2023-03-09 04:23:24,501 INFO [decode.py:657] Decoding started
2023-03-09 04:23:24,501 INFO [decode.py:658] {'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': 'r8n03', 'IP address': '10.1.8.3'}, 'beam_size': 10, 'reduction': 'sum', 'use_double_scores': True, 'search_beam': 20, 'output_beam': 8, 'min_active_states': 30, 'max_active_states': 10000, 'epoch': 23, 'iter': 0, 'avg': 5, 'use_averaged_model': True, 'method': 'ctc-decoding', 'num_paths': 100, 'nbest_scale': 0.5, 'exp_dir': PosixPath('zipformer_ctc_att/exp/v0'), 'lang_dir': PosixPath('data/lang_bpe_500'), 'num_decoder_layers': 6, 'lm_dir': PosixPath('data/lm'), 'rnn_lm_exp_dir': 'rnn_lm/exp', 'rnn_lm_epoch': 7, 'rnn_lm_avg': 2, 'rnn_lm_embedding_dim': 2048, 'rnn_lm_hidden_dim': 2048, 'rnn_lm_num_layers': 4, 'rnn_lm_tie_weights': False, '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': 200.0, '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'}
2023-03-09 04:23:24,610 INFO [lexicon.py:168] Loading pre-compiled data/lang_bpe_500/Linv.pt
2023-03-09 04:23:24,725 INFO [decode.py:669] device: cuda:0
2023-03-09 04:23:29,169 INFO [decode.py:757] About to create model
2023-03-09 04:23:29,625 INFO [zipformer.py:178] 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.
2023-03-09 04:23:29,683 INFO [decode.py:824] Calculating the averaged model over epoch range from 18 (excluded) to 23
2023-03-09 04:23:59,034 INFO [decode.py:840] Number of model parameters: 86083707
2023-03-09 04:23:59,035 INFO [asr_datamodule.py:440] About to get dev-clean cuts
2023-03-09 04:23:59,102 INFO [asr_datamodule.py:454] About to get dev-other cuts
2023-03-09 04:23:59,103 INFO [asr_datamodule.py:468] About to get test-clean cuts
2023-03-09 04:23:59,104 INFO [asr_datamodule.py:482] About to get test-other cuts
2023-03-09 04:24:00,793 INFO [decode.py:595] batch 0/?, cuts processed until now is 16
2023-03-09 04:24:20,887 INFO [zipformer.py:1447] attn_weights_entropy = tensor([3.5097, 2.4544, 3.7387, 3.2820, 2.6631, 3.5698, 3.6508, 3.5118],
       device='cuda:0'), covar=tensor([0.0233, 0.1487, 0.0209, 0.0739, 0.1421, 0.0285, 0.0247, 0.0240],
       device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0236, 0.0176, 0.0306, 0.0256, 0.0204, 0.0164, 0.0194],
       device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002],
       device='cuda:0')
2023-03-09 04:24:23,072 INFO [zipformer.py:1447] attn_weights_entropy = tensor([2.5664, 2.2849, 2.5186, 3.0238, 2.7956, 2.9528, 2.4076, 2.0466],
       device='cuda:0'), covar=tensor([0.0793, 0.2076, 0.0952, 0.0847, 0.0986, 0.0889, 0.1785, 0.2259],
       device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0209, 0.0183, 0.0211, 0.0217, 0.0170, 0.0196, 0.0182],
       device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002],
       device='cuda:0')
2023-03-09 04:24:49,384 INFO [zipformer.py:1447] attn_weights_entropy = tensor([4.8208, 5.4436, 5.4218, 4.8623, 5.9272, 4.8507, 5.3783, 3.3480],
       device='cuda:0'), covar=tensor([0.0181, 0.0106, 0.0117, 0.0305, 0.0505, 0.0170, 0.0146, 0.1411],
       device='cuda:0'), in_proj_covar=tensor([0.0151, 0.0176, 0.0177, 0.0193, 0.0361, 0.0147, 0.0166, 0.0209],
       device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002],
       device='cuda:0')
2023-03-09 04:24:52,404 INFO [decode.py:595] batch 100/?, cuts processed until now is 2335
2023-03-09 04:24:59,003 INFO [decode.py:615] The transcripts are stored in zipformer_ctc_att/exp/v0/recogs-dev-clean-ctc-decoding.txt
2023-03-09 04:24:59,072 INFO [utils.py:538] [dev-clean-ctc-decoding] %WER 2.44% [1325 / 54402, 115 ins, 113 del, 1097 sub ]
2023-03-09 04:24:59,222 INFO [decode.py:627] Wrote detailed error stats to zipformer_ctc_att/exp/v0/errs-dev-clean-ctc-decoding.txt
2023-03-09 04:24:59,223 INFO [decode.py:641] 
For dev-clean, WER of different settings are:
ctc-decoding	2.44	best for dev-clean

2023-03-09 04:25:00,276 INFO [decode.py:595] batch 0/?, cuts processed until now is 18
2023-03-09 04:25:20,446 INFO [zipformer.py:1447] attn_weights_entropy = tensor([3.1746, 3.6242, 3.1356, 3.6365, 2.3967, 3.4838, 2.4663, 1.7018],
       device='cuda:0'), covar=tensor([0.0603, 0.0347, 0.0963, 0.0240, 0.1707, 0.0248, 0.1535, 0.1882],
       device='cuda:0'), in_proj_covar=tensor([0.0182, 0.0154, 0.0252, 0.0147, 0.0215, 0.0134, 0.0225, 0.0198],
       device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002],
       device='cuda:0')
2023-03-09 04:25:50,986 INFO [decode.py:595] batch 100/?, cuts processed until now is 2625
2023-03-09 04:25:55,314 INFO [decode.py:615] The transcripts are stored in zipformer_ctc_att/exp/v0/recogs-dev-other-ctc-decoding.txt
2023-03-09 04:25:55,383 INFO [utils.py:538] [dev-other-ctc-decoding] %WER 6.20% [3157 / 50948, 254 ins, 245 del, 2658 sub ]
2023-03-09 04:25:55,534 INFO [decode.py:627] Wrote detailed error stats to zipformer_ctc_att/exp/v0/errs-dev-other-ctc-decoding.txt
2023-03-09 04:25:55,536 INFO [decode.py:641] 
For dev-other, WER of different settings are:
ctc-decoding	6.2	best for dev-other

2023-03-09 04:25:56,598 INFO [decode.py:595] batch 0/?, cuts processed until now is 14
2023-03-09 04:26:01,569 INFO [zipformer.py:1447] attn_weights_entropy = tensor([4.4120, 5.0760, 5.2997, 5.1498, 3.5879, 4.8010, 3.7668, 2.6052],
       device='cuda:0'), covar=tensor([0.0291, 0.0159, 0.0397, 0.0115, 0.1199, 0.0151, 0.1023, 0.1435],
       device='cuda:0'), in_proj_covar=tensor([0.0182, 0.0154, 0.0252, 0.0147, 0.0215, 0.0134, 0.0225, 0.0198],
       device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002],
       device='cuda:0')
2023-03-09 04:26:16,340 INFO [zipformer.py:1447] attn_weights_entropy = tensor([3.1817, 3.9030, 3.9128, 3.6124, 3.7950, 3.8524, 3.9487, 3.2508],
       device='cuda:0'), covar=tensor([0.0674, 0.0929, 0.1161, 0.1590, 0.1837, 0.0824, 0.0429, 0.2261],
       device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0177, 0.0189, 0.0245, 0.0149, 0.0250, 0.0169, 0.0209],
       device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002],
       device='cuda:0')
2023-03-09 04:26:49,057 INFO [decode.py:595] batch 100/?, cuts processed until now is 2293
2023-03-09 04:26:56,021 INFO [decode.py:615] The transcripts are stored in zipformer_ctc_att/exp/v0/recogs-test-clean-ctc-decoding.txt
2023-03-09 04:26:56,086 INFO [utils.py:538] [test-clean-ctc-decoding] %WER 2.58% [1354 / 52576, 134 ins, 101 del, 1119 sub ]
2023-03-09 04:26:56,224 INFO [decode.py:627] Wrote detailed error stats to zipformer_ctc_att/exp/v0/errs-test-clean-ctc-decoding.txt
2023-03-09 04:26:56,225 INFO [decode.py:641] 
For test-clean, WER of different settings are:
ctc-decoding	2.58	best for test-clean

2023-03-09 04:26:57,298 INFO [decode.py:595] batch 0/?, cuts processed until now is 17
2023-03-09 04:27:36,606 INFO [zipformer.py:1447] attn_weights_entropy = tensor([3.5775, 2.5343, 3.2675, 2.5832, 3.1080, 3.7384, 3.6521, 2.7340],
       device='cuda:0'), covar=tensor([0.0503, 0.1965, 0.1334, 0.1462, 0.1432, 0.1233, 0.0734, 0.1428],
       device='cuda:0'), in_proj_covar=tensor([0.0237, 0.0237, 0.0272, 0.0212, 0.0257, 0.0360, 0.0253, 0.0225],
       device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003],
       device='cuda:0')
2023-03-09 04:27:44,448 INFO [zipformer.py:1447] attn_weights_entropy = tensor([4.1154, 4.9219, 4.8740, 2.4294, 2.1290, 3.0700, 2.4395, 3.8819],
       device='cuda:0'), covar=tensor([0.0583, 0.0207, 0.0217, 0.5031, 0.5416, 0.2327, 0.3603, 0.1347],
       device='cuda:0'), in_proj_covar=tensor([0.0346, 0.0263, 0.0258, 0.0238, 0.0335, 0.0328, 0.0247, 0.0356],
       device='cuda:0'), out_proj_covar=tensor([1.4699e-04, 9.7060e-05, 1.0949e-04, 1.0262e-04, 1.4174e-04, 1.2850e-04,
        9.9139e-05, 1.4624e-04], device='cuda:0')
2023-03-09 04:27:48,477 INFO [decode.py:595] batch 100/?, cuts processed until now is 2560
2023-03-09 04:27:55,425 INFO [decode.py:615] The transcripts are stored in zipformer_ctc_att/exp/v0/recogs-test-other-ctc-decoding.txt
2023-03-09 04:27:55,496 INFO [utils.py:538] [test-other-ctc-decoding] %WER 6.25% [3271 / 52343, 306 ins, 273 del, 2692 sub ]
2023-03-09 04:27:55,651 INFO [decode.py:627] Wrote detailed error stats to zipformer_ctc_att/exp/v0/errs-test-other-ctc-decoding.txt
2023-03-09 04:27:55,652 INFO [decode.py:641] 
For test-other, WER of different settings are:
ctc-decoding	6.25	best for test-other

2023-03-09 04:27:55,652 INFO [decode.py:901] Done!