decoding-results/modified_beam_search/log-decode-epoch-30-avg-1-modified_beam_search-beam-size-4-use-averaged-model-2023-03-21-13-44-53

2023-03-21 13:44:53,273 INFO [decode.py:690] Decoding started
2023-03-21 13:44:53,274 INFO [decode.py:696] Device: cuda:0
2023-03-21 13:44:53,281 INFO [decode.py:706] {'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, 'warm_step': 2000, 'env_info': {'k2-version': '1.22', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': '96c9a2aece2a3a7633da07740e24fa3d96f5498c', 'k2-git-date': 'Thu Nov 10 08:14:02 2022', 'lhotse-version': '1.13.0.dev+git.527d964.clean', 'torch-version': '1.12.1', 'torch-cuda-available': True, 'torch-cuda-version': '11.6', 'python-version': '3.8', 'icefall-git-branch': 'zipformer_libri_small_models', 'icefall-git-sha1': 'd3145cd-dirty', 'icefall-git-date': 'Thu Feb 16 15:24:55 2023', 'icefall-path': '/ceph-data4/yangxiaoyu/softwares/icefall_development/icefall_small_models', 'k2-path': '/ceph-data4/yangxiaoyu/softwares/anaconda3/envs/k2_latest/lib/python3.8/site-packages/k2/__init__.py', 'lhotse-path': '/ceph-data4/yangxiaoyu/softwares/lhotse_development/lhotse_random_padding_left/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-10-0221105906-5745685d6b-t8zzx', 'IP address': '10.177.57.19'}, 'epoch': 30, 'iter': 0, 'avg': 1, 'use_averaged_model': True, 'exp_dir': PosixPath('pruned_transducer_stateless7_streaming_multi/exp-small-6M'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'lang_dir': PosixPath('data/lang_bpe_500'), 'decoding_method': 'modified_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, 'num_paths': 200, 'nbest_scale': 0.5, 'simulate_streaming': False, 'decode_chunk_size': 16, 'left_context': 64, 'num_encoder_layers': '2,2,2,2,2', 'feedforward_dims': '256,256,512,512,256', 'nhead': '4,4,4,4,4', 'encoder_dims': '128,128,128,128,128', 'attention_dims': '96,96,96,96,96', 'encoder_unmasked_dims': '96,96,96,96,96', 'zipformer_downsampling_factors': '1,2,4,8,2', 'cnn_module_kernels': '31,31,31,31,31', 'decoder_dim': 512, 'joiner_dim': 512, 'short_chunk_size': 50, 'num_left_chunks': 4, 'decode_chunk_len': 32, '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('pruned_transducer_stateless7_streaming_multi/exp-small-6M/modified_beam_search'), 'suffix': 'epoch-30-avg-1-modified_beam_search-beam-size-4-use-averaged-model', 'blank_id': 0, 'unk_id': 2, 'vocab_size': 500}
2023-03-21 13:44:53,281 INFO [decode.py:708] About to create model
2023-03-21 13:44:53,427 INFO [zipformer.py:405] At encoder stack 4, which has downsampling_factor=2, we will combine the outputs of layers 1 and 3, with downsampling_factors=2 and 8.
2023-03-21 13:44:53,435 INFO [train.py:536] Use giga
2023-03-21 13:44:53,438 INFO [decode.py:779] Calculating the averaged model over epoch range from 29 (excluded) to 30
2023-03-21 13:44:55,938 INFO [decode.py:813] Number of model parameters: 6061029
2023-03-21 13:44:55,938 INFO [librispeech.py:58] About to get test-clean cuts from data/fbank/librispeech_cuts_test-clean.jsonl.gz
2023-03-21 13:44:55,947 INFO [librispeech.py:63] About to get test-other cuts from data/fbank/librispeech_cuts_test-other.jsonl.gz
2023-03-21 13:45:03,046 INFO [decode.py:592] batch 0/?, cuts processed until now is 26
2023-03-21 13:46:18,846 INFO [decode.py:592] batch 20/?, cuts processed until now is 1545
2023-03-21 13:47:05,883 INFO [decode.py:592] batch 40/?, cuts processed until now is 2375
2023-03-21 13:47:07,579 INFO [zipformer.py:2441] attn_weights_entropy = tensor([1.4154, 1.2778, 1.1482, 1.5772], device='cuda:0'), covar=tensor([0.0857, 0.0405, 0.0402, 0.0997], device='cuda:0'), in_proj_covar=tensor([0.0195, 0.0123, 0.0121, 0.0232], device='cuda:0'), out_proj_covar=tensor([0.0107, 0.0076, 0.0067, 0.0117], device='cuda:0')
2023-03-21 13:47:32,569 INFO [decode.py:608] The transcripts are stored in pruned_transducer_stateless7_streaming_multi/exp-small-6M/modified_beam_search/recogs-test-clean-beam_size_4-epoch-30-avg-1-modified_beam_search-beam-size-4-use-averaged-model.txt
2023-03-21 13:47:32,668 INFO [utils.py:558] [test-clean-beam_size_4] %WER 5.79% [3044 / 52576, 344 ins, 283 del, 2417 sub ]
2023-03-21 13:47:32,999 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless7_streaming_multi/exp-small-6M/modified_beam_search/errs-test-clean-beam_size_4-epoch-30-avg-1-modified_beam_search-beam-size-4-use-averaged-model.txt
2023-03-21 13:47:33,006 INFO [decode.py:637] 
For test-clean, WER of different settings are:
beam_size_4	5.79	best for test-clean

2023-03-21 13:47:38,189 INFO [decode.py:592] batch 0/?, cuts processed until now is 30
2023-03-21 13:47:52,805 INFO [zipformer.py:2441] attn_weights_entropy = tensor([2.2750, 1.3069, 3.4282, 3.1751], device='cuda:0'), covar=tensor([0.1717, 0.2875, 0.0558, 0.0988], device='cuda:0'), in_proj_covar=tensor([0.0816, 0.0679, 0.1016, 0.0992], device='cuda:0'), out_proj_covar=tensor([0.0011, 0.0010, 0.0011, 0.0012], device='cuda:0')
2023-03-21 13:48:55,576 INFO [decode.py:592] batch 20/?, cuts processed until now is 1771
2023-03-21 13:49:08,575 INFO [zipformer.py:2441] attn_weights_entropy = tensor([2.2201, 1.2812, 3.5021, 3.1816], device='cuda:0'), covar=tensor([0.1809, 0.2933, 0.0553, 0.1024], device='cuda:0'), in_proj_covar=tensor([0.0816, 0.0679, 0.1016, 0.0992], device='cuda:0'), out_proj_covar=tensor([0.0011, 0.0010, 0.0011, 0.0012], device='cuda:0')
2023-03-21 13:49:39,149 INFO [decode.py:592] batch 40/?, cuts processed until now is 2696
2023-03-21 13:50:02,629 INFO [decode.py:608] The transcripts are stored in pruned_transducer_stateless7_streaming_multi/exp-small-6M/modified_beam_search/recogs-test-other-beam_size_4-epoch-30-avg-1-modified_beam_search-beam-size-4-use-averaged-model.txt
2023-03-21 13:50:02,738 INFO [utils.py:558] [test-other-beam_size_4] %WER 14.38% [7525 / 52343, 869 ins, 749 del, 5907 sub ]
2023-03-21 13:50:02,957 INFO [decode.py:621] Wrote detailed error stats to pruned_transducer_stateless7_streaming_multi/exp-small-6M/modified_beam_search/errs-test-other-beam_size_4-epoch-30-avg-1-modified_beam_search-beam-size-4-use-averaged-model.txt
2023-03-21 13:50:02,958 INFO [decode.py:637] 
For test-other, WER of different settings are:
beam_size_4	14.38	best for test-other

2023-03-21 13:50:02,958 INFO [decode.py:845] Done!