asahi417 commited on
Commit
e98771d
1 Parent(s): b8413a4
Files changed (2) hide show
  1. main_s2s.sh +53 -215
  2. tokenize_dataset_s2s.py +1 -2
main_s2s.sh CHANGED
@@ -1,3 +1,42 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ######################
2
  # enA-frA: 3_054_258 #
3
  ######################
@@ -8,36 +47,6 @@ export LINE_NO_START=0
8
  export LINE_NO_END=10
9
  python fetch_dataset_s2s.py
10
  # main
11
- for i in $(seq 181 200);
12
- do
13
- export N_POOL=15
14
- export DATASET_ID=${i}
15
- export DIRECTION="enA-frA"
16
- export LINE_NO_START=$(((DATASET_ID-1) * 2500))
17
- export LINE_NO_END=$((DATASET_ID * 2500))
18
- echo ${LINE_NO_START}
19
- python fetch_dataset_s2s.py
20
- done
21
- for i in $(seq 256 280);
22
- do
23
- export N_POOL=15
24
- export DATASET_ID=${i}
25
- export DIRECTION="enA-frA"
26
- export LINE_NO_START=$(((DATASET_ID-1) * 2500))
27
- export LINE_NO_END=$((DATASET_ID * 2500))
28
- echo ${LINE_NO_START}
29
- python fetch_dataset_s2s.py
30
- done
31
- for i in $(seq 281 299);
32
- do
33
- export N_POOL=15
34
- export DATASET_ID=${i}
35
- export DIRECTION="enA-frA"
36
- export LINE_NO_START=$(((DATASET_ID-1) * 2500))
37
- export LINE_NO_END=$((DATASET_ID * 2500))
38
- echo ${LINE_NO_START}
39
- python fetch_dataset_s2s.py
40
- done
41
  for i in $(seq 1 600);
42
  do
43
  export N_POOL=15
@@ -82,48 +91,7 @@ export LINE_NO_START=0
82
  export LINE_NO_END=10
83
  python fetch_dataset_s2s.py
84
  # main
85
- for i in $(seq 345 420);
86
- do
87
- export N_POOL=15
88
- export DATASET_ID=${i}
89
- export DIRECTION="enA-esA"
90
- export LINE_NO_START=$(((DATASET_ID-1) * 2500))
91
- export LINE_NO_END=$((DATASET_ID * 2500))
92
- echo ${LINE_NO_START}
93
- python fetch_dataset_s2s.py
94
- done
95
- for i in $(seq 390 420);
96
- do
97
- export N_POOL=15
98
- export DATASET_ID=${i}
99
- export DIRECTION="enA-esA"
100
- export LINE_NO_START=$(((DATASET_ID-1) * 2500))
101
- export LINE_NO_END=$((DATASET_ID * 2500))
102
- echo ${LINE_NO_START}
103
- python fetch_dataset_s2s.py
104
- done
105
- for i in $(seq 451 500);
106
- do
107
- export N_POOL=15
108
- export DATASET_ID=${i}
109
- export DIRECTION="enA-esA"
110
- export LINE_NO_START=$(((DATASET_ID-1) * 2500))
111
- export LINE_NO_END=$((DATASET_ID * 2500))
112
- echo ${LINE_NO_START}
113
- python fetch_dataset_s2s.py
114
- done
115
- for i in $(seq 500 514);
116
- do
117
- export N_POOL=15
118
- export DATASET_ID=${i}
119
- export DIRECTION="enA-esA"
120
- export LINE_NO_START=$(((DATASET_ID-1) * 2500))
121
- export LINE_NO_END=$((DATASET_ID * 2500))
122
- echo ${LINE_NO_START}
123
- python fetch_dataset_s2s.py
124
- done
125
-
126
- for i in $(seq 515 532);
127
  do
128
  export N_POOL=15
129
  export DATASET_ID=${i}
@@ -134,7 +102,7 @@ do
134
  python fetch_dataset_s2s.py
135
  done
136
  # tokenize
137
- for i in $(seq 51 210);
138
  do
139
  export N_POOL=15
140
  export DATASET_ID=${i}
@@ -145,27 +113,17 @@ do
145
  python tokenize_dataset_s2s.py
146
  done
147
  # speaker embedding
148
- for i in $(seq 251 280);
149
- do
150
- export DATASET_ID=${i}
151
- export DIRECTION="enA-esA"
152
- python attach_speaker_embedding_s2s.py
153
- done
154
-
155
- export SE_MODEL="w2vbert-600m" # running
156
- export SE_MODEL="xlsr-2b" # running
157
- export SE_MODEL="hubert-xl"
158
  export SE_MODEL="metavoice"
159
- for i in $(seq 26 125);
 
 
 
160
  do
161
  export DATASET_ID=${i}
162
  export DIRECTION="enA-esA"
163
  python attach_speaker_embedding_s2s.py
164
  done
165
 
166
-
167
-
168
-
169
  ######################
170
  # enA-deA: 1_965_186 #
171
  ######################
@@ -187,7 +145,7 @@ do
187
  python fetch_dataset_s2s.py
188
  done
189
  # tokenize
190
- for i in $(seq 321 353);
191
  do
192
  export N_POOL=15
193
  export DATASET_ID=${i}
@@ -197,122 +155,12 @@ do
197
  echo ${LINE_NO_START}
198
  python tokenize_dataset_s2s.py
199
  done
200
-
201
  # speaker embedding
202
- for i in $(seq 28 104);
203
- do
204
- export DATASET_ID=${i}
205
- export DIRECTION="deA-enA"
206
- python attach_speaker_embedding_s2s.py
207
- done
208
- # speaker embedding
209
-
210
-
211
- # running on kotoba_2
212
- #TODO1
213
- export SE_MODEL="xlsr-2b"
214
- for i in 1 2 4 5 6 10 41 50 55 340 377 379;
215
- do
216
- export DATASET_ID=${i}
217
- export DIRECTION="enA-frA"
218
- python attach_speaker_embedding_s2s.py
219
- done
220
- for i in $(seq 73 299);
221
- do
222
- export DATASET_ID=${i}
223
- export DIRECTION="enA-frA"
224
- python attach_speaker_embedding_s2s.py
225
- done
226
- for i in $(seq 381 600);
227
- do
228
- export DATASET_ID=${i}
229
- export DIRECTION="enA-frA"
230
- python attach_speaker_embedding_s2s.py
231
- done
232
- #TODO2
233
- export SE_MODEL="hubert-xl"
234
- for i in 318 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361;
235
- do
236
- export DATASET_ID=${i}
237
- export DIRECTION="enA-esA"
238
- python attach_speaker_embedding_s2s.py
239
- done
240
- #TODO3
241
- export SE_MODEL="w2vbert-600m"
242
- for i in $(seq 40 600);
243
- do
244
- export DATASET_ID=${i}
245
- export DIRECTION="enA-frA"
246
- python attach_speaker_embedding_s2s.py
247
- done
248
- #TODO4
249
- export SE_MODEL="w2vbert-600m"
250
- for i in $(seq 467 470);
251
- do
252
- export DATASET_ID=${i}
253
- export DIRECTION="enA-esA"
254
- python attach_speaker_embedding_s2s.py
255
- done
256
- export SE_MODEL="xlsr-2b"
257
- for i in $(seq 453 470);
258
- do
259
- export DATASET_ID=${i}
260
- export DIRECTION="enA-esA"
261
- python attach_speaker_embedding_s2s.py
262
- done
263
  export SE_MODEL="metavoice"
264
- for i in $(seq 453 459);
265
- do
266
- export DATASET_ID=${i}
267
- export DIRECTION="enA-esA"
268
- python attach_speaker_embedding_s2s.py
269
- done
270
-
271
-
272
- # running on kotoba_1
273
- #TODO1
274
- export SE_MODEL="hubert-xl"
275
- for i in $(seq 1 394);
276
- do
277
- export DATASET_ID=${i}
278
- export DIRECTION="deA-enA"
279
- python attach_speaker_embedding_s2s.py
280
- done
281
- #TODO2
282
  export SE_MODEL="hubert-xl"
283
- for i in $(seq 38 600);
284
- do
285
- export DATASET_ID=${i}
286
- export DIRECTION="enA-frA"
287
- python attach_speaker_embedding_s2s.py
288
- done
289
- #TODO3
290
- export SE_MODEL="metavoice"
291
- for i in $(seq 50 299);
292
- do
293
- export DATASET_ID=${i}
294
- export DIRECTION="enA-frA"
295
- python attach_speaker_embedding_s2s.py
296
- done
297
- for i in $(seq 331 398);
298
- do
299
- export DATASET_ID=${i}
300
- export DIRECTION="enA-frA"
301
- python attach_speaker_embedding_s2s.py
302
- done
303
- for i in $(seq 426 600);
304
- do
305
- export DATASET_ID=${i}
306
- export DIRECTION="enA-frA"
307
- python attach_speaker_embedding_s2s.py
308
- done
309
- #TODO4
310
  export SE_MODEL="xlsr-2b"
311
- export DATASET_ID=213
312
- export DIRECTION="deA-enA"
313
- python attach_speaker_embedding_s2s.py
314
- export SE_MODEL="hubert-xl"
315
- for i in 117 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284;
316
  do
317
  export DATASET_ID=${i}
318
  export DIRECTION="deA-enA"
@@ -320,16 +168,6 @@ do
320
  done
321
 
322
 
323
-
324
-
325
-
326
-
327
-
328
-
329
-
330
-
331
-
332
-
333
  ####################
334
  # enA-koA: 511_358 #
335
  ####################
@@ -363,9 +201,9 @@ do
363
  done
364
  # speaker embedding
365
  export SE_MODEL="metavoice"
366
- export SE_MODEL="w2vbert-600m"
367
  export SE_MODEL="hubert-xl"
368
  export SE_MODEL="xlsr-2b"
 
369
  for i in $(seq 1 100);
370
  do
371
  export DATASET_ID=${i}
@@ -406,9 +244,9 @@ do
406
  done
407
  # speaker embedding
408
  export SE_MODEL="metavoice"
409
- export SE_MODEL="w2vbert-600m"
410
  export SE_MODEL="hubert-xl"
411
  export SE_MODEL="xlsr-2b"
 
412
  for i in $(seq 1 258);
413
  do
414
  export DATASET_ID=${i}
@@ -445,9 +283,9 @@ do
445
  done
446
  # speaker embedding
447
  export SE_MODEL="metavoice"
448
- export SE_MODEL="w2vbert-600m"
449
  export SE_MODEL="hubert-xl"
450
  export SE_MODEL="xlsr-2b"
 
451
  for i in $(seq 1 144);
452
  do
453
  export DATASET_ID=${i}
@@ -484,9 +322,9 @@ do
484
  done
485
  # speaker embedding
486
  export SE_MODEL="metavoice"
487
- export SE_MODEL="w2vbert-600m"
488
  export SE_MODEL="hubert-xl"
489
  export SE_MODEL="xlsr-2b"
 
490
  for i in $(seq 1 149);
491
  do
492
  export DATASET_ID=${i}
@@ -523,9 +361,9 @@ do
523
  done
524
  # speaker embedding
525
  export SE_MODEL="metavoice"
526
- export SE_MODEL="w2vbert-600m"
527
  export SE_MODEL="hubert-xl"
528
  export SE_MODEL="xlsr-2b"
 
529
  for i in $(seq 1 91);
530
  do
531
  export DATASET_ID=${i}
 
1
+ export SE_MODEL="xlsr-2b"
2
+ for i in 517 518 519 520;
3
+ do
4
+ export DATASET_ID=${i}
5
+ export DIRECTION="enA-esA"
6
+ python attach_speaker_embedding_s2s.py
7
+ done
8
+ export SE_MODEL="xlsr-2b"
9
+ for i in 507 508 515 516;
10
+ do
11
+ export DATASET_ID=${i}
12
+ export DIRECTION="enA-esA"
13
+ python attach_speaker_embedding_s2s.py
14
+ done
15
+ export SE_MODEL="xlsr-2b"
16
+ for i in 505 506 507;
17
+ do
18
+ export DATASET_ID=${i}
19
+ export DIRECTION="enA-esA"
20
+ python attach_speaker_embedding_s2s.py
21
+ done
22
+ export SE_MODEL="xlsr-2b"
23
+ for i in 500 502 504;
24
+ do
25
+ export DATASET_ID=${i}
26
+ export DIRECTION="enA-esA"
27
+ python attach_speaker_embedding_s2s.py
28
+ done
29
+
30
+
31
+ export N_POOL=15
32
+ export DATASET_ID=${i}
33
+ export DIRECTION="enA-jaA"
34
+ export LINE_NO_START=$(((DATASET_ID-1) * 2500))
35
+ export LINE_NO_END=$((DATASET_ID * 2500))
36
+ echo ${LINE_NO_START}
37
+ python tokenize_dataset_s2s.py
38
+
39
+
40
  ######################
41
  # enA-frA: 3_054_258 #
42
  ######################
 
47
  export LINE_NO_END=10
48
  python fetch_dataset_s2s.py
49
  # main
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
50
  for i in $(seq 1 600);
51
  do
52
  export N_POOL=15
 
91
  export LINE_NO_END=10
92
  python fetch_dataset_s2s.py
93
  # main
94
+ for i in $(seq 1 532);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
95
  do
96
  export N_POOL=15
97
  export DATASET_ID=${i}
 
102
  python fetch_dataset_s2s.py
103
  done
104
  # tokenize
105
+ for i in $(seq 1 532);
106
  do
107
  export N_POOL=15
108
  export DATASET_ID=${i}
 
113
  python tokenize_dataset_s2s.py
114
  done
115
  # speaker embedding
 
 
 
 
 
 
 
 
 
 
116
  export SE_MODEL="metavoice"
117
+ export SE_MODEL="hubert-xl"
118
+ export SE_MODEL="xlsr-2b"
119
+ export SE_MODEL="w2vbert-600m"
120
+ for i in $(seq 1 532);
121
  do
122
  export DATASET_ID=${i}
123
  export DIRECTION="enA-esA"
124
  python attach_speaker_embedding_s2s.py
125
  done
126
 
 
 
 
127
  ######################
128
  # enA-deA: 1_965_186 #
129
  ######################
 
145
  python fetch_dataset_s2s.py
146
  done
147
  # tokenize
148
+ for i in $(seq 1 394);
149
  do
150
  export N_POOL=15
151
  export DATASET_ID=${i}
 
155
  echo ${LINE_NO_START}
156
  python tokenize_dataset_s2s.py
157
  done
 
158
  # speaker embedding
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
159
  export SE_MODEL="metavoice"
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
160
  export SE_MODEL="hubert-xl"
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
161
  export SE_MODEL="xlsr-2b"
162
+ export SE_MODEL="w2vbert-600m"
163
+ for i in $(seq 1 394);
 
 
 
164
  do
165
  export DATASET_ID=${i}
166
  export DIRECTION="deA-enA"
 
168
  done
169
 
170
 
 
 
 
 
 
 
 
 
 
 
171
  ####################
172
  # enA-koA: 511_358 #
173
  ####################
 
201
  done
202
  # speaker embedding
203
  export SE_MODEL="metavoice"
 
204
  export SE_MODEL="hubert-xl"
205
  export SE_MODEL="xlsr-2b"
206
+ export SE_MODEL="w2vbert-600m"
207
  for i in $(seq 1 100);
208
  do
209
  export DATASET_ID=${i}
 
244
  done
245
  # speaker embedding
246
  export SE_MODEL="metavoice"
 
247
  export SE_MODEL="hubert-xl"
248
  export SE_MODEL="xlsr-2b"
249
+ export SE_MODEL="w2vbert-600m"
250
  for i in $(seq 1 258);
251
  do
252
  export DATASET_ID=${i}
 
283
  done
284
  # speaker embedding
285
  export SE_MODEL="metavoice"
 
286
  export SE_MODEL="hubert-xl"
287
  export SE_MODEL="xlsr-2b"
288
+ export SE_MODEL="w2vbert-600m"
289
  for i in $(seq 1 144);
290
  do
291
  export DATASET_ID=${i}
 
322
  done
323
  # speaker embedding
324
  export SE_MODEL="metavoice"
 
325
  export SE_MODEL="hubert-xl"
326
  export SE_MODEL="xlsr-2b"
327
+ export SE_MODEL="w2vbert-600m"
328
  for i in $(seq 1 149);
329
  do
330
  export DATASET_ID=${i}
 
361
  done
362
  # speaker embedding
363
  export SE_MODEL="metavoice"
 
364
  export SE_MODEL="hubert-xl"
365
  export SE_MODEL="xlsr-2b"
366
+ export SE_MODEL="w2vbert-600m"
367
  for i in $(seq 1 91);
368
  do
369
  export DATASET_ID=${i}
tokenize_dataset_s2s.py CHANGED
@@ -60,8 +60,7 @@ dataset = dataset.map(
60
  num_proc=num_proc,
61
  desc="tokenize dataset"
62
  )
63
- DatasetDict({"train": dataset}).push_to_hub(f"{hf_org}/{hf_dataset}.tokenized", config_name=f"subset_{dataset_id}")
64
- # DatasetDict({"train": dataset}).push_to_hub(f"{hf_org}/{hf_dataset}.tokenized.encodec", config_name=f"subset_{dataset_id}")
65
  cache_dir = f"{expanduser('~')}/.cache/huggingface/datasets/{hf_org}___{hf_dataset}/subset_{dataset_id}"
66
  if os.path.exists(cache_dir):
67
  shutil.rmtree(cache_dir)
 
60
  num_proc=num_proc,
61
  desc="tokenize dataset"
62
  )
63
+ DatasetDict({"train": dataset}).push_to_hub(f"{hf_org}/{hf_dataset}.tokenized.encodec", config_name=f"subset_{dataset_id}")
 
64
  cache_dir = f"{expanduser('~')}/.cache/huggingface/datasets/{hf_org}___{hf_dataset}/subset_{dataset_id}"
65
  if os.path.exists(cache_dir):
66
  shutil.rmtree(cache_dir)