add models

no_specaug_no_musan/epoch-150-avg-18.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bb8bdd433e9150e2b9e6c9ce436ec35c6eca4905288b99c63294828e83c66a47
+size 258643812

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-140-avg-14-use-averaged-model-2024-08-05-12-16-21

@@ -0,0 +1,38 @@
+2024-08-05 12:16:21,182 INFO [evaluate.py:213] Evaluation started
+2024-08-05 12:16:21,183 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 140, 'iter': 0, 'avg': 14, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-140-avg-14-use-averaged-model'}
+2024-08-05 12:16:21,183 INFO [evaluate.py:221] About to create model
+2024-08-05 12:16:21,858 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 126 (excluded) to 140
+2024-08-05 12:16:27,412 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-05 12:16:27,413 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-05 12:16:27,426 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-05 12:16:28,066 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-05 12:16:32,252 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-05 12:16:36,250 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-05 12:16:38,041 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.1972, 4.1963, 3.9358, 4.4532], device='cuda:0')
+2024-08-05 12:16:39,968 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.4707, 3.3144, 2.8956, 3.0848], device='cuda:0')
+2024-08-05 12:16:40,250 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-05 12:16:44,248 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-05 12:16:48,256 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-05 12:16:52,241 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-05 12:16:56,220 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-05 12:17:00,192 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-05 12:17:04,175 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-05 12:17:08,154 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-05 12:17:12,124 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-05 12:17:16,093 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-05 12:17:20,058 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-05 12:17:23,377 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.4174, 3.0005, 2.9699, 2.9819, 2.7483, 2.7206, 2.7553, 2.7024],
+       device='cuda:0')
+2024-08-05 12:17:24,034 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-05 12:17:25,402 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.2632, 4.1797, 4.6193, 3.9889], device='cuda:0')
+2024-08-05 12:17:25,758 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.7429, 2.5259, 2.6362, 2.5923, 2.5404, 2.1183, 2.8796, 2.3396],
+       device='cuda:0')
+2024-08-05 12:17:28,005 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-05 12:17:31,986 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-05 12:17:35,949 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-05 12:17:39,917 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-05 12:17:43,891 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-05 12:17:44,948 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([6.1322, 6.0327, 5.8728, 5.3873], device='cuda:0')
+2024-08-05 12:17:45,882 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-05 12:17:50,106 INFO [evaluate.py:338] mAP for audioset eval is: 0.4953719754000955
+2024-08-05 12:17:50,106 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-140-avg-16-use-averaged-model-2024-08-05-12-14-45

@@ -0,0 +1,41 @@
+2024-08-05 12:14:45,356 INFO [evaluate.py:213] Evaluation started
+2024-08-05 12:14:45,356 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 140, 'iter': 0, 'avg': 16, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-140-avg-16-use-averaged-model'}
+2024-08-05 12:14:45,356 INFO [evaluate.py:221] About to create model
+2024-08-05 12:14:46,008 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 124 (excluded) to 140
+2024-08-05 12:14:51,335 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-05 12:14:51,335 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-05 12:14:51,347 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-05 12:14:51,976 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-05 12:14:56,086 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-05 12:14:58,028 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.3175, 2.9843, 2.9006, 3.0747, 2.7918, 2.5831, 2.6854, 2.4254],
+       device='cuda:0')
+2024-08-05 12:14:58,812 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.4553, 3.1635, 2.8824, 3.1085], device='cuda:0')
+2024-08-05 12:15:00,084 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-05 12:15:00,865 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.2392, 4.1039, 4.4622, 3.9406], device='cuda:0')
+2024-08-05 12:15:04,099 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-05 12:15:08,074 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.6900, 3.0033, 2.6632, 2.9431, 0.9440, 2.6265, 2.4517, 2.1863],
+       device='cuda:0')
+2024-08-05 12:15:08,121 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-05 12:15:12,124 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-05 12:15:16,104 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-05 12:15:19,827 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6037, 3.1144, 2.9697, 3.1100], device='cuda:0')
+2024-08-05 12:15:20,100 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-05 12:15:24,085 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-05 12:15:28,076 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-05 12:15:32,058 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-05 12:15:36,036 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-05 12:15:40,018 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-05 12:15:44,001 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-05 12:15:47,972 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-05 12:15:51,946 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-05 12:15:55,328 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.2153, 4.2929, 4.0199, 4.5434], device='cuda:0')
+2024-08-05 12:15:55,932 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-05 12:15:59,921 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-05 12:16:03,905 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-05 12:16:04,012 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.1760, 4.4085, 4.4394, 5.0269], device='cuda:0')
+2024-08-05 12:16:07,068 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.1975, 4.1968, 4.0280, 4.4636], device='cuda:0')
+2024-08-05 12:16:07,148 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([6.1271, 6.0480, 5.8890, 5.4013], device='cuda:0')
+2024-08-05 12:16:07,873 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-05 12:16:09,838 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-05 12:16:14,029 INFO [evaluate.py:338] mAP for audioset eval is: 0.49540243993758815
+2024-08-05 12:16:14,029 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-140-avg-18-use-averaged-model-2024-08-03-17-13-49

@@ -0,0 +1,44 @@
+2024-08-03 17:13:49,297 INFO [evaluate.py:213] Evaluation started
+2024-08-03 17:13:49,297 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 140, 'iter': 0, 'avg': 18, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-140-avg-18-use-averaged-model'}
+2024-08-03 17:13:49,297 INFO [evaluate.py:221] About to create model
+2024-08-03 17:13:49,955 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 122 (excluded) to 140
+2024-08-03 17:13:55,156 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-03 17:13:55,156 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-03 17:13:55,169 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-03 17:13:55,798 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-03 17:13:59,769 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-03 17:14:01,346 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6513, 2.6955, 2.8915, 3.3786], device='cuda:0')
+2024-08-03 17:14:03,797 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-03 17:14:07,819 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-03 17:14:08,375 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.8685, 2.9482, 2.8068, 3.0250, 2.5339, 2.7087, 2.5689, 2.5496],
+       device='cuda:0')
+2024-08-03 17:14:11,865 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-03 17:14:15,890 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-03 17:14:19,897 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-03 17:14:23,893 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-03 17:14:27,915 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-03 17:14:28,289 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6872, 3.5182, 3.5265, 3.2772], device='cuda:0')
+2024-08-03 17:14:31,878 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.7723, 2.6481, 2.6140, 2.6070, 2.5229, 2.2876, 2.8765, 2.1863],
+       device='cuda:0')
+2024-08-03 17:14:31,939 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-03 17:14:33,253 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.3813, 3.7167, 3.3018, 3.5505], device='cuda:0')
+2024-08-03 17:14:35,964 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-03 17:14:40,011 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-03 17:14:41,762 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.7446, 2.4907, 2.5816, 2.5119, 2.5523, 2.4023, 2.9202, 2.1411],
+       device='cuda:0')
+2024-08-03 17:14:44,032 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-03 17:14:48,024 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-03 17:14:49,757 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.4278, 3.0891, 2.9398, 3.1603, 2.8855, 2.7726, 2.8783, 2.5982],
+       device='cuda:0')
+2024-08-03 17:14:52,038 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-03 17:14:56,048 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-03 17:15:00,027 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-03 17:15:03,598 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6682, 3.4346, 3.3812, 3.4713], device='cuda:0')
+2024-08-03 17:15:04,052 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-03 17:15:08,043 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-03 17:15:09,988 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.7575, 2.5538, 2.6221, 2.5403, 2.4896, 2.2727, 2.8657, 2.4205],
+       device='cuda:0')
+2024-08-03 17:15:12,031 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-03 17:15:14,061 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-03 17:15:18,284 INFO [evaluate.py:338] mAP for audioset eval is: 0.4953638025152282
+2024-08-03 17:15:18,284 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-140-avg-20-use-averaged-model-2024-08-03-17-12-13

@@ -0,0 +1,35 @@
+2024-08-03 17:12:13,438 INFO [evaluate.py:213] Evaluation started
+2024-08-03 17:12:13,438 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 140, 'iter': 0, 'avg': 20, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-140-avg-20-use-averaged-model'}
+2024-08-03 17:12:13,438 INFO [evaluate.py:221] About to create model
+2024-08-03 17:12:14,070 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 120 (excluded) to 140
+2024-08-03 17:12:19,222 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-03 17:12:19,222 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-03 17:12:19,248 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-03 17:12:19,960 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-03 17:12:24,169 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-03 17:12:28,191 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-03 17:12:31,669 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.8876, 5.8256, 5.6792, 5.3440], device='cuda:0')
+2024-08-03 17:12:32,202 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-03 17:12:36,209 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-03 17:12:40,212 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-03 17:12:44,208 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-03 17:12:48,203 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-03 17:12:52,205 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-03 17:12:56,222 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-03 17:13:00,214 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-03 17:13:04,205 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-03 17:13:08,183 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-03 17:13:12,175 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-03 17:13:16,178 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-03 17:13:17,647 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([6.1138, 6.0267, 5.8441, 5.2048], device='cuda:0')
+2024-08-03 17:13:20,167 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-03 17:13:23,112 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.4019, 2.9898, 2.9343, 3.0354, 2.9048, 2.7597, 2.7956, 2.6496],
+       device='cuda:0')
+2024-08-03 17:13:24,168 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-03 17:13:28,146 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-03 17:13:30,544 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.1993, 4.0618, 3.9321, 4.4060], device='cuda:0')
+2024-08-03 17:13:32,151 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-03 17:13:36,143 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-03 17:13:38,158 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-03 17:13:42,396 INFO [evaluate.py:338] mAP for audioset eval is: 0.4954264337603019
+2024-08-03 17:13:42,396 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-140-avg-22-use-averaged-model-2024-08-03-17-10-38

@@ -0,0 +1,42 @@
+2024-08-03 17:10:38,274 INFO [evaluate.py:213] Evaluation started
+2024-08-03 17:10:38,274 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 140, 'iter': 0, 'avg': 22, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-140-avg-22-use-averaged-model'}
+2024-08-03 17:10:38,275 INFO [evaluate.py:221] About to create model
+2024-08-03 17:10:38,889 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 118 (excluded) to 140
+2024-08-03 17:10:44,034 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-03 17:10:44,035 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-03 17:10:44,049 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-03 17:10:44,669 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-03 17:10:48,701 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-03 17:10:52,728 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-03 17:10:56,741 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-03 17:11:00,754 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-03 17:11:04,732 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-03 17:11:08,703 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-03 17:11:10,293 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.1940, 4.1631, 3.9947, 4.4751], device='cuda:0')
+2024-08-03 17:11:10,823 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.5857, 3.2848, 3.1542, 3.2490], device='cuda:0')
+2024-08-03 17:11:12,685 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-03 17:11:15,558 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.0638, 4.7856, 4.5535, 4.9425], device='cuda:0')
+2024-08-03 17:11:16,651 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-03 17:11:16,773 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.5824, 3.3699, 3.0921, 3.2589], device='cuda:0')
+2024-08-03 17:11:20,614 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-03 17:11:24,579 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-03 17:11:28,540 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-03 17:11:32,513 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-03 17:11:36,478 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-03 17:11:38,630 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6729, 2.9674, 3.1197, 3.6051], device='cuda:0')
+2024-08-03 17:11:40,455 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-03 17:11:44,413 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-03 17:11:48,393 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-03 17:11:49,501 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.3740, 3.6672, 3.3461, 3.5702], device='cuda:0')
+2024-08-03 17:11:52,363 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-03 17:11:54,077 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6071, 3.1243, 2.9067, 3.1026], device='cuda:0')
+2024-08-03 17:11:55,688 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.8772, 3.0390, 2.7833, 2.9102, 2.4863, 2.7452, 2.6053, 2.6245],
+       device='cuda:0')
+2024-08-03 17:11:56,330 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-03 17:11:56,702 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6792, 3.6919, 3.4350, 3.3792], device='cuda:0')
+2024-08-03 17:11:59,049 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.7782, 2.4856, 2.5672, 2.6036, 2.4324, 2.2837, 2.8777, 2.2503],
+       device='cuda:0')
+2024-08-03 17:12:00,297 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-03 17:12:02,280 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-03 17:12:06,366 INFO [evaluate.py:338] mAP for audioset eval is: 0.4954528017331534
+2024-08-03 17:12:06,367 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-140-avg-24-use-averaged-model-2024-08-03-17-09-02

@@ -0,0 +1,35 @@
+2024-08-03 17:09:02,704 INFO [evaluate.py:213] Evaluation started
+2024-08-03 17:09:02,705 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 140, 'iter': 0, 'avg': 24, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-140-avg-24-use-averaged-model'}
+2024-08-03 17:09:02,705 INFO [evaluate.py:221] About to create model
+2024-08-03 17:09:03,343 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 116 (excluded) to 140
+2024-08-03 17:09:08,849 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-03 17:09:08,849 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-03 17:09:08,874 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-03 17:09:09,509 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-03 17:09:13,557 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-03 17:09:17,559 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-03 17:09:19,877 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.3918, 3.7823, 3.2083, 3.5850], device='cuda:0')
+2024-08-03 17:09:21,559 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-03 17:09:23,260 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.0245, 4.3916, 4.4437, 4.8936], device='cuda:0')
+2024-08-03 17:09:25,556 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-03 17:09:29,554 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-03 17:09:33,553 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-03 17:09:37,534 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-03 17:09:41,510 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-03 17:09:45,481 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-03 17:09:49,443 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-03 17:09:53,397 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-03 17:09:57,352 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-03 17:10:01,304 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-03 17:10:03,985 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.1695, 4.5013, 4.5157, 5.0198], device='cuda:0')
+2024-08-03 17:10:05,264 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-03 17:10:09,216 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-03 17:10:13,165 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-03 17:10:17,115 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-03 17:10:21,060 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-03 17:10:22,580 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.7259, 2.4887, 2.6226, 2.5619, 2.4542, 2.2662, 2.8771, 2.2043],
+       device='cuda:0')
+2024-08-03 17:10:25,012 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-03 17:10:26,974 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-03 17:10:31,039 INFO [evaluate.py:338] mAP for audioset eval is: 0.4953338762999663
+2024-08-03 17:10:31,040 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-140-avg-26-use-averaged-model-2024-08-03-17-07-27

@@ -0,0 +1,32 @@
+2024-08-03 17:07:27,743 INFO [evaluate.py:213] Evaluation started
+2024-08-03 17:07:27,743 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 140, 'iter': 0, 'avg': 26, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-140-avg-26-use-averaged-model'}
+2024-08-03 17:07:27,743 INFO [evaluate.py:221] About to create model
+2024-08-03 17:07:28,399 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 114 (excluded) to 140
+2024-08-03 17:07:33,375 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-03 17:07:33,375 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-03 17:07:33,387 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-03 17:07:34,005 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-03 17:07:37,959 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-03 17:07:41,653 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.9303, 5.6789, 5.6734, 5.7140], device='cuda:0')
+2024-08-03 17:07:41,967 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-03 17:07:45,962 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-03 17:07:49,954 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-03 17:07:53,920 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-03 17:07:57,882 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-03 17:08:01,855 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-03 17:08:05,827 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-03 17:08:09,784 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-03 17:08:13,739 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-03 17:08:17,698 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-03 17:08:21,666 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-03 17:08:25,641 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-03 17:08:29,611 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-03 17:08:33,569 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-03 17:08:37,530 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-03 17:08:41,098 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6730, 3.5668, 3.4243, 3.2629], device='cuda:0')
+2024-08-03 17:08:41,522 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-03 17:08:45,479 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-03 17:08:49,455 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-03 17:08:51,467 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-03 17:08:55,633 INFO [evaluate.py:338] mAP for audioset eval is: 0.4954517662187094
+2024-08-03 17:08:55,633 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-140-avg-28-use-averaged-model-2024-08-03-17-05-52

@@ -0,0 +1,39 @@
+2024-08-03 17:05:52,188 INFO [evaluate.py:213] Evaluation started
+2024-08-03 17:05:52,188 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 140, 'iter': 0, 'avg': 28, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-140-avg-28-use-averaged-model'}
+2024-08-03 17:05:52,188 INFO [evaluate.py:221] About to create model
+2024-08-03 17:05:52,838 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 112 (excluded) to 140
+2024-08-03 17:05:58,115 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-03 17:05:58,115 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-03 17:05:58,127 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-03 17:05:58,754 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-03 17:06:02,887 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-03 17:06:06,907 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-03 17:06:10,925 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-03 17:06:14,934 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-03 17:06:16,476 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.6047, 2.5282, 2.4678, 2.3822, 2.4014, 2.1456, 2.7848, 2.2641],
+       device='cuda:0')
+2024-08-03 17:06:18,929 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-03 17:06:22,906 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-03 17:06:26,880 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-03 17:06:30,839 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-03 17:06:34,805 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-03 17:06:35,514 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.3805, 3.7340, 3.4670, 3.6318], device='cuda:0')
+2024-08-03 17:06:38,787 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-03 17:06:42,751 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-03 17:06:46,486 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.8739, 2.9909, 2.7266, 3.0147, 2.4318, 2.8737, 2.6450, 2.6835],
+       device='cuda:0')
+2024-08-03 17:06:46,733 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-03 17:06:50,696 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-03 17:06:54,665 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-03 17:06:57,428 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.8812, 3.0325, 2.8090, 2.9391, 2.2232, 2.7702, 2.7030, 2.4763],
+       device='cuda:0')
+2024-08-03 17:06:58,665 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-03 17:07:02,639 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-03 17:07:04,975 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.8753, 2.9688, 2.7237, 2.9857, 2.3038, 2.6593, 2.5664, 2.6082],
+       device='cuda:0')
+2024-08-03 17:07:06,612 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-03 17:07:10,569 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-03 17:07:14,529 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-03 17:07:16,502 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-03 17:07:20,710 INFO [evaluate.py:338] mAP for audioset eval is: 0.49540708223047103
+2024-08-03 17:07:20,710 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-140-avg-30-use-averaged-model-2024-08-03-17-04-13

@@ -0,0 +1,44 @@
+2024-08-03 17:04:13,455 INFO [evaluate.py:213] Evaluation started
+2024-08-03 17:04:13,456 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 140, 'iter': 0, 'avg': 30, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-140-avg-30-use-averaged-model'}
+2024-08-03 17:04:13,456 INFO [evaluate.py:221] About to create model
+2024-08-03 17:04:14,194 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 110 (excluded) to 140
+2024-08-03 17:04:22,839 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-03 17:04:22,840 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-03 17:04:22,863 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-03 17:04:23,504 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-03 17:04:27,637 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-03 17:04:29,420 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6854, 3.6100, 3.4271, 3.3417], device='cuda:0')
+2024-08-03 17:04:31,647 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-03 17:04:33,594 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.7554, 3.0274, 2.6336, 2.9018, 0.8823, 2.5816, 2.3337, 2.2192],
+       device='cuda:0')
+2024-08-03 17:04:35,628 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-03 17:04:39,611 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-03 17:04:43,580 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-03 17:04:46,659 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.1723, 4.4627, 4.5303, 5.0152], device='cuda:0')
+2024-08-03 17:04:47,543 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-03 17:04:50,646 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.5984, 3.2958, 2.8195, 3.0006], device='cuda:0')
+2024-08-03 17:04:51,514 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-03 17:04:55,475 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-03 17:04:59,429 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-03 17:05:02,091 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.9338, 5.7336, 5.7171, 5.6731], device='cuda:0')
+2024-08-03 17:05:03,394 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-03 17:05:05,722 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.8661, 3.0290, 2.7724, 2.8615, 2.1821, 2.5600, 2.4415, 2.3920],
+       device='cuda:0')
+2024-08-03 17:05:07,349 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-03 17:05:11,308 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-03 17:05:15,282 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-03 17:05:15,859 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.2617, 4.0793, 4.5875, 4.1029], device='cuda:0')
+2024-08-03 17:05:17,928 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([6.1215, 6.0225, 5.8567, 5.3821], device='cuda:0')
+2024-08-03 17:05:19,242 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-03 17:05:23,196 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-03 17:05:27,149 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-03 17:05:29,260 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.3468, 2.9493, 2.9552, 3.0871, 2.8381, 2.7053, 2.8748, 2.7642],
+       device='cuda:0')
+2024-08-03 17:05:31,103 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-03 17:05:35,052 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-03 17:05:36,790 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.8794, 2.9864, 2.7087, 2.8776, 2.4011, 2.6972, 2.5277, 2.5453],
+       device='cuda:0')
+2024-08-03 17:05:39,020 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-03 17:05:41,002 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-03 17:05:45,192 INFO [evaluate.py:338] mAP for audioset eval is: 0.49540547716218697
+2024-08-03 17:05:45,192 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-142-avg-14-use-averaged-model-2024-08-05-12-13-09

@@ -0,0 +1,38 @@
+2024-08-05 12:13:09,379 INFO [evaluate.py:213] Evaluation started
+2024-08-05 12:13:09,380 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 142, 'iter': 0, 'avg': 14, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-142-avg-14-use-averaged-model'}
+2024-08-05 12:13:09,380 INFO [evaluate.py:221] About to create model
+2024-08-05 12:13:10,051 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 128 (excluded) to 142
+2024-08-05 12:13:15,559 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-05 12:13:15,560 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-05 12:13:15,574 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-05 12:13:16,207 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-05 12:13:20,474 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([6.1148, 5.9963, 5.8437, 5.2559], device='cuda:0')
+2024-08-05 12:13:20,604 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-05 12:13:24,615 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-05 12:13:27,036 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.2005, 4.1752, 3.9446, 4.4406], device='cuda:0')
+2024-08-05 12:13:28,661 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-05 12:13:32,683 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-05 12:13:36,006 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.3999, 3.6468, 3.2233, 3.5710], device='cuda:0')
+2024-08-05 12:13:36,703 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-05 12:13:40,722 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-05 12:13:44,707 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-05 12:13:48,689 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-05 12:13:52,690 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-05 12:13:56,704 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-05 12:14:00,685 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-05 12:14:04,624 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-05 12:14:06,304 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.0651, 4.8010, 4.5093, 4.9426], device='cuda:0')
+2024-08-05 12:14:08,074 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.9106, 5.7014, 5.7392, 5.5707], device='cuda:0')
+2024-08-05 12:14:08,581 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-05 12:14:10,858 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6026, 3.3303, 3.0362, 3.3527], device='cuda:0')
+2024-08-05 12:14:10,885 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.7336, 2.5490, 2.3241, 2.3543, 2.4712, 2.4643, 2.4387, 2.3264],
+       device='cuda:0')
+2024-08-05 12:14:12,503 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-05 12:14:16,454 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-05 12:14:20,395 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-05 12:14:24,323 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-05 12:14:28,241 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-05 12:14:32,167 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-05 12:14:34,176 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-05 12:14:38,306 INFO [evaluate.py:338] mAP for audioset eval is: 0.49560828444951577
+2024-08-05 12:14:38,306 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-142-avg-16-use-averaged-model-2024-08-05-12-11-34

@@ -0,0 +1,45 @@
+2024-08-05 12:11:34,509 INFO [evaluate.py:213] Evaluation started
+2024-08-05 12:11:34,509 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 142, 'iter': 0, 'avg': 16, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-142-avg-16-use-averaged-model'}
+2024-08-05 12:11:34,510 INFO [evaluate.py:221] About to create model
+2024-08-05 12:11:35,135 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 126 (excluded) to 142
+2024-08-05 12:11:40,241 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-05 12:11:40,242 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-05 12:11:40,255 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-05 12:11:40,901 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-05 12:11:45,006 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-05 12:11:48,991 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-05 12:11:49,507 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.1858, 3.5517, 3.2716, 3.5429], device='cuda:0')
+2024-08-05 12:11:51,331 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.7457, 2.3229, 2.2756, 2.3710, 2.4974, 2.3715, 2.5153, 2.3664],
+       device='cuda:0')
+2024-08-05 12:11:51,887 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.1881, 4.4788, 4.5326, 5.0245], device='cuda:0')
+2024-08-05 12:11:52,972 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-05 12:11:56,957 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-05 12:12:00,922 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-05 12:12:04,879 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-05 12:12:08,841 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-05 12:12:12,802 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-05 12:12:14,871 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.9386, 5.6776, 5.6934, 5.6137], device='cuda:0')
+2024-08-05 12:12:15,325 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.7730, 2.3937, 2.6136, 2.4645, 2.5213, 2.3236, 2.8531, 2.2467],
+       device='cuda:0')
+2024-08-05 12:12:16,773 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-05 12:12:20,098 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6715, 3.4990, 3.2822, 3.5169], device='cuda:0')
+2024-08-05 12:12:20,731 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-05 12:12:24,689 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-05 12:12:26,183 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.1816, 4.4432, 4.5552, 5.0365], device='cuda:0')
+2024-08-05 12:12:28,648 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-05 12:12:32,602 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-05 12:12:35,695 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.5993, 3.2790, 2.9471, 3.1890], device='cuda:0')
+2024-08-05 12:12:36,556 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-05 12:12:40,514 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-05 12:12:40,620 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.1875, 4.4189, 4.5628, 5.0477], device='cuda:0')
+2024-08-05 12:12:44,472 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-05 12:12:45,398 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.4228, 3.0727, 2.9432, 3.0214, 2.8516, 2.6376, 2.7981, 2.6150],
+       device='cuda:0')
+2024-08-05 12:12:48,430 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-05 12:12:52,382 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-05 12:12:56,333 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-05 12:12:58,164 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.6816, 2.9636, 2.6441, 2.8968, 0.9856, 2.8129, 2.3664, 2.2841],
+       device='cuda:0')
+2024-08-05 12:12:58,290 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-05 12:13:02,371 INFO [evaluate.py:338] mAP for audioset eval is: 0.49555501076204594
+2024-08-05 12:13:02,371 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-142-avg-18-use-averaged-model-2024-08-03-17-02-38

@@ -0,0 +1,35 @@
+2024-08-03 17:02:38,197 INFO [evaluate.py:213] Evaluation started
+2024-08-03 17:02:38,197 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 142, 'iter': 0, 'avg': 18, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-142-avg-18-use-averaged-model'}
+2024-08-03 17:02:38,198 INFO [evaluate.py:221] About to create model
+2024-08-03 17:02:38,851 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 124 (excluded) to 142
+2024-08-03 17:02:43,829 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-03 17:02:43,830 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-03 17:02:43,845 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-03 17:02:44,488 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-03 17:02:48,608 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-03 17:02:52,632 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-03 17:02:56,639 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-03 17:03:00,667 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-03 17:03:04,675 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-03 17:03:05,452 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6856, 3.6806, 3.4839, 3.4222], device='cuda:0')
+2024-08-03 17:03:08,675 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-03 17:03:12,652 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-03 17:03:16,623 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-03 17:03:20,592 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-03 17:03:24,570 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-03 17:03:28,544 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-03 17:03:32,527 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-03 17:03:34,297 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6834, 3.5699, 3.4347, 3.1232], device='cuda:0')
+2024-08-03 17:03:35,882 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6627, 2.9972, 3.1668, 3.5037], device='cuda:0')
+2024-08-03 17:03:36,511 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-03 17:03:38,642 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.7453, 2.5539, 2.3652, 2.2513, 2.5539, 2.4340, 2.5149, 2.2427],
+       device='cuda:0')
+2024-08-03 17:03:40,483 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-03 17:03:44,463 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-03 17:03:48,448 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-03 17:03:52,421 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-03 17:03:56,389 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-03 17:04:00,358 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-03 17:04:02,336 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-03 17:04:06,464 INFO [evaluate.py:338] mAP for audioset eval is: 0.4954569603198562
+2024-08-03 17:04:06,464 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-142-avg-20-use-averaged-model-2024-08-03-17-01-02

@@ -0,0 +1,39 @@
+2024-08-03 17:01:02,073 INFO [evaluate.py:213] Evaluation started
+2024-08-03 17:01:02,074 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 142, 'iter': 0, 'avg': 20, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-142-avg-20-use-averaged-model'}
+2024-08-03 17:01:02,074 INFO [evaluate.py:221] About to create model
+2024-08-03 17:01:02,718 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 122 (excluded) to 142
+2024-08-03 17:01:08,420 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-03 17:01:08,420 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-03 17:01:08,434 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-03 17:01:09,109 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-03 17:01:13,596 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-03 17:01:16,365 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.5248, 3.1644, 2.7088, 3.2612], device='cuda:0')
+2024-08-03 17:01:17,599 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-03 17:01:20,504 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.1846, 4.4740, 4.5335, 5.0196], device='cuda:0')
+2024-08-03 17:01:21,594 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-03 17:01:23,277 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.7867, 5.5623, 5.6068, 5.5148], device='cuda:0')
+2024-08-03 17:01:25,584 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-03 17:01:28,871 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.1657, 4.3777, 4.5368, 5.0276], device='cuda:0')
+2024-08-03 17:01:29,557 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-03 17:01:33,522 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-03 17:01:37,479 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-03 17:01:41,448 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-03 17:01:45,410 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-03 17:01:49,366 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-03 17:01:53,324 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-03 17:01:57,279 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-03 17:02:01,234 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-03 17:02:05,191 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-03 17:02:09,142 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-03 17:02:09,726 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.2280, 4.3475, 4.1260, 4.5041], device='cuda:0')
+2024-08-03 17:02:11,063 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.7341, 2.6939, 2.5924, 2.5307, 2.6329, 2.3741, 2.8942, 2.3398],
+       device='cuda:0')
+2024-08-03 17:02:13,097 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-03 17:02:16,547 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.9313, 5.6897, 5.7218, 5.6637], device='cuda:0')
+2024-08-03 17:02:17,056 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-03 17:02:21,011 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-03 17:02:22,507 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.3823, 3.7251, 3.4864, 3.6300], device='cuda:0')
+2024-08-03 17:02:24,960 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-03 17:02:26,928 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-03 17:02:31,060 INFO [evaluate.py:338] mAP for audioset eval is: 0.4955116176360669
+2024-08-03 17:02:31,060 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-142-avg-22-use-averaged-model-2024-08-03-16-59-25

@@ -0,0 +1,40 @@
+2024-08-03 16:59:25,935 INFO [evaluate.py:213] Evaluation started
+2024-08-03 16:59:25,936 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 142, 'iter': 0, 'avg': 22, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-142-avg-22-use-averaged-model'}
+2024-08-03 16:59:25,936 INFO [evaluate.py:221] About to create model
+2024-08-03 16:59:26,593 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 120 (excluded) to 142
+2024-08-03 16:59:31,862 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-03 16:59:31,862 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-03 16:59:31,875 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-03 16:59:32,501 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-03 16:59:36,692 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-03 16:59:38,012 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.1621, 4.3238, 4.5192, 5.0186], device='cuda:0')
+2024-08-03 16:59:38,049 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.7662, 2.4606, 2.6125, 2.4365, 2.5362, 2.3109, 2.8288, 2.2858],
+       device='cuda:0')
+2024-08-03 16:59:40,262 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.8772, 2.9915, 2.9160, 2.9494, 2.4503, 2.7932, 2.6748, 2.7354],
+       device='cuda:0')
+2024-08-03 16:59:40,706 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-03 16:59:43,881 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.7382, 2.5762, 2.5582, 2.5443, 2.4655, 2.3974, 2.8604, 2.1792],
+       device='cuda:0')
+2024-08-03 16:59:44,748 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-03 16:59:48,774 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-03 16:59:52,758 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-03 16:59:56,742 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-03 16:59:57,659 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.5961, 3.3785, 3.2552, 3.3564], device='cuda:0')
+2024-08-03 17:00:00,724 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-03 17:00:04,706 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-03 17:00:08,688 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-03 17:00:12,668 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-03 17:00:16,636 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-03 17:00:20,612 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-03 17:00:24,600 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-03 17:00:26,117 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6087, 3.2693, 2.9780, 3.0093], device='cuda:0')
+2024-08-03 17:00:28,603 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-03 17:00:32,585 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-03 17:00:34,255 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([6.1346, 6.0392, 5.8505, 5.4887], device='cuda:0')
+2024-08-03 17:00:36,563 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-03 17:00:40,560 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-03 17:00:44,548 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-03 17:00:48,532 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-03 17:00:50,506 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-03 17:00:54,653 INFO [evaluate.py:338] mAP for audioset eval is: 0.4955384987754096
+2024-08-03 17:00:54,653 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-142-avg-24-use-averaged-model-2024-08-03-16-57-50

@@ -0,0 +1,33 @@
+2024-08-03 16:57:50,125 INFO [evaluate.py:213] Evaluation started
+2024-08-03 16:57:50,125 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 142, 'iter': 0, 'avg': 24, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-142-avg-24-use-averaged-model'}
+2024-08-03 16:57:50,125 INFO [evaluate.py:221] About to create model
+2024-08-03 16:57:50,766 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 118 (excluded) to 142
+2024-08-03 16:57:55,988 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-03 16:57:55,988 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-03 16:57:56,002 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-03 16:57:56,628 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-03 16:58:00,648 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-03 16:58:04,663 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-03 16:58:06,037 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6672, 3.4298, 2.1943, 3.3347], device='cuda:0')
+2024-08-03 16:58:08,686 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-03 16:58:12,692 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-03 16:58:16,670 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-03 16:58:18,959 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.9376, 5.6813, 5.7228, 5.5803], device='cuda:0')
+2024-08-03 16:58:20,664 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-03 16:58:24,656 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-03 16:58:28,653 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-03 16:58:32,642 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-03 16:58:36,633 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-03 16:58:40,624 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-03 16:58:43,106 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.9413, 5.6827, 5.6972, 5.5713], device='cuda:0')
+2024-08-03 16:58:44,616 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-03 16:58:48,610 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-03 16:58:52,588 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-03 16:58:56,579 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-03 16:59:00,566 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-03 16:59:04,560 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-03 16:59:08,571 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-03 16:59:12,597 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-03 16:59:14,614 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-03 16:59:18,800 INFO [evaluate.py:338] mAP for audioset eval is: 0.4955021638347698
+2024-08-03 16:59:18,801 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-142-avg-26-use-averaged-model-2024-08-03-16-56-15

@@ -0,0 +1,37 @@
+2024-08-03 16:56:15,154 INFO [evaluate.py:213] Evaluation started
+2024-08-03 16:56:15,154 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 142, 'iter': 0, 'avg': 26, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-142-avg-26-use-averaged-model'}
+2024-08-03 16:56:15,154 INFO [evaluate.py:221] About to create model
+2024-08-03 16:56:15,780 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 116 (excluded) to 142
+2024-08-03 16:56:20,815 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-03 16:56:20,815 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-03 16:56:20,826 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-03 16:56:21,449 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-03 16:56:25,585 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-03 16:56:26,937 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.8660, 2.8799, 2.7226, 3.0114, 0.6019, 2.6425, 2.5438, 2.5049],
+       device='cuda:0')
+2024-08-03 16:56:29,596 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-03 16:56:32,738 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.7454, 2.5597, 2.1701, 2.5373, 2.5533, 2.3816, 2.3615, 2.2412],
+       device='cuda:0')
+2024-08-03 16:56:33,589 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-03 16:56:34,714 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.5963, 3.2193, 2.9263, 3.1413], device='cuda:0')
+2024-08-03 16:56:37,573 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-03 16:56:37,953 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.1197, 3.9383, 2.7340, 3.9171], device='cuda:0')
+2024-08-03 16:56:41,548 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-03 16:56:45,507 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-03 16:56:49,468 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-03 16:56:53,452 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-03 16:56:57,443 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-03 16:57:01,403 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-03 16:57:05,361 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-03 16:57:07,718 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.2541, 4.0062, 4.5016, 3.8257], device='cuda:0')
+2024-08-03 16:57:09,319 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-03 16:57:13,271 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-03 16:57:17,224 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-03 16:57:21,171 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-03 16:57:25,120 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-03 16:57:29,075 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-03 16:57:33,025 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-03 16:57:36,979 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-03 16:57:38,938 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-03 16:57:42,967 INFO [evaluate.py:338] mAP for audioset eval is: 0.49547168166533423
+2024-08-03 16:57:42,967 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-142-avg-28-use-averaged-model-2024-08-03-16-54-39

@@ -0,0 +1,37 @@
+2024-08-03 16:54:39,668 INFO [evaluate.py:213] Evaluation started
+2024-08-03 16:54:39,668 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 142, 'iter': 0, 'avg': 28, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-142-avg-28-use-averaged-model'}
+2024-08-03 16:54:39,668 INFO [evaluate.py:221] About to create model
+2024-08-03 16:54:40,295 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 114 (excluded) to 142
+2024-08-03 16:54:45,527 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-03 16:54:45,527 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-03 16:54:45,538 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-03 16:54:46,171 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-03 16:54:50,363 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-03 16:54:54,365 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-03 16:54:54,454 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.8106, 5.5617, 5.5647, 5.5923], device='cuda:0')
+2024-08-03 16:54:58,373 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-03 16:55:02,369 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-03 16:55:06,340 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-03 16:55:08,269 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.7516, 2.5714, 2.5726, 2.5714, 2.4525, 2.3192, 2.9149, 2.1216],
+       device='cuda:0')
+2024-08-03 16:55:10,311 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-03 16:55:14,267 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-03 16:55:16,618 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6800, 3.6551, 3.2524, 3.2222], device='cuda:0')
+2024-08-03 16:55:18,227 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-03 16:55:22,197 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-03 16:55:22,898 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.1875, 4.4095, 4.4959, 5.0335], device='cuda:0')
+2024-08-03 16:55:26,165 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-03 16:55:30,130 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-03 16:55:34,101 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-03 16:55:35,377 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.9141, 5.6846, 5.7110, 5.6330], device='cuda:0')
+2024-08-03 16:55:36,458 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6806, 3.5711, 3.5152, 3.2133], device='cuda:0')
+2024-08-03 16:55:38,069 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-03 16:55:42,036 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-03 16:55:46,004 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-03 16:55:49,972 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-03 16:55:53,935 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-03 16:55:57,899 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-03 16:56:01,869 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-03 16:56:03,829 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-03 16:56:07,859 INFO [evaluate.py:338] mAP for audioset eval is: 0.4954966041347471
+2024-08-03 16:56:07,859 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-142-avg-30-use-averaged-model-2024-08-03-16-52-56

@@ -0,0 +1,33 @@
+2024-08-03 16:52:56,794 INFO [evaluate.py:213] Evaluation started
+2024-08-03 16:52:56,794 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 142, 'iter': 0, 'avg': 30, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-142-avg-30-use-averaged-model'}
+2024-08-03 16:52:56,794 INFO [evaluate.py:221] About to create model
+2024-08-03 16:52:57,429 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 112 (excluded) to 142
+2024-08-03 16:53:10,142 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-03 16:53:10,143 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-03 16:53:10,157 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-03 16:53:10,824 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-03 16:53:14,919 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-03 16:53:18,924 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-03 16:53:22,922 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-03 16:53:24,510 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.0794, 3.8472, 2.7453, 3.7017], device='cuda:0')
+2024-08-03 16:53:26,931 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-03 16:53:28,101 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6782, 3.5393, 3.4097, 3.2710], device='cuda:0')
+2024-08-03 16:53:30,920 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-03 16:53:34,903 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-03 16:53:38,880 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-03 16:53:42,868 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-03 16:53:46,858 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-03 16:53:50,837 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-03 16:53:54,815 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-03 16:53:58,786 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-03 16:54:02,774 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-03 16:54:06,760 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-03 16:54:10,728 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-03 16:54:14,707 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-03 16:54:17,888 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.2429, 4.1024, 4.5717, 4.0765], device='cuda:0')
+2024-08-03 16:54:18,705 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-03 16:54:22,703 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-03 16:54:26,669 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-03 16:54:28,649 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-03 16:54:32,727 INFO [evaluate.py:338] mAP for audioset eval is: 0.4953973252515115
+2024-08-03 16:54:32,727 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-144-avg-14-use-averaged-model-2024-08-05-12-10-00

@@ -0,0 +1,32 @@
+2024-08-05 12:10:00,094 INFO [evaluate.py:213] Evaluation started
+2024-08-05 12:10:00,094 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 144, 'iter': 0, 'avg': 14, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-144-avg-14-use-averaged-model'}
+2024-08-05 12:10:00,095 INFO [evaluate.py:221] About to create model
+2024-08-05 12:10:00,718 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 130 (excluded) to 144
+2024-08-05 12:10:05,805 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-05 12:10:05,806 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-05 12:10:05,819 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-05 12:10:06,442 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-05 12:10:10,376 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-05 12:10:14,357 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-05 12:10:18,327 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-05 12:10:22,306 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-05 12:10:26,268 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-05 12:10:30,224 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-05 12:10:34,186 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-05 12:10:38,157 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-05 12:10:42,126 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-05 12:10:46,092 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-05 12:10:50,056 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-05 12:10:54,017 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-05 12:10:57,980 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-05 12:11:01,959 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-05 12:11:05,927 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-05 12:11:09,897 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-05 12:11:11,605 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.3949, 3.6432, 3.3924, 3.5383], device='cuda:0')
+2024-08-05 12:11:13,882 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-05 12:11:17,623 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6764, 3.3846, 3.2355, 3.4178], device='cuda:0')
+2024-08-05 12:11:17,862 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-05 12:11:21,820 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-05 12:11:23,783 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-05 12:11:27,859 INFO [evaluate.py:338] mAP for audioset eval is: 0.4957802861129404
+2024-08-05 12:11:27,859 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-144-avg-16-use-averaged-model-2024-08-05-12-08-24

@@ -0,0 +1,36 @@
+2024-08-05 12:08:24,941 INFO [evaluate.py:213] Evaluation started
+2024-08-05 12:08:24,941 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 144, 'iter': 0, 'avg': 16, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-144-avg-16-use-averaged-model'}
+2024-08-05 12:08:24,942 INFO [evaluate.py:221] About to create model
+2024-08-05 12:08:25,542 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 128 (excluded) to 144
+2024-08-05 12:08:30,878 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-05 12:08:30,878 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-05 12:08:30,891 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-05 12:08:31,544 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-05 12:08:35,667 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-05 12:08:39,663 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-05 12:08:43,652 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-05 12:08:47,641 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-05 12:08:51,613 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-05 12:08:55,585 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-05 12:08:59,540 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-05 12:09:03,500 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-05 12:09:03,587 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.9567, 5.6846, 5.7099, 5.5976], device='cuda:0')
+2024-08-05 12:09:07,464 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-05 12:09:09,760 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.5956, 3.0971, 3.2205, 3.3364], device='cuda:0')
+2024-08-05 12:09:11,421 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-05 12:09:12,791 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.2484, 4.1527, 4.6291, 4.1500], device='cuda:0')
+2024-08-05 12:09:15,379 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-05 12:09:19,338 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-05 12:09:23,293 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-05 12:09:27,252 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-05 12:09:31,216 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-05 12:09:35,177 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-05 12:09:39,135 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-05 12:09:40,261 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.4219, 2.9592, 2.8939, 3.0956, 2.8527, 2.7048, 2.9266, 2.6461],
+       device='cuda:0')
+2024-08-05 12:09:43,101 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-05 12:09:45,380 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.0505, 4.7954, 4.5756, 4.9377], device='cuda:0')
+2024-08-05 12:09:47,068 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-05 12:09:49,032 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-05 12:09:53,175 INFO [evaluate.py:338] mAP for audioset eval is: 0.49547797627154366
+2024-08-05 12:09:53,175 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-144-avg-18-use-averaged-model-2024-08-03-16-51-20

@@ -0,0 +1,36 @@
+2024-08-03 16:51:20,303 INFO [evaluate.py:213] Evaluation started
+2024-08-03 16:51:20,303 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 144, 'iter': 0, 'avg': 18, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-144-avg-18-use-averaged-model'}
+2024-08-03 16:51:20,303 INFO [evaluate.py:221] About to create model
+2024-08-03 16:51:20,945 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 126 (excluded) to 144
+2024-08-03 16:51:26,207 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-03 16:51:26,208 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-03 16:51:26,221 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-03 16:51:26,858 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-03 16:51:31,022 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-03 16:51:35,062 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-03 16:51:39,095 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-03 16:51:43,108 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-03 16:51:44,281 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.7545, 2.6808, 2.6675, 2.6386, 2.6105, 2.1422, 2.8838, 2.1167],
+       device='cuda:0')
+2024-08-03 16:51:47,141 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-03 16:51:48,321 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.2390, 4.1322, 4.6058, 3.9726], device='cuda:0')
+2024-08-03 16:51:51,153 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-03 16:51:55,162 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-03 16:51:59,139 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-03 16:52:03,131 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-03 16:52:03,262 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.5933, 3.2925, 3.0595, 3.0447], device='cuda:0')
+2024-08-03 16:52:04,634 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.0852, 4.8017, 4.5753, 4.9480], device='cuda:0')
+2024-08-03 16:52:07,132 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-03 16:52:10,582 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([6.1215, 6.0324, 5.8338, 5.3933], device='cuda:0')
+2024-08-03 16:52:11,113 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-03 16:52:15,103 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-03 16:52:19,103 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-03 16:52:23,100 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-03 16:52:27,085 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-03 16:52:31,058 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-03 16:52:35,042 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-03 16:52:39,025 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-03 16:52:43,009 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-03 16:52:45,038 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-03 16:52:49,204 INFO [evaluate.py:338] mAP for audioset eval is: 0.4955233751731442
+2024-08-03 16:52:49,204 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-144-avg-20-use-averaged-model-2024-08-03-16-49-44

@@ -0,0 +1,36 @@
+2024-08-03 16:49:44,660 INFO [evaluate.py:213] Evaluation started
+2024-08-03 16:49:44,660 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 144, 'iter': 0, 'avg': 20, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-144-avg-20-use-averaged-model'}
+2024-08-03 16:49:44,661 INFO [evaluate.py:221] About to create model
+2024-08-03 16:49:45,317 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 124 (excluded) to 144
+2024-08-03 16:49:50,727 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-03 16:49:50,727 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-03 16:49:50,739 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-03 16:49:51,379 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-03 16:49:55,629 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-03 16:49:59,652 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-03 16:50:03,646 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-03 16:50:06,755 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.4193, 3.7443, 3.4971, 3.7185], device='cuda:0')
+2024-08-03 16:50:07,635 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-03 16:50:10,528 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.1934, 4.3393, 4.5219, 5.0308], device='cuda:0')
+2024-08-03 16:50:11,624 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-03 16:50:15,601 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-03 16:50:19,572 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-03 16:50:23,537 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-03 16:50:27,504 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-03 16:50:31,470 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-03 16:50:35,436 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-03 16:50:39,402 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-03 16:50:43,375 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-03 16:50:47,380 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-03 16:50:51,353 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-03 16:50:53,693 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6705, 3.5476, 3.3155, 3.2292], device='cuda:0')
+2024-08-03 16:50:55,321 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-03 16:50:56,786 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([6.1203, 6.0400, 5.8450, 5.3255], device='cuda:0')
+2024-08-03 16:50:59,294 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-03 16:51:03,267 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-03 16:51:05,205 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.8689, 2.9861, 2.6712, 3.0121, 2.5062, 2.7520, 2.5930, 2.5583],
+       device='cuda:0')
+2024-08-03 16:51:07,235 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-03 16:51:09,199 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-03 16:51:13,279 INFO [evaluate.py:338] mAP for audioset eval is: 0.49554796034599485
+2024-08-03 16:51:13,279 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-144-avg-22-use-averaged-model-2024-08-03-16-48-09

@@ -0,0 +1,34 @@
+2024-08-03 16:48:09,479 INFO [evaluate.py:213] Evaluation started
+2024-08-03 16:48:09,479 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 144, 'iter': 0, 'avg': 22, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-144-avg-22-use-averaged-model'}
+2024-08-03 16:48:09,480 INFO [evaluate.py:221] About to create model
+2024-08-03 16:48:10,113 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 122 (excluded) to 144
+2024-08-03 16:48:15,131 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-03 16:48:15,132 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-03 16:48:15,143 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-03 16:48:15,780 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-03 16:48:19,761 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-03 16:48:22,271 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.1774, 4.3220, 4.5346, 5.0027], device='cuda:0')
+2024-08-03 16:48:23,762 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-03 16:48:26,230 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([6.1253, 6.0244, 5.8322, 5.4850], device='cuda:0')
+2024-08-03 16:48:27,747 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-03 16:48:31,733 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-03 16:48:35,704 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-03 16:48:39,674 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-03 16:48:43,637 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-03 16:48:47,596 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-03 16:48:47,727 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6068, 3.1524, 3.0380, 3.0874], device='cuda:0')
+2024-08-03 16:48:50,340 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6668, 3.5884, 3.3800, 3.4134], device='cuda:0')
+2024-08-03 16:48:51,624 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-03 16:48:55,620 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-03 16:48:59,590 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-03 16:49:03,578 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-03 16:49:07,541 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-03 16:49:11,508 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-03 16:49:15,471 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-03 16:49:19,441 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-03 16:49:23,409 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-03 16:49:27,367 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-03 16:49:31,330 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-03 16:49:33,292 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-03 16:49:37,411 INFO [evaluate.py:338] mAP for audioset eval is: 0.4955613493005715
+2024-08-03 16:49:37,411 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-144-avg-24-use-averaged-model-2024-08-03-16-46-34

@@ -0,0 +1,41 @@
+2024-08-03 16:46:34,482 INFO [evaluate.py:213] Evaluation started
+2024-08-03 16:46:34,482 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 144, 'iter': 0, 'avg': 24, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-144-avg-24-use-averaged-model'}
+2024-08-03 16:46:34,482 INFO [evaluate.py:221] About to create model
+2024-08-03 16:46:35,175 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 120 (excluded) to 144
+2024-08-03 16:46:40,364 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-03 16:46:40,364 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-03 16:46:40,376 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-03 16:46:41,016 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-03 16:46:45,097 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-03 16:46:49,092 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-03 16:46:53,067 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-03 16:46:57,058 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-03 16:47:00,026 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.2581, 4.1144, 4.5813, 4.0837], device='cuda:0')
+2024-08-03 16:47:01,037 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-03 16:47:05,005 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-03 16:47:06,385 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.2092, 4.3528, 4.0319, 4.5296], device='cuda:0')
+2024-08-03 16:47:08,975 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-03 16:47:10,057 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.9161, 5.6764, 5.7241, 5.6432], device='cuda:0')
+2024-08-03 16:47:11,747 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.2123, 4.2040, 3.9692, 4.5208], device='cuda:0')
+2024-08-03 16:47:12,948 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-03 16:47:16,915 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-03 16:47:17,096 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.2663, 4.0674, 4.4782, 4.0788], device='cuda:0')
+2024-08-03 16:47:17,666 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.8708, 3.1020, 2.7646, 3.0356, 2.4131, 2.7579, 2.6329, 2.5364],
+       device='cuda:0')
+2024-08-03 16:47:20,882 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-03 16:47:24,848 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-03 16:47:28,814 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-03 16:47:32,781 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-03 16:47:33,293 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.3988, 3.7095, 3.4756, 3.5958], device='cuda:0')
+2024-08-03 16:47:36,746 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-03 16:47:40,726 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-03 16:47:44,699 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-03 16:47:45,994 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.1782, 4.4358, 4.4194, 5.0423], device='cuda:0')
+2024-08-03 16:47:48,667 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-03 16:47:52,625 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-03 16:47:54,732 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.5969, 3.3586, 3.2031, 3.1662], device='cuda:0')
+2024-08-03 16:47:55,908 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.1898, 4.5394, 4.5082, 5.0375], device='cuda:0')
+2024-08-03 16:47:56,594 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-03 16:47:58,561 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-03 16:48:02,640 INFO [evaluate.py:338] mAP for audioset eval is: 0.4955825972771181
+2024-08-03 16:48:02,641 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-144-avg-26-use-averaged-model-2024-08-03-16-44-57

@@ -0,0 +1,39 @@
+2024-08-03 16:44:57,631 INFO [evaluate.py:213] Evaluation started
+2024-08-03 16:44:57,631 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 144, 'iter': 0, 'avg': 26, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-144-avg-26-use-averaged-model'}
+2024-08-03 16:44:57,631 INFO [evaluate.py:221] About to create model
+2024-08-03 16:44:58,274 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 118 (excluded) to 144
+2024-08-03 16:45:03,778 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-03 16:45:03,778 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-03 16:45:03,787 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-03 16:45:04,421 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-03 16:45:08,718 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-03 16:45:12,802 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-03 16:45:16,904 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-03 16:45:21,306 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-03 16:45:25,351 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-03 16:45:29,402 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-03 16:45:33,431 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-03 16:45:35,567 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.7610, 2.4928, 2.6504, 2.6388, 2.5720, 2.4782, 2.9229, 2.3410],
+       device='cuda:0')
+2024-08-03 16:45:37,428 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-03 16:45:41,461 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-03 16:45:42,208 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.7547, 2.6430, 2.3222, 2.4896, 2.5831, 2.4731, 2.4859, 2.2921],
+       device='cuda:0')
+2024-08-03 16:45:45,451 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-03 16:45:46,395 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.7697, 2.3544, 2.3032, 2.3921, 2.5826, 2.2967, 2.4785, 2.2608],
+       device='cuda:0')
+2024-08-03 16:45:49,429 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-03 16:45:53,397 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-03 16:45:57,361 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-03 16:46:00,206 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([6.1340, 6.0577, 5.9039, 5.4832], device='cuda:0')
+2024-08-03 16:46:01,328 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-03 16:46:05,301 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-03 16:46:09,270 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-03 16:46:13,244 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-03 16:46:17,213 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-03 16:46:19,830 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.1935, 4.2570, 4.0034, 4.4543], device='cuda:0')
+2024-08-03 16:46:20,531 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.0935, 4.8159, 4.6385, 4.9780], device='cuda:0')
+2024-08-03 16:46:21,227 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-03 16:46:23,209 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-03 16:46:27,321 INFO [evaluate.py:338] mAP for audioset eval is: 0.49560955098107645
+2024-08-03 16:46:27,321 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-144-avg-28-use-averaged-model-2024-08-03-16-43-21

@@ -0,0 +1,36 @@
+2024-08-03 16:43:21,066 INFO [evaluate.py:213] Evaluation started
+2024-08-03 16:43:21,066 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 144, 'iter': 0, 'avg': 28, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-144-avg-28-use-averaged-model'}
+2024-08-03 16:43:21,066 INFO [evaluate.py:221] About to create model
+2024-08-03 16:43:21,698 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 116 (excluded) to 144
+2024-08-03 16:43:26,737 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-03 16:43:26,737 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-03 16:43:26,747 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-03 16:43:27,374 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-03 16:43:31,413 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-03 16:43:35,412 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-03 16:43:39,399 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-03 16:43:43,468 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-03 16:43:47,449 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-03 16:43:51,415 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-03 16:43:55,404 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-03 16:43:59,395 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-03 16:44:03,418 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-03 16:44:05,176 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.7578, 2.5163, 2.1802, 2.5147, 2.4909, 2.3668, 2.5190, 2.4248],
+       device='cuda:0')
+2024-08-03 16:44:07,454 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-03 16:44:11,454 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-03 16:44:11,819 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.3652, 3.6473, 3.3666, 3.5613], device='cuda:0')
+2024-08-03 16:44:15,042 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.7345, 2.3793, 2.6292, 2.4911, 2.5730, 2.3209, 2.8312, 2.0853],
+       device='cuda:0')
+2024-08-03 16:44:15,497 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-03 16:44:19,472 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-03 16:44:23,439 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-03 16:44:27,401 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-03 16:44:31,366 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-03 16:44:34,237 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.0964, 4.7824, 4.5694, 4.9629], device='cuda:0')
+2024-08-03 16:44:35,330 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-03 16:44:39,292 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-03 16:44:43,256 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-03 16:44:45,221 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-03 16:44:49,321 INFO [evaluate.py:338] mAP for audioset eval is: 0.49559908715353496
+2024-08-03 16:44:49,321 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-144-avg-30-use-averaged-model-2024-08-03-16-41-38

@@ -0,0 +1,39 @@
+2024-08-03 16:41:38,742 INFO [evaluate.py:213] Evaluation started
+2024-08-03 16:41:38,742 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 144, 'iter': 0, 'avg': 30, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-144-avg-30-use-averaged-model'}
+2024-08-03 16:41:38,742 INFO [evaluate.py:221] About to create model
+2024-08-03 16:41:39,376 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 114 (excluded) to 144
+2024-08-03 16:41:48,955 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-03 16:41:48,955 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-03 16:41:48,965 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-03 16:41:49,594 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-03 16:41:53,785 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-03 16:41:56,577 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.7332, 2.9545, 2.7029, 2.8611, 0.9853, 2.5506, 2.4561, 2.2169],
+       device='cuda:0')
+2024-08-03 16:41:56,947 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6042, 3.2469, 3.0061, 3.2732], device='cuda:0')
+2024-08-03 16:41:57,125 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.0436, 4.7342, 4.5100, 4.9250], device='cuda:0')
+2024-08-03 16:41:57,835 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-03 16:42:01,859 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-03 16:42:05,906 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-03 16:42:09,951 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-03 16:42:13,960 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-03 16:42:18,030 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-03 16:42:22,395 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-03 16:42:26,425 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-03 16:42:30,469 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-03 16:42:30,582 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.3901, 3.7226, 3.3874, 3.5511], device='cuda:0')
+2024-08-03 16:42:33,847 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.3719, 2.9708, 3.0097, 3.1596, 2.8941, 2.6391, 2.8601, 2.4069],
+       device='cuda:0')
+2024-08-03 16:42:34,465 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.8730, 2.7951, 2.7497, 2.9622, 2.4548, 2.6930, 2.4462, 2.5450],
+       device='cuda:0')
+2024-08-03 16:42:34,512 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-03 16:42:38,534 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-03 16:42:43,317 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-03 16:42:47,350 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-03 16:42:51,374 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-03 16:42:55,378 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-03 16:42:59,416 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-03 16:43:03,421 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-03 16:43:07,419 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-03 16:43:09,418 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-03 16:43:13,571 INFO [evaluate.py:338] mAP for audioset eval is: 0.49558507451351885
+2024-08-03 16:43:13,572 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-146-avg-14-use-averaged-model-2024-08-05-12-06-49

@@ -0,0 +1,38 @@
+2024-08-05 12:06:49,899 INFO [evaluate.py:213] Evaluation started
+2024-08-05 12:06:49,899 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 146, 'iter': 0, 'avg': 14, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-146-avg-14-use-averaged-model'}
+2024-08-05 12:06:49,900 INFO [evaluate.py:221] About to create model
+2024-08-05 12:06:50,631 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 132 (excluded) to 146
+2024-08-05 12:06:56,053 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-05 12:06:56,054 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-05 12:06:56,067 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-05 12:06:56,691 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-05 12:07:00,757 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-05 12:07:04,733 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-05 12:07:08,709 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-05 12:07:12,680 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-05 12:07:16,644 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-05 12:07:20,168 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6730, 3.4318, 3.3203, 3.4919], device='cuda:0')
+2024-08-05 12:07:20,604 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-05 12:07:24,572 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-05 12:07:28,523 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-05 12:07:28,658 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.4098, 2.9478, 2.9845, 3.1545, 2.9032, 2.7761, 2.9213, 2.6937],
+       device='cuda:0')
+2024-08-05 12:07:29,890 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.2558, 4.1194, 4.4898, 4.0192], device='cuda:0')
+2024-08-05 12:07:32,483 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-05 12:07:35,233 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.2564, 4.1813, 4.5466, 3.9610], device='cuda:0')
+2024-08-05 12:07:36,437 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-05 12:07:40,391 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-05 12:07:44,336 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-05 12:07:48,292 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-05 12:07:52,244 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-05 12:07:55,558 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.8776, 3.0447, 2.6994, 2.9695, 2.4126, 2.8184, 2.5549, 2.3190],
+       device='cuda:0')
+2024-08-05 12:07:56,197 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-05 12:08:00,153 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-05 12:08:04,103 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-05 12:08:04,568 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([6.1311, 6.0524, 5.9063, 5.4400], device='cuda:0')
+2024-08-05 12:08:08,052 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-05 12:08:12,004 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-05 12:08:13,962 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-05 12:08:18,063 INFO [evaluate.py:338] mAP for audioset eval is: 0.4957936571232727
+2024-08-05 12:08:18,063 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-146-avg-16-use-averaged-model-2024-08-05-12-05-15

@@ -0,0 +1,36 @@
+2024-08-05 12:05:15,350 INFO [evaluate.py:213] Evaluation started
+2024-08-05 12:05:15,350 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 146, 'iter': 0, 'avg': 16, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-146-avg-16-use-averaged-model'}
+2024-08-05 12:05:15,350 INFO [evaluate.py:221] About to create model
+2024-08-05 12:05:16,063 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 130 (excluded) to 146
+2024-08-05 12:05:21,600 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-05 12:05:21,601 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-05 12:05:21,614 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-05 12:05:22,248 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-05 12:05:26,447 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-05 12:05:30,386 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-05 12:05:34,333 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-05 12:05:38,273 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-05 12:05:41,379 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6702, 3.3771, 3.2825, 3.2374], device='cuda:0')
+2024-08-05 12:05:42,202 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-05 12:05:46,117 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-05 12:05:47,630 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.7483, 2.5641, 2.3017, 2.4953, 2.6075, 2.3030, 2.4387, 2.2490],
+       device='cuda:0')
+2024-08-05 12:05:47,815 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.4048, 2.9232, 2.8885, 3.1155, 2.8605, 2.7155, 2.9140, 2.5237],
+       device='cuda:0')
+2024-08-05 12:05:50,031 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-05 12:05:53,948 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-05 12:05:57,870 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-05 12:06:01,788 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-05 12:06:05,728 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-05 12:06:09,672 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-05 12:06:13,588 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-05 12:06:17,503 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-05 12:06:21,419 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-05 12:06:24,339 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6870, 3.7034, 3.5166, 3.3539], device='cuda:0')
+2024-08-05 12:06:25,347 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-05 12:06:29,259 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-05 12:06:33,165 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-05 12:06:37,072 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-05 12:06:39,027 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-05 12:06:43,059 INFO [evaluate.py:338] mAP for audioset eval is: 0.49576535624904133
+2024-08-05 12:06:43,060 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-146-avg-18-use-averaged-model-2024-08-03-16-40-01

@@ -0,0 +1,33 @@
+2024-08-03 16:40:01,745 INFO [evaluate.py:213] Evaluation started
+2024-08-03 16:40:01,746 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 146, 'iter': 0, 'avg': 18, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-146-avg-18-use-averaged-model'}
+2024-08-03 16:40:01,746 INFO [evaluate.py:221] About to create model
+2024-08-03 16:40:02,363 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 128 (excluded) to 146
+2024-08-03 16:40:08,400 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-03 16:40:08,400 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-03 16:40:08,413 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-03 16:40:09,056 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-03 16:40:13,189 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-03 16:40:16,269 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.0853, 4.8091, 4.5713, 4.9644], device='cuda:0')
+2024-08-03 16:40:17,168 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-03 16:40:21,143 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-03 16:40:25,145 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-03 16:40:26,259 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.4052, 3.7019, 3.3278, 3.6948], device='cuda:0')
+2024-08-03 16:40:29,116 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-03 16:40:33,069 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-03 16:40:33,821 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6740, 3.2661, 3.4093, 3.3483], device='cuda:0')
+2024-08-03 16:40:37,021 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-03 16:40:40,975 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-03 16:40:44,949 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-03 16:40:48,953 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-03 16:40:52,935 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-03 16:40:56,928 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-03 16:41:00,933 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-03 16:41:04,905 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-03 16:41:08,874 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-03 16:41:12,858 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-03 16:41:16,845 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-03 16:41:20,840 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-03 16:41:24,852 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-03 16:41:26,827 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-03 16:41:30,965 INFO [evaluate.py:338] mAP for audioset eval is: 0.495697656159449
+2024-08-03 16:41:30,965 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-146-avg-20-use-averaged-model-2024-08-03-16-38-26

@@ -0,0 +1,34 @@
+2024-08-03 16:38:26,571 INFO [evaluate.py:213] Evaluation started
+2024-08-03 16:38:26,571 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 146, 'iter': 0, 'avg': 20, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-146-avg-20-use-averaged-model'}
+2024-08-03 16:38:26,571 INFO [evaluate.py:221] About to create model
+2024-08-03 16:38:27,184 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 126 (excluded) to 146
+2024-08-03 16:38:32,330 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-03 16:38:32,330 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-03 16:38:32,340 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-03 16:38:32,995 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-03 16:38:37,026 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-03 16:38:41,019 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-03 16:38:45,007 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-03 16:38:48,993 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-03 16:38:52,981 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-03 16:38:55,641 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([6.1288, 6.0387, 5.8533, 5.5528], device='cuda:0')
+2024-08-03 16:38:56,959 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-03 16:39:00,923 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-03 16:39:04,900 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-03 16:39:08,882 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-03 16:39:12,853 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-03 16:39:16,822 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-03 16:39:20,791 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-03 16:39:24,760 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-03 16:39:28,726 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-03 16:39:32,695 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-03 16:39:36,662 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-03 16:39:39,748 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.1769, 4.3753, 4.4685, 5.0320], device='cuda:0')
+2024-08-03 16:39:40,635 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-03 16:39:44,597 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-03 16:39:48,554 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-03 16:39:49,695 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.8868, 2.8792, 2.6552, 2.9041, 2.3867, 2.7613, 2.5953, 2.6163],
+       device='cuda:0')
+2024-08-03 16:39:50,521 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-03 16:39:54,635 INFO [evaluate.py:338] mAP for audioset eval is: 0.49570587310974645
+2024-08-03 16:39:54,636 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-146-avg-22-use-averaged-model-2024-08-03-16-36-51

@@ -0,0 +1,37 @@
+2024-08-03 16:36:51,079 INFO [evaluate.py:213] Evaluation started
+2024-08-03 16:36:51,080 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 146, 'iter': 0, 'avg': 22, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-146-avg-22-use-averaged-model'}
+2024-08-03 16:36:51,080 INFO [evaluate.py:221] About to create model
+2024-08-03 16:36:51,752 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 124 (excluded) to 146
+2024-08-03 16:36:57,152 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-03 16:36:57,152 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-03 16:36:57,162 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-03 16:36:57,792 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-03 16:37:01,945 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-03 16:37:05,923 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-03 16:37:06,521 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.0145, 4.0702, 3.6802, 4.2502], device='cuda:0')
+2024-08-03 16:37:09,641 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.4746, 3.3217, 2.9040, 3.0974], device='cuda:0')
+2024-08-03 16:37:09,916 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-03 16:37:13,909 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-03 16:37:17,894 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-03 16:37:21,873 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-03 16:37:25,847 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-03 16:37:29,817 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-03 16:37:31,524 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.3610, 3.7107, 3.4212, 3.5949], device='cuda:0')
+2024-08-03 16:37:33,787 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-03 16:37:37,753 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-03 16:37:41,723 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-03 16:37:45,689 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-03 16:37:49,650 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-03 16:37:49,935 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.9495, 5.6964, 5.7428, 5.5612], device='cuda:0')
+2024-08-03 16:37:53,109 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.9162, 5.6524, 5.6732, 5.6394], device='cuda:0')
+2024-08-03 16:37:53,557 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.3797, 2.9899, 2.9723, 3.0788, 2.8471, 2.7243, 2.9408, 2.7317],
+       device='cuda:0')
+2024-08-03 16:37:53,621 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-03 16:37:57,595 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-03 16:38:01,552 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-03 16:38:05,513 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-03 16:38:09,477 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-03 16:38:13,443 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-03 16:38:15,404 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-03 16:38:19,554 INFO [evaluate.py:338] mAP for audioset eval is: 0.4955991283104626
+2024-08-03 16:38:19,554 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-146-avg-24-use-averaged-model-2024-08-03-16-35-14

@@ -0,0 +1,37 @@
+2024-08-03 16:35:14,443 INFO [evaluate.py:213] Evaluation started
+2024-08-03 16:35:14,443 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 146, 'iter': 0, 'avg': 24, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-146-avg-24-use-averaged-model'}
+2024-08-03 16:35:14,444 INFO [evaluate.py:221] About to create model
+2024-08-03 16:35:15,068 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 122 (excluded) to 146
+2024-08-03 16:35:20,433 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-03 16:35:20,433 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-03 16:35:20,445 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-03 16:35:21,086 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-03 16:35:25,236 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-03 16:35:26,573 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.7778, 2.4734, 2.6212, 2.4396, 2.5456, 2.3200, 2.8386, 2.2829],
+       device='cuda:0')
+2024-08-03 16:35:29,216 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-03 16:35:33,190 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-03 16:35:36,636 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([6.1314, 6.0372, 5.8521, 5.4813], device='cuda:0')
+2024-08-03 16:35:37,163 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-03 16:35:41,124 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-03 16:35:45,085 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-03 16:35:48,771 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6068, 3.1226, 2.9763, 3.1037], device='cuda:0')
+2024-08-03 16:35:49,041 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-03 16:35:53,000 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-03 16:35:56,959 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-03 16:35:57,442 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.9213, 5.6771, 5.7256, 5.6075], device='cuda:0')
+2024-08-03 16:36:00,933 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-03 16:36:02,050 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6177, 3.3255, 3.1379, 3.3007], device='cuda:0')
+2024-08-03 16:36:04,898 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-03 16:36:08,863 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-03 16:36:12,816 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-03 16:36:16,783 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-03 16:36:20,755 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-03 16:36:24,712 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-03 16:36:26,267 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6767, 3.0439, 3.3201, 3.6364], device='cuda:0')
+2024-08-03 16:36:28,670 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-03 16:36:32,619 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-03 16:36:36,573 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-03 16:36:38,533 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-03 16:36:42,666 INFO [evaluate.py:338] mAP for audioset eval is: 0.49564216968114855
+2024-08-03 16:36:42,667 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-146-avg-26-use-averaged-model-2024-08-03-16-33-38

@@ -0,0 +1,35 @@
+2024-08-03 16:33:38,272 INFO [evaluate.py:213] Evaluation started
+2024-08-03 16:33:38,272 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 146, 'iter': 0, 'avg': 26, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-146-avg-26-use-averaged-model'}
+2024-08-03 16:33:38,273 INFO [evaluate.py:221] About to create model
+2024-08-03 16:33:38,917 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 120 (excluded) to 146
+2024-08-03 16:33:44,297 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-03 16:33:44,297 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-03 16:33:44,310 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-03 16:33:44,971 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-03 16:33:49,085 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-03 16:33:53,064 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-03 16:33:57,086 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-03 16:34:01,080 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-03 16:34:05,050 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-03 16:34:09,028 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-03 16:34:12,992 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-03 16:34:16,960 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-03 16:34:20,918 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-03 16:34:24,881 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-03 16:34:28,844 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-03 16:34:32,814 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-03 16:34:36,779 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-03 16:34:37,127 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.7479, 2.5372, 2.2561, 2.5155, 2.6501, 2.4239, 2.5602, 2.3003],
+       device='cuda:0')
+2024-08-03 16:34:40,752 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-03 16:34:41,276 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6082, 3.0444, 2.9369, 3.1245], device='cuda:0')
+2024-08-03 16:34:44,717 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-03 16:34:46,072 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.8756, 3.0695, 2.7880, 2.9565, 2.5400, 2.6724, 2.4859, 2.5107],
+       device='cuda:0')
+2024-08-03 16:34:48,722 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-03 16:34:52,686 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-03 16:34:56,664 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-03 16:35:00,649 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-03 16:35:02,618 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-03 16:35:06,857 INFO [evaluate.py:338] mAP for audioset eval is: 0.49566394846493195
+2024-08-03 16:35:06,858 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-146-avg-28-use-averaged-model-2024-08-03-16-32-01

@@ -0,0 +1,33 @@
+2024-08-03 16:32:01,516 INFO [evaluate.py:213] Evaluation started
+2024-08-03 16:32:01,516 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 146, 'iter': 0, 'avg': 28, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-146-avg-28-use-averaged-model'}
+2024-08-03 16:32:01,516 INFO [evaluate.py:221] About to create model
+2024-08-03 16:32:02,181 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 118 (excluded) to 146
+2024-08-03 16:32:07,966 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-03 16:32:07,967 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-03 16:32:07,980 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-03 16:32:08,608 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-03 16:32:12,714 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-03 16:32:16,716 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-03 16:32:20,723 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-03 16:32:24,732 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-03 16:32:27,209 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([6.1201, 6.0306, 5.8435, 5.3797], device='cuda:0')
+2024-08-03 16:32:27,408 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([6.1247, 6.0373, 5.8450, 5.4424], device='cuda:0')
+2024-08-03 16:32:28,733 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-03 16:32:32,711 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-03 16:32:36,688 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-03 16:32:40,682 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-03 16:32:44,673 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-03 16:32:48,654 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-03 16:32:52,663 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-03 16:32:56,649 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-03 16:33:00,638 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-03 16:33:04,613 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-03 16:33:08,591 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-03 16:33:12,582 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-03 16:33:16,563 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-03 16:33:20,542 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-03 16:33:20,912 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6637, 2.8887, 3.2705, 3.6221], device='cuda:0')
+2024-08-03 16:33:24,533 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-03 16:33:26,543 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-03 16:33:30,795 INFO [evaluate.py:338] mAP for audioset eval is: 0.49564397532353993
+2024-08-03 16:33:30,795 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-146-avg-30-use-averaged-model-2024-08-03-16-30-21

@@ -0,0 +1,35 @@
+2024-08-03 16:30:21,369 INFO [evaluate.py:213] Evaluation started
+2024-08-03 16:30:21,369 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 146, 'iter': 0, 'avg': 30, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-146-avg-30-use-averaged-model'}
+2024-08-03 16:30:21,369 INFO [evaluate.py:221] About to create model
+2024-08-03 16:30:21,984 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 116 (excluded) to 146
+2024-08-03 16:30:31,171 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-03 16:30:31,171 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-03 16:30:31,183 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-03 16:30:31,797 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-03 16:30:35,995 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-03 16:30:38,138 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6010, 3.3023, 3.0208, 3.2334], device='cuda:0')
+2024-08-03 16:30:39,759 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.5853, 3.1598, 2.9753, 3.1636], device='cuda:0')
+2024-08-03 16:30:40,031 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-03 16:30:44,046 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-03 16:30:48,077 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-03 16:30:52,107 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-03 16:30:56,128 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-03 16:30:57,663 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.3853, 2.9762, 2.9268, 2.9764, 2.8587, 2.6006, 2.7937, 2.5244],
+       device='cuda:0')
+2024-08-03 16:31:00,161 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-03 16:31:04,145 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-03 16:31:08,114 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-03 16:31:12,132 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-03 16:31:16,138 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-03 16:31:20,133 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-03 16:31:24,101 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-03 16:31:28,041 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-03 16:31:31,974 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-03 16:31:35,929 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-03 16:31:39,896 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-03 16:31:43,859 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-03 16:31:47,815 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-03 16:31:48,935 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.3667, 3.6852, 3.3186, 3.5626], device='cuda:0')
+2024-08-03 16:31:49,815 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-03 16:31:54,048 INFO [evaluate.py:338] mAP for audioset eval is: 0.4957230020650777
+2024-08-03 16:31:54,048 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-148-avg-14-use-averaged-model-2024-08-05-12-03-39

@@ -0,0 +1,33 @@
+2024-08-05 12:03:39,410 INFO [evaluate.py:213] Evaluation started
+2024-08-05 12:03:39,411 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 148, 'iter': 0, 'avg': 14, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-148-avg-14-use-averaged-model'}
+2024-08-05 12:03:39,411 INFO [evaluate.py:221] About to create model
+2024-08-05 12:03:40,088 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 134 (excluded) to 148
+2024-08-05 12:03:45,632 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-05 12:03:45,633 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-05 12:03:45,646 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-05 12:03:46,319 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-05 12:03:50,472 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-05 12:03:52,057 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.2030, 4.2559, 3.9431, 4.4063], device='cuda:0')
+2024-08-05 12:03:54,464 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-05 12:03:58,454 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-05 12:04:02,449 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-05 12:04:06,425 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-05 12:04:10,399 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-05 12:04:14,367 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-05 12:04:18,326 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-05 12:04:22,288 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-05 12:04:26,251 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-05 12:04:30,217 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-05 12:04:34,176 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-05 12:04:35,557 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.1975, 4.2098, 3.9773, 4.4633], device='cuda:0')
+2024-08-05 12:04:37,418 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([6.1190, 6.0437, 5.8684, 5.4190], device='cuda:0')
+2024-08-05 12:04:38,144 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-05 12:04:42,123 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-05 12:04:46,085 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-05 12:04:50,040 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-05 12:04:54,000 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-05 12:04:57,960 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-05 12:05:01,924 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-05 12:05:03,893 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-05 12:05:07,974 INFO [evaluate.py:338] mAP for audioset eval is: 0.49598251997679177
+2024-08-05 12:05:07,974 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-148-avg-16-use-averaged-model-2024-08-05-12-02-03

@@ -0,0 +1,34 @@
+2024-08-05 12:02:03,217 INFO [evaluate.py:213] Evaluation started
+2024-08-05 12:02:03,218 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 148, 'iter': 0, 'avg': 16, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-148-avg-16-use-averaged-model'}
+2024-08-05 12:02:03,218 INFO [evaluate.py:221] About to create model
+2024-08-05 12:02:03,847 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 132 (excluded) to 148
+2024-08-05 12:02:10,445 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-05 12:02:10,446 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-05 12:02:10,459 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-05 12:02:11,096 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-05 12:02:15,257 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-05 12:02:19,244 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-05 12:02:23,226 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-05 12:02:27,214 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-05 12:02:31,178 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-05 12:02:34,853 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.3905, 3.6893, 3.3967, 3.6242], device='cuda:0')
+2024-08-05 12:02:35,136 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-05 12:02:39,092 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-05 12:02:43,056 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-05 12:02:45,701 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([6.1295, 6.0520, 5.9010, 5.4555], device='cuda:0')
+2024-08-05 12:02:47,017 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-05 12:02:50,976 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-05 12:02:54,930 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-05 12:02:58,886 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-05 12:03:02,855 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-05 12:03:06,819 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-05 12:03:10,779 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-05 12:03:11,129 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.8876, 2.8992, 2.6370, 3.0656, 2.2815, 2.8081, 2.7139, 2.5214],
+       device='cuda:0')
+2024-08-05 12:03:14,748 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-05 12:03:18,716 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-05 12:03:22,686 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-05 12:03:26,650 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-05 12:03:28,618 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-05 12:03:32,690 INFO [evaluate.py:338] mAP for audioset eval is: 0.49597113660587616
+2024-08-05 12:03:32,691 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-148-avg-18-use-averaged-model-2024-08-03-16-28-45

@@ -0,0 +1,32 @@
+2024-08-03 16:28:45,177 INFO [evaluate.py:213] Evaluation started
+2024-08-03 16:28:45,177 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 148, 'iter': 0, 'avg': 18, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-148-avg-18-use-averaged-model'}
+2024-08-03 16:28:45,177 INFO [evaluate.py:221] About to create model
+2024-08-03 16:28:45,889 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 130 (excluded) to 148
+2024-08-03 16:28:51,344 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-03 16:28:51,344 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-03 16:28:51,355 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-03 16:28:51,989 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-03 16:28:56,229 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-03 16:29:00,205 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-03 16:29:04,180 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-03 16:29:08,071 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.9334, 4.6917, 4.3955, 4.7910], device='cuda:0')
+2024-08-03 16:29:08,173 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-03 16:29:12,147 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-03 16:29:16,123 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-03 16:29:20,085 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-03 16:29:24,043 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-03 16:29:28,001 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-03 16:29:28,123 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.5890, 3.4102, 3.0978, 3.3193], device='cuda:0')
+2024-08-03 16:29:31,959 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-03 16:29:35,916 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-03 16:29:39,875 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-03 16:29:43,838 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-03 16:29:47,802 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-03 16:29:51,764 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-03 16:29:55,728 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-03 16:29:59,696 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-03 16:30:03,651 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-03 16:30:07,617 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-03 16:30:09,577 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-03 16:30:13,700 INFO [evaluate.py:338] mAP for audioset eval is: 0.4959698809945916
+2024-08-03 16:30:13,700 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-148-avg-20-use-averaged-model-2024-08-03-16-27-09

@@ -0,0 +1,37 @@
+2024-08-03 16:27:09,671 INFO [evaluate.py:213] Evaluation started
+2024-08-03 16:27:09,671 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 148, 'iter': 0, 'avg': 20, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-148-avg-20-use-averaged-model'}
+2024-08-03 16:27:09,671 INFO [evaluate.py:221] About to create model
+2024-08-03 16:27:10,290 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 128 (excluded) to 148
+2024-08-03 16:27:15,490 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-03 16:27:15,491 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-03 16:27:15,502 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-03 16:27:16,124 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-03 16:27:20,106 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-03 16:27:24,094 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-03 16:27:28,078 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-03 16:27:32,070 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-03 16:27:36,042 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-03 16:27:40,016 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-03 16:27:40,914 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.2043, 4.4866, 4.6326, 5.0239], device='cuda:0')
+2024-08-03 16:27:43,545 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6738, 3.4112, 3.2744, 3.2264], device='cuda:0')
+2024-08-03 16:27:43,984 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-03 16:27:47,944 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-03 16:27:51,907 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-03 16:27:55,877 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-03 16:27:56,463 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.2182, 4.3327, 4.1401, 4.5849], device='cuda:0')
+2024-08-03 16:27:59,837 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-03 16:28:03,807 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-03 16:28:07,766 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-03 16:28:11,633 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.0987, 4.8243, 4.6738, 4.9833], device='cuda:0')
+2024-08-03 16:28:11,735 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-03 16:28:15,699 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-03 16:28:15,839 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.7880, 2.3470, 2.5987, 2.5738, 2.5354, 2.3242, 2.9082, 2.3937],
+       device='cuda:0')
+2024-08-03 16:28:19,657 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-03 16:28:23,620 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-03 16:28:27,583 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-03 16:28:29,347 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6902, 3.4492, 3.4422, 3.3273], device='cuda:0')
+2024-08-03 16:28:31,564 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-03 16:28:33,518 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-03 16:28:37,694 INFO [evaluate.py:338] mAP for audioset eval is: 0.49580849232139246
+2024-08-03 16:28:37,695 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-148-avg-22-use-averaged-model-2024-08-03-16-25-33

@@ -0,0 +1,37 @@
+2024-08-03 16:25:33,049 INFO [evaluate.py:213] Evaluation started
+2024-08-03 16:25:33,049 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 148, 'iter': 0, 'avg': 22, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-148-avg-22-use-averaged-model'}
+2024-08-03 16:25:33,050 INFO [evaluate.py:221] About to create model
+2024-08-03 16:25:33,703 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 126 (excluded) to 148
+2024-08-03 16:25:38,857 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-03 16:25:38,857 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-03 16:25:38,868 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-03 16:25:39,503 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-03 16:25:43,964 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-03 16:25:48,002 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-03 16:25:52,034 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-03 16:25:56,079 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-03 16:26:00,083 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-03 16:26:04,099 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-03 16:26:08,100 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-03 16:26:12,121 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-03 16:26:16,149 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-03 16:26:20,152 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-03 16:26:22,522 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.2470, 3.9393, 4.4245, 3.8483], device='cuda:0')
+2024-08-03 16:26:24,129 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-03 16:26:25,656 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.6991, 2.4687, 2.2475, 2.3648, 2.4793, 2.2836, 2.4160, 2.1463],
+       device='cuda:0')
+2024-08-03 16:26:28,106 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-03 16:26:32,143 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-03 16:26:36,147 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-03 16:26:40,162 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-03 16:26:44,106 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-03 16:26:48,043 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-03 16:26:51,159 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.4218, 3.0152, 2.9267, 3.1377, 2.9449, 2.5604, 2.6919, 2.4029],
+       device='cuda:0')
+2024-08-03 16:26:52,012 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-03 16:26:55,948 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-03 16:26:57,014 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.9300, 5.6829, 5.7204, 5.6579], device='cuda:0')
+2024-08-03 16:26:57,211 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.8970, 5.6050, 5.6807, 5.5824], device='cuda:0')
+2024-08-03 16:26:57,937 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-03 16:27:02,149 INFO [evaluate.py:338] mAP for audioset eval is: 0.4958361389578191
+2024-08-03 16:27:02,149 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-148-avg-24-use-averaged-model-2024-08-03-16-23-58

@@ -0,0 +1,42 @@
+2024-08-03 16:23:58,135 INFO [evaluate.py:213] Evaluation started
+2024-08-03 16:23:58,135 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 148, 'iter': 0, 'avg': 24, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-148-avg-24-use-averaged-model'}
+2024-08-03 16:23:58,135 INFO [evaluate.py:221] About to create model
+2024-08-03 16:23:58,777 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 124 (excluded) to 148
+2024-08-03 16:24:03,981 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-03 16:24:03,982 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-03 16:24:03,994 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-03 16:24:04,625 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-03 16:24:08,584 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-03 16:24:09,537 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.8737, 2.9984, 2.8467, 2.8448, 0.6639, 2.8149, 2.4652, 2.6173],
+       device='cuda:0')
+2024-08-03 16:24:12,575 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-03 16:24:16,551 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-03 16:24:20,535 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-03 16:24:24,507 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-03 16:24:28,475 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-03 16:24:32,445 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-03 16:24:36,434 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-03 16:24:40,428 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-03 16:24:44,387 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-03 16:24:48,344 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-03 16:24:52,318 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-03 16:24:56,286 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-03 16:25:00,252 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-03 16:25:00,350 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.0653, 4.7708, 4.5507, 4.9515], device='cuda:0')
+2024-08-03 16:25:04,219 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-03 16:25:08,186 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-03 16:25:08,322 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.4000, 2.9698, 2.9366, 3.0154, 2.8793, 2.6331, 2.8704, 2.7449],
+       device='cuda:0')
+2024-08-03 16:25:09,516 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.7852, 2.4653, 2.5897, 2.5518, 2.5862, 2.2266, 2.9041, 2.2143],
+       device='cuda:0')
+2024-08-03 16:25:10,315 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.7729, 2.6633, 2.3180, 2.3580, 2.5917, 2.1943, 2.6364, 2.4730],
+       device='cuda:0')
+2024-08-03 16:25:11,699 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.7940, 2.5213, 2.6996, 2.6148, 2.5095, 2.3230, 2.8995, 2.2759],
+       device='cuda:0')
+2024-08-03 16:25:12,154 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-03 16:25:14,638 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.1865, 4.4543, 4.5390, 5.0138], device='cuda:0')
+2024-08-03 16:25:16,120 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-03 16:25:20,089 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-03 16:25:22,046 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-03 16:25:26,209 INFO [evaluate.py:338] mAP for audioset eval is: 0.4958814308445697
+2024-08-03 16:25:26,209 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-148-avg-26-use-averaged-model-2024-08-03-16-22-21

@@ -0,0 +1,38 @@
+2024-08-03 16:22:21,201 INFO [evaluate.py:213] Evaluation started
+2024-08-03 16:22:21,202 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 148, 'iter': 0, 'avg': 26, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-148-avg-26-use-averaged-model'}
+2024-08-03 16:22:21,202 INFO [evaluate.py:221] About to create model
+2024-08-03 16:22:21,861 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 122 (excluded) to 148
+2024-08-03 16:22:27,303 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-03 16:22:27,304 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-03 16:22:27,317 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-03 16:22:27,952 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-03 16:22:32,044 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-03 16:22:35,595 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6822, 3.2898, 3.3913, 3.6270], device='cuda:0')
+2024-08-03 16:22:36,030 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-03 16:22:40,068 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-03 16:22:44,095 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-03 16:22:48,095 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-03 16:22:52,093 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-03 16:22:56,095 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-03 16:22:57,013 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6067, 3.3223, 3.0753, 3.2242], device='cuda:0')
+2024-08-03 16:23:00,082 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-03 16:23:04,066 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-03 16:23:07,566 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6114, 3.3244, 3.0233, 3.1628], device='cuda:0')
+2024-08-03 16:23:08,045 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-03 16:23:12,030 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-03 16:23:14,778 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.8783, 3.0125, 2.8332, 2.9450, 2.4349, 2.7784, 2.5442, 2.5501],
+       device='cuda:0')
+2024-08-03 16:23:16,020 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-03 16:23:20,013 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-03 16:23:24,028 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-03 16:23:24,352 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6056, 3.3071, 3.1349, 3.2521], device='cuda:0')
+2024-08-03 16:23:28,051 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-03 16:23:32,038 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-03 16:23:35,014 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.2410, 3.9137, 4.5274, 3.9442], device='cuda:0')
+2024-08-03 16:23:36,029 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-03 16:23:40,047 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-03 16:23:41,208 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6614, 2.7120, 3.3366, 3.5155], device='cuda:0')
+2024-08-03 16:23:44,033 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-03 16:23:46,107 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-03 16:23:50,350 INFO [evaluate.py:338] mAP for audioset eval is: 0.49575070336821925
+2024-08-03 16:23:50,351 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-148-avg-28-use-averaged-model-2024-08-03-16-20-45

@@ -0,0 +1,37 @@
+2024-08-03 16:20:45,577 INFO [evaluate.py:213] Evaluation started
+2024-08-03 16:20:45,577 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 148, 'iter': 0, 'avg': 28, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-148-avg-28-use-averaged-model'}
+2024-08-03 16:20:45,578 INFO [evaluate.py:221] About to create model
+2024-08-03 16:20:46,242 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 120 (excluded) to 148
+2024-08-03 16:20:51,407 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-03 16:20:51,407 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-03 16:20:51,417 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-03 16:20:52,061 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-03 16:20:56,272 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-03 16:21:00,288 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-03 16:21:04,284 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-03 16:21:08,294 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-03 16:21:10,635 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.4168, 3.0010, 2.9450, 3.0388, 2.7558, 2.6446, 2.8413, 2.4807],
+       device='cuda:0')
+2024-08-03 16:21:12,289 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-03 16:21:16,304 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-03 16:21:20,283 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-03 16:21:24,256 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-03 16:21:25,373 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6043, 3.2992, 3.0959, 3.1847], device='cuda:0')
+2024-08-03 16:21:27,746 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.3943, 3.6857, 3.4338, 3.6483], device='cuda:0')
+2024-08-03 16:21:28,231 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-03 16:21:32,201 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-03 16:21:34,572 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.2038, 4.2097, 4.0245, 4.5170], device='cuda:0')
+2024-08-03 16:21:35,934 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6643, 2.7231, 3.1782, 3.4763], device='cuda:0')
+2024-08-03 16:21:36,165 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-03 16:21:40,120 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-03 16:21:44,087 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-03 16:21:48,068 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-03 16:21:52,043 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-03 16:21:56,003 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-03 16:21:59,078 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.1724, 4.3719, 4.4702, 5.0236], device='cuda:0')
+2024-08-03 16:21:59,963 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-03 16:22:03,920 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-03 16:22:07,876 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-03 16:22:09,838 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-03 16:22:13,925 INFO [evaluate.py:338] mAP for audioset eval is: 0.4957584753728736
+2024-08-03 16:22:13,925 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-148-avg-30-use-averaged-model-2024-08-03-16-19-05

@@ -0,0 +1,38 @@
+2024-08-03 16:19:05,692 INFO [evaluate.py:213] Evaluation started
+2024-08-03 16:19:05,692 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 148, 'iter': 0, 'avg': 30, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-148-avg-30-use-averaged-model'}
+2024-08-03 16:19:05,692 INFO [evaluate.py:221] About to create model
+2024-08-03 16:19:06,359 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 118 (excluded) to 148
+2024-08-03 16:19:15,620 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-03 16:19:15,620 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-03 16:19:15,635 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-03 16:19:16,269 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-03 16:19:20,361 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-03 16:19:22,719 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.6943, 2.7592, 2.6203, 2.7325, 0.9809, 2.5457, 2.2833, 2.1434],
+       device='cuda:0')
+2024-08-03 16:19:24,381 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-03 16:19:28,429 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-03 16:19:30,221 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.1160, 3.8866, 2.7286, 3.7177], device='cuda:0')
+2024-08-03 16:19:32,451 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-03 16:19:36,456 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-03 16:19:40,423 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-03 16:19:44,355 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-03 16:19:48,288 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-03 16:19:52,237 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-03 16:19:52,543 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.1924, 4.3606, 4.5756, 5.0229], device='cuda:0')
+2024-08-03 16:19:56,181 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-03 16:20:00,125 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-03 16:20:04,075 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-03 16:20:08,018 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-03 16:20:11,108 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6007, 3.2675, 2.9375, 3.1712], device='cuda:0')
+2024-08-03 16:20:11,967 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-03 16:20:15,908 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-03 16:20:19,845 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-03 16:20:20,578 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.7838, 2.4146, 2.6654, 2.5698, 2.5224, 2.2844, 2.9548, 2.3398],
+       device='cuda:0')
+2024-08-03 16:20:20,928 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.0733, 4.7845, 4.5291, 4.9265], device='cuda:0')
+2024-08-03 16:20:23,788 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-03 16:20:27,754 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-03 16:20:31,700 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-03 16:20:33,676 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-03 16:20:37,868 INFO [evaluate.py:338] mAP for audioset eval is: 0.49575695927344926
+2024-08-03 16:20:37,868 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-150-avg-14-use-averaged-model-2024-08-05-12-00-25

@@ -0,0 +1,34 @@
+2024-08-05 12:00:25,325 INFO [evaluate.py:213] Evaluation started
+2024-08-05 12:00:25,325 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 150, 'iter': 0, 'avg': 14, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-150-avg-14-use-averaged-model'}
+2024-08-05 12:00:25,326 INFO [evaluate.py:221] About to create model
+2024-08-05 12:00:25,964 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 136 (excluded) to 150
+2024-08-05 12:00:33,912 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-05 12:00:33,912 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-05 12:00:33,925 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-05 12:00:34,561 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-05 12:00:38,636 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-05 12:00:42,645 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-05 12:00:46,636 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-05 12:00:50,627 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-05 12:00:54,611 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-05 12:00:58,583 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-05 12:01:02,549 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-05 12:01:06,509 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-05 12:01:10,487 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-05 12:01:14,410 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.7760, 2.5058, 2.3485, 2.4461, 2.6260, 2.4434, 2.5269, 2.3361],
+       device='cuda:0')
+2024-08-05 12:01:14,463 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-05 12:01:18,439 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-05 12:01:22,403 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-05 12:01:22,518 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.3948, 3.7832, 3.4100, 3.8110], device='cuda:0')
+2024-08-05 12:01:26,357 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-05 12:01:30,322 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-05 12:01:33,272 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.2660, 4.2658, 4.6835, 4.2092], device='cuda:0')
+2024-08-05 12:01:34,283 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-05 12:01:38,243 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-05 12:01:42,201 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-05 12:01:46,157 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-05 12:01:50,110 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-05 12:01:52,068 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-05 12:01:56,156 INFO [evaluate.py:338] mAP for audioset eval is: 0.4960691402341602
+2024-08-05 12:01:56,156 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-150-avg-16-use-averaged-model-2024-08-05-11-58-46

@@ -0,0 +1,43 @@
+2024-08-05 11:58:46,068 INFO [evaluate.py:213] Evaluation started
+2024-08-05 11:58:46,068 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 150, 'iter': 0, 'avg': 16, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-150-avg-16-use-averaged-model'}
+2024-08-05 11:58:46,069 INFO [evaluate.py:221] About to create model
+2024-08-05 11:58:46,705 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 134 (excluded) to 150
+2024-08-05 11:58:54,616 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-05 11:58:54,617 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-05 11:58:54,632 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-05 11:58:55,263 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-05 11:58:59,630 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-05 11:59:03,667 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-05 11:59:07,666 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-05 11:59:11,664 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-05 11:59:15,654 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-05 11:59:19,232 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.2126, 4.2173, 4.0403, 4.5491], device='cuda:0')
+2024-08-05 11:59:19,641 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-05 11:59:21,016 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.2668, 4.1739, 4.5452, 4.0878], device='cuda:0')
+2024-08-05 11:59:22,136 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.0868, 4.8241, 4.5831, 4.9380], device='cuda:0')
+2024-08-05 11:59:23,635 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-05 11:59:27,549 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.4115, 3.6065, 3.4302, 3.6516], device='cuda:0')
+2024-08-05 11:59:27,633 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-05 11:59:28,546 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.3940, 3.7362, 3.3356, 3.6220], device='cuda:0')
+2024-08-05 11:59:31,620 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-05 11:59:33,358 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.7871, 2.5198, 2.6255, 2.5511, 2.4887, 2.3397, 2.8686, 2.1934],
+       device='cuda:0')
+2024-08-05 11:59:35,609 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-05 11:59:37,933 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.5880, 3.4257, 3.1117, 3.0753], device='cuda:0')
+2024-08-05 11:59:38,285 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([6.1215, 6.0484, 5.8750, 5.3550], device='cuda:0')
+2024-08-05 11:59:39,618 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-05 11:59:43,621 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-05 11:59:47,587 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-05 11:59:50,111 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.7568, 2.4169, 2.3906, 2.2529, 2.5666, 2.1183, 2.3725, 2.3356],
+       device='cuda:0')
+2024-08-05 11:59:50,746 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.2468, 4.1537, 4.5496, 4.0453], device='cuda:0')
+2024-08-05 11:59:51,566 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-05 11:59:51,694 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6180, 3.2058, 3.0470, 2.9045], device='cuda:0')
+2024-08-05 11:59:55,547 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-05 11:59:59,527 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-05 12:00:03,505 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-05 12:00:07,459 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-05 12:00:11,436 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-05 12:00:13,458 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-05 12:00:17,824 INFO [evaluate.py:338] mAP for audioset eval is: 0.49613529016555646
+2024-08-05 12:00:17,825 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-150-avg-18-use-averaged-model-2024-08-03-16-17-26

@@ -0,0 +1,35 @@
+2024-08-03 16:17:26,792 INFO [evaluate.py:213] Evaluation started
+2024-08-03 16:17:26,792 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 150, 'iter': 0, 'avg': 18, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-150-avg-18-use-averaged-model'}
+2024-08-03 16:17:26,792 INFO [evaluate.py:221] About to create model
+2024-08-03 16:17:27,462 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 132 (excluded) to 150
+2024-08-03 16:17:34,412 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-03 16:17:34,412 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-03 16:17:34,423 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-03 16:17:35,055 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-03 16:17:39,346 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-03 16:17:43,356 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-03 16:17:47,358 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-03 16:17:51,370 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-03 16:17:52,949 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.4898, 2.5132, 2.9398, 3.4344], device='cuda:0')
+2024-08-03 16:17:55,395 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-03 16:17:56,142 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.7705, 2.6611, 2.4382, 2.4619, 2.5842, 2.4430, 2.5308, 2.3547],
+       device='cuda:0')
+2024-08-03 16:17:59,388 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-03 16:18:03,394 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-03 16:18:04,524 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.4300, 2.9651, 3.0000, 3.1868, 3.0239, 2.7786, 2.8964, 2.6545],
+       device='cuda:0')
+2024-08-03 16:18:07,391 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-03 16:18:11,401 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-03 16:18:15,446 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-03 16:18:19,463 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-03 16:18:23,466 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-03 16:18:27,481 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-03 16:18:31,486 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-03 16:18:35,501 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-03 16:18:39,525 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-03 16:18:43,511 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-03 16:18:47,494 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-03 16:18:51,488 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-03 16:18:53,525 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-03 16:18:57,826 INFO [evaluate.py:338] mAP for audioset eval is: 0.4961728170009992
+2024-08-03 16:18:57,826 INFO [evaluate.py:340] Done

no_specaug_no_musan/inference_audio_tagging/log-decode-epoch-150-avg-20-use-averaged-model-2024-08-03-16-15-49

@@ -0,0 +1,43 @@
+2024-08-03 16:15:49,334 INFO [evaluate.py:213] Evaluation started
+2024-08-03 16:15:49,334 INFO [evaluate.py:215] {'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, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.24.4', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'ff1d435a8d3c4eaa15828a84a7240678a70539a7', 'k2-git-date': 'Fri Feb 23 01:48:38 2024', 'lhotse-version': '1.24.0.dev+git.5cae6234.dirty', 'torch-version': '2.0.1+cu117', 'torch-cuda-available': True, 'torch-cuda-version': '11.7', 'python-version': '3.1', 'icefall-git-branch': 'audio_pretraining_frame_level', 'icefall-git-sha1': 'e6ae882e-dirty', 'icefall-git-date': 'Tue Jul 9 10:48:15 2024', 'icefall-path': '/star-xy/softwares/icefall_development/icefall_multi_KD_streaming', 'k2-path': '/star-xy/softwares/pyenvs/k2_cuda11/k2_cuda11/lib/python3.10/site-packages/k2/__init__.py', 'lhotse-path': '/star-xy/softwares/lhotse_development/lhotse_weighted_sampler/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1207150844-f49d8c4f4-c49d5'}, 'epoch': 150, 'iter': 0, 'avg': 20, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'num_events': 527, 'feature_dim': 128, 'audioset_subset': 'balanced', 'manifest_dir': PosixPath('data/fbank_as_ced_mAP50'), 'max_duration': 500, 'weighted_sampler': False, 'num_samples': 200000, '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, 'features_mask_size': 100, 'frames_mask_size': 27, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('zipformer/exp_KD_CED_base_full_lr_epochs_15_specaug0_frame192_feature27_musan0_weighted1_md1000/inference_audio_tagging'), 'suffix': 'epoch-150-avg-20-use-averaged-model'}
+2024-08-03 16:15:49,334 INFO [evaluate.py:221] About to create model
+2024-08-03 16:15:49,967 INFO [evaluate.py:294] Calculating the averaged model over epoch range from 130 (excluded) to 150
+2024-08-03 16:15:57,178 INFO [evaluate.py:312] Number of model parameters: 64559366
+2024-08-03 16:15:57,179 INFO [at_datamodule.py:494] About to get audioset eval all cuts
+2024-08-03 16:15:57,188 INFO [at_datamodule.py:386] About to create dev dataset
+2024-08-03 16:15:57,828 INFO [at_datamodule.py:403] About to create dev dataloader
+2024-08-03 16:16:02,185 INFO [evaluate.py:186] Processed 98 cuts already.
+2024-08-03 16:16:06,178 INFO [evaluate.py:186] Processed 1083 cuts already.
+2024-08-03 16:16:07,152 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.4989, 2.7376, 2.8083, 3.3218], device='cuda:0')
+2024-08-03 16:16:10,170 INFO [evaluate.py:186] Processed 2071 cuts already.
+2024-08-03 16:16:14,155 INFO [evaluate.py:186] Processed 3062 cuts already.
+2024-08-03 16:16:18,123 INFO [evaluate.py:186] Processed 4046 cuts already.
+2024-08-03 16:16:22,086 INFO [evaluate.py:186] Processed 5026 cuts already.
+2024-08-03 16:16:22,657 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6657, 2.9511, 3.1642, 3.4502], device='cuda:0')
+2024-08-03 16:16:26,055 INFO [evaluate.py:186] Processed 6006 cuts already.
+2024-08-03 16:16:28,788 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6769, 3.3601, 3.3066, 3.4100], device='cuda:0')
+2024-08-03 16:16:30,013 INFO [evaluate.py:186] Processed 6986 cuts already.
+2024-08-03 16:16:34,019 INFO [evaluate.py:186] Processed 7966 cuts already.
+2024-08-03 16:16:37,980 INFO [evaluate.py:186] Processed 8946 cuts already.
+2024-08-03 16:16:38,306 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.5993, 3.0337, 2.9772, 3.0119], device='cuda:0')
+2024-08-03 16:16:41,935 INFO [evaluate.py:186] Processed 9926 cuts already.
+2024-08-03 16:16:45,896 INFO [evaluate.py:186] Processed 10906 cuts already.
+2024-08-03 16:16:47,652 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([4.6675, 2.8839, 3.3383, 3.6075], device='cuda:0')
+2024-08-03 16:16:49,860 INFO [evaluate.py:186] Processed 11886 cuts already.
+2024-08-03 16:16:50,396 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.4638, 2.9462, 2.9104, 3.1240, 2.8638, 2.7295, 2.9108, 2.6232],
+       device='cuda:0')
+2024-08-03 16:16:53,825 INFO [evaluate.py:186] Processed 12866 cuts already.
+2024-08-03 16:16:54,951 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.4342, 2.9819, 2.9959, 3.1013, 2.7758, 2.7609, 2.8632, 2.6825],
+       device='cuda:0')
+2024-08-03 16:16:55,963 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([3.8909, 2.9926, 2.7039, 2.9391, 2.4477, 2.8431, 2.5322, 2.4752],
+       device='cuda:0')
+2024-08-03 16:16:57,805 INFO [evaluate.py:186] Processed 13846 cuts already.
+2024-08-03 16:17:01,771 INFO [evaluate.py:186] Processed 14826 cuts already.
+2024-08-03 16:17:02,056 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.9243, 5.6984, 5.7233, 5.6195], device='cuda:0')
+2024-08-03 16:17:05,734 INFO [evaluate.py:186] Processed 15806 cuts already.
+2024-08-03 16:17:09,712 INFO [evaluate.py:186] Processed 16786 cuts already.
+2024-08-03 16:17:13,686 INFO [evaluate.py:186] Processed 17766 cuts already.
+2024-08-03 16:17:15,241 INFO [zipformer.py:1858] name=None, attn_weights_entropy = tensor([5.2630, 4.1359, 2.8531, 4.1041], device='cuda:0')
+2024-08-03 16:17:15,663 INFO [evaluate.py:187] Finish collecting audio logits
+2024-08-03 16:17:19,796 INFO [evaluate.py:338] mAP for audioset eval is: 0.49608092660673614
+2024-08-03 16:17:19,797 INFO [evaluate.py:340] Done