baseline/modified_beam_search/log-decode-epoch-40-avg-10-modified_beam_search-beam-size-4-2022-05-28-00-51-00

2022-05-28 00:51:00,330 INFO [decode.py:495] Decoding started
2022-05-28 00:51:00,330 INFO [decode.py:501] Device: cuda:0
2022-05-28 00:51:00,333 INFO [decode.py:511] {'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, 'encoder_dim': 512, 'nhead': 8, 'dim_feedforward': 2048, 'num_encoder_layers': 12, 'decoder_dim': 512, 'joiner_dim': 512, 'model_warm_step': 3000, 'env_info': {'k2-version': '1.13', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'f4fefe4882bc0ae59af951da3f47335d5495ef71', 'k2-git-date': 'Thu Feb 10 15:16:02 2022', 'lhotse-version': '1.1.0', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'stateless6', 'icefall-git-sha1': '50641cd-dirty', 'icefall-git-date': 'Fri May 27 13:49:39 2022', 'icefall-path': '/ceph-data2/ly/open_source/vq2_icefall', 'k2-path': '/ceph-jb/yaozengwei/workspace/rnnt/k2/k2/python/k2/__init__.py', 'lhotse-path': '/ceph-ly/open-source/hubert/lhotse/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-9-0425111216-65f66bdf4-bkrql', 'IP address': '10.177.77.9'}, 'enable_distiallation': False, 'distillation_layer': 5, 'num_codebooks': 16, 'epoch': 40, 'iter': 0, 'avg': 10, 'use_averaged_model': False, 'exp_dir': PosixPath('pruned_transducer_stateless6/exp'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'decoding_method': 'modified_beam_search', 'beam_size': 4, 'beam': 4, 'max_contexts': 4, 'max_states': 8, 'context_size': 2, 'max_sym_per_frame': 1, 'full_libri': True, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 200, '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('pruned_transducer_stateless6/exp/modified_beam_search'), 'suffix': 'epoch-40-avg-10-modified_beam_search-beam-size-4', 'blank_id': 0, 'unk_id': 2, 'vocab_size': 500}
2022-05-28 00:51:00,334 INFO [decode.py:513] About to create model
2022-05-28 00:51:00,810 INFO [decode.py:542] averaging ['pruned_transducer_stateless6/exp/epoch-31.pt', 'pruned_transducer_stateless6/exp/epoch-32.pt', 'pruned_transducer_stateless6/exp/epoch-33.pt', 'pruned_transducer_stateless6/exp/epoch-34.pt', 'pruned_transducer_stateless6/exp/epoch-35.pt', 'pruned_transducer_stateless6/exp/epoch-36.pt', 'pruned_transducer_stateless6/exp/epoch-37.pt', 'pruned_transducer_stateless6/exp/epoch-38.pt', 'pruned_transducer_stateless6/exp/epoch-39.pt', 'pruned_transducer_stateless6/exp/epoch-40.pt']
2022-05-28 00:51:56,806 INFO [decode.py:602] Number of model parameters: 78648040
2022-05-28 00:51:56,807 INFO [asr_datamodule.py:439] About to get test-clean cuts
2022-05-28 00:51:57,049 INFO [asr_datamodule.py:444] About to get test-other cuts
2022-05-28 00:51:59,244 INFO [decode.py:406] batch 0/?, cuts processed until now is 41
2022-05-28 00:52:12,999 INFO [decode.py:406] batch 10/?, cuts processed until now is 345
2022-05-28 00:52:28,426 INFO [decode.py:406] batch 20/?, cuts processed until now is 515
2022-05-28 00:52:42,237 INFO [decode.py:406] batch 30/?, cuts processed until now is 791
2022-05-28 00:52:57,191 INFO [decode.py:406] batch 40/?, cuts processed until now is 1054
2022-05-28 00:53:12,786 INFO [decode.py:406] batch 50/?, cuts processed until now is 1362
2022-05-28 00:53:26,351 INFO [decode.py:406] batch 60/?, cuts processed until now is 1659
2022-05-28 00:53:39,202 INFO [decode.py:406] batch 70/?, cuts processed until now is 1888
2022-05-28 00:53:52,447 INFO [decode.py:406] batch 80/?, cuts processed until now is 2185
2022-05-28 00:54:05,879 INFO [decode.py:406] batch 90/?, cuts processed until now is 2388
2022-05-28 00:54:20,168 INFO [decode.py:406] batch 100/?, cuts processed until now is 2560
2022-05-28 00:54:28,824 INFO [decode.py:423] The transcripts are stored in pruned_transducer_stateless6/exp/modified_beam_search/recogs-test-clean-beam_size_4-epoch-40-avg-10-modified_beam_search-beam-size-4.txt
2022-05-28 00:54:28,915 INFO [utils.py:409] [test-clean-beam_size_4] %WER 6.73% [3540 / 52576, 385 ins, 348 del, 2807 sub ]
2022-05-28 00:54:29,131 INFO [decode.py:436] Wrote detailed error stats to pruned_transducer_stateless6/exp/modified_beam_search/errs-test-clean-beam_size_4-epoch-40-avg-10-modified_beam_search-beam-size-4.txt
2022-05-28 00:54:29,132 INFO [decode.py:453] 
For test-clean, WER of different settings are:
beam_size_4	6.73	best for test-clean

2022-05-28 00:54:30,590 INFO [decode.py:406] batch 0/?, cuts processed until now is 46
2022-05-28 00:54:43,833 INFO [decode.py:406] batch 10/?, cuts processed until now is 382
2022-05-28 00:54:59,115 INFO [decode.py:406] batch 20/?, cuts processed until now is 577
2022-05-28 00:55:12,249 INFO [decode.py:406] batch 30/?, cuts processed until now is 887
2022-05-28 00:55:25,749 INFO [decode.py:406] batch 40/?, cuts processed until now is 1187
2022-05-28 00:55:39,494 INFO [decode.py:406] batch 50/?, cuts processed until now is 1518
2022-05-28 00:55:52,486 INFO [decode.py:406] batch 60/?, cuts processed until now is 1848
2022-05-28 00:56:05,499 INFO [decode.py:406] batch 70/?, cuts processed until now is 2133
2022-05-28 00:56:20,351 INFO [decode.py:406] batch 80/?, cuts processed until now is 2419
2022-05-28 00:56:37,169 INFO [decode.py:406] batch 90/?, cuts processed until now is 2685
2022-05-28 00:56:47,056 INFO [decode.py:406] batch 100/?, cuts processed until now is 2868
2022-05-28 00:56:53,891 INFO [decode.py:423] The transcripts are stored in pruned_transducer_stateless6/exp/modified_beam_search/recogs-test-other-beam_size_4-epoch-40-avg-10-modified_beam_search-beam-size-4.txt
2022-05-28 00:56:53,987 INFO [utils.py:409] [test-other-beam_size_4] %WER 17.79% [9314 / 52343, 900 ins, 1096 del, 7318 sub ]
2022-05-28 00:56:54,227 INFO [decode.py:436] Wrote detailed error stats to pruned_transducer_stateless6/exp/modified_beam_search/errs-test-other-beam_size_4-epoch-40-avg-10-modified_beam_search-beam-size-4.txt
2022-05-28 00:56:54,228 INFO [decode.py:453] 
For test-other, WER of different settings are:
beam_size_4	17.79	best for test-other

2022-05-28 00:56:54,228 INFO [decode.py:630] Done!