ftshijt commited on
Commit
c4a2b8b
1 Parent(s): 5b3a684

add 20ms, 40ms, 80ms cvhubert

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/checkpoint.pth +3 -0
  2. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/config.yaml +680 -0
  3. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/images/acc_m.png +0 -0
  4. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/images/acc_u.png +0 -0
  5. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/images/backward_time.png +0 -0
  6. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/images/clip.png +0 -0
  7. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/images/correct_m.png +0 -0
  8. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/images/correct_u.png +0 -0
  9. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/images/count_m.png +0 -0
  10. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/images/count_u.png +0 -0
  11. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/images/forward_time.png +0 -0
  12. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/images/gpu_max_cached_mem_GB.png +0 -0
  13. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/images/grad_norm.png +0 -0
  14. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/images/iter_time.png +0 -0
  15. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/images/loss.png +0 -0
  16. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/images/loss_scale.png +0 -0
  17. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/images/optim0_lr0.png +0 -0
  18. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/images/optim_step_time.png +0 -0
  19. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/images/train_time.png +0 -0
  20. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/latest.pth +3 -0
  21. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687916943.learnfair1502.3712237.0 +3 -0
  22. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687916945.learnfair1487.3993792.0 +3 -0
  23. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687917438.learnfair7731.1932016.0 +3 -0
  24. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687920167.learnfair1176.2412152.0 +3 -0
  25. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687921040.learnfair1818.3720269.0 +3 -0
  26. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687922225.learnfair1176.2488951.0 +3 -0
  27. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687973980.devfair0221.1681694.0 +3 -0
  28. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687974173.devfair0221.1682873.0 +3 -0
  29. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687974338.devfair0221.1683709.0 +3 -0
  30. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687974509.devfair0221.1684568.0 +3 -0
  31. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687975323.devfair0221.1686400.0 +3 -0
  32. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687975454.devfair0221.1687214.0 +3 -0
  33. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687975549.devfair0221.1688213.0 +3 -0
  34. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687976840.devfair0221.1689705.0 +3 -0
  35. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687976933.devfair0221.1690515.0 +3 -0
  36. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687977150.devfair0221.1691707.0 +3 -0
  37. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687977407.devfair0221.1692695.0 +3 -0
  38. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687977741.devfair0221.1694145.0 +3 -0
  39. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687977803.devfair0221.1694956.0 +3 -0
  40. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687978054.devfair0221.1695835.0 +3 -0
  41. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687978307.learnfair1513.1588425.0 +3 -0
  42. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1688000555.learnfair1157.2375965.0 +3 -0
  43. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1688011576.learnfair1093.3854371.0 +3 -0
  44. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1688060719.learnfair1953.2654745.0 +3 -0
  45. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1688404967.learnfair1665.169854.0 +3 -0
  46. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/valid/events.out.tfevents.1687916943.learnfair1502.3712237.1 +3 -0
  47. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/valid/events.out.tfevents.1687916945.learnfair1487.3993792.1 +3 -0
  48. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/valid/events.out.tfevents.1687917438.learnfair7731.1932016.1 +3 -0
  49. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/valid/events.out.tfevents.1687920167.learnfair1176.2412152.1 +3 -0
  50. exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/valid/events.out.tfevents.1687921040.learnfair1818.3720269.1 +3 -0
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/checkpoint.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fdd200a95e1f7f665669af9ed1363eb675edee869e73392ce392e5a35a5ce5d4
3
+ size 1149379424
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/config.yaml ADDED
@@ -0,0 +1,680 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ config: conf/tuning/train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms.yaml
2
+ print_config: false
3
+ log_level: INFO
4
+ dry_run: false
5
+ iterator_type: sequence
6
+ output_dir: exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw
7
+ ngpu: 1
8
+ seed: 0
9
+ num_workers: 8
10
+ num_att_plot: 3
11
+ dist_backend: nccl
12
+ dist_init_method: env://
13
+ dist_world_size: 8
14
+ dist_rank: 0
15
+ local_rank: 0
16
+ dist_master_addr: localhost
17
+ dist_master_port: 54445
18
+ dist_launcher: null
19
+ multiprocessing_distributed: true
20
+ unused_parameters: true
21
+ sharded_ddp: false
22
+ cudnn_enabled: true
23
+ cudnn_benchmark: false
24
+ cudnn_deterministic: true
25
+ collect_stats: false
26
+ write_collected_feats: false
27
+ max_epoch: 250
28
+ patience: null
29
+ val_scheduler_criterion:
30
+ - valid
31
+ - loss
32
+ early_stopping_criterion:
33
+ - valid
34
+ - loss
35
+ - min
36
+ best_model_criterion:
37
+ - - valid
38
+ - loss
39
+ - min
40
+ keep_nbest_models: 10
41
+ nbest_averaging_interval: 0
42
+ grad_clip: 5.0
43
+ grad_clip_type: 2.0
44
+ grad_noise: false
45
+ accum_grad: 4
46
+ no_forward_run: false
47
+ resume: true
48
+ train_dtype: float32
49
+ use_amp: true
50
+ log_interval: null
51
+ use_matplotlib: true
52
+ use_tensorboard: true
53
+ create_graph_in_tensorboard: false
54
+ use_wandb: false
55
+ wandb_project: null
56
+ wandb_id: null
57
+ wandb_entity: null
58
+ wandb_name: null
59
+ wandb_model_log_interval: -1
60
+ detect_anomaly: false
61
+ pretrain_path: null
62
+ init_param: []
63
+ ignore_init_mismatch: false
64
+ freeze_param: []
65
+ num_iters_per_epoch: 4000
66
+ batch_size: 20
67
+ valid_batch_size: null
68
+ batch_bins: 20000000
69
+ valid_batch_bins: null
70
+ train_shape_file:
71
+ - exp/hubert_iter2_stats_raw/splits16/speech_shape
72
+ - exp/hubert_iter2_stats_raw/splits16/text_shape.word
73
+ valid_shape_file:
74
+ - exp/hubert_iter2_stats_raw/valid/speech_shape
75
+ - exp/hubert_iter2_stats_raw/valid/text_shape.word
76
+ batch_type: numel
77
+ valid_batch_type: null
78
+ fold_length:
79
+ - 80000
80
+ - 400
81
+ sort_in_batch: descending
82
+ sort_batch: descending
83
+ multiple_iterator: true
84
+ chunk_length: 500
85
+ chunk_shift_ratio: 0.5
86
+ num_cache_chunks: 1024
87
+ chunk_excluded_key_prefixes: []
88
+ train_data_path_and_name_and_type:
89
+ - - exp/hubert_iter2_stats_raw/splits16/wav.scp
90
+ - speech
91
+ - kaldi_ark
92
+ - - exp/hubert_iter2_stats_raw/splits16/text.km.kmeans_iter2_hubert_train_commonvoice_portion0.1
93
+ - text
94
+ - text
95
+ valid_data_path_and_name_and_type:
96
+ - - dump/raw/dev_all_li/wav.scp
97
+ - speech
98
+ - kaldi_ark
99
+ - - dump/raw/dev_all_li/text.km.kmeans_iter2_hubert_train_commonvoice_portion0.1
100
+ - text
101
+ - text
102
+ allow_variable_data_keys: false
103
+ max_cache_size: 0.0
104
+ max_cache_fd: 32
105
+ valid_max_cache_size: null
106
+ exclude_weight_decay: false
107
+ exclude_weight_decay_conf: {}
108
+ optim: adam
109
+ optim_conf:
110
+ lr: 0.0005
111
+ scheduler: warmuplr
112
+ scheduler_conf:
113
+ warmup_steps: 32000
114
+ token_list:
115
+ - '319'
116
+ - '427'
117
+ - '472'
118
+ - '418'
119
+ - '447'
120
+ - '188'
121
+ - '366'
122
+ - '132'
123
+ - '328'
124
+ - '193'
125
+ - '398'
126
+ - '440'
127
+ - '301'
128
+ - '433'
129
+ - '314'
130
+ - '45'
131
+ - '380'
132
+ - '383'
133
+ - '485'
134
+ - '357'
135
+ - '267'
136
+ - '348'
137
+ - '490'
138
+ - '120'
139
+ - '491'
140
+ - '303'
141
+ - '498'
142
+ - '399'
143
+ - '55'
144
+ - '437'
145
+ - '476'
146
+ - '423'
147
+ - '385'
148
+ - '351'
149
+ - '395'
150
+ - '458'
151
+ - '381'
152
+ - '222'
153
+ - '373'
154
+ - '287'
155
+ - '424'
156
+ - '487'
157
+ - '248'
158
+ - '114'
159
+ - '388'
160
+ - '232'
161
+ - '40'
162
+ - '81'
163
+ - '459'
164
+ - '171'
165
+ - '327'
166
+ - '138'
167
+ - '242'
168
+ - '121'
169
+ - '444'
170
+ - '480'
171
+ - '202'
172
+ - '147'
173
+ - '162'
174
+ - '268'
175
+ - '448'
176
+ - '429'
177
+ - '156'
178
+ - '9'
179
+ - '87'
180
+ - '207'
181
+ - '350'
182
+ - '332'
183
+ - '464'
184
+ - '77'
185
+ - '263'
186
+ - '28'
187
+ - '460'
188
+ - '258'
189
+ - '90'
190
+ - '53'
191
+ - '499'
192
+ - '400'
193
+ - '450'
194
+ - '161'
195
+ - '446'
196
+ - '49'
197
+ - '360'
198
+ - '247'
199
+ - '436'
200
+ - '389'
201
+ - '449'
202
+ - '428'
203
+ - '368'
204
+ - '131'
205
+ - '362'
206
+ - '409'
207
+ - '214'
208
+ - '422'
209
+ - '204'
210
+ - '406'
211
+ - '201'
212
+ - '21'
213
+ - '412'
214
+ - '377'
215
+ - '312'
216
+ - '285'
217
+ - '405'
218
+ - '123'
219
+ - '254'
220
+ - '497'
221
+ - '386'
222
+ - '59'
223
+ - '7'
224
+ - '382'
225
+ - '220'
226
+ - '296'
227
+ - '363'
228
+ - '496'
229
+ - '454'
230
+ - '230'
231
+ - '407'
232
+ - '307'
233
+ - '172'
234
+ - '151'
235
+ - '337'
236
+ - '286'
237
+ - '37'
238
+ - '302'
239
+ - '482'
240
+ - '414'
241
+ - '42'
242
+ - '20'
243
+ - '403'
244
+ - '375'
245
+ - '294'
246
+ - '19'
247
+ - '376'
248
+ - '31'
249
+ - '27'
250
+ - '276'
251
+ - '347'
252
+ - '10'
253
+ - '249'
254
+ - '198'
255
+ - '305'
256
+ - '425'
257
+ - '250'
258
+ - '239'
259
+ - '465'
260
+ - '44'
261
+ - '179'
262
+ - '190'
263
+ - '469'
264
+ - '236'
265
+ - '111'
266
+ - '479'
267
+ - '227'
268
+ - '478'
269
+ - '41'
270
+ - '181'
271
+ - '100'
272
+ - '420'
273
+ - '415'
274
+ - '396'
275
+ - '39'
276
+ - '316'
277
+ - '137'
278
+ - '51'
279
+ - '66'
280
+ - '311'
281
+ - '295'
282
+ - '122'
283
+ - '228'
284
+ - '113'
285
+ - '61'
286
+ - '60'
287
+ - '93'
288
+ - '397'
289
+ - '392'
290
+ - '142'
291
+ - '233'
292
+ - '110'
293
+ - '484'
294
+ - '244'
295
+ - '281'
296
+ - '394'
297
+ - '352'
298
+ - '461'
299
+ - '185'
300
+ - '470'
301
+ - '455'
302
+ - '474'
303
+ - '345'
304
+ - '451'
305
+ - '6'
306
+ - '309'
307
+ - '3'
308
+ - '445'
309
+ - '219'
310
+ - '401'
311
+ - '333'
312
+ - '163'
313
+ - '102'
314
+ - '310'
315
+ - '493'
316
+ - '199'
317
+ - '17'
318
+ - '48'
319
+ - '346'
320
+ - '489'
321
+ - '323'
322
+ - '488'
323
+ - '462'
324
+ - '260'
325
+ - '197'
326
+ - '391'
327
+ - '126'
328
+ - '251'
329
+ - '324'
330
+ - '135'
331
+ - '402'
332
+ - '262'
333
+ - '443'
334
+ - '393'
335
+ - '441'
336
+ - '417'
337
+ - '378'
338
+ - '94'
339
+ - '290'
340
+ - '38'
341
+ - '369'
342
+ - '229'
343
+ - '477'
344
+ - '195'
345
+ - '34'
346
+ - '481'
347
+ - '11'
348
+ - '231'
349
+ - '65'
350
+ - '144'
351
+ - '5'
352
+ - '97'
353
+ - '225'
354
+ - '150'
355
+ - '152'
356
+ - '339'
357
+ - '237'
358
+ - '205'
359
+ - '257'
360
+ - '367'
361
+ - '68'
362
+ - '475'
363
+ - '430'
364
+ - '140'
365
+ - '105'
366
+ - '154'
367
+ - '166'
368
+ - '349'
369
+ - '36'
370
+ - '127'
371
+ - '467'
372
+ - '192'
373
+ - '16'
374
+ - '336'
375
+ - '356'
376
+ - '71'
377
+ - '282'
378
+ - '180'
379
+ - '291'
380
+ - '30'
381
+ - '361'
382
+ - '155'
383
+ - '492'
384
+ - '306'
385
+ - '438'
386
+ - '63'
387
+ - '33'
388
+ - '240'
389
+ - '404'
390
+ - '210'
391
+ - '191'
392
+ - '434'
393
+ - '374'
394
+ - '317'
395
+ - '24'
396
+ - '125'
397
+ - '70'
398
+ - '299'
399
+ - '483'
400
+ - '177'
401
+ - '359'
402
+ - '342'
403
+ - '255'
404
+ - '211'
405
+ - '86'
406
+ - '35'
407
+ - '304'
408
+ - '334'
409
+ - '168'
410
+ - '259'
411
+ - '325'
412
+ - '297'
413
+ - '221'
414
+ - '107'
415
+ - '279'
416
+ - '453'
417
+ - '321'
418
+ - '411'
419
+ - '264'
420
+ - '176'
421
+ - '340'
422
+ - '313'
423
+ - '326'
424
+ - '384'
425
+ - '271'
426
+ - '67'
427
+ - '364'
428
+ - '1'
429
+ - '338'
430
+ - '25'
431
+ - '355'
432
+ - '106'
433
+ - '95'
434
+ - '29'
435
+ - '288'
436
+ - '74'
437
+ - '213'
438
+ - '62'
439
+ - '116'
440
+ - '47'
441
+ - '178'
442
+ - '289'
443
+ - '431'
444
+ - '50'
445
+ - '390'
446
+ - '343'
447
+ - '439'
448
+ - '266'
449
+ - '112'
450
+ - '85'
451
+ - '435'
452
+ - '413'
453
+ - '217'
454
+ - '292'
455
+ - '165'
456
+ - '153'
457
+ - '145'
458
+ - '22'
459
+ - '371'
460
+ - '370'
461
+ - '215'
462
+ - '54'
463
+ - '452'
464
+ - '82'
465
+ - '117'
466
+ - '226'
467
+ - '280'
468
+ - '13'
469
+ - '103'
470
+ - '234'
471
+ - '186'
472
+ - '241'
473
+ - '52'
474
+ - '468'
475
+ - '261'
476
+ - '256'
477
+ - '14'
478
+ - '408'
479
+ - '212'
480
+ - '88'
481
+ - '73'
482
+ - '141'
483
+ - '318'
484
+ - '139'
485
+ - '169'
486
+ - '278'
487
+ - '216'
488
+ - '128'
489
+ - '379'
490
+ - '277'
491
+ - '115'
492
+ - '56'
493
+ - '344'
494
+ - '252'
495
+ - '184'
496
+ - '15'
497
+ - '203'
498
+ - '187'
499
+ - '471'
500
+ - '245'
501
+ - '23'
502
+ - '173'
503
+ - '146'
504
+ - '175'
505
+ - '322'
506
+ - '157'
507
+ - '96'
508
+ - '72'
509
+ - '98'
510
+ - '466'
511
+ - '387'
512
+ - '315'
513
+ - '253'
514
+ - '8'
515
+ - '283'
516
+ - '243'
517
+ - '416'
518
+ - '329'
519
+ - '170'
520
+ - '134'
521
+ - '421'
522
+ - '238'
523
+ - '209'
524
+ - '43'
525
+ - '494'
526
+ - '442'
527
+ - '372'
528
+ - '308'
529
+ - '293'
530
+ - '78'
531
+ - '272'
532
+ - '473'
533
+ - '4'
534
+ - '2'
535
+ - '265'
536
+ - '167'
537
+ - '92'
538
+ - '463'
539
+ - '410'
540
+ - '298'
541
+ - '206'
542
+ - '269'
543
+ - '365'
544
+ - '189'
545
+ - '183'
546
+ - '246'
547
+ - '331'
548
+ - '273'
549
+ - '18'
550
+ - '223'
551
+ - '200'
552
+ - '194'
553
+ - '124'
554
+ - '89'
555
+ - '75'
556
+ - '426'
557
+ - '320'
558
+ - '335'
559
+ - '160'
560
+ - '495'
561
+ - '26'
562
+ - '80'
563
+ - '419'
564
+ - '196'
565
+ - '130'
566
+ - '118'
567
+ - '148'
568
+ - '353'
569
+ - '159'
570
+ - '119'
571
+ - '84'
572
+ - '284'
573
+ - '133'
574
+ - '76'
575
+ - '274'
576
+ - '270'
577
+ - '79'
578
+ - '300'
579
+ - '275'
580
+ - '218'
581
+ - '83'
582
+ - '64'
583
+ - '486'
584
+ - '235'
585
+ - '143'
586
+ - '164'
587
+ - '136'
588
+ - '129'
589
+ - '101'
590
+ - '91'
591
+ - '108'
592
+ - '46'
593
+ - '330'
594
+ - '174'
595
+ - '12'
596
+ - '149'
597
+ - '457'
598
+ - '354'
599
+ - '432'
600
+ - '208'
601
+ - '58'
602
+ - '358'
603
+ - '224'
604
+ - '104'
605
+ - '69'
606
+ - '57'
607
+ - '99'
608
+ - '456'
609
+ - '182'
610
+ - '109'
611
+ - '158'
612
+ - '32'
613
+ - <unk>
614
+ - <sos/eos>
615
+ init: null
616
+ collate_fn_conf:
617
+ label_downsampling: 2
618
+ pad: false
619
+ rand_crop: true
620
+ input_size: 1
621
+ num_classes: 500
622
+ use_preprocessor: true
623
+ token_type: word
624
+ bpemodel: null
625
+ non_linguistic_symbols: null
626
+ cleaner: null
627
+ g2p: null
628
+ speech_volume_normalize: null
629
+ rir_scp: null
630
+ rir_apply_prob: 1.0
631
+ noise_scp: null
632
+ noise_apply_prob: 1.0
633
+ noise_db_range: '13_15'
634
+ pred_masked_weight: 1.0
635
+ pred_nomask_weight: 0.0
636
+ loss_weights: 0.0
637
+ frontend: null
638
+ frontend_conf: {}
639
+ specaug: null
640
+ specaug_conf: {}
641
+ normalize: null
642
+ normalize_conf: {}
643
+ preencoder: null
644
+ preencoder_conf: {}
645
+ encoder: torchaudio_hubert
646
+ encoder_conf:
647
+ encoder_projection_dropout: 0.1
648
+ encoder_attention_dropout: 0.1
649
+ encoder_ff_interm_dropout: 0.0
650
+ encoder_dropout: 0.1
651
+ encoder_layer_drop: 0.05
652
+ extractor_conv_layer_config:
653
+ - - 512
654
+ - 10
655
+ - 5
656
+ - - 512
657
+ - 5
658
+ - 4
659
+ - - 512
660
+ - 3
661
+ - 2
662
+ - - 512
663
+ - 3
664
+ - 2
665
+ - - 512
666
+ - 3
667
+ - 2
668
+ - - 512
669
+ - 3
670
+ - 2
671
+ - - 512
672
+ - 3
673
+ - 2
674
+ model: torchaudio
675
+ model_conf: {}
676
+ required:
677
+ - output_dir
678
+ - token_list
679
+ version: '202304'
680
+ distributed: true
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/images/acc_m.png ADDED
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/images/acc_u.png ADDED
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/images/backward_time.png ADDED
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/images/clip.png ADDED
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/images/correct_m.png ADDED
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/images/correct_u.png ADDED
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/images/count_m.png ADDED
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/images/count_u.png ADDED
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/images/forward_time.png ADDED
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/images/gpu_max_cached_mem_GB.png ADDED
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/images/grad_norm.png ADDED
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/images/iter_time.png ADDED
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/images/loss.png ADDED
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/images/loss_scale.png ADDED
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/images/optim0_lr0.png ADDED
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/images/optim_step_time.png ADDED
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/images/train_time.png ADDED
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/latest.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:04eca8a28a9d0036d04acad50068818a4e0343dc27a827ed839a70411477e3c7
3
+ size 383085228
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687916943.learnfair1502.3712237.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d00ed81ae71f1a3797669763499d047289efc33dcb53f14662aa4cee06f772eb
3
+ size 88
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687916945.learnfair1487.3993792.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6ffe6428b6adc5dfe446cb5a3b1aa494ce61a6d0d2eb0fc636a1dd2d8b1137f1
3
+ size 88
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687917438.learnfair7731.1932016.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a637704c93c5fd16d54d4937c7ca05cedf0c745d35b0860a5fb149c079d7d851
3
+ size 88
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687920167.learnfair1176.2412152.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aadb4a805e834381ed40d26ff91d9874dfcb750a4e57c178c1539ac2d33dfc98
3
+ size 88
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687921040.learnfair1818.3720269.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:678014611e3541659fc492a8fef38da3e685509dd8879bd84bb80cbbca7a6294
3
+ size 88
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687922225.learnfair1176.2488951.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c14b0e4e4588925da08095e7e8850227cca3787e3d8192e5061dbe7e2bc4091b
3
+ size 88
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687973980.devfair0221.1681694.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:afc96cd55c5f55616cc2e43f929191443150f0f3a8585b78ff8724d40dc9a1af
3
+ size 88
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687974173.devfair0221.1682873.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e52c9dbb2c0ad42ea4d2c8c6812cf9e7775a45a003ee43ca4fbff5fa0538fc58
3
+ size 88
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687974338.devfair0221.1683709.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:24e3a93d16bef273830e93e98bf3830ba2aae877738049eda193aae446d9107e
3
+ size 88
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687974509.devfair0221.1684568.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b6b74a39b4a1da0f0dab8aaee6c0d01ae064758dab5a8a9fe8de7a1379e8723f
3
+ size 88
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687975323.devfair0221.1686400.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:82885396c26bdcbf69f2a28c6c19fd116fd6ad13bb8030c4e751b187de32fd65
3
+ size 88
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687975454.devfair0221.1687214.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c517d6cd693e7016d5d0e26db94b8b355d57c9cffa81ed06385a7cc277a943c3
3
+ size 88
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687975549.devfair0221.1688213.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8eff59402d67fa4f35c248f18276ded4b8863b0a7cba7fa8dfd4c4884ed85eca
3
+ size 88
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687976840.devfair0221.1689705.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3b74b2987d9841c31bc41966cacb379537a07a0adadcd1fc62f13078ae70f6b2
3
+ size 88
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687976933.devfair0221.1690515.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:eb86e503b7969a88c4152c7d7f2f84714dcbc6b165a6f1f81994d9920eacca8b
3
+ size 88
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687977150.devfair0221.1691707.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e4c68ebdb1d76b5ffed07116f76e59c490cac2bc1b5fea6c4286cb80287ec499
3
+ size 88
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687977407.devfair0221.1692695.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b36668247739cb20062188413f9877f2b6478589492719e863537a94d2558990
3
+ size 834
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687977741.devfair0221.1694145.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:76266da76d9de3f6eed7c00a4518a59f1de42932f69c60d4d579f36dae8302bd
3
+ size 88
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687977803.devfair0221.1694956.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:585f2c3b5f055549526eaa0b835341561489525af57d1b4afcde324dd3156e72
3
+ size 88
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687978054.devfair0221.1695835.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5f2f3c36283e4eccc96b23343db3a58cd2d427c87ceb0de77623d6ea138cd749
3
+ size 834
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1687978307.learnfair1513.1588425.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:edf2ce84d91069b1eafd9b1668352a10718bf7b25896f03f0b7a4fb32f8a45d2
3
+ size 343674
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1688000555.learnfair1157.2375965.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5aa35dd17e7fda422d83d108ef5e4ef1bea906a5525f54a26eeec6cb14a1a321
3
+ size 84872
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1688011576.learnfair1093.3854371.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:95ac6990f81c78848fb026244ac2bdc7495d785b4e8e7a98fce538f7c93d6232
3
+ size 374163
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1688060719.learnfair1953.2654745.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6f45672ef138201416a8ca9c9cc0a8066b14ed8dc016bd48838b933c7ced7f0d
3
+ size 2020545
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/train/events.out.tfevents.1688404967.learnfair1665.169854.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4d8ed3e7714fa9b370450a7d8e6ecd407e74d37b3369ef77c1bf72aef086982e
3
+ size 2020545
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/valid/events.out.tfevents.1687916943.learnfair1502.3712237.1 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:864c60a8989e57bce42e724efec658ff850e3645244de79eae3e1aea26d198e4
3
+ size 88
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/valid/events.out.tfevents.1687916945.learnfair1487.3993792.1 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7ed48bf69fea788b03fcad5f1fb8407741002b471067aff0994bd28ae01a3dcb
3
+ size 88
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/valid/events.out.tfevents.1687917438.learnfair7731.1932016.1 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:74b5cecd74d25471b21f23ada8825d24a07ef4be84c3b3733d64e8e87ad2c6fb
3
+ size 88
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/valid/events.out.tfevents.1687920167.learnfair1176.2412152.1 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5837a1bfc9d3c6170ec9d050f693fa9115cb69f74df43722be461e3c63bc475a
3
+ size 88
exp/hubert_iter2_train_ssl_torchaudiohubert_base_960h_pretrain_it2_40ms_raw/tensorboard/valid/events.out.tfevents.1687921040.learnfair1818.3720269.1 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6429888a705bc3eab22a263a0c3723163e8bf8c7dad47cc8e93de52bcf942c4f
3
+ size 88