2022-05-14 11:07:31,479 INFO [decode.py:503] Decoding started
2022-05-14 11:07:31,480 INFO [decode.py:509] Device: cuda:0
2022-05-14 11:07:31,505 INFO [decode.py:518] {'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': '1.14', 'k2-build-type': 'Debug', 'k2-with-cuda': True, 'k2-git-sha1': '1b29f0a946f50186aaa82df46a59f492ade9692b', 'k2-git-date': 'Tue Apr 12 20:46:49 2022', 'lhotse-version': '1.1.0', 'torch-version': '1.10.1+cu111', 'torch-cuda-available': True, 'torch-cuda-version': '11.1', 'python-version': '3.8', 'icefall-git-branch': 'spgi', 'icefall-git-sha1': '4fc1638-dirty', 'icefall-git-date': 'Sat May 14 10:41:06 2022', 'icefall-path': '/exp/draj/mini_scale_2022/icefall', 'k2-path': '/exp/draj/mini_scale_2022/k2/k2/python/k2/__init__.py', 'lhotse-path': '/exp/draj/mini_scale_2022/lhotse/lhotse/__init__.py', 'hostname': 'r7n03', 'IP address': '10.1.7.3'}, 'epoch': 20, 'iter': 696000, 'avg': 10, 'exp_dir': PosixPath('pruned_transducer_stateless2/exp/v2'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'decoding_method': 'greedy_search', 'beam_size': 4, 'beam': 4, 'max_contexts': 4, 'max_states': 8, 'context_size': 2, 'max_sym_per_frame': 1, 'manifest_dir': PosixPath('data/manifests'), 'enable_musan': True, 'concatenate_cuts': False, 'duration_factor': 1.0, 'gap': 1.0, 'max_duration': 500, 'num_buckets': 30, 'on_the_fly_feats': False, 'shuffle': True, 'num_workers': 8, 'enable_spec_aug': True, 'spec_aug_time_warp_factor': 80, 'res_dir': PosixPath('pruned_transducer_stateless2/exp/v2/greedy_search'), 'suffix': 'iter-696000-avg-10-context-2-max-sym-per-frame-1', 'blank_id': 0, 'vocab_size': 500}
2022-05-14 11:07:31,505 INFO [decode.py:520] About to create model
2022-05-14 11:07:32,313 INFO [decode.py:537] averaging ['pruned_transducer_stateless2/exp/v2/checkpoint-696000.pt', 'pruned_transducer_stateless2/exp/v2/checkpoint-688000.pt', 'pruned_transducer_stateless2/exp/v2/checkpoint-680000.pt', 'pruned_transducer_stateless2/exp/v2/checkpoint-672000.pt', 'pruned_transducer_stateless2/exp/v2/checkpoint-664000.pt', 'pruned_transducer_stateless2/exp/v2/checkpoint-656000.pt', 'pruned_transducer_stateless2/exp/v2/checkpoint-648000.pt', 'pruned_transducer_stateless2/exp/v2/checkpoint-640000.pt', 'pruned_transducer_stateless2/exp/v2/checkpoint-632000.pt', 'pruned_transducer_stateless2/exp/v2/checkpoint-624000.pt']
2022-05-14 11:09:35,201 INFO [decode.py:562] Number of model parameters: 78648040
2022-05-14 11:09:35,201 INFO [asr_datamodule.py:337] About to get SPGISpeech dev cuts
2022-05-14 11:09:35,246 INFO [asr_datamodule.py:342] About to get SPGISpeech val cuts
2022-05-14 11:09:38,613 INFO [decode.py:391] batch 0/?, cuts processed until now is 45
2022-05-14 11:10:20,509 INFO [decode.py:409] The transcripts are stored in pruned_transducer_stateless2/exp/v2/greedy_search/recogs-dev-greedy_search-iter-696000-avg-10-context-2-max-sym-per-frame-1.txt
2022-05-14 11:10:20,641 INFO [utils.py:405] [dev-greedy_search] %WER 2.46% [2390 / 97304, 412 ins, 591 del, 1387 sub ]
2022-05-14 11:10:21,330 INFO [utils.py:405] [dev-greedy_search] %WER 1.02% [4704 / 458995, 1492 ins, 2118 del, 1094 sub ]
2022-05-14 11:10:22,401 INFO [decode.py:435] Wrote detailed error stats to pruned_transducer_stateless2/exp/v2/greedy_search/wers-dev-greedy_search-iter-696000-avg-10-context-2-max-sym-per-frame-1.txt
2022-05-14 11:10:22,402 INFO [decode.py:464] 
For dev, WER/CER of different settings are:
greedy_search	2.46	1.02	best for dev

2022-05-14 11:10:24,562 INFO [decode.py:391] batch 0/?, cuts processed until now is 45
2022-05-14 11:11:20,603 INFO [decode.py:391] batch 100/?, cuts processed until now is 5355
2022-05-14 11:12:15,899 INFO [decode.py:391] batch 200/?, cuts processed until now is 10993
2022-05-14 11:13:10,704 INFO [decode.py:391] batch 300/?, cuts processed until now is 16731
2022-05-14 11:14:05,718 INFO [decode.py:391] batch 400/?, cuts processed until now is 22285
2022-05-14 11:15:01,816 INFO [decode.py:391] batch 500/?, cuts processed until now is 27573
2022-05-14 11:15:56,329 INFO [decode.py:391] batch 600/?, cuts processed until now is 33145
2022-05-14 11:16:52,772 INFO [decode.py:391] batch 700/?, cuts processed until now is 38113
2022-05-14 11:17:08,932 INFO [decode.py:409] The transcripts are stored in pruned_transducer_stateless2/exp/v2/greedy_search/recogs-val-greedy_search-iter-696000-avg-10-context-2-max-sym-per-frame-1.txt
2022-05-14 11:17:10,213 INFO [utils.py:405] [val-greedy_search] %WER 2.40% [22728 / 946469, 3838 ins, 5532 del, 13358 sub ]
2022-05-14 11:17:16,382 INFO [utils.py:405] [val-greedy_search] %WER 0.99% [44273 / 4485920, 13602 ins, 20929 del, 9742 sub ]
2022-05-14 11:17:28,262 INFO [decode.py:435] Wrote detailed error stats to pruned_transducer_stateless2/exp/v2/greedy_search/wers-val-greedy_search-iter-696000-avg-10-context-2-max-sym-per-frame-1.txt
2022-05-14 11:17:28,264 INFO [decode.py:464] 
For val, WER/CER of different settings are:
greedy_search	2.4	0.99	best for val

2022-05-14 11:17:28,286 INFO [decode.py:590] Done!