decoding-results/modified_beam_search/log-decode-iter-472000-avg-18-modified_beam_search-beam-size-4-use-averaged-model-2022-09-01-14-27-47

2022-09-01 14:27:47,985 INFO [decode.py:663] Decoding started
2022-09-01 14:27:47,985 INFO [decode.py:669] Device: cuda:0
2022-09-01 14:27:47,987 INFO [decode.py:679] {'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, 'dim_feedforward': 2048, 'decoder_dim': 512, 'joiner_dim': 512, 'model_warm_step': 3000, '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': '1.3.0.dev+missing.version.file', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'lstm-giga-libri', 'icefall-git-sha1': 'e3128cb-dirty', 'icefall-git-date': 'Mon Aug 29 19:05:41 2022', 'icefall-path': '/k2-dev/fangjun/open-source/icefall-lstm-giga', 'k2-path': '/ceph-fj/fangjun/open-source-2/k2-multi-22/k2/python/k2/__init__.py', 'lhotse-path': '/ceph-fj/fangjun/open-source-2/lhotse-jsonl/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-0602201035-5fb6d86964-mclm7', 'IP address': '10.177.74.202'}, 'epoch': 30, 'iter': 472000, 'avg': 18, 'use_averaged_model': True, 'exp_dir': PosixPath('lstm_transducer_stateless2/exp'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'lang_dir': PosixPath('data/lang_bpe_500'), 'decoding_method': 'modified_beam_search', 'beam_size': 4, 'beam': 4.0, 'ngram_lm_scale': 0.01, 'max_contexts': 4, 'max_states': 8, 'context_size': 2, 'max_sym_per_frame': 1, 'num_paths': 200, 'nbest_scale': 0.5, 'num_encoder_layers': 12, 'encoder_dim': 512, 'rnn_hidden_size': 1024, 'aux_layer_period': 0, 'max_duration': 600, 'bucketing_sampler': True, 'num_buckets': 30, 'shuffle': True, 'return_cuts': True, 'num_workers': 2, 'on_the_fly_num_workers': 0, 'enable_spec_aug': True, 'spec_aug_time_warp_factor': 80, 'enable_musan': True, 'manifest_dir': PosixPath('data/fbank'), 'on_the_fly_feats': False, 'res_dir': PosixPath('lstm_transducer_stateless2/exp/modified_beam_search'), 'suffix': 'iter-472000-avg-18-modified_beam_search-beam-size-4-use-averaged-model', 'blank_id': 0, 'unk_id': 2, 'vocab_size': 500}
2022-09-01 14:27:47,987 INFO [decode.py:681] About to create model
2022-09-01 14:27:48,327 INFO [train.py:464] Disable giga
2022-09-01 14:27:48,337 INFO [decode.py:735] Calculating the averaged model over iteration checkpoints from lstm_transducer_stateless2/exp/checkpoint-436000.pt (excluded) to lstm_transducer_stateless2/exp/checkpoint-472000.pt
2022-09-01 14:27:53,879 INFO [decode.py:791] Number of model parameters: 84689496
2022-09-01 14:27:53,880 INFO [librispeech.py:58] About to get test-clean cuts from data/fbank/librispeech_cuts_test-clean.jsonl.gz
2022-09-01 14:27:53,882 INFO [librispeech.py:63] About to get test-other cuts from data/fbank/librispeech_cuts_test-other.jsonl.gz
2022-09-01 14:27:58,373 INFO [decode.py:565] batch 0/?, cuts processed until now is 27
2022-09-01 14:29:07,720 INFO [decode.py:565] batch 20/?, cuts processed until now is 1623
2022-09-01 14:29:43,529 INFO [decode.py:565] batch 40/?, cuts processed until now is 2468
2022-09-01 14:29:58,954 INFO [decode.py:583] The transcripts are stored in lstm_transducer_stateless2/exp/modified_beam_search/recogs-test-clean-beam_size_4-iter-472000-avg-18-modified_beam_search-beam-size-4-use-averaged-model.txt
2022-09-01 14:29:59,020 INFO [utils.py:428] [test-clean-beam_size_4] %WER 2.75% [1448 / 52576, 171 ins, 105 del, 1172 sub ]
2022-09-01 14:29:59,187 INFO [decode.py:596] Wrote detailed error stats to lstm_transducer_stateless2/exp/modified_beam_search/errs-test-clean-beam_size_4-iter-472000-avg-18-modified_beam_search-beam-size-4-use-averaged-model.txt
2022-09-01 14:29:59,188 INFO [decode.py:613] 
For test-clean, WER of different settings are:
beam_size_4	2.75	best for test-clean

2022-09-01 14:30:03,309 INFO [decode.py:565] batch 0/?, cuts processed until now is 31
2022-09-01 14:32:16,602 INFO [decode.py:565] batch 20/?, cuts processed until now is 1849
2022-09-01 14:32:52,559 INFO [decode.py:565] batch 40/?, cuts processed until now is 2785
2022-09-01 14:33:06,715 INFO [decode.py:583] The transcripts are stored in lstm_transducer_stateless2/exp/modified_beam_search/recogs-test-other-beam_size_4-iter-472000-avg-18-modified_beam_search-beam-size-4-use-averaged-model.txt
2022-09-01 14:33:06,783 INFO [utils.py:428] [test-other-beam_size_4] %WER 7.08% [3707 / 52343, 406 ins, 337 del, 2964 sub ]
2022-09-01 14:33:07,000 INFO [decode.py:596] Wrote detailed error stats to lstm_transducer_stateless2/exp/modified_beam_search/errs-test-other-beam_size_4-iter-472000-avg-18-modified_beam_search-beam-size-4-use-averaged-model.txt
2022-09-01 14:33:07,001 INFO [decode.py:613] 
For test-other, WER of different settings are:
beam_size_4	7.08	best for test-other

2022-09-01 14:33:07,001 INFO [decode.py:823] Done!