add files

data/lang_bpe_500/bpe.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c53433de083c4a6ad12d034550ef22de68cec62c4f58932a7b6b8b2f1e743fa5
+size 244865

decode_results/fast_beam_search/log-decode-epoch-25-avg-3-streaming-chunk-size-16-left-context-32-beam-20.0-max-contexts-8-max-states-64-use-averaged-model-2022-06-26-17-36-52

@@ -0,0 +1,25 @@
+2022-06-26 17:36:52,448 INFO [decode.py:706] Decoding started
+2022-06-26 17:36:52,448 INFO [decode.py:712] Device: cuda:0
+2022-06-26 17:36:52,451 INFO [decode.py:727] {'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': '1.15.1', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'f8d2dba06c000ffee36aab5b66f24e7c9809f116', 'k2-git-date': 'Thu Apr 21 12:20:34 2022', 'lhotse-version': '1.1.0', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'streaming-conformer', 'icefall-git-sha1': 'af80a46-clean', 'icefall-git-date': 'Sat Jun 25 18:24:36 2022', 'icefall-path': '/ceph-kw/kangwei/code/icefall_streaming4', 'k2-path': '/ceph-hw/kangwei/code/k2_release/k2/k2/python/k2/__init__.py', 'lhotse-path': '/ceph-hw/kangwei/dev_tools/anaconda3/envs/rnnt2/lib/python3.8/site-packages/lhotse-1.1.0-py3.8.egg/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-7-0616225511-78bf4545d8-tv52r', 'IP address': '10.177.77.9'}, 'epoch': 25, 'iter': 0, 'avg': 3, 'use_averaged_model': True, 'exp_dir': PosixPath('pruned_transducer_stateless4/exp'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'lang_dir': PosixPath('data/lang_bpe_500'), '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, 'simulate_streaming': True, 'decode_chunk_size': 16, 'left_context': 32, 'num_paths': 200, 'nbest_scale': 0.5, 'dynamic_chunk_training': False, 'causal_convolution': True, 'short_chunk_size': 25, 'num_left_chunks': 4, 'full_libri': True, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 1000, '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_stateless4/exp/fast_beam_search'), 'suffix': 'epoch-25-avg-3-streaming-chunk-size-16-left-context-32-beam-20.0-max-contexts-8-max-states-64-use-averaged-model', 'blank_id': 0, 'unk_id': 2, 'vocab_size': 500}
+2022-06-26 17:36:52,452 INFO [decode.py:729] About to create model
+2022-06-26 17:36:52,899 INFO [decode.py:796] Calculating the averaged model over epoch range from 22 (excluded) to 25
+2022-06-26 17:37:00,769 INFO [decode.py:832] Number of model parameters: 78648040
+2022-06-26 17:37:00,769 INFO [asr_datamodule.py:444] About to get test-clean cuts
+2022-06-26 17:37:00,772 INFO [asr_datamodule.py:451] About to get test-other cuts
+2022-06-26 17:37:03,947 INFO [decode.py:605] batch 0/?, cuts processed until now is 79
+2022-06-26 17:37:40,301 INFO [decode.py:622] The transcripts are stored in pruned_transducer_stateless4/exp/fast_beam_search/recogs-test-clean-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-16-left-context-32-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-06-26 17:37:40,386 INFO [utils.py:418] [test-clean-beam_20.0_max_contexts_8_max_states_64] %WER 3.46% [1817 / 52576, 203 ins, 150 del, 1464 sub ]
+2022-06-26 17:37:40,612 INFO [decode.py:635] Wrote detailed error stats to pruned_transducer_stateless4/exp/fast_beam_search/errs-test-clean-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-16-left-context-32-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-06-26 17:37:40,613 INFO [decode.py:652] 
+For test-clean, WER of different settings are:
+beam_20.0_max_contexts_8_max_states_64	3.46	best for test-clean
+
+2022-06-26 17:37:43,271 INFO [decode.py:605] batch 0/?, cuts processed until now is 96
+2022-06-26 17:38:19,507 INFO [decode.py:622] The transcripts are stored in pruned_transducer_stateless4/exp/fast_beam_search/recogs-test-other-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-16-left-context-32-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-06-26 17:38:19,594 INFO [utils.py:418] [test-other-beam_20.0_max_contexts_8_max_states_64] %WER 8.91% [4665 / 52343, 468 ins, 443 del, 3754 sub ]
+2022-06-26 17:38:19,905 INFO [decode.py:635] Wrote detailed error stats to pruned_transducer_stateless4/exp/fast_beam_search/errs-test-other-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-16-left-context-32-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-06-26 17:38:19,906 INFO [decode.py:652] 
+For test-other, WER of different settings are:
+beam_20.0_max_contexts_8_max_states_64	8.91	best for test-other
+
+2022-06-26 17:38:19,906 INFO [decode.py:861] Done!

decode_results/fast_beam_search/log-decode-epoch-25-avg-3-streaming-chunk-size-16-left-context-64-beam-20.0-max-contexts-8-max-states-64-use-averaged-model-2022-06-26-17-38-23

@@ -0,0 +1,25 @@
+2022-06-26 17:38:23,296 INFO [decode.py:706] Decoding started
+2022-06-26 17:38:23,297 INFO [decode.py:712] Device: cuda:0
+2022-06-26 17:38:23,300 INFO [decode.py:727] {'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': '1.15.1', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'f8d2dba06c000ffee36aab5b66f24e7c9809f116', 'k2-git-date': 'Thu Apr 21 12:20:34 2022', 'lhotse-version': '1.1.0', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'streaming-conformer', 'icefall-git-sha1': 'af80a46-clean', 'icefall-git-date': 'Sat Jun 25 18:24:36 2022', 'icefall-path': '/ceph-kw/kangwei/code/icefall_streaming4', 'k2-path': '/ceph-hw/kangwei/code/k2_release/k2/k2/python/k2/__init__.py', 'lhotse-path': '/ceph-hw/kangwei/dev_tools/anaconda3/envs/rnnt2/lib/python3.8/site-packages/lhotse-1.1.0-py3.8.egg/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-7-0616225511-78bf4545d8-tv52r', 'IP address': '10.177.77.9'}, 'epoch': 25, 'iter': 0, 'avg': 3, 'use_averaged_model': True, 'exp_dir': PosixPath('pruned_transducer_stateless4/exp'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'lang_dir': PosixPath('data/lang_bpe_500'), '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, 'simulate_streaming': True, 'decode_chunk_size': 16, 'left_context': 64, 'num_paths': 200, 'nbest_scale': 0.5, 'dynamic_chunk_training': False, 'causal_convolution': True, 'short_chunk_size': 25, 'num_left_chunks': 4, 'full_libri': True, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 1000, '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_stateless4/exp/fast_beam_search'), 'suffix': 'epoch-25-avg-3-streaming-chunk-size-16-left-context-64-beam-20.0-max-contexts-8-max-states-64-use-averaged-model', 'blank_id': 0, 'unk_id': 2, 'vocab_size': 500}
+2022-06-26 17:38:23,301 INFO [decode.py:729] About to create model
+2022-06-26 17:38:23,762 INFO [decode.py:796] Calculating the averaged model over epoch range from 22 (excluded) to 25
+2022-06-26 17:38:31,410 INFO [decode.py:832] Number of model parameters: 78648040
+2022-06-26 17:38:31,410 INFO [asr_datamodule.py:444] About to get test-clean cuts
+2022-06-26 17:38:31,413 INFO [asr_datamodule.py:451] About to get test-other cuts
+2022-06-26 17:38:34,869 INFO [decode.py:605] batch 0/?, cuts processed until now is 79
+2022-06-26 17:39:16,844 INFO [decode.py:622] The transcripts are stored in pruned_transducer_stateless4/exp/fast_beam_search/recogs-test-clean-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-16-left-context-64-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-06-26 17:39:16,988 INFO [utils.py:418] [test-clean-beam_20.0_max_contexts_8_max_states_64] %WER 3.43% [1802 / 52576, 204 ins, 157 del, 1441 sub ]
+2022-06-26 17:39:17,351 INFO [decode.py:635] Wrote detailed error stats to pruned_transducer_stateless4/exp/fast_beam_search/errs-test-clean-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-16-left-context-64-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-06-26 17:39:17,351 INFO [decode.py:652] 
+For test-clean, WER of different settings are:
+beam_20.0_max_contexts_8_max_states_64	3.43	best for test-clean
+
+2022-06-26 17:39:20,359 INFO [decode.py:605] batch 0/?, cuts processed until now is 96
+2022-06-26 17:40:00,464 INFO [decode.py:622] The transcripts are stored in pruned_transducer_stateless4/exp/fast_beam_search/recogs-test-other-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-16-left-context-64-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-06-26 17:40:00,552 INFO [utils.py:418] [test-other-beam_20.0_max_contexts_8_max_states_64] %WER 8.78% [4598 / 52343, 463 ins, 429 del, 3706 sub ]
+2022-06-26 17:40:00,853 INFO [decode.py:635] Wrote detailed error stats to pruned_transducer_stateless4/exp/fast_beam_search/errs-test-other-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-16-left-context-64-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-06-26 17:40:00,854 INFO [decode.py:652] 
+For test-other, WER of different settings are:
+beam_20.0_max_contexts_8_max_states_64	8.78	best for test-other
+
+2022-06-26 17:40:00,854 INFO [decode.py:861] Done!

decode_results/fast_beam_search/log-decode-epoch-25-avg-3-streaming-chunk-size-2-left-context-32-beam-20.0-max-contexts-8-max-states-64-use-averaged-model-2022-06-26-17-27-32

@@ -0,0 +1,25 @@
+2022-06-26 17:27:32,253 INFO [decode.py:706] Decoding started
+2022-06-26 17:27:32,253 INFO [decode.py:712] Device: cuda:0
+2022-06-26 17:27:32,256 INFO [decode.py:727] {'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': '1.15.1', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'f8d2dba06c000ffee36aab5b66f24e7c9809f116', 'k2-git-date': 'Thu Apr 21 12:20:34 2022', 'lhotse-version': '1.1.0', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'streaming-conformer', 'icefall-git-sha1': 'af80a46-clean', 'icefall-git-date': 'Sat Jun 25 18:24:36 2022', 'icefall-path': '/ceph-kw/kangwei/code/icefall_streaming4', 'k2-path': '/ceph-hw/kangwei/code/k2_release/k2/k2/python/k2/__init__.py', 'lhotse-path': '/ceph-hw/kangwei/dev_tools/anaconda3/envs/rnnt2/lib/python3.8/site-packages/lhotse-1.1.0-py3.8.egg/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-7-0616225511-78bf4545d8-tv52r', 'IP address': '10.177.77.9'}, 'epoch': 25, 'iter': 0, 'avg': 3, 'use_averaged_model': True, 'exp_dir': PosixPath('pruned_transducer_stateless4/exp'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'lang_dir': PosixPath('data/lang_bpe_500'), '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, 'simulate_streaming': True, 'decode_chunk_size': 2, 'left_context': 32, 'num_paths': 200, 'nbest_scale': 0.5, 'dynamic_chunk_training': False, 'causal_convolution': True, 'short_chunk_size': 25, 'num_left_chunks': 4, 'full_libri': True, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 1000, '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_stateless4/exp/fast_beam_search'), 'suffix': 'epoch-25-avg-3-streaming-chunk-size-2-left-context-32-beam-20.0-max-contexts-8-max-states-64-use-averaged-model', 'blank_id': 0, 'unk_id': 2, 'vocab_size': 500}
+2022-06-26 17:27:32,256 INFO [decode.py:729] About to create model
+2022-06-26 17:27:32,890 INFO [decode.py:796] Calculating the averaged model over epoch range from 22 (excluded) to 25
+2022-06-26 17:27:43,503 INFO [decode.py:832] Number of model parameters: 78648040
+2022-06-26 17:27:43,503 INFO [asr_datamodule.py:444] About to get test-clean cuts
+2022-06-26 17:27:43,506 INFO [asr_datamodule.py:451] About to get test-other cuts
+2022-06-26 17:27:46,783 INFO [decode.py:605] batch 0/?, cuts processed until now is 79
+2022-06-26 17:28:23,830 INFO [decode.py:622] The transcripts are stored in pruned_transducer_stateless4/exp/fast_beam_search/recogs-test-clean-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-2-left-context-32-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-06-26 17:28:23,916 INFO [utils.py:418] [test-clean-beam_20.0_max_contexts_8_max_states_64] %WER 3.90% [2048 / 52576, 223 ins, 181 del, 1644 sub ]
+2022-06-26 17:28:24,144 INFO [decode.py:635] Wrote detailed error stats to pruned_transducer_stateless4/exp/fast_beam_search/errs-test-clean-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-2-left-context-32-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-06-26 17:28:24,145 INFO [decode.py:652] 
+For test-clean, WER of different settings are:
+beam_20.0_max_contexts_8_max_states_64	3.9	best for test-clean
+
+2022-06-26 17:28:26,917 INFO [decode.py:605] batch 0/?, cuts processed until now is 96
+2022-06-26 17:29:03,101 INFO [decode.py:622] The transcripts are stored in pruned_transducer_stateless4/exp/fast_beam_search/recogs-test-other-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-2-left-context-32-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-06-26 17:29:03,191 INFO [utils.py:418] [test-other-beam_20.0_max_contexts_8_max_states_64] %WER 10.09% [5281 / 52343, 490 ins, 602 del, 4189 sub ]
+2022-06-26 17:29:03,506 INFO [decode.py:635] Wrote detailed error stats to pruned_transducer_stateless4/exp/fast_beam_search/errs-test-other-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-2-left-context-32-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-06-26 17:29:03,506 INFO [decode.py:652] 
+For test-other, WER of different settings are:
+beam_20.0_max_contexts_8_max_states_64	10.09	best for test-other
+
+2022-06-26 17:29:03,506 INFO [decode.py:861] Done!

decode_results/fast_beam_search/log-decode-epoch-25-avg-3-streaming-chunk-size-2-left-context-64-beam-20.0-max-contexts-8-max-states-64-use-averaged-model-2022-06-26-17-29-06

@@ -0,0 +1,25 @@
+2022-06-26 17:29:06,750 INFO [decode.py:706] Decoding started
+2022-06-26 17:29:06,751 INFO [decode.py:712] Device: cuda:0
+2022-06-26 17:29:06,753 INFO [decode.py:727] {'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': '1.15.1', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'f8d2dba06c000ffee36aab5b66f24e7c9809f116', 'k2-git-date': 'Thu Apr 21 12:20:34 2022', 'lhotse-version': '1.1.0', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'streaming-conformer', 'icefall-git-sha1': 'af80a46-clean', 'icefall-git-date': 'Sat Jun 25 18:24:36 2022', 'icefall-path': '/ceph-kw/kangwei/code/icefall_streaming4', 'k2-path': '/ceph-hw/kangwei/code/k2_release/k2/k2/python/k2/__init__.py', 'lhotse-path': '/ceph-hw/kangwei/dev_tools/anaconda3/envs/rnnt2/lib/python3.8/site-packages/lhotse-1.1.0-py3.8.egg/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-7-0616225511-78bf4545d8-tv52r', 'IP address': '10.177.77.9'}, 'epoch': 25, 'iter': 0, 'avg': 3, 'use_averaged_model': True, 'exp_dir': PosixPath('pruned_transducer_stateless4/exp'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'lang_dir': PosixPath('data/lang_bpe_500'), '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, 'simulate_streaming': True, 'decode_chunk_size': 2, 'left_context': 64, 'num_paths': 200, 'nbest_scale': 0.5, 'dynamic_chunk_training': False, 'causal_convolution': True, 'short_chunk_size': 25, 'num_left_chunks': 4, 'full_libri': True, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 1000, '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_stateless4/exp/fast_beam_search'), 'suffix': 'epoch-25-avg-3-streaming-chunk-size-2-left-context-64-beam-20.0-max-contexts-8-max-states-64-use-averaged-model', 'blank_id': 0, 'unk_id': 2, 'vocab_size': 500}
+2022-06-26 17:29:06,754 INFO [decode.py:729] About to create model
+2022-06-26 17:29:07,201 INFO [decode.py:796] Calculating the averaged model over epoch range from 22 (excluded) to 25
+2022-06-26 17:29:14,338 INFO [decode.py:832] Number of model parameters: 78648040
+2022-06-26 17:29:14,338 INFO [asr_datamodule.py:444] About to get test-clean cuts
+2022-06-26 17:29:14,342 INFO [asr_datamodule.py:451] About to get test-other cuts
+2022-06-26 17:29:17,485 INFO [decode.py:605] batch 0/?, cuts processed until now is 79
+2022-06-26 17:29:55,400 INFO [decode.py:622] The transcripts are stored in pruned_transducer_stateless4/exp/fast_beam_search/recogs-test-clean-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-2-left-context-64-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-06-26 17:29:55,484 INFO [utils.py:418] [test-clean-beam_20.0_max_contexts_8_max_states_64] %WER 3.82% [2006 / 52576, 218 ins, 186 del, 1602 sub ]
+2022-06-26 17:29:55,710 INFO [decode.py:635] Wrote detailed error stats to pruned_transducer_stateless4/exp/fast_beam_search/errs-test-clean-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-2-left-context-64-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-06-26 17:29:55,711 INFO [decode.py:652] 
+For test-clean, WER of different settings are:
+beam_20.0_max_contexts_8_max_states_64	3.82	best for test-clean
+
+2022-06-26 17:29:58,371 INFO [decode.py:605] batch 0/?, cuts processed until now is 96
+2022-06-26 17:30:35,595 INFO [decode.py:622] The transcripts are stored in pruned_transducer_stateless4/exp/fast_beam_search/recogs-test-other-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-2-left-context-64-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-06-26 17:30:35,687 INFO [utils.py:418] [test-other-beam_20.0_max_contexts_8_max_states_64] %WER 10.03% [5248 / 52343, 477 ins, 608 del, 4163 sub ]
+2022-06-26 17:30:36,001 INFO [decode.py:635] Wrote detailed error stats to pruned_transducer_stateless4/exp/fast_beam_search/errs-test-other-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-2-left-context-64-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-06-26 17:30:36,002 INFO [decode.py:652] 
+For test-other, WER of different settings are:
+beam_20.0_max_contexts_8_max_states_64	10.03	best for test-other
+
+2022-06-26 17:30:36,002 INFO [decode.py:861] Done!

decode_results/fast_beam_search/log-decode-epoch-25-avg-3-streaming-chunk-size-4-left-context-32-beam-20.0-max-contexts-8-max-states-64-use-averaged-model-2022-06-26-17-30-39

@@ -0,0 +1,25 @@
+2022-06-26 17:30:39,392 INFO [decode.py:706] Decoding started
+2022-06-26 17:30:39,392 INFO [decode.py:712] Device: cuda:0
+2022-06-26 17:30:39,395 INFO [decode.py:727] {'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': '1.15.1', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'f8d2dba06c000ffee36aab5b66f24e7c9809f116', 'k2-git-date': 'Thu Apr 21 12:20:34 2022', 'lhotse-version': '1.1.0', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'streaming-conformer', 'icefall-git-sha1': 'af80a46-clean', 'icefall-git-date': 'Sat Jun 25 18:24:36 2022', 'icefall-path': '/ceph-kw/kangwei/code/icefall_streaming4', 'k2-path': '/ceph-hw/kangwei/code/k2_release/k2/k2/python/k2/__init__.py', 'lhotse-path': '/ceph-hw/kangwei/dev_tools/anaconda3/envs/rnnt2/lib/python3.8/site-packages/lhotse-1.1.0-py3.8.egg/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-7-0616225511-78bf4545d8-tv52r', 'IP address': '10.177.77.9'}, 'epoch': 25, 'iter': 0, 'avg': 3, 'use_averaged_model': True, 'exp_dir': PosixPath('pruned_transducer_stateless4/exp'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'lang_dir': PosixPath('data/lang_bpe_500'), '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, 'simulate_streaming': True, 'decode_chunk_size': 4, 'left_context': 32, 'num_paths': 200, 'nbest_scale': 0.5, 'dynamic_chunk_training': False, 'causal_convolution': True, 'short_chunk_size': 25, 'num_left_chunks': 4, 'full_libri': True, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 1000, '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_stateless4/exp/fast_beam_search'), 'suffix': 'epoch-25-avg-3-streaming-chunk-size-4-left-context-32-beam-20.0-max-contexts-8-max-states-64-use-averaged-model', 'blank_id': 0, 'unk_id': 2, 'vocab_size': 500}
+2022-06-26 17:30:39,395 INFO [decode.py:729] About to create model
+2022-06-26 17:30:39,828 INFO [decode.py:796] Calculating the averaged model over epoch range from 22 (excluded) to 25
+2022-06-26 17:30:49,922 INFO [decode.py:832] Number of model parameters: 78648040
+2022-06-26 17:30:49,922 INFO [asr_datamodule.py:444] About to get test-clean cuts
+2022-06-26 17:30:49,925 INFO [asr_datamodule.py:451] About to get test-other cuts
+2022-06-26 17:30:53,211 INFO [decode.py:605] batch 0/?, cuts processed until now is 79
+2022-06-26 17:31:29,915 INFO [decode.py:622] The transcripts are stored in pruned_transducer_stateless4/exp/fast_beam_search/recogs-test-clean-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-4-left-context-32-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-06-26 17:31:30,000 INFO [utils.py:418] [test-clean-beam_20.0_max_contexts_8_max_states_64] %WER 3.69% [1941 / 52576, 222 ins, 160 del, 1559 sub ]
+2022-06-26 17:31:30,229 INFO [decode.py:635] Wrote detailed error stats to pruned_transducer_stateless4/exp/fast_beam_search/errs-test-clean-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-4-left-context-32-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-06-26 17:31:30,230 INFO [decode.py:652] 
+For test-clean, WER of different settings are:
+beam_20.0_max_contexts_8_max_states_64	3.69	best for test-clean
+
+2022-06-26 17:31:33,025 INFO [decode.py:605] batch 0/?, cuts processed until now is 96
+2022-06-26 17:32:08,911 INFO [decode.py:622] The transcripts are stored in pruned_transducer_stateless4/exp/fast_beam_search/recogs-test-other-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-4-left-context-32-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-06-26 17:32:09,001 INFO [utils.py:418] [test-other-beam_20.0_max_contexts_8_max_states_64] %WER 9.65% [5053 / 52343, 499 ins, 522 del, 4032 sub ]
+2022-06-26 17:32:09,316 INFO [decode.py:635] Wrote detailed error stats to pruned_transducer_stateless4/exp/fast_beam_search/errs-test-other-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-4-left-context-32-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-06-26 17:32:09,317 INFO [decode.py:652] 
+For test-other, WER of different settings are:
+beam_20.0_max_contexts_8_max_states_64	9.65	best for test-other
+
+2022-06-26 17:32:09,317 INFO [decode.py:861] Done!

decode_results/fast_beam_search/log-decode-epoch-25-avg-3-streaming-chunk-size-4-left-context-64-beam-20.0-max-contexts-8-max-states-64-use-averaged-model-2022-06-26-17-32-12

@@ -0,0 +1,25 @@
+2022-06-26 17:32:12,692 INFO [decode.py:706] Decoding started
+2022-06-26 17:32:12,692 INFO [decode.py:712] Device: cuda:0
+2022-06-26 17:32:12,695 INFO [decode.py:727] {'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': '1.15.1', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'f8d2dba06c000ffee36aab5b66f24e7c9809f116', 'k2-git-date': 'Thu Apr 21 12:20:34 2022', 'lhotse-version': '1.1.0', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'streaming-conformer', 'icefall-git-sha1': 'af80a46-clean', 'icefall-git-date': 'Sat Jun 25 18:24:36 2022', 'icefall-path': '/ceph-kw/kangwei/code/icefall_streaming4', 'k2-path': '/ceph-hw/kangwei/code/k2_release/k2/k2/python/k2/__init__.py', 'lhotse-path': '/ceph-hw/kangwei/dev_tools/anaconda3/envs/rnnt2/lib/python3.8/site-packages/lhotse-1.1.0-py3.8.egg/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-7-0616225511-78bf4545d8-tv52r', 'IP address': '10.177.77.9'}, 'epoch': 25, 'iter': 0, 'avg': 3, 'use_averaged_model': True, 'exp_dir': PosixPath('pruned_transducer_stateless4/exp'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'lang_dir': PosixPath('data/lang_bpe_500'), '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, 'simulate_streaming': True, 'decode_chunk_size': 4, 'left_context': 64, 'num_paths': 200, 'nbest_scale': 0.5, 'dynamic_chunk_training': False, 'causal_convolution': True, 'short_chunk_size': 25, 'num_left_chunks': 4, 'full_libri': True, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 1000, '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_stateless4/exp/fast_beam_search'), 'suffix': 'epoch-25-avg-3-streaming-chunk-size-4-left-context-64-beam-20.0-max-contexts-8-max-states-64-use-averaged-model', 'blank_id': 0, 'unk_id': 2, 'vocab_size': 500}
+2022-06-26 17:32:12,695 INFO [decode.py:729] About to create model
+2022-06-26 17:32:13,134 INFO [decode.py:796] Calculating the averaged model over epoch range from 22 (excluded) to 25
+2022-06-26 17:32:20,466 INFO [decode.py:832] Number of model parameters: 78648040
+2022-06-26 17:32:20,466 INFO [asr_datamodule.py:444] About to get test-clean cuts
+2022-06-26 17:32:20,470 INFO [asr_datamodule.py:451] About to get test-other cuts
+2022-06-26 17:32:23,663 INFO [decode.py:605] batch 0/?, cuts processed until now is 79
+2022-06-26 17:33:01,450 INFO [decode.py:622] The transcripts are stored in pruned_transducer_stateless4/exp/fast_beam_search/recogs-test-clean-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-4-left-context-64-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-06-26 17:33:01,536 INFO [utils.py:418] [test-clean-beam_20.0_max_contexts_8_max_states_64] %WER 3.67% [1932 / 52576, 223 ins, 165 del, 1544 sub ]
+2022-06-26 17:33:01,764 INFO [decode.py:635] Wrote detailed error stats to pruned_transducer_stateless4/exp/fast_beam_search/errs-test-clean-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-4-left-context-64-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-06-26 17:33:01,765 INFO [decode.py:652] 
+For test-clean, WER of different settings are:
+beam_20.0_max_contexts_8_max_states_64	3.67	best for test-clean
+
+2022-06-26 17:33:04,501 INFO [decode.py:605] batch 0/?, cuts processed until now is 96
+2022-06-26 17:33:41,488 INFO [decode.py:622] The transcripts are stored in pruned_transducer_stateless4/exp/fast_beam_search/recogs-test-other-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-4-left-context-64-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-06-26 17:33:41,577 INFO [utils.py:418] [test-other-beam_20.0_max_contexts_8_max_states_64] %WER 9.56% [5006 / 52343, 494 ins, 511 del, 4001 sub ]
+2022-06-26 17:33:41,881 INFO [decode.py:635] Wrote detailed error stats to pruned_transducer_stateless4/exp/fast_beam_search/errs-test-other-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-4-left-context-64-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-06-26 17:33:41,881 INFO [decode.py:652] 
+For test-other, WER of different settings are:
+beam_20.0_max_contexts_8_max_states_64	9.56	best for test-other
+
+2022-06-26 17:33:41,881 INFO [decode.py:861] Done!

decode_results/fast_beam_search/log-decode-epoch-25-avg-3-streaming-chunk-size-8-left-context-32-beam-20.0-max-contexts-8-max-states-64-use-averaged-model-2022-06-26-17-33-45

@@ -0,0 +1,25 @@
+2022-06-26 17:33:45,237 INFO [decode.py:706] Decoding started
+2022-06-26 17:33:45,237 INFO [decode.py:712] Device: cuda:0
+2022-06-26 17:33:45,240 INFO [decode.py:727] {'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': '1.15.1', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'f8d2dba06c000ffee36aab5b66f24e7c9809f116', 'k2-git-date': 'Thu Apr 21 12:20:34 2022', 'lhotse-version': '1.1.0', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'streaming-conformer', 'icefall-git-sha1': 'af80a46-clean', 'icefall-git-date': 'Sat Jun 25 18:24:36 2022', 'icefall-path': '/ceph-kw/kangwei/code/icefall_streaming4', 'k2-path': '/ceph-hw/kangwei/code/k2_release/k2/k2/python/k2/__init__.py', 'lhotse-path': '/ceph-hw/kangwei/dev_tools/anaconda3/envs/rnnt2/lib/python3.8/site-packages/lhotse-1.1.0-py3.8.egg/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-7-0616225511-78bf4545d8-tv52r', 'IP address': '10.177.77.9'}, 'epoch': 25, 'iter': 0, 'avg': 3, 'use_averaged_model': True, 'exp_dir': PosixPath('pruned_transducer_stateless4/exp'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'lang_dir': PosixPath('data/lang_bpe_500'), '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, 'simulate_streaming': True, 'decode_chunk_size': 8, 'left_context': 32, 'num_paths': 200, 'nbest_scale': 0.5, 'dynamic_chunk_training': False, 'causal_convolution': True, 'short_chunk_size': 25, 'num_left_chunks': 4, 'full_libri': True, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 1000, '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_stateless4/exp/fast_beam_search'), 'suffix': 'epoch-25-avg-3-streaming-chunk-size-8-left-context-32-beam-20.0-max-contexts-8-max-states-64-use-averaged-model', 'blank_id': 0, 'unk_id': 2, 'vocab_size': 500}
+2022-06-26 17:33:45,240 INFO [decode.py:729] About to create model
+2022-06-26 17:33:45,669 INFO [decode.py:796] Calculating the averaged model over epoch range from 22 (excluded) to 25
+2022-06-26 17:33:52,759 INFO [decode.py:832] Number of model parameters: 78648040
+2022-06-26 17:33:52,759 INFO [asr_datamodule.py:444] About to get test-clean cuts
+2022-06-26 17:33:52,761 INFO [asr_datamodule.py:451] About to get test-other cuts
+2022-06-26 17:33:55,891 INFO [decode.py:605] batch 0/?, cuts processed until now is 79
+2022-06-26 17:34:32,561 INFO [decode.py:622] The transcripts are stored in pruned_transducer_stateless4/exp/fast_beam_search/recogs-test-clean-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-8-left-context-32-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-06-26 17:34:32,643 INFO [utils.py:418] [test-clean-beam_20.0_max_contexts_8_max_states_64] %WER 3.58% [1881 / 52576, 220 ins, 171 del, 1490 sub ]
+2022-06-26 17:34:32,870 INFO [decode.py:635] Wrote detailed error stats to pruned_transducer_stateless4/exp/fast_beam_search/errs-test-clean-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-8-left-context-32-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-06-26 17:34:32,871 INFO [decode.py:652] 
+For test-clean, WER of different settings are:
+beam_20.0_max_contexts_8_max_states_64	3.58	best for test-clean
+
+2022-06-26 17:34:35,560 INFO [decode.py:605] batch 0/?, cuts processed until now is 96
+2022-06-26 17:35:11,004 INFO [decode.py:622] The transcripts are stored in pruned_transducer_stateless4/exp/fast_beam_search/recogs-test-other-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-8-left-context-32-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-06-26 17:35:11,095 INFO [utils.py:418] [test-other-beam_20.0_max_contexts_8_max_states_64] %WER 9.28% [4857 / 52343, 489 ins, 459 del, 3909 sub ]
+2022-06-26 17:35:11,405 INFO [decode.py:635] Wrote detailed error stats to pruned_transducer_stateless4/exp/fast_beam_search/errs-test-other-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-8-left-context-32-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-06-26 17:35:11,406 INFO [decode.py:652] 
+For test-other, WER of different settings are:
+beam_20.0_max_contexts_8_max_states_64	9.28	best for test-other
+
+2022-06-26 17:35:11,406 INFO [decode.py:861] Done!

decode_results/fast_beam_search/log-decode-epoch-25-avg-3-streaming-chunk-size-8-left-context-64-beam-20.0-max-contexts-8-max-states-64-use-averaged-model-2022-06-26-17-35-14

@@ -0,0 +1,25 @@
+2022-06-26 17:35:14,624 INFO [decode.py:706] Decoding started
+2022-06-26 17:35:14,625 INFO [decode.py:712] Device: cuda:0
+2022-06-26 17:35:14,628 INFO [decode.py:727] {'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': '1.15.1', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'f8d2dba06c000ffee36aab5b66f24e7c9809f116', 'k2-git-date': 'Thu Apr 21 12:20:34 2022', 'lhotse-version': '1.1.0', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'streaming-conformer', 'icefall-git-sha1': 'af80a46-clean', 'icefall-git-date': 'Sat Jun 25 18:24:36 2022', 'icefall-path': '/ceph-kw/kangwei/code/icefall_streaming4', 'k2-path': '/ceph-hw/kangwei/code/k2_release/k2/k2/python/k2/__init__.py', 'lhotse-path': '/ceph-hw/kangwei/dev_tools/anaconda3/envs/rnnt2/lib/python3.8/site-packages/lhotse-1.1.0-py3.8.egg/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-7-0616225511-78bf4545d8-tv52r', 'IP address': '10.177.77.9'}, 'epoch': 25, 'iter': 0, 'avg': 3, 'use_averaged_model': True, 'exp_dir': PosixPath('pruned_transducer_stateless4/exp'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'lang_dir': PosixPath('data/lang_bpe_500'), '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, 'simulate_streaming': True, 'decode_chunk_size': 8, 'left_context': 64, 'num_paths': 200, 'nbest_scale': 0.5, 'dynamic_chunk_training': False, 'causal_convolution': True, 'short_chunk_size': 25, 'num_left_chunks': 4, 'full_libri': True, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 1000, '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_stateless4/exp/fast_beam_search'), 'suffix': 'epoch-25-avg-3-streaming-chunk-size-8-left-context-64-beam-20.0-max-contexts-8-max-states-64-use-averaged-model', 'blank_id': 0, 'unk_id': 2, 'vocab_size': 500}
+2022-06-26 17:35:14,628 INFO [decode.py:729] About to create model
+2022-06-26 17:35:15,069 INFO [decode.py:796] Calculating the averaged model over epoch range from 22 (excluded) to 25
+2022-06-26 17:35:22,118 INFO [decode.py:832] Number of model parameters: 78648040
+2022-06-26 17:35:22,119 INFO [asr_datamodule.py:444] About to get test-clean cuts
+2022-06-26 17:35:22,121 INFO [asr_datamodule.py:451] About to get test-other cuts
+2022-06-26 17:35:25,615 INFO [decode.py:605] batch 0/?, cuts processed until now is 79
+2022-06-26 17:36:08,710 INFO [decode.py:622] The transcripts are stored in pruned_transducer_stateless4/exp/fast_beam_search/recogs-test-clean-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-8-left-context-64-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-06-26 17:36:08,797 INFO [utils.py:418] [test-clean-beam_20.0_max_contexts_8_max_states_64] %WER 3.51% [1845 / 52576, 218 ins, 166 del, 1461 sub ]
+2022-06-26 17:36:09,025 INFO [decode.py:635] Wrote detailed error stats to pruned_transducer_stateless4/exp/fast_beam_search/errs-test-clean-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-8-left-context-64-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-06-26 17:36:09,026 INFO [decode.py:652] 
+For test-clean, WER of different settings are:
+beam_20.0_max_contexts_8_max_states_64	3.51	best for test-clean
+
+2022-06-26 17:36:11,675 INFO [decode.py:605] batch 0/?, cuts processed until now is 96
+2022-06-26 17:36:48,630 INFO [decode.py:622] The transcripts are stored in pruned_transducer_stateless4/exp/fast_beam_search/recogs-test-other-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-8-left-context-64-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-06-26 17:36:48,715 INFO [utils.py:418] [test-other-beam_20.0_max_contexts_8_max_states_64] %WER 9.18% [4806 / 52343, 480 ins, 456 del, 3870 sub ]
+2022-06-26 17:36:49,010 INFO [decode.py:635] Wrote detailed error stats to pruned_transducer_stateless4/exp/fast_beam_search/errs-test-other-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-8-left-context-64-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-06-26 17:36:49,010 INFO [decode.py:652] 
+For test-other, WER of different settings are:
+beam_20.0_max_contexts_8_max_states_64	9.18	best for test-other
+
+2022-06-26 17:36:49,011 INFO [decode.py:861] Done!

decode_results/fast_beam_search/wer-summary-test-clean-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-16-left-context-32-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt

@@ -0,0 +1,2 @@
+settings	WER
+beam_20.0_max_contexts_8_max_states_64	3.46

decode_results/fast_beam_search/wer-summary-test-clean-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-16-left-context-64-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt

@@ -0,0 +1,2 @@
+settings	WER
+beam_20.0_max_contexts_8_max_states_64	3.43

decode_results/fast_beam_search/wer-summary-test-clean-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-2-left-context-32-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt

@@ -0,0 +1,2 @@
+settings	WER
+beam_20.0_max_contexts_8_max_states_64	3.9

decode_results/fast_beam_search/wer-summary-test-clean-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-2-left-context-64-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt

@@ -0,0 +1,2 @@
+settings	WER
+beam_20.0_max_contexts_8_max_states_64	3.82

decode_results/fast_beam_search/wer-summary-test-clean-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-4-left-context-32-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt

@@ -0,0 +1,2 @@
+settings	WER
+beam_20.0_max_contexts_8_max_states_64	3.69

decode_results/fast_beam_search/wer-summary-test-clean-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-4-left-context-64-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt

@@ -0,0 +1,2 @@
+settings	WER
+beam_20.0_max_contexts_8_max_states_64	3.67

decode_results/fast_beam_search/wer-summary-test-clean-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-8-left-context-32-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt

@@ -0,0 +1,2 @@
+settings	WER
+beam_20.0_max_contexts_8_max_states_64	3.58

decode_results/fast_beam_search/wer-summary-test-clean-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-8-left-context-64-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt

@@ -0,0 +1,2 @@
+settings	WER
+beam_20.0_max_contexts_8_max_states_64	3.51

decode_results/fast_beam_search/wer-summary-test-other-beam_20.0_max_contexts_8_max_states_64-epoch-25-avg-3-streaming-chunk-size-16-left-context-32-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt

@@ -0,0 +1,2 @@
+settings	WER
+beam_20.0_max_contexts_8_max_states_64	8.91