decoding_result/fast_beam_search/log-decode-epoch-34-avg-19-beam-20.0-max-contexts-8-max-states-64-use-averaged-model-2023-11-17-17-30-48

2023-11-17 17:30:48,977 INFO [decode.py:688] Decoding started
2023-11-17 17:30:48,978 INFO [decode.py:694] Device: cuda:0
2023-11-17 17:30:48,985 INFO [decode.py:704] {'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': '1.24.3', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': '821ebc378e7fb99b8adc81950227963332821e01', 'k2-git-date': 'Wed Jul 19 15:38:25 2023', 'lhotse-version': '1.17.0.dev+git.b3dc9faf.dirty', 'torch-version': '1.11.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.9', 'icefall-git-branch': 'dev/bilingual', 'icefall-git-sha1': '4897f2c0-dirty', 'icefall-git-date': 'Thu Sep 28 11:38:28 2023', 'icefall-path': '/star-home/jinzengrui/lib/miniconda3/envs/dev39/lib/python3.9/site-packages/icefall-1.0-py3.9.egg', 'k2-path': '/star-home/jinzengrui/lib/miniconda3/envs/dev39/lib/python3.9/site-packages/k2-1.24.3.dev20230721+cuda10.2.torch1.11.0-py3.9-linux-x86_64.egg/k2/__init__.py', 'lhotse-path': '/star-home/jinzengrui/lib/miniconda3/envs/dev39/lib/python3.9/site-packages/lhotse-1.17.0.dev0+git.b3dc9faf.dirty-py3.9.egg/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-0423201334-6587bbc68d-tn554', 'IP address': '10.177.74.211'}, 'epoch': 34, 'iter': 0, 'avg': 19, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp-w-tal-csasr'), 'bpe_model': 'data/lang_bbpe_2000/bbpe.model', 'lang_dir': PosixPath('data/lang_bbpe_2000'), 'decoding_method': 'fast_beam_search', 'beam_size': 4, 'beam': 20.0, 'ngram_lm_scale': 0.01, 'max_contexts': 8, 'max_states': 64, 'context_size': 2, 'max_sym_per_frame': 1, 'num_paths': 200, 'nbest_scale': 0.5, 'use_tal_csasr': False, 'use_librispeech': False, 'use_aishell2': False, 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'decoder_dim': 512, 'joiner_dim': 512, 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'use_transducer': True, 'use_ctc': False, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 300.0, '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', 'res_dir': PosixPath('zipformer/exp-w-tal-csasr/fast_beam_search'), 'suffix': 'epoch-34-avg-19-beam-20.0-max-contexts-8-max-states-64-use-averaged-model', 'blank_id': 0, 'unk_id': 2, 'vocab_size': 2000}
2023-11-17 17:30:48,986 INFO [decode.py:706] About to create model
2023-11-17 17:30:49,738 INFO [decode.py:773] Calculating the averaged model over epoch range from 15 (excluded) to 34
2023-11-17 17:31:08,055 INFO [decode.py:807] Number of model parameters: 68625511
2023-11-17 17:31:08,055 INFO [multi_dataset.py:142] About to get multidataset test cuts
2023-11-17 17:31:08,056 INFO [multi_dataset.py:145] Loading Aishell-2 set in lazy mode
2023-11-17 17:31:08,056 INFO [multi_dataset.py:157] Loading TAL-CSASR set in lazy mode
2023-11-17 17:31:09,242 INFO [decode.py:831] Start decoding test set: tal_csasr_test
2023-11-17 17:31:15,994 INFO [decode.py:585] batch 0/?, cuts processed until now is 27
2023-11-17 17:31:21,546 INFO [zipformer.py:1853] name=None, attn_weights_entropy = tensor([3.7140, 3.4879, 3.3815, 3.3047], device='cuda:0')
2023-11-17 17:33:02,322 INFO [decode.py:585] batch 20/?, cuts processed until now is 772
2023-11-17 17:34:40,082 INFO [decode.py:585] batch 40/?, cuts processed until now is 1576
2023-11-17 17:36:09,675 INFO [decode.py:585] batch 60/?, cuts processed until now is 2563
2023-11-17 17:38:01,626 INFO [decode.py:585] batch 80/?, cuts processed until now is 3238
2023-11-17 17:39:01,426 INFO [zipformer.py:1853] name=None, attn_weights_entropy = tensor([4.2191, 4.0339, 3.7788, 3.6473], device='cuda:0')
2023-11-17 17:39:29,046 INFO [decode.py:585] batch 100/?, cuts processed until now is 4248
2023-11-17 17:41:10,049 INFO [decode.py:585] batch 120/?, cuts processed until now is 5015
2023-11-17 17:42:18,277 INFO [zipformer.py:1853] name=None, attn_weights_entropy = tensor([2.9075, 2.7398, 3.6239, 3.5886], device='cuda:0')
2023-11-17 17:42:33,083 INFO [decode.py:585] batch 140/?, cuts processed until now is 6099
2023-11-17 17:43:56,259 INFO [decode.py:585] batch 160/?, cuts processed until now is 7227
2023-11-17 17:44:34,211 INFO [zipformer.py:1853] name=None, attn_weights_entropy = tensor([6.0532, 5.9718, 5.8672, 5.3014], device='cuda:0')
2023-11-17 17:45:24,393 INFO [decode.py:585] batch 180/?, cuts processed until now is 8156
2023-11-17 17:46:17,041 INFO [zipformer.py:1853] name=None, attn_weights_entropy = tensor([5.3534, 5.0811, 4.8281, 4.8843], device='cuda:0')
2023-11-17 17:46:56,076 INFO [decode.py:585] batch 200/?, cuts processed until now is 9045
2023-11-17 17:48:36,439 INFO [decode.py:585] batch 220/?, cuts processed until now is 9901
2023-11-17 17:50:09,847 INFO [decode.py:585] batch 240/?, cuts processed until now is 10806
2023-11-17 17:51:48,201 INFO [decode.py:585] batch 260/?, cuts processed until now is 11684
2023-11-17 17:52:26,414 INFO [zipformer.py:1853] name=None, attn_weights_entropy = tensor([1.9570, 1.7653, 3.8541, 3.5736], device='cuda:0')
2023-11-17 17:53:07,522 INFO [decode.py:585] batch 280/?, cuts processed until now is 12756
2023-11-17 17:54:25,360 INFO [decode.py:585] batch 300/?, cuts processed until now is 13809
2023-11-17 17:55:17,859 INFO [zipformer.py:1853] name=None, attn_weights_entropy = tensor([3.0869, 2.9026, 2.5093, 2.5621], device='cuda:0')
2023-11-17 17:55:59,054 INFO [decode.py:585] batch 320/?, cuts processed until now is 14719
2023-11-17 17:57:05,185 INFO [decode.py:601] The transcripts are stored in zipformer/exp-w-tal-csasr/fast_beam_search/recogs-tal_csasr_test-beam_20.0_max_contexts_8_max_states_64-epoch-34-avg-19-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
2023-11-17 17:57:05,860 INFO [utils.py:565] [tal_csasr_test-beam_20.0_max_contexts_8_max_states_64] %WER 6.68% [22382 / 334989, 3704 ins, 5492 del, 13186 sub ]
2023-11-17 17:57:07,213 INFO [decode.py:614] Wrote detailed error stats to zipformer/exp-w-tal-csasr/fast_beam_search/errs-tal_csasr_test-beam_20.0_max_contexts_8_max_states_64-epoch-34-avg-19-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
2023-11-17 17:57:07,218 INFO [decode.py:630] 
For tal_csasr_test, WER of different settings are:
beam_20.0_max_contexts_8_max_states_64	6.68	best for tal_csasr_test

2023-11-17 17:57:07,219 INFO [decode.py:831] Start decoding test set: tal_csasr_dev
2023-11-17 17:57:15,137 INFO [decode.py:585] batch 0/?, cuts processed until now is 26
2023-11-17 17:57:24,392 INFO [zipformer.py:1853] name=None, attn_weights_entropy = tensor([3.0018, 3.4013, 2.8015, 3.1066, 3.1630, 1.9779, 1.8654, 3.1387],
       device='cuda:0')
2023-11-17 17:58:49,980 INFO [decode.py:585] batch 20/?, cuts processed until now is 758
2023-11-17 17:59:14,367 INFO [decode.py:585] batch 40/?, cuts processed until now is 1552
2023-11-17 17:59:28,121 INFO [zipformer.py:1853] name=None, attn_weights_entropy = tensor([1.4897, 2.4091, 3.3914, 2.1619], device='cuda:0')
2023-11-17 17:59:38,132 INFO [decode.py:585] batch 60/?, cuts processed until now is 2520
2023-11-17 18:00:59,231 INFO [decode.py:585] batch 80/?, cuts processed until now is 3309
2023-11-17 18:01:46,144 INFO [zipformer.py:1853] name=None, attn_weights_entropy = tensor([3.8113, 4.2499, 3.2566, 3.7277, 3.8991, 2.6693, 2.4118, 4.0344],
       device='cuda:0')
2023-11-17 18:02:28,898 INFO [decode.py:585] batch 100/?, cuts processed until now is 4451
2023-11-17 18:03:19,829 INFO [decode.py:601] The transcripts are stored in zipformer/exp-w-tal-csasr/fast_beam_search/recogs-tal_csasr_dev-beam_20.0_max_contexts_8_max_states_64-epoch-34-avg-19-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
2023-11-17 18:03:20,098 INFO [utils.py:565] [tal_csasr_dev-beam_20.0_max_contexts_8_max_states_64] %WER 6.57% [7489 / 113905, 1281 ins, 1798 del, 4410 sub ]
2023-11-17 18:03:20,616 INFO [decode.py:614] Wrote detailed error stats to zipformer/exp-w-tal-csasr/fast_beam_search/errs-tal_csasr_dev-beam_20.0_max_contexts_8_max_states_64-epoch-34-avg-19-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
2023-11-17 18:03:20,622 INFO [decode.py:630] 
For tal_csasr_dev, WER of different settings are:
beam_20.0_max_contexts_8_max_states_64	6.57	best for tal_csasr_dev

2023-11-17 18:03:20,623 INFO [decode.py:848] Done!