Hugging Face's logo

espnet
/
OpenBEATS-Large-i2-dcase21_task5

ESPnet
TensorBoard
audio
classification
Model card Files
xet
Metrics Community
Shikhar Bharadwaj commited on
Commit
c6e8a76
·
1 Parent(s): eeb73bc

Update model

Browse files
Files changed (10) hide show
  1. README.md +271 -0
  2. meta.yaml +8 -0
  3. work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/data/beans_dcase/token_list +20 -0
  4. work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_earlarge2/RESULTS.md +16 -0
  5. work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_earlarge2/config.yaml +219 -0
  6. work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_earlarge2/lightning_logs/version_0/events.out.tfevents.1743859441.gh105.hsn.cm.delta.internal.ncsa.edu.53473.0 +3 -0
  7. work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_earlarge2/lightning_logs/version_0/hparams.yaml +205 -0
  8. work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_earlarge2/lightning_logs/version_1/events.out.tfevents.1743910435.gh096.hsn.cm.delta.internal.ncsa.edu.1386844.0 +3 -0
  9. work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_earlarge2/lightning_logs/version_1/hparams.yaml +205 -0
  10. work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_earlarge2/valid.epoch_mAP.ave_1best.pth +3 -0
README.md ADDED
@@ -0,0 +1,271 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ tags:
3
+ - espnet
4
+ - audio
5
+ - classification
6
+ datasets:
7
+ - beans
8
+ license: cc-by-4.0
9
+ ---
10
+
11
+ ## ESPnet2 CLS model
12
+
13
+ ### `espnet/OpenBEATS-Large-i2-dcase21_task5`
14
+
15
+ This model was trained by Shikhar Bharadwaj using beans recipe in [espnet](https://github.com/espnet/espnet/).
16
+
17
+ ## CLS config
18
+
19
+ <details><summary>expand</summary>
20
+
21
+ ```
22
+ config: /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/earlarge2/conf/ear_large/beans_dcase.yaml
23
+ print_config: false
24
+ log_level: INFO
25
+ drop_last_iter: false
26
+ dry_run: false
27
+ iterator_type: sequence
28
+ valid_iterator_type: null
29
+ output_dir: /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_earlarge2
30
+ ngpu: 0
31
+ seed: 0
32
+ num_workers: 2
33
+ num_att_plot: 0
34
+ dist_backend: nccl
35
+ dist_init_method: env://
36
+ dist_world_size: null
37
+ dist_rank: null
38
+ local_rank: null
39
+ dist_master_addr: null
40
+ dist_master_port: null
41
+ dist_launcher: null
42
+ multiprocessing_distributed: false
43
+ unused_parameters: true
44
+ sharded_ddp: false
45
+ use_deepspeed: false
46
+ deepspeed_config: null
47
+ gradient_as_bucket_view: true
48
+ ddp_comm_hook: null
49
+ cudnn_enabled: true
50
+ cudnn_benchmark: false
51
+ cudnn_deterministic: true
52
+ use_tf32: false
53
+ collect_stats: false
54
+ write_collected_feats: false
55
+ max_epoch: 250
56
+ patience: null
57
+ val_scheduler_criterion:
58
+ - valid
59
+ - loss
60
+ early_stopping_criterion:
61
+ - valid
62
+ - loss
63
+ - min
64
+ best_model_criterion:
65
+ - - valid
66
+ - epoch_mAP
67
+ - max
68
+ keep_nbest_models: 1
69
+ nbest_averaging_interval: 0
70
+ grad_clip: 1
71
+ grad_clip_type: 2.0
72
+ grad_noise: false
73
+ accum_grad: 1
74
+ no_forward_run: false
75
+ resume: true
76
+ train_dtype: float32
77
+ use_amp: false
78
+ log_interval: null
79
+ use_matplotlib: true
80
+ use_tensorboard: true
81
+ create_graph_in_tensorboard: false
82
+ use_wandb: true
83
+ wandb_project: audioverse
84
+ wandb_id: null
85
+ wandb_entity: shikhar
86
+ wandb_name: beans_dcase.earlarge2
87
+ wandb_model_log_interval: -1
88
+ detect_anomaly: false
89
+ use_adapter: false
90
+ adapter: lora
91
+ save_strategy: all
92
+ adapter_conf: {}
93
+ pretrain_path: null
94
+ init_param: []
95
+ ignore_init_mismatch: false
96
+ freeze_param: []
97
+ num_iters_per_epoch: null
98
+ batch_size: 32
99
+ valid_batch_size: 32
100
+ batch_bins: 1000000
101
+ valid_batch_bins: null
102
+ category_sample_size: 10
103
+ train_shape_file:
104
+ - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_stats_16k/train/speech_shape
105
+ - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_stats_16k/train/label_shape
106
+ valid_shape_file:
107
+ - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_stats_16k/valid/speech_shape
108
+ - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_stats_16k/valid/label_shape
109
+ batch_type: folded
110
+ valid_batch_type: null
111
+ fold_length:
112
+ - 160000
113
+ - 35
114
+ sort_in_batch: descending
115
+ shuffle_within_batch: false
116
+ sort_batch: descending
117
+ multiple_iterator: false
118
+ utt2weight_file: null
119
+ chunk_length: 500
120
+ chunk_shift_ratio: 0.5
121
+ num_cache_chunks: 1024
122
+ chunk_excluded_key_prefixes: []
123
+ chunk_default_fs: null
124
+ chunk_max_abs_length: null
125
+ chunk_discard_short_samples: true
126
+ train_data_path_and_name_and_type:
127
+ - - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/dump/beans_dcase/dcase.train/wav.scp
128
+ - speech
129
+ - sound
130
+ - - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/dump/beans_dcase/dcase.train/text
131
+ - label
132
+ - text
133
+ valid_data_path_and_name_and_type:
134
+ - - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/dump/beans_dcase/dcase.dev/wav.scp
135
+ - speech
136
+ - sound
137
+ - - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/dump/beans_dcase/dcase.dev/text
138
+ - label
139
+ - text
140
+ multi_task_dataset: false
141
+ allow_variable_data_keys: false
142
+ max_cache_size: 0.0
143
+ max_cache_fd: 32
144
+ allow_multi_rates: false
145
+ valid_max_cache_size: null
146
+ exclude_weight_decay: false
147
+ exclude_weight_decay_conf: {}
148
+ optim: adamw
149
+ optim_conf:
150
+ lr: 3.0e-05
151
+ weight_decay: 0.01
152
+ betas:
153
+ - 0.9
154
+ - 0.98
155
+ scheduler: cosineannealingwarmuprestarts
156
+ scheduler_conf:
157
+ first_cycle_steps: 95000
158
+ warmup_steps: 8000
159
+ max_lr: 3.0e-05
160
+ min_lr: 5.0e-06
161
+ lightning_conf:
162
+ log_every_n_steps: 250
163
+ max_epochs: 250
164
+ strategy: ddp
165
+ strategy_conf:
166
+ find_unused_parameters: true
167
+ best_model_criterion:
168
+ - - valid/epoch_mAP
169
+ - max
170
+ - 1
171
+ devices: 1
172
+ num_nodes: 1
173
+ default_root_dir: /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_earlarge2
174
+ token_list:
175
+ - ccmk
176
+ - swth
177
+ - rbgr
178
+ - q
179
+ - wtsp
180
+ - oven
181
+ - grn
182
+ - sqt
183
+ - coye
184
+ - btbw
185
+ - gig
186
+ - gcth
187
+ - call
188
+ - savs
189
+ - snmk
190
+ - chsp
191
+ - bbwa
192
+ - amre
193
+ - <unk>
194
+ - <blank>
195
+ text_token_list: null
196
+ text_bpemodel: null
197
+ init: xavier_normal
198
+ input_size: 1
199
+ use_preprocessor: true
200
+ frontend: null
201
+ frontend_conf: {}
202
+ specaug: null
203
+ specaug_conf: {}
204
+ normalize: null
205
+ normalize_conf: {}
206
+ preencoder: null
207
+ preencoder_conf: {}
208
+ encoder: beats
209
+ encoder_conf:
210
+ beats_ckpt_path: /work/nvme/bbjs/sbharadwaj/7Msounds/exp/beats_iter1_large1.tune_lr1.0e-4_warmup40000_bins1600000_totalsteps400000/epoch_latest.pt
211
+ beats_config:
212
+ layer_wise_gradient_decay_ratio: 0.3
213
+ encoder_layerdrop: 0.1
214
+ dropout: 0.0
215
+ use_weighted_representation: false
216
+ specaug_config:
217
+ apply_time_warp: true
218
+ apply_freq_mask: false
219
+ apply_time_mask: true
220
+ time_mask_width_ratio_range:
221
+ - 0
222
+ - 0.06
223
+ num_time_mask: 1
224
+ roll_augment: false
225
+ text_encoder: null
226
+ text_encoder_conf: {}
227
+ embedding_fusion: null
228
+ embedding_fusion_conf: {}
229
+ decoder: linear
230
+ decoder_conf: {}
231
+ model: espnet
232
+ model_conf:
233
+ classification_type: multi-label
234
+ log_epoch_metrics: true
235
+ user_callbacks:
236
+ - mAP_logging
237
+ required:
238
+ - output_dir
239
+ - token_list
240
+ task: cls
241
+ ```
242
+
243
+ </details>
244
+
245
+ ### Citations
246
+
247
+ ```BibTex
248
+
249
+ @article{bharadwaj2025openbeats,
250
+ title={OpenBEATs: A Fully Open-Source General-Purpose Audio Encoder},
251
+ author={Bharadwaj, Shikhar and Cornell, Samuele and Choi, Kwanghee and Fukayama, Satoru and Shim, Hye-jin and Deshmukh, Soham and Watanabe, Shinji},
252
+ journal={arXiv preprint arXiv:2507.14129},
253
+ year={2025}
254
+ }
255
+
256
+ @inproceedings{watanabe2018espnet,
257
+ author={Shinji Watanabe and Takaaki Hori and Shigeki Karita and Tomoki Hayashi and Jiro Nishitoba and Yuya Unno and Nelson Yalta and Jahn Heymann and Matthew Wiesner and Nanxin Chen and Adithya Renduchintala and Tsubasa Ochiai},
258
+ title={{ESPnet}: End-to-End Speech Processing Toolkit},
259
+ year={2018},
260
+ booktitle={Proceedings of Interspeech},
261
+ pages={2207--2211},
262
+ doi={10.21437/Interspeech.2018-1456},
263
+ url={http://dx.doi.org/10.21437/Interspeech.2018-1456}
264
+ }
265
+
266
+
267
+
268
+
269
+
270
+
271
+ ```
meta.yaml ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ espnet: '202503'
2
+ files:
3
+ classification_model_file: /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_earlarge2/valid.epoch_mAP.ave_1best.pth
4
+ python: "3.9.18 | packaged by conda-forge | (main, Dec 23 2023, 17:20:25) \n[GCC 12.3.0]"
5
+ timestamp: 1763331724.215724
6
+ torch: 2.1.2
7
+ yaml_files:
8
+ classification_train_config: /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_earlarge2/config.yaml
work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/data/beans_dcase/token_list ADDED
@@ -0,0 +1,20 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ccmk
2
+ swth
3
+ rbgr
4
+ q
5
+ wtsp
6
+ oven
7
+ grn
8
+ sqt
9
+ coye
10
+ btbw
11
+ gig
12
+ gcth
13
+ call
14
+ savs
15
+ snmk
16
+ chsp
17
+ bbwa
18
+ amre
19
+ <unk>
20
+ <blank>
work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_earlarge2/RESULTS.md ADDED
@@ -0,0 +1,16 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ <!-- Generated by scripts/utils/show_cls_result.sh -->
2
+ # RESULTS
3
+ ## Environments
4
+ - date: `Sun Apr 6 10:18:55 CDT 2025`
5
+ - python version: `3.9.18 | packaged by conda-forge | (main, Dec 23 2023, 17:20:25) [GCC 12.3.0]`
6
+ - espnet version: `espnet 202412`
7
+ - pytorch version: `pytorch 2.6.0.dev20241210+cu124`
8
+ - Git hash: `c96433a43c5c3984889b81804becac6ebf10f7a7`
9
+ - Commit date: `Mon Mar 31 20:24:06 2025 -0500`
10
+
11
+ ## cls_earlarge2
12
+ |Split|mean_acc|mAP|mean_auc|n_labels|n_instances|
13
+ |---|---|---|---|---|---|
14
+ cls_dcase.dev|12.94|51.57|98.35|18.00|13806.00
15
+ cls_dcase.test|12.84|47.58|98.26|18.00|13530.00
16
+
work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_earlarge2/config.yaml ADDED
@@ -0,0 +1,219 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ config: /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/earlarge2/conf/ear_large/beans_dcase.yaml
2
+ print_config: false
3
+ log_level: INFO
4
+ drop_last_iter: false
5
+ dry_run: false
6
+ iterator_type: sequence
7
+ valid_iterator_type: null
8
+ output_dir: /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_earlarge2
9
+ ngpu: 0
10
+ seed: 0
11
+ num_workers: 2
12
+ num_att_plot: 0
13
+ dist_backend: nccl
14
+ dist_init_method: env://
15
+ dist_world_size: null
16
+ dist_rank: null
17
+ local_rank: null
18
+ dist_master_addr: null
19
+ dist_master_port: null
20
+ dist_launcher: null
21
+ multiprocessing_distributed: false
22
+ unused_parameters: true
23
+ sharded_ddp: false
24
+ use_deepspeed: false
25
+ deepspeed_config: null
26
+ gradient_as_bucket_view: true
27
+ ddp_comm_hook: null
28
+ cudnn_enabled: true
29
+ cudnn_benchmark: false
30
+ cudnn_deterministic: true
31
+ use_tf32: false
32
+ collect_stats: false
33
+ write_collected_feats: false
34
+ max_epoch: 250
35
+ patience: null
36
+ val_scheduler_criterion:
37
+ - valid
38
+ - loss
39
+ early_stopping_criterion:
40
+ - valid
41
+ - loss
42
+ - min
43
+ best_model_criterion:
44
+ - - valid
45
+ - epoch_mAP
46
+ - max
47
+ keep_nbest_models: 1
48
+ nbest_averaging_interval: 0
49
+ grad_clip: 1
50
+ grad_clip_type: 2.0
51
+ grad_noise: false
52
+ accum_grad: 1
53
+ no_forward_run: false
54
+ resume: true
55
+ train_dtype: float32
56
+ use_amp: false
57
+ log_interval: null
58
+ use_matplotlib: true
59
+ use_tensorboard: true
60
+ create_graph_in_tensorboard: false
61
+ use_wandb: true
62
+ wandb_project: audioverse
63
+ wandb_id: null
64
+ wandb_entity: shikhar
65
+ wandb_name: beans_dcase.earlarge2
66
+ wandb_model_log_interval: -1
67
+ detect_anomaly: false
68
+ use_adapter: false
69
+ adapter: lora
70
+ save_strategy: all
71
+ adapter_conf: {}
72
+ pretrain_path: null
73
+ init_param: []
74
+ ignore_init_mismatch: false
75
+ freeze_param: []
76
+ num_iters_per_epoch: null
77
+ batch_size: 32
78
+ valid_batch_size: 32
79
+ batch_bins: 1000000
80
+ valid_batch_bins: null
81
+ category_sample_size: 10
82
+ train_shape_file:
83
+ - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_stats_16k/train/speech_shape
84
+ - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_stats_16k/train/label_shape
85
+ valid_shape_file:
86
+ - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_stats_16k/valid/speech_shape
87
+ - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_stats_16k/valid/label_shape
88
+ batch_type: folded
89
+ valid_batch_type: null
90
+ fold_length:
91
+ - 160000
92
+ - 35
93
+ sort_in_batch: descending
94
+ shuffle_within_batch: false
95
+ sort_batch: descending
96
+ multiple_iterator: false
97
+ utt2weight_file: null
98
+ chunk_length: 500
99
+ chunk_shift_ratio: 0.5
100
+ num_cache_chunks: 1024
101
+ chunk_excluded_key_prefixes: []
102
+ chunk_default_fs: null
103
+ chunk_max_abs_length: null
104
+ chunk_discard_short_samples: true
105
+ train_data_path_and_name_and_type:
106
+ - - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/dump/beans_dcase/dcase.train/wav.scp
107
+ - speech
108
+ - sound
109
+ - - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/dump/beans_dcase/dcase.train/text
110
+ - label
111
+ - text
112
+ valid_data_path_and_name_and_type:
113
+ - - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/dump/beans_dcase/dcase.dev/wav.scp
114
+ - speech
115
+ - sound
116
+ - - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/dump/beans_dcase/dcase.dev/text
117
+ - label
118
+ - text
119
+ multi_task_dataset: false
120
+ allow_variable_data_keys: false
121
+ max_cache_size: 0.0
122
+ max_cache_fd: 32
123
+ allow_multi_rates: false
124
+ valid_max_cache_size: null
125
+ exclude_weight_decay: false
126
+ exclude_weight_decay_conf: {}
127
+ optim: adamw
128
+ optim_conf:
129
+ lr: 3.0e-05
130
+ weight_decay: 0.01
131
+ betas:
132
+ - 0.9
133
+ - 0.98
134
+ scheduler: cosineannealingwarmuprestarts
135
+ scheduler_conf:
136
+ first_cycle_steps: 95000
137
+ warmup_steps: 8000
138
+ max_lr: 3.0e-05
139
+ min_lr: 5.0e-06
140
+ lightning_conf:
141
+ log_every_n_steps: 250
142
+ max_epochs: 250
143
+ strategy: ddp
144
+ strategy_conf:
145
+ find_unused_parameters: true
146
+ best_model_criterion:
147
+ - - valid/epoch_mAP
148
+ - max
149
+ - 1
150
+ devices: 1
151
+ num_nodes: 1
152
+ default_root_dir: /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_earlarge2
153
+ token_list:
154
+ - ccmk
155
+ - swth
156
+ - rbgr
157
+ - q
158
+ - wtsp
159
+ - oven
160
+ - grn
161
+ - sqt
162
+ - coye
163
+ - btbw
164
+ - gig
165
+ - gcth
166
+ - call
167
+ - savs
168
+ - snmk
169
+ - chsp
170
+ - bbwa
171
+ - amre
172
+ - <unk>
173
+ - <blank>
174
+ text_token_list: null
175
+ text_bpemodel: null
176
+ init: xavier_normal
177
+ input_size: 1
178
+ use_preprocessor: true
179
+ frontend: null
180
+ frontend_conf: {}
181
+ specaug: null
182
+ specaug_conf: {}
183
+ normalize: null
184
+ normalize_conf: {}
185
+ preencoder: null
186
+ preencoder_conf: {}
187
+ encoder: beats
188
+ encoder_conf:
189
+ beats_ckpt_path: /work/nvme/bbjs/sbharadwaj/7Msounds/exp/beats_iter1_large1.tune_lr1.0e-4_warmup40000_bins1600000_totalsteps400000/epoch_latest.pt
190
+ beats_config:
191
+ layer_wise_gradient_decay_ratio: 0.3
192
+ encoder_layerdrop: 0.1
193
+ dropout: 0.0
194
+ use_weighted_representation: false
195
+ specaug_config:
196
+ apply_time_warp: true
197
+ apply_freq_mask: false
198
+ apply_time_mask: true
199
+ time_mask_width_ratio_range:
200
+ - 0
201
+ - 0.06
202
+ num_time_mask: 1
203
+ roll_augment: false
204
+ text_encoder: null
205
+ text_encoder_conf: {}
206
+ embedding_fusion: null
207
+ embedding_fusion_conf: {}
208
+ decoder: linear
209
+ decoder_conf: {}
210
+ model: espnet
211
+ model_conf:
212
+ classification_type: multi-label
213
+ log_epoch_metrics: true
214
+ user_callbacks:
215
+ - mAP_logging
216
+ required:
217
+ - output_dir
218
+ - token_list
219
+ task: cls
work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_earlarge2/lightning_logs/version_0/events.out.tfevents.1743859441.gh105.hsn.cm.delta.internal.ncsa.edu.53473.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dbf4a813946065fc61ba1b309e0c884422644a72d4aabc2a5c105c8475d91e63
3
+ size 40046
work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_earlarge2/lightning_logs/version_0/hparams.yaml ADDED
@@ -0,0 +1,205 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ args: !!python/object:argparse.Namespace
2
+ accum_grad: 1
3
+ adapter: lora
4
+ adapter_conf: {}
5
+ allow_multi_rates: false
6
+ allow_variable_data_keys: false
7
+ batch_bins: 1000000
8
+ batch_size: 32
9
+ batch_type: folded
10
+ best_model_criterion:
11
+ - - valid
12
+ - epoch_mAP
13
+ - max
14
+ category_sample_size: 10
15
+ chunk_default_fs: null
16
+ chunk_discard_short_samples: true
17
+ chunk_excluded_key_prefixes: []
18
+ chunk_length: 500
19
+ chunk_max_abs_length: null
20
+ chunk_shift_ratio: 0.5
21
+ collect_stats: false
22
+ config: /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/earlarge2/conf/ear_large/beans_dcase.yaml
23
+ create_graph_in_tensorboard: false
24
+ cudnn_benchmark: false
25
+ cudnn_deterministic: true
26
+ cudnn_enabled: true
27
+ ddp_comm_hook: null
28
+ decoder: linear
29
+ decoder_conf: {}
30
+ deepspeed_config: null
31
+ detect_anomaly: false
32
+ dist_backend: nccl
33
+ dist_init_method: env://
34
+ dist_launcher: null
35
+ dist_master_addr: null
36
+ dist_master_port: null
37
+ dist_rank: null
38
+ dist_world_size: null
39
+ drop_last_iter: false
40
+ dry_run: false
41
+ early_stopping_criterion: !!python/tuple
42
+ - valid
43
+ - loss
44
+ - min
45
+ embedding_fusion: null
46
+ embedding_fusion_conf: {}
47
+ encoder: beats
48
+ encoder_conf:
49
+ beats_ckpt_path: /work/nvme/bbjs/sbharadwaj/7Msounds/exp/beats_iter1_large1.tune_lr1.0e-4_warmup40000_bins1600000_totalsteps400000/epoch_latest.pt
50
+ beats_config:
51
+ dropout: 0.0
52
+ encoder_layerdrop: 0.1
53
+ layer_wise_gradient_decay_ratio: 0.3
54
+ roll_augment: false
55
+ specaug_config:
56
+ apply_freq_mask: false
57
+ apply_time_mask: true
58
+ apply_time_warp: true
59
+ num_time_mask: 1
60
+ time_mask_width_ratio_range:
61
+ - 0
62
+ - 0.06
63
+ use_weighted_representation: false
64
+ exclude_weight_decay: false
65
+ exclude_weight_decay_conf: {}
66
+ fold_length:
67
+ - 160000
68
+ - 35
69
+ freeze_param: []
70
+ frontend: null
71
+ frontend_conf:
72
+ fs: 16k
73
+ grad_clip: 1
74
+ grad_clip_type: 2.0
75
+ grad_noise: false
76
+ gradient_as_bucket_view: true
77
+ ignore_init_mismatch: false
78
+ init: xavier_normal
79
+ init_param: []
80
+ input_size: 1
81
+ iterator_type: sequence
82
+ keep_nbest_models: 1
83
+ lightning_conf:
84
+ best_model_criterion:
85
+ - - valid/epoch_mAP
86
+ - max
87
+ - 1
88
+ default_root_dir: /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_earlarge2
89
+ devices: 1
90
+ log_every_n_steps: 250
91
+ max_epochs: 250
92
+ num_nodes: 1
93
+ strategy: ddp
94
+ strategy_conf:
95
+ find_unused_parameters: true
96
+ local_rank: null
97
+ log_interval: null
98
+ log_level: INFO
99
+ max_cache_fd: 32
100
+ max_cache_size: 0.0
101
+ max_epoch: 250
102
+ model: espnet
103
+ model_conf:
104
+ classification_type: multi-label
105
+ log_epoch_metrics: true
106
+ multi_task_dataset: false
107
+ multiple_iterator: false
108
+ multiprocessing_distributed: false
109
+ nbest_averaging_interval: 0
110
+ ngpu: 0
111
+ no_forward_run: false
112
+ normalize: null
113
+ normalize_conf: {}
114
+ num_att_plot: 0
115
+ num_cache_chunks: 1024
116
+ num_iters_per_epoch: null
117
+ num_workers: 2
118
+ optim: adamw
119
+ optim_conf:
120
+ betas:
121
+ - 0.9
122
+ - 0.98
123
+ lr: 3.0e-05
124
+ weight_decay: 0.01
125
+ output_dir: /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_earlarge2
126
+ patience: null
127
+ preencoder: null
128
+ preencoder_conf: {}
129
+ pretrain_path: null
130
+ print_config: false
131
+ required:
132
+ - output_dir
133
+ - token_list
134
+ resume: true
135
+ save_strategy: all
136
+ scheduler: cosineannealingwarmuprestarts
137
+ scheduler_conf:
138
+ first_cycle_steps: 95000
139
+ max_lr: 3.0e-05
140
+ min_lr: 5.0e-06
141
+ warmup_steps: 8000
142
+ seed: 0
143
+ sharded_ddp: false
144
+ shuffle_within_batch: false
145
+ sort_batch: descending
146
+ sort_in_batch: descending
147
+ specaug: null
148
+ specaug_conf: {}
149
+ task: cls
150
+ text_bpemodel: null
151
+ text_encoder: null
152
+ text_encoder_conf: {}
153
+ text_token_list: null
154
+ token_list: /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/data/beans_dcase/token_list
155
+ train_data_path_and_name_and_type:
156
+ - !!python/tuple
157
+ - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/dump/beans_dcase/dcase.train/wav.scp
158
+ - speech
159
+ - sound
160
+ - !!python/tuple
161
+ - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/dump/beans_dcase/dcase.train/text
162
+ - label
163
+ - text
164
+ train_dtype: float32
165
+ train_shape_file:
166
+ - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_stats_16k/train/speech_shape
167
+ - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_stats_16k/train/label_shape
168
+ unused_parameters: true
169
+ use_adapter: false
170
+ use_amp: false
171
+ use_deepspeed: false
172
+ use_matplotlib: true
173
+ use_preprocessor: true
174
+ use_tensorboard: true
175
+ use_tf32: false
176
+ use_wandb: true
177
+ user_callbacks:
178
+ - mAP_logging
179
+ utt2weight_file: null
180
+ val_scheduler_criterion: !!python/tuple
181
+ - valid
182
+ - loss
183
+ valid_batch_bins: null
184
+ valid_batch_size: 32
185
+ valid_batch_type: null
186
+ valid_data_path_and_name_and_type:
187
+ - !!python/tuple
188
+ - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/dump/beans_dcase/dcase.dev/wav.scp
189
+ - speech
190
+ - sound
191
+ - !!python/tuple
192
+ - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/dump/beans_dcase/dcase.dev/text
193
+ - label
194
+ - text
195
+ valid_iterator_type: null
196
+ valid_max_cache_size: null
197
+ valid_shape_file:
198
+ - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_stats_16k/valid/speech_shape
199
+ - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_stats_16k/valid/label_shape
200
+ wandb_entity: shikhar
201
+ wandb_id: null
202
+ wandb_model_log_interval: -1
203
+ wandb_name: beans_dcase.earlarge2
204
+ wandb_project: audioverse
205
+ write_collected_feats: false
work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_earlarge2/lightning_logs/version_1/events.out.tfevents.1743910435.gh096.hsn.cm.delta.internal.ncsa.edu.1386844.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:59c6d5b70fcc1e386ec50c265d71e4cbf8f984b77b898b16ce145859b9bcf1ae
3
+ size 344417
work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_earlarge2/lightning_logs/version_1/hparams.yaml ADDED
@@ -0,0 +1,205 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ args: !!python/object:argparse.Namespace
2
+ accum_grad: 1
3
+ adapter: lora
4
+ adapter_conf: {}
5
+ allow_multi_rates: false
6
+ allow_variable_data_keys: false
7
+ batch_bins: 1000000
8
+ batch_size: 32
9
+ batch_type: folded
10
+ best_model_criterion:
11
+ - - valid
12
+ - epoch_mAP
13
+ - max
14
+ category_sample_size: 10
15
+ chunk_default_fs: null
16
+ chunk_discard_short_samples: true
17
+ chunk_excluded_key_prefixes: []
18
+ chunk_length: 500
19
+ chunk_max_abs_length: null
20
+ chunk_shift_ratio: 0.5
21
+ collect_stats: false
22
+ config: /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/earlarge2/conf/ear_large/beans_dcase.yaml
23
+ create_graph_in_tensorboard: false
24
+ cudnn_benchmark: false
25
+ cudnn_deterministic: true
26
+ cudnn_enabled: true
27
+ ddp_comm_hook: null
28
+ decoder: linear
29
+ decoder_conf: {}
30
+ deepspeed_config: null
31
+ detect_anomaly: false
32
+ dist_backend: nccl
33
+ dist_init_method: env://
34
+ dist_launcher: null
35
+ dist_master_addr: null
36
+ dist_master_port: null
37
+ dist_rank: null
38
+ dist_world_size: null
39
+ drop_last_iter: false
40
+ dry_run: false
41
+ early_stopping_criterion: !!python/tuple
42
+ - valid
43
+ - loss
44
+ - min
45
+ embedding_fusion: null
46
+ embedding_fusion_conf: {}
47
+ encoder: beats
48
+ encoder_conf:
49
+ beats_ckpt_path: /work/nvme/bbjs/sbharadwaj/7Msounds/exp/beats_iter1_large1.tune_lr1.0e-4_warmup40000_bins1600000_totalsteps400000/epoch_latest.pt
50
+ beats_config:
51
+ dropout: 0.0
52
+ encoder_layerdrop: 0.1
53
+ layer_wise_gradient_decay_ratio: 0.3
54
+ roll_augment: false
55
+ specaug_config:
56
+ apply_freq_mask: false
57
+ apply_time_mask: true
58
+ apply_time_warp: true
59
+ num_time_mask: 1
60
+ time_mask_width_ratio_range:
61
+ - 0
62
+ - 0.06
63
+ use_weighted_representation: false
64
+ exclude_weight_decay: false
65
+ exclude_weight_decay_conf: {}
66
+ fold_length:
67
+ - 160000
68
+ - 35
69
+ freeze_param: []
70
+ frontend: null
71
+ frontend_conf:
72
+ fs: 16k
73
+ grad_clip: 1
74
+ grad_clip_type: 2.0
75
+ grad_noise: false
76
+ gradient_as_bucket_view: true
77
+ ignore_init_mismatch: false
78
+ init: xavier_normal
79
+ init_param: []
80
+ input_size: 1
81
+ iterator_type: sequence
82
+ keep_nbest_models: 1
83
+ lightning_conf:
84
+ best_model_criterion:
85
+ - - valid/epoch_mAP
86
+ - max
87
+ - 1
88
+ default_root_dir: /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_earlarge2
89
+ devices: 1
90
+ log_every_n_steps: 250
91
+ max_epochs: 250
92
+ num_nodes: 1
93
+ strategy: ddp
94
+ strategy_conf:
95
+ find_unused_parameters: true
96
+ local_rank: null
97
+ log_interval: null
98
+ log_level: INFO
99
+ max_cache_fd: 32
100
+ max_cache_size: 0.0
101
+ max_epoch: 250
102
+ model: espnet
103
+ model_conf:
104
+ classification_type: multi-label
105
+ log_epoch_metrics: true
106
+ multi_task_dataset: false
107
+ multiple_iterator: false
108
+ multiprocessing_distributed: false
109
+ nbest_averaging_interval: 0
110
+ ngpu: 0
111
+ no_forward_run: false
112
+ normalize: null
113
+ normalize_conf: {}
114
+ num_att_plot: 0
115
+ num_cache_chunks: 1024
116
+ num_iters_per_epoch: null
117
+ num_workers: 2
118
+ optim: adamw
119
+ optim_conf:
120
+ betas:
121
+ - 0.9
122
+ - 0.98
123
+ lr: 3.0e-05
124
+ weight_decay: 0.01
125
+ output_dir: /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_earlarge2
126
+ patience: null
127
+ preencoder: null
128
+ preencoder_conf: {}
129
+ pretrain_path: null
130
+ print_config: false
131
+ required:
132
+ - output_dir
133
+ - token_list
134
+ resume: true
135
+ save_strategy: all
136
+ scheduler: cosineannealingwarmuprestarts
137
+ scheduler_conf:
138
+ first_cycle_steps: 95000
139
+ max_lr: 3.0e-05
140
+ min_lr: 5.0e-06
141
+ warmup_steps: 8000
142
+ seed: 0
143
+ sharded_ddp: false
144
+ shuffle_within_batch: false
145
+ sort_batch: descending
146
+ sort_in_batch: descending
147
+ specaug: null
148
+ specaug_conf: {}
149
+ task: cls
150
+ text_bpemodel: null
151
+ text_encoder: null
152
+ text_encoder_conf: {}
153
+ text_token_list: null
154
+ token_list: /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/data/beans_dcase/token_list
155
+ train_data_path_and_name_and_type:
156
+ - !!python/tuple
157
+ - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/dump/beans_dcase/dcase.train/wav.scp
158
+ - speech
159
+ - sound
160
+ - !!python/tuple
161
+ - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/dump/beans_dcase/dcase.train/text
162
+ - label
163
+ - text
164
+ train_dtype: float32
165
+ train_shape_file:
166
+ - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_stats_16k/train/speech_shape
167
+ - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_stats_16k/train/label_shape
168
+ unused_parameters: true
169
+ use_adapter: false
170
+ use_amp: false
171
+ use_deepspeed: false
172
+ use_matplotlib: true
173
+ use_preprocessor: true
174
+ use_tensorboard: true
175
+ use_tf32: false
176
+ use_wandb: true
177
+ user_callbacks:
178
+ - mAP_logging
179
+ utt2weight_file: null
180
+ val_scheduler_criterion: !!python/tuple
181
+ - valid
182
+ - loss
183
+ valid_batch_bins: null
184
+ valid_batch_size: 32
185
+ valid_batch_type: null
186
+ valid_data_path_and_name_and_type:
187
+ - !!python/tuple
188
+ - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/dump/beans_dcase/dcase.dev/wav.scp
189
+ - speech
190
+ - sound
191
+ - !!python/tuple
192
+ - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/dump/beans_dcase/dcase.dev/text
193
+ - label
194
+ - text
195
+ valid_iterator_type: null
196
+ valid_max_cache_size: null
197
+ valid_shape_file:
198
+ - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_stats_16k/valid/speech_shape
199
+ - /work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_stats_16k/valid/label_shape
200
+ wandb_entity: shikhar
201
+ wandb_id: null
202
+ wandb_model_log_interval: -1
203
+ wandb_name: beans_dcase.earlarge2
204
+ wandb_project: audioverse
205
+ write_collected_feats: false
work/nvme/bbjs/sbharadwaj/espnet/egs2/audioverse/v1/exp/beans_dcase/cls_earlarge2/valid.epoch_mAP.ave_1best.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:48b3b3446ff0a0f4556cd20ce2244d9d4e7af4c121e0383da604d4e67befa322
3
+ size 1246221850