icefall_asr_wenetspeech_pruned_transducer_stateless2 / log_trained_with_S /fast_beam_search /log-decode-epoch-29-avg-24-beam-4-max-contexts-4-max-states-8-2022-05-29-21-10-39
luomingshuang's picture
add M and S training models
6b2eb67
2022-05-29 21:10:39,882 INFO [decode.py:470] Decoding started
2022-05-29 21:10:39,882 INFO [decode.py:476] Device: cuda:0
2022-05-29 21:10:41,190 INFO [lexicon.py:176] Loading pre-compiled data/lang_char/Linv.pt
2022-05-29 21:10:41,277 INFO [decode.py:482] {'best_train_loss': inf, 'best_valid_loss': inf, 'best_train_epoch': -1, 'best_valid_epoch': -1, 'batch_idx_train': 10, 'log_interval': 1, 'reset_interval': 200, 'feature_dim': 80, 'subsampling_factor': 4, 'encoder_dim': 512, 'nhead': 8, 'dim_feedforward': 2048, 'num_encoder_layers': 12, 'decoder_dim': 512, 'joiner_dim': 512, 'env_info': {'k2-version': '1.15.1', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'f8d2dba06c000ffee36aab5b66f24e7c9809f116', 'k2-git-date': 'Thu Apr 21 12:20:34 2022', 'lhotse-version': '1.2.0.dev+git.de75634.dirty', 'torch-version': '1.11.0', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'wenetspeech-pruned-transducer-stateless2', 'icefall-git-sha1': '4b567e4-dirty', 'icefall-git-date': 'Wed Apr 27 13:43:54 2022', 'icefall-path': '/ceph-meixu/luomingshuang/icefall', 'k2-path': '/ceph-ms/luomingshuang/k2_latest/k2/python/k2/__init__.py', 'lhotse-path': '/ceph-meixu/luomingshuang/anaconda3/envs/k2-python/lib/python3.8/site-packages/lhotse-1.2.0.dev0+git.de75634.dirty-py3.8.egg/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-7-0309102938-68688b4cbd-xhtcg', 'IP address': '10.48.32.137'}, 'epoch': 29, 'batch': None, 'avg': 24, 'avg_last_n': 0, 'exp_dir': PosixPath('pruned_transducer_stateless2/exp-char-S-4-gpus'), 'lang_dir': 'data/lang_char', 'decoding_method': 'fast_beam_search', 'beam_size': 4, 'beam': 4, 'max_contexts': 4, 'max_states': 8, 'context_size': 2, 'max_sym_per_frame': 1, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 1500, 'bucketing_sampler': True, 'num_buckets': 300, 'concatenate_cuts': False, 'duration_factor': 1.0, 'gap': 1.0, 'on_the_fly_feats': False, 'shuffle': True, 'return_cuts': True, 'num_workers': 2, 'enable_spec_aug': True, 'spec_aug_time_warp_factor': 80, 'enable_musan': True, 'lazy_load': True, 'training_subset': 'L', 'res_dir': PosixPath('pruned_transducer_stateless2/exp-char-S-4-gpus/fast_beam_search'), 'suffix': 'epoch-29-avg-24-beam-4-max-contexts-4-max-states-8', 'blank_id': 0, 'vocab_size': 5537}
2022-05-29 21:10:41,277 INFO [decode.py:484] About to create model
2022-05-29 21:10:41,907 INFO [decode.py:505] averaging ['pruned_transducer_stateless2/exp-char-S-4-gpus/epoch-6.pt', 'pruned_transducer_stateless2/exp-char-S-4-gpus/epoch-7.pt', 'pruned_transducer_stateless2/exp-char-S-4-gpus/epoch-8.pt', 'pruned_transducer_stateless2/exp-char-S-4-gpus/epoch-9.pt', 'pruned_transducer_stateless2/exp-char-S-4-gpus/epoch-10.pt', 'pruned_transducer_stateless2/exp-char-S-4-gpus/epoch-11.pt', 'pruned_transducer_stateless2/exp-char-S-4-gpus/epoch-12.pt', 'pruned_transducer_stateless2/exp-char-S-4-gpus/epoch-13.pt', 'pruned_transducer_stateless2/exp-char-S-4-gpus/epoch-14.pt', 'pruned_transducer_stateless2/exp-char-S-4-gpus/epoch-15.pt', 'pruned_transducer_stateless2/exp-char-S-4-gpus/epoch-16.pt', 'pruned_transducer_stateless2/exp-char-S-4-gpus/epoch-17.pt', 'pruned_transducer_stateless2/exp-char-S-4-gpus/epoch-18.pt', 'pruned_transducer_stateless2/exp-char-S-4-gpus/epoch-19.pt', 'pruned_transducer_stateless2/exp-char-S-4-gpus/epoch-20.pt', 'pruned_transducer_stateless2/exp-char-S-4-gpus/epoch-21.pt', 'pruned_transducer_stateless2/exp-char-S-4-gpus/epoch-22.pt', 'pruned_transducer_stateless2/exp-char-S-4-gpus/epoch-23.pt', 'pruned_transducer_stateless2/exp-char-S-4-gpus/epoch-24.pt', 'pruned_transducer_stateless2/exp-char-S-4-gpus/epoch-25.pt', 'pruned_transducer_stateless2/exp-char-S-4-gpus/epoch-26.pt', 'pruned_transducer_stateless2/exp-char-S-4-gpus/epoch-27.pt', 'pruned_transducer_stateless2/exp-char-S-4-gpus/epoch-28.pt', 'pruned_transducer_stateless2/exp-char-S-4-gpus/epoch-29.pt']
2022-05-29 21:14:00,758 INFO [decode.py:523] Number of model parameters: 88978927
2022-05-29 21:14:00,894 INFO [asr_datamodule.py:353] About to create dev dataset
2022-05-29 21:14:15,621 INFO [asr_datamodule.py:374] About to create dev dataloader
2022-05-29 21:14:37,752 INFO [decode.py:390] batch 0/?, cuts processed until now is 204
2022-05-29 21:14:42,268 INFO [decode.py:390] batch 2/?, cuts processed until now is 614
2022-05-29 21:14:46,911 INFO [decode.py:390] batch 4/?, cuts processed until now is 1427
2022-05-29 21:14:51,366 INFO [decode.py:390] batch 6/?, cuts processed until now is 2282
2022-05-29 21:14:56,120 INFO [decode.py:390] batch 8/?, cuts processed until now is 2693
2022-05-29 21:15:00,895 INFO [decode.py:390] batch 10/?, cuts processed until now is 3029
2022-05-29 21:15:05,470 INFO [decode.py:390] batch 12/?, cuts processed until now is 3393
2022-05-29 21:15:09,894 INFO [decode.py:390] batch 14/?, cuts processed until now is 3831
2022-05-29 21:15:14,449 INFO [decode.py:390] batch 16/?, cuts processed until now is 4242
2022-05-29 21:15:18,918 INFO [decode.py:390] batch 18/?, cuts processed until now is 4744
2022-05-29 21:15:23,243 INFO [decode.py:390] batch 20/?, cuts processed until now is 5310
2022-05-29 21:15:27,633 INFO [decode.py:390] batch 22/?, cuts processed until now is 5961
2022-05-29 21:15:32,134 INFO [decode.py:390] batch 24/?, cuts processed until now is 6611
2022-05-29 21:15:36,657 INFO [decode.py:390] batch 26/?, cuts processed until now is 7265
2022-05-29 21:15:41,139 INFO [decode.py:390] batch 28/?, cuts processed until now is 7981
2022-05-29 21:15:45,558 INFO [decode.py:390] batch 30/?, cuts processed until now is 8609
2022-05-29 21:15:50,115 INFO [decode.py:390] batch 32/?, cuts processed until now is 9551
2022-05-29 21:15:54,691 INFO [decode.py:390] batch 34/?, cuts processed until now is 10324
2022-05-29 21:15:59,268 INFO [decode.py:390] batch 36/?, cuts processed until now is 10687
2022-05-29 21:16:03,414 INFO [decode.py:390] batch 38/?, cuts processed until now is 11164
2022-05-29 21:16:06,482 INFO [decode.py:390] batch 40/?, cuts processed until now is 11535
2022-05-29 21:16:08,810 INFO [decode.py:390] batch 42/?, cuts processed until now is 11862
2022-05-29 21:16:10,867 INFO [decode.py:390] batch 44/?, cuts processed until now is 12149
2022-05-29 21:16:13,196 INFO [decode.py:390] batch 46/?, cuts processed until now is 12328
2022-05-29 21:16:16,213 INFO [decode.py:390] batch 48/?, cuts processed until now is 12491
2022-05-29 21:16:18,751 INFO [decode.py:390] batch 50/?, cuts processed until now is 12669
2022-05-29 21:16:20,993 INFO [decode.py:390] batch 52/?, cuts processed until now is 12815
2022-05-29 21:16:23,593 INFO [decode.py:390] batch 54/?, cuts processed until now is 12980
2022-05-29 21:16:26,156 INFO [decode.py:390] batch 56/?, cuts processed until now is 13442
2022-05-29 21:16:28,167 INFO [decode.py:390] batch 58/?, cuts processed until now is 13735
2022-05-29 21:16:29,494 INFO [decode.py:407] The transcripts are stored in pruned_transducer_stateless2/exp-char-S-4-gpus/fast_beam_search/recogs-DEV-beam_4_max_contexts_4_max_states_8-epoch-29-avg-24-beam-4-max-contexts-4-max-states-8.txt
2022-05-29 21:16:29,983 INFO [utils.py:406] [DEV-beam_4_max_contexts_4_max_states_8] %WER 19.02% [62867 / 330498, 1407 ins, 19330 del, 42130 sub ]
2022-05-29 21:16:31,066 INFO [decode.py:420] Wrote detailed error stats to pruned_transducer_stateless2/exp-char-S-4-gpus/fast_beam_search/errs-DEV-beam_4_max_contexts_4_max_states_8-epoch-29-avg-24-beam-4-max-contexts-4-max-states-8.txt
2022-05-29 21:16:31,073 INFO [decode.py:437]
For DEV, WER of different settings are:
beam_4_max_contexts_4_max_states_8 19.02 best for DEV
2022-05-29 21:16:39,435 INFO [decode.py:390] batch 0/?, cuts processed until now is 240
2022-05-29 21:16:44,003 INFO [decode.py:390] batch 2/?, cuts processed until now is 720
2022-05-29 21:16:49,100 INFO [decode.py:390] batch 4/?, cuts processed until now is 2345
2022-05-29 21:16:54,191 INFO [decode.py:390] batch 6/?, cuts processed until now is 4113
2022-05-29 21:17:02,217 INFO [decode.py:390] batch 8/?, cuts processed until now is 4598
2022-05-29 21:17:11,999 INFO [decode.py:390] batch 10/?, cuts processed until now is 4927
2022-05-29 21:17:16,396 INFO [decode.py:390] batch 12/?, cuts processed until now is 5340
2022-05-29 21:17:20,799 INFO [decode.py:390] batch 14/?, cuts processed until now is 5908
2022-05-29 21:17:27,084 INFO [decode.py:390] batch 16/?, cuts processed until now is 6395
2022-05-29 21:17:34,082 INFO [decode.py:390] batch 18/?, cuts processed until now is 6750
2022-05-29 21:17:38,658 INFO [decode.py:390] batch 20/?, cuts processed until now is 7247
2022-05-29 21:17:43,107 INFO [decode.py:390] batch 22/?, cuts processed until now is 8319
2022-05-29 21:17:47,581 INFO [decode.py:390] batch 24/?, cuts processed until now is 9403
2022-05-29 21:17:52,159 INFO [decode.py:390] batch 26/?, cuts processed until now is 10482
2022-05-29 21:17:56,821 INFO [decode.py:390] batch 28/?, cuts processed until now is 11733
2022-05-29 21:18:01,170 INFO [decode.py:390] batch 30/?, cuts processed until now is 12776
2022-05-29 21:18:06,020 INFO [decode.py:390] batch 32/?, cuts processed until now is 14713
2022-05-29 21:18:10,846 INFO [decode.py:390] batch 34/?, cuts processed until now is 16235
2022-05-29 21:18:18,460 INFO [decode.py:390] batch 36/?, cuts processed until now is 16564
2022-05-29 21:18:25,278 INFO [decode.py:390] batch 38/?, cuts processed until now is 17289
2022-05-29 21:18:29,403 INFO [decode.py:390] batch 40/?, cuts processed until now is 18330
2022-05-29 21:18:33,299 INFO [decode.py:390] batch 42/?, cuts processed until now is 19370
2022-05-29 21:18:36,169 INFO [decode.py:390] batch 44/?, cuts processed until now is 20163
2022-05-29 21:18:39,316 INFO [decode.py:390] batch 46/?, cuts processed until now is 20677
2022-05-29 21:18:43,251 INFO [decode.py:390] batch 48/?, cuts processed until now is 21115
2022-05-29 21:18:47,017 INFO [decode.py:390] batch 50/?, cuts processed until now is 21505
2022-05-29 21:18:50,963 INFO [decode.py:390] batch 52/?, cuts processed until now is 22645
2022-05-29 21:18:53,993 INFO [decode.py:390] batch 54/?, cuts processed until now is 23529
2022-05-29 21:18:56,814 INFO [decode.py:390] batch 56/?, cuts processed until now is 23871
2022-05-29 21:19:01,495 INFO [decode.py:390] batch 58/?, cuts processed until now is 24173
2022-05-29 21:19:05,463 INFO [decode.py:390] batch 60/?, cuts processed until now is 24534
2022-05-29 21:19:07,047 INFO [decode.py:407] The transcripts are stored in pruned_transducer_stateless2/exp-char-S-4-gpus/fast_beam_search/recogs-TEST_NET-beam_4_max_contexts_4_max_states_8-epoch-29-avg-24-beam-4-max-contexts-4-max-states-8.txt
2022-05-29 21:19:07,599 INFO [utils.py:406] [TEST_NET-beam_4_max_contexts_4_max_states_8] %WER 24.14% [100372 / 415747, 2343 ins, 22089 del, 75940 sub ]
2022-05-29 21:19:09,057 INFO [decode.py:420] Wrote detailed error stats to pruned_transducer_stateless2/exp-char-S-4-gpus/fast_beam_search/errs-TEST_NET-beam_4_max_contexts_4_max_states_8-epoch-29-avg-24-beam-4-max-contexts-4-max-states-8.txt
2022-05-29 21:19:09,065 INFO [decode.py:437]
For TEST_NET, WER of different settings are:
beam_4_max_contexts_4_max_states_8 24.14 best for TEST_NET
2022-05-29 21:19:14,960 INFO [decode.py:390] batch 0/?, cuts processed until now is 149
2022-05-29 21:19:19,955 INFO [decode.py:390] batch 2/?, cuts processed until now is 431
2022-05-29 21:19:25,041 INFO [decode.py:390] batch 4/?, cuts processed until now is 1136
2022-05-29 21:19:29,670 INFO [decode.py:390] batch 6/?, cuts processed until now is 1932
2022-05-29 21:19:34,536 INFO [decode.py:390] batch 8/?, cuts processed until now is 2279
2022-05-29 21:19:39,505 INFO [decode.py:390] batch 10/?, cuts processed until now is 2566
2022-05-29 21:19:45,105 INFO [decode.py:390] batch 12/?, cuts processed until now is 2839
2022-05-29 21:19:51,582 INFO [decode.py:390] batch 14/?, cuts processed until now is 3215
2022-05-29 21:19:56,023 INFO [decode.py:390] batch 16/?, cuts processed until now is 3689
2022-05-29 21:20:00,350 INFO [decode.py:390] batch 18/?, cuts processed until now is 4238
2022-05-29 21:20:05,395 INFO [decode.py:390] batch 20/?, cuts processed until now is 4656
2022-05-29 21:20:10,451 INFO [decode.py:390] batch 22/?, cuts processed until now is 4977
2022-05-29 21:20:14,187 INFO [decode.py:390] batch 24/?, cuts processed until now is 5384
2022-05-29 21:20:18,004 INFO [decode.py:390] batch 26/?, cuts processed until now is 5763
2022-05-29 21:20:21,973 INFO [decode.py:390] batch 28/?, cuts processed until now is 6230
2022-05-29 21:20:25,653 INFO [decode.py:390] batch 30/?, cuts processed until now is 6928
2022-05-29 21:20:29,804 INFO [decode.py:390] batch 32/?, cuts processed until now is 7519
2022-05-29 21:20:34,327 INFO [decode.py:390] batch 34/?, cuts processed until now is 7726
2022-05-29 21:20:39,108 INFO [decode.py:390] batch 36/?, cuts processed until now is 8005
2022-05-29 21:20:42,695 INFO [decode.py:390] batch 38/?, cuts processed until now is 8229
2022-05-29 21:20:45,171 INFO [decode.py:407] The transcripts are stored in pruned_transducer_stateless2/exp-char-S-4-gpus/fast_beam_search/recogs-TEST_MEETING-beam_4_max_contexts_4_max_states_8-epoch-29-avg-24-beam-4-max-contexts-4-max-states-8.txt
2022-05-29 21:20:45,451 INFO [utils.py:406] [TEST_MEETING-beam_4_max_contexts_4_max_states_8] %WER 34.97% [77077 / 220385, 1575 ins, 30420 del, 45082 sub ]
2022-05-29 21:20:46,218 INFO [decode.py:420] Wrote detailed error stats to pruned_transducer_stateless2/exp-char-S-4-gpus/fast_beam_search/errs-TEST_MEETING-beam_4_max_contexts_4_max_states_8-epoch-29-avg-24-beam-4-max-contexts-4-max-states-8.txt
2022-05-29 21:20:46,226 INFO [decode.py:437]
For TEST_MEETING, WER of different settings are:
beam_4_max_contexts_4_max_states_8 34.97 best for TEST_MEETING
2022-05-29 21:20:46,226 INFO [decode.py:617] Done!