icefall-asr-librispeech-pruned-transducer-stateless-2022-03-12 / log /log-decode-epoch-42-avg-11-context-2-max-sym-per-frame-2-2022-03-12-10-56-46
csukuangfj's picture
Add decoding logs.
26d87f8
2022-03-12 10:56:46,784 INFO [decode.py:425] Decoding started
2022-03-12 10:56:46,784 INFO [decode.py:431] Device: cuda:0
2022-03-12 10:56:46,786 INFO [decode.py:440] {'feature_dim': 80, 'subsampling_factor': 4, 'attention_dim': 512, 'nhead': 8, 'dim_feedforward': 2048, 'num_encoder_layers': 12, 'vgg_frontend': False, 'embedding_dim': 512, 'env_info': {'k2-version': '1.13', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': '5ee082ea55f50e8bd42203ba266945ea5a236ab8', 'k2-git-date': 'Sun Feb 27 09:00:48 2022', 'lhotse-version': '1.0.0.dev+missing.version.file', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'modified-beam-search-for-pruned-rnnt', 'icefall-git-sha1': 'bd033de-clean', 'icefall-git-date': 'Sat Mar 12 10:42:25 2022', 'icefall-path': '/ceph-fj/fangjun/open-source-2/icefall-pruned-rnnt', 'k2-path': '/ceph-fj/fangjun/open-source-2/k2-rnnt-ali/k2/python/k2/__init__.py', 'lhotse-path': '/ceph-fj/fangjun/open-source-2/lhotse-rnnt-ali/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-3-0307202051-57dc848959-8tmmp', 'IP address': '10.177.24.138'}, 'epoch': 42, 'avg': 11, 'exp_dir': PosixPath('pruned_transducer_stateless/exp'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'decoding_method': 'greedy_search', 'beam_size': 4, 'context_size': 2, 'max_sym_per_frame': 2, 'full_libri': True, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 100, 'bucketing_sampler': True, 'num_buckets': 30, '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, 'res_dir': PosixPath('pruned_transducer_stateless/exp/greedy_search'), 'suffix': 'epoch-42-avg-11-context-2-max-sym-per-frame-2', 'blank_id': 0, 'vocab_size': 500}
2022-03-12 10:56:46,786 INFO [decode.py:442] About to create model
2022-03-12 10:56:47,289 INFO [decode.py:453] averaging ['pruned_transducer_stateless/exp/epoch-32.pt', 'pruned_transducer_stateless/exp/epoch-33.pt', 'pruned_transducer_stateless/exp/epoch-34.pt', 'pruned_transducer_stateless/exp/epoch-35.pt', 'pruned_transducer_stateless/exp/epoch-36.pt', 'pruned_transducer_stateless/exp/epoch-37.pt', 'pruned_transducer_stateless/exp/epoch-38.pt', 'pruned_transducer_stateless/exp/epoch-39.pt', 'pruned_transducer_stateless/exp/epoch-40.pt', 'pruned_transducer_stateless/exp/epoch-41.pt', 'pruned_transducer_stateless/exp/epoch-42.pt']
2022-03-12 10:57:53,821 INFO [decode.py:462] Number of model parameters: 84514780
2022-03-12 10:57:53,822 INFO [asr_datamodule.py:379] About to get test-clean cuts
2022-03-12 10:57:53,952 INFO [asr_datamodule.py:384] About to get test-other cuts
2022-03-12 10:57:55,219 INFO [decode.py:350] batch 0/?, cuts processed until now is 20
2022-03-12 10:59:24,003 INFO [decode.py:350] batch 100/?, cuts processed until now is 1406
2022-03-12 11:00:51,007 INFO [decode.py:350] batch 200/?, cuts processed until now is 2563
2022-03-12 11:00:59,257 INFO [decode.py:367] The transcripts are stored in pruned_transducer_stateless/exp/greedy_search/recogs-test-clean-greedy_search-epoch-42-avg-11-context-2-max-sym-per-frame-2.txt
2022-03-12 11:00:59,331 INFO [utils.py:406] [test-clean-greedy_search] %WER 2.62% [1380 / 52576, 147 ins, 115 del, 1118 sub ]
2022-03-12 11:00:59,498 INFO [decode.py:380] Wrote detailed error stats to pruned_transducer_stateless/exp/greedy_search/errs-test-clean-greedy_search-epoch-42-avg-11-context-2-max-sym-per-frame-2.txt
2022-03-12 11:00:59,499 INFO [decode.py:397]
For test-clean, WER of different settings are:
greedy_search 2.62 best for test-clean
2022-03-12 11:01:00,615 INFO [decode.py:350] batch 0/?, cuts processed until now is 23
2022-03-12 11:02:28,395 INFO [decode.py:350] batch 100/?, cuts processed until now is 1614
2022-03-12 11:03:52,897 INFO [decode.py:350] batch 200/?, cuts processed until now is 2899
2022-03-12 11:03:57,774 INFO [decode.py:367] The transcripts are stored in pruned_transducer_stateless/exp/greedy_search/recogs-test-other-greedy_search-epoch-42-avg-11-context-2-max-sym-per-frame-2.txt
2022-03-12 11:03:57,863 INFO [utils.py:406] [test-other-greedy_search] %WER 6.37% [3334 / 52343, 326 ins, 326 del, 2682 sub ]
2022-03-12 11:03:58,039 INFO [decode.py:380] Wrote detailed error stats to pruned_transducer_stateless/exp/greedy_search/errs-test-other-greedy_search-epoch-42-avg-11-context-2-max-sym-per-frame-2.txt
2022-03-12 11:03:58,040 INFO [decode.py:397]
For test-other, WER of different settings are:
greedy_search 6.37 best for test-other
2022-03-12 11:03:58,040 INFO [decode.py:489] Done!