.6 frames, utt_pad_proportion=0.1292, over 99.00 utterances.], tot_loss[ctc_loss=0.06767, att_loss=0.233, loss=0.1999, over 3268070.94 frames. utt_duration=1240 frames, utt_pad_proportion=0.05671, over 10558.24 utterances.], batch size: 99, lr: 4.14e-03, grad_scale: 16.0 2023-03-09 07:28:51,468 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.196e+02 1.935e+02 2.207e+02 2.665e+02 5.260e+02, threshold=4.414e+02, percent-clipped=2.0 2023-03-09 07:29:55,490 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8555, 3.6437, 3.9973, 3.6039, 3.9580, 4.8696, 4.7422, 3.6819], device='cuda:1'), covar=tensor([0.0286, 0.1259, 0.0976, 0.1127, 0.0846, 0.0812, 0.0471, 0.1057], device='cuda:1'), in_proj_covar=tensor([0.0247, 0.0249, 0.0287, 0.0221, 0.0267, 0.0378, 0.0270, 0.0235], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-09 07:30:04,563 INFO [train2.py:809] (1/4) Epoch 26, batch 1400, loss[ctc_loss=0.05633, att_loss=0.2276, loss=0.1934, over 16318.00 frames. utt_duration=1452 frames, utt_pad_proportion=0.007022, over 45.00 utterances.], tot_loss[ctc_loss=0.06727, att_loss=0.2328, loss=0.1997, over 3270967.84 frames. utt_duration=1266 frames, utt_pad_proportion=0.04937, over 10346.17 utterances.], batch size: 45, lr: 4.14e-03, grad_scale: 16.0 2023-03-09 07:30:14,605 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7224, 4.9915, 4.9896, 4.9770, 4.9994, 4.9806, 4.5645, 4.4733], device='cuda:1'), covar=tensor([0.1117, 0.0586, 0.0364, 0.0577, 0.0321, 0.0339, 0.0484, 0.0359], device='cuda:1'), in_proj_covar=tensor([0.0533, 0.0382, 0.0364, 0.0376, 0.0438, 0.0447, 0.0376, 0.0409], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:1') 2023-03-09 07:30:59,628 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-03-09 07:31:24,471 INFO [train2.py:809] (1/4) Epoch 26, batch 1450, loss[ctc_loss=0.05874, att_loss=0.232, loss=0.1973, over 16386.00 frames. utt_duration=1492 frames, utt_pad_proportion=0.008281, over 44.00 utterances.], tot_loss[ctc_loss=0.06731, att_loss=0.233, loss=0.1999, over 3276420.50 frames. utt_duration=1256 frames, utt_pad_proportion=0.05061, over 10448.87 utterances.], batch size: 44, lr: 4.14e-03, grad_scale: 16.0 2023-03-09 07:31:32,130 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.264e+02 1.754e+02 2.064e+02 2.618e+02 5.187e+02, threshold=4.127e+02, percent-clipped=2.0 2023-03-09 07:31:45,731 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=101058.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 07:32:30,534 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=101086.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 07:32:33,803 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.6889, 3.4069, 3.3119, 2.9273, 3.3581, 3.3518, 3.4599, 2.4755], device='cuda:1'), covar=tensor([0.1166, 0.1197, 0.2258, 0.3220, 0.1184, 0.2763, 0.1039, 0.3353], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0199, 0.0214, 0.0264, 0.0176, 0.0275, 0.0195, 0.0226], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 07:32:44,606 INFO [train2.py:809] (1/4) Epoch 26, batch 1500, loss[ctc_loss=0.07463, att_loss=0.2281, loss=0.1974, over 15987.00 frames. utt_duration=1600 frames, utt_pad_proportion=0.007936, over 40.00 utterances.], tot_loss[ctc_loss=0.06681, att_loss=0.2324, loss=0.1993, over 3278074.45 frames. utt_duration=1275 frames, utt_pad_proportion=0.04548, over 10296.92 utterances.], batch size: 40, lr: 4.14e-03, grad_scale: 16.0 2023-03-09 07:33:02,325 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=101106.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 07:33:04,131 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101107.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 07:34:03,291 INFO [train2.py:809] (1/4) Epoch 26, batch 1550, loss[ctc_loss=0.05727, att_loss=0.2338, loss=0.1985, over 16475.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.006669, over 46.00 utterances.], tot_loss[ctc_loss=0.06702, att_loss=0.2323, loss=0.1993, over 3280137.24 frames. utt_duration=1278 frames, utt_pad_proportion=0.04438, over 10279.61 utterances.], batch size: 46, lr: 4.14e-03, grad_scale: 16.0 2023-03-09 07:34:11,011 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.086e+02 1.859e+02 2.187e+02 2.715e+02 5.591e+02, threshold=4.373e+02, percent-clipped=3.0 2023-03-09 07:34:39,802 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=101168.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 07:35:22,402 INFO [train2.py:809] (1/4) Epoch 26, batch 1600, loss[ctc_loss=0.05944, att_loss=0.238, loss=0.2023, over 16884.00 frames. utt_duration=1380 frames, utt_pad_proportion=0.007312, over 49.00 utterances.], tot_loss[ctc_loss=0.06742, att_loss=0.2331, loss=0.1999, over 3280931.04 frames. utt_duration=1250 frames, utt_pad_proportion=0.05085, over 10512.67 utterances.], batch size: 49, lr: 4.14e-03, grad_scale: 16.0 2023-03-09 07:35:48,922 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-03-09 07:36:42,605 INFO [train2.py:809] (1/4) Epoch 26, batch 1650, loss[ctc_loss=0.06006, att_loss=0.213, loss=0.1824, over 16179.00 frames. utt_duration=1580 frames, utt_pad_proportion=0.006807, over 41.00 utterances.], tot_loss[ctc_loss=0.0673, att_loss=0.2329, loss=0.1998, over 3285477.26 frames. utt_duration=1263 frames, utt_pad_proportion=0.04739, over 10420.14 utterances.], batch size: 41, lr: 4.13e-03, grad_scale: 16.0 2023-03-09 07:36:50,258 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.244e+02 1.931e+02 2.160e+02 2.616e+02 4.577e+02, threshold=4.320e+02, percent-clipped=2.0 2023-03-09 07:37:12,590 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.6872, 5.9270, 5.4365, 5.6491, 5.5794, 5.1002, 5.3662, 5.1470], device='cuda:1'), covar=tensor([0.1321, 0.0992, 0.0859, 0.0882, 0.1011, 0.1553, 0.2184, 0.2405], device='cuda:1'), in_proj_covar=tensor([0.0550, 0.0634, 0.0485, 0.0475, 0.0447, 0.0487, 0.0638, 0.0547], 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') 2023-03-09 07:38:02,610 INFO [train2.py:809] (1/4) Epoch 26, batch 1700, loss[ctc_loss=0.07004, att_loss=0.2135, loss=0.1848, over 14463.00 frames. utt_duration=1810 frames, utt_pad_proportion=0.04403, over 32.00 utterances.], tot_loss[ctc_loss=0.06706, att_loss=0.2326, loss=0.1995, over 3275388.75 frames. utt_duration=1257 frames, utt_pad_proportion=0.05124, over 10432.81 utterances.], batch size: 32, lr: 4.13e-03, grad_scale: 16.0 2023-03-09 07:39:22,763 INFO [train2.py:809] (1/4) Epoch 26, batch 1750, loss[ctc_loss=0.04316, att_loss=0.2026, loss=0.1707, over 15943.00 frames. utt_duration=1557 frames, utt_pad_proportion=0.007015, over 41.00 utterances.], tot_loss[ctc_loss=0.06716, att_loss=0.2321, loss=0.1991, over 3265479.52 frames. utt_duration=1238 frames, utt_pad_proportion=0.05857, over 10561.50 utterances.], batch size: 41, lr: 4.13e-03, grad_scale: 16.0 2023-03-09 07:39:30,365 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.455e+02 1.899e+02 2.202e+02 2.819e+02 5.865e+02, threshold=4.404e+02, percent-clipped=4.0 2023-03-09 07:39:42,120 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7592, 2.5024, 2.5576, 2.6827, 2.8788, 2.7253, 2.4473, 3.1884], device='cuda:1'), covar=tensor([0.1724, 0.2609, 0.1983, 0.1542, 0.1732, 0.1418, 0.2184, 0.1254], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0139, 0.0133, 0.0128, 0.0146, 0.0125, 0.0147, 0.0124], device='cuda:1'), out_proj_covar=tensor([1.0512e-04, 1.0984e-04, 1.0860e-04, 1.0036e-04, 1.0963e-04, 1.0063e-04, 1.1256e-04, 9.8637e-05], device='cuda:1') 2023-03-09 07:40:09,020 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4993, 2.4706, 4.9798, 3.9360, 2.9372, 4.2780, 4.7977, 4.6505], device='cuda:1'), covar=tensor([0.0319, 0.1602, 0.0216, 0.0933, 0.1767, 0.0277, 0.0186, 0.0291], device='cuda:1'), in_proj_covar=tensor([0.0225, 0.0252, 0.0217, 0.0329, 0.0275, 0.0237, 0.0206, 0.0234], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 07:40:26,892 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=101386.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 07:40:41,617 INFO [train2.py:809] (1/4) Epoch 26, batch 1800, loss[ctc_loss=0.07056, att_loss=0.2392, loss=0.2055, over 17323.00 frames. utt_duration=1176 frames, utt_pad_proportion=0.02326, over 59.00 utterances.], tot_loss[ctc_loss=0.06683, att_loss=0.2316, loss=0.1986, over 3268297.04 frames. utt_duration=1274 frames, utt_pad_proportion=0.0497, over 10269.97 utterances.], batch size: 59, lr: 4.13e-03, grad_scale: 16.0 2023-03-09 07:41:43,372 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=101434.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 07:42:01,375 INFO [train2.py:809] (1/4) Epoch 26, batch 1850, loss[ctc_loss=0.0444, att_loss=0.2206, loss=0.1853, over 16186.00 frames. utt_duration=1581 frames, utt_pad_proportion=0.006561, over 41.00 utterances.], tot_loss[ctc_loss=0.0668, att_loss=0.232, loss=0.1989, over 3270393.61 frames. utt_duration=1272 frames, utt_pad_proportion=0.0508, over 10300.35 utterances.], batch size: 41, lr: 4.13e-03, grad_scale: 16.0 2023-03-09 07:42:08,708 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.271e+02 1.881e+02 2.093e+02 3.016e+02 9.107e+02, threshold=4.186e+02, percent-clipped=4.0 2023-03-09 07:42:29,652 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=101463.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 07:43:19,786 INFO [train2.py:809] (1/4) Epoch 26, batch 1900, loss[ctc_loss=0.05265, att_loss=0.2125, loss=0.1806, over 16030.00 frames. utt_duration=1604 frames, utt_pad_proportion=0.005948, over 40.00 utterances.], tot_loss[ctc_loss=0.06721, att_loss=0.2327, loss=0.1996, over 3277341.41 frames. utt_duration=1274 frames, utt_pad_proportion=0.04773, over 10304.01 utterances.], batch size: 40, lr: 4.13e-03, grad_scale: 16.0 2023-03-09 07:44:39,310 INFO [train2.py:809] (1/4) Epoch 26, batch 1950, loss[ctc_loss=0.05102, att_loss=0.2167, loss=0.1836, over 15750.00 frames. utt_duration=1659 frames, utt_pad_proportion=0.009379, over 38.00 utterances.], tot_loss[ctc_loss=0.06741, att_loss=0.2328, loss=0.1997, over 3273929.79 frames. utt_duration=1265 frames, utt_pad_proportion=0.05021, over 10368.43 utterances.], batch size: 38, lr: 4.13e-03, grad_scale: 16.0 2023-03-09 07:44:47,406 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.265e+02 1.785e+02 2.199e+02 2.810e+02 5.538e+02, threshold=4.398e+02, percent-clipped=5.0 2023-03-09 07:45:57,906 INFO [train2.py:809] (1/4) Epoch 26, batch 2000, loss[ctc_loss=0.06282, att_loss=0.2251, loss=0.1926, over 16415.00 frames. utt_duration=1494 frames, utt_pad_proportion=0.006154, over 44.00 utterances.], tot_loss[ctc_loss=0.06648, att_loss=0.2321, loss=0.199, over 3273924.55 frames. utt_duration=1285 frames, utt_pad_proportion=0.04637, over 10203.88 utterances.], batch size: 44, lr: 4.13e-03, grad_scale: 16.0 2023-03-09 07:46:03,389 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.24 vs. limit=5.0 2023-03-09 07:47:05,504 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-03-09 07:47:17,705 INFO [train2.py:809] (1/4) Epoch 26, batch 2050, loss[ctc_loss=0.07547, att_loss=0.249, loss=0.2143, over 16475.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.005457, over 46.00 utterances.], tot_loss[ctc_loss=0.06624, att_loss=0.2321, loss=0.1989, over 3276238.75 frames. utt_duration=1263 frames, utt_pad_proportion=0.05044, over 10388.24 utterances.], batch size: 46, lr: 4.13e-03, grad_scale: 16.0 2023-03-09 07:47:26,103 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.028e+02 1.755e+02 2.108e+02 2.595e+02 4.367e+02, threshold=4.215e+02, percent-clipped=0.0 2023-03-09 07:47:50,891 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5101, 3.3702, 3.7520, 3.1179, 3.4800, 4.6190, 4.4704, 3.3967], device='cuda:1'), covar=tensor([0.0374, 0.1295, 0.1135, 0.1243, 0.1121, 0.0951, 0.0493, 0.1134], device='cuda:1'), in_proj_covar=tensor([0.0247, 0.0248, 0.0286, 0.0221, 0.0267, 0.0376, 0.0270, 0.0234], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-09 07:47:59,070 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.4117, 2.8234, 3.4508, 4.4310, 3.9173, 3.9338, 3.0432, 2.4125], device='cuda:1'), covar=tensor([0.0747, 0.2111, 0.0897, 0.0601, 0.0921, 0.0522, 0.1432, 0.2086], device='cuda:1'), in_proj_covar=tensor([0.0188, 0.0220, 0.0188, 0.0223, 0.0234, 0.0189, 0.0204, 0.0192], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 07:48:19,670 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101684.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 07:48:37,218 INFO [train2.py:809] (1/4) Epoch 26, batch 2100, loss[ctc_loss=0.0646, att_loss=0.2371, loss=0.2026, over 17033.00 frames. utt_duration=1337 frames, utt_pad_proportion=0.007118, over 51.00 utterances.], tot_loss[ctc_loss=0.06663, att_loss=0.2328, loss=0.1996, over 3285255.31 frames. utt_duration=1271 frames, utt_pad_proportion=0.04662, over 10352.47 utterances.], batch size: 51, lr: 4.13e-03, grad_scale: 16.0 2023-03-09 07:48:46,596 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1887, 5.0984, 4.9803, 3.1528, 4.9834, 4.7664, 4.5147, 3.1890], device='cuda:1'), covar=tensor([0.0109, 0.0095, 0.0261, 0.0849, 0.0090, 0.0165, 0.0268, 0.1014], device='cuda:1'), in_proj_covar=tensor([0.0078, 0.0107, 0.0110, 0.0113, 0.0089, 0.0118, 0.0103, 0.0105], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-09 07:49:26,819 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4476, 4.8697, 4.6873, 4.8577, 4.9087, 4.6809, 3.4942, 4.8028], device='cuda:1'), covar=tensor([0.0127, 0.0112, 0.0151, 0.0083, 0.0103, 0.0124, 0.0693, 0.0180], device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0094, 0.0119, 0.0074, 0.0080, 0.0092, 0.0108, 0.0113], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-09 07:49:55,491 INFO [train2.py:809] (1/4) Epoch 26, batch 2150, loss[ctc_loss=0.08038, att_loss=0.2448, loss=0.2119, over 17084.00 frames. utt_duration=1222 frames, utt_pad_proportion=0.01711, over 56.00 utterances.], tot_loss[ctc_loss=0.06692, att_loss=0.2327, loss=0.1995, over 3277237.25 frames. utt_duration=1257 frames, utt_pad_proportion=0.05034, over 10438.17 utterances.], batch size: 56, lr: 4.12e-03, grad_scale: 16.0 2023-03-09 07:49:55,859 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=101745.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 07:50:03,017 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.203e+02 1.940e+02 2.256e+02 2.810e+02 5.136e+02, threshold=4.511e+02, percent-clipped=2.0 2023-03-09 07:50:23,092 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=101763.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 07:51:13,848 INFO [train2.py:809] (1/4) Epoch 26, batch 2200, loss[ctc_loss=0.05485, att_loss=0.2073, loss=0.1768, over 15636.00 frames. utt_duration=1692 frames, utt_pad_proportion=0.009494, over 37.00 utterances.], tot_loss[ctc_loss=0.06694, att_loss=0.2326, loss=0.1995, over 3280651.91 frames. utt_duration=1258 frames, utt_pad_proportion=0.0493, over 10440.98 utterances.], batch size: 37, lr: 4.12e-03, grad_scale: 16.0 2023-03-09 07:51:33,262 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-03-09 07:51:38,367 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=101811.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 07:51:54,474 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2128, 4.2885, 4.4249, 4.4794, 4.9865, 4.3346, 4.2913, 2.4528], device='cuda:1'), covar=tensor([0.0299, 0.0474, 0.0380, 0.0383, 0.0685, 0.0278, 0.0443, 0.1833], device='cuda:1'), in_proj_covar=tensor([0.0184, 0.0212, 0.0207, 0.0224, 0.0379, 0.0183, 0.0197, 0.0219], 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') 2023-03-09 07:52:32,210 INFO [train2.py:809] (1/4) Epoch 26, batch 2250, loss[ctc_loss=0.05176, att_loss=0.2115, loss=0.1796, over 15773.00 frames. utt_duration=1662 frames, utt_pad_proportion=0.007965, over 38.00 utterances.], tot_loss[ctc_loss=0.06686, att_loss=0.2331, loss=0.1998, over 3280552.91 frames. utt_duration=1261 frames, utt_pad_proportion=0.04822, over 10419.19 utterances.], batch size: 38, lr: 4.12e-03, grad_scale: 16.0 2023-03-09 07:52:39,681 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.215e+02 1.832e+02 2.219e+02 2.925e+02 8.188e+02, threshold=4.439e+02, percent-clipped=2.0 2023-03-09 07:52:44,835 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3341, 2.6104, 4.7132, 3.8219, 3.2332, 4.1827, 4.4011, 4.5006], device='cuda:1'), covar=tensor([0.0269, 0.1555, 0.0192, 0.0876, 0.1410, 0.0267, 0.0254, 0.0249], device='cuda:1'), in_proj_covar=tensor([0.0221, 0.0245, 0.0213, 0.0321, 0.0269, 0.0232, 0.0203, 0.0229], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 07:53:51,084 INFO [train2.py:809] (1/4) Epoch 26, batch 2300, loss[ctc_loss=0.06621, att_loss=0.2427, loss=0.2074, over 17298.00 frames. utt_duration=1174 frames, utt_pad_proportion=0.02378, over 59.00 utterances.], tot_loss[ctc_loss=0.06638, att_loss=0.2328, loss=0.1996, over 3282967.14 frames. utt_duration=1256 frames, utt_pad_proportion=0.0483, over 10465.79 utterances.], batch size: 59, lr: 4.12e-03, grad_scale: 16.0 2023-03-09 07:54:30,199 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101920.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 07:54:59,229 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7452, 3.3401, 3.9474, 3.4089, 3.6669, 4.7435, 4.6854, 3.6522], device='cuda:1'), covar=tensor([0.0316, 0.1483, 0.1102, 0.1143, 0.1064, 0.1099, 0.0444, 0.1012], device='cuda:1'), in_proj_covar=tensor([0.0247, 0.0249, 0.0286, 0.0222, 0.0268, 0.0378, 0.0271, 0.0234], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-09 07:55:09,346 INFO [train2.py:809] (1/4) Epoch 26, batch 2350, loss[ctc_loss=0.07307, att_loss=0.2199, loss=0.1905, over 16006.00 frames. utt_duration=1602 frames, utt_pad_proportion=0.007327, over 40.00 utterances.], tot_loss[ctc_loss=0.0667, att_loss=0.2329, loss=0.1997, over 3279980.26 frames. utt_duration=1256 frames, utt_pad_proportion=0.05054, over 10456.11 utterances.], batch size: 40, lr: 4.12e-03, grad_scale: 16.0 2023-03-09 07:55:16,746 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.309e+02 1.850e+02 2.322e+02 2.866e+02 7.956e+02, threshold=4.644e+02, percent-clipped=4.0 2023-03-09 07:55:17,255 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4465, 4.4902, 4.6060, 4.5645, 5.1301, 4.5288, 4.4479, 2.6681], device='cuda:1'), covar=tensor([0.0292, 0.0378, 0.0353, 0.0440, 0.0639, 0.0260, 0.0396, 0.1724], device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0213, 0.0207, 0.0225, 0.0380, 0.0184, 0.0198, 0.0220], 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') 2023-03-09 07:55:41,434 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-03-09 07:56:05,577 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=101981.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 07:56:15,048 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101987.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 07:56:27,652 INFO [train2.py:809] (1/4) Epoch 26, batch 2400, loss[ctc_loss=0.06494, att_loss=0.2397, loss=0.2048, over 17082.00 frames. utt_duration=684.7 frames, utt_pad_proportion=0.1344, over 100.00 utterances.], tot_loss[ctc_loss=0.0676, att_loss=0.2337, loss=0.2005, over 3285220.22 frames. utt_duration=1226 frames, utt_pad_proportion=0.05669, over 10732.09 utterances.], batch size: 100, lr: 4.12e-03, grad_scale: 16.0 2023-03-09 07:57:44,271 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=102040.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 07:57:51,788 INFO [train2.py:809] (1/4) Epoch 26, batch 2450, loss[ctc_loss=0.05549, att_loss=0.22, loss=0.1871, over 16190.00 frames. utt_duration=1581 frames, utt_pad_proportion=0.005691, over 41.00 utterances.], tot_loss[ctc_loss=0.06825, att_loss=0.2343, loss=0.2011, over 3286000.23 frames. utt_duration=1213 frames, utt_pad_proportion=0.06007, over 10846.43 utterances.], batch size: 41, lr: 4.12e-03, grad_scale: 16.0 2023-03-09 07:57:53,841 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2760, 4.3417, 4.4989, 4.4657, 4.9916, 4.3257, 4.3049, 2.6666], device='cuda:1'), covar=tensor([0.0314, 0.0455, 0.0385, 0.0367, 0.0656, 0.0285, 0.0402, 0.1690], device='cuda:1'), in_proj_covar=tensor([0.0185, 0.0213, 0.0208, 0.0225, 0.0380, 0.0184, 0.0198, 0.0220], 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') 2023-03-09 07:57:56,710 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=102048.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 07:57:59,372 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.331e+02 1.902e+02 2.246e+02 2.761e+02 8.723e+02, threshold=4.491e+02, percent-clipped=3.0 2023-03-09 07:59:11,508 INFO [train2.py:809] (1/4) Epoch 26, batch 2500, loss[ctc_loss=0.08551, att_loss=0.2556, loss=0.2216, over 13747.00 frames. utt_duration=377.9 frames, utt_pad_proportion=0.3416, over 146.00 utterances.], tot_loss[ctc_loss=0.06747, att_loss=0.2333, loss=0.2002, over 3267401.78 frames. utt_duration=1242 frames, utt_pad_proportion=0.05711, over 10539.72 utterances.], batch size: 146, lr: 4.12e-03, grad_scale: 16.0 2023-03-09 08:00:29,995 INFO [train2.py:809] (1/4) Epoch 26, batch 2550, loss[ctc_loss=0.05163, att_loss=0.2231, loss=0.1888, over 15966.00 frames. utt_duration=1559 frames, utt_pad_proportion=0.006311, over 41.00 utterances.], tot_loss[ctc_loss=0.06737, att_loss=0.2334, loss=0.2002, over 3266651.13 frames. utt_duration=1234 frames, utt_pad_proportion=0.0596, over 10602.73 utterances.], batch size: 41, lr: 4.12e-03, grad_scale: 16.0 2023-03-09 08:00:37,741 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.009e+02 1.703e+02 2.113e+02 2.564e+02 5.440e+02, threshold=4.227e+02, percent-clipped=2.0 2023-03-09 08:01:48,436 INFO [train2.py:809] (1/4) Epoch 26, batch 2600, loss[ctc_loss=0.0572, att_loss=0.2113, loss=0.1805, over 15490.00 frames. utt_duration=1723 frames, utt_pad_proportion=0.009376, over 36.00 utterances.], tot_loss[ctc_loss=0.06695, att_loss=0.2332, loss=0.2, over 3265124.73 frames. utt_duration=1241 frames, utt_pad_proportion=0.05757, over 10539.52 utterances.], batch size: 36, lr: 4.12e-03, grad_scale: 16.0 2023-03-09 08:02:09,272 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.4834, 2.9481, 3.4339, 4.3738, 3.9784, 3.9499, 2.9850, 2.3589], device='cuda:1'), covar=tensor([0.0718, 0.1956, 0.0904, 0.0600, 0.0900, 0.0534, 0.1513, 0.2187], device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0220, 0.0189, 0.0224, 0.0234, 0.0189, 0.0205, 0.0192], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 08:03:06,804 INFO [train2.py:809] (1/4) Epoch 26, batch 2650, loss[ctc_loss=0.07939, att_loss=0.2408, loss=0.2086, over 16916.00 frames. utt_duration=684.9 frames, utt_pad_proportion=0.1407, over 99.00 utterances.], tot_loss[ctc_loss=0.06683, att_loss=0.2328, loss=0.1996, over 3269031.55 frames. utt_duration=1250 frames, utt_pad_proportion=0.05451, over 10472.99 utterances.], batch size: 99, lr: 4.11e-03, grad_scale: 16.0 2023-03-09 08:03:14,512 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.238e+02 1.711e+02 1.973e+02 2.429e+02 4.776e+02, threshold=3.946e+02, percent-clipped=2.0 2023-03-09 08:03:33,898 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4673, 3.0619, 3.7074, 3.1967, 3.5877, 4.5984, 4.4643, 3.3539], device='cuda:1'), covar=tensor([0.0386, 0.1619, 0.1123, 0.1262, 0.1007, 0.0775, 0.0495, 0.1138], device='cuda:1'), in_proj_covar=tensor([0.0249, 0.0250, 0.0289, 0.0223, 0.0269, 0.0380, 0.0272, 0.0236], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-09 08:03:55,843 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=102276.0, num_to_drop=1, layers_to_drop={3} 2023-03-09 08:04:18,255 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.2250, 5.4829, 5.4223, 5.4281, 5.5186, 5.4568, 5.0815, 4.8979], device='cuda:1'), covar=tensor([0.0953, 0.0504, 0.0294, 0.0440, 0.0264, 0.0276, 0.0405, 0.0352], device='cuda:1'), in_proj_covar=tensor([0.0529, 0.0376, 0.0364, 0.0374, 0.0435, 0.0444, 0.0375, 0.0406], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004], device='cuda:1') 2023-03-09 08:04:25,661 INFO [train2.py:809] (1/4) Epoch 26, batch 2700, loss[ctc_loss=0.05024, att_loss=0.1995, loss=0.1697, over 15529.00 frames. utt_duration=1727 frames, utt_pad_proportion=0.006837, over 36.00 utterances.], tot_loss[ctc_loss=0.06652, att_loss=0.2327, loss=0.1995, over 3272368.26 frames. utt_duration=1250 frames, utt_pad_proportion=0.05416, over 10482.46 utterances.], batch size: 36, lr: 4.11e-03, grad_scale: 16.0 2023-03-09 08:05:37,438 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102340.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:05:41,915 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=102343.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:05:44,682 INFO [train2.py:809] (1/4) Epoch 26, batch 2750, loss[ctc_loss=0.05544, att_loss=0.2448, loss=0.2069, over 17030.00 frames. utt_duration=1337 frames, utt_pad_proportion=0.007307, over 51.00 utterances.], tot_loss[ctc_loss=0.06625, att_loss=0.2329, loss=0.1996, over 3280016.15 frames. utt_duration=1265 frames, utt_pad_proportion=0.04844, over 10384.41 utterances.], batch size: 51, lr: 4.11e-03, grad_scale: 16.0 2023-03-09 08:05:52,186 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.173e+02 1.911e+02 2.234e+02 2.672e+02 6.955e+02, threshold=4.468e+02, percent-clipped=5.0 2023-03-09 08:06:52,286 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=102388.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:06:55,604 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7526, 2.4836, 5.1514, 4.1795, 3.2344, 4.5602, 4.9178, 4.8723], device='cuda:1'), covar=tensor([0.0213, 0.1449, 0.0171, 0.0757, 0.1453, 0.0193, 0.0127, 0.0195], device='cuda:1'), in_proj_covar=tensor([0.0222, 0.0247, 0.0215, 0.0323, 0.0271, 0.0234, 0.0206, 0.0232], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 08:07:02,736 INFO [train2.py:809] (1/4) Epoch 26, batch 2800, loss[ctc_loss=0.06793, att_loss=0.2483, loss=0.2122, over 16769.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.006348, over 48.00 utterances.], tot_loss[ctc_loss=0.0656, att_loss=0.2323, loss=0.199, over 3277396.98 frames. utt_duration=1276 frames, utt_pad_proportion=0.04658, over 10282.37 utterances.], batch size: 48, lr: 4.11e-03, grad_scale: 16.0 2023-03-09 08:08:18,159 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-03-09 08:08:21,715 INFO [train2.py:809] (1/4) Epoch 26, batch 2850, loss[ctc_loss=0.0715, att_loss=0.2206, loss=0.1908, over 15998.00 frames. utt_duration=1601 frames, utt_pad_proportion=0.007884, over 40.00 utterances.], tot_loss[ctc_loss=0.06612, att_loss=0.2323, loss=0.199, over 3274869.06 frames. utt_duration=1262 frames, utt_pad_proportion=0.0507, over 10390.80 utterances.], batch size: 40, lr: 4.11e-03, grad_scale: 16.0 2023-03-09 08:08:29,301 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.109e+02 1.936e+02 2.281e+02 2.866e+02 5.493e+02, threshold=4.561e+02, percent-clipped=4.0 2023-03-09 08:08:51,710 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-03-09 08:08:52,838 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.75 vs. limit=2.0 2023-03-09 08:09:40,753 INFO [train2.py:809] (1/4) Epoch 26, batch 2900, loss[ctc_loss=0.07254, att_loss=0.2454, loss=0.2108, over 17270.00 frames. utt_duration=1258 frames, utt_pad_proportion=0.01359, over 55.00 utterances.], tot_loss[ctc_loss=0.0662, att_loss=0.2327, loss=0.1994, over 3279473.45 frames. utt_duration=1258 frames, utt_pad_proportion=0.05131, over 10438.90 utterances.], batch size: 55, lr: 4.11e-03, grad_scale: 16.0 2023-03-09 08:11:00,386 INFO [train2.py:809] (1/4) Epoch 26, batch 2950, loss[ctc_loss=0.0749, att_loss=0.2541, loss=0.2183, over 17461.00 frames. utt_duration=1014 frames, utt_pad_proportion=0.04446, over 69.00 utterances.], tot_loss[ctc_loss=0.06688, att_loss=0.2337, loss=0.2003, over 3289687.04 frames. utt_duration=1250 frames, utt_pad_proportion=0.05002, over 10541.57 utterances.], batch size: 69, lr: 4.11e-03, grad_scale: 16.0 2023-03-09 08:11:01,614 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-03-09 08:11:08,373 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.205e+02 1.971e+02 2.322e+02 2.777e+02 5.083e+02, threshold=4.643e+02, percent-clipped=3.0 2023-03-09 08:11:33,075 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.9122, 3.7109, 3.7395, 3.2743, 3.6604, 3.7274, 3.8094, 3.0012], device='cuda:1'), covar=tensor([0.0908, 0.1071, 0.1365, 0.2527, 0.1285, 0.1579, 0.0771, 0.2470], device='cuda:1'), in_proj_covar=tensor([0.0198, 0.0203, 0.0218, 0.0269, 0.0178, 0.0279, 0.0200, 0.0229], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 08:11:54,084 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102576.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 08:12:24,263 INFO [train2.py:809] (1/4) Epoch 26, batch 3000, loss[ctc_loss=0.07192, att_loss=0.2492, loss=0.2137, over 17307.00 frames. utt_duration=1260 frames, utt_pad_proportion=0.0108, over 55.00 utterances.], tot_loss[ctc_loss=0.06607, att_loss=0.2325, loss=0.1992, over 3288787.63 frames. utt_duration=1280 frames, utt_pad_proportion=0.04327, over 10289.95 utterances.], batch size: 55, lr: 4.11e-03, grad_scale: 16.0 2023-03-09 08:12:24,264 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-09 08:12:38,600 INFO [train2.py:843] (1/4) Epoch 26, validation: ctc_loss=0.04046, att_loss=0.2348, loss=0.1959, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-09 08:12:38,600 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-09 08:13:13,543 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0107, 4.9952, 4.9020, 2.3411, 1.9876, 2.9395, 2.3522, 3.9428], device='cuda:1'), covar=tensor([0.0759, 0.0345, 0.0260, 0.4701, 0.5662, 0.2407, 0.3786, 0.1524], device='cuda:1'), in_proj_covar=tensor([0.0358, 0.0294, 0.0276, 0.0248, 0.0336, 0.0328, 0.0259, 0.0368], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-03-09 08:13:27,791 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=102624.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 08:14:00,124 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102643.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:14:03,012 INFO [train2.py:809] (1/4) Epoch 26, batch 3050, loss[ctc_loss=0.07037, att_loss=0.2491, loss=0.2133, over 16607.00 frames. utt_duration=1415 frames, utt_pad_proportion=0.00626, over 47.00 utterances.], tot_loss[ctc_loss=0.06588, att_loss=0.2319, loss=0.1987, over 3288281.24 frames. utt_duration=1296 frames, utt_pad_proportion=0.03919, over 10159.16 utterances.], batch size: 47, lr: 4.11e-03, grad_scale: 16.0 2023-03-09 08:14:10,826 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.303e+02 1.917e+02 2.316e+02 2.727e+02 9.288e+02, threshold=4.632e+02, percent-clipped=2.0 2023-03-09 08:14:39,087 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-03-09 08:15:18,663 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=102691.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:15:24,943 INFO [train2.py:809] (1/4) Epoch 26, batch 3100, loss[ctc_loss=0.06121, att_loss=0.2303, loss=0.1965, over 17009.00 frames. utt_duration=1336 frames, utt_pad_proportion=0.008385, over 51.00 utterances.], tot_loss[ctc_loss=0.06623, att_loss=0.2314, loss=0.1984, over 3280602.87 frames. utt_duration=1309 frames, utt_pad_proportion=0.03707, over 10039.25 utterances.], batch size: 51, lr: 4.11e-03, grad_scale: 16.0 2023-03-09 08:16:47,153 INFO [train2.py:809] (1/4) Epoch 26, batch 3150, loss[ctc_loss=0.07, att_loss=0.2093, loss=0.1815, over 15631.00 frames. utt_duration=1692 frames, utt_pad_proportion=0.009621, over 37.00 utterances.], tot_loss[ctc_loss=0.06576, att_loss=0.2312, loss=0.1981, over 3276296.30 frames. utt_duration=1285 frames, utt_pad_proportion=0.04345, over 10208.71 utterances.], batch size: 37, lr: 4.10e-03, grad_scale: 16.0 2023-03-09 08:16:54,911 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.334e+02 1.851e+02 2.183e+02 2.802e+02 6.228e+02, threshold=4.366e+02, percent-clipped=1.0 2023-03-09 08:16:57,369 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-03-09 08:18:06,655 INFO [train2.py:809] (1/4) Epoch 26, batch 3200, loss[ctc_loss=0.0412, att_loss=0.2072, loss=0.174, over 15949.00 frames. utt_duration=1557 frames, utt_pad_proportion=0.007415, over 41.00 utterances.], tot_loss[ctc_loss=0.0658, att_loss=0.2314, loss=0.1983, over 3279713.68 frames. utt_duration=1297 frames, utt_pad_proportion=0.03917, over 10124.18 utterances.], batch size: 41, lr: 4.10e-03, grad_scale: 32.0 2023-03-09 08:18:08,489 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0322, 6.2553, 5.7132, 5.8955, 5.8872, 5.3604, 5.7294, 5.4197], device='cuda:1'), covar=tensor([0.1225, 0.0784, 0.0967, 0.0817, 0.0988, 0.1479, 0.2035, 0.2206], device='cuda:1'), in_proj_covar=tensor([0.0557, 0.0637, 0.0493, 0.0484, 0.0453, 0.0490, 0.0641, 0.0549], 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') 2023-03-09 08:18:11,850 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.0441, 3.7142, 3.0737, 3.4040, 3.9373, 3.5441, 3.1188, 4.2083], device='cuda:1'), covar=tensor([0.1014, 0.0532, 0.1122, 0.0752, 0.0770, 0.0845, 0.0817, 0.0445], device='cuda:1'), in_proj_covar=tensor([0.0207, 0.0224, 0.0230, 0.0207, 0.0290, 0.0248, 0.0203, 0.0295], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-09 08:19:00,007 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.3948, 5.6911, 5.1480, 5.4356, 5.3511, 4.8276, 5.0931, 4.8891], device='cuda:1'), covar=tensor([0.1449, 0.1003, 0.1033, 0.0927, 0.0954, 0.1630, 0.2674, 0.2424], device='cuda:1'), in_proj_covar=tensor([0.0558, 0.0639, 0.0494, 0.0485, 0.0454, 0.0492, 0.0643, 0.0551], 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') 2023-03-09 08:19:25,813 INFO [train2.py:809] (1/4) Epoch 26, batch 3250, loss[ctc_loss=0.05281, att_loss=0.2304, loss=0.1949, over 16993.00 frames. utt_duration=1334 frames, utt_pad_proportion=0.008018, over 51.00 utterances.], tot_loss[ctc_loss=0.06555, att_loss=0.2315, loss=0.1983, over 3279523.71 frames. utt_duration=1272 frames, utt_pad_proportion=0.04514, over 10325.23 utterances.], batch size: 51, lr: 4.10e-03, grad_scale: 32.0 2023-03-09 08:19:33,382 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.262e+02 1.810e+02 2.101e+02 2.825e+02 6.306e+02, threshold=4.202e+02, percent-clipped=5.0 2023-03-09 08:19:33,686 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9325, 5.2028, 4.8316, 5.2407, 4.6407, 4.9216, 5.3693, 5.1589], device='cuda:1'), covar=tensor([0.0645, 0.0318, 0.0766, 0.0328, 0.0453, 0.0262, 0.0222, 0.0208], device='cuda:1'), in_proj_covar=tensor([0.0398, 0.0335, 0.0377, 0.0368, 0.0336, 0.0244, 0.0315, 0.0299], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-09 08:19:35,346 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=102851.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:20:32,523 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=102887.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:20:44,488 INFO [train2.py:809] (1/4) Epoch 26, batch 3300, loss[ctc_loss=0.05986, att_loss=0.2361, loss=0.2008, over 16885.00 frames. utt_duration=1380 frames, utt_pad_proportion=0.006612, over 49.00 utterances.], tot_loss[ctc_loss=0.06523, att_loss=0.2308, loss=0.1977, over 3270887.55 frames. utt_duration=1274 frames, utt_pad_proportion=0.04816, over 10284.05 utterances.], batch size: 49, lr: 4.10e-03, grad_scale: 32.0 2023-03-09 08:20:57,599 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-03-09 08:21:11,627 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=102912.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:22:03,745 INFO [train2.py:809] (1/4) Epoch 26, batch 3350, loss[ctc_loss=0.05903, att_loss=0.2393, loss=0.2033, over 16891.00 frames. utt_duration=1380 frames, utt_pad_proportion=0.006886, over 49.00 utterances.], tot_loss[ctc_loss=0.06579, att_loss=0.231, loss=0.198, over 3273330.15 frames. utt_duration=1273 frames, utt_pad_proportion=0.04912, over 10301.62 utterances.], batch size: 49, lr: 4.10e-03, grad_scale: 32.0 2023-03-09 08:22:07,227 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5012, 2.7230, 4.9904, 3.8939, 3.2050, 4.2988, 4.7729, 4.7025], device='cuda:1'), covar=tensor([0.0297, 0.1507, 0.0232, 0.0898, 0.1573, 0.0255, 0.0188, 0.0264], device='cuda:1'), in_proj_covar=tensor([0.0222, 0.0247, 0.0215, 0.0324, 0.0271, 0.0234, 0.0207, 0.0232], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 08:22:08,714 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=102948.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:22:11,378 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.207e+02 1.718e+02 2.165e+02 2.625e+02 3.837e+02, threshold=4.330e+02, percent-clipped=0.0 2023-03-09 08:23:22,225 INFO [train2.py:809] (1/4) Epoch 26, batch 3400, loss[ctc_loss=0.07049, att_loss=0.2451, loss=0.2101, over 16630.00 frames. utt_duration=1417 frames, utt_pad_proportion=0.005095, over 47.00 utterances.], tot_loss[ctc_loss=0.06599, att_loss=0.2316, loss=0.1985, over 3275464.48 frames. utt_duration=1264 frames, utt_pad_proportion=0.04927, over 10374.06 utterances.], batch size: 47, lr: 4.10e-03, grad_scale: 16.0 2023-03-09 08:23:50,433 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.20 vs. limit=5.0 2023-03-09 08:24:11,101 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0559, 5.2852, 5.2524, 5.2317, 5.3038, 5.2871, 4.9333, 4.7675], device='cuda:1'), covar=tensor([0.1074, 0.0584, 0.0323, 0.0479, 0.0315, 0.0329, 0.0386, 0.0341], device='cuda:1'), in_proj_covar=tensor([0.0533, 0.0377, 0.0367, 0.0376, 0.0439, 0.0444, 0.0375, 0.0410], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:1') 2023-03-09 08:24:24,602 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2571, 4.4290, 4.7571, 4.7686, 3.1414, 4.4593, 3.3858, 1.9664], device='cuda:1'), covar=tensor([0.0505, 0.0351, 0.0544, 0.0232, 0.1385, 0.0254, 0.1054, 0.1631], device='cuda:1'), in_proj_covar=tensor([0.0213, 0.0185, 0.0266, 0.0177, 0.0225, 0.0165, 0.0234, 0.0207], 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') 2023-03-09 08:24:40,908 INFO [train2.py:809] (1/4) Epoch 26, batch 3450, loss[ctc_loss=0.08021, att_loss=0.2381, loss=0.2065, over 15958.00 frames. utt_duration=1558 frames, utt_pad_proportion=0.006238, over 41.00 utterances.], tot_loss[ctc_loss=0.06606, att_loss=0.2318, loss=0.1987, over 3272154.24 frames. utt_duration=1279 frames, utt_pad_proportion=0.04707, over 10242.02 utterances.], batch size: 41, lr: 4.10e-03, grad_scale: 16.0 2023-03-09 08:24:50,036 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.313e+02 1.811e+02 2.239e+02 2.771e+02 6.342e+02, threshold=4.479e+02, percent-clipped=4.0 2023-03-09 08:26:00,924 INFO [train2.py:809] (1/4) Epoch 26, batch 3500, loss[ctc_loss=0.08377, att_loss=0.2472, loss=0.2145, over 16769.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.006466, over 48.00 utterances.], tot_loss[ctc_loss=0.06648, att_loss=0.2322, loss=0.1991, over 3269727.95 frames. utt_duration=1254 frames, utt_pad_proportion=0.05331, over 10439.08 utterances.], batch size: 48, lr: 4.10e-03, grad_scale: 16.0 2023-03-09 08:26:04,127 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0959, 5.3865, 5.6633, 5.3645, 5.5958, 6.0638, 5.2763, 6.1405], device='cuda:1'), covar=tensor([0.0657, 0.0729, 0.0744, 0.1488, 0.1660, 0.0809, 0.0780, 0.0648], device='cuda:1'), in_proj_covar=tensor([0.0915, 0.0526, 0.0632, 0.0687, 0.0905, 0.0658, 0.0511, 0.0645], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 08:26:05,870 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103098.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:26:59,851 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103132.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:27:15,619 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9267, 4.0330, 4.0225, 4.3837, 2.6564, 4.1970, 2.8603, 2.0801], device='cuda:1'), covar=tensor([0.0559, 0.0298, 0.0719, 0.0240, 0.1779, 0.0256, 0.1394, 0.1675], device='cuda:1'), in_proj_covar=tensor([0.0212, 0.0184, 0.0263, 0.0176, 0.0223, 0.0164, 0.0232, 0.0206], 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') 2023-03-09 08:27:19,712 INFO [train2.py:809] (1/4) Epoch 26, batch 3550, loss[ctc_loss=0.07248, att_loss=0.2485, loss=0.2133, over 17132.00 frames. utt_duration=1225 frames, utt_pad_proportion=0.01414, over 56.00 utterances.], tot_loss[ctc_loss=0.06652, att_loss=0.2324, loss=0.1992, over 3271399.90 frames. utt_duration=1255 frames, utt_pad_proportion=0.05372, over 10442.15 utterances.], batch size: 56, lr: 4.10e-03, grad_scale: 16.0 2023-03-09 08:27:28,876 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.344e+02 1.850e+02 2.221e+02 2.819e+02 5.198e+02, threshold=4.442e+02, percent-clipped=1.0 2023-03-09 08:27:42,864 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103159.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:28:29,942 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.5708, 2.4651, 2.2703, 2.4896, 2.6064, 2.5486, 2.3114, 2.9515], device='cuda:1'), covar=tensor([0.1513, 0.2122, 0.2025, 0.1384, 0.1718, 0.1179, 0.2067, 0.1059], device='cuda:1'), in_proj_covar=tensor([0.0136, 0.0140, 0.0135, 0.0130, 0.0146, 0.0125, 0.0148, 0.0124], device='cuda:1'), out_proj_covar=tensor([1.0522e-04, 1.1070e-04, 1.1000e-04, 1.0137e-04, 1.1024e-04, 1.0073e-04, 1.1310e-04, 9.8431e-05], device='cuda:1') 2023-03-09 08:28:36,287 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103193.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:28:38,762 INFO [train2.py:809] (1/4) Epoch 26, batch 3600, loss[ctc_loss=0.08965, att_loss=0.2553, loss=0.2222, over 17345.00 frames. utt_duration=1177 frames, utt_pad_proportion=0.02202, over 59.00 utterances.], tot_loss[ctc_loss=0.06648, att_loss=0.2328, loss=0.1996, over 3270088.87 frames. utt_duration=1263 frames, utt_pad_proportion=0.05153, over 10372.39 utterances.], batch size: 59, lr: 4.10e-03, grad_scale: 16.0 2023-03-09 08:28:57,789 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103207.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:29:55,728 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103243.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:29:58,626 INFO [train2.py:809] (1/4) Epoch 26, batch 3650, loss[ctc_loss=0.05401, att_loss=0.2043, loss=0.1743, over 15783.00 frames. utt_duration=1663 frames, utt_pad_proportion=0.007882, over 38.00 utterances.], tot_loss[ctc_loss=0.06632, att_loss=0.2319, loss=0.1988, over 3258771.90 frames. utt_duration=1248 frames, utt_pad_proportion=0.05889, over 10458.58 utterances.], batch size: 38, lr: 4.09e-03, grad_scale: 16.0 2023-03-09 08:29:58,897 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8400, 5.0610, 4.9894, 4.9986, 5.0738, 5.0548, 4.6753, 4.5346], device='cuda:1'), covar=tensor([0.1039, 0.0591, 0.0382, 0.0592, 0.0307, 0.0351, 0.0465, 0.0366], device='cuda:1'), in_proj_covar=tensor([0.0530, 0.0376, 0.0367, 0.0375, 0.0438, 0.0444, 0.0374, 0.0409], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004], device='cuda:1') 2023-03-09 08:30:07,795 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.026e+02 1.817e+02 2.258e+02 2.684e+02 6.066e+02, threshold=4.515e+02, percent-clipped=3.0 2023-03-09 08:30:18,820 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0474, 2.6043, 4.3492, 3.5843, 2.8965, 3.9024, 3.9923, 4.0876], device='cuda:1'), covar=tensor([0.0313, 0.1500, 0.0215, 0.0943, 0.1691, 0.0337, 0.0324, 0.0366], device='cuda:1'), in_proj_covar=tensor([0.0223, 0.0247, 0.0216, 0.0324, 0.0272, 0.0235, 0.0207, 0.0234], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 08:31:17,482 INFO [train2.py:809] (1/4) Epoch 26, batch 3700, loss[ctc_loss=0.066, att_loss=0.237, loss=0.2028, over 17004.00 frames. utt_duration=1335 frames, utt_pad_proportion=0.008734, over 51.00 utterances.], tot_loss[ctc_loss=0.06611, att_loss=0.2324, loss=0.1991, over 3268905.93 frames. utt_duration=1223 frames, utt_pad_proportion=0.06081, over 10706.77 utterances.], batch size: 51, lr: 4.09e-03, grad_scale: 16.0 2023-03-09 08:31:55,555 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5501, 5.0286, 4.8511, 4.9700, 5.0971, 4.6532, 3.6092, 5.0212], device='cuda:1'), covar=tensor([0.0138, 0.0118, 0.0143, 0.0098, 0.0095, 0.0166, 0.0663, 0.0190], device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0095, 0.0119, 0.0074, 0.0080, 0.0092, 0.0108, 0.0113], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-09 08:32:30,491 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5611, 5.0896, 4.8399, 5.0309, 5.1862, 4.7503, 3.5792, 4.9634], device='cuda:1'), covar=tensor([0.0137, 0.0101, 0.0139, 0.0084, 0.0075, 0.0106, 0.0678, 0.0169], device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0095, 0.0119, 0.0074, 0.0080, 0.0091, 0.0108, 0.0113], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-09 08:32:36,354 INFO [train2.py:809] (1/4) Epoch 26, batch 3750, loss[ctc_loss=0.05869, att_loss=0.2199, loss=0.1877, over 15620.00 frames. utt_duration=1690 frames, utt_pad_proportion=0.01049, over 37.00 utterances.], tot_loss[ctc_loss=0.06598, att_loss=0.2322, loss=0.1989, over 3265892.53 frames. utt_duration=1237 frames, utt_pad_proportion=0.05682, over 10572.06 utterances.], batch size: 37, lr: 4.09e-03, grad_scale: 16.0 2023-03-09 08:32:45,490 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.324e+02 1.935e+02 2.309e+02 3.024e+02 8.462e+02, threshold=4.618e+02, percent-clipped=2.0 2023-03-09 08:33:10,139 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103366.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 08:33:11,620 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.8126, 3.6142, 3.6418, 3.1170, 3.5411, 3.6179, 3.6336, 2.6205], device='cuda:1'), covar=tensor([0.1016, 0.1009, 0.1504, 0.2914, 0.1187, 0.2513, 0.0836, 0.3260], device='cuda:1'), in_proj_covar=tensor([0.0199, 0.0204, 0.0220, 0.0270, 0.0180, 0.0281, 0.0201, 0.0229], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 08:33:54,195 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-03-09 08:33:56,328 INFO [train2.py:809] (1/4) Epoch 26, batch 3800, loss[ctc_loss=0.04976, att_loss=0.2358, loss=0.1986, over 16890.00 frames. utt_duration=1380 frames, utt_pad_proportion=0.006945, over 49.00 utterances.], tot_loss[ctc_loss=0.06527, att_loss=0.232, loss=0.1986, over 3272202.64 frames. utt_duration=1254 frames, utt_pad_proportion=0.05257, over 10448.72 utterances.], batch size: 49, lr: 4.09e-03, grad_scale: 16.0 2023-03-09 08:34:45,439 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0601, 5.0904, 4.9448, 2.3022, 2.0857, 2.9314, 2.5163, 3.8929], device='cuda:1'), covar=tensor([0.0742, 0.0358, 0.0264, 0.4963, 0.5419, 0.2612, 0.3904, 0.1648], device='cuda:1'), in_proj_covar=tensor([0.0360, 0.0298, 0.0277, 0.0249, 0.0336, 0.0329, 0.0260, 0.0366], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-03-09 08:34:48,464 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103427.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 08:34:55,138 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2700, 4.7334, 4.5936, 4.7180, 4.7619, 4.4608, 3.2988, 4.6571], device='cuda:1'), covar=tensor([0.0157, 0.0126, 0.0149, 0.0093, 0.0101, 0.0131, 0.0743, 0.0201], device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0094, 0.0118, 0.0074, 0.0080, 0.0091, 0.0108, 0.0113], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 08:35:10,633 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.0006, 3.7403, 3.8159, 3.1919, 3.7750, 3.7487, 3.7527, 2.7029], device='cuda:1'), covar=tensor([0.0931, 0.0930, 0.1267, 0.2671, 0.0692, 0.1638, 0.0933, 0.3013], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0204, 0.0218, 0.0269, 0.0179, 0.0279, 0.0200, 0.0228], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 08:35:16,368 INFO [train2.py:809] (1/4) Epoch 26, batch 3850, loss[ctc_loss=0.05264, att_loss=0.2148, loss=0.1823, over 16023.00 frames. utt_duration=1604 frames, utt_pad_proportion=0.006811, over 40.00 utterances.], tot_loss[ctc_loss=0.06524, att_loss=0.2317, loss=0.1984, over 3274273.69 frames. utt_duration=1258 frames, utt_pad_proportion=0.0504, over 10423.47 utterances.], batch size: 40, lr: 4.09e-03, grad_scale: 8.0 2023-03-09 08:35:26,997 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.253e+02 1.752e+02 2.146e+02 2.591e+02 4.293e+02, threshold=4.292e+02, percent-clipped=0.0 2023-03-09 08:35:30,347 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103454.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:36:22,166 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103488.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:36:32,462 INFO [train2.py:809] (1/4) Epoch 26, batch 3900, loss[ctc_loss=0.08097, att_loss=0.2494, loss=0.2157, over 17082.00 frames. utt_duration=1291 frames, utt_pad_proportion=0.007962, over 53.00 utterances.], tot_loss[ctc_loss=0.06572, att_loss=0.2319, loss=0.1986, over 3274799.44 frames. utt_duration=1223 frames, utt_pad_proportion=0.05877, over 10726.93 utterances.], batch size: 53, lr: 4.09e-03, grad_scale: 8.0 2023-03-09 08:36:50,977 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103507.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:37:47,333 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103543.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:37:50,085 INFO [train2.py:809] (1/4) Epoch 26, batch 3950, loss[ctc_loss=0.0595, att_loss=0.2253, loss=0.1922, over 16326.00 frames. utt_duration=1453 frames, utt_pad_proportion=0.006338, over 45.00 utterances.], tot_loss[ctc_loss=0.0655, att_loss=0.232, loss=0.1987, over 3282954.19 frames. utt_duration=1225 frames, utt_pad_proportion=0.05592, over 10733.39 utterances.], batch size: 45, lr: 4.09e-03, grad_scale: 8.0 2023-03-09 08:38:00,647 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.092e+02 1.745e+02 2.083e+02 2.654e+02 4.442e+02, threshold=4.166e+02, percent-clipped=1.0 2023-03-09 08:38:05,191 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=103555.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:39:05,365 INFO [train2.py:809] (1/4) Epoch 27, batch 0, loss[ctc_loss=0.04849, att_loss=0.2136, loss=0.1806, over 14601.00 frames. utt_duration=1827 frames, utt_pad_proportion=0.03855, over 32.00 utterances.], tot_loss[ctc_loss=0.04849, att_loss=0.2136, loss=0.1806, over 14601.00 frames. utt_duration=1827 frames, utt_pad_proportion=0.03855, over 32.00 utterances.], batch size: 32, lr: 4.01e-03, grad_scale: 8.0 2023-03-09 08:39:05,365 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-09 08:39:17,375 INFO [train2.py:843] (1/4) Epoch 27, validation: ctc_loss=0.04075, att_loss=0.2342, loss=0.1955, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-09 08:39:17,376 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16092MB 2023-03-09 08:39:36,974 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=103591.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:40:35,286 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3968, 4.6176, 4.6820, 4.6776, 5.2409, 4.4629, 4.6611, 2.7113], device='cuda:1'), covar=tensor([0.0318, 0.0355, 0.0327, 0.0331, 0.0756, 0.0282, 0.0331, 0.1692], device='cuda:1'), in_proj_covar=tensor([0.0190, 0.0219, 0.0215, 0.0232, 0.0389, 0.0190, 0.0203, 0.0225], 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') 2023-03-09 08:40:36,385 INFO [train2.py:809] (1/4) Epoch 27, batch 50, loss[ctc_loss=0.05274, att_loss=0.2088, loss=0.1776, over 15628.00 frames. utt_duration=1691 frames, utt_pad_proportion=0.009985, over 37.00 utterances.], tot_loss[ctc_loss=0.06568, att_loss=0.2305, loss=0.1975, over 731025.83 frames. utt_duration=1265 frames, utt_pad_proportion=0.05922, over 2313.45 utterances.], batch size: 37, lr: 4.01e-03, grad_scale: 8.0 2023-03-09 08:41:14,260 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.229e+02 1.910e+02 2.259e+02 2.713e+02 6.930e+02, threshold=4.518e+02, percent-clipped=6.0 2023-03-09 08:41:56,146 INFO [train2.py:809] (1/4) Epoch 27, batch 100, loss[ctc_loss=0.08014, att_loss=0.2368, loss=0.2055, over 16485.00 frames. utt_duration=1435 frames, utt_pad_proportion=0.006248, over 46.00 utterances.], tot_loss[ctc_loss=0.06463, att_loss=0.2291, loss=0.1962, over 1289958.66 frames. utt_duration=1295 frames, utt_pad_proportion=0.05026, over 3987.57 utterances.], batch size: 46, lr: 4.01e-03, grad_scale: 8.0 2023-03-09 08:42:01,080 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5670, 4.9829, 4.7936, 4.9536, 4.9816, 4.5974, 3.4182, 4.9110], device='cuda:1'), covar=tensor([0.0121, 0.0116, 0.0143, 0.0076, 0.0094, 0.0157, 0.0720, 0.0157], device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0094, 0.0117, 0.0074, 0.0079, 0.0091, 0.0107, 0.0112], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 08:42:45,231 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1097, 5.0840, 4.8175, 2.7309, 4.9185, 4.6995, 4.5342, 2.8984], device='cuda:1'), covar=tensor([0.0127, 0.0114, 0.0275, 0.1212, 0.0109, 0.0214, 0.0265, 0.1351], device='cuda:1'), in_proj_covar=tensor([0.0078, 0.0107, 0.0110, 0.0113, 0.0090, 0.0118, 0.0101, 0.0105], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-09 08:43:04,841 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103722.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 08:43:06,402 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5569, 2.8377, 3.6780, 3.1171, 3.5006, 4.6773, 4.4725, 3.3311], device='cuda:1'), covar=tensor([0.0376, 0.1855, 0.1254, 0.1296, 0.1122, 0.0805, 0.0644, 0.1167], device='cuda:1'), in_proj_covar=tensor([0.0249, 0.0251, 0.0290, 0.0223, 0.0271, 0.0382, 0.0274, 0.0236], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-09 08:43:11,057 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0430, 4.4360, 4.6041, 4.6065, 3.0227, 4.2605, 3.0390, 1.8700], device='cuda:1'), covar=tensor([0.0517, 0.0297, 0.0540, 0.0263, 0.1348, 0.0281, 0.1193, 0.1629], device='cuda:1'), in_proj_covar=tensor([0.0210, 0.0183, 0.0261, 0.0175, 0.0221, 0.0163, 0.0230, 0.0204], 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') 2023-03-09 08:43:15,221 INFO [train2.py:809] (1/4) Epoch 27, batch 150, loss[ctc_loss=0.08594, att_loss=0.2493, loss=0.2167, over 17134.00 frames. utt_duration=1225 frames, utt_pad_proportion=0.01412, over 56.00 utterances.], tot_loss[ctc_loss=0.06517, att_loss=0.23, loss=0.197, over 1731897.50 frames. utt_duration=1313 frames, utt_pad_proportion=0.04389, over 5281.75 utterances.], batch size: 56, lr: 4.01e-03, grad_scale: 8.0 2023-03-09 08:43:33,687 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103740.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:43:52,273 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.205e+02 1.770e+02 2.085e+02 2.452e+02 6.368e+02, threshold=4.170e+02, percent-clipped=1.0 2023-03-09 08:43:56,604 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103754.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:44:13,418 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103765.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:44:25,448 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0388, 5.3475, 5.0053, 5.4280, 4.8265, 5.1073, 5.5355, 5.2985], device='cuda:1'), covar=tensor([0.0703, 0.0317, 0.0748, 0.0319, 0.0436, 0.0233, 0.0244, 0.0204], device='cuda:1'), in_proj_covar=tensor([0.0406, 0.0339, 0.0385, 0.0377, 0.0341, 0.0249, 0.0321, 0.0302], device='cuda:1'), out_proj_covar=tensor([0.0007, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-09 08:44:34,217 INFO [train2.py:809] (1/4) Epoch 27, batch 200, loss[ctc_loss=0.07727, att_loss=0.2505, loss=0.2159, over 17353.00 frames. utt_duration=1178 frames, utt_pad_proportion=0.01999, over 59.00 utterances.], tot_loss[ctc_loss=0.06556, att_loss=0.2311, loss=0.198, over 2074355.17 frames. utt_duration=1263 frames, utt_pad_proportion=0.05502, over 6579.85 utterances.], batch size: 59, lr: 4.01e-03, grad_scale: 8.0 2023-03-09 08:44:49,607 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103788.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:44:49,734 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4321, 4.5916, 4.6936, 4.6830, 5.3476, 4.4165, 4.6420, 2.8093], device='cuda:1'), covar=tensor([0.0286, 0.0447, 0.0347, 0.0374, 0.0648, 0.0296, 0.0348, 0.1557], device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0219, 0.0215, 0.0232, 0.0388, 0.0190, 0.0203, 0.0224], 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') 2023-03-09 08:45:10,655 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103801.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:45:11,943 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=103802.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:45:49,611 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103826.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:45:53,753 INFO [train2.py:809] (1/4) Epoch 27, batch 250, loss[ctc_loss=0.05309, att_loss=0.2333, loss=0.1972, over 17119.00 frames. utt_duration=1224 frames, utt_pad_proportion=0.01341, over 56.00 utterances.], tot_loss[ctc_loss=0.0655, att_loss=0.231, loss=0.1979, over 2344149.55 frames. utt_duration=1275 frames, utt_pad_proportion=0.04953, over 7364.71 utterances.], batch size: 56, lr: 4.01e-03, grad_scale: 8.0 2023-03-09 08:46:05,261 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=103836.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:46:15,355 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4019, 2.7609, 3.6070, 3.0178, 3.4931, 4.5203, 4.3727, 3.0997], device='cuda:1'), covar=tensor([0.0383, 0.1859, 0.1341, 0.1326, 0.1117, 0.0971, 0.0617, 0.1366], device='cuda:1'), in_proj_covar=tensor([0.0249, 0.0252, 0.0291, 0.0224, 0.0272, 0.0383, 0.0274, 0.0237], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-09 08:46:30,829 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.121e+02 1.879e+02 2.239e+02 2.738e+02 5.518e+02, threshold=4.478e+02, percent-clipped=1.0 2023-03-09 08:46:57,887 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.1988, 2.7673, 3.1528, 4.3122, 3.7620, 3.7988, 2.9111, 2.0423], device='cuda:1'), covar=tensor([0.0817, 0.1906, 0.0925, 0.0582, 0.0950, 0.0521, 0.1517, 0.2348], device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0220, 0.0189, 0.0226, 0.0236, 0.0191, 0.0206, 0.0194], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 08:47:08,875 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103876.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:47:13,081 INFO [train2.py:809] (1/4) Epoch 27, batch 300, loss[ctc_loss=0.0495, att_loss=0.2266, loss=0.1912, over 16535.00 frames. utt_duration=1471 frames, utt_pad_proportion=0.005109, over 45.00 utterances.], tot_loss[ctc_loss=0.06561, att_loss=0.2317, loss=0.1985, over 2553624.26 frames. utt_duration=1277 frames, utt_pad_proportion=0.04719, over 8010.55 utterances.], batch size: 45, lr: 4.00e-03, grad_scale: 8.0 2023-03-09 08:48:20,463 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1333, 5.0300, 4.8070, 2.8888, 4.8900, 4.8016, 4.3038, 2.5282], device='cuda:1'), covar=tensor([0.0121, 0.0124, 0.0295, 0.1064, 0.0116, 0.0187, 0.0322, 0.1443], device='cuda:1'), in_proj_covar=tensor([0.0078, 0.0106, 0.0110, 0.0113, 0.0090, 0.0118, 0.0101, 0.0105], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-09 08:48:32,103 INFO [train2.py:809] (1/4) Epoch 27, batch 350, loss[ctc_loss=0.08107, att_loss=0.2185, loss=0.191, over 15345.00 frames. utt_duration=1755 frames, utt_pad_proportion=0.009894, over 35.00 utterances.], tot_loss[ctc_loss=0.06709, att_loss=0.2322, loss=0.1992, over 2710361.56 frames. utt_duration=1251 frames, utt_pad_proportion=0.0547, over 8675.59 utterances.], batch size: 35, lr: 4.00e-03, grad_scale: 8.0 2023-03-09 08:48:45,731 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103937.0, num_to_drop=1, layers_to_drop={3} 2023-03-09 08:49:08,960 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.349e+02 1.861e+02 2.238e+02 2.752e+02 4.825e+02, threshold=4.477e+02, percent-clipped=1.0 2023-03-09 08:49:14,468 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0113, 4.9846, 4.7939, 2.6663, 4.8731, 4.6392, 4.3051, 2.6857], device='cuda:1'), covar=tensor([0.0119, 0.0110, 0.0266, 0.1212, 0.0103, 0.0206, 0.0321, 0.1437], device='cuda:1'), in_proj_covar=tensor([0.0079, 0.0107, 0.0111, 0.0114, 0.0091, 0.0119, 0.0102, 0.0106], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-09 08:49:28,376 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4834, 2.6028, 4.9264, 3.8549, 3.1123, 4.2284, 4.7306, 4.6143], device='cuda:1'), covar=tensor([0.0310, 0.1565, 0.0211, 0.0974, 0.1641, 0.0273, 0.0203, 0.0291], device='cuda:1'), in_proj_covar=tensor([0.0225, 0.0249, 0.0217, 0.0325, 0.0271, 0.0237, 0.0208, 0.0234], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 08:49:30,495 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=2.72 vs. limit=5.0 2023-03-09 08:49:51,233 INFO [train2.py:809] (1/4) Epoch 27, batch 400, loss[ctc_loss=0.04849, att_loss=0.2085, loss=0.1765, over 16122.00 frames. utt_duration=1536 frames, utt_pad_proportion=0.006787, over 42.00 utterances.], tot_loss[ctc_loss=0.06603, att_loss=0.2312, loss=0.1982, over 2838600.49 frames. utt_duration=1283 frames, utt_pad_proportion=0.04608, over 8857.62 utterances.], batch size: 42, lr: 4.00e-03, grad_scale: 8.0 2023-03-09 08:50:09,007 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103990.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:50:19,658 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103997.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:50:53,740 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9967, 4.0698, 4.0499, 4.1385, 4.1839, 4.1759, 3.9153, 3.8326], device='cuda:1'), covar=tensor([0.0994, 0.0735, 0.1003, 0.0557, 0.0379, 0.0404, 0.0544, 0.0407], device='cuda:1'), in_proj_covar=tensor([0.0535, 0.0381, 0.0370, 0.0378, 0.0441, 0.0446, 0.0378, 0.0411], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:1') 2023-03-09 08:51:04,769 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104022.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 08:51:15,266 INFO [train2.py:809] (1/4) Epoch 27, batch 450, loss[ctc_loss=0.06167, att_loss=0.2359, loss=0.2011, over 17311.00 frames. utt_duration=1175 frames, utt_pad_proportion=0.02232, over 59.00 utterances.], tot_loss[ctc_loss=0.06692, att_loss=0.2316, loss=0.1987, over 2916949.59 frames. utt_duration=1227 frames, utt_pad_proportion=0.06376, over 9521.30 utterances.], batch size: 59, lr: 4.00e-03, grad_scale: 8.0 2023-03-09 08:51:48,284 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=104049.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:51:51,436 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104051.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:51:52,513 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.211e+02 1.790e+02 2.279e+02 2.671e+02 6.527e+02, threshold=4.558e+02, percent-clipped=3.0 2023-03-09 08:52:02,869 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104058.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:52:20,943 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=104070.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 08:52:34,617 INFO [train2.py:809] (1/4) Epoch 27, batch 500, loss[ctc_loss=0.05777, att_loss=0.2209, loss=0.1883, over 15769.00 frames. utt_duration=1661 frames, utt_pad_proportion=0.008201, over 38.00 utterances.], tot_loss[ctc_loss=0.06722, att_loss=0.2329, loss=0.1998, over 3009753.40 frames. utt_duration=1212 frames, utt_pad_proportion=0.06073, over 9946.47 utterances.], batch size: 38, lr: 4.00e-03, grad_scale: 8.0 2023-03-09 08:53:02,103 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104096.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:53:16,166 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-03-09 08:53:23,999 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104110.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:53:41,003 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104121.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:53:53,122 INFO [train2.py:809] (1/4) Epoch 27, batch 550, loss[ctc_loss=0.05879, att_loss=0.2371, loss=0.2014, over 17032.00 frames. utt_duration=1337 frames, utt_pad_proportion=0.007075, over 51.00 utterances.], tot_loss[ctc_loss=0.06728, att_loss=0.2333, loss=0.2001, over 3078101.26 frames. utt_duration=1225 frames, utt_pad_proportion=0.05537, over 10059.22 utterances.], batch size: 51, lr: 4.00e-03, grad_scale: 8.0 2023-03-09 08:54:15,636 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.0396, 3.7843, 3.1812, 3.3659, 4.0368, 3.6307, 2.8666, 4.2286], device='cuda:1'), covar=tensor([0.1048, 0.0486, 0.1074, 0.0753, 0.0713, 0.0768, 0.0972, 0.0506], device='cuda:1'), in_proj_covar=tensor([0.0206, 0.0222, 0.0231, 0.0206, 0.0288, 0.0246, 0.0202, 0.0293], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-09 08:54:29,374 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.176e+02 1.746e+02 2.054e+02 2.540e+02 6.808e+02, threshold=4.109e+02, percent-clipped=2.0 2023-03-09 08:55:10,801 INFO [train2.py:809] (1/4) Epoch 27, batch 600, loss[ctc_loss=0.04786, att_loss=0.219, loss=0.1848, over 16001.00 frames. utt_duration=1602 frames, utt_pad_proportion=0.007683, over 40.00 utterances.], tot_loss[ctc_loss=0.06719, att_loss=0.2333, loss=0.2, over 3130776.82 frames. utt_duration=1253 frames, utt_pad_proportion=0.04737, over 10006.72 utterances.], batch size: 40, lr: 4.00e-03, grad_scale: 8.0 2023-03-09 08:55:41,911 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7467, 2.5976, 2.4762, 2.6814, 2.8801, 2.9492, 2.3668, 3.1422], device='cuda:1'), covar=tensor([0.1438, 0.2116, 0.2025, 0.1361, 0.1397, 0.0930, 0.2093, 0.1033], device='cuda:1'), in_proj_covar=tensor([0.0138, 0.0142, 0.0137, 0.0132, 0.0149, 0.0126, 0.0150, 0.0127], 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') 2023-03-09 08:56:17,658 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9365, 5.2032, 5.4736, 5.3013, 5.4376, 5.9001, 5.2523, 5.9400], device='cuda:1'), covar=tensor([0.0767, 0.0784, 0.0900, 0.1360, 0.1806, 0.0925, 0.0734, 0.0796], device='cuda:1'), in_proj_covar=tensor([0.0901, 0.0524, 0.0634, 0.0682, 0.0901, 0.0661, 0.0507, 0.0638], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 08:56:29,436 INFO [train2.py:809] (1/4) Epoch 27, batch 650, loss[ctc_loss=0.05532, att_loss=0.2185, loss=0.1858, over 16152.00 frames. utt_duration=1577 frames, utt_pad_proportion=0.007869, over 41.00 utterances.], tot_loss[ctc_loss=0.06792, att_loss=0.2336, loss=0.2005, over 3170731.60 frames. utt_duration=1243 frames, utt_pad_proportion=0.04897, over 10216.49 utterances.], batch size: 41, lr: 4.00e-03, grad_scale: 8.0 2023-03-09 08:56:34,218 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104232.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 08:57:05,924 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.363e+02 1.860e+02 2.240e+02 2.877e+02 8.849e+02, threshold=4.479e+02, percent-clipped=3.0 2023-03-09 08:57:47,596 INFO [train2.py:809] (1/4) Epoch 27, batch 700, loss[ctc_loss=0.06305, att_loss=0.2388, loss=0.2036, over 16772.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.006244, over 48.00 utterances.], tot_loss[ctc_loss=0.06769, att_loss=0.2333, loss=0.2001, over 3194445.14 frames. utt_duration=1216 frames, utt_pad_proportion=0.05597, over 10521.31 utterances.], batch size: 48, lr: 4.00e-03, grad_scale: 8.0 2023-03-09 08:58:14,124 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5626, 4.6555, 4.7661, 4.6566, 5.2614, 4.5178, 4.7015, 2.5323], device='cuda:1'), covar=tensor([0.0327, 0.0371, 0.0321, 0.0370, 0.0697, 0.0327, 0.0323, 0.1826], device='cuda:1'), in_proj_covar=tensor([0.0190, 0.0219, 0.0215, 0.0233, 0.0388, 0.0189, 0.0204, 0.0224], 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') 2023-03-09 08:59:04,931 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-03-09 08:59:06,610 INFO [train2.py:809] (1/4) Epoch 27, batch 750, loss[ctc_loss=0.04003, att_loss=0.201, loss=0.1688, over 15733.00 frames. utt_duration=1658 frames, utt_pad_proportion=0.009699, over 38.00 utterances.], tot_loss[ctc_loss=0.06769, att_loss=0.2331, loss=0.2, over 3208091.07 frames. utt_duration=1210 frames, utt_pad_proportion=0.06029, over 10617.55 utterances.], batch size: 38, lr: 4.00e-03, grad_scale: 8.0 2023-03-09 08:59:34,728 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104346.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 08:59:43,522 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.260e+02 1.900e+02 2.289e+02 2.673e+02 5.162e+02, threshold=4.578e+02, percent-clipped=1.0 2023-03-09 08:59:45,860 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104353.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:00:26,056 INFO [train2.py:809] (1/4) Epoch 27, batch 800, loss[ctc_loss=0.04985, att_loss=0.2114, loss=0.1791, over 15882.00 frames. utt_duration=1630 frames, utt_pad_proportion=0.00944, over 39.00 utterances.], tot_loss[ctc_loss=0.06763, att_loss=0.2324, loss=0.1995, over 3213322.44 frames. utt_duration=1217 frames, utt_pad_proportion=0.06218, over 10574.61 utterances.], batch size: 39, lr: 3.99e-03, grad_scale: 8.0 2023-03-09 09:00:54,085 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104396.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:01:07,269 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4418, 2.3569, 4.9258, 3.8646, 3.0852, 4.1742, 4.6863, 4.6433], device='cuda:1'), covar=tensor([0.0287, 0.1690, 0.0237, 0.0895, 0.1619, 0.0295, 0.0188, 0.0268], device='cuda:1'), in_proj_covar=tensor([0.0225, 0.0247, 0.0217, 0.0323, 0.0270, 0.0236, 0.0209, 0.0234], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 09:01:08,571 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104405.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:01:18,844 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.15 vs. limit=5.0 2023-03-09 09:01:33,312 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104421.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:01:46,498 INFO [train2.py:809] (1/4) Epoch 27, batch 850, loss[ctc_loss=0.08234, att_loss=0.2269, loss=0.198, over 16181.00 frames. utt_duration=1580 frames, utt_pad_proportion=0.006473, over 41.00 utterances.], tot_loss[ctc_loss=0.06705, att_loss=0.2318, loss=0.1989, over 3219240.66 frames. utt_duration=1213 frames, utt_pad_proportion=0.06578, over 10632.24 utterances.], batch size: 41, lr: 3.99e-03, grad_scale: 8.0 2023-03-09 09:02:10,583 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=104444.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:02:23,113 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.335e+02 1.830e+02 2.139e+02 2.642e+02 5.994e+02, threshold=4.278e+02, percent-clipped=1.0 2023-03-09 09:02:31,192 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1866, 5.4319, 5.3867, 5.3756, 5.4996, 5.4242, 5.1030, 4.8921], device='cuda:1'), covar=tensor([0.0911, 0.0514, 0.0324, 0.0534, 0.0282, 0.0331, 0.0390, 0.0310], device='cuda:1'), in_proj_covar=tensor([0.0534, 0.0379, 0.0369, 0.0376, 0.0439, 0.0444, 0.0377, 0.0409], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004], device='cuda:1') 2023-03-09 09:02:49,618 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=104469.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:03:05,897 INFO [train2.py:809] (1/4) Epoch 27, batch 900, loss[ctc_loss=0.07565, att_loss=0.2362, loss=0.2041, over 16538.00 frames. utt_duration=1471 frames, utt_pad_proportion=0.006527, over 45.00 utterances.], tot_loss[ctc_loss=0.06705, att_loss=0.232, loss=0.199, over 3233801.40 frames. utt_duration=1241 frames, utt_pad_proportion=0.05875, over 10437.32 utterances.], batch size: 45, lr: 3.99e-03, grad_scale: 8.0 2023-03-09 09:03:39,653 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1834, 4.2241, 4.1987, 4.1633, 4.6464, 4.1354, 4.1911, 2.5022], device='cuda:1'), covar=tensor([0.0330, 0.0467, 0.0482, 0.0471, 0.0947, 0.0348, 0.0446, 0.1826], device='cuda:1'), in_proj_covar=tensor([0.0191, 0.0221, 0.0216, 0.0234, 0.0389, 0.0190, 0.0205, 0.0225], 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') 2023-03-09 09:04:02,493 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0406, 5.0329, 5.0283, 2.0445, 1.9005, 2.6652, 2.1253, 3.8026], device='cuda:1'), covar=tensor([0.0789, 0.0336, 0.0209, 0.5286, 0.5865, 0.2762, 0.4019, 0.1724], device='cuda:1'), in_proj_covar=tensor([0.0366, 0.0302, 0.0280, 0.0254, 0.0344, 0.0335, 0.0263, 0.0373], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-09 09:04:24,928 INFO [train2.py:809] (1/4) Epoch 27, batch 950, loss[ctc_loss=0.0642, att_loss=0.228, loss=0.1952, over 16157.00 frames. utt_duration=1578 frames, utt_pad_proportion=0.008294, over 41.00 utterances.], tot_loss[ctc_loss=0.06641, att_loss=0.2317, loss=0.1986, over 3243782.53 frames. utt_duration=1259 frames, utt_pad_proportion=0.05412, over 10319.93 utterances.], batch size: 41, lr: 3.99e-03, grad_scale: 8.0 2023-03-09 09:04:30,380 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104532.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 09:04:39,537 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9006, 5.2085, 5.3745, 5.2671, 5.4017, 5.8549, 5.2637, 5.9529], device='cuda:1'), covar=tensor([0.0732, 0.0750, 0.0942, 0.1451, 0.1824, 0.0983, 0.0761, 0.0702], device='cuda:1'), in_proj_covar=tensor([0.0913, 0.0530, 0.0643, 0.0691, 0.0908, 0.0668, 0.0513, 0.0646], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 09:05:01,190 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.309e+02 1.806e+02 2.235e+02 2.788e+02 5.642e+02, threshold=4.470e+02, percent-clipped=2.0 2023-03-09 09:05:10,807 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.1329, 3.7896, 3.8193, 3.2386, 3.7873, 3.8933, 3.8600, 2.7058], device='cuda:1'), covar=tensor([0.0837, 0.1166, 0.1335, 0.3067, 0.1391, 0.1873, 0.0833, 0.3356], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0204, 0.0219, 0.0267, 0.0179, 0.0278, 0.0200, 0.0227], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 09:05:43,845 INFO [train2.py:809] (1/4) Epoch 27, batch 1000, loss[ctc_loss=0.06177, att_loss=0.2182, loss=0.1869, over 13701.00 frames. utt_duration=1829 frames, utt_pad_proportion=0.07362, over 30.00 utterances.], tot_loss[ctc_loss=0.06673, att_loss=0.2322, loss=0.1991, over 3245944.85 frames. utt_duration=1257 frames, utt_pad_proportion=0.05523, over 10340.64 utterances.], batch size: 30, lr: 3.99e-03, grad_scale: 8.0 2023-03-09 09:05:45,435 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=104580.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:07:03,585 INFO [train2.py:809] (1/4) Epoch 27, batch 1050, loss[ctc_loss=0.05532, att_loss=0.2158, loss=0.1837, over 16133.00 frames. utt_duration=1538 frames, utt_pad_proportion=0.005741, over 42.00 utterances.], tot_loss[ctc_loss=0.06659, att_loss=0.232, loss=0.1989, over 3254969.08 frames. utt_duration=1261 frames, utt_pad_proportion=0.05206, over 10333.61 utterances.], batch size: 42, lr: 3.99e-03, grad_scale: 8.0 2023-03-09 09:07:30,293 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104646.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:07:40,080 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.329e+02 1.873e+02 2.248e+02 2.692e+02 7.712e+02, threshold=4.496e+02, percent-clipped=6.0 2023-03-09 09:07:41,953 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104653.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:08:22,948 INFO [train2.py:809] (1/4) Epoch 27, batch 1100, loss[ctc_loss=0.07038, att_loss=0.2349, loss=0.202, over 17277.00 frames. utt_duration=876.3 frames, utt_pad_proportion=0.08244, over 79.00 utterances.], tot_loss[ctc_loss=0.06619, att_loss=0.2317, loss=0.1986, over 3262934.10 frames. utt_duration=1264 frames, utt_pad_proportion=0.0486, over 10340.69 utterances.], batch size: 79, lr: 3.99e-03, grad_scale: 8.0 2023-03-09 09:08:46,597 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=104694.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:08:58,150 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=104701.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:09:04,419 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104705.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:09:42,150 INFO [train2.py:809] (1/4) Epoch 27, batch 1150, loss[ctc_loss=0.05551, att_loss=0.2149, loss=0.183, over 15744.00 frames. utt_duration=1659 frames, utt_pad_proportion=0.008945, over 38.00 utterances.], tot_loss[ctc_loss=0.06594, att_loss=0.2313, loss=0.1983, over 3255182.94 frames. utt_duration=1249 frames, utt_pad_proportion=0.05334, over 10435.63 utterances.], batch size: 38, lr: 3.99e-03, grad_scale: 8.0 2023-03-09 09:09:49,650 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1397, 2.7929, 3.4434, 2.9590, 3.3729, 4.2450, 4.0437, 3.1394], device='cuda:1'), covar=tensor([0.0376, 0.1663, 0.1280, 0.1228, 0.1098, 0.0870, 0.0655, 0.1138], device='cuda:1'), in_proj_covar=tensor([0.0248, 0.0249, 0.0288, 0.0221, 0.0270, 0.0378, 0.0273, 0.0233], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-09 09:10:19,016 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.235e+02 1.763e+02 2.147e+02 2.589e+02 3.796e+02, threshold=4.293e+02, percent-clipped=0.0 2023-03-09 09:10:20,703 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=104753.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:10:25,604 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.5356, 2.5805, 2.6260, 2.4300, 2.5588, 2.5186, 2.5781, 1.9809], device='cuda:1'), covar=tensor([0.1206, 0.1514, 0.1900, 0.3055, 0.1216, 0.2454, 0.1488, 0.3465], device='cuda:1'), in_proj_covar=tensor([0.0197, 0.0204, 0.0219, 0.0268, 0.0180, 0.0278, 0.0200, 0.0227], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 09:11:01,672 INFO [train2.py:809] (1/4) Epoch 27, batch 1200, loss[ctc_loss=0.04154, att_loss=0.2015, loss=0.1695, over 15890.00 frames. utt_duration=1631 frames, utt_pad_proportion=0.008292, over 39.00 utterances.], tot_loss[ctc_loss=0.06558, att_loss=0.231, loss=0.1979, over 3252346.52 frames. utt_duration=1243 frames, utt_pad_proportion=0.05656, over 10474.53 utterances.], batch size: 39, lr: 3.99e-03, grad_scale: 8.0 2023-03-09 09:11:49,104 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-03-09 09:12:21,905 INFO [train2.py:809] (1/4) Epoch 27, batch 1250, loss[ctc_loss=0.05466, att_loss=0.2228, loss=0.1892, over 15946.00 frames. utt_duration=1557 frames, utt_pad_proportion=0.006969, over 41.00 utterances.], tot_loss[ctc_loss=0.06509, att_loss=0.2303, loss=0.1973, over 3256996.40 frames. utt_duration=1274 frames, utt_pad_proportion=0.04906, over 10236.75 utterances.], batch size: 41, lr: 3.99e-03, grad_scale: 8.0 2023-03-09 09:12:58,971 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.231e+02 1.878e+02 2.209e+02 2.737e+02 5.339e+02, threshold=4.418e+02, percent-clipped=1.0 2023-03-09 09:13:42,099 INFO [train2.py:809] (1/4) Epoch 27, batch 1300, loss[ctc_loss=0.06181, att_loss=0.2227, loss=0.1905, over 15959.00 frames. utt_duration=1558 frames, utt_pad_proportion=0.006206, over 41.00 utterances.], tot_loss[ctc_loss=0.06478, att_loss=0.2302, loss=0.1971, over 3249867.51 frames. utt_duration=1265 frames, utt_pad_proportion=0.05424, over 10287.23 utterances.], batch size: 41, lr: 3.99e-03, grad_scale: 8.0 2023-03-09 09:14:11,608 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.97 vs. limit=5.0 2023-03-09 09:15:01,661 INFO [train2.py:809] (1/4) Epoch 27, batch 1350, loss[ctc_loss=0.06411, att_loss=0.2308, loss=0.1974, over 16991.00 frames. utt_duration=1334 frames, utt_pad_proportion=0.009433, over 51.00 utterances.], tot_loss[ctc_loss=0.06508, att_loss=0.2304, loss=0.1973, over 3257933.58 frames. utt_duration=1266 frames, utt_pad_proportion=0.0518, over 10303.96 utterances.], batch size: 51, lr: 3.98e-03, grad_scale: 8.0 2023-03-09 09:15:37,626 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.130e+02 1.826e+02 2.285e+02 3.088e+02 1.440e+03, threshold=4.571e+02, percent-clipped=4.0 2023-03-09 09:15:57,048 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7853, 4.8102, 4.9300, 4.9216, 5.3453, 4.8207, 4.7927, 2.4239], device='cuda:1'), covar=tensor([0.0164, 0.0188, 0.0189, 0.0181, 0.0507, 0.0156, 0.0206, 0.1733], device='cuda:1'), in_proj_covar=tensor([0.0191, 0.0221, 0.0216, 0.0234, 0.0389, 0.0191, 0.0206, 0.0225], 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') 2023-03-09 09:16:20,668 INFO [train2.py:809] (1/4) Epoch 27, batch 1400, loss[ctc_loss=0.06236, att_loss=0.2115, loss=0.1816, over 15364.00 frames. utt_duration=1757 frames, utt_pad_proportion=0.01042, over 35.00 utterances.], tot_loss[ctc_loss=0.06534, att_loss=0.2309, loss=0.1978, over 3260779.66 frames. utt_duration=1263 frames, utt_pad_proportion=0.05089, over 10337.54 utterances.], batch size: 35, lr: 3.98e-03, grad_scale: 8.0 2023-03-09 09:17:40,904 INFO [train2.py:809] (1/4) Epoch 27, batch 1450, loss[ctc_loss=0.06322, att_loss=0.239, loss=0.2039, over 16888.00 frames. utt_duration=1380 frames, utt_pad_proportion=0.006553, over 49.00 utterances.], tot_loss[ctc_loss=0.06514, att_loss=0.2308, loss=0.1977, over 3260164.78 frames. utt_duration=1281 frames, utt_pad_proportion=0.04772, over 10192.60 utterances.], batch size: 49, lr: 3.98e-03, grad_scale: 8.0 2023-03-09 09:17:59,355 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0724, 6.2523, 5.7463, 5.9796, 5.9101, 5.4026, 5.8007, 5.4303], device='cuda:1'), covar=tensor([0.1139, 0.0981, 0.0851, 0.0755, 0.0783, 0.1785, 0.2319, 0.2319], device='cuda:1'), in_proj_covar=tensor([0.0550, 0.0627, 0.0481, 0.0472, 0.0445, 0.0477, 0.0634, 0.0542], 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') 2023-03-09 09:18:16,918 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.239e+02 1.845e+02 2.127e+02 2.675e+02 5.625e+02, threshold=4.253e+02, percent-clipped=2.0 2023-03-09 09:19:00,492 INFO [train2.py:809] (1/4) Epoch 27, batch 1500, loss[ctc_loss=0.08136, att_loss=0.2478, loss=0.2145, over 17057.00 frames. utt_duration=1288 frames, utt_pad_proportion=0.009659, over 53.00 utterances.], tot_loss[ctc_loss=0.06537, att_loss=0.2309, loss=0.1978, over 3264334.71 frames. utt_duration=1267 frames, utt_pad_proportion=0.05106, over 10319.48 utterances.], batch size: 53, lr: 3.98e-03, grad_scale: 8.0 2023-03-09 09:19:45,682 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6073, 3.2628, 3.7435, 3.2890, 3.5610, 4.6551, 4.4705, 3.5187], device='cuda:1'), covar=tensor([0.0317, 0.1463, 0.1204, 0.1127, 0.1117, 0.0883, 0.0552, 0.0982], device='cuda:1'), in_proj_covar=tensor([0.0250, 0.0250, 0.0291, 0.0223, 0.0273, 0.0383, 0.0276, 0.0235], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-09 09:20:18,686 INFO [train2.py:809] (1/4) Epoch 27, batch 1550, loss[ctc_loss=0.06117, att_loss=0.2321, loss=0.1979, over 16425.00 frames. utt_duration=1495 frames, utt_pad_proportion=0.006286, over 44.00 utterances.], tot_loss[ctc_loss=0.06531, att_loss=0.2312, loss=0.198, over 3273693.61 frames. utt_duration=1290 frames, utt_pad_proportion=0.0437, over 10163.53 utterances.], batch size: 44, lr: 3.98e-03, grad_scale: 8.0 2023-03-09 09:20:55,285 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.109e+02 1.873e+02 2.333e+02 2.814e+02 5.399e+02, threshold=4.665e+02, percent-clipped=5.0 2023-03-09 09:21:38,750 INFO [train2.py:809] (1/4) Epoch 27, batch 1600, loss[ctc_loss=0.05773, att_loss=0.2033, loss=0.1741, over 15751.00 frames. utt_duration=1659 frames, utt_pad_proportion=0.009301, over 38.00 utterances.], tot_loss[ctc_loss=0.06584, att_loss=0.2313, loss=0.1982, over 3271402.95 frames. utt_duration=1258 frames, utt_pad_proportion=0.0525, over 10416.41 utterances.], batch size: 38, lr: 3.98e-03, grad_scale: 8.0 2023-03-09 09:22:58,649 INFO [train2.py:809] (1/4) Epoch 27, batch 1650, loss[ctc_loss=0.05278, att_loss=0.2266, loss=0.1918, over 16527.00 frames. utt_duration=1471 frames, utt_pad_proportion=0.006947, over 45.00 utterances.], tot_loss[ctc_loss=0.06681, att_loss=0.2323, loss=0.1992, over 3272990.08 frames. utt_duration=1241 frames, utt_pad_proportion=0.05692, over 10560.37 utterances.], batch size: 45, lr: 3.98e-03, grad_scale: 8.0 2023-03-09 09:23:34,348 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.134e+02 2.028e+02 2.397e+02 2.825e+02 1.110e+03, threshold=4.794e+02, percent-clipped=4.0 2023-03-09 09:23:55,192 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9969, 3.9665, 3.7878, 2.7772, 3.7990, 3.8288, 3.4404, 2.6168], device='cuda:1'), covar=tensor([0.0115, 0.0149, 0.0264, 0.0916, 0.0151, 0.0371, 0.0387, 0.1307], device='cuda:1'), in_proj_covar=tensor([0.0078, 0.0106, 0.0110, 0.0112, 0.0090, 0.0118, 0.0101, 0.0104], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-09 09:24:17,723 INFO [train2.py:809] (1/4) Epoch 27, batch 1700, loss[ctc_loss=0.08465, att_loss=0.2537, loss=0.2199, over 17384.00 frames. utt_duration=1105 frames, utt_pad_proportion=0.03401, over 63.00 utterances.], tot_loss[ctc_loss=0.06696, att_loss=0.2325, loss=0.1994, over 3275176.58 frames. utt_duration=1259 frames, utt_pad_proportion=0.05258, over 10416.00 utterances.], batch size: 63, lr: 3.98e-03, grad_scale: 8.0 2023-03-09 09:24:21,274 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.5943, 2.1920, 2.3775, 2.3936, 2.4493, 2.6365, 2.1275, 2.5612], device='cuda:1'), covar=tensor([0.1260, 0.2370, 0.1830, 0.1205, 0.2492, 0.1296, 0.1637, 0.1509], device='cuda:1'), in_proj_covar=tensor([0.0142, 0.0144, 0.0138, 0.0135, 0.0152, 0.0128, 0.0152, 0.0128], 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') 2023-03-09 09:24:40,923 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8765, 5.2041, 4.8056, 5.2323, 4.5984, 4.8363, 5.3492, 5.1092], device='cuda:1'), covar=tensor([0.0627, 0.0297, 0.0772, 0.0332, 0.0454, 0.0299, 0.0242, 0.0223], device='cuda:1'), in_proj_covar=tensor([0.0399, 0.0336, 0.0378, 0.0374, 0.0337, 0.0246, 0.0316, 0.0300], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-09 09:25:07,312 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5049, 4.5558, 4.6869, 4.6983, 5.2175, 4.5849, 4.5622, 2.7991], device='cuda:1'), covar=tensor([0.0279, 0.0384, 0.0340, 0.0306, 0.0619, 0.0261, 0.0378, 0.1503], device='cuda:1'), in_proj_covar=tensor([0.0191, 0.0220, 0.0215, 0.0232, 0.0385, 0.0190, 0.0205, 0.0222], 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') 2023-03-09 09:25:35,433 INFO [train2.py:809] (1/4) Epoch 27, batch 1750, loss[ctc_loss=0.05316, att_loss=0.2241, loss=0.1899, over 16706.00 frames. utt_duration=1454 frames, utt_pad_proportion=0.005531, over 46.00 utterances.], tot_loss[ctc_loss=0.06697, att_loss=0.2321, loss=0.1991, over 3279547.36 frames. utt_duration=1267 frames, utt_pad_proportion=0.04879, over 10364.64 utterances.], batch size: 46, lr: 3.98e-03, grad_scale: 8.0 2023-03-09 09:26:11,488 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.189e+02 1.890e+02 2.121e+02 2.571e+02 3.982e+02, threshold=4.242e+02, percent-clipped=0.0 2023-03-09 09:26:30,779 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6087, 4.6946, 4.7560, 4.7244, 5.2910, 4.5844, 4.6699, 2.9303], device='cuda:1'), covar=tensor([0.0279, 0.0383, 0.0325, 0.0418, 0.0649, 0.0266, 0.0381, 0.1585], device='cuda:1'), in_proj_covar=tensor([0.0191, 0.0220, 0.0216, 0.0232, 0.0384, 0.0190, 0.0205, 0.0223], 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') 2023-03-09 09:26:54,623 INFO [train2.py:809] (1/4) Epoch 27, batch 1800, loss[ctc_loss=0.05094, att_loss=0.2356, loss=0.1986, over 16634.00 frames. utt_duration=1417 frames, utt_pad_proportion=0.004901, over 47.00 utterances.], tot_loss[ctc_loss=0.06644, att_loss=0.2319, loss=0.1988, over 3280363.44 frames. utt_duration=1273 frames, utt_pad_proportion=0.0481, over 10319.12 utterances.], batch size: 47, lr: 3.98e-03, grad_scale: 8.0 2023-03-09 09:27:08,786 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105388.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:28:15,235 INFO [train2.py:809] (1/4) Epoch 27, batch 1850, loss[ctc_loss=0.07624, att_loss=0.241, loss=0.2081, over 17384.00 frames. utt_duration=1105 frames, utt_pad_proportion=0.03305, over 63.00 utterances.], tot_loss[ctc_loss=0.06668, att_loss=0.2321, loss=0.199, over 3274293.21 frames. utt_duration=1260 frames, utt_pad_proportion=0.05298, over 10410.52 utterances.], batch size: 63, lr: 3.97e-03, grad_scale: 16.0 2023-03-09 09:28:20,271 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4908, 5.1008, 5.1975, 5.2285, 5.0539, 5.1640, 4.8842, 4.6348], device='cuda:1'), covar=tensor([0.1677, 0.0910, 0.0403, 0.0615, 0.0752, 0.0434, 0.0542, 0.0451], device='cuda:1'), in_proj_covar=tensor([0.0538, 0.0383, 0.0371, 0.0377, 0.0442, 0.0448, 0.0380, 0.0412], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:1') 2023-03-09 09:28:47,148 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105449.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 09:28:51,301 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.205e+02 1.738e+02 2.150e+02 2.856e+02 4.488e+02, threshold=4.299e+02, percent-clipped=4.0 2023-03-09 09:29:34,710 INFO [train2.py:809] (1/4) Epoch 27, batch 1900, loss[ctc_loss=0.07724, att_loss=0.2475, loss=0.2134, over 17331.00 frames. utt_duration=1176 frames, utt_pad_proportion=0.02205, over 59.00 utterances.], tot_loss[ctc_loss=0.0665, att_loss=0.2327, loss=0.1994, over 3277180.07 frames. utt_duration=1241 frames, utt_pad_proportion=0.05723, over 10575.04 utterances.], batch size: 59, lr: 3.97e-03, grad_scale: 8.0 2023-03-09 09:29:45,585 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105486.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:29:56,076 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-03-09 09:30:11,366 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105502.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:30:34,527 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.8273, 2.4964, 2.7453, 2.7378, 2.8571, 3.0247, 2.6865, 3.1377], device='cuda:1'), covar=tensor([0.1458, 0.2430, 0.1649, 0.1438, 0.1768, 0.1011, 0.1798, 0.1356], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0145, 0.0140, 0.0136, 0.0153, 0.0129, 0.0153, 0.0129], 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') 2023-03-09 09:30:54,635 INFO [train2.py:809] (1/4) Epoch 27, batch 1950, loss[ctc_loss=0.06089, att_loss=0.2461, loss=0.2091, over 17433.00 frames. utt_duration=884.3 frames, utt_pad_proportion=0.07399, over 79.00 utterances.], tot_loss[ctc_loss=0.06634, att_loss=0.2323, loss=0.1991, over 3281512.38 frames. utt_duration=1254 frames, utt_pad_proportion=0.05317, over 10478.33 utterances.], batch size: 79, lr: 3.97e-03, grad_scale: 8.0 2023-03-09 09:31:23,659 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105547.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:31:32,461 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.251e+02 1.913e+02 2.076e+02 2.767e+02 9.695e+02, threshold=4.152e+02, percent-clipped=4.0 2023-03-09 09:31:49,354 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105563.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:32:01,136 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-03-09 09:32:14,217 INFO [train2.py:809] (1/4) Epoch 27, batch 2000, loss[ctc_loss=0.06983, att_loss=0.2283, loss=0.1966, over 15938.00 frames. utt_duration=1557 frames, utt_pad_proportion=0.007773, over 41.00 utterances.], tot_loss[ctc_loss=0.06584, att_loss=0.2313, loss=0.1982, over 3268333.30 frames. utt_duration=1265 frames, utt_pad_proportion=0.05386, over 10344.89 utterances.], batch size: 41, lr: 3.97e-03, grad_scale: 8.0 2023-03-09 09:32:17,588 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.8527, 2.5638, 2.3853, 2.7070, 3.0587, 2.9722, 2.6401, 3.3160], device='cuda:1'), covar=tensor([0.1252, 0.2431, 0.1904, 0.1293, 0.1397, 0.0878, 0.1827, 0.0866], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0146, 0.0140, 0.0136, 0.0153, 0.0130, 0.0154, 0.0130], 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') 2023-03-09 09:33:34,517 INFO [train2.py:809] (1/4) Epoch 27, batch 2050, loss[ctc_loss=0.07771, att_loss=0.2536, loss=0.2185, over 17036.00 frames. utt_duration=1287 frames, utt_pad_proportion=0.01057, over 53.00 utterances.], tot_loss[ctc_loss=0.06525, att_loss=0.231, loss=0.1978, over 3270611.16 frames. utt_duration=1268 frames, utt_pad_proportion=0.05144, over 10328.61 utterances.], batch size: 53, lr: 3.97e-03, grad_scale: 8.0 2023-03-09 09:34:02,092 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-03-09 09:34:03,450 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.62 vs. limit=5.0 2023-03-09 09:34:11,424 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.161e+02 1.673e+02 2.035e+02 2.438e+02 4.266e+02, threshold=4.071e+02, percent-clipped=1.0 2023-03-09 09:34:31,425 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6514, 2.3078, 5.1117, 4.1431, 3.1984, 4.4962, 4.8509, 4.8629], device='cuda:1'), covar=tensor([0.0211, 0.1593, 0.0149, 0.0758, 0.1533, 0.0194, 0.0132, 0.0184], device='cuda:1'), in_proj_covar=tensor([0.0228, 0.0246, 0.0220, 0.0326, 0.0273, 0.0237, 0.0210, 0.0235], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 09:34:33,463 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=2.90 vs. limit=5.0 2023-03-09 09:34:53,529 INFO [train2.py:809] (1/4) Epoch 27, batch 2100, loss[ctc_loss=0.06302, att_loss=0.2528, loss=0.2148, over 16613.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.005782, over 47.00 utterances.], tot_loss[ctc_loss=0.06552, att_loss=0.2315, loss=0.1983, over 3274284.18 frames. utt_duration=1273 frames, utt_pad_proportion=0.04957, over 10304.43 utterances.], batch size: 47, lr: 3.97e-03, grad_scale: 8.0 2023-03-09 09:35:07,702 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.8968, 3.6519, 3.0563, 3.1806, 3.8184, 3.5186, 2.6833, 4.0377], device='cuda:1'), covar=tensor([0.1085, 0.0551, 0.1173, 0.0866, 0.0799, 0.0800, 0.1026, 0.0467], device='cuda:1'), in_proj_covar=tensor([0.0207, 0.0227, 0.0231, 0.0207, 0.0290, 0.0249, 0.0204, 0.0297], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-09 09:36:13,362 INFO [train2.py:809] (1/4) Epoch 27, batch 2150, loss[ctc_loss=0.1104, att_loss=0.2639, loss=0.2332, over 14006.00 frames. utt_duration=387.8 frames, utt_pad_proportion=0.3256, over 145.00 utterances.], tot_loss[ctc_loss=0.06472, att_loss=0.2311, loss=0.1979, over 3275407.51 frames. utt_duration=1265 frames, utt_pad_proportion=0.05014, over 10373.02 utterances.], batch size: 145, lr: 3.97e-03, grad_scale: 8.0 2023-03-09 09:36:37,237 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105744.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 09:36:49,064 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-03-09 09:36:51,251 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.182e+02 1.815e+02 2.188e+02 2.690e+02 5.528e+02, threshold=4.376e+02, percent-clipped=1.0 2023-03-09 09:37:12,935 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-03-09 09:37:32,706 INFO [train2.py:809] (1/4) Epoch 27, batch 2200, loss[ctc_loss=0.06466, att_loss=0.2452, loss=0.2091, over 17376.00 frames. utt_duration=1179 frames, utt_pad_proportion=0.02038, over 59.00 utterances.], tot_loss[ctc_loss=0.06484, att_loss=0.232, loss=0.1986, over 3287012.70 frames. utt_duration=1246 frames, utt_pad_proportion=0.04919, over 10561.94 utterances.], batch size: 59, lr: 3.97e-03, grad_scale: 8.0 2023-03-09 09:37:51,448 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.3661, 5.3001, 5.1419, 3.2900, 5.1159, 4.9840, 4.6110, 3.1945], device='cuda:1'), covar=tensor([0.0132, 0.0086, 0.0264, 0.0880, 0.0098, 0.0173, 0.0285, 0.1111], device='cuda:1'), in_proj_covar=tensor([0.0078, 0.0106, 0.0110, 0.0112, 0.0090, 0.0118, 0.0101, 0.0104], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-09 09:37:53,014 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105792.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:38:04,355 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105799.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:38:33,066 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0719, 5.4096, 4.9854, 5.4405, 4.8399, 5.0096, 5.5245, 5.2844], device='cuda:1'), covar=tensor([0.0613, 0.0301, 0.0779, 0.0360, 0.0420, 0.0250, 0.0231, 0.0210], device='cuda:1'), in_proj_covar=tensor([0.0400, 0.0338, 0.0380, 0.0374, 0.0336, 0.0247, 0.0318, 0.0301], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-09 09:38:52,246 INFO [train2.py:809] (1/4) Epoch 27, batch 2250, loss[ctc_loss=0.06046, att_loss=0.2229, loss=0.1904, over 16391.00 frames. utt_duration=1492 frames, utt_pad_proportion=0.00747, over 44.00 utterances.], tot_loss[ctc_loss=0.06567, att_loss=0.2324, loss=0.199, over 3280261.79 frames. utt_duration=1261 frames, utt_pad_proportion=0.04727, over 10420.47 utterances.], batch size: 44, lr: 3.97e-03, grad_scale: 8.0 2023-03-09 09:39:13,153 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105842.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:39:30,859 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.345e+02 1.815e+02 2.086e+02 2.504e+02 6.374e+02, threshold=4.173e+02, percent-clipped=3.0 2023-03-09 09:39:31,351 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105853.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:39:39,174 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105858.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:39:43,177 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105860.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:40:12,505 INFO [train2.py:809] (1/4) Epoch 27, batch 2300, loss[ctc_loss=0.05207, att_loss=0.2377, loss=0.2006, over 16974.00 frames. utt_duration=1359 frames, utt_pad_proportion=0.006345, over 50.00 utterances.], tot_loss[ctc_loss=0.06588, att_loss=0.233, loss=0.1996, over 3294907.57 frames. utt_duration=1259 frames, utt_pad_proportion=0.04329, over 10484.26 utterances.], batch size: 50, lr: 3.97e-03, grad_scale: 8.0 2023-03-09 09:41:08,220 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8836, 5.1518, 4.7580, 5.1978, 4.5966, 4.8050, 5.2821, 5.0673], device='cuda:1'), covar=tensor([0.0595, 0.0327, 0.0863, 0.0335, 0.0457, 0.0324, 0.0224, 0.0231], device='cuda:1'), in_proj_covar=tensor([0.0400, 0.0338, 0.0381, 0.0374, 0.0336, 0.0247, 0.0318, 0.0301], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-09 09:41:31,987 INFO [train2.py:809] (1/4) Epoch 27, batch 2350, loss[ctc_loss=0.1109, att_loss=0.2647, loss=0.234, over 17046.00 frames. utt_duration=683.5 frames, utt_pad_proportion=0.136, over 100.00 utterances.], tot_loss[ctc_loss=0.06716, att_loss=0.2336, loss=0.2003, over 3295523.71 frames. utt_duration=1254 frames, utt_pad_proportion=0.04487, over 10528.54 utterances.], batch size: 100, lr: 3.97e-03, grad_scale: 8.0 2023-03-09 09:41:58,741 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1843, 4.4880, 4.5060, 4.5643, 5.0549, 4.4189, 4.4749, 2.4867], device='cuda:1'), covar=tensor([0.0371, 0.0362, 0.0378, 0.0319, 0.0634, 0.0275, 0.0346, 0.1915], device='cuda:1'), in_proj_covar=tensor([0.0193, 0.0221, 0.0217, 0.0233, 0.0384, 0.0192, 0.0207, 0.0224], 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') 2023-03-09 09:42:09,624 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.259e+02 1.971e+02 2.209e+02 2.522e+02 5.529e+02, threshold=4.418e+02, percent-clipped=2.0 2023-03-09 09:42:28,864 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-03-09 09:42:49,237 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-03-09 09:42:51,498 INFO [train2.py:809] (1/4) Epoch 27, batch 2400, loss[ctc_loss=0.08321, att_loss=0.2426, loss=0.2107, over 16966.00 frames. utt_duration=686.9 frames, utt_pad_proportion=0.1392, over 99.00 utterances.], tot_loss[ctc_loss=0.06751, att_loss=0.2338, loss=0.2005, over 3291795.18 frames. utt_duration=1220 frames, utt_pad_proportion=0.05401, over 10808.72 utterances.], batch size: 99, lr: 3.96e-03, grad_scale: 8.0 2023-03-09 09:44:16,149 INFO [train2.py:809] (1/4) Epoch 27, batch 2450, loss[ctc_loss=0.05737, att_loss=0.2421, loss=0.2052, over 16673.00 frames. utt_duration=1451 frames, utt_pad_proportion=0.006577, over 46.00 utterances.], tot_loss[ctc_loss=0.06739, att_loss=0.2338, loss=0.2005, over 3299925.81 frames. utt_duration=1229 frames, utt_pad_proportion=0.04927, over 10750.91 utterances.], batch size: 46, lr: 3.96e-03, grad_scale: 4.0 2023-03-09 09:44:40,225 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106044.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:44:55,919 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.307e+02 1.876e+02 2.325e+02 2.817e+02 5.594e+02, threshold=4.649e+02, percent-clipped=2.0 2023-03-09 09:44:59,517 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3057, 2.9874, 3.5946, 2.7675, 3.4262, 4.4115, 4.2704, 3.0041], device='cuda:1'), covar=tensor([0.0377, 0.1651, 0.1275, 0.1566, 0.1173, 0.0948, 0.0576, 0.1480], device='cuda:1'), in_proj_covar=tensor([0.0249, 0.0249, 0.0290, 0.0221, 0.0271, 0.0382, 0.0274, 0.0235], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-09 09:45:33,190 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106077.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:45:35,775 INFO [train2.py:809] (1/4) Epoch 27, batch 2500, loss[ctc_loss=0.08616, att_loss=0.245, loss=0.2132, over 17285.00 frames. utt_duration=1259 frames, utt_pad_proportion=0.01199, over 55.00 utterances.], tot_loss[ctc_loss=0.06727, att_loss=0.2339, loss=0.2006, over 3298521.65 frames. utt_duration=1231 frames, utt_pad_proportion=0.05036, over 10731.93 utterances.], batch size: 55, lr: 3.96e-03, grad_scale: 4.0 2023-03-09 09:45:37,574 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.7601, 4.9360, 5.4453, 4.8770, 4.8645, 5.6369, 5.1005, 5.6422], device='cuda:1'), covar=tensor([0.1197, 0.1559, 0.1313, 0.2496, 0.2995, 0.1521, 0.1166, 0.1182], device='cuda:1'), in_proj_covar=tensor([0.0911, 0.0523, 0.0640, 0.0683, 0.0907, 0.0661, 0.0508, 0.0641], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 09:45:57,173 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=106092.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:46:08,433 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-03-09 09:46:56,810 INFO [train2.py:809] (1/4) Epoch 27, batch 2550, loss[ctc_loss=0.1194, att_loss=0.2631, loss=0.2343, over 14462.00 frames. utt_duration=400.5 frames, utt_pad_proportion=0.3035, over 145.00 utterances.], tot_loss[ctc_loss=0.06704, att_loss=0.234, loss=0.2006, over 3301052.62 frames. utt_duration=1227 frames, utt_pad_proportion=0.05126, over 10772.75 utterances.], batch size: 145, lr: 3.96e-03, grad_scale: 4.0 2023-03-09 09:47:12,049 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106138.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:47:18,016 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106142.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:47:26,189 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9642, 5.2480, 5.4886, 5.2804, 5.4255, 5.9214, 5.1740, 5.9945], device='cuda:1'), covar=tensor([0.0671, 0.0771, 0.0878, 0.1394, 0.1637, 0.0862, 0.0762, 0.0608], device='cuda:1'), in_proj_covar=tensor([0.0914, 0.0523, 0.0641, 0.0683, 0.0908, 0.0663, 0.0509, 0.0642], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 09:47:27,608 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106148.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:47:36,815 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.077e+02 1.781e+02 2.163e+02 2.660e+02 6.949e+02, threshold=4.326e+02, percent-clipped=5.0 2023-03-09 09:47:39,216 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106155.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:47:44,022 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106158.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:48:16,576 INFO [train2.py:809] (1/4) Epoch 27, batch 2600, loss[ctc_loss=0.09885, att_loss=0.2583, loss=0.2264, over 13945.00 frames. utt_duration=383.6 frames, utt_pad_proportion=0.3306, over 146.00 utterances.], tot_loss[ctc_loss=0.06593, att_loss=0.2327, loss=0.1994, over 3284605.85 frames. utt_duration=1224 frames, utt_pad_proportion=0.05657, over 10751.08 utterances.], batch size: 146, lr: 3.96e-03, grad_scale: 4.0 2023-03-09 09:48:20,506 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.38 vs. limit=5.0 2023-03-09 09:48:33,840 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=106190.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:49:01,024 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=106206.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:49:02,893 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.8483, 2.4499, 2.7220, 2.7141, 2.7851, 3.0933, 2.7347, 3.2393], device='cuda:1'), covar=tensor([0.2470, 0.2178, 0.1486, 0.1270, 0.1508, 0.0956, 0.1571, 0.1209], device='cuda:1'), in_proj_covar=tensor([0.0143, 0.0145, 0.0140, 0.0135, 0.0153, 0.0129, 0.0151, 0.0129], 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') 2023-03-09 09:49:37,263 INFO [train2.py:809] (1/4) Epoch 27, batch 2650, loss[ctc_loss=0.0532, att_loss=0.2156, loss=0.1831, over 14964.00 frames. utt_duration=1816 frames, utt_pad_proportion=0.0275, over 33.00 utterances.], tot_loss[ctc_loss=0.06552, att_loss=0.2316, loss=0.1984, over 3273884.53 frames. utt_duration=1228 frames, utt_pad_proportion=0.05899, over 10677.89 utterances.], batch size: 33, lr: 3.96e-03, grad_scale: 4.0 2023-03-09 09:50:17,425 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.073e+02 1.781e+02 2.178e+02 2.674e+02 4.972e+02, threshold=4.357e+02, percent-clipped=1.0 2023-03-09 09:50:39,612 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106268.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:50:55,949 INFO [train2.py:809] (1/4) Epoch 27, batch 2700, loss[ctc_loss=0.07945, att_loss=0.25, loss=0.2159, over 17263.00 frames. utt_duration=1257 frames, utt_pad_proportion=0.01399, over 55.00 utterances.], tot_loss[ctc_loss=0.06587, att_loss=0.2315, loss=0.1984, over 3267817.48 frames. utt_duration=1250 frames, utt_pad_proportion=0.05476, over 10472.71 utterances.], batch size: 55, lr: 3.96e-03, grad_scale: 4.0 2023-03-09 09:52:15,257 INFO [train2.py:809] (1/4) Epoch 27, batch 2750, loss[ctc_loss=0.06269, att_loss=0.2263, loss=0.1935, over 16893.00 frames. utt_duration=1380 frames, utt_pad_proportion=0.006901, over 49.00 utterances.], tot_loss[ctc_loss=0.06561, att_loss=0.2313, loss=0.1981, over 3271943.72 frames. utt_duration=1261 frames, utt_pad_proportion=0.05133, over 10394.90 utterances.], batch size: 49, lr: 3.96e-03, grad_scale: 4.0 2023-03-09 09:52:15,618 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106329.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:52:27,156 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106336.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:52:55,847 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.096e+02 1.858e+02 2.130e+02 2.638e+02 1.636e+03, threshold=4.261e+02, percent-clipped=5.0 2023-03-09 09:53:34,595 INFO [train2.py:809] (1/4) Epoch 27, batch 2800, loss[ctc_loss=0.06677, att_loss=0.2194, loss=0.1889, over 16187.00 frames. utt_duration=1580 frames, utt_pad_proportion=0.006623, over 41.00 utterances.], tot_loss[ctc_loss=0.06586, att_loss=0.2314, loss=0.1983, over 3280162.24 frames. utt_duration=1271 frames, utt_pad_proportion=0.04642, over 10335.78 utterances.], batch size: 41, lr: 3.96e-03, grad_scale: 8.0 2023-03-09 09:53:36,929 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-03-09 09:54:04,686 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106397.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:54:55,105 INFO [train2.py:809] (1/4) Epoch 27, batch 2850, loss[ctc_loss=0.05406, att_loss=0.2065, loss=0.176, over 15504.00 frames. utt_duration=1724 frames, utt_pad_proportion=0.008514, over 36.00 utterances.], tot_loss[ctc_loss=0.06512, att_loss=0.2308, loss=0.1977, over 3275541.27 frames. utt_duration=1268 frames, utt_pad_proportion=0.0486, over 10348.20 utterances.], batch size: 36, lr: 3.96e-03, grad_scale: 8.0 2023-03-09 09:54:58,877 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.8247, 2.4041, 2.4234, 2.7753, 2.7284, 2.9354, 2.6281, 3.1335], device='cuda:1'), covar=tensor([0.1348, 0.1847, 0.1489, 0.1193, 0.1813, 0.0822, 0.1492, 0.0928], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0143, 0.0139, 0.0133, 0.0151, 0.0128, 0.0150, 0.0128], 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') 2023-03-09 09:55:01,774 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106433.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:55:23,167 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.94 vs. limit=5.0 2023-03-09 09:55:26,913 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106448.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:55:36,025 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.102e+02 1.809e+02 2.032e+02 2.602e+02 5.247e+02, threshold=4.065e+02, percent-clipped=2.0 2023-03-09 09:55:36,428 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106454.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:55:37,997 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106455.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:56:15,401 INFO [train2.py:809] (1/4) Epoch 27, batch 2900, loss[ctc_loss=0.06894, att_loss=0.2325, loss=0.1998, over 16765.00 frames. utt_duration=678.9 frames, utt_pad_proportion=0.1493, over 99.00 utterances.], tot_loss[ctc_loss=0.06525, att_loss=0.2315, loss=0.1983, over 3287452.60 frames. utt_duration=1277 frames, utt_pad_proportion=0.04326, over 10309.53 utterances.], batch size: 99, lr: 3.96e-03, grad_scale: 8.0 2023-03-09 09:56:42,922 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=106496.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:56:53,780 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=106503.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:57:13,234 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106515.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 09:57:13,792 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-03-09 09:57:34,533 INFO [train2.py:809] (1/4) Epoch 27, batch 2950, loss[ctc_loss=0.07747, att_loss=0.2425, loss=0.2095, over 17465.00 frames. utt_duration=885.8 frames, utt_pad_proportion=0.07245, over 79.00 utterances.], tot_loss[ctc_loss=0.06612, att_loss=0.2321, loss=0.1989, over 3287200.39 frames. utt_duration=1258 frames, utt_pad_proportion=0.04758, over 10467.46 utterances.], batch size: 79, lr: 3.95e-03, grad_scale: 8.0 2023-03-09 09:58:15,060 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.306e+02 2.110e+02 2.427e+02 2.893e+02 4.636e+02, threshold=4.854e+02, percent-clipped=4.0 2023-03-09 09:58:52,852 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9912, 2.7973, 3.5239, 2.6563, 3.3334, 4.2042, 4.0622, 2.8686], device='cuda:1'), covar=tensor([0.0466, 0.1650, 0.1116, 0.1518, 0.1209, 0.0906, 0.0612, 0.1459], device='cuda:1'), in_proj_covar=tensor([0.0250, 0.0249, 0.0291, 0.0222, 0.0270, 0.0382, 0.0275, 0.0237], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-09 09:58:53,987 INFO [train2.py:809] (1/4) Epoch 27, batch 3000, loss[ctc_loss=0.07013, att_loss=0.2411, loss=0.2069, over 17260.00 frames. utt_duration=875.1 frames, utt_pad_proportion=0.07884, over 79.00 utterances.], tot_loss[ctc_loss=0.06674, att_loss=0.2326, loss=0.1995, over 3286772.28 frames. utt_duration=1230 frames, utt_pad_proportion=0.0557, over 10699.10 utterances.], batch size: 79, lr: 3.95e-03, grad_scale: 8.0 2023-03-09 09:58:53,987 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-09 09:59:07,590 INFO [train2.py:843] (1/4) Epoch 27, validation: ctc_loss=0.04056, att_loss=0.2346, loss=0.1958, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-09 09:59:07,590 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16129MB 2023-03-09 10:00:20,316 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106624.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 10:00:28,229 INFO [train2.py:809] (1/4) Epoch 27, batch 3050, loss[ctc_loss=0.04411, att_loss=0.2051, loss=0.1729, over 15360.00 frames. utt_duration=1757 frames, utt_pad_proportion=0.01181, over 35.00 utterances.], tot_loss[ctc_loss=0.06643, att_loss=0.2325, loss=0.1993, over 3280629.45 frames. utt_duration=1246 frames, utt_pad_proportion=0.05269, over 10545.89 utterances.], batch size: 35, lr: 3.95e-03, grad_scale: 8.0 2023-03-09 10:01:08,157 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.208e+02 1.917e+02 2.259e+02 2.641e+02 5.465e+02, threshold=4.518e+02, percent-clipped=3.0 2023-03-09 10:01:46,963 INFO [train2.py:809] (1/4) Epoch 27, batch 3100, loss[ctc_loss=0.05877, att_loss=0.2135, loss=0.1826, over 15784.00 frames. utt_duration=1663 frames, utt_pad_proportion=0.007615, over 38.00 utterances.], tot_loss[ctc_loss=0.06628, att_loss=0.2329, loss=0.1996, over 3286782.26 frames. utt_duration=1258 frames, utt_pad_proportion=0.04712, over 10464.22 utterances.], batch size: 38, lr: 3.95e-03, grad_scale: 8.0 2023-03-09 10:02:08,305 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106692.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 10:02:11,570 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-03-09 10:02:42,492 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7851, 2.4993, 2.5823, 2.6714, 2.7502, 2.8969, 2.2478, 3.0185], device='cuda:1'), covar=tensor([0.1562, 0.2174, 0.1608, 0.1173, 0.1959, 0.0915, 0.2081, 0.1026], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0142, 0.0138, 0.0133, 0.0150, 0.0128, 0.0149, 0.0128], 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') 2023-03-09 10:03:06,633 INFO [train2.py:809] (1/4) Epoch 27, batch 3150, loss[ctc_loss=0.04085, att_loss=0.2049, loss=0.1721, over 15517.00 frames. utt_duration=1726 frames, utt_pad_proportion=0.007619, over 36.00 utterances.], tot_loss[ctc_loss=0.06652, att_loss=0.2327, loss=0.1995, over 3279078.93 frames. utt_duration=1231 frames, utt_pad_proportion=0.05723, over 10666.68 utterances.], batch size: 36, lr: 3.95e-03, grad_scale: 8.0 2023-03-09 10:03:10,732 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7472, 2.4812, 2.6255, 3.2860, 3.0122, 3.1696, 2.4986, 2.3891], device='cuda:1'), covar=tensor([0.0734, 0.1612, 0.0912, 0.0848, 0.1150, 0.0578, 0.1430, 0.1576], device='cuda:1'), in_proj_covar=tensor([0.0187, 0.0216, 0.0186, 0.0224, 0.0232, 0.0190, 0.0203, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 10:03:13,694 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106733.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 10:03:47,289 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.396e+02 1.911e+02 2.328e+02 2.721e+02 4.436e+02, threshold=4.656e+02, percent-clipped=0.0 2023-03-09 10:04:26,573 INFO [train2.py:809] (1/4) Epoch 27, batch 3200, loss[ctc_loss=0.04437, att_loss=0.2206, loss=0.1853, over 16382.00 frames. utt_duration=1491 frames, utt_pad_proportion=0.007399, over 44.00 utterances.], tot_loss[ctc_loss=0.06568, att_loss=0.2321, loss=0.1989, over 3273939.59 frames. utt_duration=1219 frames, utt_pad_proportion=0.06253, over 10760.62 utterances.], batch size: 44, lr: 3.95e-03, grad_scale: 8.0 2023-03-09 10:04:30,371 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=106781.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 10:05:17,429 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106810.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 10:05:47,180 INFO [train2.py:809] (1/4) Epoch 27, batch 3250, loss[ctc_loss=0.04681, att_loss=0.208, loss=0.1758, over 15749.00 frames. utt_duration=1659 frames, utt_pad_proportion=0.007807, over 38.00 utterances.], tot_loss[ctc_loss=0.06544, att_loss=0.2315, loss=0.1983, over 3277582.08 frames. utt_duration=1240 frames, utt_pad_proportion=0.05703, over 10585.23 utterances.], batch size: 38, lr: 3.95e-03, grad_scale: 8.0 2023-03-09 10:06:27,323 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.302e+02 1.765e+02 2.095e+02 2.388e+02 3.844e+02, threshold=4.190e+02, percent-clipped=0.0 2023-03-09 10:07:07,567 INFO [train2.py:809] (1/4) Epoch 27, batch 3300, loss[ctc_loss=0.0708, att_loss=0.2472, loss=0.212, over 16469.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.006222, over 46.00 utterances.], tot_loss[ctc_loss=0.06582, att_loss=0.2325, loss=0.1992, over 3283602.64 frames. utt_duration=1224 frames, utt_pad_proportion=0.05977, over 10741.20 utterances.], batch size: 46, lr: 3.95e-03, grad_scale: 8.0 2023-03-09 10:08:19,449 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106924.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 10:08:27,393 INFO [train2.py:809] (1/4) Epoch 27, batch 3350, loss[ctc_loss=0.04543, att_loss=0.2038, loss=0.1721, over 15387.00 frames. utt_duration=1760 frames, utt_pad_proportion=0.009744, over 35.00 utterances.], tot_loss[ctc_loss=0.06633, att_loss=0.2326, loss=0.1993, over 3282480.26 frames. utt_duration=1226 frames, utt_pad_proportion=0.05965, over 10720.85 utterances.], batch size: 35, lr: 3.95e-03, grad_scale: 8.0 2023-03-09 10:09:07,450 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.083e+02 1.952e+02 2.382e+02 3.093e+02 6.808e+02, threshold=4.763e+02, percent-clipped=6.0 2023-03-09 10:09:35,730 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=106972.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 10:09:47,293 INFO [train2.py:809] (1/4) Epoch 27, batch 3400, loss[ctc_loss=0.05744, att_loss=0.2315, loss=0.1967, over 16128.00 frames. utt_duration=1537 frames, utt_pad_proportion=0.006264, over 42.00 utterances.], tot_loss[ctc_loss=0.06625, att_loss=0.2326, loss=0.1993, over 3283664.62 frames. utt_duration=1223 frames, utt_pad_proportion=0.06067, over 10751.39 utterances.], batch size: 42, lr: 3.95e-03, grad_scale: 8.0 2023-03-09 10:10:09,118 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106992.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 10:10:09,864 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-03-09 10:10:28,464 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.8994, 5.1917, 5.4440, 5.2221, 5.4057, 5.8882, 5.2237, 5.9490], device='cuda:1'), covar=tensor([0.0740, 0.0773, 0.0832, 0.1450, 0.1700, 0.0861, 0.0740, 0.0644], device='cuda:1'), in_proj_covar=tensor([0.0913, 0.0523, 0.0644, 0.0679, 0.0907, 0.0665, 0.0510, 0.0643], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 10:11:07,807 INFO [train2.py:809] (1/4) Epoch 27, batch 3450, loss[ctc_loss=0.07201, att_loss=0.2221, loss=0.1921, over 15387.00 frames. utt_duration=1760 frames, utt_pad_proportion=0.00888, over 35.00 utterances.], tot_loss[ctc_loss=0.06639, att_loss=0.233, loss=0.1997, over 3282133.00 frames. utt_duration=1210 frames, utt_pad_proportion=0.06399, over 10868.00 utterances.], batch size: 35, lr: 3.95e-03, grad_scale: 8.0 2023-03-09 10:11:25,847 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=107040.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 10:11:48,077 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.134e+02 1.835e+02 2.250e+02 2.844e+02 5.706e+02, threshold=4.501e+02, percent-clipped=2.0 2023-03-09 10:12:26,816 INFO [train2.py:809] (1/4) Epoch 27, batch 3500, loss[ctc_loss=0.08534, att_loss=0.2539, loss=0.2202, over 17126.00 frames. utt_duration=1225 frames, utt_pad_proportion=0.01452, over 56.00 utterances.], tot_loss[ctc_loss=0.06683, att_loss=0.2328, loss=0.1996, over 3279829.51 frames. utt_duration=1220 frames, utt_pad_proportion=0.06202, over 10766.94 utterances.], batch size: 56, lr: 3.94e-03, grad_scale: 8.0 2023-03-09 10:13:15,816 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=107110.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 10:13:45,760 INFO [train2.py:809] (1/4) Epoch 27, batch 3550, loss[ctc_loss=0.09396, att_loss=0.267, loss=0.2324, over 17333.00 frames. utt_duration=1177 frames, utt_pad_proportion=0.02117, over 59.00 utterances.], tot_loss[ctc_loss=0.06648, att_loss=0.2324, loss=0.1992, over 3276238.67 frames. utt_duration=1233 frames, utt_pad_proportion=0.05966, over 10640.66 utterances.], batch size: 59, lr: 3.94e-03, grad_scale: 8.0 2023-03-09 10:14:03,570 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107140.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 10:14:11,402 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1929, 5.2468, 5.0327, 2.4733, 1.9792, 2.9019, 2.3248, 4.0642], device='cuda:1'), covar=tensor([0.0649, 0.0344, 0.0247, 0.4211, 0.5748, 0.2506, 0.3926, 0.1545], device='cuda:1'), in_proj_covar=tensor([0.0360, 0.0298, 0.0278, 0.0250, 0.0337, 0.0330, 0.0261, 0.0367], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-03-09 10:14:22,407 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107152.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 10:14:24,974 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 9.809e+01 1.862e+02 2.158e+02 2.587e+02 5.069e+02, threshold=4.316e+02, percent-clipped=3.0 2023-03-09 10:14:31,572 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=107158.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 10:15:05,782 INFO [train2.py:809] (1/4) Epoch 27, batch 3600, loss[ctc_loss=0.05714, att_loss=0.2131, loss=0.1819, over 16190.00 frames. utt_duration=1581 frames, utt_pad_proportion=0.005829, over 41.00 utterances.], tot_loss[ctc_loss=0.06638, att_loss=0.2325, loss=0.1992, over 3283184.71 frames. utt_duration=1232 frames, utt_pad_proportion=0.05693, over 10671.15 utterances.], batch size: 41, lr: 3.94e-03, grad_scale: 8.0 2023-03-09 10:15:41,163 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107201.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 10:15:59,488 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107213.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 10:16:24,612 INFO [train2.py:809] (1/4) Epoch 27, batch 3650, loss[ctc_loss=0.05875, att_loss=0.239, loss=0.203, over 17453.00 frames. utt_duration=1110 frames, utt_pad_proportion=0.03007, over 63.00 utterances.], tot_loss[ctc_loss=0.06568, att_loss=0.2314, loss=0.1982, over 3278422.89 frames. utt_duration=1261 frames, utt_pad_proportion=0.05012, over 10411.34 utterances.], batch size: 63, lr: 3.94e-03, grad_scale: 8.0 2023-03-09 10:17:03,103 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.137e+02 1.892e+02 2.154e+02 2.728e+02 3.735e+02, threshold=4.308e+02, percent-clipped=0.0 2023-03-09 10:17:35,352 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4867, 4.9336, 3.8101, 5.2075, 4.6764, 4.7391, 4.8580, 4.7779], device='cuda:1'), covar=tensor([0.0796, 0.0540, 0.1662, 0.0423, 0.0375, 0.0417, 0.0663, 0.0391], device='cuda:1'), in_proj_covar=tensor([0.0398, 0.0339, 0.0377, 0.0377, 0.0336, 0.0245, 0.0318, 0.0301], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-09 10:17:43,437 INFO [train2.py:809] (1/4) Epoch 27, batch 3700, loss[ctc_loss=0.08316, att_loss=0.2481, loss=0.2151, over 16467.00 frames. utt_duration=1433 frames, utt_pad_proportion=0.006121, over 46.00 utterances.], tot_loss[ctc_loss=0.0649, att_loss=0.2302, loss=0.1972, over 3276296.46 frames. utt_duration=1283 frames, utt_pad_proportion=0.04488, over 10230.05 utterances.], batch size: 46, lr: 3.94e-03, grad_scale: 8.0 2023-03-09 10:19:01,340 INFO [train2.py:809] (1/4) Epoch 27, batch 3750, loss[ctc_loss=0.06124, att_loss=0.2095, loss=0.1799, over 15398.00 frames. utt_duration=1761 frames, utt_pad_proportion=0.008763, over 35.00 utterances.], tot_loss[ctc_loss=0.06433, att_loss=0.2299, loss=0.1968, over 3275505.67 frames. utt_duration=1295 frames, utt_pad_proportion=0.04139, over 10131.04 utterances.], batch size: 35, lr: 3.94e-03, grad_scale: 8.0 2023-03-09 10:19:40,407 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.252e+02 1.917e+02 2.215e+02 2.702e+02 1.402e+03, threshold=4.430e+02, percent-clipped=2.0 2023-03-09 10:19:52,937 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.5266, 2.9528, 3.5639, 4.4762, 3.9377, 3.9507, 2.9265, 2.4020], device='cuda:1'), covar=tensor([0.0712, 0.1846, 0.0810, 0.0486, 0.0868, 0.0471, 0.1641, 0.2040], device='cuda:1'), in_proj_covar=tensor([0.0190, 0.0219, 0.0189, 0.0226, 0.0233, 0.0192, 0.0206, 0.0193], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 10:20:19,854 INFO [train2.py:809] (1/4) Epoch 27, batch 3800, loss[ctc_loss=0.05566, att_loss=0.2101, loss=0.1792, over 14519.00 frames. utt_duration=1816 frames, utt_pad_proportion=0.03945, over 32.00 utterances.], tot_loss[ctc_loss=0.06528, att_loss=0.2307, loss=0.1976, over 3267352.56 frames. utt_duration=1278 frames, utt_pad_proportion=0.04656, over 10241.94 utterances.], batch size: 32, lr: 3.94e-03, grad_scale: 8.0 2023-03-09 10:20:54,525 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.2760, 5.4956, 5.4360, 5.4560, 5.5299, 5.5023, 5.1569, 4.9810], device='cuda:1'), covar=tensor([0.1008, 0.0572, 0.0319, 0.0517, 0.0298, 0.0323, 0.0406, 0.0326], device='cuda:1'), in_proj_covar=tensor([0.0542, 0.0385, 0.0378, 0.0382, 0.0448, 0.0453, 0.0380, 0.0418], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:1') 2023-03-09 10:21:00,447 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0483, 5.3728, 5.5808, 5.3325, 5.6226, 5.9917, 5.2768, 6.1109], device='cuda:1'), covar=tensor([0.0744, 0.0710, 0.0927, 0.1501, 0.1814, 0.0976, 0.0710, 0.0685], device='cuda:1'), in_proj_covar=tensor([0.0919, 0.0527, 0.0646, 0.0685, 0.0915, 0.0670, 0.0511, 0.0644], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 10:21:03,400 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0158, 5.3076, 5.5682, 5.3737, 5.5686, 6.0063, 5.2780, 6.0971], device='cuda:1'), covar=tensor([0.0732, 0.0719, 0.0863, 0.1404, 0.1765, 0.0847, 0.0712, 0.0643], device='cuda:1'), in_proj_covar=tensor([0.0919, 0.0527, 0.0645, 0.0685, 0.0914, 0.0670, 0.0511, 0.0644], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 10:21:08,237 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7250, 3.3643, 3.3831, 2.9944, 3.3520, 3.3745, 3.3936, 2.4668], device='cuda:1'), covar=tensor([0.1071, 0.1132, 0.1532, 0.2833, 0.1523, 0.1379, 0.0929, 0.3057], device='cuda:1'), in_proj_covar=tensor([0.0202, 0.0207, 0.0223, 0.0271, 0.0184, 0.0284, 0.0206, 0.0228], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 10:21:20,658 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.3639, 2.4512, 3.1402, 2.6367, 3.0592, 3.5061, 3.4202, 2.7842], device='cuda:1'), covar=tensor([0.0532, 0.1761, 0.1141, 0.1109, 0.1001, 0.1453, 0.0661, 0.1146], device='cuda:1'), in_proj_covar=tensor([0.0249, 0.0251, 0.0293, 0.0220, 0.0270, 0.0381, 0.0274, 0.0237], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-09 10:21:39,086 INFO [train2.py:809] (1/4) Epoch 27, batch 3850, loss[ctc_loss=0.04239, att_loss=0.2015, loss=0.1697, over 15885.00 frames. utt_duration=1631 frames, utt_pad_proportion=0.009191, over 39.00 utterances.], tot_loss[ctc_loss=0.06571, att_loss=0.2308, loss=0.1978, over 3269157.84 frames. utt_duration=1275 frames, utt_pad_proportion=0.04812, over 10270.81 utterances.], batch size: 39, lr: 3.94e-03, grad_scale: 8.0 2023-03-09 10:22:13,216 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.9182, 2.5724, 2.6954, 2.6893, 2.9054, 2.9302, 2.4672, 3.1552], device='cuda:1'), covar=tensor([0.1335, 0.1977, 0.1698, 0.1328, 0.1231, 0.0916, 0.1891, 0.0938], device='cuda:1'), in_proj_covar=tensor([0.0141, 0.0141, 0.0139, 0.0133, 0.0149, 0.0129, 0.0149, 0.0127], 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') 2023-03-09 10:22:17,272 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.331e+02 1.947e+02 2.251e+02 2.768e+02 8.015e+02, threshold=4.503e+02, percent-clipped=5.0 2023-03-09 10:22:51,506 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-03-09 10:22:55,311 INFO [train2.py:809] (1/4) Epoch 27, batch 3900, loss[ctc_loss=0.07, att_loss=0.2352, loss=0.2022, over 17042.00 frames. utt_duration=1313 frames, utt_pad_proportion=0.008599, over 52.00 utterances.], tot_loss[ctc_loss=0.06555, att_loss=0.231, loss=0.1979, over 3274012.11 frames. utt_duration=1262 frames, utt_pad_proportion=0.05055, over 10385.68 utterances.], batch size: 52, lr: 3.94e-03, grad_scale: 8.0 2023-03-09 10:23:22,182 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107496.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 10:23:40,496 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107508.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 10:23:43,591 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107510.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 10:23:57,097 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5054, 4.8751, 4.0771, 4.9784, 4.4175, 4.6048, 4.8751, 4.7592], device='cuda:1'), covar=tensor([0.0731, 0.0406, 0.1103, 0.0468, 0.0405, 0.0464, 0.0391, 0.0254], device='cuda:1'), in_proj_covar=tensor([0.0399, 0.0340, 0.0378, 0.0381, 0.0337, 0.0246, 0.0319, 0.0302], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-09 10:24:12,009 INFO [train2.py:809] (1/4) Epoch 27, batch 3950, loss[ctc_loss=0.05726, att_loss=0.2067, loss=0.1768, over 15797.00 frames. utt_duration=1664 frames, utt_pad_proportion=0.007658, over 38.00 utterances.], tot_loss[ctc_loss=0.06537, att_loss=0.2312, loss=0.1981, over 3277425.08 frames. utt_duration=1265 frames, utt_pad_proportion=0.04879, over 10377.34 utterances.], batch size: 38, lr: 3.94e-03, grad_scale: 8.0 2023-03-09 10:24:49,848 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.368e+02 1.912e+02 2.157e+02 2.759e+02 5.743e+02, threshold=4.314e+02, percent-clipped=3.0 2023-03-09 10:25:20,653 INFO [train2.py:809] (1/4) Epoch 28, batch 0, loss[ctc_loss=0.06194, att_loss=0.2446, loss=0.2081, over 16475.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.006252, over 46.00 utterances.], tot_loss[ctc_loss=0.06194, att_loss=0.2446, loss=0.2081, over 16475.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.006252, over 46.00 utterances.], batch size: 46, lr: 3.86e-03, grad_scale: 8.0 2023-03-09 10:25:20,653 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-09 10:25:32,828 INFO [train2.py:843] (1/4) Epoch 28, validation: ctc_loss=0.04041, att_loss=0.2344, loss=0.1956, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-09 10:25:32,829 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16129MB 2023-03-09 10:25:47,315 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107571.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 10:26:27,064 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4419, 2.5299, 4.9905, 3.8420, 3.1138, 4.2454, 4.8164, 4.6689], device='cuda:1'), covar=tensor([0.0335, 0.1494, 0.0247, 0.0962, 0.1597, 0.0289, 0.0262, 0.0309], device='cuda:1'), in_proj_covar=tensor([0.0227, 0.0246, 0.0221, 0.0321, 0.0267, 0.0237, 0.0212, 0.0234], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 10:26:52,867 INFO [train2.py:809] (1/4) Epoch 28, batch 50, loss[ctc_loss=0.05395, att_loss=0.2172, loss=0.1846, over 16272.00 frames. utt_duration=1515 frames, utt_pad_proportion=0.007645, over 43.00 utterances.], tot_loss[ctc_loss=0.06259, att_loss=0.2278, loss=0.1948, over 740634.12 frames. utt_duration=1341 frames, utt_pad_proportion=0.02629, over 2211.10 utterances.], batch size: 43, lr: 3.86e-03, grad_scale: 8.0 2023-03-09 10:27:12,704 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.83 vs. limit=2.0 2023-03-09 10:28:00,269 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.063e+02 1.779e+02 2.036e+02 2.554e+02 4.414e+02, threshold=4.071e+02, percent-clipped=2.0 2023-03-09 10:28:08,378 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107659.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 10:28:12,548 INFO [train2.py:809] (1/4) Epoch 28, batch 100, loss[ctc_loss=0.06903, att_loss=0.2354, loss=0.2021, over 16116.00 frames. utt_duration=1536 frames, utt_pad_proportion=0.006926, over 42.00 utterances.], tot_loss[ctc_loss=0.06354, att_loss=0.2304, loss=0.197, over 1296512.62 frames. utt_duration=1263 frames, utt_pad_proportion=0.05319, over 4111.28 utterances.], batch size: 42, lr: 3.86e-03, grad_scale: 8.0 2023-03-09 10:28:41,357 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-03-09 10:29:27,486 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0216, 5.2984, 5.5917, 5.3915, 5.5107, 5.9837, 5.2626, 6.0442], device='cuda:1'), covar=tensor([0.0756, 0.0797, 0.0860, 0.1324, 0.1894, 0.0910, 0.0723, 0.0737], device='cuda:1'), in_proj_covar=tensor([0.0917, 0.0524, 0.0644, 0.0684, 0.0914, 0.0665, 0.0512, 0.0644], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 10:29:32,145 INFO [train2.py:809] (1/4) Epoch 28, batch 150, loss[ctc_loss=0.05591, att_loss=0.2096, loss=0.1789, over 13182.00 frames. utt_duration=1820 frames, utt_pad_proportion=0.04785, over 29.00 utterances.], tot_loss[ctc_loss=0.06573, att_loss=0.231, loss=0.1979, over 1726254.15 frames. utt_duration=1233 frames, utt_pad_proportion=0.06263, over 5606.51 utterances.], batch size: 29, lr: 3.86e-03, grad_scale: 8.0 2023-03-09 10:29:40,192 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0775, 5.1900, 4.9251, 2.3398, 2.1027, 3.1551, 2.4198, 4.0002], device='cuda:1'), covar=tensor([0.0704, 0.0390, 0.0337, 0.5372, 0.5345, 0.2201, 0.3962, 0.1632], device='cuda:1'), in_proj_covar=tensor([0.0367, 0.0303, 0.0283, 0.0255, 0.0342, 0.0337, 0.0267, 0.0374], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-09 10:29:45,379 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107720.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 10:30:02,330 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.2265, 5.1908, 4.9522, 2.9027, 4.9857, 4.7872, 4.5612, 2.9049], device='cuda:1'), covar=tensor([0.0099, 0.0083, 0.0252, 0.1075, 0.0098, 0.0192, 0.0255, 0.1299], device='cuda:1'), in_proj_covar=tensor([0.0080, 0.0108, 0.0112, 0.0114, 0.0091, 0.0120, 0.0103, 0.0105], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-09 10:30:11,821 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107737.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 10:30:22,158 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-03-09 10:30:38,889 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.211e+02 1.937e+02 2.331e+02 2.865e+02 6.453e+02, threshold=4.662e+02, percent-clipped=4.0 2023-03-09 10:30:51,739 INFO [train2.py:809] (1/4) Epoch 28, batch 200, loss[ctc_loss=0.04703, att_loss=0.213, loss=0.1798, over 15899.00 frames. utt_duration=1632 frames, utt_pad_proportion=0.008319, over 39.00 utterances.], tot_loss[ctc_loss=0.06464, att_loss=0.23, loss=0.1969, over 2066303.81 frames. utt_duration=1236 frames, utt_pad_proportion=0.06135, over 6695.34 utterances.], batch size: 39, lr: 3.86e-03, grad_scale: 8.0 2023-03-09 10:31:45,231 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=107796.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 10:31:48,982 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107798.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 10:32:05,093 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=107808.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 10:32:10,870 INFO [train2.py:809] (1/4) Epoch 28, batch 250, loss[ctc_loss=0.04826, att_loss=0.2131, loss=0.1801, over 15968.00 frames. utt_duration=1559 frames, utt_pad_proportion=0.006218, over 41.00 utterances.], tot_loss[ctc_loss=0.06552, att_loss=0.2306, loss=0.1975, over 2337139.32 frames. utt_duration=1261 frames, utt_pad_proportion=0.05329, over 7420.87 utterances.], batch size: 41, lr: 3.86e-03, grad_scale: 8.0 2023-03-09 10:33:01,108 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=107844.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 10:33:16,918 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.125e+02 1.778e+02 2.118e+02 2.564e+02 4.844e+02, threshold=4.236e+02, percent-clipped=1.0 2023-03-09 10:33:20,621 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=107856.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 10:33:29,917 INFO [train2.py:809] (1/4) Epoch 28, batch 300, loss[ctc_loss=0.0676, att_loss=0.241, loss=0.2063, over 17239.00 frames. utt_duration=874.5 frames, utt_pad_proportion=0.08236, over 79.00 utterances.], tot_loss[ctc_loss=0.06525, att_loss=0.2306, loss=0.1976, over 2543924.19 frames. utt_duration=1247 frames, utt_pad_proportion=0.05525, over 8171.25 utterances.], batch size: 79, lr: 3.86e-03, grad_scale: 8.0 2023-03-09 10:33:36,241 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107866.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 10:34:49,119 INFO [train2.py:809] (1/4) Epoch 28, batch 350, loss[ctc_loss=0.04871, att_loss=0.2062, loss=0.1747, over 15507.00 frames. utt_duration=1724 frames, utt_pad_proportion=0.007831, over 36.00 utterances.], tot_loss[ctc_loss=0.06486, att_loss=0.2303, loss=0.1972, over 2703515.18 frames. utt_duration=1268 frames, utt_pad_proportion=0.05095, over 8536.46 utterances.], batch size: 36, lr: 3.86e-03, grad_scale: 8.0 2023-03-09 10:34:55,380 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9668, 6.2050, 5.6374, 5.9196, 5.9248, 5.3656, 5.7280, 5.4230], device='cuda:1'), covar=tensor([0.1232, 0.0873, 0.0989, 0.0736, 0.0891, 0.1520, 0.2030, 0.2218], device='cuda:1'), in_proj_covar=tensor([0.0558, 0.0638, 0.0487, 0.0475, 0.0452, 0.0481, 0.0642, 0.0542], 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') 2023-03-09 10:34:59,462 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4124, 2.3639, 4.8562, 3.8666, 2.9440, 4.1155, 4.7243, 4.5365], device='cuda:1'), covar=tensor([0.0315, 0.1801, 0.0327, 0.0894, 0.1732, 0.0291, 0.0200, 0.0297], device='cuda:1'), in_proj_covar=tensor([0.0229, 0.0248, 0.0223, 0.0323, 0.0269, 0.0239, 0.0213, 0.0236], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 10:35:55,816 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.165e+02 1.778e+02 2.080e+02 2.502e+02 5.851e+02, threshold=4.160e+02, percent-clipped=2.0 2023-03-09 10:36:02,636 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0531, 5.3025, 5.5725, 5.4121, 5.5655, 5.9712, 5.2486, 6.0722], device='cuda:1'), covar=tensor([0.0741, 0.0832, 0.0914, 0.1375, 0.1831, 0.1043, 0.0824, 0.0716], device='cuda:1'), in_proj_covar=tensor([0.0912, 0.0524, 0.0645, 0.0684, 0.0913, 0.0665, 0.0510, 0.0644], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 10:36:08,361 INFO [train2.py:809] (1/4) Epoch 28, batch 400, loss[ctc_loss=0.05118, att_loss=0.2364, loss=0.1993, over 16627.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.005349, over 47.00 utterances.], tot_loss[ctc_loss=0.06423, att_loss=0.2304, loss=0.1972, over 2833738.09 frames. utt_duration=1265 frames, utt_pad_proportion=0.05019, over 8971.83 utterances.], batch size: 47, lr: 3.86e-03, grad_scale: 8.0 2023-03-09 10:36:30,767 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0178, 4.9859, 4.7571, 2.9191, 4.7424, 4.6931, 4.2489, 2.8906], device='cuda:1'), covar=tensor([0.0122, 0.0118, 0.0278, 0.0995, 0.0116, 0.0201, 0.0304, 0.1218], device='cuda:1'), in_proj_covar=tensor([0.0079, 0.0107, 0.0112, 0.0114, 0.0091, 0.0120, 0.0102, 0.0105], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-09 10:37:33,051 INFO [train2.py:809] (1/4) Epoch 28, batch 450, loss[ctc_loss=0.07879, att_loss=0.2555, loss=0.2201, over 16897.00 frames. utt_duration=1381 frames, utt_pad_proportion=0.006024, over 49.00 utterances.], tot_loss[ctc_loss=0.06455, att_loss=0.2312, loss=0.1979, over 2930319.99 frames. utt_duration=1258 frames, utt_pad_proportion=0.053, over 9328.99 utterances.], batch size: 49, lr: 3.86e-03, grad_scale: 8.0 2023-03-09 10:37:37,598 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108015.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 10:38:20,532 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7028, 5.0145, 5.0022, 5.1284, 5.1526, 4.8524, 3.5522, 5.0594], device='cuda:1'), covar=tensor([0.0138, 0.0150, 0.0189, 0.0084, 0.0121, 0.0136, 0.0775, 0.0246], device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0091, 0.0116, 0.0072, 0.0078, 0.0089, 0.0104, 0.0108], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 10:38:39,469 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.152e+02 1.861e+02 2.240e+02 2.601e+02 4.871e+02, threshold=4.480e+02, percent-clipped=1.0 2023-03-09 10:38:51,889 INFO [train2.py:809] (1/4) Epoch 28, batch 500, loss[ctc_loss=0.07538, att_loss=0.231, loss=0.1999, over 16295.00 frames. utt_duration=1517 frames, utt_pad_proportion=0.006488, over 43.00 utterances.], tot_loss[ctc_loss=0.0644, att_loss=0.2314, loss=0.198, over 3016367.75 frames. utt_duration=1265 frames, utt_pad_proportion=0.04888, over 9551.35 utterances.], batch size: 43, lr: 3.85e-03, grad_scale: 16.0 2023-03-09 10:38:55,277 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4207, 2.8919, 3.6094, 2.9203, 3.4954, 4.4900, 4.3490, 3.1115], device='cuda:1'), covar=tensor([0.0353, 0.1867, 0.1259, 0.1337, 0.1152, 0.0936, 0.0629, 0.1355], device='cuda:1'), in_proj_covar=tensor([0.0248, 0.0249, 0.0289, 0.0218, 0.0269, 0.0379, 0.0272, 0.0236], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-09 10:39:05,316 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.1718, 3.8673, 3.9091, 3.4041, 3.9099, 3.9474, 3.9195, 3.1476], device='cuda:1'), covar=tensor([0.0943, 0.1130, 0.1506, 0.2561, 0.1252, 0.1743, 0.0720, 0.2791], device='cuda:1'), in_proj_covar=tensor([0.0201, 0.0206, 0.0222, 0.0270, 0.0182, 0.0280, 0.0206, 0.0227], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 10:39:41,180 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108093.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 10:39:58,067 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1649, 5.2429, 5.0392, 2.4259, 2.2109, 3.0718, 2.6330, 3.9334], device='cuda:1'), covar=tensor([0.0671, 0.0325, 0.0283, 0.5001, 0.5016, 0.2276, 0.3752, 0.1727], device='cuda:1'), in_proj_covar=tensor([0.0365, 0.0302, 0.0282, 0.0254, 0.0341, 0.0335, 0.0265, 0.0373], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-09 10:40:05,562 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.0670, 3.8350, 3.8711, 3.3100, 3.8494, 3.8672, 3.8503, 2.9384], device='cuda:1'), covar=tensor([0.0880, 0.0960, 0.1142, 0.2547, 0.0897, 0.1710, 0.0645, 0.2822], device='cuda:1'), in_proj_covar=tensor([0.0202, 0.0206, 0.0222, 0.0270, 0.0182, 0.0280, 0.0206, 0.0227], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 10:40:11,190 INFO [train2.py:809] (1/4) Epoch 28, batch 550, loss[ctc_loss=0.05081, att_loss=0.2142, loss=0.1815, over 16008.00 frames. utt_duration=1602 frames, utt_pad_proportion=0.007265, over 40.00 utterances.], tot_loss[ctc_loss=0.06392, att_loss=0.2303, loss=0.1971, over 3068141.79 frames. utt_duration=1266 frames, utt_pad_proportion=0.05077, over 9708.21 utterances.], batch size: 40, lr: 3.85e-03, grad_scale: 16.0 2023-03-09 10:40:43,355 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6988, 4.9235, 4.4097, 4.7703, 4.6205, 4.0609, 4.4117, 4.1660], device='cuda:1'), covar=tensor([0.1335, 0.1243, 0.1129, 0.1132, 0.1208, 0.1933, 0.2366, 0.2775], device='cuda:1'), in_proj_covar=tensor([0.0563, 0.0643, 0.0490, 0.0478, 0.0457, 0.0485, 0.0645, 0.0550], 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') 2023-03-09 10:41:18,473 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.142e+02 1.757e+02 2.157e+02 2.611e+02 5.249e+02, threshold=4.315e+02, percent-clipped=2.0 2023-03-09 10:41:30,579 INFO [train2.py:809] (1/4) Epoch 28, batch 600, loss[ctc_loss=0.0535, att_loss=0.221, loss=0.1875, over 16122.00 frames. utt_duration=1537 frames, utt_pad_proportion=0.006526, over 42.00 utterances.], tot_loss[ctc_loss=0.06431, att_loss=0.2302, loss=0.197, over 3107433.22 frames. utt_duration=1242 frames, utt_pad_proportion=0.05845, over 10018.26 utterances.], batch size: 42, lr: 3.85e-03, grad_scale: 16.0 2023-03-09 10:41:37,584 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108166.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 10:42:49,656 INFO [train2.py:809] (1/4) Epoch 28, batch 650, loss[ctc_loss=0.06823, att_loss=0.2405, loss=0.2061, over 16468.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.006523, over 46.00 utterances.], tot_loss[ctc_loss=0.06489, att_loss=0.2307, loss=0.1975, over 3144945.22 frames. utt_duration=1254 frames, utt_pad_proportion=0.05428, over 10045.66 utterances.], batch size: 46, lr: 3.85e-03, grad_scale: 16.0 2023-03-09 10:42:53,338 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=108214.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 10:43:56,527 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.058e+02 1.814e+02 2.237e+02 2.665e+02 5.467e+02, threshold=4.474e+02, percent-clipped=5.0 2023-03-09 10:44:08,564 INFO [train2.py:809] (1/4) Epoch 28, batch 700, loss[ctc_loss=0.1014, att_loss=0.2587, loss=0.2272, over 14396.00 frames. utt_duration=398.6 frames, utt_pad_proportion=0.3067, over 145.00 utterances.], tot_loss[ctc_loss=0.06545, att_loss=0.2312, loss=0.1981, over 3167933.58 frames. utt_duration=1233 frames, utt_pad_proportion=0.06027, over 10292.72 utterances.], batch size: 145, lr: 3.85e-03, grad_scale: 16.0 2023-03-09 10:44:08,945 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.9182, 2.5099, 2.9615, 2.7722, 2.9397, 2.9419, 2.4951, 3.0004], device='cuda:1'), covar=tensor([0.1906, 0.2998, 0.2123, 0.1690, 0.2207, 0.1504, 0.2651, 0.1653], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0147, 0.0142, 0.0137, 0.0154, 0.0132, 0.0153, 0.0131], 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') 2023-03-09 10:44:46,751 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4994, 3.0736, 3.6178, 3.1190, 3.4878, 4.5848, 4.4013, 3.3580], device='cuda:1'), covar=tensor([0.0403, 0.1779, 0.1340, 0.1315, 0.1249, 0.0940, 0.0589, 0.1238], device='cuda:1'), in_proj_covar=tensor([0.0251, 0.0251, 0.0293, 0.0221, 0.0273, 0.0383, 0.0275, 0.0238], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-09 10:44:50,541 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.40 vs. limit=5.0 2023-03-09 10:45:27,466 INFO [train2.py:809] (1/4) Epoch 28, batch 750, loss[ctc_loss=0.0536, att_loss=0.2313, loss=0.1957, over 16707.00 frames. utt_duration=1454 frames, utt_pad_proportion=0.004657, over 46.00 utterances.], tot_loss[ctc_loss=0.06518, att_loss=0.231, loss=0.1979, over 3187642.49 frames. utt_duration=1238 frames, utt_pad_proportion=0.05937, over 10312.34 utterances.], batch size: 46, lr: 3.85e-03, grad_scale: 16.0 2023-03-09 10:45:32,960 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108315.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 10:45:48,263 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108325.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 10:46:33,463 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.198e+02 1.877e+02 2.184e+02 2.631e+02 4.478e+02, threshold=4.368e+02, percent-clipped=1.0 2023-03-09 10:46:45,999 INFO [train2.py:809] (1/4) Epoch 28, batch 800, loss[ctc_loss=0.0745, att_loss=0.2295, loss=0.1985, over 15355.00 frames. utt_duration=1756 frames, utt_pad_proportion=0.01212, over 35.00 utterances.], tot_loss[ctc_loss=0.06579, att_loss=0.2317, loss=0.1985, over 3207491.32 frames. utt_duration=1206 frames, utt_pad_proportion=0.06649, over 10655.53 utterances.], batch size: 35, lr: 3.85e-03, grad_scale: 16.0 2023-03-09 10:46:47,659 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=108363.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 10:47:10,574 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5224, 2.8101, 4.9065, 3.8776, 2.9815, 4.2250, 4.8054, 4.6841], device='cuda:1'), covar=tensor([0.0298, 0.1452, 0.0290, 0.0915, 0.1746, 0.0285, 0.0191, 0.0275], device='cuda:1'), in_proj_covar=tensor([0.0233, 0.0250, 0.0227, 0.0329, 0.0273, 0.0242, 0.0217, 0.0241], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 10:47:23,509 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108386.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 10:47:34,688 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108393.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 10:48:04,322 INFO [train2.py:809] (1/4) Epoch 28, batch 850, loss[ctc_loss=0.05227, att_loss=0.2062, loss=0.1754, over 15747.00 frames. utt_duration=1659 frames, utt_pad_proportion=0.008866, over 38.00 utterances.], tot_loss[ctc_loss=0.06549, att_loss=0.2315, loss=0.1983, over 3225956.90 frames. utt_duration=1221 frames, utt_pad_proportion=0.06112, over 10583.15 utterances.], batch size: 38, lr: 3.85e-03, grad_scale: 16.0 2023-03-09 10:48:27,377 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.8041, 6.0891, 5.5637, 5.7517, 5.7618, 5.2331, 5.5711, 5.2458], device='cuda:1'), covar=tensor([0.1301, 0.0933, 0.0954, 0.0832, 0.0909, 0.1605, 0.2244, 0.2496], device='cuda:1'), in_proj_covar=tensor([0.0564, 0.0646, 0.0492, 0.0481, 0.0459, 0.0487, 0.0648, 0.0552], 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') 2023-03-09 10:48:35,099 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108432.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 10:48:49,723 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=108441.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 10:49:03,428 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.26 vs. limit=5.0 2023-03-09 10:49:10,179 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.210e+02 1.750e+02 2.196e+02 2.771e+02 5.305e+02, threshold=4.392e+02, percent-clipped=2.0 2023-03-09 10:49:23,200 INFO [train2.py:809] (1/4) Epoch 28, batch 900, loss[ctc_loss=0.0607, att_loss=0.2392, loss=0.2035, over 17158.00 frames. utt_duration=1227 frames, utt_pad_proportion=0.0128, over 56.00 utterances.], tot_loss[ctc_loss=0.0655, att_loss=0.2325, loss=0.1991, over 3247422.94 frames. utt_duration=1212 frames, utt_pad_proportion=0.06015, over 10731.71 utterances.], batch size: 56, lr: 3.85e-03, grad_scale: 16.0 2023-03-09 10:50:12,177 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108493.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 10:50:37,564 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.02 vs. limit=5.0 2023-03-09 10:50:42,018 INFO [train2.py:809] (1/4) Epoch 28, batch 950, loss[ctc_loss=0.05532, att_loss=0.2406, loss=0.2036, over 16625.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.005469, over 47.00 utterances.], tot_loss[ctc_loss=0.06529, att_loss=0.2322, loss=0.1989, over 3255622.22 frames. utt_duration=1218 frames, utt_pad_proportion=0.05917, over 10706.12 utterances.], batch size: 47, lr: 3.85e-03, grad_scale: 16.0 2023-03-09 10:51:36,946 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2599, 2.9678, 3.3183, 4.3433, 3.8710, 3.8803, 2.9199, 2.2729], device='cuda:1'), covar=tensor([0.0839, 0.1862, 0.0904, 0.0619, 0.0914, 0.0537, 0.1470, 0.2191], device='cuda:1'), in_proj_covar=tensor([0.0191, 0.0221, 0.0188, 0.0228, 0.0236, 0.0194, 0.0206, 0.0194], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 10:51:47,830 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.119e+02 1.763e+02 2.072e+02 2.486e+02 7.743e+02, threshold=4.143e+02, percent-clipped=3.0 2023-03-09 10:52:00,761 INFO [train2.py:809] (1/4) Epoch 28, batch 1000, loss[ctc_loss=0.05752, att_loss=0.2115, loss=0.1807, over 11449.00 frames. utt_duration=1834 frames, utt_pad_proportion=0.1883, over 25.00 utterances.], tot_loss[ctc_loss=0.06517, att_loss=0.2318, loss=0.1985, over 3253334.05 frames. utt_duration=1209 frames, utt_pad_proportion=0.06273, over 10778.07 utterances.], batch size: 25, lr: 3.85e-03, grad_scale: 16.0 2023-03-09 10:52:08,899 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.0818, 3.6768, 3.2223, 3.3327, 3.9826, 3.6067, 2.9981, 4.1898], device='cuda:1'), covar=tensor([0.1038, 0.0542, 0.1083, 0.0819, 0.0724, 0.0798, 0.0886, 0.0579], device='cuda:1'), in_proj_covar=tensor([0.0208, 0.0228, 0.0233, 0.0210, 0.0293, 0.0252, 0.0207, 0.0299], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-09 10:52:24,733 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108577.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 10:53:13,952 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-03-09 10:53:20,718 INFO [train2.py:809] (1/4) Epoch 28, batch 1050, loss[ctc_loss=0.07158, att_loss=0.2321, loss=0.2, over 17026.00 frames. utt_duration=1286 frames, utt_pad_proportion=0.01121, over 53.00 utterances.], tot_loss[ctc_loss=0.06514, att_loss=0.2319, loss=0.1985, over 3260931.04 frames. utt_duration=1218 frames, utt_pad_proportion=0.05906, over 10720.34 utterances.], batch size: 53, lr: 3.84e-03, grad_scale: 16.0 2023-03-09 10:53:50,547 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-03-09 10:54:02,774 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108638.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 10:54:27,412 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.257e+02 1.905e+02 2.250e+02 2.754e+02 7.007e+02, threshold=4.500e+02, percent-clipped=3.0 2023-03-09 10:54:40,464 INFO [train2.py:809] (1/4) Epoch 28, batch 1100, loss[ctc_loss=0.07644, att_loss=0.2472, loss=0.213, over 16776.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.006155, over 48.00 utterances.], tot_loss[ctc_loss=0.06496, att_loss=0.2315, loss=0.1982, over 3259147.53 frames. utt_duration=1229 frames, utt_pad_proportion=0.05857, over 10618.45 utterances.], batch size: 48, lr: 3.84e-03, grad_scale: 16.0 2023-03-09 10:55:10,658 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108681.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 10:55:16,910 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3741, 4.3050, 4.4331, 4.4266, 5.0052, 4.3729, 4.3009, 2.4709], device='cuda:1'), covar=tensor([0.0305, 0.0484, 0.0397, 0.0370, 0.0538, 0.0280, 0.0398, 0.1776], device='cuda:1'), in_proj_covar=tensor([0.0192, 0.0220, 0.0215, 0.0231, 0.0377, 0.0191, 0.0206, 0.0219], 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') 2023-03-09 10:56:00,160 INFO [train2.py:809] (1/4) Epoch 28, batch 1150, loss[ctc_loss=0.04724, att_loss=0.1945, loss=0.1651, over 15389.00 frames. utt_duration=1760 frames, utt_pad_proportion=0.01022, over 35.00 utterances.], tot_loss[ctc_loss=0.06531, att_loss=0.2316, loss=0.1984, over 3256142.45 frames. utt_duration=1214 frames, utt_pad_proportion=0.06463, over 10740.89 utterances.], batch size: 35, lr: 3.84e-03, grad_scale: 16.0 2023-03-09 10:56:08,455 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.6909, 2.4671, 2.4610, 2.4808, 2.8719, 2.8098, 2.5225, 3.1366], device='cuda:1'), covar=tensor([0.1435, 0.2396, 0.1761, 0.1616, 0.1516, 0.0993, 0.1848, 0.1312], device='cuda:1'), in_proj_covar=tensor([0.0146, 0.0147, 0.0143, 0.0138, 0.0155, 0.0132, 0.0154, 0.0132], 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') 2023-03-09 10:57:07,178 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.269e+02 1.925e+02 2.338e+02 2.824e+02 7.096e+02, threshold=4.677e+02, percent-clipped=6.0 2023-03-09 10:57:10,353 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4061, 2.9661, 3.5959, 2.9036, 3.4591, 4.5138, 4.3795, 3.2553], device='cuda:1'), covar=tensor([0.0392, 0.1716, 0.1289, 0.1414, 0.1152, 0.0900, 0.0600, 0.1291], device='cuda:1'), in_proj_covar=tensor([0.0246, 0.0248, 0.0287, 0.0217, 0.0268, 0.0377, 0.0270, 0.0234], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-09 10:57:19,312 INFO [train2.py:809] (1/4) Epoch 28, batch 1200, loss[ctc_loss=0.07147, att_loss=0.2437, loss=0.2092, over 17278.00 frames. utt_duration=1258 frames, utt_pad_proportion=0.01256, over 55.00 utterances.], tot_loss[ctc_loss=0.06597, att_loss=0.232, loss=0.1988, over 3249133.03 frames. utt_duration=1193 frames, utt_pad_proportion=0.07227, over 10904.57 utterances.], batch size: 55, lr: 3.84e-03, grad_scale: 16.0 2023-03-09 10:58:00,765 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108788.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 10:58:05,465 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108791.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 10:58:39,211 INFO [train2.py:809] (1/4) Epoch 28, batch 1250, loss[ctc_loss=0.06699, att_loss=0.2041, loss=0.1766, over 15480.00 frames. utt_duration=1722 frames, utt_pad_proportion=0.01005, over 36.00 utterances.], tot_loss[ctc_loss=0.06594, att_loss=0.2327, loss=0.1993, over 3258121.44 frames. utt_duration=1189 frames, utt_pad_proportion=0.06922, over 10976.29 utterances.], batch size: 36, lr: 3.84e-03, grad_scale: 16.0 2023-03-09 10:59:17,538 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9366, 6.1866, 5.6533, 5.8831, 5.8974, 5.3249, 5.6447, 5.3186], device='cuda:1'), covar=tensor([0.1216, 0.0850, 0.0916, 0.0846, 0.1045, 0.1617, 0.2149, 0.2369], device='cuda:1'), in_proj_covar=tensor([0.0557, 0.0643, 0.0488, 0.0475, 0.0454, 0.0482, 0.0644, 0.0548], 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') 2023-03-09 10:59:28,369 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.0551, 3.7521, 3.6640, 3.2372, 3.7103, 3.7831, 3.7525, 2.8056], device='cuda:1'), covar=tensor([0.1087, 0.1078, 0.2274, 0.3027, 0.0897, 0.2281, 0.0949, 0.3251], device='cuda:1'), in_proj_covar=tensor([0.0204, 0.0209, 0.0226, 0.0275, 0.0185, 0.0286, 0.0208, 0.0232], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-03-09 10:59:42,571 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108852.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 10:59:45,115 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.146e+02 1.829e+02 2.128e+02 2.551e+02 3.725e+02, threshold=4.255e+02, percent-clipped=0.0 2023-03-09 10:59:58,169 INFO [train2.py:809] (1/4) Epoch 28, batch 1300, loss[ctc_loss=0.05799, att_loss=0.2089, loss=0.1787, over 15335.00 frames. utt_duration=1754 frames, utt_pad_proportion=0.01358, over 35.00 utterances.], tot_loss[ctc_loss=0.06499, att_loss=0.232, loss=0.1986, over 3259614.28 frames. utt_duration=1227 frames, utt_pad_proportion=0.06032, over 10637.81 utterances.], batch size: 35, lr: 3.84e-03, grad_scale: 16.0 2023-03-09 11:00:34,517 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7188, 5.0689, 4.9088, 5.0502, 5.1519, 4.7672, 3.4720, 5.0431], device='cuda:1'), covar=tensor([0.0122, 0.0104, 0.0146, 0.0076, 0.0080, 0.0116, 0.0697, 0.0181], device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0092, 0.0116, 0.0073, 0.0079, 0.0089, 0.0105, 0.0109], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 11:01:13,691 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.8650, 3.9589, 3.9275, 4.2527, 2.8072, 4.1698, 2.7503, 2.0300], device='cuda:1'), covar=tensor([0.0506, 0.0267, 0.0736, 0.0261, 0.1356, 0.0258, 0.1388, 0.1538], device='cuda:1'), in_proj_covar=tensor([0.0219, 0.0187, 0.0268, 0.0183, 0.0224, 0.0169, 0.0236, 0.0204], 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') 2023-03-09 11:01:17,644 INFO [train2.py:809] (1/4) Epoch 28, batch 1350, loss[ctc_loss=0.07187, att_loss=0.2425, loss=0.2084, over 16335.00 frames. utt_duration=1453 frames, utt_pad_proportion=0.005867, over 45.00 utterances.], tot_loss[ctc_loss=0.06526, att_loss=0.2322, loss=0.1988, over 3265199.47 frames. utt_duration=1214 frames, utt_pad_proportion=0.06195, over 10768.31 utterances.], batch size: 45, lr: 3.84e-03, grad_scale: 16.0 2023-03-09 11:01:23,448 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-03-09 11:01:50,631 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108933.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:02:24,360 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.034e+02 1.787e+02 2.090e+02 2.740e+02 6.253e+02, threshold=4.180e+02, percent-clipped=4.0 2023-03-09 11:02:36,714 INFO [train2.py:809] (1/4) Epoch 28, batch 1400, loss[ctc_loss=0.04865, att_loss=0.2118, loss=0.1792, over 16388.00 frames. utt_duration=1492 frames, utt_pad_proportion=0.008251, over 44.00 utterances.], tot_loss[ctc_loss=0.06536, att_loss=0.2323, loss=0.1989, over 3262191.70 frames. utt_duration=1199 frames, utt_pad_proportion=0.06825, over 10899.26 utterances.], batch size: 44, lr: 3.84e-03, grad_scale: 16.0 2023-03-09 11:03:07,520 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108981.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:03:26,012 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0703, 4.4683, 4.6946, 4.6004, 2.8715, 4.5239, 2.8677, 2.0142], device='cuda:1'), covar=tensor([0.0497, 0.0305, 0.0531, 0.0262, 0.1416, 0.0237, 0.1356, 0.1546], device='cuda:1'), in_proj_covar=tensor([0.0221, 0.0189, 0.0269, 0.0184, 0.0226, 0.0171, 0.0237, 0.0206], 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') 2023-03-09 11:03:56,710 INFO [train2.py:809] (1/4) Epoch 28, batch 1450, loss[ctc_loss=0.05299, att_loss=0.1961, loss=0.1675, over 14585.00 frames. utt_duration=1825 frames, utt_pad_proportion=0.03957, over 32.00 utterances.], tot_loss[ctc_loss=0.06521, att_loss=0.232, loss=0.1986, over 3262142.76 frames. utt_duration=1209 frames, utt_pad_proportion=0.06623, over 10805.40 utterances.], batch size: 32, lr: 3.84e-03, grad_scale: 16.0 2023-03-09 11:04:24,431 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=109029.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:04:36,950 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109037.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 11:04:38,427 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7067, 3.2519, 3.2357, 2.8933, 3.2618, 3.2424, 3.2967, 2.4675], device='cuda:1'), covar=tensor([0.1083, 0.1177, 0.1900, 0.2724, 0.1061, 0.2525, 0.0985, 0.2851], device='cuda:1'), in_proj_covar=tensor([0.0204, 0.0209, 0.0225, 0.0275, 0.0185, 0.0286, 0.0207, 0.0232], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-03-09 11:05:03,519 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.105e+02 1.756e+02 2.072e+02 2.557e+02 4.820e+02, threshold=4.143e+02, percent-clipped=2.0 2023-03-09 11:05:06,499 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-03-09 11:05:16,007 INFO [train2.py:809] (1/4) Epoch 28, batch 1500, loss[ctc_loss=0.06103, att_loss=0.2426, loss=0.2063, over 16470.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.007151, over 46.00 utterances.], tot_loss[ctc_loss=0.06484, att_loss=0.232, loss=0.1986, over 3273144.73 frames. utt_duration=1240 frames, utt_pad_proportion=0.0564, over 10570.05 utterances.], batch size: 46, lr: 3.84e-03, grad_scale: 16.0 2023-03-09 11:05:29,907 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.7479, 5.0589, 5.3179, 5.0961, 5.2708, 5.6971, 5.1230, 5.7818], device='cuda:1'), covar=tensor([0.0823, 0.0821, 0.0921, 0.1517, 0.1918, 0.1016, 0.0780, 0.0804], device='cuda:1'), in_proj_covar=tensor([0.0924, 0.0531, 0.0651, 0.0688, 0.0923, 0.0670, 0.0517, 0.0652], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 11:05:57,492 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109088.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:06:13,823 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109098.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 11:06:34,720 INFO [train2.py:809] (1/4) Epoch 28, batch 1550, loss[ctc_loss=0.05082, att_loss=0.2114, loss=0.1793, over 16178.00 frames. utt_duration=1579 frames, utt_pad_proportion=0.007266, over 41.00 utterances.], tot_loss[ctc_loss=0.06455, att_loss=0.2319, loss=0.1984, over 3267986.14 frames. utt_duration=1227 frames, utt_pad_proportion=0.06172, over 10664.13 utterances.], batch size: 41, lr: 3.84e-03, grad_scale: 16.0 2023-03-09 11:07:05,323 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0815, 5.1135, 4.9089, 2.3254, 2.1201, 3.3768, 2.5417, 3.8985], device='cuda:1'), covar=tensor([0.0683, 0.0351, 0.0279, 0.4893, 0.5182, 0.1926, 0.3779, 0.1654], device='cuda:1'), in_proj_covar=tensor([0.0364, 0.0303, 0.0281, 0.0255, 0.0342, 0.0334, 0.0266, 0.0373], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-09 11:07:12,458 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=109136.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:07:30,584 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109147.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:07:41,452 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.161e+02 1.952e+02 2.217e+02 2.637e+02 4.191e+02, threshold=4.435e+02, percent-clipped=1.0 2023-03-09 11:07:53,854 INFO [train2.py:809] (1/4) Epoch 28, batch 1600, loss[ctc_loss=0.08032, att_loss=0.2448, loss=0.2119, over 17267.00 frames. utt_duration=1172 frames, utt_pad_proportion=0.02556, over 59.00 utterances.], tot_loss[ctc_loss=0.06428, att_loss=0.2316, loss=0.1981, over 3267501.49 frames. utt_duration=1223 frames, utt_pad_proportion=0.0634, over 10704.23 utterances.], batch size: 59, lr: 3.83e-03, grad_scale: 16.0 2023-03-09 11:08:44,700 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109194.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:08:54,863 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.1295, 3.7942, 3.2364, 3.3843, 3.9441, 3.6216, 2.9392, 4.1359], device='cuda:1'), covar=tensor([0.0991, 0.0528, 0.1066, 0.0752, 0.0796, 0.0770, 0.0989, 0.0516], device='cuda:1'), in_proj_covar=tensor([0.0208, 0.0229, 0.0232, 0.0211, 0.0291, 0.0252, 0.0207, 0.0301], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:1') 2023-03-09 11:09:12,892 INFO [train2.py:809] (1/4) Epoch 28, batch 1650, loss[ctc_loss=0.08216, att_loss=0.2505, loss=0.2169, over 16878.00 frames. utt_duration=1379 frames, utt_pad_proportion=0.006964, over 49.00 utterances.], tot_loss[ctc_loss=0.06441, att_loss=0.2313, loss=0.1979, over 3266681.45 frames. utt_duration=1241 frames, utt_pad_proportion=0.05871, over 10539.17 utterances.], batch size: 49, lr: 3.83e-03, grad_scale: 16.0 2023-03-09 11:09:22,379 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1362, 5.0390, 4.7060, 2.7723, 4.8271, 4.7626, 4.3350, 2.7877], device='cuda:1'), covar=tensor([0.0130, 0.0131, 0.0390, 0.1292, 0.0142, 0.0244, 0.0385, 0.1555], device='cuda:1'), in_proj_covar=tensor([0.0079, 0.0106, 0.0112, 0.0113, 0.0089, 0.0118, 0.0102, 0.0104], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-09 11:09:47,306 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109233.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:10:22,240 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.344e+02 1.779e+02 2.256e+02 2.612e+02 6.095e+02, threshold=4.513e+02, percent-clipped=4.0 2023-03-09 11:10:22,712 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109255.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:10:33,280 INFO [train2.py:809] (1/4) Epoch 28, batch 1700, loss[ctc_loss=0.05613, att_loss=0.2369, loss=0.2007, over 16472.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.007016, over 46.00 utterances.], tot_loss[ctc_loss=0.0643, att_loss=0.2312, loss=0.1978, over 3274303.03 frames. utt_duration=1233 frames, utt_pad_proportion=0.05711, over 10636.73 utterances.], batch size: 46, lr: 3.83e-03, grad_scale: 8.0 2023-03-09 11:11:03,711 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=109281.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:11:23,972 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109294.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:11:52,255 INFO [train2.py:809] (1/4) Epoch 28, batch 1750, loss[ctc_loss=0.05338, att_loss=0.2109, loss=0.1794, over 16401.00 frames. utt_duration=1493 frames, utt_pad_proportion=0.006865, over 44.00 utterances.], tot_loss[ctc_loss=0.06469, att_loss=0.2308, loss=0.1976, over 3276103.96 frames. utt_duration=1244 frames, utt_pad_proportion=0.05427, over 10544.86 utterances.], batch size: 44, lr: 3.83e-03, grad_scale: 8.0 2023-03-09 11:12:02,562 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109318.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:12:02,672 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0648, 4.9042, 5.0251, 1.9557, 1.9908, 2.5746, 2.2395, 3.7684], device='cuda:1'), covar=tensor([0.0796, 0.0475, 0.0275, 0.5054, 0.6050, 0.3385, 0.4413, 0.1832], device='cuda:1'), in_proj_covar=tensor([0.0362, 0.0303, 0.0280, 0.0253, 0.0340, 0.0333, 0.0264, 0.0372], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-09 11:12:47,097 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7099, 5.1018, 4.9565, 5.0609, 5.1283, 4.8272, 3.8210, 5.0777], device='cuda:1'), covar=tensor([0.0128, 0.0102, 0.0129, 0.0083, 0.0115, 0.0111, 0.0564, 0.0243], device='cuda:1'), in_proj_covar=tensor([0.0096, 0.0091, 0.0116, 0.0072, 0.0079, 0.0089, 0.0104, 0.0110], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 11:12:48,676 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6234, 3.2732, 3.7314, 2.9203, 3.5727, 4.7203, 4.5385, 3.3897], device='cuda:1'), covar=tensor([0.0340, 0.1542, 0.1286, 0.1431, 0.1057, 0.0803, 0.0606, 0.1174], device='cuda:1'), in_proj_covar=tensor([0.0248, 0.0250, 0.0290, 0.0220, 0.0271, 0.0380, 0.0274, 0.0236], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-09 11:12:50,892 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-03-09 11:13:00,661 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.203e+02 1.738e+02 2.027e+02 2.395e+02 5.989e+02, threshold=4.055e+02, percent-clipped=1.0 2023-03-09 11:13:01,104 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109355.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:13:07,457 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6442, 3.1278, 3.7318, 3.0087, 3.5901, 4.7139, 4.5500, 3.4830], device='cuda:1'), covar=tensor([0.0370, 0.1716, 0.1325, 0.1419, 0.1128, 0.0998, 0.0634, 0.1093], device='cuda:1'), in_proj_covar=tensor([0.0248, 0.0250, 0.0290, 0.0220, 0.0271, 0.0380, 0.0274, 0.0236], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-09 11:13:11,726 INFO [train2.py:809] (1/4) Epoch 28, batch 1800, loss[ctc_loss=0.1059, att_loss=0.2582, loss=0.2277, over 17307.00 frames. utt_duration=1175 frames, utt_pad_proportion=0.02326, over 59.00 utterances.], tot_loss[ctc_loss=0.0646, att_loss=0.231, loss=0.1977, over 3272554.75 frames. utt_duration=1214 frames, utt_pad_proportion=0.06342, over 10797.05 utterances.], batch size: 59, lr: 3.83e-03, grad_scale: 8.0 2023-03-09 11:13:39,561 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109379.0, num_to_drop=1, layers_to_drop={3} 2023-03-09 11:14:01,821 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109393.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 11:14:31,015 INFO [train2.py:809] (1/4) Epoch 28, batch 1850, loss[ctc_loss=0.05943, att_loss=0.2376, loss=0.202, over 17059.00 frames. utt_duration=1314 frames, utt_pad_proportion=0.008476, over 52.00 utterances.], tot_loss[ctc_loss=0.06473, att_loss=0.2314, loss=0.1981, over 3276001.68 frames. utt_duration=1233 frames, utt_pad_proportion=0.05796, over 10644.47 utterances.], batch size: 52, lr: 3.83e-03, grad_scale: 8.0 2023-03-09 11:15:13,199 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7599, 2.6020, 2.8386, 2.8205, 3.0376, 2.8732, 2.5579, 3.1690], device='cuda:1'), covar=tensor([0.1693, 0.2138, 0.1673, 0.1455, 0.1432, 0.1086, 0.2001, 0.1339], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0149, 0.0144, 0.0139, 0.0155, 0.0133, 0.0154, 0.0133], 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') 2023-03-09 11:15:26,811 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109447.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:15:39,066 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.065e+02 1.724e+02 2.158e+02 2.874e+02 5.021e+02, threshold=4.317e+02, percent-clipped=1.0 2023-03-09 11:15:42,589 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9348, 5.2865, 5.4944, 5.3160, 5.3835, 5.9214, 5.2688, 5.9849], device='cuda:1'), covar=tensor([0.0863, 0.0808, 0.0970, 0.1526, 0.2149, 0.1015, 0.0701, 0.0867], device='cuda:1'), in_proj_covar=tensor([0.0931, 0.0534, 0.0654, 0.0692, 0.0923, 0.0669, 0.0517, 0.0653], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 11:15:50,146 INFO [train2.py:809] (1/4) Epoch 28, batch 1900, loss[ctc_loss=0.05728, att_loss=0.2357, loss=0.2, over 16405.00 frames. utt_duration=1493 frames, utt_pad_proportion=0.007359, over 44.00 utterances.], tot_loss[ctc_loss=0.06517, att_loss=0.232, loss=0.1987, over 3281781.02 frames. utt_duration=1238 frames, utt_pad_proportion=0.05626, over 10617.07 utterances.], batch size: 44, lr: 3.83e-03, grad_scale: 8.0 2023-03-09 11:16:07,931 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-03-09 11:16:12,015 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.4093, 2.8673, 3.3532, 4.4526, 4.0160, 3.9753, 2.8160, 2.4926], device='cuda:1'), covar=tensor([0.0788, 0.1909, 0.0939, 0.0495, 0.0813, 0.0519, 0.1566, 0.1949], device='cuda:1'), in_proj_covar=tensor([0.0189, 0.0218, 0.0187, 0.0224, 0.0234, 0.0192, 0.0206, 0.0192], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 11:16:23,181 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-03-09 11:16:43,558 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=109495.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:17:09,536 INFO [train2.py:809] (1/4) Epoch 28, batch 1950, loss[ctc_loss=0.05931, att_loss=0.2414, loss=0.205, over 16980.00 frames. utt_duration=1360 frames, utt_pad_proportion=0.006706, over 50.00 utterances.], tot_loss[ctc_loss=0.065, att_loss=0.2319, loss=0.1985, over 3286020.85 frames. utt_duration=1262 frames, utt_pad_proportion=0.04893, over 10428.46 utterances.], batch size: 50, lr: 3.83e-03, grad_scale: 8.0 2023-03-09 11:18:01,990 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.32 vs. limit=5.0 2023-03-09 11:18:10,595 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109550.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:18:17,572 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.372e+02 1.926e+02 2.337e+02 2.750e+02 5.694e+02, threshold=4.673e+02, percent-clipped=4.0 2023-03-09 11:18:28,476 INFO [train2.py:809] (1/4) Epoch 28, batch 2000, loss[ctc_loss=0.06244, att_loss=0.2267, loss=0.1938, over 16117.00 frames. utt_duration=1537 frames, utt_pad_proportion=0.006772, over 42.00 utterances.], tot_loss[ctc_loss=0.06551, att_loss=0.2321, loss=0.1987, over 3289617.27 frames. utt_duration=1250 frames, utt_pad_proportion=0.05042, over 10538.27 utterances.], batch size: 42, lr: 3.83e-03, grad_scale: 8.0 2023-03-09 11:18:55,942 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109579.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:18:57,878 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.72 vs. limit=2.0 2023-03-09 11:19:47,009 INFO [train2.py:809] (1/4) Epoch 28, batch 2050, loss[ctc_loss=0.05326, att_loss=0.2302, loss=0.1948, over 16419.00 frames. utt_duration=1494 frames, utt_pad_proportion=0.005731, over 44.00 utterances.], tot_loss[ctc_loss=0.06549, att_loss=0.2321, loss=0.1988, over 3291251.52 frames. utt_duration=1246 frames, utt_pad_proportion=0.05054, over 10580.12 utterances.], batch size: 44, lr: 3.83e-03, grad_scale: 8.0 2023-03-09 11:20:20,231 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-03-09 11:20:33,229 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109640.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:20:48,873 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109650.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:20:56,143 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.260e+02 1.837e+02 2.171e+02 2.517e+02 4.514e+02, threshold=4.342e+02, percent-clipped=0.0 2023-03-09 11:21:02,505 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109659.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:21:06,773 INFO [train2.py:809] (1/4) Epoch 28, batch 2100, loss[ctc_loss=0.05758, att_loss=0.2272, loss=0.1933, over 16315.00 frames. utt_duration=1452 frames, utt_pad_proportion=0.007129, over 45.00 utterances.], tot_loss[ctc_loss=0.06537, att_loss=0.2319, loss=0.1986, over 3285011.85 frames. utt_duration=1249 frames, utt_pad_proportion=0.05163, over 10536.52 utterances.], batch size: 45, lr: 3.83e-03, grad_scale: 8.0 2023-03-09 11:21:07,021 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.3420, 4.7144, 4.9066, 4.7561, 4.8649, 5.2808, 4.7559, 5.3589], device='cuda:1'), covar=tensor([0.0843, 0.0842, 0.0911, 0.1448, 0.1941, 0.0931, 0.1327, 0.0708], device='cuda:1'), in_proj_covar=tensor([0.0916, 0.0528, 0.0646, 0.0684, 0.0917, 0.0664, 0.0511, 0.0643], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 11:21:18,073 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8198, 5.1231, 4.6825, 5.1664, 4.5083, 4.7699, 5.2158, 4.9976], device='cuda:1'), covar=tensor([0.0576, 0.0300, 0.0867, 0.0317, 0.0454, 0.0331, 0.0265, 0.0235], device='cuda:1'), in_proj_covar=tensor([0.0408, 0.0346, 0.0386, 0.0384, 0.0342, 0.0250, 0.0327, 0.0308], device='cuda:1'), out_proj_covar=tensor([0.0007, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0006, 0.0005], device='cuda:1') 2023-03-09 11:21:21,188 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.3812, 5.2808, 5.1019, 3.4937, 5.1518, 5.0117, 4.8691, 3.1360], device='cuda:1'), covar=tensor([0.0105, 0.0119, 0.0252, 0.0791, 0.0093, 0.0158, 0.0224, 0.1192], device='cuda:1'), in_proj_covar=tensor([0.0079, 0.0106, 0.0112, 0.0112, 0.0089, 0.0118, 0.0101, 0.0104], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-09 11:21:26,411 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109674.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 11:21:57,826 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109693.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 11:22:26,688 INFO [train2.py:809] (1/4) Epoch 28, batch 2150, loss[ctc_loss=0.04555, att_loss=0.2, loss=0.1691, over 15643.00 frames. utt_duration=1693 frames, utt_pad_proportion=0.008988, over 37.00 utterances.], tot_loss[ctc_loss=0.06612, att_loss=0.2325, loss=0.1993, over 3287773.94 frames. utt_duration=1237 frames, utt_pad_proportion=0.053, over 10640.43 utterances.], batch size: 37, lr: 3.83e-03, grad_scale: 8.0 2023-03-09 11:22:40,159 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109720.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:23:12,295 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-03-09 11:23:13,175 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=109741.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 11:23:34,196 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.122e+02 1.864e+02 2.379e+02 2.851e+02 5.187e+02, threshold=4.758e+02, percent-clipped=2.0 2023-03-09 11:23:45,113 INFO [train2.py:809] (1/4) Epoch 28, batch 2200, loss[ctc_loss=0.07991, att_loss=0.241, loss=0.2088, over 14217.00 frames. utt_duration=393.7 frames, utt_pad_proportion=0.3153, over 145.00 utterances.], tot_loss[ctc_loss=0.06604, att_loss=0.2323, loss=0.199, over 3280804.45 frames. utt_duration=1217 frames, utt_pad_proportion=0.05948, over 10793.24 utterances.], batch size: 145, lr: 3.82e-03, grad_scale: 8.0 2023-03-09 11:25:03,656 INFO [train2.py:809] (1/4) Epoch 28, batch 2250, loss[ctc_loss=0.06637, att_loss=0.2164, loss=0.1864, over 15507.00 frames. utt_duration=1725 frames, utt_pad_proportion=0.008226, over 36.00 utterances.], tot_loss[ctc_loss=0.06569, att_loss=0.2319, loss=0.1986, over 3272930.72 frames. utt_duration=1236 frames, utt_pad_proportion=0.05816, over 10606.43 utterances.], batch size: 36, lr: 3.82e-03, grad_scale: 8.0 2023-03-09 11:26:03,236 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109850.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:26:10,587 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.320e+02 1.849e+02 2.212e+02 2.625e+02 5.541e+02, threshold=4.423e+02, percent-clipped=2.0 2023-03-09 11:26:21,267 INFO [train2.py:809] (1/4) Epoch 28, batch 2300, loss[ctc_loss=0.07763, att_loss=0.2409, loss=0.2082, over 17456.00 frames. utt_duration=885.4 frames, utt_pad_proportion=0.07188, over 79.00 utterances.], tot_loss[ctc_loss=0.0658, att_loss=0.2314, loss=0.1983, over 3279254.36 frames. utt_duration=1263 frames, utt_pad_proportion=0.04982, over 10399.60 utterances.], batch size: 79, lr: 3.82e-03, grad_scale: 8.0 2023-03-09 11:26:51,216 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5535, 2.7106, 5.0450, 3.9880, 3.2946, 4.3472, 4.8481, 4.6330], device='cuda:1'), covar=tensor([0.0308, 0.1393, 0.0287, 0.0805, 0.1442, 0.0256, 0.0200, 0.0332], device='cuda:1'), in_proj_covar=tensor([0.0231, 0.0248, 0.0226, 0.0325, 0.0270, 0.0242, 0.0217, 0.0241], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 11:27:17,539 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=109898.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:27:38,455 INFO [train2.py:809] (1/4) Epoch 28, batch 2350, loss[ctc_loss=0.06804, att_loss=0.2259, loss=0.1943, over 16536.00 frames. utt_duration=1471 frames, utt_pad_proportion=0.005961, over 45.00 utterances.], tot_loss[ctc_loss=0.06575, att_loss=0.2307, loss=0.1977, over 3264296.20 frames. utt_duration=1241 frames, utt_pad_proportion=0.05927, over 10530.29 utterances.], batch size: 45, lr: 3.82e-03, grad_scale: 8.0 2023-03-09 11:27:47,027 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9394, 5.0370, 4.8920, 2.0962, 2.1190, 3.0085, 2.3690, 3.9056], device='cuda:1'), covar=tensor([0.0749, 0.0337, 0.0306, 0.5527, 0.5132, 0.2228, 0.3952, 0.1497], device='cuda:1'), in_proj_covar=tensor([0.0360, 0.0300, 0.0278, 0.0251, 0.0338, 0.0331, 0.0263, 0.0368], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-03-09 11:27:47,642 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.20 vs. limit=5.0 2023-03-09 11:28:16,094 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109935.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:28:23,953 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0317, 4.3330, 4.1864, 4.6309, 2.6569, 4.3467, 2.7515, 1.8359], device='cuda:1'), covar=tensor([0.0532, 0.0298, 0.0768, 0.0254, 0.1693, 0.0268, 0.1499, 0.1759], device='cuda:1'), in_proj_covar=tensor([0.0217, 0.0188, 0.0265, 0.0181, 0.0222, 0.0170, 0.0233, 0.0203], 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') 2023-03-09 11:28:39,557 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109950.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:28:46,856 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.356e+02 1.839e+02 2.224e+02 2.680e+02 5.661e+02, threshold=4.448e+02, percent-clipped=2.0 2023-03-09 11:28:57,736 INFO [train2.py:809] (1/4) Epoch 28, batch 2400, loss[ctc_loss=0.08558, att_loss=0.2482, loss=0.2157, over 17363.00 frames. utt_duration=1104 frames, utt_pad_proportion=0.03432, over 63.00 utterances.], tot_loss[ctc_loss=0.06541, att_loss=0.2309, loss=0.1978, over 3272558.37 frames. utt_duration=1248 frames, utt_pad_proportion=0.05563, over 10504.71 utterances.], batch size: 63, lr: 3.82e-03, grad_scale: 8.0 2023-03-09 11:29:17,583 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109974.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:29:55,817 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=109998.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:30:22,867 INFO [train2.py:809] (1/4) Epoch 28, batch 2450, loss[ctc_loss=0.05612, att_loss=0.2372, loss=0.201, over 16882.00 frames. utt_duration=1380 frames, utt_pad_proportion=0.006715, over 49.00 utterances.], tot_loss[ctc_loss=0.06477, att_loss=0.2306, loss=0.1974, over 3266448.93 frames. utt_duration=1248 frames, utt_pad_proportion=0.05621, over 10481.10 utterances.], batch size: 49, lr: 3.82e-03, grad_scale: 8.0 2023-03-09 11:30:28,520 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110015.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:30:28,598 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7390, 5.1092, 4.9653, 4.9962, 5.1592, 4.8408, 3.7848, 5.1298], device='cuda:1'), covar=tensor([0.0117, 0.0118, 0.0136, 0.0123, 0.0106, 0.0106, 0.0601, 0.0207], device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0092, 0.0117, 0.0073, 0.0080, 0.0090, 0.0105, 0.0110], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 11:30:31,954 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110017.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:30:39,407 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=110022.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:30:41,099 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7845, 5.1343, 4.9802, 5.0597, 5.2516, 4.9015, 4.0166, 5.2205], device='cuda:1'), covar=tensor([0.0132, 0.0147, 0.0132, 0.0093, 0.0096, 0.0107, 0.0531, 0.0216], device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0093, 0.0118, 0.0073, 0.0080, 0.0090, 0.0105, 0.0110], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 11:31:12,218 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-03-09 11:31:14,784 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110044.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:31:30,266 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.8171, 3.6583, 3.0768, 3.2744, 3.8055, 3.5057, 2.9086, 3.9359], device='cuda:1'), covar=tensor([0.1155, 0.0483, 0.1133, 0.0772, 0.0859, 0.0760, 0.0954, 0.0460], device='cuda:1'), in_proj_covar=tensor([0.0209, 0.0229, 0.0233, 0.0210, 0.0291, 0.0252, 0.0205, 0.0299], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-09 11:31:31,406 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.139e+02 1.872e+02 2.152e+02 2.645e+02 5.223e+02, threshold=4.304e+02, percent-clipped=1.0 2023-03-09 11:31:41,822 INFO [train2.py:809] (1/4) Epoch 28, batch 2500, loss[ctc_loss=0.07015, att_loss=0.2385, loss=0.2048, over 16705.00 frames. utt_duration=1454 frames, utt_pad_proportion=0.005383, over 46.00 utterances.], tot_loss[ctc_loss=0.06498, att_loss=0.2315, loss=0.1982, over 3270476.87 frames. utt_duration=1223 frames, utt_pad_proportion=0.06071, over 10711.35 utterances.], batch size: 46, lr: 3.82e-03, grad_scale: 8.0 2023-03-09 11:32:07,612 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110078.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:32:50,570 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110105.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:32:58,104 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110110.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:33:01,505 INFO [train2.py:809] (1/4) Epoch 28, batch 2550, loss[ctc_loss=0.06223, att_loss=0.236, loss=0.2013, over 17122.00 frames. utt_duration=693.2 frames, utt_pad_proportion=0.127, over 99.00 utterances.], tot_loss[ctc_loss=0.06438, att_loss=0.2318, loss=0.1983, over 3273887.32 frames. utt_duration=1226 frames, utt_pad_proportion=0.05811, over 10690.30 utterances.], batch size: 99, lr: 3.82e-03, grad_scale: 8.0 2023-03-09 11:33:07,174 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110115.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:33:41,641 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.8971, 4.2961, 4.2610, 4.5316, 2.7177, 4.2549, 2.5787, 1.7579], device='cuda:1'), covar=tensor([0.0598, 0.0354, 0.0738, 0.0269, 0.1635, 0.0284, 0.1649, 0.1762], device='cuda:1'), in_proj_covar=tensor([0.0221, 0.0191, 0.0270, 0.0184, 0.0226, 0.0172, 0.0236, 0.0206], 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') 2023-03-09 11:34:08,243 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.183e+02 1.726e+02 2.160e+02 2.628e+02 5.720e+02, threshold=4.319e+02, percent-clipped=3.0 2023-03-09 11:34:19,906 INFO [train2.py:809] (1/4) Epoch 28, batch 2600, loss[ctc_loss=0.0511, att_loss=0.2245, loss=0.1899, over 15946.00 frames. utt_duration=1557 frames, utt_pad_proportion=0.006969, over 41.00 utterances.], tot_loss[ctc_loss=0.06513, att_loss=0.2327, loss=0.1992, over 3275857.83 frames. utt_duration=1214 frames, utt_pad_proportion=0.05973, over 10810.09 utterances.], batch size: 41, lr: 3.82e-03, grad_scale: 8.0 2023-03-09 11:34:34,600 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110171.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:34:42,167 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110176.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:35:38,095 INFO [train2.py:809] (1/4) Epoch 28, batch 2650, loss[ctc_loss=0.04966, att_loss=0.2101, loss=0.178, over 10652.00 frames. utt_duration=1854 frames, utt_pad_proportion=0.2261, over 23.00 utterances.], tot_loss[ctc_loss=0.06503, att_loss=0.2321, loss=0.1987, over 3265488.74 frames. utt_duration=1215 frames, utt_pad_proportion=0.06299, over 10765.13 utterances.], batch size: 23, lr: 3.82e-03, grad_scale: 8.0 2023-03-09 11:36:02,623 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9667, 4.1663, 4.0383, 4.4541, 2.4880, 4.2569, 2.5782, 1.6948], device='cuda:1'), covar=tensor([0.0546, 0.0332, 0.0870, 0.0537, 0.1856, 0.0276, 0.1724, 0.1877], device='cuda:1'), in_proj_covar=tensor([0.0219, 0.0190, 0.0268, 0.0183, 0.0225, 0.0171, 0.0234, 0.0205], 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') 2023-03-09 11:36:16,241 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110235.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:36:46,734 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.140e+02 1.822e+02 2.163e+02 2.650e+02 7.141e+02, threshold=4.325e+02, percent-clipped=4.0 2023-03-09 11:36:47,020 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9555, 4.0543, 4.0670, 4.0615, 4.1603, 4.1602, 3.8709, 3.8539], device='cuda:1'), covar=tensor([0.1026, 0.0753, 0.0793, 0.0610, 0.0346, 0.0400, 0.0489, 0.0343], device='cuda:1'), in_proj_covar=tensor([0.0531, 0.0375, 0.0371, 0.0378, 0.0438, 0.0444, 0.0376, 0.0411], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004], device='cuda:1') 2023-03-09 11:36:57,431 INFO [train2.py:809] (1/4) Epoch 28, batch 2700, loss[ctc_loss=0.06898, att_loss=0.2226, loss=0.1919, over 16180.00 frames. utt_duration=1580 frames, utt_pad_proportion=0.006458, over 41.00 utterances.], tot_loss[ctc_loss=0.06517, att_loss=0.2314, loss=0.1982, over 3258378.80 frames. utt_duration=1175 frames, utt_pad_proportion=0.07514, over 11104.13 utterances.], batch size: 41, lr: 3.82e-03, grad_scale: 8.0 2023-03-09 11:37:32,565 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=110283.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:37:32,848 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110283.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:38:03,457 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110303.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:38:16,816 INFO [train2.py:809] (1/4) Epoch 28, batch 2750, loss[ctc_loss=0.05482, att_loss=0.2176, loss=0.1851, over 15916.00 frames. utt_duration=1634 frames, utt_pad_proportion=0.006968, over 39.00 utterances.], tot_loss[ctc_loss=0.06425, att_loss=0.2301, loss=0.1969, over 3251597.64 frames. utt_duration=1192 frames, utt_pad_proportion=0.07377, over 10924.88 utterances.], batch size: 39, lr: 3.81e-03, grad_scale: 8.0 2023-03-09 11:38:22,356 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110315.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:38:27,634 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110318.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:38:36,813 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0341, 4.3425, 4.2848, 4.5415, 2.5622, 4.2911, 2.6299, 1.8275], device='cuda:1'), covar=tensor([0.0509, 0.0309, 0.0711, 0.0293, 0.1681, 0.0296, 0.1556, 0.1706], device='cuda:1'), in_proj_covar=tensor([0.0219, 0.0190, 0.0268, 0.0182, 0.0224, 0.0171, 0.0233, 0.0204], 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') 2023-03-09 11:39:08,941 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110344.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:39:13,557 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7681, 2.4555, 2.8485, 2.8045, 3.1113, 3.0430, 2.4482, 3.0608], device='cuda:1'), covar=tensor([0.1224, 0.1946, 0.1324, 0.1032, 0.1039, 0.0856, 0.1612, 0.1290], device='cuda:1'), in_proj_covar=tensor([0.0147, 0.0149, 0.0144, 0.0139, 0.0156, 0.0134, 0.0154, 0.0134], 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') 2023-03-09 11:39:25,560 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.237e+02 1.842e+02 2.112e+02 2.497e+02 5.270e+02, threshold=4.224e+02, percent-clipped=1.0 2023-03-09 11:39:34,249 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.45 vs. limit=5.0 2023-03-09 11:39:36,861 INFO [train2.py:809] (1/4) Epoch 28, batch 2800, loss[ctc_loss=0.07758, att_loss=0.2412, loss=0.2084, over 17494.00 frames. utt_duration=876 frames, utt_pad_proportion=0.0759, over 80.00 utterances.], tot_loss[ctc_loss=0.06402, att_loss=0.2301, loss=0.1969, over 3256600.90 frames. utt_duration=1210 frames, utt_pad_proportion=0.06898, over 10780.41 utterances.], batch size: 80, lr: 3.81e-03, grad_scale: 8.0 2023-03-09 11:39:38,493 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=110363.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:39:40,864 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110364.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:39:54,210 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110373.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:40:04,289 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110379.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:40:37,361 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110400.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:40:55,888 INFO [train2.py:809] (1/4) Epoch 28, batch 2850, loss[ctc_loss=0.05323, att_loss=0.2101, loss=0.1787, over 15372.00 frames. utt_duration=1758 frames, utt_pad_proportion=0.01068, over 35.00 utterances.], tot_loss[ctc_loss=0.06445, att_loss=0.2301, loss=0.197, over 3252089.12 frames. utt_duration=1197 frames, utt_pad_proportion=0.07267, over 10879.85 utterances.], batch size: 35, lr: 3.81e-03, grad_scale: 8.0 2023-03-09 11:42:02,948 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.137e+02 1.889e+02 2.295e+02 2.895e+02 8.128e+02, threshold=4.591e+02, percent-clipped=8.0 2023-03-09 11:42:14,321 INFO [train2.py:809] (1/4) Epoch 28, batch 2900, loss[ctc_loss=0.0422, att_loss=0.2226, loss=0.1866, over 16125.00 frames. utt_duration=1537 frames, utt_pad_proportion=0.005652, over 42.00 utterances.], tot_loss[ctc_loss=0.06474, att_loss=0.2305, loss=0.1974, over 3253720.10 frames. utt_duration=1218 frames, utt_pad_proportion=0.0676, over 10694.44 utterances.], batch size: 42, lr: 3.81e-03, grad_scale: 8.0 2023-03-09 11:42:21,198 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110466.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:42:28,781 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110471.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:42:35,019 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7123, 2.6023, 2.7876, 2.7861, 3.0278, 3.0628, 2.6511, 3.0908], device='cuda:1'), covar=tensor([0.1619, 0.2123, 0.1689, 0.1362, 0.1381, 0.0981, 0.1749, 0.1349], device='cuda:1'), in_proj_covar=tensor([0.0145, 0.0146, 0.0142, 0.0137, 0.0154, 0.0132, 0.0153, 0.0132], 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') 2023-03-09 11:43:32,999 INFO [train2.py:809] (1/4) Epoch 28, batch 2950, loss[ctc_loss=0.04832, att_loss=0.2122, loss=0.1794, over 16135.00 frames. utt_duration=1538 frames, utt_pad_proportion=0.005541, over 42.00 utterances.], tot_loss[ctc_loss=0.06478, att_loss=0.2311, loss=0.1979, over 3270314.05 frames. utt_duration=1228 frames, utt_pad_proportion=0.06069, over 10667.09 utterances.], batch size: 42, lr: 3.81e-03, grad_scale: 8.0 2023-03-09 11:44:31,437 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0712, 5.3176, 5.2070, 5.2796, 5.3783, 5.3523, 4.9967, 4.8342], device='cuda:1'), covar=tensor([0.1036, 0.0560, 0.0349, 0.0500, 0.0270, 0.0304, 0.0373, 0.0314], device='cuda:1'), in_proj_covar=tensor([0.0531, 0.0374, 0.0370, 0.0376, 0.0436, 0.0440, 0.0374, 0.0409], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004], device='cuda:1') 2023-03-09 11:44:39,994 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 9.946e+01 1.771e+02 2.102e+02 2.638e+02 6.862e+02, threshold=4.204e+02, percent-clipped=2.0 2023-03-09 11:44:52,456 INFO [train2.py:809] (1/4) Epoch 28, batch 3000, loss[ctc_loss=0.06735, att_loss=0.2367, loss=0.2028, over 17196.00 frames. utt_duration=696.3 frames, utt_pad_proportion=0.1264, over 99.00 utterances.], tot_loss[ctc_loss=0.06486, att_loss=0.2313, loss=0.198, over 3277176.23 frames. utt_duration=1232 frames, utt_pad_proportion=0.05573, over 10654.07 utterances.], batch size: 99, lr: 3.81e-03, grad_scale: 8.0 2023-03-09 11:44:52,456 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-09 11:45:06,797 INFO [train2.py:843] (1/4) Epoch 28, validation: ctc_loss=0.04082, att_loss=0.2346, loss=0.1958, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-09 11:45:06,798 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16129MB 2023-03-09 11:46:25,765 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110611.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:46:26,961 INFO [train2.py:809] (1/4) Epoch 28, batch 3050, loss[ctc_loss=0.07166, att_loss=0.2483, loss=0.213, over 17044.00 frames. utt_duration=1313 frames, utt_pad_proportion=0.009129, over 52.00 utterances.], tot_loss[ctc_loss=0.06467, att_loss=0.2311, loss=0.1978, over 3275549.29 frames. utt_duration=1253 frames, utt_pad_proportion=0.05204, over 10469.21 utterances.], batch size: 52, lr: 3.81e-03, grad_scale: 8.0 2023-03-09 11:47:10,745 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110639.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:47:35,121 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.322e+02 1.799e+02 2.141e+02 2.649e+02 5.384e+02, threshold=4.281e+02, percent-clipped=1.0 2023-03-09 11:47:42,928 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110659.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:47:47,514 INFO [train2.py:809] (1/4) Epoch 28, batch 3100, loss[ctc_loss=0.07099, att_loss=0.2173, loss=0.188, over 14136.00 frames. utt_duration=1826 frames, utt_pad_proportion=0.06424, over 31.00 utterances.], tot_loss[ctc_loss=0.06445, att_loss=0.231, loss=0.1977, over 3278959.52 frames. utt_duration=1265 frames, utt_pad_proportion=0.04851, over 10380.18 utterances.], batch size: 31, lr: 3.81e-03, grad_scale: 8.0 2023-03-09 11:47:57,988 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-03-09 11:48:03,593 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110672.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:48:05,217 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110673.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:48:06,660 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110674.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:48:48,122 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110700.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:49:07,619 INFO [train2.py:809] (1/4) Epoch 28, batch 3150, loss[ctc_loss=0.04896, att_loss=0.203, loss=0.1722, over 15497.00 frames. utt_duration=1723 frames, utt_pad_proportion=0.008961, over 36.00 utterances.], tot_loss[ctc_loss=0.06421, att_loss=0.2308, loss=0.1974, over 3276931.06 frames. utt_duration=1262 frames, utt_pad_proportion=0.05044, over 10395.45 utterances.], batch size: 36, lr: 3.81e-03, grad_scale: 8.0 2023-03-09 11:49:21,659 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=110721.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:49:46,536 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110736.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:50:04,804 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=110748.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:50:15,627 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.358e+02 1.871e+02 2.189e+02 2.637e+02 5.370e+02, threshold=4.378e+02, percent-clipped=2.0 2023-03-09 11:50:27,803 INFO [train2.py:809] (1/4) Epoch 28, batch 3200, loss[ctc_loss=0.04734, att_loss=0.2287, loss=0.1925, over 16877.00 frames. utt_duration=1379 frames, utt_pad_proportion=0.007229, over 49.00 utterances.], tot_loss[ctc_loss=0.06378, att_loss=0.2308, loss=0.1974, over 3283416.54 frames. utt_duration=1271 frames, utt_pad_proportion=0.04702, over 10341.56 utterances.], batch size: 49, lr: 3.81e-03, grad_scale: 8.0 2023-03-09 11:50:34,376 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110766.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:50:42,150 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110771.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:50:45,149 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3517, 5.0309, 5.0594, 5.1651, 4.9936, 5.0861, 4.8548, 4.6196], device='cuda:1'), covar=tensor([0.1792, 0.0825, 0.0417, 0.0527, 0.0691, 0.0516, 0.0479, 0.0413], device='cuda:1'), in_proj_covar=tensor([0.0531, 0.0375, 0.0371, 0.0375, 0.0436, 0.0442, 0.0375, 0.0409], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005, 0.0004, 0.0004], device='cuda:1') 2023-03-09 11:51:01,885 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.20 vs. limit=5.0 2023-03-09 11:51:23,963 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110797.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 11:51:48,162 INFO [train2.py:809] (1/4) Epoch 28, batch 3250, loss[ctc_loss=0.0822, att_loss=0.2472, loss=0.2142, over 17055.00 frames. utt_duration=1288 frames, utt_pad_proportion=0.009644, over 53.00 utterances.], tot_loss[ctc_loss=0.06418, att_loss=0.2311, loss=0.1977, over 3282718.14 frames. utt_duration=1242 frames, utt_pad_proportion=0.05392, over 10582.60 utterances.], batch size: 53, lr: 3.81e-03, grad_scale: 8.0 2023-03-09 11:51:51,345 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=110814.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:51:59,335 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=110819.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:52:34,682 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9796, 5.0914, 4.8936, 2.3498, 1.9210, 3.0415, 2.4844, 3.8813], device='cuda:1'), covar=tensor([0.0772, 0.0362, 0.0326, 0.4812, 0.5873, 0.2423, 0.4030, 0.1647], device='cuda:1'), in_proj_covar=tensor([0.0364, 0.0302, 0.0280, 0.0253, 0.0341, 0.0333, 0.0267, 0.0371], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-09 11:52:55,398 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.243e+02 1.746e+02 2.093e+02 2.617e+02 5.212e+02, threshold=4.186e+02, percent-clipped=3.0 2023-03-09 11:53:07,980 INFO [train2.py:809] (1/4) Epoch 28, batch 3300, loss[ctc_loss=0.05387, att_loss=0.2101, loss=0.1789, over 15486.00 frames. utt_duration=1722 frames, utt_pad_proportion=0.009856, over 36.00 utterances.], tot_loss[ctc_loss=0.06438, att_loss=0.231, loss=0.1977, over 3285755.51 frames. utt_duration=1264 frames, utt_pad_proportion=0.04758, over 10411.09 utterances.], batch size: 36, lr: 3.81e-03, grad_scale: 8.0 2023-03-09 11:53:17,694 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110868.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:54:27,399 INFO [train2.py:809] (1/4) Epoch 28, batch 3350, loss[ctc_loss=0.08353, att_loss=0.2499, loss=0.2166, over 17266.00 frames. utt_duration=1257 frames, utt_pad_proportion=0.0139, over 55.00 utterances.], tot_loss[ctc_loss=0.06508, att_loss=0.2316, loss=0.1983, over 3278314.25 frames. utt_duration=1213 frames, utt_pad_proportion=0.06283, over 10820.48 utterances.], batch size: 55, lr: 3.80e-03, grad_scale: 8.0 2023-03-09 11:54:54,327 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110929.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:55:10,270 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110939.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:55:36,026 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.217e+02 1.730e+02 2.154e+02 2.752e+02 9.084e+02, threshold=4.307e+02, percent-clipped=4.0 2023-03-09 11:55:41,061 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110958.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:55:42,522 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110959.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:55:46,721 INFO [train2.py:809] (1/4) Epoch 28, batch 3400, loss[ctc_loss=0.07582, att_loss=0.2192, loss=0.1905, over 16003.00 frames. utt_duration=1602 frames, utt_pad_proportion=0.007528, over 40.00 utterances.], tot_loss[ctc_loss=0.06494, att_loss=0.2315, loss=0.1982, over 3277860.73 frames. utt_duration=1208 frames, utt_pad_proportion=0.0628, over 10863.63 utterances.], batch size: 40, lr: 3.80e-03, grad_scale: 8.0 2023-03-09 11:55:54,574 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110967.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:56:05,572 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110974.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:56:24,427 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-03-09 11:56:26,761 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=110987.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:56:58,895 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=111007.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:57:06,335 INFO [train2.py:809] (1/4) Epoch 28, batch 3450, loss[ctc_loss=0.05304, att_loss=0.2288, loss=0.1936, over 16539.00 frames. utt_duration=1472 frames, utt_pad_proportion=0.006302, over 45.00 utterances.], tot_loss[ctc_loss=0.0646, att_loss=0.2314, loss=0.198, over 3276057.93 frames. utt_duration=1221 frames, utt_pad_proportion=0.06031, over 10745.54 utterances.], batch size: 45, lr: 3.80e-03, grad_scale: 8.0 2023-03-09 11:57:17,587 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111019.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:57:21,906 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=111022.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 11:58:14,707 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.190e+02 1.799e+02 2.104e+02 2.788e+02 6.212e+02, threshold=4.209e+02, percent-clipped=4.0 2023-03-09 11:58:25,586 INFO [train2.py:809] (1/4) Epoch 28, batch 3500, loss[ctc_loss=0.06714, att_loss=0.2515, loss=0.2146, over 17298.00 frames. utt_duration=1004 frames, utt_pad_proportion=0.05242, over 69.00 utterances.], tot_loss[ctc_loss=0.06502, att_loss=0.232, loss=0.1986, over 3280867.88 frames. utt_duration=1232 frames, utt_pad_proportion=0.05583, over 10663.80 utterances.], batch size: 69, lr: 3.80e-03, grad_scale: 8.0 2023-03-09 11:58:33,858 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9395, 5.2566, 5.1356, 5.2318, 5.3693, 4.9835, 3.8435, 5.2787], device='cuda:1'), covar=tensor([0.0131, 0.0152, 0.0207, 0.0106, 0.0132, 0.0169, 0.0656, 0.0234], device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0093, 0.0118, 0.0073, 0.0080, 0.0091, 0.0106, 0.0111], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 11:59:14,526 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111092.0, num_to_drop=1, layers_to_drop={3} 2023-03-09 11:59:46,739 INFO [train2.py:809] (1/4) Epoch 28, batch 3550, loss[ctc_loss=0.06507, att_loss=0.2093, loss=0.1805, over 15492.00 frames. utt_duration=1723 frames, utt_pad_proportion=0.008124, over 36.00 utterances.], tot_loss[ctc_loss=0.06396, att_loss=0.2314, loss=0.1979, over 3283426.06 frames. utt_duration=1238 frames, utt_pad_proportion=0.05367, over 10625.60 utterances.], batch size: 36, lr: 3.80e-03, grad_scale: 8.0 2023-03-09 12:00:55,331 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.124e+02 1.811e+02 2.154e+02 2.613e+02 6.278e+02, threshold=4.308e+02, percent-clipped=2.0 2023-03-09 12:01:05,922 INFO [train2.py:809] (1/4) Epoch 28, batch 3600, loss[ctc_loss=0.06705, att_loss=0.2179, loss=0.1877, over 15958.00 frames. utt_duration=1558 frames, utt_pad_proportion=0.00616, over 41.00 utterances.], tot_loss[ctc_loss=0.0646, att_loss=0.2314, loss=0.198, over 3267663.60 frames. utt_duration=1185 frames, utt_pad_proportion=0.07162, over 11040.47 utterances.], batch size: 41, lr: 3.80e-03, grad_scale: 8.0 2023-03-09 12:01:11,512 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9665, 5.2636, 4.7902, 5.3505, 4.7343, 4.9551, 5.3796, 5.1490], device='cuda:1'), covar=tensor([0.0640, 0.0300, 0.0833, 0.0335, 0.0422, 0.0292, 0.0277, 0.0230], device='cuda:1'), in_proj_covar=tensor([0.0403, 0.0339, 0.0378, 0.0380, 0.0338, 0.0245, 0.0319, 0.0301], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-09 12:01:22,751 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-03-09 12:02:25,883 INFO [train2.py:809] (1/4) Epoch 28, batch 3650, loss[ctc_loss=0.0673, att_loss=0.234, loss=0.2007, over 16684.00 frames. utt_duration=1452 frames, utt_pad_proportion=0.005952, over 46.00 utterances.], tot_loss[ctc_loss=0.06504, att_loss=0.2319, loss=0.1985, over 3275517.43 frames. utt_duration=1196 frames, utt_pad_proportion=0.06774, over 10970.32 utterances.], batch size: 46, lr: 3.80e-03, grad_scale: 8.0 2023-03-09 12:02:45,387 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111224.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:02:47,133 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111225.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:03:34,857 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.213e+02 1.792e+02 2.128e+02 2.694e+02 5.056e+02, threshold=4.257e+02, percent-clipped=2.0 2023-03-09 12:03:45,875 INFO [train2.py:809] (1/4) Epoch 28, batch 3700, loss[ctc_loss=0.0701, att_loss=0.228, loss=0.1964, over 15885.00 frames. utt_duration=1631 frames, utt_pad_proportion=0.009378, over 39.00 utterances.], tot_loss[ctc_loss=0.06565, att_loss=0.2324, loss=0.199, over 3281729.36 frames. utt_duration=1212 frames, utt_pad_proportion=0.06211, over 10843.50 utterances.], batch size: 39, lr: 3.80e-03, grad_scale: 16.0 2023-03-09 12:03:54,499 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111267.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:04:13,825 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5025, 2.7521, 4.9497, 3.8386, 3.1264, 4.2646, 4.7858, 4.6312], device='cuda:1'), covar=tensor([0.0326, 0.1437, 0.0261, 0.1133, 0.1668, 0.0279, 0.0247, 0.0334], device='cuda:1'), in_proj_covar=tensor([0.0236, 0.0252, 0.0231, 0.0330, 0.0275, 0.0246, 0.0222, 0.0246], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 12:04:24,256 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111286.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:04:50,393 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2546, 5.2287, 5.1713, 2.4423, 2.2065, 2.9768, 2.5254, 3.9902], device='cuda:1'), covar=tensor([0.0642, 0.0349, 0.0190, 0.4640, 0.5275, 0.2372, 0.3683, 0.1475], device='cuda:1'), in_proj_covar=tensor([0.0366, 0.0304, 0.0281, 0.0254, 0.0342, 0.0334, 0.0267, 0.0372], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-09 12:05:06,060 INFO [train2.py:809] (1/4) Epoch 28, batch 3750, loss[ctc_loss=0.09724, att_loss=0.256, loss=0.2243, over 16471.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.006448, over 46.00 utterances.], tot_loss[ctc_loss=0.06611, att_loss=0.2329, loss=0.1995, over 3278143.67 frames. utt_duration=1192 frames, utt_pad_proportion=0.06836, over 11012.28 utterances.], batch size: 46, lr: 3.80e-03, grad_scale: 16.0 2023-03-09 12:05:09,377 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111314.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:05:10,248 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.88 vs. limit=5.0 2023-03-09 12:05:10,824 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=111315.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:06:13,694 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.228e+02 1.865e+02 2.265e+02 2.758e+02 6.117e+02, threshold=4.530e+02, percent-clipped=1.0 2023-03-09 12:06:25,560 INFO [train2.py:809] (1/4) Epoch 28, batch 3800, loss[ctc_loss=0.05351, att_loss=0.2373, loss=0.2006, over 16905.00 frames. utt_duration=684.6 frames, utt_pad_proportion=0.1355, over 99.00 utterances.], tot_loss[ctc_loss=0.06591, att_loss=0.2326, loss=0.1992, over 3278988.15 frames. utt_duration=1196 frames, utt_pad_proportion=0.06647, over 10979.04 utterances.], batch size: 99, lr: 3.80e-03, grad_scale: 16.0 2023-03-09 12:07:13,895 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111392.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 12:07:44,853 INFO [train2.py:809] (1/4) Epoch 28, batch 3850, loss[ctc_loss=0.04623, att_loss=0.2013, loss=0.1703, over 15642.00 frames. utt_duration=1693 frames, utt_pad_proportion=0.008392, over 37.00 utterances.], tot_loss[ctc_loss=0.06508, att_loss=0.2313, loss=0.1981, over 3264075.71 frames. utt_duration=1225 frames, utt_pad_proportion=0.06064, over 10669.52 utterances.], batch size: 37, lr: 3.80e-03, grad_scale: 16.0 2023-03-09 12:08:28,210 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=111440.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:08:45,310 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-03-09 12:08:51,740 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.233e+02 1.814e+02 2.141e+02 2.602e+02 6.153e+02, threshold=4.282e+02, percent-clipped=2.0 2023-03-09 12:09:02,534 INFO [train2.py:809] (1/4) Epoch 28, batch 3900, loss[ctc_loss=0.06765, att_loss=0.2331, loss=0.2, over 16479.00 frames. utt_duration=1435 frames, utt_pad_proportion=0.006489, over 46.00 utterances.], tot_loss[ctc_loss=0.06491, att_loss=0.2317, loss=0.1983, over 3278707.45 frames. utt_duration=1242 frames, utt_pad_proportion=0.05296, over 10569.76 utterances.], batch size: 46, lr: 3.80e-03, grad_scale: 16.0 2023-03-09 12:09:12,019 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111468.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:09:44,236 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1992, 5.4858, 5.0643, 5.5475, 4.8801, 5.1174, 5.6030, 5.3628], device='cuda:1'), covar=tensor([0.0597, 0.0303, 0.0743, 0.0376, 0.0411, 0.0232, 0.0280, 0.0212], device='cuda:1'), in_proj_covar=tensor([0.0405, 0.0341, 0.0381, 0.0383, 0.0340, 0.0246, 0.0322, 0.0303], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-09 12:10:19,199 INFO [train2.py:809] (1/4) Epoch 28, batch 3950, loss[ctc_loss=0.06082, att_loss=0.2318, loss=0.1976, over 17144.00 frames. utt_duration=694.1 frames, utt_pad_proportion=0.1302, over 99.00 utterances.], tot_loss[ctc_loss=0.06568, att_loss=0.2323, loss=0.199, over 3282691.40 frames. utt_duration=1227 frames, utt_pad_proportion=0.05561, over 10716.06 utterances.], batch size: 99, lr: 3.79e-03, grad_scale: 16.0 2023-03-09 12:10:37,694 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111524.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:10:40,781 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0616, 4.3577, 4.1931, 4.5101, 2.7858, 4.3984, 2.6423, 1.6810], device='cuda:1'), covar=tensor([0.0526, 0.0332, 0.0816, 0.0271, 0.1573, 0.0267, 0.1564, 0.1715], device='cuda:1'), in_proj_covar=tensor([0.0226, 0.0195, 0.0276, 0.0187, 0.0231, 0.0176, 0.0239, 0.0210], 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') 2023-03-09 12:10:45,240 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111529.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:11:24,275 INFO [train2.py:809] (1/4) Epoch 29, batch 0, loss[ctc_loss=0.06338, att_loss=0.2204, loss=0.189, over 16020.00 frames. utt_duration=1603 frames, utt_pad_proportion=0.007136, over 40.00 utterances.], tot_loss[ctc_loss=0.06338, att_loss=0.2204, loss=0.189, over 16020.00 frames. utt_duration=1603 frames, utt_pad_proportion=0.007136, over 40.00 utterances.], batch size: 40, lr: 3.73e-03, grad_scale: 8.0 2023-03-09 12:11:24,276 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-09 12:11:36,590 INFO [train2.py:843] (1/4) Epoch 29, validation: ctc_loss=0.04125, att_loss=0.2346, loss=0.1959, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-09 12:11:36,591 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16129MB 2023-03-09 12:11:54,773 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.196e+02 1.789e+02 2.164e+02 2.545e+02 5.866e+02, threshold=4.328e+02, percent-clipped=2.0 2023-03-09 12:12:19,804 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=111572.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:12:34,352 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111581.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:12:44,467 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-03-09 12:12:55,763 INFO [train2.py:809] (1/4) Epoch 29, batch 50, loss[ctc_loss=0.06759, att_loss=0.2359, loss=0.2022, over 17322.00 frames. utt_duration=878.3 frames, utt_pad_proportion=0.0803, over 79.00 utterances.], tot_loss[ctc_loss=0.06378, att_loss=0.2316, loss=0.1981, over 738605.37 frames. utt_duration=1213 frames, utt_pad_proportion=0.0598, over 2439.48 utterances.], batch size: 79, lr: 3.73e-03, grad_scale: 8.0 2023-03-09 12:13:26,139 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.50 vs. limit=5.0 2023-03-09 12:13:26,782 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111614.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:14:02,216 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0588, 4.2563, 4.3211, 4.5217, 2.6018, 4.3665, 2.7974, 1.7973], device='cuda:1'), covar=tensor([0.0531, 0.0352, 0.0715, 0.0258, 0.1670, 0.0291, 0.1436, 0.1651], device='cuda:1'), in_proj_covar=tensor([0.0222, 0.0192, 0.0270, 0.0183, 0.0227, 0.0173, 0.0234, 0.0206], 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') 2023-03-09 12:14:15,536 INFO [train2.py:809] (1/4) Epoch 29, batch 100, loss[ctc_loss=0.06437, att_loss=0.2282, loss=0.1954, over 16874.00 frames. utt_duration=1379 frames, utt_pad_proportion=0.007141, over 49.00 utterances.], tot_loss[ctc_loss=0.06425, att_loss=0.2328, loss=0.1991, over 1306700.02 frames. utt_duration=1191 frames, utt_pad_proportion=0.06105, over 4392.51 utterances.], batch size: 49, lr: 3.72e-03, grad_scale: 8.0 2023-03-09 12:14:33,660 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.291e+02 1.812e+02 2.121e+02 2.575e+02 9.495e+02, threshold=4.242e+02, percent-clipped=2.0 2023-03-09 12:14:40,262 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0786, 5.1569, 5.0230, 2.4040, 2.0257, 2.8188, 2.3082, 4.0597], device='cuda:1'), covar=tensor([0.0745, 0.0306, 0.0209, 0.4610, 0.5773, 0.2825, 0.4137, 0.1393], device='cuda:1'), in_proj_covar=tensor([0.0366, 0.0304, 0.0281, 0.0253, 0.0342, 0.0335, 0.0268, 0.0373], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-09 12:14:42,934 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=111662.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:15:35,272 INFO [train2.py:809] (1/4) Epoch 29, batch 150, loss[ctc_loss=0.08389, att_loss=0.2518, loss=0.2182, over 17220.00 frames. utt_duration=873.5 frames, utt_pad_proportion=0.08632, over 79.00 utterances.], tot_loss[ctc_loss=0.06372, att_loss=0.2312, loss=0.1977, over 1744162.62 frames. utt_duration=1238 frames, utt_pad_proportion=0.05289, over 5643.87 utterances.], batch size: 79, lr: 3.72e-03, grad_scale: 8.0 2023-03-09 12:16:20,885 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0236, 4.0103, 3.8451, 2.7601, 3.8030, 3.8610, 3.5223, 2.7360], device='cuda:1'), covar=tensor([0.0141, 0.0162, 0.0260, 0.0934, 0.0150, 0.0377, 0.0355, 0.1235], device='cuda:1'), in_proj_covar=tensor([0.0079, 0.0107, 0.0113, 0.0113, 0.0090, 0.0119, 0.0101, 0.0105], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-09 12:16:25,621 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.7494, 4.9711, 4.9022, 4.9385, 4.9992, 5.0020, 4.6390, 4.4837], device='cuda:1'), covar=tensor([0.1049, 0.0538, 0.0401, 0.0521, 0.0321, 0.0388, 0.0484, 0.0372], device='cuda:1'), in_proj_covar=tensor([0.0539, 0.0379, 0.0377, 0.0380, 0.0446, 0.0448, 0.0380, 0.0417], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:1') 2023-03-09 12:16:58,001 INFO [train2.py:809] (1/4) Epoch 29, batch 200, loss[ctc_loss=0.05325, att_loss=0.1967, loss=0.168, over 14906.00 frames. utt_duration=1809 frames, utt_pad_proportion=0.02925, over 33.00 utterances.], tot_loss[ctc_loss=0.0644, att_loss=0.2316, loss=0.1981, over 2085833.48 frames. utt_duration=1199 frames, utt_pad_proportion=0.06094, over 6964.43 utterances.], batch size: 33, lr: 3.72e-03, grad_scale: 8.0 2023-03-09 12:17:15,320 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-03-09 12:17:15,891 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.262e+02 1.846e+02 2.300e+02 2.696e+02 5.049e+02, threshold=4.600e+02, percent-clipped=1.0 2023-03-09 12:17:45,424 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0919, 5.3969, 5.6208, 5.4200, 5.6060, 6.0527, 5.3350, 6.1395], device='cuda:1'), covar=tensor([0.0713, 0.0754, 0.0871, 0.1392, 0.1786, 0.0909, 0.0700, 0.0639], device='cuda:1'), in_proj_covar=tensor([0.0927, 0.0538, 0.0656, 0.0694, 0.0926, 0.0672, 0.0516, 0.0656], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 12:18:01,249 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4555, 4.4318, 4.2204, 2.8263, 4.3129, 4.2126, 3.8183, 2.6494], device='cuda:1'), covar=tensor([0.0130, 0.0132, 0.0300, 0.0981, 0.0115, 0.0275, 0.0348, 0.1298], device='cuda:1'), in_proj_covar=tensor([0.0079, 0.0107, 0.0112, 0.0112, 0.0090, 0.0118, 0.0101, 0.0105], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-09 12:18:12,041 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111791.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 12:18:17,857 INFO [train2.py:809] (1/4) Epoch 29, batch 250, loss[ctc_loss=0.06833, att_loss=0.2107, loss=0.1823, over 15503.00 frames. utt_duration=1724 frames, utt_pad_proportion=0.008071, over 36.00 utterances.], tot_loss[ctc_loss=0.06472, att_loss=0.2307, loss=0.1975, over 2342849.28 frames. utt_duration=1217 frames, utt_pad_proportion=0.06138, over 7711.40 utterances.], batch size: 36, lr: 3.72e-03, grad_scale: 8.0 2023-03-09 12:19:05,336 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111824.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:19:37,439 INFO [train2.py:809] (1/4) Epoch 29, batch 300, loss[ctc_loss=0.07772, att_loss=0.2501, loss=0.2156, over 17082.00 frames. utt_duration=1315 frames, utt_pad_proportion=0.007213, over 52.00 utterances.], tot_loss[ctc_loss=0.06508, att_loss=0.2312, loss=0.198, over 2552614.83 frames. utt_duration=1239 frames, utt_pad_proportion=0.05503, over 8247.58 utterances.], batch size: 52, lr: 3.72e-03, grad_scale: 8.0 2023-03-09 12:19:49,919 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111852.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 12:19:55,511 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.235e+02 1.741e+02 2.135e+02 2.604e+02 1.132e+03, threshold=4.270e+02, percent-clipped=2.0 2023-03-09 12:20:35,435 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111881.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:20:56,611 INFO [train2.py:809] (1/4) Epoch 29, batch 350, loss[ctc_loss=0.06727, att_loss=0.2368, loss=0.2029, over 17439.00 frames. utt_duration=1109 frames, utt_pad_proportion=0.03183, over 63.00 utterances.], tot_loss[ctc_loss=0.06403, att_loss=0.2301, loss=0.1969, over 2709524.27 frames. utt_duration=1281 frames, utt_pad_proportion=0.04845, over 8471.33 utterances.], batch size: 63, lr: 3.72e-03, grad_scale: 8.0 2023-03-09 12:21:50,815 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=111929.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:22:16,052 INFO [train2.py:809] (1/4) Epoch 29, batch 400, loss[ctc_loss=0.04678, att_loss=0.2039, loss=0.1724, over 14579.00 frames. utt_duration=1824 frames, utt_pad_proportion=0.03411, over 32.00 utterances.], tot_loss[ctc_loss=0.0643, att_loss=0.2298, loss=0.1967, over 2833633.94 frames. utt_duration=1288 frames, utt_pad_proportion=0.04685, over 8809.23 utterances.], batch size: 32, lr: 3.72e-03, grad_scale: 8.0 2023-03-09 12:22:17,646 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0488, 6.3158, 5.7432, 6.0230, 5.9607, 5.5163, 5.7576, 5.4127], device='cuda:1'), covar=tensor([0.1228, 0.0726, 0.1005, 0.0775, 0.0773, 0.1500, 0.1870, 0.2118], device='cuda:1'), in_proj_covar=tensor([0.0563, 0.0639, 0.0490, 0.0475, 0.0454, 0.0483, 0.0641, 0.0547], 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') 2023-03-09 12:22:31,915 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.2828, 5.1922, 5.0570, 3.2963, 5.0490, 4.9100, 4.6170, 2.9108], device='cuda:1'), covar=tensor([0.0099, 0.0101, 0.0223, 0.0815, 0.0088, 0.0183, 0.0254, 0.1230], device='cuda:1'), in_proj_covar=tensor([0.0079, 0.0106, 0.0112, 0.0112, 0.0089, 0.0118, 0.0100, 0.0104], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-09 12:22:34,680 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.149e+02 1.870e+02 2.171e+02 2.633e+02 6.586e+02, threshold=4.343e+02, percent-clipped=4.0 2023-03-09 12:23:38,447 INFO [train2.py:809] (1/4) Epoch 29, batch 450, loss[ctc_loss=0.0934, att_loss=0.2487, loss=0.2176, over 14189.00 frames. utt_duration=390.2 frames, utt_pad_proportion=0.3178, over 146.00 utterances.], tot_loss[ctc_loss=0.06432, att_loss=0.2294, loss=0.1964, over 2926596.50 frames. utt_duration=1270 frames, utt_pad_proportion=0.05113, over 9231.09 utterances.], batch size: 146, lr: 3.72e-03, grad_scale: 4.0 2023-03-09 12:23:43,568 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7731, 2.5298, 2.4641, 2.6300, 2.8388, 2.8159, 2.4860, 3.0191], device='cuda:1'), covar=tensor([0.1484, 0.2156, 0.1917, 0.1326, 0.1661, 0.1114, 0.1955, 0.1248], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0149, 0.0146, 0.0140, 0.0158, 0.0135, 0.0158, 0.0135], 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') 2023-03-09 12:25:02,538 INFO [train2.py:809] (1/4) Epoch 29, batch 500, loss[ctc_loss=0.06422, att_loss=0.235, loss=0.2008, over 16545.00 frames. utt_duration=1472 frames, utt_pad_proportion=0.006032, over 45.00 utterances.], tot_loss[ctc_loss=0.06398, att_loss=0.229, loss=0.196, over 3005997.23 frames. utt_duration=1290 frames, utt_pad_proportion=0.04516, over 9335.02 utterances.], batch size: 45, lr: 3.72e-03, grad_scale: 4.0 2023-03-09 12:25:22,139 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.395e+02 1.941e+02 2.244e+02 2.736e+02 5.323e+02, threshold=4.488e+02, percent-clipped=4.0 2023-03-09 12:25:37,184 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112066.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:26:06,145 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-03-09 12:26:22,022 INFO [train2.py:809] (1/4) Epoch 29, batch 550, loss[ctc_loss=0.05478, att_loss=0.2359, loss=0.1996, over 16887.00 frames. utt_duration=1380 frames, utt_pad_proportion=0.007077, over 49.00 utterances.], tot_loss[ctc_loss=0.06415, att_loss=0.2296, loss=0.1965, over 3062761.60 frames. utt_duration=1251 frames, utt_pad_proportion=0.05526, over 9802.63 utterances.], batch size: 49, lr: 3.72e-03, grad_scale: 4.0 2023-03-09 12:27:08,918 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.29 vs. limit=5.0 2023-03-09 12:27:10,027 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=112124.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:27:14,822 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112127.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:27:42,787 INFO [train2.py:809] (1/4) Epoch 29, batch 600, loss[ctc_loss=0.06217, att_loss=0.2219, loss=0.1899, over 16172.00 frames. utt_duration=1579 frames, utt_pad_proportion=0.006796, over 41.00 utterances.], tot_loss[ctc_loss=0.06452, att_loss=0.23, loss=0.1969, over 3114854.32 frames. utt_duration=1227 frames, utt_pad_proportion=0.05797, over 10168.05 utterances.], batch size: 41, lr: 3.72e-03, grad_scale: 4.0 2023-03-09 12:27:45,922 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112147.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 12:28:01,253 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.227e+02 2.016e+02 2.437e+02 2.973e+02 9.680e+02, threshold=4.873e+02, percent-clipped=6.0 2023-03-09 12:28:26,109 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=112172.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:29:03,068 INFO [train2.py:809] (1/4) Epoch 29, batch 650, loss[ctc_loss=0.05137, att_loss=0.1929, loss=0.1646, over 15362.00 frames. utt_duration=1757 frames, utt_pad_proportion=0.009964, over 35.00 utterances.], tot_loss[ctc_loss=0.06439, att_loss=0.2306, loss=0.1973, over 3154348.27 frames. utt_duration=1223 frames, utt_pad_proportion=0.05803, over 10325.87 utterances.], batch size: 35, lr: 3.72e-03, grad_scale: 4.0 2023-03-09 12:29:47,393 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-03-09 12:30:18,647 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.95 vs. limit=5.0 2023-03-09 12:30:22,503 INFO [train2.py:809] (1/4) Epoch 29, batch 700, loss[ctc_loss=0.07816, att_loss=0.2486, loss=0.2145, over 17616.00 frames. utt_duration=893.7 frames, utt_pad_proportion=0.06423, over 79.00 utterances.], tot_loss[ctc_loss=0.06534, att_loss=0.2312, loss=0.198, over 3168578.09 frames. utt_duration=1177 frames, utt_pad_proportion=0.07261, over 10784.99 utterances.], batch size: 79, lr: 3.71e-03, grad_scale: 4.0 2023-03-09 12:30:41,003 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.343e+02 1.769e+02 2.156e+02 2.659e+02 7.287e+02, threshold=4.312e+02, percent-clipped=5.0 2023-03-09 12:31:34,934 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-03-09 12:31:41,969 INFO [train2.py:809] (1/4) Epoch 29, batch 750, loss[ctc_loss=0.06603, att_loss=0.2332, loss=0.1997, over 16479.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.006086, over 46.00 utterances.], tot_loss[ctc_loss=0.06556, att_loss=0.2316, loss=0.1984, over 3187115.28 frames. utt_duration=1179 frames, utt_pad_proportion=0.07184, over 10827.50 utterances.], batch size: 46, lr: 3.71e-03, grad_scale: 4.0 2023-03-09 12:32:23,409 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=2.76 vs. limit=5.0 2023-03-09 12:33:02,101 INFO [train2.py:809] (1/4) Epoch 29, batch 800, loss[ctc_loss=0.05754, att_loss=0.2242, loss=0.1908, over 16392.00 frames. utt_duration=1492 frames, utt_pad_proportion=0.007425, over 44.00 utterances.], tot_loss[ctc_loss=0.06608, att_loss=0.2314, loss=0.1983, over 3199447.24 frames. utt_duration=1179 frames, utt_pad_proportion=0.07442, over 10865.67 utterances.], batch size: 44, lr: 3.71e-03, grad_scale: 8.0 2023-03-09 12:33:21,112 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.436e+02 1.905e+02 2.158e+02 2.747e+02 4.535e+02, threshold=4.315e+02, percent-clipped=2.0 2023-03-09 12:34:13,332 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112389.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:34:22,116 INFO [train2.py:809] (1/4) Epoch 29, batch 850, loss[ctc_loss=0.06834, att_loss=0.2266, loss=0.195, over 16409.00 frames. utt_duration=1493 frames, utt_pad_proportion=0.007132, over 44.00 utterances.], tot_loss[ctc_loss=0.06562, att_loss=0.2309, loss=0.1978, over 3207766.44 frames. utt_duration=1198 frames, utt_pad_proportion=0.071, over 10722.05 utterances.], batch size: 44, lr: 3.71e-03, grad_scale: 8.0 2023-03-09 12:35:06,141 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112422.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:35:29,382 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7050, 2.4449, 2.5580, 2.5654, 3.0035, 2.8932, 2.5239, 3.2016], device='cuda:1'), covar=tensor([0.1551, 0.2072, 0.1529, 0.1227, 0.1458, 0.0940, 0.1747, 0.0967], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0149, 0.0147, 0.0140, 0.0158, 0.0136, 0.0158, 0.0136], 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') 2023-03-09 12:35:42,815 INFO [train2.py:809] (1/4) Epoch 29, batch 900, loss[ctc_loss=0.06403, att_loss=0.2052, loss=0.177, over 15655.00 frames. utt_duration=1694 frames, utt_pad_proportion=0.007552, over 37.00 utterances.], tot_loss[ctc_loss=0.06548, att_loss=0.2309, loss=0.1978, over 3217007.14 frames. utt_duration=1204 frames, utt_pad_proportion=0.0691, over 10700.58 utterances.], batch size: 37, lr: 3.71e-03, grad_scale: 8.0 2023-03-09 12:35:46,220 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=112447.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 12:35:51,276 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112450.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:36:02,471 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.247e+02 1.980e+02 2.255e+02 2.971e+02 7.744e+02, threshold=4.510e+02, percent-clipped=6.0 2023-03-09 12:36:16,732 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-03-09 12:36:34,596 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.9860, 3.6750, 3.6988, 3.2108, 3.7049, 3.7544, 3.7557, 2.7738], device='cuda:1'), covar=tensor([0.1011, 0.1045, 0.1583, 0.2580, 0.0849, 0.4152, 0.1006, 0.3004], device='cuda:1'), in_proj_covar=tensor([0.0207, 0.0210, 0.0226, 0.0276, 0.0187, 0.0287, 0.0210, 0.0232], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-03-09 12:37:03,515 INFO [train2.py:809] (1/4) Epoch 29, batch 950, loss[ctc_loss=0.07164, att_loss=0.251, loss=0.2151, over 16768.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.005701, over 48.00 utterances.], tot_loss[ctc_loss=0.06506, att_loss=0.2312, loss=0.198, over 3235084.64 frames. utt_duration=1227 frames, utt_pad_proportion=0.06205, over 10554.96 utterances.], batch size: 48, lr: 3.71e-03, grad_scale: 8.0 2023-03-09 12:37:03,633 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=112495.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 12:38:07,066 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2729, 3.8166, 3.2957, 3.4583, 4.0785, 3.6542, 3.1842, 4.3048], device='cuda:1'), covar=tensor([0.0993, 0.0598, 0.1043, 0.0795, 0.0729, 0.0758, 0.0875, 0.0421], device='cuda:1'), in_proj_covar=tensor([0.0210, 0.0232, 0.0233, 0.0212, 0.0293, 0.0253, 0.0208, 0.0303], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:1') 2023-03-09 12:38:23,730 INFO [train2.py:809] (1/4) Epoch 29, batch 1000, loss[ctc_loss=0.06468, att_loss=0.2269, loss=0.1944, over 16277.00 frames. utt_duration=1515 frames, utt_pad_proportion=0.00701, over 43.00 utterances.], tot_loss[ctc_loss=0.06513, att_loss=0.2315, loss=0.1982, over 3243897.88 frames. utt_duration=1222 frames, utt_pad_proportion=0.06231, over 10635.24 utterances.], batch size: 43, lr: 3.71e-03, grad_scale: 8.0 2023-03-09 12:38:30,192 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.8322, 3.8221, 3.2150, 3.1946, 3.9352, 3.6119, 2.6124, 4.1410], device='cuda:1'), covar=tensor([0.1414, 0.0568, 0.1137, 0.0977, 0.0840, 0.0807, 0.1274, 0.0547], device='cuda:1'), in_proj_covar=tensor([0.0210, 0.0232, 0.0233, 0.0212, 0.0293, 0.0252, 0.0208, 0.0303], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:1') 2023-03-09 12:38:42,361 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.260e+02 1.873e+02 2.288e+02 2.757e+02 5.242e+02, threshold=4.577e+02, percent-clipped=3.0 2023-03-09 12:38:51,132 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112562.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:39:14,933 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112577.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:39:35,933 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.39 vs. limit=5.0 2023-03-09 12:39:42,799 INFO [train2.py:809] (1/4) Epoch 29, batch 1050, loss[ctc_loss=0.08973, att_loss=0.2455, loss=0.2144, over 16349.00 frames. utt_duration=1454 frames, utt_pad_proportion=0.005138, over 45.00 utterances.], tot_loss[ctc_loss=0.06475, att_loss=0.231, loss=0.1978, over 3253681.05 frames. utt_duration=1238 frames, utt_pad_proportion=0.05594, over 10522.74 utterances.], batch size: 45, lr: 3.71e-03, grad_scale: 8.0 2023-03-09 12:40:19,254 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112617.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:40:19,676 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.45 vs. limit=5.0 2023-03-09 12:40:28,661 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112623.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:40:52,264 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112638.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 12:41:03,006 INFO [train2.py:809] (1/4) Epoch 29, batch 1100, loss[ctc_loss=0.04615, att_loss=0.1966, loss=0.1665, over 15486.00 frames. utt_duration=1722 frames, utt_pad_proportion=0.008049, over 36.00 utterances.], tot_loss[ctc_loss=0.06468, att_loss=0.2302, loss=0.1971, over 3249586.11 frames. utt_duration=1233 frames, utt_pad_proportion=0.05951, over 10552.26 utterances.], batch size: 36, lr: 3.71e-03, grad_scale: 8.0 2023-03-09 12:41:22,207 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.269e+02 1.863e+02 2.140e+02 2.769e+02 5.111e+02, threshold=4.279e+02, percent-clipped=1.0 2023-03-09 12:41:43,837 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-03-09 12:41:55,462 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112678.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 12:42:17,483 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=2.86 vs. limit=5.0 2023-03-09 12:42:21,112 INFO [train2.py:809] (1/4) Epoch 29, batch 1150, loss[ctc_loss=0.04528, att_loss=0.204, loss=0.1723, over 15869.00 frames. utt_duration=1629 frames, utt_pad_proportion=0.008459, over 39.00 utterances.], tot_loss[ctc_loss=0.0649, att_loss=0.2303, loss=0.1973, over 3249913.16 frames. utt_duration=1228 frames, utt_pad_proportion=0.06291, over 10600.46 utterances.], batch size: 39, lr: 3.71e-03, grad_scale: 8.0 2023-03-09 12:43:04,794 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=112722.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:43:40,475 INFO [train2.py:809] (1/4) Epoch 29, batch 1200, loss[ctc_loss=0.06127, att_loss=0.2287, loss=0.1952, over 16278.00 frames. utt_duration=1516 frames, utt_pad_proportion=0.006797, over 43.00 utterances.], tot_loss[ctc_loss=0.06459, att_loss=0.2304, loss=0.1973, over 3259800.19 frames. utt_duration=1247 frames, utt_pad_proportion=0.05681, over 10467.65 utterances.], batch size: 43, lr: 3.71e-03, grad_scale: 8.0 2023-03-09 12:43:40,700 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112745.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:43:59,909 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.017e+02 1.786e+02 2.116e+02 2.747e+02 6.920e+02, threshold=4.232e+02, percent-clipped=7.0 2023-03-09 12:44:21,236 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=112770.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:45:00,878 INFO [train2.py:809] (1/4) Epoch 29, batch 1250, loss[ctc_loss=0.09844, att_loss=0.2546, loss=0.2234, over 14316.00 frames. utt_duration=393.8 frames, utt_pad_proportion=0.3128, over 146.00 utterances.], tot_loss[ctc_loss=0.06459, att_loss=0.2303, loss=0.1972, over 3260987.36 frames. utt_duration=1245 frames, utt_pad_proportion=0.05818, over 10492.09 utterances.], batch size: 146, lr: 3.71e-03, grad_scale: 8.0 2023-03-09 12:45:11,408 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.6148, 2.3575, 2.3606, 2.5606, 2.9012, 2.8680, 2.3552, 3.0938], device='cuda:1'), covar=tensor([0.1792, 0.2277, 0.1662, 0.1287, 0.1257, 0.0987, 0.2058, 0.1213], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0150, 0.0146, 0.0140, 0.0157, 0.0135, 0.0158, 0.0136], 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') 2023-03-09 12:45:16,783 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4708, 2.7305, 4.9967, 3.9933, 3.1900, 4.3667, 4.9009, 4.7300], device='cuda:1'), covar=tensor([0.0332, 0.1434, 0.0291, 0.0926, 0.1558, 0.0271, 0.0208, 0.0290], device='cuda:1'), in_proj_covar=tensor([0.0237, 0.0249, 0.0231, 0.0328, 0.0273, 0.0244, 0.0224, 0.0246], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 12:46:20,451 INFO [train2.py:809] (1/4) Epoch 29, batch 1300, loss[ctc_loss=0.05103, att_loss=0.2191, loss=0.1854, over 16537.00 frames. utt_duration=1471 frames, utt_pad_proportion=0.005841, over 45.00 utterances.], tot_loss[ctc_loss=0.06418, att_loss=0.2299, loss=0.1968, over 3265967.11 frames. utt_duration=1248 frames, utt_pad_proportion=0.05467, over 10481.97 utterances.], batch size: 45, lr: 3.70e-03, grad_scale: 8.0 2023-03-09 12:46:24,466 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9590, 5.1802, 5.2032, 5.1589, 5.2555, 5.2208, 4.8974, 4.7039], device='cuda:1'), covar=tensor([0.1076, 0.0570, 0.0323, 0.0498, 0.0284, 0.0334, 0.0411, 0.0315], device='cuda:1'), in_proj_covar=tensor([0.0542, 0.0384, 0.0378, 0.0383, 0.0447, 0.0453, 0.0382, 0.0418], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:1') 2023-03-09 12:46:31,277 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.9185, 3.6706, 3.6449, 3.1355, 3.6309, 3.6716, 3.6923, 2.6810], device='cuda:1'), covar=tensor([0.0845, 0.0933, 0.1227, 0.2833, 0.1445, 0.1603, 0.0735, 0.2957], device='cuda:1'), in_proj_covar=tensor([0.0207, 0.0209, 0.0225, 0.0276, 0.0187, 0.0286, 0.0208, 0.0231], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-03-09 12:46:40,102 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.148e+02 1.906e+02 2.330e+02 2.870e+02 1.155e+03, threshold=4.661e+02, percent-clipped=5.0 2023-03-09 12:47:40,654 INFO [train2.py:809] (1/4) Epoch 29, batch 1350, loss[ctc_loss=0.06159, att_loss=0.2374, loss=0.2022, over 16864.00 frames. utt_duration=1378 frames, utt_pad_proportion=0.007802, over 49.00 utterances.], tot_loss[ctc_loss=0.06393, att_loss=0.2294, loss=0.1963, over 3261742.93 frames. utt_duration=1279 frames, utt_pad_proportion=0.04734, over 10209.56 utterances.], batch size: 49, lr: 3.70e-03, grad_scale: 8.0 2023-03-09 12:48:12,930 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0131, 5.2097, 5.2553, 5.2125, 5.2826, 5.2562, 4.8845, 4.7151], device='cuda:1'), covar=tensor([0.1039, 0.0560, 0.0317, 0.0527, 0.0297, 0.0327, 0.0442, 0.0337], device='cuda:1'), in_proj_covar=tensor([0.0541, 0.0384, 0.0377, 0.0383, 0.0446, 0.0453, 0.0381, 0.0418], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:1') 2023-03-09 12:48:17,508 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112918.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:48:22,235 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112921.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:48:41,458 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112933.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 12:49:00,131 INFO [train2.py:809] (1/4) Epoch 29, batch 1400, loss[ctc_loss=0.07646, att_loss=0.2423, loss=0.2091, over 17060.00 frames. utt_duration=1289 frames, utt_pad_proportion=0.009383, over 53.00 utterances.], tot_loss[ctc_loss=0.0641, att_loss=0.2302, loss=0.197, over 3262537.98 frames. utt_duration=1264 frames, utt_pad_proportion=0.05113, over 10338.25 utterances.], batch size: 53, lr: 3.70e-03, grad_scale: 8.0 2023-03-09 12:49:19,015 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.160e+02 1.843e+02 2.238e+02 2.535e+02 6.230e+02, threshold=4.475e+02, percent-clipped=2.0 2023-03-09 12:49:42,539 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-03-09 12:49:44,610 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112973.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 12:49:59,358 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112982.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:50:02,266 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.6174, 2.9557, 3.5166, 4.5412, 4.0738, 4.0664, 3.0110, 2.7146], device='cuda:1'), covar=tensor([0.0641, 0.1928, 0.0856, 0.0563, 0.0910, 0.0520, 0.1519, 0.1846], device='cuda:1'), in_proj_covar=tensor([0.0188, 0.0219, 0.0184, 0.0227, 0.0233, 0.0192, 0.0204, 0.0190], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 12:50:19,501 INFO [train2.py:809] (1/4) Epoch 29, batch 1450, loss[ctc_loss=0.07289, att_loss=0.2406, loss=0.2071, over 15965.00 frames. utt_duration=1559 frames, utt_pad_proportion=0.006358, over 41.00 utterances.], tot_loss[ctc_loss=0.06493, att_loss=0.2313, loss=0.198, over 3263956.32 frames. utt_duration=1232 frames, utt_pad_proportion=0.06006, over 10610.85 utterances.], batch size: 41, lr: 3.70e-03, grad_scale: 8.0 2023-03-09 12:51:39,662 INFO [train2.py:809] (1/4) Epoch 29, batch 1500, loss[ctc_loss=0.06382, att_loss=0.2339, loss=0.1999, over 17157.00 frames. utt_duration=870.3 frames, utt_pad_proportion=0.08873, over 79.00 utterances.], tot_loss[ctc_loss=0.06453, att_loss=0.2313, loss=0.198, over 3271042.73 frames. utt_duration=1229 frames, utt_pad_proportion=0.05828, over 10663.27 utterances.], batch size: 79, lr: 3.70e-03, grad_scale: 8.0 2023-03-09 12:51:39,991 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113045.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:51:58,818 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.135e+02 1.764e+02 2.089e+02 2.547e+02 5.820e+02, threshold=4.178e+02, percent-clipped=2.0 2023-03-09 12:52:04,148 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4503, 2.4040, 4.8298, 3.9328, 3.0513, 4.1370, 4.6461, 4.5551], device='cuda:1'), covar=tensor([0.0284, 0.1692, 0.0217, 0.0863, 0.1618, 0.0305, 0.0207, 0.0282], device='cuda:1'), in_proj_covar=tensor([0.0236, 0.0248, 0.0230, 0.0326, 0.0272, 0.0244, 0.0223, 0.0244], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 12:52:55,238 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=113093.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:52:58,840 INFO [train2.py:809] (1/4) Epoch 29, batch 1550, loss[ctc_loss=0.05793, att_loss=0.2069, loss=0.1771, over 15372.00 frames. utt_duration=1758 frames, utt_pad_proportion=0.0111, over 35.00 utterances.], tot_loss[ctc_loss=0.06483, att_loss=0.2314, loss=0.1981, over 3263881.94 frames. utt_duration=1214 frames, utt_pad_proportion=0.0649, over 10770.69 utterances.], batch size: 35, lr: 3.70e-03, grad_scale: 8.0 2023-03-09 12:54:19,352 INFO [train2.py:809] (1/4) Epoch 29, batch 1600, loss[ctc_loss=0.05403, att_loss=0.2141, loss=0.1821, over 16123.00 frames. utt_duration=1537 frames, utt_pad_proportion=0.006526, over 42.00 utterances.], tot_loss[ctc_loss=0.06427, att_loss=0.2311, loss=0.1977, over 3259186.01 frames. utt_duration=1217 frames, utt_pad_proportion=0.06456, over 10728.32 utterances.], batch size: 42, lr: 3.70e-03, grad_scale: 8.0 2023-03-09 12:54:30,430 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.14 vs. limit=5.0 2023-03-09 12:54:38,748 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.261e+02 1.848e+02 2.168e+02 2.655e+02 4.318e+02, threshold=4.336e+02, percent-clipped=2.0 2023-03-09 12:55:38,883 INFO [train2.py:809] (1/4) Epoch 29, batch 1650, loss[ctc_loss=0.04399, att_loss=0.2249, loss=0.1887, over 17018.00 frames. utt_duration=1336 frames, utt_pad_proportion=0.007904, over 51.00 utterances.], tot_loss[ctc_loss=0.06412, att_loss=0.2312, loss=0.1978, over 3268339.06 frames. utt_duration=1238 frames, utt_pad_proportion=0.05829, over 10575.02 utterances.], batch size: 51, lr: 3.70e-03, grad_scale: 8.0 2023-03-09 12:56:13,483 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9026, 4.8785, 4.5976, 2.9511, 4.5879, 4.4778, 4.1126, 2.5212], device='cuda:1'), covar=tensor([0.0122, 0.0127, 0.0307, 0.1056, 0.0140, 0.0252, 0.0358, 0.1564], device='cuda:1'), in_proj_covar=tensor([0.0079, 0.0107, 0.0112, 0.0113, 0.0091, 0.0120, 0.0102, 0.0106], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-09 12:56:14,950 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113218.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:56:38,823 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113233.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:56:40,404 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113234.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:56:58,265 INFO [train2.py:809] (1/4) Epoch 29, batch 1700, loss[ctc_loss=0.04698, att_loss=0.219, loss=0.1846, over 16012.00 frames. utt_duration=1603 frames, utt_pad_proportion=0.007079, over 40.00 utterances.], tot_loss[ctc_loss=0.06334, att_loss=0.2307, loss=0.1972, over 3263250.10 frames. utt_duration=1216 frames, utt_pad_proportion=0.06516, over 10749.90 utterances.], batch size: 40, lr: 3.70e-03, grad_scale: 8.0 2023-03-09 12:57:17,572 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.337e+02 1.796e+02 2.106e+02 2.556e+02 4.663e+02, threshold=4.212e+02, percent-clipped=1.0 2023-03-09 12:57:27,315 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4302, 2.5847, 4.8808, 3.8869, 3.1572, 4.1912, 4.6009, 4.6269], device='cuda:1'), covar=tensor([0.0282, 0.1538, 0.0198, 0.0895, 0.1510, 0.0279, 0.0214, 0.0256], device='cuda:1'), in_proj_covar=tensor([0.0235, 0.0248, 0.0230, 0.0326, 0.0272, 0.0244, 0.0223, 0.0244], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 12:57:31,608 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=113266.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:57:32,325 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-03-09 12:57:43,269 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113273.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 12:57:49,282 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=113277.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:57:55,519 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=113281.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:58:18,083 INFO [train2.py:809] (1/4) Epoch 29, batch 1750, loss[ctc_loss=0.06061, att_loss=0.2239, loss=0.1912, over 15947.00 frames. utt_duration=1557 frames, utt_pad_proportion=0.007353, over 41.00 utterances.], tot_loss[ctc_loss=0.06342, att_loss=0.2304, loss=0.197, over 3264139.49 frames. utt_duration=1237 frames, utt_pad_proportion=0.05967, over 10567.24 utterances.], batch size: 41, lr: 3.70e-03, grad_scale: 8.0 2023-03-09 12:58:18,472 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113295.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:58:49,045 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-03-09 12:58:59,077 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=113321.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 12:59:37,380 INFO [train2.py:809] (1/4) Epoch 29, batch 1800, loss[ctc_loss=0.05844, att_loss=0.221, loss=0.1885, over 16177.00 frames. utt_duration=1580 frames, utt_pad_proportion=0.006565, over 41.00 utterances.], tot_loss[ctc_loss=0.0633, att_loss=0.2308, loss=0.1973, over 3273679.84 frames. utt_duration=1246 frames, utt_pad_proportion=0.05522, over 10526.06 utterances.], batch size: 41, lr: 3.70e-03, grad_scale: 4.0 2023-03-09 12:59:57,720 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.115e+02 1.881e+02 2.263e+02 2.651e+02 5.025e+02, threshold=4.526e+02, percent-clipped=2.0 2023-03-09 13:00:27,335 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4195, 2.5291, 3.6868, 2.4888, 3.5396, 4.6616, 4.6134, 2.8536], device='cuda:1'), covar=tensor([0.0445, 0.2238, 0.0942, 0.1932, 0.0946, 0.0572, 0.0423, 0.1608], device='cuda:1'), in_proj_covar=tensor([0.0254, 0.0254, 0.0297, 0.0223, 0.0278, 0.0389, 0.0278, 0.0239], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0005, 0.0003, 0.0003], device='cuda:1') 2023-03-09 13:00:55,305 INFO [train2.py:809] (1/4) Epoch 29, batch 1850, loss[ctc_loss=0.06147, att_loss=0.214, loss=0.1835, over 15640.00 frames. utt_duration=1692 frames, utt_pad_proportion=0.008708, over 37.00 utterances.], tot_loss[ctc_loss=0.06497, att_loss=0.2316, loss=0.1983, over 3272871.59 frames. utt_duration=1213 frames, utt_pad_proportion=0.0637, over 10806.57 utterances.], batch size: 37, lr: 3.70e-03, grad_scale: 4.0 2023-03-09 13:02:15,435 INFO [train2.py:809] (1/4) Epoch 29, batch 1900, loss[ctc_loss=0.04906, att_loss=0.2225, loss=0.1878, over 16098.00 frames. utt_duration=1535 frames, utt_pad_proportion=0.007757, over 42.00 utterances.], tot_loss[ctc_loss=0.06475, att_loss=0.2309, loss=0.1977, over 3274438.41 frames. utt_duration=1237 frames, utt_pad_proportion=0.05699, over 10603.50 utterances.], batch size: 42, lr: 3.70e-03, grad_scale: 4.0 2023-03-09 13:02:35,445 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.243e+02 1.919e+02 2.347e+02 2.814e+02 3.658e+03, threshold=4.695e+02, percent-clipped=4.0 2023-03-09 13:03:25,858 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.4336, 5.6848, 5.1898, 5.4840, 5.3795, 4.9083, 5.1836, 4.8854], device='cuda:1'), covar=tensor([0.1287, 0.1092, 0.1165, 0.0890, 0.1249, 0.1507, 0.2250, 0.2283], device='cuda:1'), in_proj_covar=tensor([0.0567, 0.0645, 0.0496, 0.0479, 0.0460, 0.0486, 0.0651, 0.0550], 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') 2023-03-09 13:03:34,135 INFO [train2.py:809] (1/4) Epoch 29, batch 1950, loss[ctc_loss=0.06288, att_loss=0.2465, loss=0.2098, over 17021.00 frames. utt_duration=1336 frames, utt_pad_proportion=0.007861, over 51.00 utterances.], tot_loss[ctc_loss=0.06413, att_loss=0.2307, loss=0.1973, over 3265298.70 frames. utt_duration=1248 frames, utt_pad_proportion=0.05626, over 10482.17 utterances.], batch size: 51, lr: 3.69e-03, grad_scale: 4.0 2023-03-09 13:03:38,830 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-03-09 13:04:54,506 INFO [train2.py:809] (1/4) Epoch 29, batch 2000, loss[ctc_loss=0.04554, att_loss=0.2014, loss=0.1702, over 15632.00 frames. utt_duration=1691 frames, utt_pad_proportion=0.009167, over 37.00 utterances.], tot_loss[ctc_loss=0.06303, att_loss=0.2295, loss=0.1962, over 3264700.73 frames. utt_duration=1274 frames, utt_pad_proportion=0.04966, over 10265.04 utterances.], batch size: 37, lr: 3.69e-03, grad_scale: 8.0 2023-03-09 13:05:15,511 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.032e+02 1.751e+02 2.072e+02 2.530e+02 5.087e+02, threshold=4.144e+02, percent-clipped=1.0 2023-03-09 13:05:45,753 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113577.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 13:06:05,667 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=113590.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 13:06:14,832 INFO [train2.py:809] (1/4) Epoch 29, batch 2050, loss[ctc_loss=0.05067, att_loss=0.2235, loss=0.1889, over 16404.00 frames. utt_duration=1493 frames, utt_pad_proportion=0.007389, over 44.00 utterances.], tot_loss[ctc_loss=0.0632, att_loss=0.23, loss=0.1966, over 3269139.84 frames. utt_duration=1279 frames, utt_pad_proportion=0.04779, over 10233.49 utterances.], batch size: 44, lr: 3.69e-03, grad_scale: 8.0 2023-03-09 13:06:49,204 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113617.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 13:06:54,482 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4749, 4.7185, 4.6464, 4.6834, 4.8096, 4.4826, 3.3991, 4.7348], device='cuda:1'), covar=tensor([0.0117, 0.0111, 0.0152, 0.0091, 0.0097, 0.0131, 0.0669, 0.0171], device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0092, 0.0117, 0.0073, 0.0080, 0.0091, 0.0106, 0.0111], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 13:07:02,078 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=113625.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 13:07:35,392 INFO [train2.py:809] (1/4) Epoch 29, batch 2100, loss[ctc_loss=0.0546, att_loss=0.2066, loss=0.1762, over 14046.00 frames. utt_duration=1814 frames, utt_pad_proportion=0.03514, over 31.00 utterances.], tot_loss[ctc_loss=0.06327, att_loss=0.2307, loss=0.1972, over 3271463.82 frames. utt_duration=1237 frames, utt_pad_proportion=0.05623, over 10594.80 utterances.], batch size: 31, lr: 3.69e-03, grad_scale: 8.0 2023-03-09 13:07:37,265 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113646.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 13:07:44,806 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0653, 4.9637, 4.7187, 2.8096, 4.6854, 4.7434, 4.2187, 2.8380], device='cuda:1'), covar=tensor([0.0114, 0.0126, 0.0306, 0.1162, 0.0140, 0.0185, 0.0361, 0.1367], device='cuda:1'), in_proj_covar=tensor([0.0080, 0.0108, 0.0113, 0.0113, 0.0092, 0.0120, 0.0102, 0.0106], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-09 13:07:48,854 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-03-09 13:07:55,042 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.156e+02 1.850e+02 2.211e+02 2.646e+02 5.386e+02, threshold=4.422e+02, percent-clipped=4.0 2023-03-09 13:07:55,468 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.6933, 2.5893, 2.4370, 2.6065, 2.7700, 2.8275, 2.3956, 2.9652], device='cuda:1'), covar=tensor([0.1501, 0.2018, 0.1868, 0.1425, 0.1466, 0.0993, 0.1822, 0.1169], device='cuda:1'), in_proj_covar=tensor([0.0150, 0.0148, 0.0146, 0.0141, 0.0157, 0.0136, 0.0158, 0.0136], 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') 2023-03-09 13:08:26,922 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113678.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 13:08:55,294 INFO [train2.py:809] (1/4) Epoch 29, batch 2150, loss[ctc_loss=0.06817, att_loss=0.2365, loss=0.2028, over 16679.00 frames. utt_duration=1452 frames, utt_pad_proportion=0.006116, over 46.00 utterances.], tot_loss[ctc_loss=0.06401, att_loss=0.2314, loss=0.1979, over 3271264.20 frames. utt_duration=1224 frames, utt_pad_proportion=0.06107, over 10702.62 utterances.], batch size: 46, lr: 3.69e-03, grad_scale: 8.0 2023-03-09 13:09:13,863 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113707.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 13:10:13,178 INFO [train2.py:809] (1/4) Epoch 29, batch 2200, loss[ctc_loss=0.06634, att_loss=0.2401, loss=0.2054, over 17342.00 frames. utt_duration=1177 frames, utt_pad_proportion=0.02212, over 59.00 utterances.], tot_loss[ctc_loss=0.06386, att_loss=0.2315, loss=0.1979, over 3276812.88 frames. utt_duration=1235 frames, utt_pad_proportion=0.05622, over 10623.86 utterances.], batch size: 59, lr: 3.69e-03, grad_scale: 8.0 2023-03-09 13:10:30,453 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.19 vs. limit=5.0 2023-03-09 13:10:32,605 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.369e+02 1.987e+02 2.347e+02 2.821e+02 4.899e+02, threshold=4.694e+02, percent-clipped=4.0 2023-03-09 13:11:32,258 INFO [train2.py:809] (1/4) Epoch 29, batch 2250, loss[ctc_loss=0.0516, att_loss=0.2104, loss=0.1786, over 15397.00 frames. utt_duration=1761 frames, utt_pad_proportion=0.008924, over 35.00 utterances.], tot_loss[ctc_loss=0.06404, att_loss=0.2311, loss=0.1977, over 3270981.76 frames. utt_duration=1220 frames, utt_pad_proportion=0.06255, over 10736.41 utterances.], batch size: 35, lr: 3.69e-03, grad_scale: 8.0 2023-03-09 13:12:01,340 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-03-09 13:12:18,341 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-03-09 13:12:51,817 INFO [train2.py:809] (1/4) Epoch 29, batch 2300, loss[ctc_loss=0.1104, att_loss=0.2562, loss=0.227, over 13912.00 frames. utt_duration=385.3 frames, utt_pad_proportion=0.3299, over 145.00 utterances.], tot_loss[ctc_loss=0.06414, att_loss=0.2308, loss=0.1975, over 3262318.31 frames. utt_duration=1218 frames, utt_pad_proportion=0.0652, over 10729.61 utterances.], batch size: 145, lr: 3.69e-03, grad_scale: 8.0 2023-03-09 13:13:12,812 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.222e+02 1.846e+02 2.164e+02 2.611e+02 9.969e+02, threshold=4.329e+02, percent-clipped=2.0 2023-03-09 13:13:13,833 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=2.98 vs. limit=5.0 2023-03-09 13:14:03,910 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113890.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 13:14:11,233 INFO [train2.py:809] (1/4) Epoch 29, batch 2350, loss[ctc_loss=0.06402, att_loss=0.2438, loss=0.2078, over 17394.00 frames. utt_duration=1106 frames, utt_pad_proportion=0.03344, over 63.00 utterances.], tot_loss[ctc_loss=0.06414, att_loss=0.2311, loss=0.1977, over 3264895.68 frames. utt_duration=1207 frames, utt_pad_proportion=0.06752, over 10836.80 utterances.], batch size: 63, lr: 3.69e-03, grad_scale: 8.0 2023-03-09 13:14:27,155 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-03-09 13:14:32,510 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4904, 2.8886, 4.9432, 4.0676, 3.1681, 4.2586, 4.7093, 4.6935], device='cuda:1'), covar=tensor([0.0339, 0.1322, 0.0313, 0.0803, 0.1509, 0.0297, 0.0272, 0.0314], device='cuda:1'), in_proj_covar=tensor([0.0237, 0.0248, 0.0230, 0.0324, 0.0271, 0.0244, 0.0224, 0.0244], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 13:15:19,182 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=113938.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 13:15:29,855 INFO [train2.py:809] (1/4) Epoch 29, batch 2400, loss[ctc_loss=0.04775, att_loss=0.2267, loss=0.1909, over 16333.00 frames. utt_duration=1453 frames, utt_pad_proportion=0.00608, over 45.00 utterances.], tot_loss[ctc_loss=0.0638, att_loss=0.2303, loss=0.197, over 3264010.92 frames. utt_duration=1248 frames, utt_pad_proportion=0.05791, over 10470.44 utterances.], batch size: 45, lr: 3.69e-03, grad_scale: 8.0 2023-03-09 13:15:50,423 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.212e+02 1.796e+02 2.124e+02 2.708e+02 5.253e+02, threshold=4.249e+02, percent-clipped=4.0 2023-03-09 13:16:13,555 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=113973.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 13:16:49,308 INFO [train2.py:809] (1/4) Epoch 29, batch 2450, loss[ctc_loss=0.07539, att_loss=0.2426, loss=0.2091, over 17345.00 frames. utt_duration=879.6 frames, utt_pad_proportion=0.07893, over 79.00 utterances.], tot_loss[ctc_loss=0.06346, att_loss=0.23, loss=0.1967, over 3271311.78 frames. utt_duration=1265 frames, utt_pad_proportion=0.05124, over 10354.59 utterances.], batch size: 79, lr: 3.69e-03, grad_scale: 4.0 2023-03-09 13:17:04,660 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=114002.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 13:18:12,393 INFO [train2.py:809] (1/4) Epoch 29, batch 2500, loss[ctc_loss=0.06365, att_loss=0.2393, loss=0.2042, over 17018.00 frames. utt_duration=1336 frames, utt_pad_proportion=0.008742, over 51.00 utterances.], tot_loss[ctc_loss=0.06441, att_loss=0.2306, loss=0.1973, over 3274705.03 frames. utt_duration=1270 frames, utt_pad_proportion=0.04911, over 10326.27 utterances.], batch size: 51, lr: 3.69e-03, grad_scale: 4.0 2023-03-09 13:18:31,991 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.1028, 5.2982, 5.3015, 5.2631, 5.3810, 5.3599, 4.9941, 4.7857], device='cuda:1'), covar=tensor([0.1014, 0.0549, 0.0316, 0.0486, 0.0330, 0.0310, 0.0426, 0.0323], device='cuda:1'), in_proj_covar=tensor([0.0538, 0.0382, 0.0379, 0.0384, 0.0448, 0.0451, 0.0382, 0.0421], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:1') 2023-03-09 13:18:34,822 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.371e+02 1.923e+02 2.146e+02 2.615e+02 3.828e+02, threshold=4.291e+02, percent-clipped=0.0 2023-03-09 13:19:02,132 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0691, 4.3045, 4.2432, 4.5216, 2.6643, 4.5002, 2.6630, 1.9599], device='cuda:1'), covar=tensor([0.0505, 0.0377, 0.0762, 0.0270, 0.1747, 0.0247, 0.1591, 0.1719], device='cuda:1'), in_proj_covar=tensor([0.0222, 0.0193, 0.0266, 0.0183, 0.0224, 0.0173, 0.0234, 0.0204], 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') 2023-03-09 13:19:32,907 INFO [train2.py:809] (1/4) Epoch 29, batch 2550, loss[ctc_loss=0.05518, att_loss=0.2265, loss=0.1923, over 16416.00 frames. utt_duration=1494 frames, utt_pad_proportion=0.0068, over 44.00 utterances.], tot_loss[ctc_loss=0.06411, att_loss=0.2297, loss=0.1966, over 3264528.80 frames. utt_duration=1264 frames, utt_pad_proportion=0.05525, over 10345.55 utterances.], batch size: 44, lr: 3.68e-03, grad_scale: 4.0 2023-03-09 13:20:23,968 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=114127.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 13:20:38,873 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5749, 4.8735, 4.7473, 4.9249, 4.9897, 4.6580, 3.3895, 4.8069], device='cuda:1'), covar=tensor([0.0120, 0.0110, 0.0145, 0.0063, 0.0089, 0.0113, 0.0731, 0.0168], device='cuda:1'), in_proj_covar=tensor([0.0097, 0.0093, 0.0118, 0.0073, 0.0080, 0.0091, 0.0107, 0.0112], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 13:20:52,674 INFO [train2.py:809] (1/4) Epoch 29, batch 2600, loss[ctc_loss=0.05468, att_loss=0.2245, loss=0.1905, over 16422.00 frames. utt_duration=1495 frames, utt_pad_proportion=0.006256, over 44.00 utterances.], tot_loss[ctc_loss=0.06345, att_loss=0.2292, loss=0.1961, over 3262702.42 frames. utt_duration=1281 frames, utt_pad_proportion=0.05144, over 10202.01 utterances.], batch size: 44, lr: 3.68e-03, grad_scale: 4.0 2023-03-09 13:21:15,102 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 9.450e+01 1.839e+02 2.178e+02 2.659e+02 5.439e+02, threshold=4.356e+02, percent-clipped=3.0 2023-03-09 13:21:21,337 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.3250, 5.3158, 5.1379, 3.2809, 5.1334, 4.9963, 4.7322, 2.8915], device='cuda:1'), covar=tensor([0.0115, 0.0089, 0.0230, 0.0960, 0.0093, 0.0183, 0.0265, 0.1340], device='cuda:1'), in_proj_covar=tensor([0.0080, 0.0107, 0.0113, 0.0113, 0.0091, 0.0121, 0.0103, 0.0107], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-09 13:22:01,499 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=114188.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 13:22:11,801 INFO [train2.py:809] (1/4) Epoch 29, batch 2650, loss[ctc_loss=0.07846, att_loss=0.2146, loss=0.1874, over 16008.00 frames. utt_duration=1602 frames, utt_pad_proportion=0.007234, over 40.00 utterances.], tot_loss[ctc_loss=0.06335, att_loss=0.2289, loss=0.1958, over 3258038.13 frames. utt_duration=1294 frames, utt_pad_proportion=0.04858, over 10085.84 utterances.], batch size: 40, lr: 3.68e-03, grad_scale: 4.0 2023-03-09 13:23:31,707 INFO [train2.py:809] (1/4) Epoch 29, batch 2700, loss[ctc_loss=0.07025, att_loss=0.2494, loss=0.2135, over 16970.00 frames. utt_duration=1359 frames, utt_pad_proportion=0.006608, over 50.00 utterances.], tot_loss[ctc_loss=0.06336, att_loss=0.2295, loss=0.1963, over 3267396.08 frames. utt_duration=1269 frames, utt_pad_proportion=0.05146, over 10308.83 utterances.], batch size: 50, lr: 3.68e-03, grad_scale: 4.0 2023-03-09 13:23:53,983 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.272e+02 1.816e+02 2.102e+02 2.530e+02 4.306e+02, threshold=4.205e+02, percent-clipped=0.0 2023-03-09 13:24:16,605 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=114273.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 13:24:51,547 INFO [train2.py:809] (1/4) Epoch 29, batch 2750, loss[ctc_loss=0.04092, att_loss=0.2272, loss=0.19, over 16458.00 frames. utt_duration=1433 frames, utt_pad_proportion=0.007111, over 46.00 utterances.], tot_loss[ctc_loss=0.06314, att_loss=0.229, loss=0.1959, over 3264671.24 frames. utt_duration=1291 frames, utt_pad_proportion=0.048, over 10125.23 utterances.], batch size: 46, lr: 3.68e-03, grad_scale: 4.0 2023-03-09 13:25:03,021 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=114302.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 13:25:32,591 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=114321.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 13:25:51,117 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.1417, 6.3456, 5.8990, 6.0833, 6.0707, 5.5467, 5.8252, 5.4889], device='cuda:1'), covar=tensor([0.1202, 0.0910, 0.0923, 0.0805, 0.0859, 0.1546, 0.2150, 0.2252], device='cuda:1'), in_proj_covar=tensor([0.0568, 0.0645, 0.0495, 0.0479, 0.0459, 0.0486, 0.0646, 0.0552], 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') 2023-03-09 13:26:11,320 INFO [train2.py:809] (1/4) Epoch 29, batch 2800, loss[ctc_loss=0.06438, att_loss=0.2468, loss=0.2103, over 17376.00 frames. utt_duration=881.2 frames, utt_pad_proportion=0.07829, over 79.00 utterances.], tot_loss[ctc_loss=0.06264, att_loss=0.2287, loss=0.1955, over 3259161.66 frames. utt_duration=1298 frames, utt_pad_proportion=0.04662, over 10059.18 utterances.], batch size: 79, lr: 3.68e-03, grad_scale: 8.0 2023-03-09 13:26:19,152 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=114350.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 13:26:19,446 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.5394, 2.4238, 2.4478, 2.6292, 2.7173, 2.6666, 2.3117, 2.9000], device='cuda:1'), covar=tensor([0.1336, 0.2021, 0.1592, 0.1097, 0.1625, 0.0898, 0.1839, 0.1511], device='cuda:1'), in_proj_covar=tensor([0.0152, 0.0150, 0.0148, 0.0142, 0.0159, 0.0136, 0.0161, 0.0138], 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') 2023-03-09 13:26:33,215 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.204e+02 1.822e+02 2.190e+02 2.901e+02 6.134e+02, threshold=4.379e+02, percent-clipped=9.0 2023-03-09 13:27:30,409 INFO [train2.py:809] (1/4) Epoch 29, batch 2850, loss[ctc_loss=0.08039, att_loss=0.2524, loss=0.218, over 17337.00 frames. utt_duration=1262 frames, utt_pad_proportion=0.009882, over 55.00 utterances.], tot_loss[ctc_loss=0.06333, att_loss=0.2292, loss=0.196, over 3255067.66 frames. utt_duration=1280 frames, utt_pad_proportion=0.05061, over 10183.88 utterances.], batch size: 55, lr: 3.68e-03, grad_scale: 8.0 2023-03-09 13:28:50,608 INFO [train2.py:809] (1/4) Epoch 29, batch 2900, loss[ctc_loss=0.0544, att_loss=0.225, loss=0.1909, over 16280.00 frames. utt_duration=1516 frames, utt_pad_proportion=0.007219, over 43.00 utterances.], tot_loss[ctc_loss=0.06299, att_loss=0.2294, loss=0.1961, over 3263182.37 frames. utt_duration=1263 frames, utt_pad_proportion=0.05332, over 10346.05 utterances.], batch size: 43, lr: 3.68e-03, grad_scale: 8.0 2023-03-09 13:29:12,892 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.369e+02 1.896e+02 2.215e+02 2.552e+02 6.599e+02, threshold=4.430e+02, percent-clipped=4.0 2023-03-09 13:29:52,206 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=114483.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 13:30:10,925 INFO [train2.py:809] (1/4) Epoch 29, batch 2950, loss[ctc_loss=0.05898, att_loss=0.2425, loss=0.2058, over 17007.00 frames. utt_duration=1335 frames, utt_pad_proportion=0.008559, over 51.00 utterances.], tot_loss[ctc_loss=0.0632, att_loss=0.2303, loss=0.1969, over 3277410.07 frames. utt_duration=1273 frames, utt_pad_proportion=0.04704, over 10312.28 utterances.], batch size: 51, lr: 3.68e-03, grad_scale: 8.0 2023-03-09 13:31:30,959 INFO [train2.py:809] (1/4) Epoch 29, batch 3000, loss[ctc_loss=0.06845, att_loss=0.2337, loss=0.2006, over 17313.00 frames. utt_duration=1101 frames, utt_pad_proportion=0.03782, over 63.00 utterances.], tot_loss[ctc_loss=0.06314, att_loss=0.2299, loss=0.1965, over 3274799.37 frames. utt_duration=1258 frames, utt_pad_proportion=0.0507, over 10421.75 utterances.], batch size: 63, lr: 3.68e-03, grad_scale: 8.0 2023-03-09 13:31:30,959 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-09 13:31:44,959 INFO [train2.py:843] (1/4) Epoch 29, validation: ctc_loss=0.04115, att_loss=0.2348, loss=0.1961, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-09 13:31:44,960 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16129MB 2023-03-09 13:32:06,763 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.233e+02 1.877e+02 2.271e+02 2.617e+02 4.276e+02, threshold=4.543e+02, percent-clipped=0.0 2023-03-09 13:33:04,604 INFO [train2.py:809] (1/4) Epoch 29, batch 3050, loss[ctc_loss=0.1066, att_loss=0.256, loss=0.2261, over 14530.00 frames. utt_duration=397 frames, utt_pad_proportion=0.3047, over 147.00 utterances.], tot_loss[ctc_loss=0.06457, att_loss=0.2314, loss=0.198, over 3282285.36 frames. utt_duration=1224 frames, utt_pad_proportion=0.05717, over 10737.16 utterances.], batch size: 147, lr: 3.68e-03, grad_scale: 8.0 2023-03-09 13:33:06,407 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9505, 5.2627, 5.4594, 5.3268, 5.4398, 5.8988, 5.2742, 6.0072], device='cuda:1'), covar=tensor([0.0792, 0.0768, 0.0984, 0.1462, 0.1999, 0.0979, 0.0737, 0.0734], device='cuda:1'), in_proj_covar=tensor([0.0912, 0.0530, 0.0641, 0.0676, 0.0903, 0.0662, 0.0509, 0.0635], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 13:34:24,749 INFO [train2.py:809] (1/4) Epoch 29, batch 3100, loss[ctc_loss=0.06138, att_loss=0.2145, loss=0.1839, over 15749.00 frames. utt_duration=1659 frames, utt_pad_proportion=0.009364, over 38.00 utterances.], tot_loss[ctc_loss=0.06422, att_loss=0.2312, loss=0.1978, over 3280193.40 frames. utt_duration=1217 frames, utt_pad_proportion=0.06033, over 10791.82 utterances.], batch size: 38, lr: 3.68e-03, grad_scale: 8.0 2023-03-09 13:34:46,590 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.249e+02 1.785e+02 2.252e+02 2.835e+02 6.156e+02, threshold=4.505e+02, percent-clipped=2.0 2023-03-09 13:35:44,840 INFO [train2.py:809] (1/4) Epoch 29, batch 3150, loss[ctc_loss=0.04987, att_loss=0.2427, loss=0.2041, over 17291.00 frames. utt_duration=877 frames, utt_pad_proportion=0.08164, over 79.00 utterances.], tot_loss[ctc_loss=0.06443, att_loss=0.2318, loss=0.1983, over 3277255.64 frames. utt_duration=1198 frames, utt_pad_proportion=0.06547, over 10951.88 utterances.], batch size: 79, lr: 3.67e-03, grad_scale: 8.0 2023-03-09 13:36:18,407 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=114716.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 13:36:28,380 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-03-09 13:36:34,210 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.1831, 2.7231, 3.0783, 4.2628, 3.8881, 3.8555, 2.8548, 2.2685], device='cuda:1'), covar=tensor([0.0836, 0.2047, 0.0978, 0.0623, 0.0894, 0.0493, 0.1473, 0.2144], device='cuda:1'), in_proj_covar=tensor([0.0192, 0.0222, 0.0189, 0.0233, 0.0239, 0.0196, 0.0209, 0.0194], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:1') 2023-03-09 13:37:04,184 INFO [train2.py:809] (1/4) Epoch 29, batch 3200, loss[ctc_loss=0.09455, att_loss=0.257, loss=0.2245, over 14270.00 frames. utt_duration=392.6 frames, utt_pad_proportion=0.3161, over 146.00 utterances.], tot_loss[ctc_loss=0.06454, att_loss=0.2319, loss=0.1984, over 3277451.78 frames. utt_duration=1194 frames, utt_pad_proportion=0.06661, over 10992.55 utterances.], batch size: 146, lr: 3.67e-03, grad_scale: 8.0 2023-03-09 13:37:18,349 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.3362, 2.9378, 3.3533, 4.5008, 4.0753, 3.9845, 3.0503, 2.4075], device='cuda:1'), covar=tensor([0.0847, 0.1945, 0.0919, 0.0471, 0.0851, 0.0556, 0.1408, 0.2143], device='cuda:1'), in_proj_covar=tensor([0.0192, 0.0222, 0.0188, 0.0232, 0.0239, 0.0196, 0.0209, 0.0194], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:1') 2023-03-09 13:37:27,015 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.042e+02 1.814e+02 2.180e+02 2.558e+02 5.727e+02, threshold=4.361e+02, percent-clipped=1.0 2023-03-09 13:37:43,899 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.15 vs. limit=5.0 2023-03-09 13:37:52,371 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.3371, 2.7957, 3.3936, 4.5120, 4.0219, 3.9634, 2.9536, 2.3594], device='cuda:1'), covar=tensor([0.0883, 0.2108, 0.0922, 0.0530, 0.0888, 0.0562, 0.1576, 0.2170], device='cuda:1'), in_proj_covar=tensor([0.0193, 0.0223, 0.0189, 0.0233, 0.0240, 0.0196, 0.0210, 0.0195], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:1') 2023-03-09 13:37:54,475 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=114777.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 13:38:04,238 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=114783.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 13:38:07,291 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0433, 5.4455, 5.5981, 5.3642, 5.5495, 6.0423, 5.3536, 6.1301], device='cuda:1'), covar=tensor([0.0768, 0.0724, 0.0959, 0.1441, 0.1992, 0.0829, 0.0723, 0.0615], device='cuda:1'), in_proj_covar=tensor([0.0914, 0.0530, 0.0640, 0.0678, 0.0906, 0.0661, 0.0511, 0.0635], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 13:38:22,854 INFO [train2.py:809] (1/4) Epoch 29, batch 3250, loss[ctc_loss=0.07339, att_loss=0.2335, loss=0.2014, over 16945.00 frames. utt_duration=686.2 frames, utt_pad_proportion=0.1401, over 99.00 utterances.], tot_loss[ctc_loss=0.06445, att_loss=0.2316, loss=0.1981, over 3278877.63 frames. utt_duration=1211 frames, utt_pad_proportion=0.06281, over 10839.81 utterances.], batch size: 99, lr: 3.67e-03, grad_scale: 8.0 2023-03-09 13:39:20,496 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=114831.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 13:39:42,520 INFO [train2.py:809] (1/4) Epoch 29, batch 3300, loss[ctc_loss=0.07039, att_loss=0.2372, loss=0.2038, over 17380.00 frames. utt_duration=1009 frames, utt_pad_proportion=0.04732, over 69.00 utterances.], tot_loss[ctc_loss=0.06412, att_loss=0.2312, loss=0.1978, over 3278615.37 frames. utt_duration=1219 frames, utt_pad_proportion=0.06128, over 10775.30 utterances.], batch size: 69, lr: 3.67e-03, grad_scale: 8.0 2023-03-09 13:40:05,657 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.268e+02 1.794e+02 2.073e+02 2.411e+02 5.454e+02, threshold=4.146e+02, percent-clipped=1.0 2023-03-09 13:40:23,004 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=114871.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 13:41:01,525 INFO [train2.py:809] (1/4) Epoch 29, batch 3350, loss[ctc_loss=0.05861, att_loss=0.2055, loss=0.1761, over 15778.00 frames. utt_duration=1662 frames, utt_pad_proportion=0.008243, over 38.00 utterances.], tot_loss[ctc_loss=0.06421, att_loss=0.2311, loss=0.1977, over 3274796.55 frames. utt_duration=1229 frames, utt_pad_proportion=0.06009, over 10670.20 utterances.], batch size: 38, lr: 3.67e-03, grad_scale: 8.0 2023-03-09 13:42:00,796 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=114932.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 13:42:13,672 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2468, 4.3848, 4.4421, 4.4438, 4.9376, 4.3397, 4.3384, 2.5459], device='cuda:1'), covar=tensor([0.0359, 0.0441, 0.0376, 0.0378, 0.0668, 0.0302, 0.0407, 0.1721], device='cuda:1'), in_proj_covar=tensor([0.0201, 0.0228, 0.0224, 0.0241, 0.0384, 0.0197, 0.0217, 0.0223], 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') 2023-03-09 13:42:20,955 INFO [train2.py:809] (1/4) Epoch 29, batch 3400, loss[ctc_loss=0.06232, att_loss=0.2425, loss=0.2064, over 16413.00 frames. utt_duration=1494 frames, utt_pad_proportion=0.006906, over 44.00 utterances.], tot_loss[ctc_loss=0.06409, att_loss=0.2313, loss=0.1979, over 3281091.74 frames. utt_duration=1232 frames, utt_pad_proportion=0.05724, over 10663.57 utterances.], batch size: 44, lr: 3.67e-03, grad_scale: 8.0 2023-03-09 13:42:44,850 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.258e+02 1.858e+02 2.229e+02 2.688e+02 7.167e+02, threshold=4.457e+02, percent-clipped=6.0 2023-03-09 13:43:27,106 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.88 vs. limit=5.0 2023-03-09 13:43:35,019 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-03-09 13:43:41,692 INFO [train2.py:809] (1/4) Epoch 29, batch 3450, loss[ctc_loss=0.07317, att_loss=0.2344, loss=0.2022, over 16267.00 frames. utt_duration=1515 frames, utt_pad_proportion=0.0073, over 43.00 utterances.], tot_loss[ctc_loss=0.06446, att_loss=0.2318, loss=0.1983, over 3287323.78 frames. utt_duration=1216 frames, utt_pad_proportion=0.05869, over 10823.77 utterances.], batch size: 43, lr: 3.67e-03, grad_scale: 8.0 2023-03-09 13:43:43,104 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-03-09 13:45:02,579 INFO [train2.py:809] (1/4) Epoch 29, batch 3500, loss[ctc_loss=0.06233, att_loss=0.2363, loss=0.2015, over 16483.00 frames. utt_duration=1435 frames, utt_pad_proportion=0.006278, over 46.00 utterances.], tot_loss[ctc_loss=0.06352, att_loss=0.2305, loss=0.1971, over 3275642.43 frames. utt_duration=1244 frames, utt_pad_proportion=0.05523, over 10546.95 utterances.], batch size: 46, lr: 3.67e-03, grad_scale: 8.0 2023-03-09 13:45:26,258 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.209e+02 1.733e+02 2.269e+02 2.703e+02 9.473e+02, threshold=4.539e+02, percent-clipped=4.0 2023-03-09 13:45:43,155 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.3149, 3.0859, 3.4238, 4.4151, 3.9100, 3.9582, 2.8465, 2.4051], device='cuda:1'), covar=tensor([0.0803, 0.1680, 0.0809, 0.0567, 0.0885, 0.0500, 0.1593, 0.2028], device='cuda:1'), in_proj_covar=tensor([0.0193, 0.0223, 0.0189, 0.0233, 0.0240, 0.0196, 0.0210, 0.0195], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:1') 2023-03-09 13:45:46,134 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115072.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 13:46:06,720 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.73 vs. limit=2.0 2023-03-09 13:46:09,451 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115086.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 13:46:22,695 INFO [train2.py:809] (1/4) Epoch 29, batch 3550, loss[ctc_loss=0.07232, att_loss=0.2378, loss=0.2047, over 17331.00 frames. utt_duration=1102 frames, utt_pad_proportion=0.03597, over 63.00 utterances.], tot_loss[ctc_loss=0.06423, att_loss=0.2309, loss=0.1975, over 3268050.56 frames. utt_duration=1246 frames, utt_pad_proportion=0.05619, over 10506.68 utterances.], batch size: 63, lr: 3.67e-03, grad_scale: 8.0 2023-03-09 13:46:33,877 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6067, 4.9708, 4.7906, 4.8644, 5.0068, 4.6345, 3.6266, 4.9454], device='cuda:1'), covar=tensor([0.0140, 0.0121, 0.0155, 0.0098, 0.0102, 0.0144, 0.0645, 0.0183], device='cuda:1'), in_proj_covar=tensor([0.0099, 0.0095, 0.0120, 0.0075, 0.0082, 0.0093, 0.0109, 0.0114], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003, 0.0004], device='cuda:1') 2023-03-09 13:47:27,455 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.87 vs. limit=2.0 2023-03-09 13:47:42,043 INFO [train2.py:809] (1/4) Epoch 29, batch 3600, loss[ctc_loss=0.04555, att_loss=0.2221, loss=0.1868, over 16777.00 frames. utt_duration=1400 frames, utt_pad_proportion=0.005741, over 48.00 utterances.], tot_loss[ctc_loss=0.06412, att_loss=0.231, loss=0.1977, over 3273105.59 frames. utt_duration=1254 frames, utt_pad_proportion=0.05132, over 10457.20 utterances.], batch size: 48, lr: 3.67e-03, grad_scale: 8.0 2023-03-09 13:47:45,610 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115147.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 13:48:05,806 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.296e+02 1.777e+02 2.035e+02 2.403e+02 5.227e+02, threshold=4.069e+02, percent-clipped=1.0 2023-03-09 13:49:02,352 INFO [train2.py:809] (1/4) Epoch 29, batch 3650, loss[ctc_loss=0.0632, att_loss=0.2158, loss=0.1853, over 15950.00 frames. utt_duration=1558 frames, utt_pad_proportion=0.00644, over 41.00 utterances.], tot_loss[ctc_loss=0.0637, att_loss=0.2309, loss=0.1975, over 3277124.38 frames. utt_duration=1249 frames, utt_pad_proportion=0.05225, over 10507.28 utterances.], batch size: 41, lr: 3.67e-03, grad_scale: 8.0 2023-03-09 13:49:54,460 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115227.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 13:50:23,432 INFO [train2.py:809] (1/4) Epoch 29, batch 3700, loss[ctc_loss=0.05901, att_loss=0.2467, loss=0.2092, over 17336.00 frames. utt_duration=1007 frames, utt_pad_proportion=0.05026, over 69.00 utterances.], tot_loss[ctc_loss=0.06296, att_loss=0.2304, loss=0.1969, over 3275260.46 frames. utt_duration=1251 frames, utt_pad_proportion=0.05161, over 10481.07 utterances.], batch size: 69, lr: 3.67e-03, grad_scale: 8.0 2023-03-09 13:50:41,839 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.2298, 5.2036, 4.9818, 3.1596, 5.0346, 4.7828, 4.6154, 3.1243], device='cuda:1'), covar=tensor([0.0116, 0.0108, 0.0284, 0.0966, 0.0101, 0.0213, 0.0253, 0.1176], device='cuda:1'), in_proj_covar=tensor([0.0079, 0.0106, 0.0112, 0.0112, 0.0091, 0.0120, 0.0102, 0.0106], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-09 13:50:47,790 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.294e+02 1.820e+02 2.082e+02 2.598e+02 4.716e+02, threshold=4.165e+02, percent-clipped=2.0 2023-03-09 13:51:44,492 INFO [train2.py:809] (1/4) Epoch 29, batch 3750, loss[ctc_loss=0.05504, att_loss=0.2116, loss=0.1803, over 15496.00 frames. utt_duration=1723 frames, utt_pad_proportion=0.009265, over 36.00 utterances.], tot_loss[ctc_loss=0.06293, att_loss=0.2306, loss=0.1971, over 3273783.31 frames. utt_duration=1235 frames, utt_pad_proportion=0.05685, over 10612.31 utterances.], batch size: 36, lr: 3.67e-03, grad_scale: 8.0 2023-03-09 13:52:27,449 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1845, 4.4097, 4.4928, 4.6091, 3.0446, 4.6526, 3.1415, 2.0857], device='cuda:1'), covar=tensor([0.0530, 0.0316, 0.0662, 0.0311, 0.1465, 0.0267, 0.1290, 0.1663], device='cuda:1'), in_proj_covar=tensor([0.0224, 0.0196, 0.0269, 0.0185, 0.0227, 0.0176, 0.0238, 0.0207], 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') 2023-03-09 13:53:04,020 INFO [train2.py:809] (1/4) Epoch 29, batch 3800, loss[ctc_loss=0.1054, att_loss=0.2615, loss=0.2303, over 14145.00 frames. utt_duration=386.4 frames, utt_pad_proportion=0.3232, over 147.00 utterances.], tot_loss[ctc_loss=0.06287, att_loss=0.2305, loss=0.197, over 3269284.69 frames. utt_duration=1221 frames, utt_pad_proportion=0.06308, over 10727.30 utterances.], batch size: 147, lr: 3.66e-03, grad_scale: 8.0 2023-03-09 13:53:27,651 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.195e+02 1.885e+02 2.239e+02 2.673e+02 5.789e+02, threshold=4.479e+02, percent-clipped=6.0 2023-03-09 13:53:43,987 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115370.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 13:53:46,174 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115371.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 13:53:47,650 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=115372.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 13:54:02,269 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2956, 4.4773, 4.7223, 4.7149, 2.9477, 4.8013, 3.3467, 2.0790], device='cuda:1'), covar=tensor([0.0473, 0.0301, 0.0540, 0.0276, 0.1449, 0.0203, 0.1076, 0.1521], device='cuda:1'), in_proj_covar=tensor([0.0223, 0.0195, 0.0268, 0.0185, 0.0226, 0.0175, 0.0237, 0.0205], 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') 2023-03-09 13:54:23,178 INFO [train2.py:809] (1/4) Epoch 29, batch 3850, loss[ctc_loss=0.05539, att_loss=0.2265, loss=0.1923, over 16171.00 frames. utt_duration=1579 frames, utt_pad_proportion=0.007435, over 41.00 utterances.], tot_loss[ctc_loss=0.06236, att_loss=0.2295, loss=0.1961, over 3263957.12 frames. utt_duration=1252 frames, utt_pad_proportion=0.05763, over 10439.82 utterances.], batch size: 41, lr: 3.66e-03, grad_scale: 8.0 2023-03-09 13:55:01,793 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=115420.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 13:55:18,396 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115431.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 13:55:20,026 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115432.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 13:55:26,056 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.4922, 3.9217, 3.4329, 3.6157, 4.1597, 3.8718, 3.1541, 4.4654], device='cuda:1'), covar=tensor([0.0772, 0.0545, 0.1008, 0.0720, 0.0667, 0.0676, 0.0868, 0.0414], device='cuda:1'), in_proj_covar=tensor([0.0207, 0.0227, 0.0230, 0.0207, 0.0289, 0.0249, 0.0205, 0.0297], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:1') 2023-03-09 13:55:34,964 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115442.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 13:55:35,239 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4236, 2.6130, 4.8641, 3.7368, 3.1540, 4.1789, 4.4847, 4.5692], device='cuda:1'), covar=tensor([0.0286, 0.1478, 0.0219, 0.0935, 0.1469, 0.0284, 0.0238, 0.0281], device='cuda:1'), in_proj_covar=tensor([0.0238, 0.0247, 0.0232, 0.0324, 0.0272, 0.0244, 0.0226, 0.0245], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 13:55:39,327 INFO [train2.py:809] (1/4) Epoch 29, batch 3900, loss[ctc_loss=0.0435, att_loss=0.2152, loss=0.1808, over 16161.00 frames. utt_duration=1578 frames, utt_pad_proportion=0.00683, over 41.00 utterances.], tot_loss[ctc_loss=0.06248, att_loss=0.2302, loss=0.1967, over 3272970.11 frames. utt_duration=1261 frames, utt_pad_proportion=0.05296, over 10397.23 utterances.], batch size: 41, lr: 3.66e-03, grad_scale: 8.0 2023-03-09 13:55:59,608 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-03-09 13:56:02,815 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.234e+02 1.900e+02 2.175e+02 2.636e+02 4.480e+02, threshold=4.350e+02, percent-clipped=1.0 2023-03-09 13:56:44,183 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.0552, 5.3788, 4.9592, 5.3975, 4.7850, 5.0230, 5.4774, 5.2453], device='cuda:1'), covar=tensor([0.0580, 0.0345, 0.0803, 0.0393, 0.0441, 0.0237, 0.0239, 0.0207], device='cuda:1'), in_proj_covar=tensor([0.0408, 0.0343, 0.0385, 0.0384, 0.0340, 0.0246, 0.0322, 0.0306], device='cuda:1'), out_proj_covar=tensor([0.0007, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-09 13:56:55,929 INFO [train2.py:809] (1/4) Epoch 29, batch 3950, loss[ctc_loss=0.05592, att_loss=0.2371, loss=0.2009, over 16762.00 frames. utt_duration=1398 frames, utt_pad_proportion=0.00691, over 48.00 utterances.], tot_loss[ctc_loss=0.06269, att_loss=0.2307, loss=0.1971, over 3276113.54 frames. utt_duration=1252 frames, utt_pad_proportion=0.05241, over 10483.38 utterances.], batch size: 48, lr: 3.66e-03, grad_scale: 8.0 2023-03-09 13:57:08,658 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5275, 2.5446, 4.9811, 3.8800, 3.1455, 4.3074, 4.7434, 4.7051], device='cuda:1'), covar=tensor([0.0317, 0.1577, 0.0243, 0.0947, 0.1588, 0.0259, 0.0234, 0.0279], device='cuda:1'), in_proj_covar=tensor([0.0238, 0.0246, 0.0232, 0.0323, 0.0271, 0.0244, 0.0225, 0.0244], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 13:57:14,607 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2993, 4.5397, 4.5573, 4.7247, 2.9444, 4.7147, 3.0172, 1.9588], device='cuda:1'), covar=tensor([0.0505, 0.0287, 0.0585, 0.0285, 0.1487, 0.0237, 0.1258, 0.1628], device='cuda:1'), in_proj_covar=tensor([0.0223, 0.0194, 0.0268, 0.0184, 0.0225, 0.0175, 0.0236, 0.0204], 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') 2023-03-09 13:57:28,068 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.8991, 3.4877, 3.6181, 3.0625, 3.5591, 3.6600, 3.6103, 2.5760], device='cuda:1'), covar=tensor([0.1255, 0.1723, 0.1558, 0.3032, 0.0975, 0.1878, 0.1048, 0.3562], device='cuda:1'), in_proj_covar=tensor([0.0212, 0.0215, 0.0228, 0.0279, 0.0191, 0.0290, 0.0214, 0.0235], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-03-09 13:57:44,616 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=115527.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 13:58:12,588 INFO [train2.py:809] (1/4) Epoch 30, batch 0, loss[ctc_loss=0.06945, att_loss=0.2321, loss=0.1995, over 17323.00 frames. utt_duration=878.7 frames, utt_pad_proportion=0.07897, over 79.00 utterances.], tot_loss[ctc_loss=0.06945, att_loss=0.2321, loss=0.1995, over 17323.00 frames. utt_duration=878.7 frames, utt_pad_proportion=0.07897, over 79.00 utterances.], batch size: 79, lr: 3.60e-03, grad_scale: 8.0 2023-03-09 13:58:12,589 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-09 13:58:17,395 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9431, 4.9749, 4.8397, 2.3408, 2.1433, 3.2328, 2.4074, 3.7412], device='cuda:1'), covar=tensor([0.0736, 0.0311, 0.0303, 0.5089, 0.5456, 0.2213, 0.4140, 0.1533], device='cuda:1'), in_proj_covar=tensor([0.0365, 0.0307, 0.0280, 0.0253, 0.0340, 0.0332, 0.0265, 0.0373], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-09 13:58:24,678 INFO [train2.py:843] (1/4) Epoch 30, validation: ctc_loss=0.03959, att_loss=0.2341, loss=0.1952, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-09 13:58:24,679 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16129MB 2023-03-09 13:58:49,320 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2509, 2.7817, 3.2974, 4.4327, 3.9296, 3.8498, 2.7806, 2.1644], device='cuda:1'), covar=tensor([0.0954, 0.2125, 0.0997, 0.0583, 0.0948, 0.0613, 0.1793, 0.2476], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0223, 0.0189, 0.0232, 0.0241, 0.0196, 0.0209, 0.0195], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:1') 2023-03-09 13:59:05,303 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-03-09 13:59:13,838 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.172e+02 1.763e+02 2.122e+02 2.494e+02 5.580e+02, threshold=4.245e+02, percent-clipped=4.0 2023-03-09 13:59:38,542 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=115575.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 13:59:44,601 INFO [train2.py:809] (1/4) Epoch 30, batch 50, loss[ctc_loss=0.05719, att_loss=0.2265, loss=0.1926, over 16625.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.005379, over 47.00 utterances.], tot_loss[ctc_loss=0.06454, att_loss=0.2323, loss=0.1988, over 735133.79 frames. utt_duration=1148 frames, utt_pad_proportion=0.08419, over 2565.41 utterances.], batch size: 47, lr: 3.60e-03, grad_scale: 8.0 2023-03-09 14:01:04,302 INFO [train2.py:809] (1/4) Epoch 30, batch 100, loss[ctc_loss=0.04227, att_loss=0.2096, loss=0.1762, over 16004.00 frames. utt_duration=1602 frames, utt_pad_proportion=0.007636, over 40.00 utterances.], tot_loss[ctc_loss=0.06344, att_loss=0.2316, loss=0.198, over 1307128.14 frames. utt_duration=1233 frames, utt_pad_proportion=0.05438, over 4246.30 utterances.], batch size: 40, lr: 3.60e-03, grad_scale: 8.0 2023-03-09 14:01:06,165 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.6046, 3.0622, 3.5889, 4.5522, 4.0300, 4.0359, 3.0582, 2.5306], device='cuda:1'), covar=tensor([0.0737, 0.1925, 0.0896, 0.0539, 0.0928, 0.0522, 0.1527, 0.2120], device='cuda:1'), in_proj_covar=tensor([0.0193, 0.0222, 0.0189, 0.0232, 0.0241, 0.0196, 0.0209, 0.0195], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:1') 2023-03-09 14:01:54,048 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.157e+02 1.758e+02 2.072e+02 2.595e+02 5.402e+02, threshold=4.143e+02, percent-clipped=3.0 2023-03-09 14:02:13,421 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.4090, 2.9433, 3.4281, 4.4400, 3.8709, 3.9148, 2.8747, 2.3572], device='cuda:1'), covar=tensor([0.0777, 0.1853, 0.0909, 0.0577, 0.1035, 0.0604, 0.1673, 0.2147], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0223, 0.0190, 0.0233, 0.0242, 0.0197, 0.0210, 0.0196], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:1') 2023-03-09 14:02:22,600 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.2343, 5.4378, 5.4087, 5.4183, 5.5382, 5.4983, 5.1160, 4.9578], device='cuda:1'), covar=tensor([0.0931, 0.0535, 0.0321, 0.0426, 0.0265, 0.0295, 0.0403, 0.0321], device='cuda:1'), in_proj_covar=tensor([0.0543, 0.0387, 0.0380, 0.0386, 0.0453, 0.0456, 0.0385, 0.0425], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:1') 2023-03-09 14:02:23,894 INFO [train2.py:809] (1/4) Epoch 30, batch 150, loss[ctc_loss=0.05348, att_loss=0.2056, loss=0.1752, over 15501.00 frames. utt_duration=1724 frames, utt_pad_proportion=0.008311, over 36.00 utterances.], tot_loss[ctc_loss=0.06413, att_loss=0.2312, loss=0.1978, over 1745478.05 frames. utt_duration=1233 frames, utt_pad_proportion=0.05246, over 5669.94 utterances.], batch size: 36, lr: 3.60e-03, grad_scale: 8.0 2023-03-09 14:03:04,244 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.6785, 5.9361, 5.4545, 5.6363, 5.6074, 5.0403, 5.3272, 5.0814], device='cuda:1'), covar=tensor([0.1253, 0.0841, 0.0965, 0.0931, 0.0970, 0.1834, 0.2315, 0.2175], device='cuda:1'), in_proj_covar=tensor([0.0563, 0.0638, 0.0492, 0.0476, 0.0454, 0.0486, 0.0641, 0.0543], 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') 2023-03-09 14:03:39,622 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115726.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:03:39,880 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4892, 2.7732, 4.8957, 3.9384, 3.0602, 4.3096, 4.7404, 4.6651], device='cuda:1'), covar=tensor([0.0354, 0.1353, 0.0269, 0.0866, 0.1618, 0.0289, 0.0257, 0.0310], device='cuda:1'), in_proj_covar=tensor([0.0239, 0.0247, 0.0233, 0.0324, 0.0272, 0.0246, 0.0226, 0.0245], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 14:03:41,133 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115727.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:03:44,045 INFO [train2.py:809] (1/4) Epoch 30, batch 200, loss[ctc_loss=0.0441, att_loss=0.1996, loss=0.1685, over 15391.00 frames. utt_duration=1760 frames, utt_pad_proportion=0.009979, over 35.00 utterances.], tot_loss[ctc_loss=0.0628, att_loss=0.2305, loss=0.197, over 2083358.77 frames. utt_duration=1256 frames, utt_pad_proportion=0.0486, over 6643.37 utterances.], batch size: 35, lr: 3.60e-03, grad_scale: 8.0 2023-03-09 14:04:04,995 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=115742.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:04:23,483 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.5308, 2.5071, 2.5923, 2.3728, 2.5520, 2.4599, 2.6133, 2.0198], device='cuda:1'), covar=tensor([0.1082, 0.1384, 0.1606, 0.2801, 0.1223, 0.2470, 0.1366, 0.3029], device='cuda:1'), in_proj_covar=tensor([0.0211, 0.0214, 0.0227, 0.0279, 0.0190, 0.0289, 0.0212, 0.0235], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-03-09 14:04:32,779 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.110e+02 1.790e+02 2.190e+02 2.640e+02 5.774e+02, threshold=4.380e+02, percent-clipped=3.0 2023-03-09 14:05:03,783 INFO [train2.py:809] (1/4) Epoch 30, batch 250, loss[ctc_loss=0.04017, att_loss=0.1991, loss=0.1673, over 15615.00 frames. utt_duration=1689 frames, utt_pad_proportion=0.0105, over 37.00 utterances.], tot_loss[ctc_loss=0.06326, att_loss=0.231, loss=0.1974, over 2350167.27 frames. utt_duration=1252 frames, utt_pad_proportion=0.05032, over 7517.91 utterances.], batch size: 37, lr: 3.60e-03, grad_scale: 8.0 2023-03-09 14:05:15,099 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4198, 2.4376, 4.8098, 3.7723, 2.9976, 4.0811, 4.5738, 4.5777], device='cuda:1'), covar=tensor([0.0302, 0.1638, 0.0249, 0.0890, 0.1634, 0.0316, 0.0265, 0.0280], device='cuda:1'), in_proj_covar=tensor([0.0240, 0.0247, 0.0233, 0.0324, 0.0273, 0.0246, 0.0226, 0.0246], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 14:05:18,790 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9637, 5.0830, 4.8055, 2.2600, 1.9490, 3.1567, 2.4053, 3.8885], device='cuda:1'), covar=tensor([0.0793, 0.0317, 0.0362, 0.5228, 0.5711, 0.2226, 0.4058, 0.1593], device='cuda:1'), in_proj_covar=tensor([0.0367, 0.0308, 0.0281, 0.0254, 0.0341, 0.0332, 0.0265, 0.0374], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-09 14:05:21,468 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=115790.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:06:23,344 INFO [train2.py:809] (1/4) Epoch 30, batch 300, loss[ctc_loss=0.05753, att_loss=0.2134, loss=0.1822, over 14543.00 frames. utt_duration=1820 frames, utt_pad_proportion=0.04183, over 32.00 utterances.], tot_loss[ctc_loss=0.06261, att_loss=0.2306, loss=0.197, over 2553451.89 frames. utt_duration=1253 frames, utt_pad_proportion=0.05201, over 8158.83 utterances.], batch size: 32, lr: 3.59e-03, grad_scale: 8.0 2023-03-09 14:07:12,539 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.192e+02 1.839e+02 2.070e+02 2.435e+02 4.420e+02, threshold=4.140e+02, percent-clipped=2.0 2023-03-09 14:07:43,743 INFO [train2.py:809] (1/4) Epoch 30, batch 350, loss[ctc_loss=0.04834, att_loss=0.2155, loss=0.182, over 16397.00 frames. utt_duration=1492 frames, utt_pad_proportion=0.007949, over 44.00 utterances.], tot_loss[ctc_loss=0.06303, att_loss=0.2312, loss=0.1975, over 2715495.59 frames. utt_duration=1229 frames, utt_pad_proportion=0.05859, over 8852.16 utterances.], batch size: 44, lr: 3.59e-03, grad_scale: 8.0 2023-03-09 14:08:05,274 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115892.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 14:09:04,300 INFO [train2.py:809] (1/4) Epoch 30, batch 400, loss[ctc_loss=0.04308, att_loss=0.2148, loss=0.1805, over 16147.00 frames. utt_duration=1539 frames, utt_pad_proportion=0.005125, over 42.00 utterances.], tot_loss[ctc_loss=0.06249, att_loss=0.2306, loss=0.197, over 2840331.95 frames. utt_duration=1227 frames, utt_pad_proportion=0.05855, over 9268.47 utterances.], batch size: 42, lr: 3.59e-03, grad_scale: 8.0 2023-03-09 14:09:43,448 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115953.0, num_to_drop=1, layers_to_drop={3} 2023-03-09 14:09:53,661 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.410e+02 1.924e+02 2.363e+02 2.895e+02 9.842e+02, threshold=4.725e+02, percent-clipped=4.0 2023-03-09 14:10:24,230 INFO [train2.py:809] (1/4) Epoch 30, batch 450, loss[ctc_loss=0.0692, att_loss=0.2266, loss=0.1951, over 15945.00 frames. utt_duration=1557 frames, utt_pad_proportion=0.007539, over 41.00 utterances.], tot_loss[ctc_loss=0.06235, att_loss=0.2294, loss=0.196, over 2927095.98 frames. utt_duration=1270 frames, utt_pad_proportion=0.05053, over 9227.81 utterances.], batch size: 41, lr: 3.59e-03, grad_scale: 8.0 2023-03-09 14:11:25,123 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116014.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:11:43,444 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116026.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:11:45,012 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116027.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:11:48,381 INFO [train2.py:809] (1/4) Epoch 30, batch 500, loss[ctc_loss=0.05579, att_loss=0.2078, loss=0.1774, over 15506.00 frames. utt_duration=1724 frames, utt_pad_proportion=0.00837, over 36.00 utterances.], tot_loss[ctc_loss=0.06208, att_loss=0.2287, loss=0.1953, over 3005058.73 frames. utt_duration=1296 frames, utt_pad_proportion=0.04527, over 9286.50 utterances.], batch size: 36, lr: 3.59e-03, grad_scale: 8.0 2023-03-09 14:12:37,213 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.075e+02 1.798e+02 2.149e+02 2.710e+02 6.262e+02, threshold=4.298e+02, percent-clipped=5.0 2023-03-09 14:12:59,509 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=116074.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:13:01,050 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=116075.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:13:01,327 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116075.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 14:13:07,556 INFO [train2.py:809] (1/4) Epoch 30, batch 550, loss[ctc_loss=0.07104, att_loss=0.2208, loss=0.1908, over 16143.00 frames. utt_duration=1539 frames, utt_pad_proportion=0.00531, over 42.00 utterances.], tot_loss[ctc_loss=0.06227, att_loss=0.2281, loss=0.1949, over 3057018.89 frames. utt_duration=1291 frames, utt_pad_proportion=0.04778, over 9481.58 utterances.], batch size: 42, lr: 3.59e-03, grad_scale: 8.0 2023-03-09 14:13:47,220 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8084, 4.9229, 4.3138, 2.5935, 4.6467, 4.5588, 3.9564, 2.1683], device='cuda:1'), covar=tensor([0.0237, 0.0154, 0.0583, 0.1602, 0.0169, 0.0292, 0.0582, 0.2656], device='cuda:1'), in_proj_covar=tensor([0.0079, 0.0108, 0.0113, 0.0114, 0.0091, 0.0121, 0.0103, 0.0107], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-09 14:13:48,822 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116105.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:14:06,613 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116116.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:14:27,141 INFO [train2.py:809] (1/4) Epoch 30, batch 600, loss[ctc_loss=0.07506, att_loss=0.239, loss=0.2062, over 17523.00 frames. utt_duration=888.8 frames, utt_pad_proportion=0.06741, over 79.00 utterances.], tot_loss[ctc_loss=0.06253, att_loss=0.2288, loss=0.1955, over 3106017.54 frames. utt_duration=1290 frames, utt_pad_proportion=0.04622, over 9644.10 utterances.], batch size: 79, lr: 3.59e-03, grad_scale: 8.0 2023-03-09 14:15:15,793 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 9.620e+01 1.813e+02 2.125e+02 2.957e+02 6.068e+02, threshold=4.250e+02, percent-clipped=9.0 2023-03-09 14:15:26,199 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116166.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:15:44,338 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116177.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:15:47,046 INFO [train2.py:809] (1/4) Epoch 30, batch 650, loss[ctc_loss=0.07647, att_loss=0.2532, loss=0.2178, over 17322.00 frames. utt_duration=1261 frames, utt_pad_proportion=0.01085, over 55.00 utterances.], tot_loss[ctc_loss=0.0629, att_loss=0.229, loss=0.1958, over 3145862.90 frames. utt_duration=1305 frames, utt_pad_proportion=0.0412, over 9653.73 utterances.], batch size: 55, lr: 3.59e-03, grad_scale: 8.0 2023-03-09 14:17:07,962 INFO [train2.py:809] (1/4) Epoch 30, batch 700, loss[ctc_loss=0.06045, att_loss=0.2283, loss=0.1947, over 16790.00 frames. utt_duration=1401 frames, utt_pad_proportion=0.005179, over 48.00 utterances.], tot_loss[ctc_loss=0.06299, att_loss=0.2294, loss=0.1961, over 3173999.30 frames. utt_duration=1292 frames, utt_pad_proportion=0.04287, over 9839.99 utterances.], batch size: 48, lr: 3.59e-03, grad_scale: 8.0 2023-03-09 14:17:38,396 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116248.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 14:17:57,488 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.158e+02 1.730e+02 1.988e+02 2.420e+02 4.405e+02, threshold=3.977e+02, percent-clipped=2.0 2023-03-09 14:18:28,318 INFO [train2.py:809] (1/4) Epoch 30, batch 750, loss[ctc_loss=0.07886, att_loss=0.2492, loss=0.2151, over 16879.00 frames. utt_duration=1380 frames, utt_pad_proportion=0.006817, over 49.00 utterances.], tot_loss[ctc_loss=0.06303, att_loss=0.2293, loss=0.196, over 3194419.98 frames. utt_duration=1278 frames, utt_pad_proportion=0.04657, over 10011.07 utterances.], batch size: 49, lr: 3.59e-03, grad_scale: 8.0 2023-03-09 14:19:37,994 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116323.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:19:47,387 INFO [train2.py:809] (1/4) Epoch 30, batch 800, loss[ctc_loss=0.04133, att_loss=0.2089, loss=0.1754, over 15964.00 frames. utt_duration=1559 frames, utt_pad_proportion=0.005709, over 41.00 utterances.], tot_loss[ctc_loss=0.06288, att_loss=0.2298, loss=0.1964, over 3215988.22 frames. utt_duration=1291 frames, utt_pad_proportion=0.04223, over 9976.46 utterances.], batch size: 41, lr: 3.59e-03, grad_scale: 8.0 2023-03-09 14:20:11,256 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4293, 2.9295, 3.6810, 3.0737, 3.5659, 4.5878, 4.4451, 3.2000], device='cuda:1'), covar=tensor([0.0441, 0.1925, 0.1139, 0.1401, 0.1051, 0.1057, 0.0551, 0.1407], device='cuda:1'), in_proj_covar=tensor([0.0252, 0.0254, 0.0295, 0.0222, 0.0274, 0.0385, 0.0276, 0.0240], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-09 14:20:30,374 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0721, 2.8269, 3.4410, 2.9165, 3.3795, 4.2046, 4.0829, 2.9589], device='cuda:1'), covar=tensor([0.0464, 0.1781, 0.1362, 0.1348, 0.1103, 0.1000, 0.0674, 0.1456], device='cuda:1'), in_proj_covar=tensor([0.0252, 0.0254, 0.0295, 0.0222, 0.0274, 0.0385, 0.0276, 0.0240], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-09 14:20:36,144 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.359e+02 1.849e+02 2.173e+02 2.485e+02 4.246e+02, threshold=4.345e+02, percent-clipped=3.0 2023-03-09 14:20:37,004 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-03-09 14:20:52,675 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116370.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 14:21:06,896 INFO [train2.py:809] (1/4) Epoch 30, batch 850, loss[ctc_loss=0.05245, att_loss=0.2266, loss=0.1917, over 16878.00 frames. utt_duration=1379 frames, utt_pad_proportion=0.007737, over 49.00 utterances.], tot_loss[ctc_loss=0.06289, att_loss=0.2296, loss=0.1963, over 3229447.27 frames. utt_duration=1294 frames, utt_pad_proportion=0.0416, over 9993.92 utterances.], batch size: 49, lr: 3.59e-03, grad_scale: 8.0 2023-03-09 14:21:15,099 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116384.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:22:27,524 INFO [train2.py:809] (1/4) Epoch 30, batch 900, loss[ctc_loss=0.05049, att_loss=0.2204, loss=0.1864, over 16190.00 frames. utt_duration=1581 frames, utt_pad_proportion=0.00586, over 41.00 utterances.], tot_loss[ctc_loss=0.0628, att_loss=0.2298, loss=0.1964, over 3242250.83 frames. utt_duration=1284 frames, utt_pad_proportion=0.04291, over 10114.10 utterances.], batch size: 41, lr: 3.59e-03, grad_scale: 8.0 2023-03-09 14:22:54,797 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.9702, 3.7783, 3.2042, 3.3614, 3.9325, 3.7100, 2.9313, 4.1435], device='cuda:1'), covar=tensor([0.1102, 0.0460, 0.1073, 0.0761, 0.0782, 0.0682, 0.0941, 0.0503], device='cuda:1'), in_proj_covar=tensor([0.0212, 0.0230, 0.0234, 0.0211, 0.0294, 0.0253, 0.0208, 0.0302], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:1') 2023-03-09 14:22:54,805 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116446.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:23:16,729 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.211e+02 1.807e+02 2.134e+02 2.722e+02 5.477e+02, threshold=4.269e+02, percent-clipped=4.0 2023-03-09 14:23:18,527 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116461.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:23:36,267 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116472.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:23:47,730 INFO [train2.py:809] (1/4) Epoch 30, batch 950, loss[ctc_loss=0.03618, att_loss=0.2082, loss=0.1738, over 15769.00 frames. utt_duration=1662 frames, utt_pad_proportion=0.008542, over 38.00 utterances.], tot_loss[ctc_loss=0.06276, att_loss=0.2299, loss=0.1965, over 3244118.47 frames. utt_duration=1289 frames, utt_pad_proportion=0.04341, over 10076.02 utterances.], batch size: 38, lr: 3.58e-03, grad_scale: 8.0 2023-03-09 14:24:29,861 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116505.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:24:33,187 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116507.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 14:25:07,879 INFO [train2.py:809] (1/4) Epoch 30, batch 1000, loss[ctc_loss=0.05428, att_loss=0.2149, loss=0.1828, over 15628.00 frames. utt_duration=1691 frames, utt_pad_proportion=0.01019, over 37.00 utterances.], tot_loss[ctc_loss=0.06245, att_loss=0.2303, loss=0.1967, over 3256977.17 frames. utt_duration=1284 frames, utt_pad_proportion=0.04409, over 10158.56 utterances.], batch size: 37, lr: 3.58e-03, grad_scale: 8.0 2023-03-09 14:25:38,089 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116548.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 14:25:56,768 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.166e+02 1.757e+02 2.029e+02 2.635e+02 6.922e+02, threshold=4.057e+02, percent-clipped=2.0 2023-03-09 14:26:07,190 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116566.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:26:21,837 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9342, 5.0133, 4.7885, 2.3815, 2.0617, 3.1160, 2.3231, 3.8170], device='cuda:1'), covar=tensor([0.0802, 0.0338, 0.0308, 0.4888, 0.5209, 0.2233, 0.4218, 0.1659], device='cuda:1'), in_proj_covar=tensor([0.0362, 0.0304, 0.0277, 0.0250, 0.0335, 0.0330, 0.0262, 0.0370], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-03-09 14:26:27,612 INFO [train2.py:809] (1/4) Epoch 30, batch 1050, loss[ctc_loss=0.05552, att_loss=0.2122, loss=0.1809, over 15892.00 frames. utt_duration=1632 frames, utt_pad_proportion=0.008755, over 39.00 utterances.], tot_loss[ctc_loss=0.06243, att_loss=0.2305, loss=0.1969, over 3267730.08 frames. utt_duration=1278 frames, utt_pad_proportion=0.04421, over 10240.98 utterances.], batch size: 39, lr: 3.58e-03, grad_scale: 8.0 2023-03-09 14:26:37,707 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3836, 4.6440, 3.9143, 4.7869, 4.2208, 4.3343, 4.6888, 4.5306], device='cuda:1'), covar=tensor([0.0651, 0.0396, 0.1182, 0.0460, 0.0423, 0.0648, 0.0360, 0.0268], device='cuda:1'), in_proj_covar=tensor([0.0410, 0.0344, 0.0386, 0.0388, 0.0341, 0.0248, 0.0323, 0.0307], device='cuda:1'), out_proj_covar=tensor([0.0007, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-09 14:26:54,737 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=116596.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 14:27:30,156 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116618.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:27:47,630 INFO [train2.py:809] (1/4) Epoch 30, batch 1100, loss[ctc_loss=0.06595, att_loss=0.2346, loss=0.2008, over 16806.00 frames. utt_duration=680.6 frames, utt_pad_proportion=0.1461, over 99.00 utterances.], tot_loss[ctc_loss=0.06243, att_loss=0.23, loss=0.1965, over 3267924.60 frames. utt_duration=1284 frames, utt_pad_proportion=0.04403, over 10194.50 utterances.], batch size: 99, lr: 3.58e-03, grad_scale: 8.0 2023-03-09 14:27:53,159 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9672, 6.1833, 5.6909, 5.8451, 5.8576, 5.2965, 5.6573, 5.3111], device='cuda:1'), covar=tensor([0.1159, 0.0931, 0.0829, 0.0963, 0.0962, 0.1544, 0.2261, 0.2200], device='cuda:1'), in_proj_covar=tensor([0.0568, 0.0641, 0.0495, 0.0477, 0.0458, 0.0486, 0.0649, 0.0545], 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') 2023-03-09 14:28:34,267 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8254, 5.0215, 4.9373, 5.0140, 5.0615, 5.0746, 4.6618, 4.5378], device='cuda:1'), covar=tensor([0.0935, 0.0576, 0.0452, 0.0552, 0.0303, 0.0350, 0.0477, 0.0359], device='cuda:1'), in_proj_covar=tensor([0.0539, 0.0385, 0.0381, 0.0386, 0.0451, 0.0454, 0.0384, 0.0422], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:1') 2023-03-09 14:28:37,014 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.262e+02 1.803e+02 2.274e+02 2.882e+02 6.074e+02, threshold=4.549e+02, percent-clipped=3.0 2023-03-09 14:28:54,117 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116670.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 14:29:07,781 INFO [train2.py:809] (1/4) Epoch 30, batch 1150, loss[ctc_loss=0.06775, att_loss=0.2155, loss=0.186, over 14578.00 frames. utt_duration=1823 frames, utt_pad_proportion=0.03319, over 32.00 utterances.], tot_loss[ctc_loss=0.06293, att_loss=0.2302, loss=0.1967, over 3266445.88 frames. utt_duration=1260 frames, utt_pad_proportion=0.05083, over 10382.21 utterances.], batch size: 32, lr: 3.58e-03, grad_scale: 8.0 2023-03-09 14:29:08,014 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116679.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:29:08,196 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116679.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:29:47,287 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116704.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:30:09,470 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=116718.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:30:26,869 INFO [train2.py:809] (1/4) Epoch 30, batch 1200, loss[ctc_loss=0.0477, att_loss=0.225, loss=0.1895, over 17075.00 frames. utt_duration=691.5 frames, utt_pad_proportion=0.1324, over 99.00 utterances.], tot_loss[ctc_loss=0.06273, att_loss=0.2303, loss=0.1968, over 3263714.49 frames. utt_duration=1245 frames, utt_pad_proportion=0.05506, over 10501.13 utterances.], batch size: 99, lr: 3.58e-03, grad_scale: 8.0 2023-03-09 14:30:56,380 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9313, 5.2542, 5.4532, 5.2922, 5.4756, 5.8860, 5.2169, 6.0126], device='cuda:1'), covar=tensor([0.0765, 0.0826, 0.0898, 0.1438, 0.1751, 0.1085, 0.0680, 0.0706], device='cuda:1'), in_proj_covar=tensor([0.0930, 0.0538, 0.0652, 0.0689, 0.0910, 0.0681, 0.0523, 0.0651], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 14:31:16,204 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.120e+02 1.678e+02 2.013e+02 2.421e+02 9.091e+02, threshold=4.026e+02, percent-clipped=5.0 2023-03-09 14:31:18,142 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116761.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:31:25,116 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116765.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:31:35,984 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116772.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:31:47,173 INFO [train2.py:809] (1/4) Epoch 30, batch 1250, loss[ctc_loss=0.06935, att_loss=0.2147, loss=0.1856, over 16154.00 frames. utt_duration=1578 frames, utt_pad_proportion=0.007762, over 41.00 utterances.], tot_loss[ctc_loss=0.06274, att_loss=0.2306, loss=0.197, over 3273068.46 frames. utt_duration=1245 frames, utt_pad_proportion=0.05115, over 10528.99 utterances.], batch size: 41, lr: 3.58e-03, grad_scale: 16.0 2023-03-09 14:31:48,085 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.63 vs. limit=5.0 2023-03-09 14:32:23,515 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116802.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 14:32:23,653 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116802.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:32:35,295 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=116809.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:32:53,097 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=116820.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:32:56,199 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.8115, 6.0937, 5.5829, 5.7922, 5.7388, 5.2426, 5.4702, 5.2460], device='cuda:1'), covar=tensor([0.1319, 0.0928, 0.0959, 0.0852, 0.1067, 0.1664, 0.2324, 0.2104], device='cuda:1'), in_proj_covar=tensor([0.0569, 0.0645, 0.0496, 0.0479, 0.0460, 0.0487, 0.0651, 0.0548], 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') 2023-03-09 14:33:07,493 INFO [train2.py:809] (1/4) Epoch 30, batch 1300, loss[ctc_loss=0.06483, att_loss=0.2359, loss=0.2017, over 17411.00 frames. utt_duration=1107 frames, utt_pad_proportion=0.03127, over 63.00 utterances.], tot_loss[ctc_loss=0.06306, att_loss=0.2305, loss=0.197, over 3257045.09 frames. utt_duration=1232 frames, utt_pad_proportion=0.05808, over 10583.95 utterances.], batch size: 63, lr: 3.58e-03, grad_scale: 16.0 2023-03-09 14:33:57,069 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.212e+02 1.796e+02 2.133e+02 2.719e+02 8.820e+02, threshold=4.267e+02, percent-clipped=2.0 2023-03-09 14:33:58,854 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116861.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:34:02,721 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116863.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:34:28,215 INFO [train2.py:809] (1/4) Epoch 30, batch 1350, loss[ctc_loss=0.05104, att_loss=0.2383, loss=0.2008, over 16882.00 frames. utt_duration=1380 frames, utt_pad_proportion=0.007473, over 49.00 utterances.], tot_loss[ctc_loss=0.0627, att_loss=0.2307, loss=0.1971, over 3262196.04 frames. utt_duration=1235 frames, utt_pad_proportion=0.05779, over 10582.38 utterances.], batch size: 49, lr: 3.58e-03, grad_scale: 16.0 2023-03-09 14:34:35,081 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9437, 5.0003, 4.6895, 2.2050, 2.1227, 3.0710, 2.5951, 3.8451], device='cuda:1'), covar=tensor([0.0812, 0.0332, 0.0377, 0.5492, 0.5152, 0.2296, 0.3494, 0.1648], device='cuda:1'), in_proj_covar=tensor([0.0363, 0.0306, 0.0279, 0.0252, 0.0337, 0.0332, 0.0262, 0.0371], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-03-09 14:34:55,317 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.8269, 4.5781, 4.5536, 2.1430, 2.1138, 2.8696, 2.3531, 3.7138], device='cuda:1'), covar=tensor([0.0787, 0.0333, 0.0294, 0.5712, 0.4820, 0.2455, 0.3738, 0.1476], device='cuda:1'), in_proj_covar=tensor([0.0364, 0.0307, 0.0279, 0.0253, 0.0337, 0.0333, 0.0263, 0.0372], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-09 14:35:07,885 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.6192, 3.5398, 3.4595, 3.8325, 2.8230, 3.7316, 2.8541, 2.3026], device='cuda:1'), covar=tensor([0.0558, 0.0479, 0.0766, 0.0436, 0.1307, 0.0329, 0.1183, 0.1375], device='cuda:1'), in_proj_covar=tensor([0.0225, 0.0196, 0.0267, 0.0184, 0.0225, 0.0176, 0.0236, 0.0205], 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') 2023-03-09 14:35:43,044 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.38 vs. limit=5.0 2023-03-09 14:35:48,163 INFO [train2.py:809] (1/4) Epoch 30, batch 1400, loss[ctc_loss=0.07003, att_loss=0.2384, loss=0.2047, over 16775.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.005405, over 48.00 utterances.], tot_loss[ctc_loss=0.06269, att_loss=0.2308, loss=0.1972, over 3266590.70 frames. utt_duration=1238 frames, utt_pad_proportion=0.05676, over 10564.05 utterances.], batch size: 48, lr: 3.58e-03, grad_scale: 16.0 2023-03-09 14:36:37,551 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.312e+02 1.819e+02 2.157e+02 2.707e+02 5.079e+02, threshold=4.315e+02, percent-clipped=2.0 2023-03-09 14:37:01,011 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116974.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:37:01,605 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.66 vs. limit=5.0 2023-03-09 14:37:08,669 INFO [train2.py:809] (1/4) Epoch 30, batch 1450, loss[ctc_loss=0.04635, att_loss=0.207, loss=0.1748, over 16169.00 frames. utt_duration=1579 frames, utt_pad_proportion=0.007512, over 41.00 utterances.], tot_loss[ctc_loss=0.06249, att_loss=0.2304, loss=0.1968, over 3266518.59 frames. utt_duration=1244 frames, utt_pad_proportion=0.05659, over 10513.97 utterances.], batch size: 41, lr: 3.58e-03, grad_scale: 16.0 2023-03-09 14:37:09,002 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116979.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:37:13,819 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2191, 4.3395, 4.3563, 4.4207, 4.9057, 4.3377, 4.3120, 2.6381], device='cuda:1'), covar=tensor([0.0339, 0.0432, 0.0422, 0.0365, 0.0627, 0.0290, 0.0429, 0.1680], device='cuda:1'), in_proj_covar=tensor([0.0207, 0.0235, 0.0227, 0.0247, 0.0390, 0.0202, 0.0222, 0.0226], 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') 2023-03-09 14:38:25,651 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=117027.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:38:28,587 INFO [train2.py:809] (1/4) Epoch 30, batch 1500, loss[ctc_loss=0.05898, att_loss=0.2269, loss=0.1933, over 16542.00 frames. utt_duration=1472 frames, utt_pad_proportion=0.005495, over 45.00 utterances.], tot_loss[ctc_loss=0.06177, att_loss=0.2292, loss=0.1957, over 3265939.14 frames. utt_duration=1276 frames, utt_pad_proportion=0.04962, over 10251.19 utterances.], batch size: 45, lr: 3.58e-03, grad_scale: 16.0 2023-03-09 14:39:00,530 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.1012, 5.0882, 4.7980, 2.3785, 2.0525, 3.4312, 2.4939, 3.9635], device='cuda:1'), covar=tensor([0.0728, 0.0378, 0.0335, 0.5089, 0.5450, 0.1912, 0.3891, 0.1578], device='cuda:1'), in_proj_covar=tensor([0.0362, 0.0306, 0.0278, 0.0252, 0.0336, 0.0331, 0.0262, 0.0371], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:1') 2023-03-09 14:39:18,346 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.257e+02 1.806e+02 2.089e+02 2.584e+02 4.922e+02, threshold=4.178e+02, percent-clipped=1.0 2023-03-09 14:39:18,625 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117060.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:39:48,449 INFO [train2.py:809] (1/4) Epoch 30, batch 1550, loss[ctc_loss=0.04907, att_loss=0.2247, loss=0.1896, over 16401.00 frames. utt_duration=1493 frames, utt_pad_proportion=0.006926, over 44.00 utterances.], tot_loss[ctc_loss=0.06199, att_loss=0.2296, loss=0.1961, over 3271218.11 frames. utt_duration=1279 frames, utt_pad_proportion=0.04743, over 10239.75 utterances.], batch size: 44, lr: 3.58e-03, grad_scale: 16.0 2023-03-09 14:40:25,854 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117102.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:40:27,291 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4936, 5.1882, 5.2604, 5.3475, 5.1612, 5.2302, 4.9855, 4.7273], device='cuda:1'), covar=tensor([0.1832, 0.0805, 0.0384, 0.0532, 0.0619, 0.0484, 0.0467, 0.0427], device='cuda:1'), in_proj_covar=tensor([0.0548, 0.0390, 0.0388, 0.0393, 0.0455, 0.0462, 0.0391, 0.0426], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:1') 2023-03-09 14:40:49,905 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.8577, 2.5102, 2.7990, 2.8421, 3.0024, 3.0033, 2.4231, 3.1816], device='cuda:1'), covar=tensor([0.1426, 0.2055, 0.1455, 0.1042, 0.1543, 0.0837, 0.1706, 0.0950], device='cuda:1'), in_proj_covar=tensor([0.0156, 0.0153, 0.0150, 0.0147, 0.0162, 0.0140, 0.0164, 0.0140], 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') 2023-03-09 14:41:08,937 INFO [train2.py:809] (1/4) Epoch 30, batch 1600, loss[ctc_loss=0.04974, att_loss=0.2262, loss=0.1909, over 16398.00 frames. utt_duration=1493 frames, utt_pad_proportion=0.00691, over 44.00 utterances.], tot_loss[ctc_loss=0.06221, att_loss=0.23, loss=0.1964, over 3265292.21 frames. utt_duration=1250 frames, utt_pad_proportion=0.05848, over 10459.80 utterances.], batch size: 44, lr: 3.57e-03, grad_scale: 16.0 2023-03-09 14:41:22,328 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.70 vs. limit=5.0 2023-03-09 14:41:29,978 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-03-09 14:41:43,258 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=117150.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:41:56,337 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117158.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:41:59,198 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.078e+02 1.817e+02 2.106e+02 2.742e+02 6.475e+02, threshold=4.211e+02, percent-clipped=3.0 2023-03-09 14:42:00,965 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117161.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:42:29,624 INFO [train2.py:809] (1/4) Epoch 30, batch 1650, loss[ctc_loss=0.03176, att_loss=0.202, loss=0.168, over 16171.00 frames. utt_duration=1579 frames, utt_pad_proportion=0.006857, over 41.00 utterances.], tot_loss[ctc_loss=0.06217, att_loss=0.2299, loss=0.1964, over 3271336.54 frames. utt_duration=1247 frames, utt_pad_proportion=0.05638, over 10502.61 utterances.], batch size: 41, lr: 3.57e-03, grad_scale: 16.0 2023-03-09 14:43:14,722 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.0294, 3.7211, 3.7130, 3.2904, 3.7574, 3.8435, 3.8051, 2.8791], device='cuda:1'), covar=tensor([0.0863, 0.0947, 0.1628, 0.2694, 0.0898, 0.1485, 0.0747, 0.2600], device='cuda:1'), in_proj_covar=tensor([0.0218, 0.0223, 0.0236, 0.0290, 0.0197, 0.0299, 0.0219, 0.0242], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-03-09 14:43:19,240 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=117209.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:43:39,158 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=117221.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:43:51,228 INFO [train2.py:809] (1/4) Epoch 30, batch 1700, loss[ctc_loss=0.06728, att_loss=0.2405, loss=0.2059, over 16763.00 frames. utt_duration=1399 frames, utt_pad_proportion=0.005908, over 48.00 utterances.], tot_loss[ctc_loss=0.06241, att_loss=0.2305, loss=0.1969, over 3275098.42 frames. utt_duration=1251 frames, utt_pad_proportion=0.05519, over 10483.22 utterances.], batch size: 48, lr: 3.57e-03, grad_scale: 16.0 2023-03-09 14:44:43,555 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.065e+02 1.821e+02 2.138e+02 2.705e+02 3.860e+02, threshold=4.276e+02, percent-clipped=0.0 2023-03-09 14:45:04,992 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117274.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:45:12,573 INFO [train2.py:809] (1/4) Epoch 30, batch 1750, loss[ctc_loss=0.08744, att_loss=0.2488, loss=0.2165, over 16469.00 frames. utt_duration=1434 frames, utt_pad_proportion=0.00588, over 46.00 utterances.], tot_loss[ctc_loss=0.06247, att_loss=0.231, loss=0.1973, over 3277098.26 frames. utt_duration=1240 frames, utt_pad_proportion=0.05646, over 10582.72 utterances.], batch size: 46, lr: 3.57e-03, grad_scale: 8.0 2023-03-09 14:45:17,629 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=117282.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:45:33,953 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.55 vs. limit=5.0 2023-03-09 14:46:23,177 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=117322.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:46:29,941 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8396, 5.1490, 5.1036, 5.1669, 5.1585, 4.8413, 3.7290, 5.2008], device='cuda:1'), covar=tensor([0.0123, 0.0117, 0.0143, 0.0073, 0.0129, 0.0135, 0.0628, 0.0187], device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0094, 0.0120, 0.0074, 0.0081, 0.0092, 0.0107, 0.0113], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 14:46:34,461 INFO [train2.py:809] (1/4) Epoch 30, batch 1800, loss[ctc_loss=0.07066, att_loss=0.2172, loss=0.1879, over 15882.00 frames. utt_duration=1630 frames, utt_pad_proportion=0.008375, over 39.00 utterances.], tot_loss[ctc_loss=0.06231, att_loss=0.2311, loss=0.1973, over 3282211.27 frames. utt_duration=1248 frames, utt_pad_proportion=0.05308, over 10533.01 utterances.], batch size: 39, lr: 3.57e-03, grad_scale: 8.0 2023-03-09 14:46:34,860 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=117329.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:47:25,088 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117360.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:47:26,234 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.207e+02 1.761e+02 2.042e+02 2.458e+02 5.519e+02, threshold=4.084e+02, percent-clipped=2.0 2023-03-09 14:47:54,417 INFO [train2.py:809] (1/4) Epoch 30, batch 1850, loss[ctc_loss=0.05904, att_loss=0.2328, loss=0.198, over 16532.00 frames. utt_duration=1471 frames, utt_pad_proportion=0.006096, over 45.00 utterances.], tot_loss[ctc_loss=0.0621, att_loss=0.2304, loss=0.1967, over 3273706.92 frames. utt_duration=1260 frames, utt_pad_proportion=0.0509, over 10407.12 utterances.], batch size: 45, lr: 3.57e-03, grad_scale: 8.0 2023-03-09 14:48:12,231 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=117390.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:48:41,542 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=117408.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:48:41,891 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2786, 4.3424, 4.4186, 4.5058, 5.0353, 4.4240, 4.3508, 2.6728], device='cuda:1'), covar=tensor([0.0330, 0.0494, 0.0415, 0.0334, 0.0592, 0.0308, 0.0449, 0.1633], device='cuda:1'), in_proj_covar=tensor([0.0206, 0.0235, 0.0227, 0.0246, 0.0389, 0.0203, 0.0222, 0.0226], 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') 2023-03-09 14:49:09,949 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.1886, 2.7351, 3.0598, 4.2196, 3.7376, 3.7940, 2.9100, 2.2473], device='cuda:1'), covar=tensor([0.0869, 0.1979, 0.0955, 0.0499, 0.0955, 0.0519, 0.1477, 0.2120], device='cuda:1'), in_proj_covar=tensor([0.0193, 0.0223, 0.0188, 0.0232, 0.0243, 0.0196, 0.0208, 0.0195], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 14:49:14,297 INFO [train2.py:809] (1/4) Epoch 30, batch 1900, loss[ctc_loss=0.05574, att_loss=0.2219, loss=0.1886, over 16295.00 frames. utt_duration=1517 frames, utt_pad_proportion=0.006549, over 43.00 utterances.], tot_loss[ctc_loss=0.06221, att_loss=0.231, loss=0.1972, over 3279113.79 frames. utt_duration=1263 frames, utt_pad_proportion=0.04694, over 10400.41 utterances.], batch size: 43, lr: 3.57e-03, grad_scale: 8.0 2023-03-09 14:49:17,126 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-03-09 14:50:00,408 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117458.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:50:04,885 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.123e+02 1.786e+02 2.185e+02 2.601e+02 6.892e+02, threshold=4.370e+02, percent-clipped=2.0 2023-03-09 14:50:21,976 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-03-09 14:50:33,475 INFO [train2.py:809] (1/4) Epoch 30, batch 1950, loss[ctc_loss=0.05094, att_loss=0.211, loss=0.179, over 15359.00 frames. utt_duration=1757 frames, utt_pad_proportion=0.01184, over 35.00 utterances.], tot_loss[ctc_loss=0.06257, att_loss=0.2308, loss=0.1971, over 3274867.25 frames. utt_duration=1240 frames, utt_pad_proportion=0.05416, over 10578.15 utterances.], batch size: 35, lr: 3.57e-03, grad_scale: 8.0 2023-03-09 14:51:17,139 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=117506.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:51:53,201 INFO [train2.py:809] (1/4) Epoch 30, batch 2000, loss[ctc_loss=0.06715, att_loss=0.2181, loss=0.1879, over 14604.00 frames. utt_duration=1827 frames, utt_pad_proportion=0.03982, over 32.00 utterances.], tot_loss[ctc_loss=0.06242, att_loss=0.2305, loss=0.1969, over 3277702.08 frames. utt_duration=1268 frames, utt_pad_proportion=0.04696, over 10349.64 utterances.], batch size: 32, lr: 3.57e-03, grad_scale: 8.0 2023-03-09 14:52:29,677 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.4170, 2.4731, 4.8061, 3.7848, 2.9833, 4.0867, 4.5797, 4.5909], device='cuda:1'), covar=tensor([0.0325, 0.1576, 0.0261, 0.0888, 0.1695, 0.0323, 0.0238, 0.0311], device='cuda:1'), in_proj_covar=tensor([0.0247, 0.0252, 0.0240, 0.0330, 0.0277, 0.0252, 0.0234, 0.0254], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:1') 2023-03-09 14:52:43,007 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.243e+02 1.738e+02 2.282e+02 2.861e+02 7.444e+02, threshold=4.565e+02, percent-clipped=5.0 2023-03-09 14:53:08,580 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117577.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:53:11,454 INFO [train2.py:809] (1/4) Epoch 30, batch 2050, loss[ctc_loss=0.06005, att_loss=0.2257, loss=0.1926, over 15961.00 frames. utt_duration=1559 frames, utt_pad_proportion=0.006498, over 41.00 utterances.], tot_loss[ctc_loss=0.06338, att_loss=0.2311, loss=0.1976, over 3269159.93 frames. utt_duration=1224 frames, utt_pad_proportion=0.06088, over 10697.50 utterances.], batch size: 41, lr: 3.57e-03, grad_scale: 8.0 2023-03-09 14:54:32,200 INFO [train2.py:809] (1/4) Epoch 30, batch 2100, loss[ctc_loss=0.08289, att_loss=0.245, loss=0.2126, over 16697.00 frames. utt_duration=1453 frames, utt_pad_proportion=0.005888, over 46.00 utterances.], tot_loss[ctc_loss=0.06354, att_loss=0.2311, loss=0.1976, over 3272537.85 frames. utt_duration=1226 frames, utt_pad_proportion=0.06006, over 10692.16 utterances.], batch size: 46, lr: 3.57e-03, grad_scale: 8.0 2023-03-09 14:55:24,880 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.295e+02 1.716e+02 1.992e+02 2.490e+02 6.469e+02, threshold=3.983e+02, percent-clipped=1.0 2023-03-09 14:55:52,866 INFO [train2.py:809] (1/4) Epoch 30, batch 2150, loss[ctc_loss=0.0561, att_loss=0.2281, loss=0.1937, over 16624.00 frames. utt_duration=1416 frames, utt_pad_proportion=0.005409, over 47.00 utterances.], tot_loss[ctc_loss=0.06346, att_loss=0.2313, loss=0.1978, over 3273804.00 frames. utt_duration=1229 frames, utt_pad_proportion=0.05868, over 10669.68 utterances.], batch size: 47, lr: 3.57e-03, grad_scale: 8.0 2023-03-09 14:56:02,888 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117685.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 14:56:40,201 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.2494, 3.7316, 3.3241, 3.4484, 3.9970, 3.6460, 2.9295, 4.2459], device='cuda:1'), covar=tensor([0.0958, 0.0517, 0.1002, 0.0737, 0.0762, 0.0776, 0.0994, 0.0540], device='cuda:1'), in_proj_covar=tensor([0.0213, 0.0231, 0.0236, 0.0212, 0.0296, 0.0256, 0.0209, 0.0303], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:1') 2023-03-09 14:57:00,161 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9373, 4.2171, 4.4972, 4.5939, 2.6885, 4.2594, 2.9801, 1.9225], device='cuda:1'), covar=tensor([0.0596, 0.0398, 0.0637, 0.0312, 0.1671, 0.0364, 0.1365, 0.1719], device='cuda:1'), in_proj_covar=tensor([0.0223, 0.0196, 0.0265, 0.0184, 0.0224, 0.0175, 0.0234, 0.0205], 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') 2023-03-09 14:57:12,172 INFO [train2.py:809] (1/4) Epoch 30, batch 2200, loss[ctc_loss=0.077, att_loss=0.211, loss=0.1842, over 15642.00 frames. utt_duration=1692 frames, utt_pad_proportion=0.008502, over 37.00 utterances.], tot_loss[ctc_loss=0.06265, att_loss=0.2304, loss=0.1969, over 3276075.32 frames. utt_duration=1254 frames, utt_pad_proportion=0.05334, over 10464.99 utterances.], batch size: 37, lr: 3.57e-03, grad_scale: 8.0 2023-03-09 14:58:03,079 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.299e+02 1.825e+02 2.078e+02 2.480e+02 5.999e+02, threshold=4.155e+02, percent-clipped=3.0 2023-03-09 14:58:30,569 INFO [train2.py:809] (1/4) Epoch 30, batch 2250, loss[ctc_loss=0.06169, att_loss=0.2355, loss=0.2007, over 16890.00 frames. utt_duration=1380 frames, utt_pad_proportion=0.006288, over 49.00 utterances.], tot_loss[ctc_loss=0.06213, att_loss=0.2293, loss=0.1959, over 3262932.05 frames. utt_duration=1294 frames, utt_pad_proportion=0.04629, over 10094.27 utterances.], batch size: 49, lr: 3.56e-03, grad_scale: 8.0 2023-03-09 14:58:30,724 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9529, 6.1808, 5.6616, 5.8390, 5.8285, 5.2864, 5.6350, 5.4083], device='cuda:1'), covar=tensor([0.1222, 0.0883, 0.0958, 0.0922, 0.1038, 0.1725, 0.2715, 0.2425], device='cuda:1'), in_proj_covar=tensor([0.0566, 0.0644, 0.0496, 0.0482, 0.0460, 0.0489, 0.0650, 0.0550], 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') 2023-03-09 14:59:37,757 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-03-09 14:59:50,946 INFO [train2.py:809] (1/4) Epoch 30, batch 2300, loss[ctc_loss=0.06913, att_loss=0.2508, loss=0.2144, over 17080.00 frames. utt_duration=1315 frames, utt_pad_proportion=0.006623, over 52.00 utterances.], tot_loss[ctc_loss=0.06239, att_loss=0.2295, loss=0.1961, over 3268049.25 frames. utt_duration=1277 frames, utt_pad_proportion=0.04956, over 10248.62 utterances.], batch size: 52, lr: 3.56e-03, grad_scale: 8.0 2023-03-09 15:00:42,226 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.175e+02 1.803e+02 2.047e+02 2.475e+02 6.158e+02, threshold=4.093e+02, percent-clipped=3.0 2023-03-09 15:00:46,944 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.2772, 5.5573, 5.8608, 5.6452, 5.8664, 6.1994, 5.3819, 6.3324], device='cuda:1'), covar=tensor([0.0659, 0.0706, 0.0752, 0.1341, 0.1629, 0.1023, 0.0689, 0.0546], device='cuda:1'), in_proj_covar=tensor([0.0947, 0.0549, 0.0660, 0.0700, 0.0930, 0.0687, 0.0526, 0.0662], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 15:01:07,123 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117877.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 15:01:09,865 INFO [train2.py:809] (1/4) Epoch 30, batch 2350, loss[ctc_loss=0.06146, att_loss=0.239, loss=0.2035, over 16725.00 frames. utt_duration=684 frames, utt_pad_proportion=0.1429, over 98.00 utterances.], tot_loss[ctc_loss=0.06305, att_loss=0.2305, loss=0.197, over 3272366.64 frames. utt_duration=1275 frames, utt_pad_proportion=0.04648, over 10275.74 utterances.], batch size: 98, lr: 3.56e-03, grad_scale: 8.0 2023-03-09 15:01:54,304 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.8224, 2.4161, 2.7181, 2.7829, 3.0639, 2.9114, 2.3102, 3.0459], device='cuda:1'), covar=tensor([0.1837, 0.2206, 0.1608, 0.1244, 0.1488, 0.1088, 0.2021, 0.1177], device='cuda:1'), in_proj_covar=tensor([0.0156, 0.0153, 0.0150, 0.0147, 0.0162, 0.0141, 0.0162, 0.0141], 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') 2023-03-09 15:02:12,209 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-03-09 15:02:23,390 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=117925.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 15:02:29,491 INFO [train2.py:809] (1/4) Epoch 30, batch 2400, loss[ctc_loss=0.06848, att_loss=0.2389, loss=0.2048, over 16881.00 frames. utt_duration=1380 frames, utt_pad_proportion=0.007503, over 49.00 utterances.], tot_loss[ctc_loss=0.06307, att_loss=0.2301, loss=0.1967, over 3266231.89 frames. utt_duration=1269 frames, utt_pad_proportion=0.04959, over 10311.01 utterances.], batch size: 49, lr: 3.56e-03, grad_scale: 8.0 2023-03-09 15:03:20,931 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.310e+02 1.808e+02 2.132e+02 2.494e+02 4.997e+02, threshold=4.263e+02, percent-clipped=2.0 2023-03-09 15:03:48,710 INFO [train2.py:809] (1/4) Epoch 30, batch 2450, loss[ctc_loss=0.06374, att_loss=0.2398, loss=0.2046, over 17447.00 frames. utt_duration=1013 frames, utt_pad_proportion=0.04442, over 69.00 utterances.], tot_loss[ctc_loss=0.06289, att_loss=0.2302, loss=0.1967, over 3272705.77 frames. utt_duration=1260 frames, utt_pad_proportion=0.0502, over 10403.33 utterances.], batch size: 69, lr: 3.56e-03, grad_scale: 8.0 2023-03-09 15:03:59,595 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117985.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 15:05:01,289 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.2562, 5.5057, 5.4192, 5.4815, 5.5687, 5.4929, 5.1468, 4.9824], device='cuda:1'), covar=tensor([0.0934, 0.0515, 0.0356, 0.0430, 0.0232, 0.0302, 0.0421, 0.0301], device='cuda:1'), in_proj_covar=tensor([0.0541, 0.0383, 0.0382, 0.0385, 0.0446, 0.0453, 0.0386, 0.0421], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:1') 2023-03-09 15:05:13,630 INFO [train2.py:809] (1/4) Epoch 30, batch 2500, loss[ctc_loss=0.1235, att_loss=0.2635, loss=0.2355, over 14338.00 frames. utt_duration=397 frames, utt_pad_proportion=0.3096, over 145.00 utterances.], tot_loss[ctc_loss=0.06226, att_loss=0.2297, loss=0.1962, over 3270186.14 frames. utt_duration=1227 frames, utt_pad_proportion=0.05907, over 10678.09 utterances.], batch size: 145, lr: 3.56e-03, grad_scale: 8.0 2023-03-09 15:05:21,116 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=118033.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 15:05:32,770 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118040.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 15:06:07,288 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.222e+02 1.938e+02 2.187e+02 2.706e+02 8.579e+02, threshold=4.374e+02, percent-clipped=4.0 2023-03-09 15:06:35,076 INFO [train2.py:809] (1/4) Epoch 30, batch 2550, loss[ctc_loss=0.04621, att_loss=0.2132, loss=0.1798, over 16271.00 frames. utt_duration=1515 frames, utt_pad_proportion=0.007752, over 43.00 utterances.], tot_loss[ctc_loss=0.06244, att_loss=0.2296, loss=0.1962, over 3265773.74 frames. utt_duration=1209 frames, utt_pad_proportion=0.06567, over 10815.37 utterances.], batch size: 43, lr: 3.56e-03, grad_scale: 4.0 2023-03-09 15:07:11,140 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118101.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 15:07:54,646 INFO [train2.py:809] (1/4) Epoch 30, batch 2600, loss[ctc_loss=0.06562, att_loss=0.2287, loss=0.1961, over 16778.00 frames. utt_duration=1400 frames, utt_pad_proportion=0.005948, over 48.00 utterances.], tot_loss[ctc_loss=0.0617, att_loss=0.229, loss=0.1955, over 3268772.26 frames. utt_duration=1231 frames, utt_pad_proportion=0.0603, over 10632.23 utterances.], batch size: 48, lr: 3.56e-03, grad_scale: 4.0 2023-03-09 15:08:46,943 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.196e+02 1.796e+02 2.196e+02 2.749e+02 4.358e+02, threshold=4.391e+02, percent-clipped=0.0 2023-03-09 15:08:56,489 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118168.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 15:09:14,092 INFO [train2.py:809] (1/4) Epoch 30, batch 2650, loss[ctc_loss=0.04724, att_loss=0.2227, loss=0.1876, over 16125.00 frames. utt_duration=1537 frames, utt_pad_proportion=0.005852, over 42.00 utterances.], tot_loss[ctc_loss=0.06128, att_loss=0.2286, loss=0.1951, over 3267505.05 frames. utt_duration=1255 frames, utt_pad_proportion=0.05369, over 10423.97 utterances.], batch size: 42, lr: 3.56e-03, grad_scale: 4.0 2023-03-09 15:09:43,372 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.3924, 5.4336, 5.2892, 3.5181, 5.2562, 5.0461, 4.9073, 3.3757], device='cuda:1'), covar=tensor([0.0094, 0.0082, 0.0197, 0.0870, 0.0087, 0.0169, 0.0218, 0.1097], device='cuda:1'), in_proj_covar=tensor([0.0080, 0.0108, 0.0113, 0.0114, 0.0091, 0.0120, 0.0103, 0.0105], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:1') 2023-03-09 15:09:43,566 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.3541, 5.2047, 4.8361, 2.8604, 2.5761, 3.6371, 3.1298, 4.1592], device='cuda:1'), covar=tensor([0.0624, 0.0358, 0.0358, 0.4735, 0.4497, 0.1784, 0.2982, 0.1420], device='cuda:1'), in_proj_covar=tensor([0.0369, 0.0313, 0.0284, 0.0257, 0.0344, 0.0338, 0.0268, 0.0377], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-09 15:10:33,465 INFO [train2.py:809] (1/4) Epoch 30, batch 2700, loss[ctc_loss=0.05144, att_loss=0.2121, loss=0.18, over 15655.00 frames. utt_duration=1694 frames, utt_pad_proportion=0.007742, over 37.00 utterances.], tot_loss[ctc_loss=0.06214, att_loss=0.2291, loss=0.1957, over 3261695.84 frames. utt_duration=1255 frames, utt_pad_proportion=0.05494, over 10411.65 utterances.], batch size: 37, lr: 3.56e-03, grad_scale: 4.0 2023-03-09 15:10:33,855 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118229.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 15:10:39,349 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.70 vs. limit=2.0 2023-03-09 15:11:25,934 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.097e+02 1.836e+02 2.266e+02 2.685e+02 4.437e+02, threshold=4.533e+02, percent-clipped=1.0 2023-03-09 15:11:37,233 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118269.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 15:11:52,828 INFO [train2.py:809] (1/4) Epoch 30, batch 2750, loss[ctc_loss=0.04115, att_loss=0.214, loss=0.1794, over 16118.00 frames. utt_duration=1536 frames, utt_pad_proportion=0.006849, over 42.00 utterances.], tot_loss[ctc_loss=0.06189, att_loss=0.2285, loss=0.1951, over 3261498.55 frames. utt_duration=1283 frames, utt_pad_proportion=0.04935, over 10176.71 utterances.], batch size: 42, lr: 3.56e-03, grad_scale: 4.0 2023-03-09 15:12:10,161 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118289.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 15:12:23,931 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5738, 2.5973, 3.6376, 2.6955, 3.5482, 4.7391, 4.5856, 3.1673], device='cuda:1'), covar=tensor([0.0488, 0.2217, 0.1173, 0.1745, 0.0995, 0.0869, 0.0538, 0.1532], device='cuda:1'), in_proj_covar=tensor([0.0253, 0.0252, 0.0294, 0.0220, 0.0275, 0.0386, 0.0278, 0.0239], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-09 15:12:26,646 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-03-09 15:12:28,663 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5080, 4.9154, 4.7640, 4.8633, 4.8935, 4.6380, 3.3862, 4.7792], device='cuda:1'), covar=tensor([0.0150, 0.0159, 0.0158, 0.0104, 0.0152, 0.0142, 0.0756, 0.0275], device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0094, 0.0121, 0.0075, 0.0082, 0.0093, 0.0108, 0.0113], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 15:13:12,412 INFO [train2.py:809] (1/4) Epoch 30, batch 2800, loss[ctc_loss=0.06789, att_loss=0.2397, loss=0.2053, over 17285.00 frames. utt_duration=876.6 frames, utt_pad_proportion=0.08205, over 79.00 utterances.], tot_loss[ctc_loss=0.06261, att_loss=0.2298, loss=0.1963, over 3274420.81 frames. utt_duration=1265 frames, utt_pad_proportion=0.05071, over 10367.12 utterances.], batch size: 79, lr: 3.56e-03, grad_scale: 8.0 2023-03-09 15:13:12,756 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118329.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 15:13:15,017 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118330.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 15:13:47,152 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118350.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 15:14:05,715 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.224e+02 1.819e+02 2.314e+02 2.830e+02 4.183e+02, threshold=4.629e+02, percent-clipped=0.0 2023-03-09 15:14:32,832 INFO [train2.py:809] (1/4) Epoch 30, batch 2850, loss[ctc_loss=0.04638, att_loss=0.2056, loss=0.1738, over 15635.00 frames. utt_duration=1692 frames, utt_pad_proportion=0.008835, over 37.00 utterances.], tot_loss[ctc_loss=0.06184, att_loss=0.2291, loss=0.1956, over 3274383.38 frames. utt_duration=1285 frames, utt_pad_proportion=0.04562, over 10207.47 utterances.], batch size: 37, lr: 3.56e-03, grad_scale: 8.0 2023-03-09 15:14:43,878 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.9608, 5.2232, 5.1997, 5.1874, 5.2400, 5.2150, 4.8091, 4.6846], device='cuda:1'), covar=tensor([0.1044, 0.0584, 0.0326, 0.0555, 0.0307, 0.0361, 0.0468, 0.0344], device='cuda:1'), in_proj_covar=tensor([0.0542, 0.0384, 0.0383, 0.0386, 0.0449, 0.0455, 0.0386, 0.0423], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:1') 2023-03-09 15:14:51,897 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118390.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 15:15:01,129 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118396.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 15:15:53,108 INFO [train2.py:809] (1/4) Epoch 30, batch 2900, loss[ctc_loss=0.09169, att_loss=0.2552, loss=0.2225, over 14040.00 frames. utt_duration=383.5 frames, utt_pad_proportion=0.326, over 147.00 utterances.], tot_loss[ctc_loss=0.06174, att_loss=0.2294, loss=0.1959, over 3267926.90 frames. utt_duration=1262 frames, utt_pad_proportion=0.05299, over 10367.18 utterances.], batch size: 147, lr: 3.55e-03, grad_scale: 8.0 2023-03-09 15:16:01,993 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.1011, 3.8130, 3.3097, 3.4733, 4.0080, 3.6963, 3.2379, 4.2401], device='cuda:1'), covar=tensor([0.1149, 0.0564, 0.1124, 0.0820, 0.0777, 0.0848, 0.0824, 0.0564], device='cuda:1'), in_proj_covar=tensor([0.0214, 0.0233, 0.0236, 0.0213, 0.0296, 0.0256, 0.0209, 0.0304], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0005], device='cuda:1') 2023-03-09 15:16:46,256 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.238e+02 1.677e+02 2.045e+02 2.519e+02 5.925e+02, threshold=4.089e+02, percent-clipped=1.0 2023-03-09 15:17:12,896 INFO [train2.py:809] (1/4) Epoch 30, batch 2950, loss[ctc_loss=0.04971, att_loss=0.2074, loss=0.1759, over 15929.00 frames. utt_duration=1556 frames, utt_pad_proportion=0.007964, over 41.00 utterances.], tot_loss[ctc_loss=0.06133, att_loss=0.2292, loss=0.1957, over 3271066.19 frames. utt_duration=1276 frames, utt_pad_proportion=0.04973, over 10266.37 utterances.], batch size: 41, lr: 3.55e-03, grad_scale: 8.0 2023-03-09 15:17:28,475 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.5521, 3.0201, 3.4022, 4.6522, 4.1869, 4.2812, 3.2856, 2.4641], device='cuda:1'), covar=tensor([0.0709, 0.2013, 0.0999, 0.0466, 0.0770, 0.0389, 0.1233, 0.2065], device='cuda:1'), in_proj_covar=tensor([0.0192, 0.0222, 0.0187, 0.0230, 0.0241, 0.0196, 0.0205, 0.0193], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 15:18:23,848 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118524.0, num_to_drop=1, layers_to_drop={1} 2023-03-09 15:18:31,907 INFO [train2.py:809] (1/4) Epoch 30, batch 3000, loss[ctc_loss=0.04275, att_loss=0.2017, loss=0.1699, over 12359.00 frames. utt_duration=1832 frames, utt_pad_proportion=0.1279, over 27.00 utterances.], tot_loss[ctc_loss=0.06172, att_loss=0.2296, loss=0.196, over 3265236.53 frames. utt_duration=1268 frames, utt_pad_proportion=0.05225, over 10314.60 utterances.], batch size: 27, lr: 3.55e-03, grad_scale: 8.0 2023-03-09 15:18:31,907 INFO [train2.py:834] (1/4) Computing validation loss 2023-03-09 15:18:46,180 INFO [train2.py:843] (1/4) Epoch 30, validation: ctc_loss=0.04128, att_loss=0.2346, loss=0.196, over 944034.00 frames. utt_duration=679.8 frames, utt_pad_proportion=0.1349, over 5567.00 utterances. 2023-03-09 15:18:46,181 INFO [train2.py:844] (1/4) Maximum memory allocated so far is 16129MB 2023-03-09 15:18:52,047 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-03-09 15:18:58,550 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=3.80 vs. limit=5.0 2023-03-09 15:19:09,437 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.2340, 5.5065, 5.4447, 5.4300, 5.5145, 5.4731, 5.1007, 4.8991], device='cuda:1'), covar=tensor([0.0935, 0.0481, 0.0286, 0.0477, 0.0278, 0.0298, 0.0429, 0.0342], device='cuda:1'), in_proj_covar=tensor([0.0538, 0.0383, 0.0382, 0.0385, 0.0448, 0.0454, 0.0384, 0.0421], device='cuda:1'), out_proj_covar=tensor([0.0005, 0.0004, 0.0004, 0.0004, 0.0005, 0.0005, 0.0004, 0.0004], device='cuda:1') 2023-03-09 15:19:15,761 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.6475, 5.9147, 5.3646, 5.5676, 5.5356, 5.0281, 5.3402, 5.0725], device='cuda:1'), covar=tensor([0.1381, 0.0893, 0.1059, 0.0985, 0.1132, 0.1655, 0.2449, 0.2427], device='cuda:1'), in_proj_covar=tensor([0.0569, 0.0645, 0.0494, 0.0484, 0.0460, 0.0486, 0.0649, 0.0551], 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') 2023-03-09 15:19:38,944 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.222e+02 1.708e+02 2.016e+02 2.374e+02 4.392e+02, threshold=4.031e+02, percent-clipped=1.0 2023-03-09 15:20:05,705 INFO [train2.py:809] (1/4) Epoch 30, batch 3050, loss[ctc_loss=0.05468, att_loss=0.2362, loss=0.1999, over 17487.00 frames. utt_duration=887.1 frames, utt_pad_proportion=0.07115, over 79.00 utterances.], tot_loss[ctc_loss=0.06125, att_loss=0.2296, loss=0.1959, over 3270701.10 frames. utt_duration=1272 frames, utt_pad_proportion=0.04783, over 10294.07 utterances.], batch size: 79, lr: 3.55e-03, grad_scale: 8.0 2023-03-09 15:20:40,617 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0554, 4.3048, 4.4032, 4.6056, 2.7920, 4.4288, 3.0624, 1.7834], device='cuda:1'), covar=tensor([0.0633, 0.0382, 0.0608, 0.0296, 0.1592, 0.0263, 0.1256, 0.1803], device='cuda:1'), in_proj_covar=tensor([0.0226, 0.0198, 0.0267, 0.0186, 0.0225, 0.0176, 0.0235, 0.0206], 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') 2023-03-09 15:21:20,123 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118625.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 15:21:26,774 INFO [train2.py:809] (1/4) Epoch 30, batch 3100, loss[ctc_loss=0.04414, att_loss=0.2071, loss=0.1745, over 15760.00 frames. utt_duration=1660 frames, utt_pad_proportion=0.008096, over 38.00 utterances.], tot_loss[ctc_loss=0.06135, att_loss=0.2295, loss=0.1959, over 3270638.68 frames. utt_duration=1272 frames, utt_pad_proportion=0.04729, over 10299.69 utterances.], batch size: 38, lr: 3.55e-03, grad_scale: 8.0 2023-03-09 15:21:30,281 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118631.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 15:21:30,307 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.7369, 2.4915, 2.5306, 2.6753, 2.9899, 2.8261, 2.3198, 2.9988], device='cuda:1'), covar=tensor([0.1684, 0.2102, 0.1770, 0.1322, 0.1889, 0.1106, 0.2034, 0.1529], device='cuda:1'), in_proj_covar=tensor([0.0154, 0.0151, 0.0148, 0.0147, 0.0161, 0.0139, 0.0160, 0.0140], 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') 2023-03-09 15:21:53,098 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118645.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 15:22:19,882 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.328e+02 1.896e+02 2.204e+02 2.593e+02 6.390e+02, threshold=4.408e+02, percent-clipped=1.0 2023-03-09 15:22:47,064 INFO [train2.py:809] (1/4) Epoch 30, batch 3150, loss[ctc_loss=0.0577, att_loss=0.2414, loss=0.2047, over 16958.00 frames. utt_duration=1358 frames, utt_pad_proportion=0.00821, over 50.00 utterances.], tot_loss[ctc_loss=0.06105, att_loss=0.2296, loss=0.1959, over 3270881.84 frames. utt_duration=1270 frames, utt_pad_proportion=0.04887, over 10311.11 utterances.], batch size: 50, lr: 3.55e-03, grad_scale: 8.0 2023-03-09 15:22:56,708 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118685.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 15:23:08,829 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118692.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 15:23:14,668 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118696.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 15:23:36,829 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.8867, 5.1832, 4.8050, 5.2579, 4.6649, 4.9242, 5.3295, 5.1122], device='cuda:1'), covar=tensor([0.0630, 0.0328, 0.0789, 0.0346, 0.0422, 0.0287, 0.0230, 0.0205], device='cuda:1'), in_proj_covar=tensor([0.0408, 0.0344, 0.0384, 0.0388, 0.0342, 0.0248, 0.0324, 0.0307], device='cuda:1'), out_proj_covar=tensor([0.0006, 0.0006, 0.0006, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005], device='cuda:1') 2023-03-09 15:23:50,252 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.30 vs. limit=5.0 2023-03-09 15:24:06,682 INFO [train2.py:809] (1/4) Epoch 30, batch 3200, loss[ctc_loss=0.07212, att_loss=0.2319, loss=0.1999, over 16121.00 frames. utt_duration=1537 frames, utt_pad_proportion=0.006633, over 42.00 utterances.], tot_loss[ctc_loss=0.0611, att_loss=0.2292, loss=0.1956, over 3268151.80 frames. utt_duration=1279 frames, utt_pad_proportion=0.04928, over 10235.61 utterances.], batch size: 42, lr: 3.55e-03, grad_scale: 8.0 2023-03-09 15:24:26,572 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6395, 3.1461, 3.7395, 3.2069, 3.6869, 4.6930, 4.5676, 3.4819], device='cuda:1'), covar=tensor([0.0316, 0.1609, 0.1242, 0.1257, 0.1024, 0.0857, 0.0549, 0.1108], device='cuda:1'), in_proj_covar=tensor([0.0250, 0.0250, 0.0292, 0.0220, 0.0273, 0.0384, 0.0276, 0.0237], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003], device='cuda:1') 2023-03-09 15:24:28,200 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.6249, 2.8591, 5.0978, 4.1868, 3.1383, 4.4174, 4.9275, 4.8284], device='cuda:1'), covar=tensor([0.0315, 0.1485, 0.0255, 0.0807, 0.1608, 0.0259, 0.0220, 0.0257], device='cuda:1'), in_proj_covar=tensor([0.0244, 0.0248, 0.0237, 0.0328, 0.0274, 0.0249, 0.0232, 0.0251], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 15:24:30,974 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=118744.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 15:24:59,446 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.119e+02 1.732e+02 2.054e+02 2.597e+02 6.136e+02, threshold=4.108e+02, percent-clipped=5.0 2023-03-09 15:25:14,403 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-03-09 15:25:26,532 INFO [train2.py:809] (1/4) Epoch 30, batch 3250, loss[ctc_loss=0.06243, att_loss=0.2308, loss=0.1971, over 16287.00 frames. utt_duration=1517 frames, utt_pad_proportion=0.006716, over 43.00 utterances.], tot_loss[ctc_loss=0.06141, att_loss=0.2294, loss=0.1958, over 3266510.71 frames. utt_duration=1258 frames, utt_pad_proportion=0.05476, over 10399.75 utterances.], batch size: 43, lr: 3.55e-03, grad_scale: 8.0 2023-03-09 15:25:41,909 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-03-09 15:26:33,174 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=2.91 vs. limit=5.0 2023-03-09 15:26:39,457 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118824.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 15:26:46,841 INFO [train2.py:809] (1/4) Epoch 30, batch 3300, loss[ctc_loss=0.05953, att_loss=0.242, loss=0.2055, over 16962.00 frames. utt_duration=1358 frames, utt_pad_proportion=0.007904, over 50.00 utterances.], tot_loss[ctc_loss=0.06152, att_loss=0.2296, loss=0.196, over 3265987.18 frames. utt_duration=1250 frames, utt_pad_proportion=0.05633, over 10460.63 utterances.], batch size: 50, lr: 3.55e-03, grad_scale: 8.0 2023-03-09 15:27:39,412 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.286e+02 1.786e+02 1.999e+02 2.429e+02 6.941e+02, threshold=3.998e+02, percent-clipped=3.0 2023-03-09 15:27:55,369 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=118872.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 15:28:06,047 INFO [train2.py:809] (1/4) Epoch 30, batch 3350, loss[ctc_loss=0.06724, att_loss=0.2076, loss=0.1796, over 11915.00 frames. utt_duration=1835 frames, utt_pad_proportion=0.1596, over 26.00 utterances.], tot_loss[ctc_loss=0.06145, att_loss=0.23, loss=0.1963, over 3270866.46 frames. utt_duration=1258 frames, utt_pad_proportion=0.05167, over 10413.41 utterances.], batch size: 26, lr: 3.55e-03, grad_scale: 8.0 2023-03-09 15:28:34,634 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.5074, 4.8602, 4.7508, 4.8536, 4.9728, 4.6161, 3.2752, 4.8525], device='cuda:1'), covar=tensor([0.0139, 0.0146, 0.0167, 0.0103, 0.0107, 0.0133, 0.0807, 0.0228], device='cuda:1'), in_proj_covar=tensor([0.0098, 0.0094, 0.0121, 0.0075, 0.0081, 0.0092, 0.0108, 0.0112], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 15:28:55,147 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2108, 4.3768, 4.4602, 4.5495, 5.0634, 4.3968, 4.4322, 2.7706], device='cuda:1'), covar=tensor([0.0398, 0.0495, 0.0464, 0.0443, 0.0660, 0.0301, 0.0421, 0.1612], device='cuda:1'), in_proj_covar=tensor([0.0208, 0.0237, 0.0231, 0.0248, 0.0390, 0.0204, 0.0223, 0.0225], 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') 2023-03-09 15:29:20,614 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118925.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 15:29:26,687 INFO [train2.py:809] (1/4) Epoch 30, batch 3400, loss[ctc_loss=0.06334, att_loss=0.2376, loss=0.2028, over 17395.00 frames. utt_duration=1106 frames, utt_pad_proportion=0.03184, over 63.00 utterances.], tot_loss[ctc_loss=0.06122, att_loss=0.2299, loss=0.1961, over 3271555.98 frames. utt_duration=1268 frames, utt_pad_proportion=0.04926, over 10335.30 utterances.], batch size: 63, lr: 3.55e-03, grad_scale: 8.0 2023-03-09 15:29:52,892 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118945.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 15:29:54,900 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.7657, 6.0134, 5.5254, 5.7059, 5.6616, 5.2106, 5.4656, 5.1948], device='cuda:1'), covar=tensor([0.1266, 0.0902, 0.0895, 0.0818, 0.1119, 0.1432, 0.2447, 0.2325], device='cuda:1'), in_proj_covar=tensor([0.0561, 0.0641, 0.0492, 0.0479, 0.0457, 0.0480, 0.0646, 0.0547], 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') 2023-03-09 15:30:19,284 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.021e+02 1.939e+02 2.194e+02 2.759e+02 6.591e+02, threshold=4.388e+02, percent-clipped=2.0 2023-03-09 15:30:21,673 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-03-09 15:30:37,148 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=118973.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 15:30:46,800 INFO [train2.py:809] (1/4) Epoch 30, batch 3450, loss[ctc_loss=0.05254, att_loss=0.2134, loss=0.1813, over 15353.00 frames. utt_duration=1756 frames, utt_pad_proportion=0.01234, over 35.00 utterances.], tot_loss[ctc_loss=0.06109, att_loss=0.2297, loss=0.196, over 3267761.20 frames. utt_duration=1270 frames, utt_pad_proportion=0.04919, over 10307.89 utterances.], batch size: 35, lr: 3.55e-03, grad_scale: 8.0 2023-03-09 15:30:57,738 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118985.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 15:31:00,750 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118987.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 15:31:09,772 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=118993.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 15:32:06,394 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.9524, 4.8812, 4.6849, 2.2212, 2.0365, 2.9997, 2.1079, 3.8760], device='cuda:1'), covar=tensor([0.0798, 0.0369, 0.0344, 0.5197, 0.5279, 0.2353, 0.4347, 0.1512], device='cuda:1'), in_proj_covar=tensor([0.0366, 0.0312, 0.0283, 0.0256, 0.0342, 0.0336, 0.0268, 0.0375], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-09 15:32:07,440 INFO [train2.py:809] (1/4) Epoch 30, batch 3500, loss[ctc_loss=0.06469, att_loss=0.2418, loss=0.2064, over 17152.00 frames. utt_duration=1227 frames, utt_pad_proportion=0.01316, over 56.00 utterances.], tot_loss[ctc_loss=0.06172, att_loss=0.2305, loss=0.1967, over 3279972.69 frames. utt_duration=1271 frames, utt_pad_proportion=0.04625, over 10335.34 utterances.], batch size: 56, lr: 3.55e-03, grad_scale: 8.0 2023-03-09 15:32:15,052 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=119033.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 15:32:48,176 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0688, 5.0727, 4.8810, 2.5728, 2.0918, 3.1723, 2.3042, 3.9362], device='cuda:1'), covar=tensor([0.0735, 0.0380, 0.0320, 0.4558, 0.5320, 0.2166, 0.4090, 0.1620], device='cuda:1'), in_proj_covar=tensor([0.0366, 0.0312, 0.0283, 0.0256, 0.0342, 0.0336, 0.0267, 0.0375], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002], device='cuda:1') 2023-03-09 15:32:59,920 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.231e+02 1.837e+02 2.167e+02 2.597e+02 5.563e+02, threshold=4.335e+02, percent-clipped=3.0 2023-03-09 15:33:20,406 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119075.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 15:33:27,091 INFO [train2.py:809] (1/4) Epoch 30, batch 3550, loss[ctc_loss=0.07295, att_loss=0.2154, loss=0.1869, over 14491.00 frames. utt_duration=1813 frames, utt_pad_proportion=0.04627, over 32.00 utterances.], tot_loss[ctc_loss=0.06216, att_loss=0.2306, loss=0.1969, over 3281445.67 frames. utt_duration=1259 frames, utt_pad_proportion=0.04921, over 10437.60 utterances.], batch size: 32, lr: 3.54e-03, grad_scale: 8.0 2023-03-09 15:33:50,479 INFO [zipformer.py:625] (1/4) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119093.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 15:34:46,590 INFO [train2.py:809] (1/4) Epoch 30, batch 3600, loss[ctc_loss=0.06156, att_loss=0.2173, loss=0.1862, over 16275.00 frames. utt_duration=1516 frames, utt_pad_proportion=0.007356, over 43.00 utterances.], tot_loss[ctc_loss=0.06231, att_loss=0.2302, loss=0.1966, over 3277885.72 frames. utt_duration=1242 frames, utt_pad_proportion=0.05465, over 10573.09 utterances.], batch size: 43, lr: 3.54e-03, grad_scale: 8.0 2023-03-09 15:34:51,489 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.0883, 4.2362, 4.3962, 4.6498, 2.8988, 4.2560, 3.0239, 1.7181], device='cuda:1'), covar=tensor([0.0612, 0.0357, 0.0627, 0.0278, 0.1507, 0.0309, 0.1287, 0.1755], device='cuda:1'), in_proj_covar=tensor([0.0226, 0.0199, 0.0266, 0.0187, 0.0226, 0.0178, 0.0236, 0.0206], 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') 2023-03-09 15:34:57,977 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119136.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 15:35:02,994 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-03-09 15:35:26,387 INFO [zipformer.py:625] (1/4) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119154.0, num_to_drop=1, layers_to_drop={2} 2023-03-09 15:35:38,264 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.140e+02 1.787e+02 2.179e+02 2.834e+02 1.118e+03, threshold=4.357e+02, percent-clipped=5.0 2023-03-09 15:36:05,602 INFO [train2.py:809] (1/4) Epoch 30, batch 3650, loss[ctc_loss=0.06164, att_loss=0.2129, loss=0.1827, over 15780.00 frames. utt_duration=1663 frames, utt_pad_proportion=0.007431, over 38.00 utterances.], tot_loss[ctc_loss=0.06251, att_loss=0.2302, loss=0.1966, over 3266570.59 frames. utt_duration=1239 frames, utt_pad_proportion=0.05691, over 10558.69 utterances.], batch size: 38, lr: 3.54e-03, grad_scale: 8.0 2023-03-09 15:37:24,891 INFO [train2.py:809] (1/4) Epoch 30, batch 3700, loss[ctc_loss=0.07924, att_loss=0.2397, loss=0.2076, over 16965.00 frames. utt_duration=1359 frames, utt_pad_proportion=0.007538, over 50.00 utterances.], tot_loss[ctc_loss=0.06227, att_loss=0.23, loss=0.1964, over 3270208.20 frames. utt_duration=1239 frames, utt_pad_proportion=0.05721, over 10573.98 utterances.], batch size: 50, lr: 3.54e-03, grad_scale: 8.0 2023-03-09 15:38:13,212 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([2.9922, 3.7127, 3.7112, 3.2932, 3.6342, 3.7579, 3.7623, 2.8524], device='cuda:1'), covar=tensor([0.1060, 0.1223, 0.1770, 0.2587, 0.2003, 0.1831, 0.0826, 0.2798], device='cuda:1'), in_proj_covar=tensor([0.0216, 0.0221, 0.0234, 0.0283, 0.0198, 0.0295, 0.0218, 0.0237], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:1') 2023-03-09 15:38:18,046 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 9.417e+01 1.774e+02 2.178e+02 2.630e+02 5.010e+02, threshold=4.357e+02, percent-clipped=3.0 2023-03-09 15:38:32,625 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.0258, 2.6260, 2.9531, 4.2350, 3.7704, 3.7951, 2.7598, 1.9831], device='cuda:1'), covar=tensor([0.0963, 0.2086, 0.1080, 0.0553, 0.0914, 0.0557, 0.1659, 0.2507], device='cuda:1'), in_proj_covar=tensor([0.0194, 0.0223, 0.0188, 0.0232, 0.0243, 0.0198, 0.0206, 0.0195], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 15:38:34,076 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([6.0913, 5.3833, 5.7010, 5.5090, 5.6002, 6.0261, 5.3518, 6.1490], device='cuda:1'), covar=tensor([0.0779, 0.0808, 0.0764, 0.1318, 0.1889, 0.1060, 0.0741, 0.0684], device='cuda:1'), in_proj_covar=tensor([0.0926, 0.0540, 0.0652, 0.0689, 0.0915, 0.0678, 0.0516, 0.0654], device='cuda:1'), out_proj_covar=tensor([0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:1') 2023-03-09 15:38:45,450 INFO [train2.py:809] (1/4) Epoch 30, batch 3750, loss[ctc_loss=0.07251, att_loss=0.2359, loss=0.2032, over 13666.00 frames. utt_duration=375.8 frames, utt_pad_proportion=0.343, over 146.00 utterances.], tot_loss[ctc_loss=0.06238, att_loss=0.2302, loss=0.1967, over 3263465.88 frames. utt_duration=1199 frames, utt_pad_proportion=0.06603, over 10902.01 utterances.], batch size: 146, lr: 3.54e-03, grad_scale: 8.0 2023-03-09 15:38:50,205 INFO [scaling.py:679] (1/4) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-03-09 15:38:59,220 INFO [zipformer.py:625] (1/4) warmup_begin=2000.0, warmup_end=2666.7, batch_count=119287.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 15:40:05,288 INFO [train2.py:809] (1/4) Epoch 30, batch 3800, loss[ctc_loss=0.07349, att_loss=0.2321, loss=0.2003, over 17349.00 frames. utt_duration=1178 frames, utt_pad_proportion=0.02106, over 59.00 utterances.], tot_loss[ctc_loss=0.06191, att_loss=0.2296, loss=0.1961, over 3262828.82 frames. utt_duration=1214 frames, utt_pad_proportion=0.06402, over 10766.84 utterances.], batch size: 59, lr: 3.54e-03, grad_scale: 8.0 2023-03-09 15:40:15,893 INFO [zipformer.py:625] (1/4) warmup_begin=666.7, warmup_end=1333.3, batch_count=119335.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 15:40:58,011 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.145e+02 1.731e+02 2.132e+02 2.565e+02 6.876e+02, threshold=4.265e+02, percent-clipped=4.0 2023-03-09 15:41:24,905 INFO [train2.py:809] (1/4) Epoch 30, batch 3850, loss[ctc_loss=0.06147, att_loss=0.2406, loss=0.2048, over 17245.00 frames. utt_duration=874.9 frames, utt_pad_proportion=0.07908, over 79.00 utterances.], tot_loss[ctc_loss=0.0618, att_loss=0.2296, loss=0.1961, over 3273662.30 frames. utt_duration=1231 frames, utt_pad_proportion=0.05683, over 10651.02 utterances.], batch size: 79, lr: 3.54e-03, grad_scale: 8.0 2023-03-09 15:41:26,345 INFO [scaling.py:679] (1/4) Whitening: num_groups=1, num_channels=384, metric=4.30 vs. limit=5.0 2023-03-09 15:42:43,262 INFO [train2.py:809] (1/4) Epoch 30, batch 3900, loss[ctc_loss=0.05745, att_loss=0.2355, loss=0.1999, over 17036.00 frames. utt_duration=1337 frames, utt_pad_proportion=0.007089, over 51.00 utterances.], tot_loss[ctc_loss=0.06247, att_loss=0.23, loss=0.1965, over 3271123.19 frames. utt_duration=1231 frames, utt_pad_proportion=0.05759, over 10639.39 utterances.], batch size: 51, lr: 3.54e-03, grad_scale: 8.0 2023-03-09 15:42:46,453 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=119431.0, num_to_drop=0, layers_to_drop=set() 2023-03-09 15:43:14,049 INFO [zipformer.py:625] (1/4) warmup_begin=1333.3, warmup_end=2000.0, batch_count=119449.0, num_to_drop=1, layers_to_drop={0} 2023-03-09 15:43:33,708 INFO [optim.py:369] (1/4) Clipping_scale=2.0, grad-norm quartiles 1.098e+02 1.815e+02 2.229e+02 2.704e+02 7.416e+02, threshold=4.458e+02, percent-clipped=4.0 2023-03-09 15:43:50,629 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([5.9764, 6.2259, 5.6421, 5.8641, 5.8563, 5.4081, 5.7181, 5.2902], device='cuda:1'), covar=tensor([0.1364, 0.0899, 0.1126, 0.0835, 0.1022, 0.1560, 0.2094, 0.2181], device='cuda:1'), in_proj_covar=tensor([0.0567, 0.0650, 0.0500, 0.0484, 0.0459, 0.0483, 0.0655, 0.0553], 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') 2023-03-09 15:43:58,560 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([4.2185, 2.6552, 4.5833, 3.7592, 2.9708, 3.9896, 4.1418, 4.2802], device='cuda:1'), covar=tensor([0.0316, 0.1450, 0.0242, 0.0815, 0.1531, 0.0334, 0.0356, 0.0336], device='cuda:1'), in_proj_covar=tensor([0.0242, 0.0245, 0.0236, 0.0325, 0.0271, 0.0248, 0.0232, 0.0250], device='cuda:1'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:1') 2023-03-09 15:43:59,617 INFO [train2.py:809] (1/4) Epoch 30, batch 3950, loss[ctc_loss=0.08109, att_loss=0.2449, loss=0.2121, over 17319.00 frames. utt_duration=878.6 frames, utt_pad_proportion=0.08001, over 79.00 utterances.], tot_loss[ctc_loss=0.0624, att_loss=0.2295, loss=0.1961, over 3272374.59 frames. utt_duration=1244 frames, utt_pad_proportion=0.05402, over 10534.77 utterances.], batch size: 79, lr: 3.54e-03, grad_scale: 8.0 2023-03-09 15:44:40,489 INFO [zipformer.py:1447] (1/4) attn_weights_entropy = tensor([3.3495, 2.5116, 3.1362, 2.6323, 3.0987, 3.4905, 3.3547, 2.6789], device='cuda:1'), covar=tensor([0.0472, 0.1619, 0.1100, 0.1087, 0.0882, 0.1203, 0.0676, 0.1233], device='cuda:1'), in_proj_covar=tensor([0.0253, 0.0252, 0.0297, 0.0220, 0.0277, 0.0388, 0.0277, 0.0240], device='cuda:1'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0005, 0.0003, 0.0003], device='cuda:1') 2023-03-09 15:44:52,276 INFO [train2.py:1037] (1/4) Done!