Jiatong Shi commited on
Commit
25e729d
·
1 Parent(s): a1b5f52

Update model

Browse files
Files changed (31) hide show
  1. README.md +797 -3
  2. data/token_list/bpe_unigram500/bpe.model +3 -0
  3. exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp/30epoch.pth +3 -0
  4. exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp/RESULTS.md +32 -0
  5. exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp/config.yaml +697 -0
  6. exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp/images/acc.png +0 -0
  7. exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp/images/backward_time.png +0 -0
  8. exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp/images/cer.png +0 -0
  9. exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp/images/cer_ctc.png +0 -0
  10. exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp/images/forward_time.png +0 -0
  11. exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp/images/gpu_max_cached_mem_GB.png +0 -0
  12. exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp/images/iter_time.png +0 -0
  13. exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp/images/loss.png +0 -0
  14. exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp/images/loss_att.png +0 -0
  15. exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp/images/loss_ctc.png +0 -0
  16. exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp/images/optim0_lr0.png +0 -0
  17. exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp/images/optim_step_time.png +0 -0
  18. exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp/images/train_time.png +0 -0
  19. exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp/images/wer.png +0 -0
  20. exp/lm_train_lm_bpe500/19epoch.pth +3 -0
  21. exp/lm_train_lm_bpe500/config.yaml +617 -0
  22. exp/lm_train_lm_bpe500/images/backward_time.png +0 -0
  23. exp/lm_train_lm_bpe500/images/forward_time.png +0 -0
  24. exp/lm_train_lm_bpe500/images/gpu_max_cached_mem_GB.png +0 -0
  25. exp/lm_train_lm_bpe500/images/iter_time.png +0 -0
  26. exp/lm_train_lm_bpe500/images/loss.png +0 -0
  27. exp/lm_train_lm_bpe500/images/optim0_lr0.png +0 -0
  28. exp/lm_train_lm_bpe500/images/optim_step_time.png +0 -0
  29. exp/lm_train_lm_bpe500/images/train_time.png +0 -0
  30. exp/lm_train_lm_bpe500/perplexity_test/ppl +1 -0
  31. meta.yaml +10 -0
README.md CHANGED
@@ -1,3 +1,797 @@
1
- ---
2
- license: apache-2.0
3
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ tags:
3
+ - espnet
4
+ - audio
5
+ - automatic-speech-recognition
6
+ language: noinfo
7
+ datasets:
8
+ - dsing
9
+ license: cc-by-4.0
10
+ ---
11
+
12
+ ## ESPnet2 ASR model
13
+
14
+ ### `espnet/ftshijt_espnet2_asr_dsing_hubert_conformer`
15
+
16
+ This model was trained by jiatong using dsing recipe in [espnet](https://github.com/espnet/espnet/).
17
+
18
+ ### Demo: How to use in ESPnet2
19
+
20
+ ```bash
21
+ cd espnet
22
+
23
+ pip install -e .
24
+ cd egs2/dsing/asr1
25
+ ./run.sh --skip_data_prep false --skip_train true --download_model espnet/ftshijt_espnet2_asr_dsing_hubert_conformer
26
+ ```
27
+
28
+ <!-- Generated by scripts/utils/show_asr_result.sh -->
29
+ # RESULTS
30
+ ## Environments
31
+ - date: `Sat Mar 19 23:02:37 EDT 2022`
32
+ - python version: `3.9.7 (default, Sep 16 2021, 13:09:58) [GCC 7.5.0]`
33
+ - espnet version: `espnet 0.10.7a1`
34
+ - pytorch version: `pytorch 1.10.1`
35
+ - Git hash: `c1ed71c6899e54c0b3dad82687886b1183cd0885`
36
+ - Commit date: `Wed Mar 16 23:34:49 2022 -0400`
37
+
38
+ ## asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp
39
+ ### WER
40
+
41
+ |dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err|
42
+ |---|---|---|---|---|---|---|---|---|
43
+ |decode_asr_lm_lm_train_lm_bpe500_valid.loss.ave_asr_model_latest/dev|482|4018|83.6|9.4|7.0|6.4|22.8|58.3|
44
+ |decode_asr_lm_lm_train_lm_bpe500_valid.loss.ave_asr_model_latest/test|480|4632|81.4|12.3|6.3|4.5|23.1|52.1|
45
+
46
+ ### CER
47
+
48
+ |dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err|
49
+ |---|---|---|---|---|---|---|---|---|
50
+ |decode_asr_lm_lm_train_lm_bpe500_valid.loss.ave_asr_model_latest/dev|482|18692|88.5|3.1|8.4|5.9|17.4|58.3|
51
+ |decode_asr_lm_lm_train_lm_bpe500_valid.loss.ave_asr_model_latest/test|480|21787|87.9|4.3|7.8|4.5|16.6|52.1|
52
+
53
+ ### TER
54
+
55
+ |dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err|
56
+ |---|---|---|---|---|---|---|---|---|
57
+ |decode_asr_lm_lm_train_lm_bpe500_valid.loss.ave_asr_model_latest/dev|482|6097|82.2|7.1|10.7|5.7|23.5|58.3|
58
+ |decode_asr_lm_lm_train_lm_bpe500_valid.loss.ave_asr_model_latest/test|480|7736|81.7|9.2|9.1|4.0|22.3|52.1|
59
+
60
+ ## ASR config
61
+
62
+ <details><summary>expand</summary>
63
+
64
+ ```
65
+ config: conf/tuning/train_asr_conformer7_hubert_ll60k_large.yaml
66
+ print_config: false
67
+ log_level: INFO
68
+ dry_run: false
69
+ iterator_type: sequence
70
+ output_dir: exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp
71
+ ngpu: 1
72
+ seed: 0
73
+ num_workers: 1
74
+ num_att_plot: 3
75
+ dist_backend: nccl
76
+ dist_init_method: env://
77
+ dist_world_size: null
78
+ dist_rank: null
79
+ local_rank: 0
80
+ dist_master_addr: null
81
+ dist_master_port: null
82
+ dist_launcher: null
83
+ multiprocessing_distributed: false
84
+ unused_parameters: true
85
+ sharded_ddp: false
86
+ cudnn_enabled: true
87
+ cudnn_benchmark: false
88
+ cudnn_deterministic: true
89
+ collect_stats: false
90
+ write_collected_feats: false
91
+ max_epoch: 35
92
+ patience: null
93
+ val_scheduler_criterion:
94
+ - valid
95
+ - loss
96
+ early_stopping_criterion:
97
+ - valid
98
+ - loss
99
+ - min
100
+ best_model_criterion:
101
+ - - valid
102
+ - acc
103
+ - max
104
+ keep_nbest_models: 10
105
+ nbest_averaging_interval: 0
106
+ grad_clip: 5.0
107
+ grad_clip_type: 2.0
108
+ grad_noise: false
109
+ accum_grad: 8
110
+ no_forward_run: false
111
+ resume: true
112
+ train_dtype: float32
113
+ use_amp: false
114
+ log_interval: null
115
+ use_matplotlib: true
116
+ use_tensorboard: true
117
+ use_wandb: false
118
+ wandb_project: null
119
+ wandb_id: null
120
+ wandb_entity: null
121
+ wandb_name: null
122
+ wandb_model_log_interval: -1
123
+ detect_anomaly: false
124
+ pretrain_path: null
125
+ init_param: []
126
+ ignore_init_mismatch: false
127
+ freeze_param:
128
+ - frontend.upstream
129
+ num_iters_per_epoch: null
130
+ batch_size: 20
131
+ valid_batch_size: null
132
+ batch_bins: 1000000
133
+ valid_batch_bins: null
134
+ train_shape_file:
135
+ - exp/asr_stats_raw_bpe500_sp/train/speech_shape
136
+ - exp/asr_stats_raw_bpe500_sp/train/text_shape.bpe
137
+ valid_shape_file:
138
+ - exp/asr_stats_raw_bpe500_sp/valid/speech_shape
139
+ - exp/asr_stats_raw_bpe500_sp/valid/text_shape.bpe
140
+ batch_type: numel
141
+ valid_batch_type: null
142
+ fold_length:
143
+ - 80000
144
+ - 150
145
+ sort_in_batch: descending
146
+ sort_batch: descending
147
+ multiple_iterator: false
148
+ chunk_length: 500
149
+ chunk_shift_ratio: 0.5
150
+ num_cache_chunks: 1024
151
+ train_data_path_and_name_and_type:
152
+ - - dump/raw/train30_sp/wav.scp
153
+ - speech
154
+ - kaldi_ark
155
+ - - dump/raw/train30_sp/text
156
+ - text
157
+ - text
158
+ valid_data_path_and_name_and_type:
159
+ - - dump/raw/dev/wav.scp
160
+ - speech
161
+ - kaldi_ark
162
+ - - dump/raw/dev/text
163
+ - text
164
+ - text
165
+ allow_variable_data_keys: false
166
+ max_cache_size: 0.0
167
+ max_cache_fd: 32
168
+ valid_max_cache_size: null
169
+ optim: adam
170
+ optim_conf:
171
+ lr: 0.0025
172
+ scheduler: warmuplr
173
+ scheduler_conf:
174
+ warmup_steps: 40000
175
+ token_list:
176
+ - <blank>
177
+ - <unk>
178
+ - ▁I
179
+ - ''''
180
+ - ▁YOU
181
+ - S
182
+ - T
183
+ - ▁THE
184
+ - M
185
+ - ▁ME
186
+ - ▁A
187
+ - ▁AND
188
+ - ▁TO
189
+ - E
190
+ - A
191
+ - ING
192
+ - D
193
+ - ▁MY
194
+ - ▁
195
+ - O
196
+ - ▁IT
197
+ - I
198
+ - N
199
+ - RE
200
+ - Y
201
+ - ▁BE
202
+ - ▁IN
203
+ - ▁ON
204
+ - ▁LOVE
205
+ - U
206
+ - ▁WE
207
+ - LL
208
+ - H
209
+ - ▁YOUR
210
+ - ▁S
211
+ - IN
212
+ - ▁OF
213
+ - ▁DO
214
+ - ▁THAT
215
+ - ▁ALL
216
+ - L
217
+ - ▁DON
218
+ - ▁OH
219
+ - ▁LIKE
220
+ - ▁KNOW
221
+ - ▁FOR
222
+ - ▁CAN
223
+ - ▁JUST
224
+ - P
225
+ - ▁BUT
226
+ - ED
227
+ - K
228
+ - ▁WHEN
229
+ - ▁SO
230
+ - R
231
+ - ▁GO
232
+ - ▁WHAT
233
+ - ▁C
234
+ - ▁WITH
235
+ - W
236
+ - ▁F
237
+ - C
238
+ - ▁NO
239
+ - ER
240
+ - ▁ONE
241
+ - ▁LET
242
+ - VE
243
+ - ES
244
+ - ▁NOW
245
+ - ▁BABY
246
+ - G
247
+ - ▁GOT
248
+ - ▁COME
249
+ - CAUSE
250
+ - LE
251
+ - B
252
+ - ▁B
253
+ - AR
254
+ - ▁UP
255
+ - ▁'
256
+ - ▁W
257
+ - ��SEE
258
+ - ▁TIME
259
+ - ▁ARE
260
+ - ▁G
261
+ - ▁LOOK
262
+ - ▁THIS
263
+ - F
264
+ - ▁IS
265
+ - ▁NEVER
266
+ - ▁M
267
+ - ▁P
268
+ - AN
269
+ - ▁WAS
270
+ - ▁WAY
271
+ - ▁IF
272
+ - OR
273
+ - ▁SAY
274
+ - V
275
+ - ▁R
276
+ - ▁T
277
+ - ▁DOWN
278
+ - RA
279
+ - ▁THERE
280
+ - ▁HEART
281
+ - ▁NOT
282
+ - RO
283
+ - ▁WILL
284
+ - ▁OUT
285
+ - CE
286
+ - ▁WANT
287
+ - ▁YEAH
288
+ - ▁HAVE
289
+ - ▁GIVE
290
+ - ▁TOO
291
+ - ▁GONNA
292
+ - ▁HOW
293
+ - ▁NEED
294
+ - ▁GET
295
+ - ▁TAKE
296
+ - ▁EVERY
297
+ - ▁FEEL
298
+ - ▁HE
299
+ - EN
300
+ - ▁FROM
301
+ - ▁HA
302
+ - ▁K
303
+ - ▁SHE
304
+ - 'ON'
305
+ - ▁DI
306
+ - RI
307
+ - ▁ONLY
308
+ - NE
309
+ - ▁WHO
310
+ - ▁AWAY
311
+ - ▁E
312
+ - ▁D
313
+ - ▁LIFE
314
+ - ▁MAKE
315
+ - IC
316
+ - ▁BACK
317
+ - ▁WHERE
318
+ - ▁MADE
319
+ - ▁DAY
320
+ - ▁HERE
321
+ - ▁LO
322
+ - ▁HER
323
+ - ▁AS
324
+ - ▁GOOD
325
+ - ▁WANNA
326
+ - ▁OOH
327
+ - ▁TELL
328
+ - LY
329
+ - TH
330
+ - ▁WON
331
+ - ▁LIGHT
332
+ - ▁KEEP
333
+ - ▁MA
334
+ - ▁LA
335
+ - ▁SH
336
+ - ▁WORLD
337
+ - ▁MORE
338
+ - ▁LI
339
+ - AL
340
+ - ▁COULD
341
+ - ▁GIRL
342
+ - ▁NOTHING
343
+ - ▁EVER
344
+ - ▁THINK
345
+ - IE
346
+ - ▁BY
347
+ - ▁AT
348
+ - ▁TONIGHT
349
+ - ▁THEY
350
+ - ▁CALL
351
+ - ▁HO
352
+ - ▁WOULD
353
+ - IL
354
+ - ▁OUR
355
+ - ▁FALL
356
+ - ▁NIGHT
357
+ - ▁THAN
358
+ - ▁DE
359
+ - ▁SOME
360
+ - ▁WAIT
361
+ - ▁RIGHT
362
+ - ▁RE
363
+ - ▁HALLELUJAH
364
+ - ▁TH
365
+ - NG
366
+ - ▁CO
367
+ - ▁WERE
368
+ - ▁TALK
369
+ - ET
370
+ - ▁BO
371
+ - ▁HOLD
372
+ - UR
373
+ - ▁BEEN
374
+ - ▁US
375
+ - ▁PA
376
+ - VER
377
+ - ▁EYES
378
+ - ▁DREAM
379
+ - ▁SONG
380
+ - ▁SHOULD
381
+ - ▁STILL
382
+ - ▁OVER
383
+ - TA
384
+ - ▁ANYMORE
385
+ - IGHT
386
+ - ▁STAY
387
+ - ▁BETTER
388
+ - LESS
389
+ - ▁THROUGH
390
+ - ▁LITTLE
391
+ - X
392
+ - ▁GONE
393
+ - ▁AIN
394
+ - ▁DA
395
+ - ▁HOLDING
396
+ - ▁HURT
397
+ - ▁TRY
398
+ - ▁FIND
399
+ - Z
400
+ - DE
401
+ - ▁LAST
402
+ - ▁SAID
403
+ - ▁ALWAYS
404
+ - ▁BODY
405
+ - ▁MIND
406
+ - ▁CRY
407
+ - ▁EVEN
408
+ - ▁RUN
409
+ - ▁HOPE
410
+ - ▁WITHOUT
411
+ - ▁MISS
412
+ - ▁ABOUT
413
+ - ▁HAND
414
+ - ▁J
415
+ - ▁AGAIN
416
+ - ▁THOUGH
417
+ - ▁NAH
418
+ - ▁LIVE
419
+ - ▁BA
420
+ - ▁OLD
421
+ - ▁HEAD
422
+ - ▁FIRE
423
+ - ▁MAN
424
+ - ▁SOMETHING
425
+ - ▁WHY
426
+ - THER
427
+ - ▁HOME
428
+ - ▁OR
429
+ - ▁INSIDE
430
+ - ▁NEW
431
+ - ▁HEY
432
+ - TION
433
+ - ▁EVERYTHING
434
+ - ▁HAD
435
+ - ▁SOMETIMES
436
+ - ▁HARD
437
+ - ▁TOUCH
438
+ - ▁HEAR
439
+ - ▁AM
440
+ - ▁MUCH
441
+ - ▁LONG
442
+ - ▁STAR
443
+ - GETTING
444
+ - ▁WALK
445
+ - ▁PEOPLE
446
+ - ▁BEFORE
447
+ - ▁CLOSE
448
+ - ▁TWO
449
+ - ▁FAR
450
+ - ▁SHOW
451
+ - ▁STAND
452
+ - ▁LOSE
453
+ - ▁HELP
454
+ - ▁NAME
455
+ - ▁BOY
456
+ - ▁TRUE
457
+ - ▁PLAY
458
+ - ▁DARK
459
+ - ▁THINGS
460
+ - ▁NA
461
+ - ▁TEAR
462
+ - ▁END
463
+ - ▁NOBODY
464
+ - ▁SEA
465
+ - ▁ROCKABYE
466
+ - ▁BELIEVE
467
+ - ▁BROKE
468
+ - ▁AROUND
469
+ - ▁START
470
+ - ▁KISS
471
+ - ▁FEELING
472
+ - ▁BREAK
473
+ - ▁SOMEONE
474
+ - ▁FRIEND
475
+ - ▁ALONE
476
+ - ▁BEAUTIFUL
477
+ - ▁CRAZY
478
+ - ▁OWN
479
+ - OSE
480
+ - ▁STOP
481
+ - ▁LOST
482
+ - ▁HIM
483
+ - ▁BAD
484
+ - ▁CHANCE
485
+ - ▁REALLY
486
+ - ▁WISH
487
+ - ▁MOVE
488
+ - ▁SKY
489
+ - ▁PLACE
490
+ - AKE
491
+ - ▁LEAVE
492
+ - ▁YA
493
+ - ▁STRONG
494
+ - ▁PUT
495
+ - ▁OPEN
496
+ - ▁WRONG
497
+ - ▁COLD
498
+ - OCK
499
+ - ▁USED
500
+ - ▁FOUND
501
+ - ▁LONELY
502
+ - ▁DANCE
503
+ - EACH
504
+ - ▁ANOTHER
505
+ - ▁SIDE
506
+ - ▁UNDER
507
+ - ▁MATTER
508
+ - ▁THESE
509
+ - ▁CARE
510
+ - ▁MINE
511
+ - ▁SHINE
512
+ - ▁AFRAID
513
+ - ▁TURN
514
+ - ▁PLEASE
515
+ - ▁SUN
516
+ - ▁DIAMOND
517
+ - ▁UNTIL
518
+ - ▁FACE
519
+ - ▁LEARN
520
+ - ▁TRUST
521
+ - ▁WONDER
522
+ - ▁BREATH
523
+ - ATE
524
+ - ▁SORRY
525
+ - ▁HU
526
+ - ▁WATCH
527
+ - ▁LATE
528
+ - ROUND
529
+ - ▁ARMS
530
+ - ▁PERFECT
531
+ - ▁MAYBE
532
+ - ▁PULL
533
+ - ▁REMEMBER
534
+ - ▁FIGHT
535
+ - ▁MYSELF
536
+ - ▁INTO
537
+ - ▁DARLING
538
+ - ▁THUNDER
539
+ - ▁FOLLOW
540
+ - ▁REASON
541
+ - ▁BURN
542
+ - ▁HIS
543
+ - ▁MUST
544
+ - ▁FREE
545
+ - ▁FLASHLIGHT
546
+ - ▁1
547
+ - ▁ENOUGH
548
+ - ▁DRINK
549
+ - ▁WORDS
550
+ - ▁HIDE
551
+ - ▁UN
552
+ - ▁FORGET
553
+ - ▁SURE
554
+ - ▁CHANGE
555
+ - ▁SMILE
556
+ - ▁PROMISE
557
+ - ▁FOREVER
558
+ - '2'
559
+ - ▁SWEET
560
+ - ▁SAME
561
+ - ▁OOOH
562
+ - ▁PART
563
+ - ▁SOMEBODY
564
+ - NESS
565
+ - ▁BRIGHT
566
+ - ▁HEAVEN
567
+ - ▁DEEP
568
+ - ▁HIGH
569
+ - ▁INSTEAD
570
+ - ▁MOMENT
571
+ - ▁ALONG
572
+ - ▁ALRIGHT
573
+ - ▁SLOW
574
+ - ▁TOMORROW
575
+ - ▁SOUL
576
+ - ▁QU
577
+ - ▁PUSH
578
+ - ▁CHANDELIER
579
+ - ▁LEFT
580
+ - SIDE
581
+ - ▁TOLD
582
+ - ▁KNEW
583
+ - READY
584
+ - ▁LOVING
585
+ - ▁SAW
586
+ - '3'
587
+ - ▁WORK
588
+ - ▁DANCING
589
+ - ▁THREE
590
+ - ▁SAVE
591
+ - ▁SHOOT
592
+ - ▁LEAD
593
+ - ▁SKI
594
+ - ▁WILD
595
+ - ▁WIND
596
+ - ▁WHILE
597
+ - ▁EDGE
598
+ - ▁HAPPY
599
+ - ▁FEAR
600
+ - STUCK
601
+ - ▁MOST
602
+ - ▁LISTEN
603
+ - ▁WOAH
604
+ - ▁FIRST
605
+ - ▁JOLENE
606
+ - ▁VOICE
607
+ - ▁COMP
608
+ - ▁MILLION
609
+ - FUL
610
+ - ▁OOOOOH
611
+ - ▁CAME
612
+ - ▁RISE
613
+ - ▁NEXT
614
+ - ▁COUNT
615
+ - ▁MOUNTAIN
616
+ - ▁ROOM
617
+ - ▁BLUE
618
+ - ▁HIT
619
+ - ▁RAISE
620
+ - J
621
+ - ▁THOUSAND
622
+ - ▁SHAP
623
+ - ▁TREAT
624
+ - ▁DRY
625
+ - ▁FINALLY
626
+ - ▁TITANIUM
627
+ - ▁CARRY
628
+ - ▁TRUTH
629
+ - ▁WATER
630
+ - ▁MORNING
631
+ - TIME
632
+ - ▁BELONG
633
+ - ▁UMA
634
+ - ▁ALIVE
635
+ - ▁ELSE
636
+ - ▁ANGEL
637
+ - ▁BRAND
638
+ - ▁APART
639
+ - ▁EVERYBODY
640
+ - ▁SOUND
641
+ - ▁GUESS
642
+ - ▁PRAY
643
+ - ▁FAITH
644
+ - ▁AFTER
645
+ - ▁THROW
646
+ - ▁TRIED
647
+ - ▁SLEEP
648
+ - ▁FOOL
649
+ - ▁DISCOVERING
650
+ - ▁FUCK
651
+ - ▁TASTE
652
+ - ▁UNDERSTAND
653
+ - ▁SHAME
654
+ - ▁POWER
655
+ - ▁WELCOME
656
+ - ▁FELT
657
+ - ▁SAFE
658
+ - ▁DESERVE
659
+ - ▁GAME
660
+ - ▁SUPERMA
661
+ - ▁SWEAR
662
+ - ▁BETWEEN
663
+ - ▁GLASS
664
+ - ▁CATCH
665
+ - ▁TOGETHER
666
+ - '0'
667
+ - '4'
668
+ - '6'
669
+ - '5'
670
+ - '1'
671
+ - '8'
672
+ - '7'
673
+ - '9'
674
+ - Q
675
+ - <sos/eos>
676
+ init: null
677
+ input_size: null
678
+ ctc_conf:
679
+ dropout_rate: 0.0
680
+ ctc_type: builtin
681
+ reduce: true
682
+ ignore_nan_grad: true
683
+ joint_net_conf: null
684
+ model_conf:
685
+ ctc_weight: 0.3
686
+ lsm_weight: 0.1
687
+ length_normalized_loss: false
688
+ extract_feats_in_collect_stats: false
689
+ use_preprocessor: true
690
+ token_type: bpe
691
+ bpemodel: data/token_list/bpe_unigram500/bpe.model
692
+ non_linguistic_symbols: null
693
+ cleaner: null
694
+ g2p: null
695
+ speech_volume_normalize: null
696
+ rir_scp: null
697
+ rir_apply_prob: 1.0
698
+ noise_scp: null
699
+ noise_apply_prob: 1.0
700
+ noise_db_range: '13_15'
701
+ frontend: s3prl
702
+ frontend_conf:
703
+ frontend_conf:
704
+ upstream: hubert_large_ll60k
705
+ download_dir: ./hub
706
+ multilayer_feature: true
707
+ fs: 16k
708
+ specaug: specaug
709
+ specaug_conf:
710
+ apply_time_warp: true
711
+ time_warp_window: 5
712
+ time_warp_mode: bicubic
713
+ apply_freq_mask: true
714
+ freq_mask_width_range:
715
+ - 0
716
+ - 30
717
+ num_freq_mask: 2
718
+ apply_time_mask: true
719
+ time_mask_width_range:
720
+ - 0
721
+ - 40
722
+ num_time_mask: 2
723
+ normalize: utterance_mvn
724
+ normalize_conf: {}
725
+ preencoder: linear
726
+ preencoder_conf:
727
+ input_size: 1024
728
+ output_size: 80
729
+ encoder: conformer
730
+ encoder_conf:
731
+ output_size: 512
732
+ attention_heads: 8
733
+ linear_units: 2048
734
+ num_blocks: 12
735
+ dropout_rate: 0.1
736
+ positional_dropout_rate: 0.1
737
+ attention_dropout_rate: 0.1
738
+ input_layer: conv2d2
739
+ normalize_before: true
740
+ macaron_style: true
741
+ pos_enc_layer_type: rel_pos
742
+ selfattention_layer_type: rel_selfattn
743
+ activation_type: swish
744
+ use_cnn_module: true
745
+ cnn_module_kernel: 31
746
+ postencoder: null
747
+ postencoder_conf: {}
748
+ decoder: transformer
749
+ decoder_conf:
750
+ attention_heads: 8
751
+ linear_units: 2048
752
+ num_blocks: 6
753
+ dropout_rate: 0.1
754
+ positional_dropout_rate: 0.1
755
+ self_attention_dropout_rate: 0.1
756
+ src_attention_dropout_rate: 0.1
757
+ required:
758
+ - output_dir
759
+ - token_list
760
+ version: 0.10.7a1
761
+ distributed: false
762
+ ```
763
+
764
+ </details>
765
+
766
+
767
+
768
+ ### Citing ESPnet
769
+
770
+ ```BibTex
771
+ @inproceedings{watanabe2018espnet,
772
+ 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},
773
+ title={{ESPnet}: End-to-End Speech Processing Toolkit},
774
+ year={2018},
775
+ booktitle={Proceedings of Interspeech},
776
+ pages={2207--2211},
777
+ doi={10.21437/Interspeech.2018-1456},
778
+ url={http://dx.doi.org/10.21437/Interspeech.2018-1456}
779
+ }
780
+
781
+
782
+
783
+
784
+ ```
785
+
786
+ or arXiv:
787
+
788
+ ```bibtex
789
+ @misc{watanabe2018espnet,
790
+ title={ESPnet: End-to-End Speech Processing Toolkit},
791
+ 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},
792
+ year={2018},
793
+ eprint={1804.00015},
794
+ archivePrefix={arXiv},
795
+ primaryClass={cs.CL}
796
+ }
797
+ ```
data/token_list/bpe_unigram500/bpe.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ce4280f9977384de1303b1f0586b426caf9080a6f75c2b187a26cf1c87142550
3
+ size 245351
exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp/30epoch.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:776fa87066bbcf087368e5900e39e2c52ce51875c623f0689a64b1994363fc95
3
+ size 1723018857
exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp/RESULTS.md ADDED
@@ -0,0 +1,32 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ <!-- Generated by scripts/utils/show_asr_result.sh -->
2
+ # RESULTS
3
+ ## Environments
4
+ - date: `Sat Mar 19 23:02:37 EDT 2022`
5
+ - python version: `3.9.7 (default, Sep 16 2021, 13:09:58) [GCC 7.5.0]`
6
+ - espnet version: `espnet 0.10.7a1`
7
+ - pytorch version: `pytorch 1.10.1`
8
+ - Git hash: `c1ed71c6899e54c0b3dad82687886b1183cd0885`
9
+ - Commit date: `Wed Mar 16 23:34:49 2022 -0400`
10
+
11
+ ## asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp
12
+ ### WER
13
+
14
+ |dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err|
15
+ |---|---|---|---|---|---|---|---|---|
16
+ |decode_asr_lm_lm_train_lm_bpe500_valid.loss.ave_asr_model_latest/dev|482|4018|83.6|9.4|7.0|6.4|22.8|58.3|
17
+ |decode_asr_lm_lm_train_lm_bpe500_valid.loss.ave_asr_model_latest/test|480|4632|81.4|12.3|6.3|4.5|23.1|52.1|
18
+
19
+ ### CER
20
+
21
+ |dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err|
22
+ |---|---|---|---|---|---|---|---|---|
23
+ |decode_asr_lm_lm_train_lm_bpe500_valid.loss.ave_asr_model_latest/dev|482|18692|88.5|3.1|8.4|5.9|17.4|58.3|
24
+ |decode_asr_lm_lm_train_lm_bpe500_valid.loss.ave_asr_model_latest/test|480|21787|87.9|4.3|7.8|4.5|16.6|52.1|
25
+
26
+ ### TER
27
+
28
+ |dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err|
29
+ |---|---|---|---|---|---|---|---|---|
30
+ |decode_asr_lm_lm_train_lm_bpe500_valid.loss.ave_asr_model_latest/dev|482|6097|82.2|7.1|10.7|5.7|23.5|58.3|
31
+ |decode_asr_lm_lm_train_lm_bpe500_valid.loss.ave_asr_model_latest/test|480|7736|81.7|9.2|9.1|4.0|22.3|52.1|
32
+
exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp/config.yaml ADDED
@@ -0,0 +1,697 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ config: conf/tuning/train_asr_conformer7_hubert_ll60k_large.yaml
2
+ print_config: false
3
+ log_level: INFO
4
+ dry_run: false
5
+ iterator_type: sequence
6
+ output_dir: exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp
7
+ ngpu: 1
8
+ seed: 0
9
+ num_workers: 1
10
+ num_att_plot: 3
11
+ dist_backend: nccl
12
+ dist_init_method: env://
13
+ dist_world_size: null
14
+ dist_rank: null
15
+ local_rank: 0
16
+ dist_master_addr: null
17
+ dist_master_port: null
18
+ dist_launcher: null
19
+ multiprocessing_distributed: false
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: 35
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
+ - acc
39
+ - max
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: 8
46
+ no_forward_run: false
47
+ resume: true
48
+ train_dtype: float32
49
+ use_amp: false
50
+ log_interval: null
51
+ use_matplotlib: true
52
+ use_tensorboard: true
53
+ use_wandb: false
54
+ wandb_project: null
55
+ wandb_id: null
56
+ wandb_entity: null
57
+ wandb_name: null
58
+ wandb_model_log_interval: -1
59
+ detect_anomaly: false
60
+ pretrain_path: null
61
+ init_param: []
62
+ ignore_init_mismatch: false
63
+ freeze_param:
64
+ - frontend.upstream
65
+ num_iters_per_epoch: null
66
+ batch_size: 20
67
+ valid_batch_size: null
68
+ batch_bins: 1000000
69
+ valid_batch_bins: null
70
+ train_shape_file:
71
+ - exp/asr_stats_raw_bpe500_sp/train/speech_shape
72
+ - exp/asr_stats_raw_bpe500_sp/train/text_shape.bpe
73
+ valid_shape_file:
74
+ - exp/asr_stats_raw_bpe500_sp/valid/speech_shape
75
+ - exp/asr_stats_raw_bpe500_sp/valid/text_shape.bpe
76
+ batch_type: numel
77
+ valid_batch_type: null
78
+ fold_length:
79
+ - 80000
80
+ - 150
81
+ sort_in_batch: descending
82
+ sort_batch: descending
83
+ multiple_iterator: false
84
+ chunk_length: 500
85
+ chunk_shift_ratio: 0.5
86
+ num_cache_chunks: 1024
87
+ train_data_path_and_name_and_type:
88
+ - - dump/raw/train30_sp/wav.scp
89
+ - speech
90
+ - kaldi_ark
91
+ - - dump/raw/train30_sp/text
92
+ - text
93
+ - text
94
+ valid_data_path_and_name_and_type:
95
+ - - dump/raw/dev/wav.scp
96
+ - speech
97
+ - kaldi_ark
98
+ - - dump/raw/dev/text
99
+ - text
100
+ - text
101
+ allow_variable_data_keys: false
102
+ max_cache_size: 0.0
103
+ max_cache_fd: 32
104
+ valid_max_cache_size: null
105
+ optim: adam
106
+ optim_conf:
107
+ lr: 0.0025
108
+ scheduler: warmuplr
109
+ scheduler_conf:
110
+ warmup_steps: 40000
111
+ token_list:
112
+ - <blank>
113
+ - <unk>
114
+ - ▁I
115
+ - ''''
116
+ - ▁YOU
117
+ - S
118
+ - T
119
+ - ▁THE
120
+ - M
121
+ - ▁ME
122
+ - ▁A
123
+ - ▁AND
124
+ - ▁TO
125
+ - E
126
+ - A
127
+ - ING
128
+ - D
129
+ - ▁MY
130
+ - ▁
131
+ - O
132
+ - ▁IT
133
+ - I
134
+ - N
135
+ - RE
136
+ - Y
137
+ - ▁BE
138
+ - ▁IN
139
+ - ▁ON
140
+ - ▁LOVE
141
+ - U
142
+ - ▁WE
143
+ - LL
144
+ - H
145
+ - ▁YOUR
146
+ - ▁S
147
+ - IN
148
+ - ▁OF
149
+ - ▁DO
150
+ - ▁THAT
151
+ - ▁ALL
152
+ - L
153
+ - ▁DON
154
+ - ▁OH
155
+ - ▁LIKE
156
+ - ▁KNOW
157
+ - ▁FOR
158
+ - ▁CAN
159
+ - ▁JUST
160
+ - P
161
+ - ▁BUT
162
+ - ED
163
+ - K
164
+ - ▁WHEN
165
+ - ▁SO
166
+ - R
167
+ - ▁GO
168
+ - ▁WHAT
169
+ - ▁C
170
+ - ▁WITH
171
+ - W
172
+ - ▁F
173
+ - C
174
+ - ▁NO
175
+ - ER
176
+ - ▁ONE
177
+ - ▁LET
178
+ - VE
179
+ - ES
180
+ - ▁NOW
181
+ - ▁BABY
182
+ - G
183
+ - ▁GOT
184
+ - ▁COME
185
+ - CAUSE
186
+ - LE
187
+ - B
188
+ - ▁B
189
+ - AR
190
+ - ▁UP
191
+ - ▁'
192
+ - ▁W
193
+ - ▁SEE
194
+ - ▁TIME
195
+ - ▁ARE
196
+ - ▁G
197
+ - ▁LOOK
198
+ - ▁THIS
199
+ - F
200
+ - ▁IS
201
+ - ▁NEVER
202
+ - ▁M
203
+ - ▁P
204
+ - AN
205
+ - ▁WAS
206
+ - ▁WAY
207
+ - ▁IF
208
+ - OR
209
+ - ▁SAY
210
+ - V
211
+ - ▁R
212
+ - ▁T
213
+ - ▁DOWN
214
+ - RA
215
+ - ▁THERE
216
+ - ▁HEART
217
+ - ▁NOT
218
+ - RO
219
+ - ▁WILL
220
+ - ▁OUT
221
+ - CE
222
+ - ▁WANT
223
+ - ▁YEAH
224
+ - ▁HAVE
225
+ - ▁GIVE
226
+ - ▁TOO
227
+ - ▁GONNA
228
+ - ▁HOW
229
+ - ▁NEED
230
+ - ▁GET
231
+ - ▁TAKE
232
+ - ▁EVERY
233
+ - ▁FEEL
234
+ - ▁HE
235
+ - EN
236
+ - ▁FROM
237
+ - ▁HA
238
+ - ▁K
239
+ - ▁SHE
240
+ - 'ON'
241
+ - ▁DI
242
+ - RI
243
+ - ▁ONLY
244
+ - NE
245
+ - ▁WHO
246
+ - ▁AWAY
247
+ - ▁E
248
+ - ▁D
249
+ - ▁LIFE
250
+ - ▁MAKE
251
+ - IC
252
+ - ▁BACK
253
+ - ▁WHERE
254
+ - ▁MADE
255
+ - ▁DAY
256
+ - ▁HERE
257
+ - ▁LO
258
+ - ▁HER
259
+ - ▁AS
260
+ - ▁GOOD
261
+ - ▁WANNA
262
+ - ▁OOH
263
+ - ▁TELL
264
+ - LY
265
+ - TH
266
+ - ▁WON
267
+ - ▁LIGHT
268
+ - ▁KEEP
269
+ - ▁MA
270
+ - ▁LA
271
+ - ▁SH
272
+ - ▁WORLD
273
+ - ▁MORE
274
+ - ▁LI
275
+ - AL
276
+ - ▁COULD
277
+ - ▁GIRL
278
+ - ▁NOTHING
279
+ - ▁EVER
280
+ - ▁THINK
281
+ - IE
282
+ - ▁BY
283
+ - ▁AT
284
+ - ▁TONIGHT
285
+ - ▁THEY
286
+ - ▁CALL
287
+ - ▁HO
288
+ - ▁WOULD
289
+ - IL
290
+ - ▁OUR
291
+ - ▁FALL
292
+ - ▁NIGHT
293
+ - ▁THAN
294
+ - ▁DE
295
+ - ▁SOME
296
+ - ▁WAIT
297
+ - ▁RIGHT
298
+ - ▁RE
299
+ - ▁HALLELUJAH
300
+ - ▁TH
301
+ - NG
302
+ - ▁CO
303
+ - ▁WERE
304
+ - ▁TALK
305
+ - ET
306
+ - ▁BO
307
+ - ▁HOLD
308
+ - UR
309
+ - ▁BEEN
310
+ - ▁US
311
+ - ▁PA
312
+ - VER
313
+ - ▁EYES
314
+ - ▁DREAM
315
+ - ▁SONG
316
+ - ▁SHOULD
317
+ - ▁STILL
318
+ - ▁OVER
319
+ - TA
320
+ - ▁ANYMORE
321
+ - IGHT
322
+ - ▁STAY
323
+ - ▁BETTER
324
+ - LESS
325
+ - ▁THROUGH
326
+ - ▁LITTLE
327
+ - X
328
+ - ▁GONE
329
+ - ▁AIN
330
+ - ▁DA
331
+ - ▁HOLDING
332
+ - ▁HURT
333
+ - ▁TRY
334
+ - ▁FIND
335
+ - Z
336
+ - DE
337
+ - ▁LAST
338
+ - ▁SAID
339
+ - ▁ALWAYS
340
+ - ▁BODY
341
+ - ▁MIND
342
+ - ▁CRY
343
+ - ▁EVEN
344
+ - ▁RUN
345
+ - ▁HOPE
346
+ - ▁WITHOUT
347
+ - ▁MISS
348
+ - ▁ABOUT
349
+ - ▁HAND
350
+ - ▁J
351
+ - ▁AGAIN
352
+ - ▁THOUGH
353
+ - ▁NAH
354
+ - ▁LIVE
355
+ - ▁BA
356
+ - ▁OLD
357
+ - ▁HEAD
358
+ - ▁FIRE
359
+ - ▁MAN
360
+ - ▁SOMETHING
361
+ - ▁WHY
362
+ - THER
363
+ - ▁HOME
364
+ - ▁OR
365
+ - ▁INSIDE
366
+ - ▁NEW
367
+ - ▁HEY
368
+ - TION
369
+ - ▁EVERYTHING
370
+ - ▁HAD
371
+ - ▁SOMETIMES
372
+ - ▁HARD
373
+ - ▁TOUCH
374
+ - ▁HEAR
375
+ - ▁AM
376
+ - ▁MUCH
377
+ - ▁LONG
378
+ - ▁STAR
379
+ - GETTING
380
+ - ▁WALK
381
+ - ▁PEOPLE
382
+ - ▁BEFORE
383
+ - ▁CLOSE
384
+ - ▁TWO
385
+ - ▁FAR
386
+ - ▁SHOW
387
+ - ▁STAND
388
+ - ▁LOSE
389
+ - ▁HELP
390
+ - ▁NAME
391
+ - ▁BOY
392
+ - ▁TRUE
393
+ - ▁PLAY
394
+ - ▁DARK
395
+ - ▁THINGS
396
+ - ▁NA
397
+ - ▁TEAR
398
+ - ▁END
399
+ - ▁NOBODY
400
+ - ▁SEA
401
+ - ▁ROCKABYE
402
+ - ▁BELIEVE
403
+ - ▁BROKE
404
+ - ▁AROUND
405
+ - ▁START
406
+ - ▁KISS
407
+ - ▁FEELING
408
+ - ▁BREAK
409
+ - ▁SOMEONE
410
+ - ▁FRIEND
411
+ - ▁ALONE
412
+ - ▁BEAUTIFUL
413
+ - ▁CRAZY
414
+ - ▁OWN
415
+ - OSE
416
+ - ▁STOP
417
+ - ▁LOST
418
+ - ▁HIM
419
+ - ▁BAD
420
+ - ▁CHANCE
421
+ - ▁REALLY
422
+ - ▁WISH
423
+ - ▁MOVE
424
+ - ▁SKY
425
+ - ▁PLACE
426
+ - AKE
427
+ - ▁LEAVE
428
+ - ▁YA
429
+ - ▁STRONG
430
+ - ▁PUT
431
+ - ▁OPEN
432
+ - ▁WRONG
433
+ - ▁COLD
434
+ - OCK
435
+ - ▁USED
436
+ - ▁FOUND
437
+ - ▁LONELY
438
+ - ▁DANCE
439
+ - EACH
440
+ - ▁ANOTHER
441
+ - ▁SIDE
442
+ - ▁UNDER
443
+ - ▁MATTER
444
+ - ▁THESE
445
+ - ▁CARE
446
+ - ▁MINE
447
+ - ▁SHINE
448
+ - ▁AFRAID
449
+ - ▁TURN
450
+ - ▁PLEASE
451
+ - ▁SUN
452
+ - ▁DIAMOND
453
+ - ▁UNTIL
454
+ - ▁FACE
455
+ - ▁LEARN
456
+ - ▁TRUST
457
+ - ▁WONDER
458
+ - ▁BREATH
459
+ - ATE
460
+ - ▁SORRY
461
+ - ▁HU
462
+ - ▁WATCH
463
+ - ▁LATE
464
+ - ROUND
465
+ - ▁ARMS
466
+ - ▁PERFECT
467
+ - ▁MAYBE
468
+ - ▁PULL
469
+ - ▁REMEMBER
470
+ - ▁FIGHT
471
+ - ▁MYSELF
472
+ - ▁INTO
473
+ - ▁DARLING
474
+ - ▁THUNDER
475
+ - ▁FOLLOW
476
+ - ▁REASON
477
+ - ▁BURN
478
+ - ▁HIS
479
+ - ▁MUST
480
+ - ▁FREE
481
+ - ▁FLASHLIGHT
482
+ - ▁1
483
+ - ▁ENOUGH
484
+ - ▁DRINK
485
+ - ▁WORDS
486
+ - ▁HIDE
487
+ - ▁UN
488
+ - ▁FORGET
489
+ - ▁SURE
490
+ - ▁CHANGE
491
+ - ▁SMILE
492
+ - ▁PROMISE
493
+ - ▁FOREVER
494
+ - '2'
495
+ - ▁SWEET
496
+ - ▁SAME
497
+ - ▁OOOH
498
+ - ▁PART
499
+ - ▁SOMEBODY
500
+ - NESS
501
+ - ▁BRIGHT
502
+ - ▁HEAVEN
503
+ - ▁DEEP
504
+ - ▁HIGH
505
+ - ▁INSTEAD
506
+ - ▁MOMENT
507
+ - ▁ALONG
508
+ - ▁ALRIGHT
509
+ - ▁SLOW
510
+ - ▁TOMORROW
511
+ - ▁SOUL
512
+ - ▁QU
513
+ - ▁PUSH
514
+ - ▁CHANDELIER
515
+ - ▁LEFT
516
+ - SIDE
517
+ - ▁TOLD
518
+ - ▁KNEW
519
+ - READY
520
+ - ▁LOVING
521
+ - ▁SAW
522
+ - '3'
523
+ - ▁WORK
524
+ - ▁DANCING
525
+ - ▁THREE
526
+ - ▁SAVE
527
+ - ▁SHOOT
528
+ - ▁LEAD
529
+ - ▁SKI
530
+ - ▁WILD
531
+ - ▁WIND
532
+ - ▁WHILE
533
+ - ▁EDGE
534
+ - ▁HAPPY
535
+ - ▁FEAR
536
+ - STUCK
537
+ - ▁MOST
538
+ - ▁LISTEN
539
+ - ▁WOAH
540
+ - ▁FIRST
541
+ - ▁JOLENE
542
+ - ▁VOICE
543
+ - ▁COMP
544
+ - ▁MILLION
545
+ - FUL
546
+ - ▁OOOOOH
547
+ - ▁CAME
548
+ - ▁RISE
549
+ - ▁NEXT
550
+ - ▁COUNT
551
+ - ▁MOUNTAIN
552
+ - ▁ROOM
553
+ - ▁BLUE
554
+ - ▁HIT
555
+ - ▁RAISE
556
+ - J
557
+ - ▁THOUSAND
558
+ - ▁SHAP
559
+ - ▁TREAT
560
+ - ▁DRY
561
+ - ▁FINALLY
562
+ - ▁TITANIUM
563
+ - ▁CARRY
564
+ - ▁TRUTH
565
+ - ▁WATER
566
+ - ▁MORNING
567
+ - TIME
568
+ - ▁BELONG
569
+ - ▁UMA
570
+ - ▁ALIVE
571
+ - ▁ELSE
572
+ - ▁ANGEL
573
+ - ▁BRAND
574
+ - ▁APART
575
+ - ▁EVERYBODY
576
+ - ▁SOUND
577
+ - ▁GUESS
578
+ - ▁PRAY
579
+ - ▁FAITH
580
+ - ▁AFTER
581
+ - ▁THROW
582
+ - ▁TRIED
583
+ - ▁SLEEP
584
+ - ▁FOOL
585
+ - ▁DISCOVERING
586
+ - ▁FUCK
587
+ - ▁TASTE
588
+ - ▁UNDERSTAND
589
+ - ▁SHAME
590
+ - ▁POWER
591
+ - ▁WELCOME
592
+ - ▁FELT
593
+ - ▁SAFE
594
+ - ▁DESERVE
595
+ - ▁GAME
596
+ - ▁SUPERMA
597
+ - ▁SWEAR
598
+ - ▁BETWEEN
599
+ - ▁GLASS
600
+ - ▁CATCH
601
+ - ▁TOGETHER
602
+ - '0'
603
+ - '4'
604
+ - '6'
605
+ - '5'
606
+ - '1'
607
+ - '8'
608
+ - '7'
609
+ - '9'
610
+ - Q
611
+ - <sos/eos>
612
+ init: null
613
+ input_size: null
614
+ ctc_conf:
615
+ dropout_rate: 0.0
616
+ ctc_type: builtin
617
+ reduce: true
618
+ ignore_nan_grad: true
619
+ joint_net_conf: null
620
+ model_conf:
621
+ ctc_weight: 0.3
622
+ lsm_weight: 0.1
623
+ length_normalized_loss: false
624
+ extract_feats_in_collect_stats: false
625
+ use_preprocessor: true
626
+ token_type: bpe
627
+ bpemodel: data/token_list/bpe_unigram500/bpe.model
628
+ non_linguistic_symbols: null
629
+ cleaner: null
630
+ g2p: null
631
+ speech_volume_normalize: null
632
+ rir_scp: null
633
+ rir_apply_prob: 1.0
634
+ noise_scp: null
635
+ noise_apply_prob: 1.0
636
+ noise_db_range: '13_15'
637
+ frontend: s3prl
638
+ frontend_conf:
639
+ frontend_conf:
640
+ upstream: hubert_large_ll60k
641
+ download_dir: ./hub
642
+ multilayer_feature: true
643
+ fs: 16k
644
+ specaug: specaug
645
+ specaug_conf:
646
+ apply_time_warp: true
647
+ time_warp_window: 5
648
+ time_warp_mode: bicubic
649
+ apply_freq_mask: true
650
+ freq_mask_width_range:
651
+ - 0
652
+ - 30
653
+ num_freq_mask: 2
654
+ apply_time_mask: true
655
+ time_mask_width_range:
656
+ - 0
657
+ - 40
658
+ num_time_mask: 2
659
+ normalize: utterance_mvn
660
+ normalize_conf: {}
661
+ preencoder: linear
662
+ preencoder_conf:
663
+ input_size: 1024
664
+ output_size: 80
665
+ encoder: conformer
666
+ encoder_conf:
667
+ output_size: 512
668
+ attention_heads: 8
669
+ linear_units: 2048
670
+ num_blocks: 12
671
+ dropout_rate: 0.1
672
+ positional_dropout_rate: 0.1
673
+ attention_dropout_rate: 0.1
674
+ input_layer: conv2d2
675
+ normalize_before: true
676
+ macaron_style: true
677
+ pos_enc_layer_type: rel_pos
678
+ selfattention_layer_type: rel_selfattn
679
+ activation_type: swish
680
+ use_cnn_module: true
681
+ cnn_module_kernel: 31
682
+ postencoder: null
683
+ postencoder_conf: {}
684
+ decoder: transformer
685
+ decoder_conf:
686
+ attention_heads: 8
687
+ linear_units: 2048
688
+ num_blocks: 6
689
+ dropout_rate: 0.1
690
+ positional_dropout_rate: 0.1
691
+ self_attention_dropout_rate: 0.1
692
+ src_attention_dropout_rate: 0.1
693
+ required:
694
+ - output_dir
695
+ - token_list
696
+ version: 0.10.7a1
697
+ distributed: false
exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp/images/acc.png ADDED
exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp/images/backward_time.png ADDED
exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp/images/cer.png ADDED
exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp/images/cer_ctc.png ADDED
exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp/images/forward_time.png ADDED
exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp/images/gpu_max_cached_mem_GB.png ADDED
exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp/images/iter_time.png ADDED
exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp/images/loss.png ADDED
exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp/images/loss_att.png ADDED
exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp/images/loss_ctc.png ADDED
exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp/images/optim0_lr0.png ADDED
exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp/images/optim_step_time.png ADDED
exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp/images/train_time.png ADDED
exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp/images/wer.png ADDED
exp/lm_train_lm_bpe500/19epoch.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6f6ced3992cb3c9f0055c829aa9c8820b61556dc2679da2a279363db8973f472
3
+ size 29685855
exp/lm_train_lm_bpe500/config.yaml ADDED
@@ -0,0 +1,617 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ config: conf/train_lm.yaml
2
+ print_config: false
3
+ log_level: INFO
4
+ dry_run: false
5
+ iterator_type: sequence
6
+ output_dir: exp/lm_train_lm_bpe500
7
+ ngpu: 1
8
+ seed: 0
9
+ num_workers: 1
10
+ num_att_plot: 3
11
+ dist_backend: nccl
12
+ dist_init_method: env://
13
+ dist_world_size: null
14
+ dist_rank: null
15
+ local_rank: 0
16
+ dist_master_addr: null
17
+ dist_master_port: null
18
+ dist_launcher: null
19
+ multiprocessing_distributed: false
20
+ unused_parameters: false
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: 20
28
+ patience: 3
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: 1
41
+ nbest_averaging_interval: 0
42
+ grad_clip: 5.0
43
+ grad_clip_type: 2.0
44
+ grad_noise: false
45
+ accum_grad: 1
46
+ no_forward_run: false
47
+ resume: true
48
+ train_dtype: float32
49
+ use_amp: false
50
+ log_interval: null
51
+ use_matplotlib: true
52
+ use_tensorboard: true
53
+ use_wandb: false
54
+ wandb_project: null
55
+ wandb_id: null
56
+ wandb_entity: null
57
+ wandb_name: null
58
+ wandb_model_log_interval: -1
59
+ detect_anomaly: false
60
+ pretrain_path: null
61
+ init_param: []
62
+ ignore_init_mismatch: false
63
+ freeze_param: []
64
+ num_iters_per_epoch: null
65
+ batch_size: 64
66
+ valid_batch_size: null
67
+ batch_bins: 1000000
68
+ valid_batch_bins: null
69
+ train_shape_file:
70
+ - exp/lm_stats_bpe500/train/text_shape.bpe
71
+ valid_shape_file:
72
+ - exp/lm_stats_bpe500/valid/text_shape.bpe
73
+ batch_type: folded
74
+ valid_batch_type: null
75
+ fold_length:
76
+ - 150
77
+ sort_in_batch: descending
78
+ sort_batch: descending
79
+ multiple_iterator: false
80
+ chunk_length: 500
81
+ chunk_shift_ratio: 0.5
82
+ num_cache_chunks: 1024
83
+ train_data_path_and_name_and_type:
84
+ - - dump/raw/lm_train.txt
85
+ - text
86
+ - text
87
+ valid_data_path_and_name_and_type:
88
+ - - dump/raw/dev/text
89
+ - text
90
+ - text
91
+ allow_variable_data_keys: false
92
+ max_cache_size: 0.0
93
+ max_cache_fd: 32
94
+ valid_max_cache_size: null
95
+ optim: sgd
96
+ optim_conf: {}
97
+ scheduler: null
98
+ scheduler_conf: {}
99
+ token_list:
100
+ - <blank>
101
+ - <unk>
102
+ - ▁I
103
+ - ''''
104
+ - ▁YOU
105
+ - S
106
+ - T
107
+ - ▁THE
108
+ - M
109
+ - ▁ME
110
+ - ▁A
111
+ - ▁AND
112
+ - ▁TO
113
+ - E
114
+ - A
115
+ - ING
116
+ - D
117
+ - ▁MY
118
+ - ▁
119
+ - O
120
+ - ▁IT
121
+ - I
122
+ - N
123
+ - RE
124
+ - Y
125
+ - ▁BE
126
+ - ▁IN
127
+ - ▁ON
128
+ - ▁LOVE
129
+ - U
130
+ - ▁WE
131
+ - LL
132
+ - H
133
+ - ▁YOUR
134
+ - ▁S
135
+ - IN
136
+ - ▁OF
137
+ - ▁DO
138
+ - ▁THAT
139
+ - ▁ALL
140
+ - L
141
+ - ▁DON
142
+ - ▁OH
143
+ - ▁LIKE
144
+ - ▁KNOW
145
+ - ▁FOR
146
+ - ▁CAN
147
+ - ▁JUST
148
+ - P
149
+ - ▁BUT
150
+ - ED
151
+ - K
152
+ - ▁WHEN
153
+ - ▁SO
154
+ - R
155
+ - ▁GO
156
+ - ▁WHAT
157
+ - ▁C
158
+ - ▁WITH
159
+ - W
160
+ - ▁F
161
+ - C
162
+ - ▁NO
163
+ - ER
164
+ - ▁ONE
165
+ - ▁LET
166
+ - VE
167
+ - ES
168
+ - ▁NOW
169
+ - ▁BABY
170
+ - G
171
+ - ▁GOT
172
+ - ▁COME
173
+ - CAUSE
174
+ - LE
175
+ - B
176
+ - ▁B
177
+ - AR
178
+ - ▁UP
179
+ - ▁'
180
+ - ▁W
181
+ - ▁SEE
182
+ - ▁TIME
183
+ - ▁ARE
184
+ - ▁G
185
+ - ▁LOOK
186
+ - ▁THIS
187
+ - F
188
+ - ▁IS
189
+ - ▁NEVER
190
+ - ▁M
191
+ - ▁P
192
+ - AN
193
+ - ▁WAS
194
+ - ▁WAY
195
+ - ▁IF
196
+ - OR
197
+ - ▁SAY
198
+ - V
199
+ - ▁R
200
+ - ▁T
201
+ - ▁DOWN
202
+ - RA
203
+ - ▁THERE
204
+ - ▁HEART
205
+ - ▁NOT
206
+ - RO
207
+ - ▁WILL
208
+ - ▁OUT
209
+ - CE
210
+ - ▁WANT
211
+ - ▁YEAH
212
+ - ▁HAVE
213
+ - ▁GIVE
214
+ - ▁TOO
215
+ - ▁GONNA
216
+ - ▁HOW
217
+ - ▁NEED
218
+ - ▁GET
219
+ - ▁TAKE
220
+ - ▁EVERY
221
+ - ▁FEEL
222
+ - ▁HE
223
+ - EN
224
+ - ▁FROM
225
+ - ▁HA
226
+ - ▁K
227
+ - ▁SHE
228
+ - 'ON'
229
+ - ▁DI
230
+ - RI
231
+ - ▁ONLY
232
+ - NE
233
+ - ▁WHO
234
+ - ▁AWAY
235
+ - ▁E
236
+ - ▁D
237
+ - ▁LIFE
238
+ - ▁MAKE
239
+ - IC
240
+ - ▁BACK
241
+ - ▁WHERE
242
+ - ▁MADE
243
+ - ▁DAY
244
+ - ▁HERE
245
+ - ▁LO
246
+ - ▁HER
247
+ - ▁AS
248
+ - ▁GOOD
249
+ - ▁WANNA
250
+ - ▁OOH
251
+ - ▁TELL
252
+ - LY
253
+ - TH
254
+ - ▁WON
255
+ - ▁LIGHT
256
+ - ▁KEEP
257
+ - ▁MA
258
+ - ▁LA
259
+ - ▁SH
260
+ - ▁WORLD
261
+ - ▁MORE
262
+ - ▁LI
263
+ - AL
264
+ - ▁COULD
265
+ - ▁GIRL
266
+ - ▁NOTHING
267
+ - ▁EVER
268
+ - ▁THINK
269
+ - IE
270
+ - ▁BY
271
+ - ▁AT
272
+ - ▁TONIGHT
273
+ - ▁THEY
274
+ - ▁CALL
275
+ - ▁HO
276
+ - ▁WOULD
277
+ - IL
278
+ - ▁OUR
279
+ - ▁FALL
280
+ - ▁NIGHT
281
+ - ▁THAN
282
+ - ▁DE
283
+ - ▁SOME
284
+ - ▁WAIT
285
+ - ▁RIGHT
286
+ - ▁RE
287
+ - ▁HALLELUJAH
288
+ - ▁TH
289
+ - NG
290
+ - ▁CO
291
+ - ▁WERE
292
+ - ▁TALK
293
+ - ET
294
+ - ▁BO
295
+ - ▁HOLD
296
+ - UR
297
+ - ▁BEEN
298
+ - ▁US
299
+ - ▁PA
300
+ - VER
301
+ - ▁EYES
302
+ - ▁DREAM
303
+ - ▁SONG
304
+ - ▁SHOULD
305
+ - ▁STILL
306
+ - ▁OVER
307
+ - TA
308
+ - ▁ANYMORE
309
+ - IGHT
310
+ - ▁STAY
311
+ - ▁BETTER
312
+ - LESS
313
+ - ▁THROUGH
314
+ - ▁LITTLE
315
+ - X
316
+ - ▁GONE
317
+ - ▁AIN
318
+ - ▁DA
319
+ - ▁HOLDING
320
+ - ▁HURT
321
+ - ▁TRY
322
+ - ▁FIND
323
+ - Z
324
+ - DE
325
+ - ▁LAST
326
+ - ▁SAID
327
+ - ▁ALWAYS
328
+ - ▁BODY
329
+ - ▁MIND
330
+ - ▁CRY
331
+ - ▁EVEN
332
+ - ▁RUN
333
+ - ▁HOPE
334
+ - ▁WITHOUT
335
+ - ▁MISS
336
+ - ▁ABOUT
337
+ - ▁HAND
338
+ - ▁J
339
+ - ▁AGAIN
340
+ - ▁THOUGH
341
+ - ▁NAH
342
+ - ▁LIVE
343
+ - ▁BA
344
+ - ▁OLD
345
+ - ▁HEAD
346
+ - ▁FIRE
347
+ - ▁MAN
348
+ - ▁SOMETHING
349
+ - ▁WHY
350
+ - THER
351
+ - ▁HOME
352
+ - ▁OR
353
+ - ▁INSIDE
354
+ - ▁NEW
355
+ - ▁HEY
356
+ - TION
357
+ - ▁EVERYTHING
358
+ - ▁HAD
359
+ - ▁SOMETIMES
360
+ - ▁HARD
361
+ - ▁TOUCH
362
+ - ▁HEAR
363
+ - ▁AM
364
+ - ▁MUCH
365
+ - ▁LONG
366
+ - ▁STAR
367
+ - GETTING
368
+ - ▁WALK
369
+ - ▁PEOPLE
370
+ - ▁BEFORE
371
+ - ▁CLOSE
372
+ - ▁TWO
373
+ - ▁FAR
374
+ - ▁SHOW
375
+ - ▁STAND
376
+ - ▁LOSE
377
+ - ▁HELP
378
+ - ▁NAME
379
+ - ▁BOY
380
+ - ▁TRUE
381
+ - ▁PLAY
382
+ - ▁DARK
383
+ - ▁THINGS
384
+ - ▁NA
385
+ - ▁TEAR
386
+ - ▁END
387
+ - ▁NOBODY
388
+ - ▁SEA
389
+ - ▁ROCKABYE
390
+ - ▁BELIEVE
391
+ - ▁BROKE
392
+ - ▁AROUND
393
+ - ▁START
394
+ - ▁KISS
395
+ - ▁FEELING
396
+ - ▁BREAK
397
+ - ▁SOMEONE
398
+ - ▁FRIEND
399
+ - ▁ALONE
400
+ - ▁BEAUTIFUL
401
+ - ▁CRAZY
402
+ - ▁OWN
403
+ - OSE
404
+ - ▁STOP
405
+ - ▁LOST
406
+ - ▁HIM
407
+ - ▁BAD
408
+ - ▁CHANCE
409
+ - ▁REALLY
410
+ - ▁WISH
411
+ - ▁MOVE
412
+ - ▁SKY
413
+ - ▁PLACE
414
+ - AKE
415
+ - ▁LEAVE
416
+ - ▁YA
417
+ - ▁STRONG
418
+ - ▁PUT
419
+ - ▁OPEN
420
+ - ▁WRONG
421
+ - ▁COLD
422
+ - OCK
423
+ - ▁USED
424
+ - ▁FOUND
425
+ - ▁LONELY
426
+ - ▁DANCE
427
+ - EACH
428
+ - ▁ANOTHER
429
+ - ▁SIDE
430
+ - ▁UNDER
431
+ - ▁MATTER
432
+ - ▁THESE
433
+ - ▁CARE
434
+ - ▁MINE
435
+ - ▁SHINE
436
+ - ▁AFRAID
437
+ - ▁TURN
438
+ - ▁PLEASE
439
+ - ▁SUN
440
+ - ▁DIAMOND
441
+ - ▁UNTIL
442
+ - ▁FACE
443
+ - ▁LEARN
444
+ - ▁TRUST
445
+ - ▁WONDER
446
+ - ▁BREATH
447
+ - ATE
448
+ - ▁SORRY
449
+ - ▁HU
450
+ - ▁WATCH
451
+ - ▁LATE
452
+ - ROUND
453
+ - ▁ARMS
454
+ - ▁PERFECT
455
+ - ▁MAYBE
456
+ - ▁PULL
457
+ - ▁REMEMBER
458
+ - ▁FIGHT
459
+ - ▁MYSELF
460
+ - ▁INTO
461
+ - ▁DARLING
462
+ - ▁THUNDER
463
+ - ▁FOLLOW
464
+ - ▁REASON
465
+ - ▁BURN
466
+ - ▁HIS
467
+ - ▁MUST
468
+ - ▁FREE
469
+ - ▁FLASHLIGHT
470
+ - ▁1
471
+ - ▁ENOUGH
472
+ - ▁DRINK
473
+ - ▁WORDS
474
+ - ▁HIDE
475
+ - ▁UN
476
+ - ▁FORGET
477
+ - ▁SURE
478
+ - ▁CHANGE
479
+ - ▁SMILE
480
+ - ▁PROMISE
481
+ - ▁FOREVER
482
+ - '2'
483
+ - ▁SWEET
484
+ - ▁SAME
485
+ - ▁OOOH
486
+ - ▁PART
487
+ - ▁SOMEBODY
488
+ - NESS
489
+ - ▁BRIGHT
490
+ - ▁HEAVEN
491
+ - ▁DEEP
492
+ - ▁HIGH
493
+ - ▁INSTEAD
494
+ - ▁MOMENT
495
+ - ▁ALONG
496
+ - ▁ALRIGHT
497
+ - ▁SLOW
498
+ - ▁TOMORROW
499
+ - ▁SOUL
500
+ - ▁QU
501
+ - ▁PUSH
502
+ - ▁CHANDELIER
503
+ - ▁LEFT
504
+ - SIDE
505
+ - ▁TOLD
506
+ - ▁KNEW
507
+ - READY
508
+ - ▁LOVING
509
+ - ▁SAW
510
+ - '3'
511
+ - ▁WORK
512
+ - ▁DANCING
513
+ - ▁THREE
514
+ - ▁SAVE
515
+ - ▁SHOOT
516
+ - ▁LEAD
517
+ - ▁SKI
518
+ - ▁WILD
519
+ - ▁WIND
520
+ - ▁WHILE
521
+ - ▁EDGE
522
+ - ▁HAPPY
523
+ - ▁FEAR
524
+ - STUCK
525
+ - ▁MOST
526
+ - ▁LISTEN
527
+ - ▁WOAH
528
+ - ▁FIRST
529
+ - ▁JOLENE
530
+ - ▁VOICE
531
+ - ▁COMP
532
+ - ▁MILLION
533
+ - FUL
534
+ - ▁OOOOOH
535
+ - ▁CAME
536
+ - ▁RISE
537
+ - ▁NEXT
538
+ - ▁COUNT
539
+ - ▁MOUNTAIN
540
+ - ▁ROOM
541
+ - ▁BLUE
542
+ - ▁HIT
543
+ - ▁RAISE
544
+ - J
545
+ - ▁THOUSAND
546
+ - ▁SHAP
547
+ - ▁TREAT
548
+ - ▁DRY
549
+ - ▁FINALLY
550
+ - ▁TITANIUM
551
+ - ▁CARRY
552
+ - ▁TRUTH
553
+ - ▁WATER
554
+ - ▁MORNING
555
+ - TIME
556
+ - ▁BELONG
557
+ - ▁UMA
558
+ - ▁ALIVE
559
+ - ▁ELSE
560
+ - ▁ANGEL
561
+ - ▁BRAND
562
+ - ▁APART
563
+ - ▁EVERYBODY
564
+ - ▁SOUND
565
+ - ▁GUESS
566
+ - ▁PRAY
567
+ - ▁FAITH
568
+ - ▁AFTER
569
+ - ▁THROW
570
+ - ▁TRIED
571
+ - ▁SLEEP
572
+ - ▁FOOL
573
+ - ▁DISCOVERING
574
+ - ▁FUCK
575
+ - ▁TASTE
576
+ - ▁UNDERSTAND
577
+ - ▁SHAME
578
+ - ▁POWER
579
+ - ▁WELCOME
580
+ - ▁FELT
581
+ - ▁SAFE
582
+ - ▁DESERVE
583
+ - ▁GAME
584
+ - ▁SUPERMA
585
+ - ▁SWEAR
586
+ - ▁BETWEEN
587
+ - ▁GLASS
588
+ - ▁CATCH
589
+ - ▁TOGETHER
590
+ - '0'
591
+ - '4'
592
+ - '6'
593
+ - '5'
594
+ - '1'
595
+ - '8'
596
+ - '7'
597
+ - '9'
598
+ - Q
599
+ - <sos/eos>
600
+ init: null
601
+ model_conf:
602
+ ignore_id: 0
603
+ use_preprocessor: true
604
+ token_type: bpe
605
+ bpemodel: data/token_list/bpe_unigram500/bpe.model
606
+ non_linguistic_symbols: null
607
+ cleaner: null
608
+ g2p: null
609
+ lm: seq_rnn
610
+ lm_conf:
611
+ nlayers: 2
612
+ unit: 650
613
+ required:
614
+ - output_dir
615
+ - token_list
616
+ version: 0.10.7a1
617
+ distributed: false
exp/lm_train_lm_bpe500/images/backward_time.png ADDED
exp/lm_train_lm_bpe500/images/forward_time.png ADDED
exp/lm_train_lm_bpe500/images/gpu_max_cached_mem_GB.png ADDED
exp/lm_train_lm_bpe500/images/iter_time.png ADDED
exp/lm_train_lm_bpe500/images/loss.png ADDED
exp/lm_train_lm_bpe500/images/optim0_lr0.png ADDED
exp/lm_train_lm_bpe500/images/optim_step_time.png ADDED
exp/lm_train_lm_bpe500/images/train_time.png ADDED
exp/lm_train_lm_bpe500/perplexity_test/ppl ADDED
@@ -0,0 +1 @@
 
 
1
+ 10.403652830281487
meta.yaml ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ espnet: 0.10.7a1
2
+ files:
3
+ asr_model_file: exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp/30epoch.pth
4
+ lm_file: exp/lm_train_lm_bpe500/19epoch.pth
5
+ python: "3.9.7 (default, Sep 16 2021, 13:09:58) \n[GCC 7.5.0]"
6
+ timestamp: 1647751540.320012
7
+ torch: 1.10.1
8
+ yaml_files:
9
+ asr_train_config: exp/asr_train_asr_conformer7_hubert_ll60k_large_raw_bpe500_sp/config.yaml
10
+ lm_train_config: exp/lm_train_lm_bpe500/config.yaml