Pu Wang commited on
Commit
cbf161b
·
1 Parent(s): 84d19d3

new file: exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/RESULTS.md

Browse files

new file: exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/config.yaml
new file: exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/acc.png
new file: exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/backward_time.png
new file: exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/cer.png
new file: exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/cer_ctc.png
new file: exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/clip.png
new file: exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/forward_time.png
new file: exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/gpu_max_cached_mem_GB.png
new file: exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/grad_norm.png
new file: exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/iter_time.png
new file: exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/loss.png
new file: exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/loss_att.png
new file: exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/loss_ctc.png
new file: exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/loss_scale.png
new file: exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/optim0_lr0.png
new file: exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/optim_step_time.png
new file: exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/train_time.png
new file: exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/wer.png
new file: exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/valid.acc.ave_4best.pth

Files changed (21) hide show
  1. exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/.config.yaml.swp +0 -0
  2. exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/RESULTS.md +29 -0
  3. exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/config.yaml +1123 -0
  4. exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/acc.png +0 -0
  5. exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/backward_time.png +0 -0
  6. exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/cer.png +0 -0
  7. exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/cer_ctc.png +0 -0
  8. exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/clip.png +0 -0
  9. exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/forward_time.png +0 -0
  10. exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/gpu_max_cached_mem_GB.png +0 -0
  11. exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/grad_norm.png +0 -0
  12. exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/iter_time.png +0 -0
  13. exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/loss.png +0 -0
  14. exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/loss_att.png +0 -0
  15. exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/loss_ctc.png +0 -0
  16. exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/loss_scale.png +0 -0
  17. exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/optim0_lr0.png +0 -0
  18. exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/optim_step_time.png +0 -0
  19. exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/train_time.png +0 -0
  20. exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/wer.png +0 -0
  21. exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/valid.acc.ave_4best.pth +3 -0
exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/.config.yaml.swp ADDED
Binary file (16.4 kB). View file
 
exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/RESULTS.md ADDED
@@ -0,0 +1,29 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ <!-- Generated by scripts/utils/show_asr_result.sh -->
2
+ # RESULTS
3
+ ## Environments
4
+ - date: `Sun Jan 19 06:55:28 EST 2025`
5
+ - python version: `3.10.16 (main, Dec 11 2024, 16:24:50) [GCC 11.2.0]`
6
+ - espnet version: `espnet 202412`
7
+ - pytorch version: `pytorch 2.4.0`
8
+ - Git hash: `0fe7b8581fbc68841eb48776f052aa9a5989108c`
9
+ - Commit date: `Tue Jan 14 20:06:15 2025 -0500`
10
+
11
+ ## exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp
12
+ ### WER
13
+
14
+ |dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err|
15
+ |---|---|---|---|---|---|---|---|---|
16
+ |decode_asr_asr_model_valid.acc.best/test|754|6005|98.3|0.7|1.0|0.6|2.3|6.6|
17
+
18
+ ### CER
19
+
20
+ |dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err|
21
+ |---|---|---|---|---|---|---|---|---|
22
+ |decode_asr_asr_model_valid.acc.best/test|754|31847|98.7|0.3|1.1|0.6|1.9|6.6|
23
+
24
+ ### TER
25
+
26
+ |dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err|
27
+ |---|---|---|---|---|---|---|---|---|
28
+ |decode_asr_asr_model_valid.acc.best/test|754|9046|98.4|0.4|1.1|0.7|2.3|6.6|
29
+
exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/config.yaml ADDED
@@ -0,0 +1,1123 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ config: conf/tuning/train_asr_wavlm_transformer.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: exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp
9
+ ngpu: 1
10
+ seed: 2022
11
+ num_workers: 4
12
+ num_att_plot: 0
13
+ dist_backend: nccl
14
+ dist_init_method: env://
15
+ dist_world_size: 2
16
+ dist_rank: 0
17
+ local_rank: 0
18
+ dist_master_addr: localhost
19
+ dist_master_port: 35147
20
+ dist_launcher: null
21
+ multiprocessing_distributed: true
22
+ unused_parameters: false
23
+ sharded_ddp: false
24
+ use_deepspeed: false
25
+ deepspeed_config: null
26
+ cudnn_enabled: true
27
+ cudnn_benchmark: false
28
+ cudnn_deterministic: false
29
+ use_tf32: false
30
+ collect_stats: false
31
+ write_collected_feats: false
32
+ max_epoch: 100
33
+ patience: null
34
+ val_scheduler_criterion:
35
+ - valid
36
+ - loss
37
+ early_stopping_criterion:
38
+ - valid
39
+ - loss
40
+ - min
41
+ best_model_criterion:
42
+ - - valid
43
+ - acc
44
+ - max
45
+ keep_nbest_models: 4
46
+ nbest_averaging_interval: 0
47
+ grad_clip: 5.0
48
+ grad_clip_type: 2.0
49
+ grad_noise: false
50
+ accum_grad: 4
51
+ no_forward_run: false
52
+ resume: true
53
+ train_dtype: float32
54
+ use_amp: true
55
+ log_interval: 400
56
+ use_matplotlib: true
57
+ use_tensorboard: true
58
+ create_graph_in_tensorboard: false
59
+ use_wandb: false
60
+ wandb_project: null
61
+ wandb_id: null
62
+ wandb_entity: null
63
+ wandb_name: null
64
+ wandb_model_log_interval: -1
65
+ detect_anomaly: false
66
+ use_adapter: false
67
+ adapter: lora
68
+ save_strategy: all
69
+ adapter_conf: {}
70
+ pretrain_path: null
71
+ init_param: []
72
+ ignore_init_mismatch: false
73
+ freeze_param:
74
+ - frontend.upstream
75
+ num_iters_per_epoch: null
76
+ batch_size: 20
77
+ valid_batch_size: null
78
+ batch_bins: 16000000
79
+ valid_batch_bins: null
80
+ category_sample_size: 10
81
+ train_shape_file:
82
+ - exp/asr_stats_raw_en_bpe900_sp/train/speech_shape
83
+ - exp/asr_stats_raw_en_bpe900_sp/train/text_shape.bpe
84
+ valid_shape_file:
85
+ - exp/asr_stats_raw_en_bpe900_sp/valid/speech_shape
86
+ - exp/asr_stats_raw_en_bpe900_sp/valid/text_shape.bpe
87
+ batch_type: numel
88
+ valid_batch_type: null
89
+ fold_length:
90
+ - 80000
91
+ - 150
92
+ sort_in_batch: descending
93
+ shuffle_within_batch: false
94
+ sort_batch: descending
95
+ multiple_iterator: false
96
+ chunk_length: 500
97
+ chunk_shift_ratio: 0.5
98
+ num_cache_chunks: 1024
99
+ chunk_excluded_key_prefixes: []
100
+ chunk_default_fs: null
101
+ chunk_max_abs_length: null
102
+ chunk_discard_short_samples: true
103
+ train_data_path_and_name_and_type:
104
+ - - dump/raw/train_sp/wav.scp
105
+ - speech
106
+ - sound
107
+ - - dump/raw/train_sp/text
108
+ - text
109
+ - text
110
+ valid_data_path_and_name_and_type:
111
+ - - dump/raw/dev/wav.scp
112
+ - speech
113
+ - sound
114
+ - - dump/raw/dev/text
115
+ - text
116
+ - text
117
+ multi_task_dataset: false
118
+ allow_variable_data_keys: false
119
+ max_cache_size: 0.0
120
+ max_cache_fd: 32
121
+ allow_multi_rates: false
122
+ valid_max_cache_size: null
123
+ exclude_weight_decay: false
124
+ exclude_weight_decay_conf: {}
125
+ optim: adam
126
+ optim_conf:
127
+ lr: 0.002
128
+ weight_decay: 1.0e-06
129
+ scheduler: warmuplr
130
+ scheduler_conf:
131
+ warmup_steps: 15000
132
+ token_list:
133
+ - <blank>
134
+ - <unk>
135
+ - ▁THE
136
+ - S
137
+ - ▁
138
+ - Y
139
+ - ▁A
140
+ - ▁IN
141
+ - ▁TO
142
+ - N
143
+ - ▁OF
144
+ - ▁PEOPLE
145
+ - AND
146
+ - ▁IT
147
+ - ▁ON
148
+ - ▁IS
149
+ - T
150
+ - RE
151
+ - ING
152
+ - ▁LIVE
153
+ - ▁SOME
154
+ - E
155
+ - D
156
+ - ▁CAN
157
+ - ▁WATER
158
+ - IR
159
+ - ▁MA
160
+ - ▁BUTTERFLIES
161
+ - ▁USE
162
+ - ST
163
+ - ▁M
164
+ - ▁MO
165
+ - ▁SCIENTIST
166
+ - ARE
167
+ - ▁T
168
+ - ▁RAIN
169
+ - ▁EAT
170
+ - M
171
+ - NOT
172
+ - ▁THEM
173
+ - ▁HA
174
+ - ▁FOOD
175
+ - ▁THAT
176
+ - ▁GO
177
+ - ▁RWANDA
178
+ - ▁HELP
179
+ - ▁HAVE
180
+ - OR
181
+ - ▁G
182
+ - ED
183
+ - ▁OCEAN
184
+ - ▁LIGHTNING
185
+ - ▁ANIMALS
186
+ - ▁EGGS
187
+ - ▁MAKE
188
+ - ▁BE
189
+ - ▁CATERPILLAR
190
+ - ▁FRO
191
+ - ▁C
192
+ - ▁BUTTERFL
193
+ - ▁KEEP
194
+ - ▁FOR
195
+ - OW
196
+ - ▁AROUND
197
+ - ▁RECYCLE
198
+ - ▁W
199
+ - ▁LAY
200
+ - ▁OUT
201
+ - ▁OTHER
202
+ - ▁GET
203
+ - R
204
+ - ▁NEW
205
+ - ▁WINDS
206
+ - ▁GROW
207
+ - ▁PLANTS
208
+ - ▁YOU
209
+ - W
210
+ - ES
211
+ - ▁TOO
212
+ - TO
213
+ - ▁F
214
+ - ▁EGG
215
+ - ▁STORMS
216
+ - ▁NO
217
+ - ▁KIND
218
+ - ▁B
219
+ - AN
220
+ - LY
221
+ - ▁PAPER
222
+ - ▁SAY
223
+ - ALL
224
+ - ▁DO
225
+ - ▁SEAL
226
+ - ▁PENGUIN
227
+ - SIDE
228
+ - ND
229
+ - ▁DU
230
+ - THER
231
+ - ▁AN
232
+ - SHELL
233
+ - L
234
+ - ▁THINGS
235
+ - ▁TURN
236
+ - TER
237
+ - ''''
238
+ - LE
239
+ - TIME
240
+ - ▁UP
241
+ - ▁HE
242
+ - ▁FOREST
243
+ - ▁BLU
244
+ - ▁WHE
245
+ - ▁BRING
246
+ - ▁STA
247
+ - ▁WORK
248
+ - ▁LONG
249
+ - ▁COLOR
250
+ - ▁LAND
251
+ - ▁FISH
252
+ - ▁HOMES
253
+ - MP
254
+ - ▁BOTTLES
255
+ - ▁AWAY
256
+ - ▁TREES
257
+ - ONE
258
+ - ▁AFRICA
259
+ - ▁WITH
260
+ - ▁BIKES
261
+ - ▁HARD
262
+ - ▁COVER
263
+ - ▁EARTH
264
+ - LED
265
+ - G
266
+ - ▁OCEANS
267
+ - ▁H
268
+ - 'ON'
269
+ - ▁CAL
270
+ - ▁FR
271
+ - ▁KIDS
272
+ - ▁USED
273
+ - ▁DOCTOR
274
+ - ▁ANTARCTICA
275
+ - IF
276
+ - ▁CARR
277
+ - CAUSE
278
+ - ▁WARM
279
+ - ▁SWIM
280
+ - X
281
+ - ▁TH
282
+ - DE
283
+ - ▁BUT
284
+ - EW
285
+ - ▁FIND
286
+ - TS
287
+ - ▁RECYCL
288
+ - OM
289
+ - ▁CHILDREN
290
+ - F
291
+ - ▁BI
292
+ - ▁SA
293
+ - ▁LE
294
+ - ROUND
295
+ - ▁PLACE
296
+ - ▁AF
297
+ - ▁PUT
298
+ - ▁JUICE
299
+ - ▁S
300
+ - ▁LOT
301
+ - ▁CAGE
302
+ - IG
303
+ - OWN
304
+ - ▁FLOWERS
305
+ - ▁BACK
306
+ - LES
307
+ - ▁COME
308
+ - OUR
309
+ - ▁SEA
310
+ - ▁RIVER
311
+ - ▁TRA
312
+ - SH
313
+ - FE
314
+ - ▁GAR
315
+ - OOD
316
+ - ▁WAR
317
+ - ▁WA
318
+ - DIED
319
+ - ▁ST
320
+ - TOPS
321
+ - ▁N
322
+ - AGE
323
+ - ▁SIGN
324
+ - ▁THROUGH
325
+ - ▁GIVE
326
+ - ▁BAD
327
+ - ▁RO
328
+ - ▁D
329
+ - ER
330
+ - ▁SOIL
331
+ - ▁SPR
332
+ - ▁SCHOOL
333
+ - ▁PLANE
334
+ - ORMS
335
+ - ▁LEARN
336
+ - ▁BRIGHT
337
+ - ▁DOWN
338
+ - EEP
339
+ - ▁FLY
340
+ - IN
341
+ - ▁O
342
+ - ▁WHA
343
+ - ▁AS
344
+ - ▁I
345
+ - VISI
346
+ - ▁MOUNTAIN
347
+ - ▁GORILLAS
348
+ - LL
349
+ - ▁TIME
350
+ - UCH
351
+ - ▁ICE
352
+ - ▁TELE
353
+ - ▁LIKE
354
+ - ▁WALK
355
+ - ▁WANT
356
+ - ▁COUNTR
357
+ - ▁HOUSES
358
+ - ▁SEE
359
+ - ILL
360
+ - ▁HARM
361
+ - ▁JU
362
+ - ▁TREE
363
+ - ▁ABOUT
364
+ - ASTES
365
+ - ▁OLD
366
+ - ▁BAB
367
+ - EACH
368
+ - ▁COMPUTER
369
+ - ▁NEAR
370
+ - ▁PLAY
371
+ - EN
372
+ - ▁BL
373
+ - ERS
374
+ - ▁NEED
375
+ - OULD
376
+ - ▁CLEAN
377
+ - ▁FL
378
+ - ▁START
379
+ - DOOR
380
+ - ICK
381
+ - ▁TUR
382
+ - OUND
383
+ - ▁MOVE
384
+ - ▁NECTAR
385
+ - ▁FARMERS
386
+ - ▁CALIFORNIA
387
+ - A
388
+ - TLES
389
+ - ▁DI
390
+ - LET
391
+ - U
392
+ - ▁FIELD
393
+ - ▁SHEATH
394
+ - P
395
+ - ▁COLD
396
+ - ▁FORESTS
397
+ - ▁THING
398
+ - THING
399
+ - ▁SPE
400
+ - SE
401
+ - ▁COUNTRIES
402
+ - ▁DID
403
+ - ▁AT
404
+ - GO
405
+ - ▁SP
406
+ - ▁HUR
407
+ - ▁PLA
408
+ - ▁MARKET
409
+ - CI
410
+ - ▁PARTS
411
+ - FORE
412
+ - ▁ANOTHER
413
+ - ▁AR
414
+ - ▁THOU
415
+ - ELEPH
416
+ - ▁COM
417
+ - ▁DINOSAURS
418
+ - ▁SMALL
419
+ - ▁HEART
420
+ - GHT
421
+ - IES
422
+ - ▁BREAK
423
+ - ▁INSECTS
424
+ - ▁WATCH
425
+ - ▁SICK
426
+ - ID
427
+ - ▁FI
428
+ - PEN
429
+ - B
430
+ - VERY
431
+ - ▁BUILDING
432
+ - ▁SEEDS
433
+ - ▁BACKS
434
+ - HANG
435
+ - O
436
+ - OTHER
437
+ - ▁GA
438
+ - ▁STORM
439
+ - ▁FAMILIES
440
+ - ▁CLEANER
441
+ - ▁HI
442
+ - ▁CA
443
+ - ▁POL
444
+ - ▁BODIES
445
+ - ▁EXPER
446
+ - LOUD
447
+ - MES
448
+ - ▁EVERY
449
+ - EY
450
+ - RT
451
+ - ▁END
452
+ - ▁AIR
453
+ - PER
454
+ - ▁EX
455
+ - ▁EN
456
+ - ▁YEARS
457
+ - ▁CONTINENT
458
+ - ▁WHITE
459
+ - HOW
460
+ - ▁FROG
461
+ - AV
462
+ - ▁CLO
463
+ - EM
464
+ - ▁THIN
465
+ - ▁DRY
466
+ - ▁FATHER
467
+ - ▁FOU
468
+ - NS
469
+ - ▁PRO
470
+ - BILLS
471
+ - ▁MEDICINES
472
+ - GE
473
+ - I
474
+ - ▁DINOSAUR
475
+ - ▁BUG
476
+ - ▁CO
477
+ - ▁WEA
478
+ - TWO
479
+ - BY
480
+ - TEN
481
+ - ▁WE
482
+ - EST
483
+ - ▁ALSO
484
+ - AL
485
+ - ▁SHOW
486
+ - ▁BELL
487
+ - TREET
488
+ - OES
489
+ - ▁OT
490
+ - ▁KN
491
+ - DLE
492
+ - UMP
493
+ - ▁FILL
494
+ - ANT
495
+ - OUGH
496
+ - STIC
497
+ - ▁FIR
498
+ - ▁BODY
499
+ - ▁FAT
500
+ - OD
501
+ - ▁BONES
502
+ - ▁PU
503
+ - ▁RE
504
+ - ORT
505
+ - ▁PI
506
+ - ▁FE
507
+ - ECT
508
+ - RING
509
+ - HIN
510
+ - ▁UN
511
+ - ARB
512
+ - NK
513
+ - UI
514
+ - IL
515
+ - CE
516
+ - TERS
517
+ - ▁BILLS
518
+ - ▁EVE
519
+ - ▁AL
520
+ - ▁TOP
521
+ - ▁BY
522
+ - ▁SL
523
+ - DOES
524
+ - INE
525
+ - ▁BANANA
526
+ - ▁FRIEND
527
+ - ▁INSTEAD
528
+ - ▁LOOK
529
+ - ▁ROCK
530
+ - 'OFF'
531
+ - TIL
532
+ - ▁PAR
533
+ - ▁DIFFEREN
534
+ - ▁WORLD
535
+ - ▁HATCHE
536
+ - ▁MEDICINE
537
+ - ▁WELL
538
+ - ▁SI
539
+ - ▁PA
540
+ - ▁SO
541
+ - ORE
542
+ - WHE
543
+ - ▁LO
544
+ - ▁YELL
545
+ - ▁SOON
546
+ - LIVE
547
+ - PA
548
+ - MO
549
+ - ▁FLOWER
550
+ - ▁SUR
551
+ - ▁VO
552
+ - ▁PLANT
553
+ - ROP
554
+ - FFE
555
+ - ▁TRY
556
+ - ACK
557
+ - IRT
558
+ - VES
559
+ - ▁EXTINCT
560
+ - ▁TRAVEL
561
+ - TURN
562
+ - BACK
563
+ - ▁P
564
+ - LOW
565
+ - ▁SEVE
566
+ - ▁WIL
567
+ - ▁HATCH
568
+ - ▁ASK
569
+ - ▁WHI
570
+ - ▁BR
571
+ - ADS
572
+ - FT
573
+ - ▁BUIL
574
+ - ▁HIGH
575
+ - PRI
576
+ - ▁BLO
577
+ - PLACE
578
+ - ILLS
579
+ - ▁LOUD
580
+ - ▁LIV
581
+ - GS
582
+ - ▁CHI
583
+ - WAY
584
+ - ▁GR
585
+ - ▁SN
586
+ - ▁SCHOOLS
587
+ - ▁HEAR
588
+ - EL
589
+ - CAN
590
+ - ▁FEET
591
+ - LP
592
+ - TTLE
593
+ - ▁LA
594
+ - ▁WIN
595
+ - URS
596
+ - PERS
597
+ - PPE
598
+ - ▁EXAM
599
+ - ▁TAKE
600
+ - CAL
601
+ - ▁HOME
602
+ - ▁STI
603
+ - TTER
604
+ - ▁RAI
605
+ - '-'
606
+ - ▁SPACE
607
+ - ▁SKIN
608
+ - ▁SHOULD
609
+ - ATCH
610
+ - ▁BLOOD
611
+ - HAR
612
+ - IT
613
+ - NA
614
+ - ▁NOISES
615
+ - AVE
616
+ - ▁WIND
617
+ - ▁MU
618
+ - ▁FEE
619
+ - ▁FLOO
620
+ - IDE
621
+ - ▁COAST
622
+ - ▁SOU
623
+ - OSAURUS
624
+ - ▁R
625
+ - ▁LEA
626
+ - ALK
627
+ - LONG
628
+ - OOK
629
+ - ME
630
+ - SED
631
+ - ▁PO
632
+ - ADE
633
+ - DS
634
+ - ▁SU
635
+ - ▁MUSIC
636
+ - MMER
637
+ - ▁CRACK
638
+ - ▁CRA
639
+ - CLE
640
+ - ▁WHO
641
+ - ▁EARS
642
+ - NE
643
+ - ▁SPI
644
+ - RS
645
+ - CKS
646
+ - CIA
647
+ - ▁HIS
648
+ - AD
649
+ - ▁WORKE
650
+ - DANGER
651
+ - IVE
652
+ - ▁ZOOS
653
+ - OP
654
+ - ▁RUN
655
+ - ▁STATION
656
+ - ▁ROPE
657
+ - RAN
658
+ - MERS
659
+ - TES
660
+ - BBE
661
+ - FISH
662
+ - GG
663
+ - RO
664
+ - ENER
665
+ - ATHER
666
+ - CH
667
+ - VE
668
+ - ▁ASTRONAUTS
669
+ - ▁TA
670
+ - CK
671
+ - ▁L
672
+ - OSS
673
+ - RD
674
+ - ▁SH
675
+ - IS
676
+ - RN
677
+ - NU
678
+ - ▁WH
679
+ - ▁POLE
680
+ - ▁WI
681
+ - IP
682
+ - ▁J
683
+ - K
684
+ - ISE
685
+ - LOCK
686
+ - ▁FO
687
+ - ▁V
688
+ - IKES
689
+ - FIGHT
690
+ - OTH
691
+ - EG
692
+ - ET
693
+ - ▁MUSCLE
694
+ - AY
695
+ - ▁STRONG
696
+ - WS
697
+ - UNCH
698
+ - CA
699
+ - AF
700
+ - AME
701
+ - ▁LI
702
+ - APPE
703
+ - TION
704
+ - ▁NOISE
705
+ - ▁EVER
706
+ - OTS
707
+ - ERA
708
+ - CR
709
+ - IME
710
+ - Q
711
+ - ▁AMERICAN
712
+ - UND
713
+ - ▁HOLES
714
+ - ▁MARKS
715
+ - ▁BONE
716
+ - ICA
717
+ - ▁HAIR
718
+ - BE
719
+ - AT
720
+ - RICAN
721
+ - OUT
722
+ - TRIC
723
+ - ▁E
724
+ - SCHOOL
725
+ - C
726
+ - ▁THIS
727
+ - FLI
728
+ - TH
729
+ - ▁CLA
730
+ - ICE
731
+ - ATER
732
+ - V
733
+ - ▁HU
734
+ - ▁Z
735
+ - ▁FAMILY
736
+ - HICK
737
+ - ROT
738
+ - ▁DAY
739
+ - UNT
740
+ - LD
741
+ - WO
742
+ - KE
743
+ - EX
744
+ - RA
745
+ - ERE
746
+ - ▁HEA
747
+ - UR
748
+ - BRA
749
+ - ▁MON
750
+ - HELL
751
+ - ▁SOMET
752
+ - OLLE
753
+ - ▁RA
754
+ - ▁TODAY
755
+ - ▁SLI
756
+ - ▁SEND
757
+ - ▁MONEY
758
+ - NTS
759
+ - ▁TR
760
+ - ▁HEARTS
761
+ - BEA
762
+ - UM
763
+ - AR
764
+ - BA
765
+ - EAT
766
+ - WORK
767
+ - LOOD
768
+ - ENT
769
+ - THE
770
+ - ▁KILL
771
+ - ORN
772
+ - ▁MEAT
773
+ - ▁MUS
774
+ - ELS
775
+ - ▁OIL
776
+ - ADO
777
+ - AS
778
+ - ANGER
779
+ - OIS
780
+ - ▁VI
781
+ - ▁CAT
782
+ - ▁PE
783
+ - ▁WO
784
+ - PI
785
+ - WER
786
+ - ▁CH
787
+ - HE
788
+ - WELL
789
+ - OT
790
+ - RP
791
+ - OF
792
+ - BONE
793
+ - ECI
794
+ - ▁POTATO
795
+ - ▁PROBLEM
796
+ - FUL
797
+ - OLD
798
+ - AC
799
+ - ULD
800
+ - ▁GRO
801
+ - ▁TRADITIONS
802
+ - IGHT
803
+ - ▁PUMP
804
+ - RIC
805
+ - ANTS
806
+ - TIN
807
+ - LAY
808
+ - ▁PLAN
809
+ - EXT
810
+ - ERED
811
+ - ▁HO
812
+ - ▁SOUND
813
+ - SMA
814
+ - ▁POLES
815
+ - PIC
816
+ - ▁AC
817
+ - ▁CHE
818
+ - ▁CARV
819
+ - ATCHED
820
+ - RC
821
+ - MS
822
+ - LS
823
+ - PIL
824
+ - ▁FINISHED
825
+ - ▁RUSSIA
826
+ - ▁TRIBE
827
+ - ▁CHIPS
828
+ - IC
829
+ - H
830
+ - ▁SHOWER
831
+ - ANGE
832
+ - ▁BOY
833
+ - ESSA
834
+ - TE
835
+ - ▁HEARING
836
+ - ▁GI
837
+ - ▁LUN
838
+ - ▁LEAD
839
+ - TCH
840
+ - ▁MAC
841
+ - MAL
842
+ - PIN
843
+ - TU
844
+ - ▁GRA
845
+ - HIP
846
+ - HA
847
+ - UE
848
+ - UG
849
+ - ▁WOR
850
+ - ROW
851
+ - AW
852
+ - NCE
853
+ - LECT
854
+ - GES
855
+ - ▁TAL
856
+ - CHED
857
+ - REET
858
+ - ▁MI
859
+ - OLL
860
+ - FATHER
861
+ - OMES
862
+ - ▁TRADITION
863
+ - GOT
864
+ - TRA
865
+ - ▁CHEETAH
866
+ - ▁CRUMBS
867
+ - ▁PUNISH
868
+ - ▁DANCES
869
+ - ▁SUPPL
870
+ - ▁HEALTH
871
+ - ▁SCIEN
872
+ - ▁POUCHES
873
+ - ▁FLOAT
874
+ - ▁MONTH
875
+ - ▁CARVING
876
+ - ▁FARM
877
+ - ILE
878
+ - ▁POTATOES
879
+ - ▁BRAIN
880
+ - ▁CUT
881
+ - ▁DRILLS
882
+ - ▁BUILD
883
+ - ▁K
884
+ - ▁DRI
885
+ - TING
886
+ - ▁PICK
887
+ - EED
888
+ - OPS
889
+ - ALE
890
+ - ULT
891
+ - ▁EA
892
+ - OK
893
+ - ADD
894
+ - ▁FIRE
895
+ - ▁AG
896
+ - VEN
897
+ - ▁TEACH
898
+ - ORM
899
+ - NEW
900
+ - REE
901
+ - OWS
902
+ - ▁COL
903
+ - PP
904
+ - EE
905
+ - YS
906
+ - ION
907
+ - ▁OU
908
+ - BLE
909
+ - FEE
910
+ - ▁COU
911
+ - MPE
912
+ - KS
913
+ - HIT
914
+ - FF
915
+ - LAR
916
+ - FACE
917
+ - TLE
918
+ - SES
919
+ - WE
920
+ - URT
921
+ - ▁SE
922
+ - TCHED
923
+ - ▁STOR
924
+ - ERAT
925
+ - TRO
926
+ - ATION
927
+ - ���DIE
928
+ - TORIES
929
+ - OU
930
+ - ▁BU
931
+ - TREE
932
+ - ▁CL
933
+ - PLE
934
+ - LLE
935
+ - MOVE
936
+ - ▁ANIMAL
937
+ - ▁ME
938
+ - ▁PR
939
+ - NG
940
+ - STRONG
941
+ - ▁ASTRONAUT
942
+ - ▁DRILL
943
+ - J
944
+ - EARING
945
+ - OSE
946
+ - ASK
947
+ - ARM
948
+ - ▁BOTT
949
+ - ▁BA
950
+ - ▁DIFF
951
+ - ▁EAR
952
+ - ▁POUCH
953
+ - MOUNT
954
+ - ▁LIGHT
955
+ - UN
956
+ - ▁NEC
957
+ - CHI
958
+ - GRA
959
+ - ALS
960
+ - ITION
961
+ - ▁POW
962
+ - ARK
963
+ - ▁US
964
+ - OVER
965
+ - EVE
966
+ - NGU
967
+ - ▁AB
968
+ - ▁THO
969
+ - MENT
970
+ - ▁DR
971
+ - RY
972
+ - ▁CAR
973
+ - PE
974
+ - TI
975
+ - BER
976
+ - SSI
977
+ - RM
978
+ - KES
979
+ - ▁ALON
980
+ - TIONS
981
+ - ▁WON
982
+ - ▁FLO
983
+ - PAR
984
+ - CHIN
985
+ - DD
986
+ - OL
987
+ - ▁PART
988
+ - LLS
989
+ - END
990
+ - ▁BET
991
+ - AK
992
+ - UTE
993
+ - EP
994
+ - OUS
995
+ - GET
996
+ - CT
997
+ - RMS
998
+ - DAY
999
+ - RENT
1000
+ - MB
1001
+ - SON
1002
+ - LAND
1003
+ - SI
1004
+ - OLS
1005
+ - AIN
1006
+ - ▁NE
1007
+ - ARING
1008
+ - ▁DE
1009
+ - MA
1010
+ - WH
1011
+ - DI
1012
+ - TOR
1013
+ - HAT
1014
+ - ▁CARRIE
1015
+ - USE
1016
+ - OCK
1017
+ - RKING
1018
+ - EARS
1019
+ - ATE
1020
+ - AP
1021
+ - RB
1022
+ - KING
1023
+ - WAR
1024
+ - NGS
1025
+ - EAD
1026
+ - RU
1027
+ - PEOPLE
1028
+ - ACES
1029
+ - EAR
1030
+ - ▁FA
1031
+ - Z
1032
+ - <sos/eos>
1033
+ init: null
1034
+ input_size: null
1035
+ ctc_conf:
1036
+ dropout_rate: 0.0
1037
+ ctc_type: builtin
1038
+ reduce: true
1039
+ ignore_nan_grad: null
1040
+ zero_infinity: true
1041
+ brctc_risk_strategy: exp
1042
+ brctc_group_strategy: end
1043
+ brctc_risk_factor: 0.0
1044
+ joint_net_conf: null
1045
+ use_preprocessor: true
1046
+ use_lang_prompt: false
1047
+ use_nlp_prompt: false
1048
+ token_type: bpe
1049
+ bpemodel: data/en_token_list/bpe_unigram900/bpe.model
1050
+ non_linguistic_symbols: null
1051
+ cleaner: null
1052
+ g2p: null
1053
+ speech_volume_normalize: null
1054
+ rir_scp: null
1055
+ rir_apply_prob: 1.0
1056
+ noise_scp: null
1057
+ noise_apply_prob: 1.0
1058
+ noise_db_range: '13_15'
1059
+ short_noise_thres: 0.5
1060
+ aux_ctc_tasks: []
1061
+ frontend: s3prl
1062
+ frontend_conf:
1063
+ frontend_conf:
1064
+ upstream: wavlm_large
1065
+ download_dir: ./hub
1066
+ multilayer_feature: true
1067
+ fs: 16k
1068
+ specaug: specaug
1069
+ specaug_conf:
1070
+ apply_time_warp: true
1071
+ time_warp_window: 5
1072
+ time_warp_mode: bicubic
1073
+ apply_freq_mask: true
1074
+ freq_mask_width_range:
1075
+ - 0
1076
+ - 27
1077
+ num_freq_mask: 2
1078
+ apply_time_mask: true
1079
+ time_mask_width_ratio_range:
1080
+ - 0.0
1081
+ - 0.05
1082
+ num_time_mask: 5
1083
+ normalize: utterance_mvn
1084
+ normalize_conf: {}
1085
+ model: espnet
1086
+ model_conf:
1087
+ ctc_weight: 0.3
1088
+ lsm_weight: 0.1
1089
+ length_normalized_loss: false
1090
+ extract_feats_in_collect_stats: false
1091
+ preencoder: linear
1092
+ preencoder_conf:
1093
+ input_size: 1024
1094
+ output_size: 80
1095
+ encoder: transformer
1096
+ encoder_conf:
1097
+ output_size: 256
1098
+ attention_heads: 4
1099
+ linear_units: 1024
1100
+ num_blocks: 18
1101
+ dropout_rate: 0.1
1102
+ positional_dropout_rate: 0.1
1103
+ attention_dropout_rate: 0.1
1104
+ input_layer: conv2d2
1105
+ normalize_before: true
1106
+ postencoder: null
1107
+ postencoder_conf: {}
1108
+ decoder: transformer
1109
+ decoder_conf:
1110
+ attention_heads: 4
1111
+ linear_units: 2048
1112
+ num_blocks: 6
1113
+ dropout_rate: 0.1
1114
+ positional_dropout_rate: 0.1
1115
+ self_attention_dropout_rate: 0.1
1116
+ src_attention_dropout_rate: 0.1
1117
+ preprocessor: default
1118
+ preprocessor_conf: {}
1119
+ required:
1120
+ - output_dir
1121
+ - token_list
1122
+ version: '202412'
1123
+ distributed: true
exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/acc.png ADDED
exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/backward_time.png ADDED
exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/cer.png ADDED
exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/cer_ctc.png ADDED
exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/clip.png ADDED
exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/forward_time.png ADDED
exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/gpu_max_cached_mem_GB.png ADDED
exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/grad_norm.png ADDED
exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/iter_time.png ADDED
exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/loss.png ADDED
exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/loss_att.png ADDED
exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/loss_ctc.png ADDED
exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/loss_scale.png ADDED
exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/optim0_lr0.png ADDED
exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/optim_step_time.png ADDED
exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/train_time.png ADDED
exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/images/wer.png ADDED
exp/asr_train_asr_wavlm_transformer_raw_en_bpe900_sp/valid.acc.ave_4best.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:383a26a00928e3363f6d360e8f57d13379a5d1e7e7b18a5906fe073e08988980
3
+ size 1372132286