Hugging Face's logo

AbstractPhil
/
beeper-rose-v2

Model card Files
xet
Community
AbstractPhil commited on
Commit
34f6fa1
·
verified ·
1 Parent(s): e17ee1d

Upload folder using huggingface_hub

Browse files
Files changed (6) hide show
  1. beeper_final.pt +3 -0
  2. beeper_final.safetensors +3 -0
  3. beeper_rose_final.pt +3 -0
  4. beeper_rose_final.safetensors +3 -0
  5. config.json +990 -0
  6. tokenizer.json +0 -0
beeper_final.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6433188d0356f329f4380f1eca132c76cd057367e039e2860280917945253da0
3
+ size 116810550
beeper_final.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ccbf844a23d53739cbbe79713fd7e3f50a7e34b73a0af2cd76aed6e77529f326
3
+ size 116795584
beeper_rose_final.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ee0f8495abba954d7125d2cb282d32b18465a3fb5d66550716b6e16ed52aa900
3
+ size 111301146
beeper_rose_final.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3b8b682beba4d8e3a7ad2a4977bc429c7f83120fd2707879d85cd704030c68a2
3
+ size 111286288
config.json ADDED
@@ -0,0 +1,990 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "name": "Rose-Beeper",
3
+ "context": 512,
4
+ "vocab_size": 8192,
5
+ "dim": 512,
6
+ "n_layers": 6,
7
+ "n_heads": 8,
8
+ "mlp_ratio": 4.0,
9
+ "dropout": 0.0,
10
+ "resid_dropout": 0.1,
11
+ "grad_checkpoint": false,
12
+ "compile_model": false,
13
+ "tokenizer_path": "beeper.tokenizer.json",
14
+ "add_bos_eos": true,
15
+ "span_corrupt_frac": 0.0,
16
+ "val_ratio": 0.01,
17
+ "test_ratio": 0.01,
18
+ "max_rows_per_dataset": null,
19
+ "dataset_cache_verbose": true,
20
+ "batch_size": 32,
21
+ "grad_accum_steps": 1,
22
+ "epochs": 3,
23
+ "lr": 0.0003,
24
+ "betas": [
25
+ 0.9,
26
+ 0.95
27
+ ],
28
+ "weight_decay": 0.1,
29
+ "warmup_steps": 500,
30
+ "max_steps": null,
31
+ "clip_grad": 1.0,
32
+ "min_lr": 1e-06,
33
+ "label_smoothing": 0.0,
34
+ "mixed_precision": "bf16",
35
+ "log_dir": "./runs/rose_beeper",
36
+ "log_interval": 50,
37
+ "ckpt_dir": "./beeper_checkpoints",
38
+ "export_dir": "./beeper_export",
39
+ "temperature": 0.9,
40
+ "top_k": 40,
41
+ "top_p": 0.9,
42
+ "repetition_penalty": 1.1,
43
+ "presence_penalty": 0.6,
44
+ "frequency_penalty": 0.0,
45
+ "hf_repo": "AbstractPhil/beeper-rose-v2",
46
+ "upload_to_hub": true,
47
+ "pent_level": "medium",
48
+ "lambda_contrast": 0.25,
49
+ "pent_min_edge": 0.5,
50
+ "pent_temp": 0.1,
51
+ "contrast_warmup": 800,
52
+ "lambda_rose": 0.1,
53
+ "rose_scale": 1.8,
54
+ "lambda_geom_sep": 0.5,
55
+ "geom_sep_margin": 0.9,
56
+ "lambda_geom": 0.3,
57
+ "lambda_geom_angle": 0.8,
58
+ "lambda_geom_var": 0.3,
59
+ "lambda_geom_edge": 0.3,
60
+ "lambda_geom_vol": 0.6,
61
+ "lambda_geom_minrel": 1.0,
62
+ "geom_min_edge_rel": 0.6,
63
+ "geom_vol_lower_frac": 0.85,
64
+ "geom_sample_classes": 64,
65
+ "geom_sample_k": 64,
66
+ "seed": 1337,
67
+ "corpus": [
68
+ {
69
+ "name": "TinyStories",
70
+ "path": "roneneldan/TinyStories",
71
+ "split": "train[10%:20%]",
72
+ "weight": 0.1,
73
+ "dialect": [
74
+ 0.6,
75
+ 0.1,
76
+ 0.05,
77
+ 0.05,
78
+ 0.2
79
+ ]
80
+ },
81
+ {
82
+ "name": "WikipediaEN",
83
+ "path": "wikimedia/wikipedia",
84
+ "config": "20231101.en",
85
+ "split": "train[1%:2%]",
86
+ "weight": 0.6,
87
+ "dialect": [
88
+ 0.12,
89
+ 0.58,
90
+ 0.1,
91
+ 0.1,
92
+ 0.1
93
+ ]
94
+ },
95
+ {
96
+ "name": "AGNews",
97
+ "path": "ag_news",
98
+ "split": "train[:50%]",
99
+ "weight": 0.4,
100
+ "dialect": [
101
+ 0.2,
102
+ 0.5,
103
+ 0.1,
104
+ 0.1,
105
+ 0.1
106
+ ]
107
+ },
108
+ {
109
+ "name": "GSM8K",
110
+ "path": "openai/gsm8k",
111
+ "config": "main",
112
+ "split": "train[10%:20%]",
113
+ "weight": 0.6,
114
+ "dialect": [
115
+ 0.1,
116
+ 0.15,
117
+ 0.5,
118
+ 0.15,
119
+ 0.1
120
+ ]
121
+ },
122
+ {
123
+ "name": "AI2-ARC-Easy",
124
+ "path": "allenai/ai2_arc",
125
+ "config": "ARC-Easy",
126
+ "split": "train[10%:20%]",
127
+ "weight": 0.5,
128
+ "dialect": [
129
+ 0.05,
130
+ 0.15,
131
+ 0.4,
132
+ 0.25,
133
+ 0.15
134
+ ]
135
+ },
136
+ {
137
+ "name": "HH-RLHF",
138
+ "path": "Anthropic/hh-rlhf",
139
+ "split": "train[1%:2%]",
140
+ "weight": 0.6,
141
+ "dialect": [
142
+ 0.1,
143
+ 0.25,
144
+ 0.2,
145
+ 0.25,
146
+ 0.2
147
+ ]
148
+ },
149
+ {
150
+ "name": "SVAMP",
151
+ "path": "ChilleD/SVAMP",
152
+ "split": "train",
153
+ "weight": 0.45,
154
+ "dialect": [
155
+ 0.1,
156
+ 0.15,
157
+ 0.55,
158
+ 0.15,
159
+ 0.05
160
+ ]
161
+ },
162
+ {
163
+ "name": "MATH-500",
164
+ "path": "HuggingFaceH4/MATH-500",
165
+ "split": "test",
166
+ "weight": 0.15,
167
+ "dialect": [
168
+ 0.05,
169
+ 0.15,
170
+ 0.6,
171
+ 0.15,
172
+ 0.05
173
+ ]
174
+ },
175
+ {
176
+ "name": "SEP",
177
+ "path": "AiresPucrs/stanford-encyclopedia-philosophy",
178
+ "split": "train",
179
+ "weight": 0.3,
180
+ "dialect": [
181
+ 0.05,
182
+ 0.45,
183
+ 0.18,
184
+ 0.22,
185
+ 0.1
186
+ ]
187
+ },
188
+ {
189
+ "name": "ETHICS-commonsense",
190
+ "path": "hendrycks/ethics",
191
+ "config": "commonsense",
192
+ "split": "train",
193
+ "weight": 0.45,
194
+ "dialect": [
195
+ 0.1,
196
+ 0.3,
197
+ 0.18,
198
+ 0.24,
199
+ 0.18
200
+ ]
201
+ },
202
+ {
203
+ "name": "ETHICS-deontology",
204
+ "path": "hendrycks/ethics",
205
+ "config": "deontology",
206
+ "split": "train",
207
+ "weight": 0.35,
208
+ "dialect": [
209
+ 0.1,
210
+ 0.3,
211
+ 0.18,
212
+ 0.24,
213
+ 0.18
214
+ ]
215
+ },
216
+ {
217
+ "name": "ETHICS-justice",
218
+ "path": "hendrycks/ethics",
219
+ "config": "justice",
220
+ "split": "train",
221
+ "weight": 0.35,
222
+ "dialect": [
223
+ 0.1,
224
+ 0.3,
225
+ 0.18,
226
+ 0.24,
227
+ 0.18
228
+ ]
229
+ },
230
+ {
231
+ "name": "ETHICS-utilitarianism",
232
+ "path": "hendrycks/ethics",
233
+ "config": "utilitarianism",
234
+ "split": "train",
235
+ "weight": 0.35,
236
+ "dialect": [
237
+ 0.1,
238
+ 0.3,
239
+ 0.18,
240
+ 0.24,
241
+ 0.18
242
+ ]
243
+ },
244
+ {
245
+ "name": "ETHICS-virtue",
246
+ "path": "hendrycks/ethics",
247
+ "config": "virtue",
248
+ "split": "train",
249
+ "weight": 0.35,
250
+ "dialect": [
251
+ 0.1,
252
+ 0.3,
253
+ 0.18,
254
+ 0.24,
255
+ 0.18
256
+ ]
257
+ },
258
+ {
259
+ "name": "SocialChem101",
260
+ "path": "allenai/social-chemistry-101",
261
+ "split": "train",
262
+ "weight": 0.65,
263
+ "dialect": [
264
+ 0.15,
265
+ 0.25,
266
+ 0.2,
267
+ 0.2,
268
+ 0.2
269
+ ]
270
+ },
271
+ {
272
+ "name": "MoralStories",
273
+ "path": "demelin/moral_stories",
274
+ "split": "train",
275
+ "weight": 0.35,
276
+ "dialect": [
277
+ 0.2,
278
+ 0.2,
279
+ 0.2,
280
+ 0.2,
281
+ 0.2
282
+ ]
283
+ },
284
+ {
285
+ "name": "ART-AbductiveNLI",
286
+ "path": "allenai/art",
287
+ "split": "train",
288
+ "weight": 0.3,
289
+ "dialect": [
290
+ 0.05,
291
+ 0.2,
292
+ 0.45,
293
+ 0.2,
294
+ 0.1
295
+ ]
296
+ },
297
+ {
298
+ "name": "EntailmentBankV3",
299
+ "path": "ariesutiono/entailment-bank-v3",
300
+ "split": "train",
301
+ "weight": 0.35,
302
+ "dialect": [
303
+ 0.05,
304
+ 0.25,
305
+ 0.45,
306
+ 0.15,
307
+ 0.1
308
+ ]
309
+ },
310
+ {
311
+ "name": "LogiQA2.0NLI",
312
+ "path": "tasksource/logiqa-2.0-nli",
313
+ "split": "train",
314
+ "weight": 0.45,
315
+ "dialect": [
316
+ 0.05,
317
+ 0.25,
318
+ 0.45,
319
+ 0.15,
320
+ 0.1
321
+ ]
322
+ },
323
+ {
324
+ "name": "TruthfulQA-MC",
325
+ "path": "EleutherAI/truthful_qa_mc",
326
+ "split": "validation",
327
+ "weight": 0.25,
328
+ "dialect": [
329
+ 0.05,
330
+ 0.35,
331
+ 0.25,
332
+ 0.25,
333
+ 0.1
334
+ ]
335
+ },
336
+ {
337
+ "name": "VUA20-Metaphor",
338
+ "path": "CreativeLang/vua20_metaphor",
339
+ "split": "train[:5%]",
340
+ "weight": 0.3,
341
+ "dialect": [
342
+ 0.3,
343
+ 0.1,
344
+ 0.1,
345
+ 0.15,
346
+ 0.35
347
+ ]
348
+ }
349
+ ],
350
+ "capoera": {
351
+ "enable": true,
352
+ "topic_bins": 512,
353
+ "mood_bins": 7
354
+ },
355
+ "_ok_entries": [
356
+ {
357
+ "name": "TinyStories",
358
+ "path": "roneneldan/TinyStories",
359
+ "split": "train[10%:20%]",
360
+ "weight": 0.1,
361
+ "dialect": [
362
+ 0.6000000238418579,
363
+ 0.10000000149011612,
364
+ 0.05000000074505806,
365
+ 0.05000000074505806,
366
+ 0.20000000298023224
367
+ ],
368
+ "class_id": 0,
369
+ "p": 0.012195121951219514
370
+ },
371
+ {
372
+ "name": "WikipediaEN",
373
+ "path": "wikimedia/wikipedia",
374
+ "config": "20231101.en",
375
+ "split": "train[1%:2%]",
376
+ "weight": 0.6,
377
+ "dialect": [
378
+ 0.11999999731779099,
379
+ 0.5799999833106995,
380
+ 0.10000000149011612,
381
+ 0.10000000149011612,
382
+ 0.10000000149011612
383
+ ],
384
+ "class_id": 1,
385
+ "p": 0.07317073170731708
386
+ },
387
+ {
388
+ "name": "AGNews",
389
+ "path": "ag_news",
390
+ "split": "train[:50%]",
391
+ "weight": 0.4,
392
+ "dialect": [
393
+ 0.20000000298023224,
394
+ 0.5,
395
+ 0.10000000149011612,
396
+ 0.10000000149011612,
397
+ 0.10000000149011612
398
+ ],
399
+ "class_id": 2,
400
+ "p": 0.04878048780487806
401
+ },
402
+ {
403
+ "name": "GSM8K",
404
+ "path": "openai/gsm8k",
405
+ "config": "main",
406
+ "split": "train[10%:20%]",
407
+ "weight": 0.6,
408
+ "dialect": [
409
+ 0.10000000149011612,
410
+ 0.15000000596046448,
411
+ 0.5,
412
+ 0.15000000596046448,
413
+ 0.10000000149011612
414
+ ],
415
+ "class_id": 3,
416
+ "p": 0.07317073170731708
417
+ },
418
+ {
419
+ "name": "AI2-ARC-Easy",
420
+ "path": "allenai/ai2_arc",
421
+ "config": "ARC-Easy",
422
+ "split": "train[10%:20%]",
423
+ "weight": 0.5,
424
+ "dialect": [
425
+ 0.05000000074505806,
426
+ 0.15000000596046448,
427
+ 0.4000000059604645,
428
+ 0.25,
429
+ 0.15000000596046448
430
+ ],
431
+ "class_id": 4,
432
+ "p": 0.06097560975609757
433
+ },
434
+ {
435
+ "name": "HH-RLHF",
436
+ "path": "Anthropic/hh-rlhf",
437
+ "split": "train[1%:2%]",
438
+ "weight": 0.6,
439
+ "dialect": [
440
+ 0.10000000149011612,
441
+ 0.25,
442
+ 0.20000000298023224,
443
+ 0.25,
444
+ 0.20000000298023224
445
+ ],
446
+ "class_id": 5,
447
+ "p": 0.07317073170731708
448
+ },
449
+ {
450
+ "name": "SVAMP",
451
+ "path": "ChilleD/SVAMP",
452
+ "split": "train",
453
+ "weight": 0.45,
454
+ "dialect": [
455
+ 0.10000000149011612,
456
+ 0.15000000596046448,
457
+ 0.550000011920929,
458
+ 0.15000000596046448,
459
+ 0.05000000074505806
460
+ ],
461
+ "class_id": 6,
462
+ "p": 0.05487804878048781
463
+ },
464
+ {
465
+ "name": "MATH-500",
466
+ "path": "HuggingFaceH4/MATH-500",
467
+ "split": "test",
468
+ "weight": 0.15,
469
+ "dialect": [
470
+ 0.05000000074505806,
471
+ 0.15000000596046448,
472
+ 0.6000000238418579,
473
+ 0.15000000596046448,
474
+ 0.05000000074505806
475
+ ],
476
+ "class_id": 7,
477
+ "p": 0.01829268292682927
478
+ },
479
+ {
480
+ "name": "SEP",
481
+ "path": "AiresPucrs/stanford-encyclopedia-philosophy",
482
+ "split": "train",
483
+ "weight": 0.3,
484
+ "dialect": [
485
+ 0.05000000074505806,
486
+ 0.44999998807907104,
487
+ 0.18000000715255737,
488
+ 0.2199999988079071,
489
+ 0.10000000149011612
490
+ ],
491
+ "class_id": 8,
492
+ "p": 0.03658536585365854
493
+ },
494
+ {
495
+ "name": "ETHICS-commonsense",
496
+ "path": "hendrycks/ethics",
497
+ "config": "commonsense",
498
+ "split": "train",
499
+ "weight": 0.45,
500
+ "dialect": [
501
+ 0.10000000149011612,
502
+ 0.30000001192092896,
503
+ 0.18000000715255737,
504
+ 0.23999999463558197,
505
+ 0.18000000715255737
506
+ ],
507
+ "class_id": 9,
508
+ "p": 0.05487804878048781
509
+ },
510
+ {
511
+ "name": "ETHICS-deontology",
512
+ "path": "hendrycks/ethics",
513
+ "config": "deontology",
514
+ "split": "train",
515
+ "weight": 0.35,
516
+ "dialect": [
517
+ 0.10000000149011612,
518
+ 0.30000001192092896,
519
+ 0.18000000715255737,
520
+ 0.23999999463558197,
521
+ 0.18000000715255737
522
+ ],
523
+ "class_id": 10,
524
+ "p": 0.042682926829268296
525
+ },
526
+ {
527
+ "name": "ETHICS-justice",
528
+ "path": "hendrycks/ethics",
529
+ "config": "justice",
530
+ "split": "train",
531
+ "weight": 0.35,
532
+ "dialect": [
533
+ 0.10000000149011612,
534
+ 0.30000001192092896,
535
+ 0.18000000715255737,
536
+ 0.23999999463558197,
537
+ 0.18000000715255737
538
+ ],
539
+ "class_id": 11,
540
+ "p": 0.042682926829268296
541
+ },
542
+ {
543
+ "name": "ETHICS-utilitarianism",
544
+ "path": "hendrycks/ethics",
545
+ "config": "utilitarianism",
546
+ "split": "train",
547
+ "weight": 0.35,
548
+ "dialect": [
549
+ 0.10000000149011612,
550
+ 0.30000001192092896,
551
+ 0.18000000715255737,
552
+ 0.23999999463558197,
553
+ 0.18000000715255737
554
+ ],
555
+ "class_id": 12,
556
+ "p": 0.042682926829268296
557
+ },
558
+ {
559
+ "name": "ETHICS-virtue",
560
+ "path": "hendrycks/ethics",
561
+ "config": "virtue",
562
+ "split": "train",
563
+ "weight": 0.35,
564
+ "dialect": [
565
+ 0.10000000149011612,
566
+ 0.30000001192092896,
567
+ 0.18000000715255737,
568
+ 0.23999999463558197,
569
+ 0.18000000715255737
570
+ ],
571
+ "class_id": 13,
572
+ "p": 0.042682926829268296
573
+ },
574
+ {
575
+ "name": "SocialChem101",
576
+ "path": "allenai/social-chemistry-101",
577
+ "split": "train",
578
+ "weight": 0.65,
579
+ "dialect": [
580
+ 0.15000000596046448,
581
+ 0.25,
582
+ 0.20000000298023224,
583
+ 0.20000000298023224,
584
+ 0.20000000298023224
585
+ ],
586
+ "class_id": 14,
587
+ "p": 0.07926829268292684
588
+ },
589
+ {
590
+ "name": "MoralStories",
591
+ "path": "demelin/moral_stories",
592
+ "split": "train",
593
+ "weight": 0.35,
594
+ "dialect": [
595
+ 0.20000000298023224,
596
+ 0.20000000298023224,
597
+ 0.20000000298023224,
598
+ 0.20000000298023224,
599
+ 0.20000000298023224
600
+ ],
601
+ "class_id": 15,
602
+ "p": 0.042682926829268296
603
+ },
604
+ {
605
+ "name": "ART-AbductiveNLI",
606
+ "path": "allenai/art",
607
+ "split": "train",
608
+ "weight": 0.3,
609
+ "dialect": [
610
+ 0.05000000074505806,
611
+ 0.20000000298023224,
612
+ 0.44999998807907104,
613
+ 0.20000000298023224,
614
+ 0.10000000149011612
615
+ ],
616
+ "class_id": 16,
617
+ "p": 0.03658536585365854
618
+ },
619
+ {
620
+ "name": "EntailmentBankV3",
621
+ "path": "ariesutiono/entailment-bank-v3",
622
+ "split": "train",
623
+ "weight": 0.35,
624
+ "dialect": [
625
+ 0.05000000074505806,
626
+ 0.25,
627
+ 0.44999998807907104,
628
+ 0.15000000596046448,
629
+ 0.10000000149011612
630
+ ],
631
+ "class_id": 17,
632
+ "p": 0.042682926829268296
633
+ },
634
+ {
635
+ "name": "LogiQA2.0NLI",
636
+ "path": "tasksource/logiqa-2.0-nli",
637
+ "split": "train",
638
+ "weight": 0.45,
639
+ "dialect": [
640
+ 0.05000000074505806,
641
+ 0.25,
642
+ 0.44999998807907104,
643
+ 0.15000000596046448,
644
+ 0.10000000149011612
645
+ ],
646
+ "class_id": 18,
647
+ "p": 0.05487804878048781
648
+ },
649
+ {
650
+ "name": "TruthfulQA-MC",
651
+ "path": "EleutherAI/truthful_qa_mc",
652
+ "split": "validation",
653
+ "weight": 0.25,
654
+ "dialect": [
655
+ 0.05000000074505806,
656
+ 0.3499999940395355,
657
+ 0.25,
658
+ 0.25,
659
+ 0.10000000149011612
660
+ ],
661
+ "class_id": 19,
662
+ "p": 0.030487804878048783
663
+ },
664
+ {
665
+ "name": "VUA20-Metaphor",
666
+ "path": "CreativeLang/vua20_metaphor",
667
+ "split": "train[:5%]",
668
+ "weight": 0.3,
669
+ "dialect": [
670
+ 0.30000001192092896,
671
+ 0.10000000149011612,
672
+ 0.10000000149011612,
673
+ 0.15000000596046448,
674
+ 0.3499999940395355
675
+ ],
676
+ "class_id": 20,
677
+ "p": 0.03658536585365854
678
+ }
679
+ ],
680
+ "_alive_entries": [
681
+ {
682
+ "name": "TinyStories",
683
+ "path": "roneneldan/TinyStories",
684
+ "split": "train[10%:20%]",
685
+ "weight": 0.1,
686
+ "dialect": [
687
+ 0.6000000238418579,
688
+ 0.10000000149011612,
689
+ 0.05000000074505806,
690
+ 0.05000000074505806,
691
+ 0.20000000298023224
692
+ ],
693
+ "class_id": 0,
694
+ "p": 0.012195121951219514
695
+ },
696
+ {
697
+ "name": "WikipediaEN",
698
+ "path": "wikimedia/wikipedia",
699
+ "config": "20231101.en",
700
+ "split": "train[1%:2%]",
701
+ "weight": 0.6,
702
+ "dialect": [
703
+ 0.11999999731779099,
704
+ 0.5799999833106995,
705
+ 0.10000000149011612,
706
+ 0.10000000149011612,
707
+ 0.10000000149011612
708
+ ],
709
+ "class_id": 1,
710
+ "p": 0.07317073170731708
711
+ },
712
+ {
713
+ "name": "AGNews",
714
+ "path": "ag_news",
715
+ "split": "train[:50%]",
716
+ "weight": 0.4,
717
+ "dialect": [
718
+ 0.20000000298023224,
719
+ 0.5,
720
+ 0.10000000149011612,
721
+ 0.10000000149011612,
722
+ 0.10000000149011612
723
+ ],
724
+ "class_id": 2,
725
+ "p": 0.04878048780487806
726
+ },
727
+ {
728
+ "name": "GSM8K",
729
+ "path": "openai/gsm8k",
730
+ "config": "main",
731
+ "split": "train[10%:20%]",
732
+ "weight": 0.6,
733
+ "dialect": [
734
+ 0.10000000149011612,
735
+ 0.15000000596046448,
736
+ 0.5,
737
+ 0.15000000596046448,
738
+ 0.10000000149011612
739
+ ],
740
+ "class_id": 3,
741
+ "p": 0.07317073170731708
742
+ },
743
+ {
744
+ "name": "AI2-ARC-Easy",
745
+ "path": "allenai/ai2_arc",
746
+ "config": "ARC-Easy",
747
+ "split": "train[10%:20%]",
748
+ "weight": 0.5,
749
+ "dialect": [
750
+ 0.05000000074505806,
751
+ 0.15000000596046448,
752
+ 0.4000000059604645,
753
+ 0.25,
754
+ 0.15000000596046448
755
+ ],
756
+ "class_id": 4,
757
+ "p": 0.06097560975609757
758
+ },
759
+ {
760
+ "name": "HH-RLHF",
761
+ "path": "Anthropic/hh-rlhf",
762
+ "split": "train[1%:2%]",
763
+ "weight": 0.6,
764
+ "dialect": [
765
+ 0.10000000149011612,
766
+ 0.25,
767
+ 0.20000000298023224,
768
+ 0.25,
769
+ 0.20000000298023224
770
+ ],
771
+ "class_id": 5,
772
+ "p": 0.07317073170731708
773
+ },
774
+ {
775
+ "name": "SVAMP",
776
+ "path": "ChilleD/SVAMP",
777
+ "split": "train",
778
+ "weight": 0.45,
779
+ "dialect": [
780
+ 0.10000000149011612,
781
+ 0.15000000596046448,
782
+ 0.550000011920929,
783
+ 0.15000000596046448,
784
+ 0.05000000074505806
785
+ ],
786
+ "class_id": 6,
787
+ "p": 0.05487804878048781
788
+ },
789
+ {
790
+ "name": "MATH-500",
791
+ "path": "HuggingFaceH4/MATH-500",
792
+ "split": "test",
793
+ "weight": 0.15,
794
+ "dialect": [
795
+ 0.05000000074505806,
796
+ 0.15000000596046448,
797
+ 0.6000000238418579,
798
+ 0.15000000596046448,
799
+ 0.05000000074505806
800
+ ],
801
+ "class_id": 7,
802
+ "p": 0.01829268292682927
803
+ },
804
+ {
805
+ "name": "SEP",
806
+ "path": "AiresPucrs/stanford-encyclopedia-philosophy",
807
+ "split": "train",
808
+ "weight": 0.3,
809
+ "dialect": [
810
+ 0.05000000074505806,
811
+ 0.44999998807907104,
812
+ 0.18000000715255737,
813
+ 0.2199999988079071,
814
+ 0.10000000149011612
815
+ ],
816
+ "class_id": 8,
817
+ "p": 0.03658536585365854
818
+ },
819
+ {
820
+ "name": "ETHICS-commonsense",
821
+ "path": "hendrycks/ethics",
822
+ "config": "commonsense",
823
+ "split": "train",
824
+ "weight": 0.45,
825
+ "dialect": [
826
+ 0.10000000149011612,
827
+ 0.30000001192092896,
828
+ 0.18000000715255737,
829
+ 0.23999999463558197,
830
+ 0.18000000715255737
831
+ ],
832
+ "class_id": 9,
833
+ "p": 0.05487804878048781
834
+ },
835
+ {
836
+ "name": "ETHICS-deontology",
837
+ "path": "hendrycks/ethics",
838
+ "config": "deontology",
839
+ "split": "train",
840
+ "weight": 0.35,
841
+ "dialect": [
842
+ 0.10000000149011612,
843
+ 0.30000001192092896,
844
+ 0.18000000715255737,
845
+ 0.23999999463558197,
846
+ 0.18000000715255737
847
+ ],
848
+ "class_id": 10,
849
+ "p": 0.042682926829268296
850
+ },
851
+ {
852
+ "name": "ETHICS-justice",
853
+ "path": "hendrycks/ethics",
854
+ "config": "justice",
855
+ "split": "train",
856
+ "weight": 0.35,
857
+ "dialect": [
858
+ 0.10000000149011612,
859
+ 0.30000001192092896,
860
+ 0.18000000715255737,
861
+ 0.23999999463558197,
862
+ 0.18000000715255737
863
+ ],
864
+ "class_id": 11,
865
+ "p": 0.042682926829268296
866
+ },
867
+ {
868
+ "name": "ETHICS-utilitarianism",
869
+ "path": "hendrycks/ethics",
870
+ "config": "utilitarianism",
871
+ "split": "train",
872
+ "weight": 0.35,
873
+ "dialect": [
874
+ 0.10000000149011612,
875
+ 0.30000001192092896,
876
+ 0.18000000715255737,
877
+ 0.23999999463558197,
878
+ 0.18000000715255737
879
+ ],
880
+ "class_id": 12,
881
+ "p": 0.042682926829268296
882
+ },
883
+ {
884
+ "name": "ETHICS-virtue",
885
+ "path": "hendrycks/ethics",
886
+ "config": "virtue",
887
+ "split": "train",
888
+ "weight": 0.35,
889
+ "dialect": [
890
+ 0.10000000149011612,
891
+ 0.30000001192092896,
892
+ 0.18000000715255737,
893
+ 0.23999999463558197,
894
+ 0.18000000715255737
895
+ ],
896
+ "class_id": 13,
897
+ "p": 0.042682926829268296
898
+ },
899
+ {
900
+ "name": "SocialChem101",
901
+ "path": "allenai/social-chemistry-101",
902
+ "split": "train",
903
+ "weight": 0.65,
904
+ "dialect": [
905
+ 0.15000000596046448,
906
+ 0.25,
907
+ 0.20000000298023224,
908
+ 0.20000000298023224,
909
+ 0.20000000298023224
910
+ ],
911
+ "class_id": 14,
912
+ "p": 0.07926829268292684
913
+ },
914
+ {
915
+ "name": "MoralStories",
916
+ "path": "demelin/moral_stories",
917
+ "split": "train",
918
+ "weight": 0.35,
919
+ "dialect": [
920
+ 0.20000000298023224,
921
+ 0.20000000298023224,
922
+ 0.20000000298023224,
923
+ 0.20000000298023224,
924
+ 0.20000000298023224
925
+ ],
926
+ "class_id": 15,
927
+ "p": 0.042682926829268296
928
+ },
929
+ {
930
+ "name": "ART-AbductiveNLI",
931
+ "path": "allenai/art",
932
+ "split": "train",
933
+ "weight": 0.3,
934
+ "dialect": [
935
+ 0.05000000074505806,
936
+ 0.20000000298023224,
937
+ 0.44999998807907104,
938
+ 0.20000000298023224,
939
+ 0.10000000149011612
940
+ ],
941
+ "class_id": 16,
942
+ "p": 0.03658536585365854
943
+ },
944
+ {
945
+ "name": "LogiQA2.0NLI",
946
+ "path": "tasksource/logiqa-2.0-nli",
947
+ "split": "train",
948
+ "weight": 0.45,
949
+ "dialect": [
950
+ 0.05000000074505806,
951
+ 0.25,
952
+ 0.44999998807907104,
953
+ 0.15000000596046448,
954
+ 0.10000000149011612
955
+ ],
956
+ "class_id": 18,
957
+ "p": 0.05487804878048781
958
+ },
959
+ {
960
+ "name": "TruthfulQA-MC",
961
+ "path": "EleutherAI/truthful_qa_mc",
962
+ "split": "validation",
963
+ "weight": 0.25,
964
+ "dialect": [
965
+ 0.05000000074505806,
966
+ 0.3499999940395355,
967
+ 0.25,
968
+ 0.25,
969
+ 0.10000000149011612
970
+ ],
971
+ "class_id": 19,
972
+ "p": 0.030487804878048783
973
+ },
974
+ {
975
+ "name": "VUA20-Metaphor",
976
+ "path": "CreativeLang/vua20_metaphor",
977
+ "split": "train[:5%]",
978
+ "weight": 0.3,
979
+ "dialect": [
980
+ 0.30000001192092896,
981
+ 0.10000000149011612,
982
+ 0.10000000149011612,
983
+ 0.15000000596046448,
984
+ 0.3499999940395355
985
+ ],
986
+ "class_id": 20,
987
+ "p": 0.03658536585365854
988
+ }
989
+ ]
990
+ }
tokenizer.json ADDED
The diff for this file is too large to render. See raw diff