Hugging Face's logo

Datasets:
SpeechPPL
/
SALMon_pGSLM-normalized

Modalities:
Audio
Text
Time-series
Formats:
parquet
Size:
1K - 10K
Libraries:
Datasets
pandas
Polars
Dataset card Data Studio Files
xet
 Community
Splend1dchan commited on
Commit
779bd11
·
verified ·
1 Parent(s): 83f4076

Upload README.md for normalized SALMon repo SpeechPPL/SALMon_pGSLM

Browse files
Files changed (1) hide show
  1. README.md +52 -76
README.md CHANGED
@@ -1,16 +1,17 @@
1
  ---
2
  dataset_info:
3
- - features:
 
4
  - name: task
5
  dtype: string
6
  - name: ind
7
  dtype: int64
8
  - name: positive_sample_tokenwise_loss
9
- dtype: '[Value(dtype=''float32'', id=None)]'
10
  - name: negative_sample_tokenwise_loss
11
- dtype: '[Value(dtype=''float32'', id=None)]'
12
  - name: prompt_sample_tokenwise_loss
13
- dtype: '[Value(dtype=''float32'', id=None)]'
14
  - name: model_generated_continuation
15
  dtype:
16
  audio:
@@ -67,23 +68,20 @@ dataset_info:
67
  - name: train
68
  num_bytes: 87019899
69
  num_examples: 200
70
- num_shards: 1
71
- shards:
72
- - train-00000-of-00001.parquet
73
  download_size: 87019899
74
  dataset_size: 87019899
75
- config_name: bg_alignment
76
- - features:
77
  - name: task
78
  dtype: string
79
  - name: ind
80
  dtype: int64
81
  - name: positive_sample_tokenwise_loss
82
- dtype: '[Value(dtype=''float32'', id=None)]'
83
  - name: negative_sample_tokenwise_loss
84
- dtype: '[Value(dtype=''float32'', id=None)]'
85
  - name: prompt_sample_tokenwise_loss
86
- dtype: '[Value(dtype=''float32'', id=None)]'
87
  - name: model_generated_continuation
88
  dtype:
89
  audio:
@@ -117,9 +115,9 @@ dataset_info:
117
  - name: ppl_sanity
118
  dtype: int64
119
  - name: positive_continuation_tokenwise_loss
120
- dtype: '[Value(dtype=''float64'', id=None)]'
121
  - name: negative_continuation_tokenwise_loss
122
- dtype: '[Value(dtype=''float64'', id=None)]'
123
  - name: positive_continuation_raw_units
124
  dtype:
125
  - name: hubert
@@ -140,23 +138,20 @@ dataset_info:
140
  - name: train
141
  num_bytes: 207143109
142
  num_examples: 200
143
- num_shards: 1
144
- shards:
145
- - train-00000-of-00001.parquet
146
  download_size: 207143109
147
  dataset_size: 207143109
148
- config_name: bg_all_consistency
149
- - features:
150
  - name: task
151
  dtype: string
152
  - name: ind
153
  dtype: int64
154
  - name: positive_sample_tokenwise_loss
155
- dtype: '[Value(dtype=''float32'', id=None)]'
156
  - name: negative_sample_tokenwise_loss
157
- dtype: '[Value(dtype=''float32'', id=None)]'
158
  - name: prompt_sample_tokenwise_loss
159
- dtype: '[Value(dtype=''float32'', id=None)]'
160
  - name: model_generated_continuation
161
  dtype:
162
  audio:
@@ -190,9 +185,9 @@ dataset_info:
190
  - name: ppl_sanity
191
  dtype: int64
192
  - name: positive_continuation_tokenwise_loss
193
- dtype: '[Value(dtype=''float64'', id=None)]'
194
  - name: negative_continuation_tokenwise_loss
195
- dtype: '[Value(dtype=''float64'', id=None)]'
196
  - name: positive_continuation_raw_units
197
  dtype:
198
  - name: hubert
@@ -213,23 +208,20 @@ dataset_info:
213
  - name: train
214
  num_bytes: 210303767
215
  num_examples: 200
216
- num_shards: 1
217
- shards:
218
- - train-00000-of-00001.parquet
219
  download_size: 210303767
220
  dataset_size: 210303767
221
- config_name: bg_domain_consistency
222
- - features:
223
  - name: task
224
  dtype: string
225
  - name: ind
226
  dtype: int64
227
  - name: positive_sample_tokenwise_loss
228
- dtype: '[Value(dtype=''float32'', id=None)]'
229
  - name: negative_sample_tokenwise_loss
230
- dtype: '[Value(dtype=''float32'', id=None)]'
231
  - name: prompt_sample_tokenwise_loss
232
- dtype: '[Value(dtype=''float32'', id=None)]'
233
  - name: model_generated_continuation
234
  dtype:
235
  audio:
@@ -263,9 +255,9 @@ dataset_info:
263
  - name: ppl_sanity
264
  dtype: int64
265
  - name: positive_continuation_tokenwise_loss
266
- dtype: '[Value(dtype=''float64'', id=None)]'
267
  - name: negative_continuation_tokenwise_loss
268
- dtype: '[Value(dtype=''float64'', id=None)]'
269
  - name: positive_continuation_raw_units
270
  dtype:
271
  - name: hubert
@@ -286,23 +278,20 @@ dataset_info:
286
  - name: train
287
  num_bytes: 209062383
288
  num_examples: 200
289
- num_shards: 1
290
- shards:
291
- - train-00000-of-00001.parquet
292
  download_size: 209062383
293
  dataset_size: 209062383
294
- config_name: gender_consistency
295
- - features:
296
  - name: task
297
  dtype: string
298
  - name: ind
299
  dtype: int64
300
  - name: positive_sample_tokenwise_loss
301
- dtype: '[Value(dtype=''float32'', id=None)]'
302
  - name: negative_sample_tokenwise_loss
303
- dtype: '[Value(dtype=''float32'', id=None)]'
304
  - name: prompt_sample_tokenwise_loss
305
- dtype: '[Value(dtype=''float32'', id=None)]'
306
  - name: model_generated_continuation
307
  dtype:
308
  audio:
@@ -336,9 +325,9 @@ dataset_info:
336
  - name: ppl_sanity
337
  dtype: int64
338
  - name: positive_continuation_tokenwise_loss
339
- dtype: '[Value(dtype=''float64'', id=None)]'
340
  - name: negative_continuation_tokenwise_loss
341
- dtype: '[Value(dtype=''float64'', id=None)]'
342
  - name: positive_continuation_raw_units
343
  dtype:
344
  - name: hubert
@@ -359,23 +348,20 @@ dataset_info:
359
  - name: train
360
  num_bytes: 202129101
361
  num_examples: 200
362
- num_shards: 1
363
- shards:
364
- - train-00000-of-00001.parquet
365
  download_size: 202129101
366
  dataset_size: 202129101
367
- config_name: rir_consistency
368
- - features:
369
  - name: task
370
  dtype: string
371
  - name: ind
372
  dtype: int64
373
  - name: positive_sample_tokenwise_loss
374
- dtype: '[Value(dtype=''float32'', id=None)]'
375
  - name: negative_sample_tokenwise_loss
376
- dtype: '[Value(dtype=''float32'', id=None)]'
377
  - name: prompt_sample_tokenwise_loss
378
- dtype: '[Value(dtype=''float32'', id=None)]'
379
  - name: model_generated_continuation
380
  dtype:
381
  audio:
@@ -432,23 +418,20 @@ dataset_info:
432
  - name: train
433
  num_bytes: 46742156
434
  num_examples: 200
435
- num_shards: 1
436
- shards:
437
- - train-00000-of-00001.parquet
438
  download_size: 46742156
439
  dataset_size: 46742156
440
- config_name: sentiment_alignment
441
- - features:
442
  - name: task
443
  dtype: string
444
  - name: ind
445
  dtype: int64
446
  - name: positive_sample_tokenwise_loss
447
- dtype: '[Value(dtype=''float32'', id=None)]'
448
  - name: negative_sample_tokenwise_loss
449
- dtype: '[Value(dtype=''float32'', id=None)]'
450
  - name: prompt_sample_tokenwise_loss
451
- dtype: '[Value(dtype=''float32'', id=None)]'
452
  - name: model_generated_continuation
453
  dtype:
454
  audio:
@@ -482,9 +465,9 @@ dataset_info:
482
  - name: ppl_sanity
483
  dtype: int64
484
  - name: positive_continuation_tokenwise_loss
485
- dtype: '[Value(dtype=''float64'', id=None)]'
486
  - name: negative_continuation_tokenwise_loss
487
- dtype: '[Value(dtype=''float64'', id=None)]'
488
  - name: positive_continuation_raw_units
489
  dtype:
490
  - name: hubert
@@ -505,23 +488,20 @@ dataset_info:
505
  - name: train
506
  num_bytes: 210124373
507
  num_examples: 200
508
- num_shards: 1
509
- shards:
510
- - train-00000-of-00001.parquet
511
  download_size: 210124373
512
  dataset_size: 210124373
513
- config_name: sentiment_consistency
514
- - features:
515
  - name: task
516
  dtype: string
517
  - name: ind
518
  dtype: int64
519
  - name: positive_sample_tokenwise_loss
520
- dtype: '[Value(dtype=''float32'', id=None)]'
521
  - name: negative_sample_tokenwise_loss
522
- dtype: '[Value(dtype=''float32'', id=None)]'
523
  - name: prompt_sample_tokenwise_loss
524
- dtype: '[Value(dtype=''float32'', id=None)]'
525
  - name: model_generated_continuation
526
  dtype:
527
  audio:
@@ -555,9 +535,9 @@ dataset_info:
555
  - name: ppl_sanity
556
  dtype: int64
557
  - name: positive_continuation_tokenwise_loss
558
- dtype: '[Value(dtype=''float64'', id=None)]'
559
  - name: negative_continuation_tokenwise_loss
560
- dtype: '[Value(dtype=''float64'', id=None)]'
561
  - name: positive_continuation_raw_units
562
  dtype:
563
  - name: hubert
@@ -578,12 +558,8 @@ dataset_info:
578
  - name: train
579
  num_bytes: 209864717
580
  num_examples: 200
581
- num_shards: 1
582
- shards:
583
- - train-00000-of-00001.parquet
584
  download_size: 209864717
585
  dataset_size: 209864717
586
- config_name: speaker_consistency
587
  ---
588
 
589
  # SALMon Normalized Dataset
 
1
  ---
2
  dataset_info:
3
+ - config_name: bg_alignment
4
+ features:
5
  - name: task
6
  dtype: string
7
  - name: ind
8
  dtype: int64
9
  - name: positive_sample_tokenwise_loss
10
+ list: float32
11
  - name: negative_sample_tokenwise_loss
12
+ list: float32
13
  - name: prompt_sample_tokenwise_loss
14
+ list: float32
15
  - name: model_generated_continuation
16
  dtype:
17
  audio:
 
68
  - name: train
69
  num_bytes: 87019899
70
  num_examples: 200
 
 
 
71
  download_size: 87019899
72
  dataset_size: 87019899
73
+ - config_name: bg_all_consistency
74
+ features:
75
  - name: task
76
  dtype: string
77
  - name: ind
78
  dtype: int64
79
  - name: positive_sample_tokenwise_loss
80
+ list: float32
81
  - name: negative_sample_tokenwise_loss
82
+ list: float32
83
  - name: prompt_sample_tokenwise_loss
84
+ list: float32
85
  - name: model_generated_continuation
86
  dtype:
87
  audio:
 
115
  - name: ppl_sanity
116
  dtype: int64
117
  - name: positive_continuation_tokenwise_loss
118
+ list: float64
119
  - name: negative_continuation_tokenwise_loss
120
+ list: float64
121
  - name: positive_continuation_raw_units
122
  dtype:
123
  - name: hubert
 
138
  - name: train
139
  num_bytes: 207143109
140
  num_examples: 200
 
 
 
141
  download_size: 207143109
142
  dataset_size: 207143109
143
+ - config_name: bg_domain_consistency
144
+ features:
145
  - name: task
146
  dtype: string
147
  - name: ind
148
  dtype: int64
149
  - name: positive_sample_tokenwise_loss
150
+ list: float32
151
  - name: negative_sample_tokenwise_loss
152
+ list: float32
153
  - name: prompt_sample_tokenwise_loss
154
+ list: float32
155
  - name: model_generated_continuation
156
  dtype:
157
  audio:
 
185
  - name: ppl_sanity
186
  dtype: int64
187
  - name: positive_continuation_tokenwise_loss
188
+ list: float64
189
  - name: negative_continuation_tokenwise_loss
190
+ list: float64
191
  - name: positive_continuation_raw_units
192
  dtype:
193
  - name: hubert
 
208
  - name: train
209
  num_bytes: 210303767
210
  num_examples: 200
 
 
 
211
  download_size: 210303767
212
  dataset_size: 210303767
213
+ - config_name: gender_consistency
214
+ features:
215
  - name: task
216
  dtype: string
217
  - name: ind
218
  dtype: int64
219
  - name: positive_sample_tokenwise_loss
220
+ list: float32
221
  - name: negative_sample_tokenwise_loss
222
+ list: float32
223
  - name: prompt_sample_tokenwise_loss
224
+ list: float32
225
  - name: model_generated_continuation
226
  dtype:
227
  audio:
 
255
  - name: ppl_sanity
256
  dtype: int64
257
  - name: positive_continuation_tokenwise_loss
258
+ list: float64
259
  - name: negative_continuation_tokenwise_loss
260
+ list: float64
261
  - name: positive_continuation_raw_units
262
  dtype:
263
  - name: hubert
 
278
  - name: train
279
  num_bytes: 209062383
280
  num_examples: 200
 
 
 
281
  download_size: 209062383
282
  dataset_size: 209062383
283
+ - config_name: rir_consistency
284
+ features:
285
  - name: task
286
  dtype: string
287
  - name: ind
288
  dtype: int64
289
  - name: positive_sample_tokenwise_loss
290
+ list: float32
291
  - name: negative_sample_tokenwise_loss
292
+ list: float32
293
  - name: prompt_sample_tokenwise_loss
294
+ list: float32
295
  - name: model_generated_continuation
296
  dtype:
297
  audio:
 
325
  - name: ppl_sanity
326
  dtype: int64
327
  - name: positive_continuation_tokenwise_loss
328
+ list: float64
329
  - name: negative_continuation_tokenwise_loss
330
+ list: float64
331
  - name: positive_continuation_raw_units
332
  dtype:
333
  - name: hubert
 
348
  - name: train
349
  num_bytes: 202129101
350
  num_examples: 200
 
 
 
351
  download_size: 202129101
352
  dataset_size: 202129101
353
+ - config_name: sentiment_alignment
354
+ features:
355
  - name: task
356
  dtype: string
357
  - name: ind
358
  dtype: int64
359
  - name: positive_sample_tokenwise_loss
360
+ list: float32
361
  - name: negative_sample_tokenwise_loss
362
+ list: float32
363
  - name: prompt_sample_tokenwise_loss
364
+ list: float32
365
  - name: model_generated_continuation
366
  dtype:
367
  audio:
 
418
  - name: train
419
  num_bytes: 46742156
420
  num_examples: 200
 
 
 
421
  download_size: 46742156
422
  dataset_size: 46742156
423
+ - config_name: sentiment_consistency
424
+ features:
425
  - name: task
426
  dtype: string
427
  - name: ind
428
  dtype: int64
429
  - name: positive_sample_tokenwise_loss
430
+ list: float32
431
  - name: negative_sample_tokenwise_loss
432
+ list: float32
433
  - name: prompt_sample_tokenwise_loss
434
+ list: float32
435
  - name: model_generated_continuation
436
  dtype:
437
  audio:
 
465
  - name: ppl_sanity
466
  dtype: int64
467
  - name: positive_continuation_tokenwise_loss
468
+ list: float64
469
  - name: negative_continuation_tokenwise_loss
470
+ list: float64
471
  - name: positive_continuation_raw_units
472
  dtype:
473
  - name: hubert
 
488
  - name: train
489
  num_bytes: 210124373
490
  num_examples: 200
 
 
 
491
  download_size: 210124373
492
  dataset_size: 210124373
493
+ - config_name: speaker_consistency
494
+ features:
495
  - name: task
496
  dtype: string
497
  - name: ind
498
  dtype: int64
499
  - name: positive_sample_tokenwise_loss
500
+ list: float32
501
  - name: negative_sample_tokenwise_loss
502
+ list: float32
503
  - name: prompt_sample_tokenwise_loss
504
+ list: float32
505
  - name: model_generated_continuation
506
  dtype:
507
  audio:
 
535
  - name: ppl_sanity
536
  dtype: int64
537
  - name: positive_continuation_tokenwise_loss
538
+ list: float64
539
  - name: negative_continuation_tokenwise_loss
540
+ list: float64
541
  - name: positive_continuation_raw_units
542
  dtype:
543
  - name: hubert
 
558
  - name: train
559
  num_bytes: 209864717
560
  num_examples: 200
 
 
 
561
  download_size: 209864717
562
  dataset_size: 209864717
 
563
  ---
564
 
565
  # SALMon Normalized Dataset