-
Notifications
You must be signed in to change notification settings - Fork 57
/
Copy patheval_configs.yaml
894 lines (894 loc) · 26.1 KB
/
eval_configs.yaml
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
# This file contains default evaluation parameters assuming access to a single A100-80GB
openbmb/UltraRM-13b:
model: 'openbmb/UltraRM-13b'
tokenizer: 'openbmb/UltraRM-13b'
chat_template: 'openbmb'
batch_size: 8
trust_remote_code: False
dpo: False
OpenAssistant/oasst-rm-2.1-pythia-1.4b-epoch-2.5:
model: 'OpenAssistant/oasst-rm-2.1-pythia-1.4b-epoch-2.5'
tokenizer: 'OpenAssistant/oasst-rm-2.1-pythia-1.4b-epoch-2.5'
chat_template: 'oasst_pythia'
batch_size: 64
trust_remote_code: False
dpo: False
OpenAssistant/oasst-rm-2-pythia-6.9b-epoch-1:
model: 'OpenAssistant/oasst-rm-2-pythia-6.9b-epoch-1'
tokenizer: 'OpenAssistant/oasst-rm-2-pythia-6.9b-epoch-1'
chat_template: 'oasst_pythia'
batch_size: 16
trust_remote_code: False
dpo: False
OpenAssistant/reward-model-deberta-v3-large-v2:
model: 'OpenAssistant/reward-model-deberta-v3-large-v2'
tokenizer: 'OpenAssistant/reward-model-deberta-v3-large-v2'
chat_template: 'raw'
batch_size: 64
trust_remote_code: False
dpo: False
weqweasdas/hh_rlhf_rm_open_llama_3b:
model: 'weqweasdas/hh_rlhf_rm_open_llama_3b'
tokenizer: 'weqweasdas/hh_rlhf_rm_open_llama_3b'
chat_template: 'Robin'
batch_size: 64
trust_remote_code: False
dpo: False
llm-blender/PairRM-hf:
model: 'llm-blender/PairRM-hf'
tokenizer: 'llm-blender/PairRM-hf'
chat_template: 'tulu'
batch_size: 64
trust_remote_code: False
dpo: False
mightbe/Better-PairRM:
model: 'mightbe/Better-PairRM'
tokenizer: 'mightbe/Better-PairRM'
chat_template: 'tulu'
batch_size: 64
max_length: 3370
trust_remote_code: False
dpo: False
berkeley-nest/Starling-RM-7B-alpha:
model: 'berkeley-nest/Starling-RM-7B-alpha'
tokenizer: 'meta-llama/Llama-2-7b-chat-hf'
chat_template: 'llama-2'
batch_size: 16
trust_remote_code: False
dpo: False
stanfordnlp/SteamSHP-flan-t5-xl:
model: 'stanfordnlp/SteamSHP-flan-t5-xl'
tokenizer: 'stanfordnlp/SteamSHP-flan-t5-xl'
chat_template: 'tulu'
batch_size: 32
trust_remote_code: False
dpo: False
stanfordnlp/SteamSHP-flan-t5-large:
model: 'stanfordnlp/SteamSHP-flan-t5-large'
tokenizer: 'stanfordnlp/SteamSHP-flan-t5-large'
chat_template: 'tulu'
batch_size: 32
trust_remote_code: False
dpo: False
PKU-Alignment/beaver-7b-v1.0-reward:
model: 'PKU-Alignment/beaver-7b-v1.0-reward'
tokenizer: 'PKU-Alignment/beaver-7b-v1.0-reward'
chat_template: 'pku-align'
batch_size: 16
trust_remote_code: False
dpo: False
PKU-Alignment/beaver-7b-v1.0-cost:
model: 'PKU-Alignment/beaver-7b-v1.0-cost'
tokenizer: 'PKU-Alignment/beaver-7b-v1.0-cost'
chat_template: 'pku-align'
batch_size: 16
trust_remote_code: False
dpo: False
IDEA-CCNL/Ziya-LLaMA-7B-Reward:
model: 'IDEA-CCNL/Ziya-LLaMA-7B-Reward'
tokenizer: 'IDEA-CCNL/Ziya-LLaMA-7B-Reward'
chat_template: 'Ziya'
batch_size: 16
trust_remote_code: True
dpo: False
Nexusflow/Starling-RM-34B:
model: 'Nexusflow/Starling-RM-34B'
tokenizer: '01-ai/Yi-34B-Chat'
chat_template: 'Yi-34b-chat'
num_gpus: 2
batch_size: 2
trust_remote_code: False
dpo: False
stabilityai/stablelm-zephyr-3b:
ref_model: stabilityai/stablelm-3b-4e1t
tokenizer: stabilityai/stablelm-zephyr-3b
chat_template:
batch_size: 12
trust_remote_code: False
dpo: True
stabilityai/stablelm-2-zephyr-1_6b:
ref_model: stabilityai/stablelm-2-1_6b
tokenizer: stabilityai/stablelm-2-zephyr-1_6b
chat_template:
batch_size: 6
trust_remote_code: True
dpo: True
HuggingFaceH4/zephyr-7b-beta:
ref_model: HuggingFaceH4/mistral-7b-sft-beta
tokenizer: HuggingFaceH4/zephyr-7b-beta
chat_template:
batch_size: 4
trust_remote_code: False
dpo: True
HuggingFaceH4/zephyr-7b-alpha:
ref_model: HuggingFaceH4/mistral-7b-sft-alpha
tokenizer: HuggingFaceH4/zephyr-7b-alpha
chat_template:
batch_size: 4
trust_remote_code: False
dpo: True
Qwen/Qwen1.5-0.5B-Chat:
ref_model: Qwen/Qwen1.5-0.5B
tokenizer: Qwen/Qwen1.5-0.5B-Chat
chat_template:
batch_size: 6
trust_remote_code: False
dpo: True
Qwen/Qwen1.5-1.8B-Chat:
ref_model: Qwen/Qwen1.5-1.8B
tokenizer: Qwen/Qwen1.5-1.8B-Chat
chat_template:
batch_size: 3
trust_remote_code: False
dpo: True
Qwen/Qwen1.5-4B-Chat:
ref_model: Qwen/Qwen1.5-4B
tokenizer: Qwen/Qwen1.5-4B-Chat
chat_template:
batch_size: 2
trust_remote_code: False
dpo: True
Qwen/Qwen1.5-7B-Chat:
ref_model: Qwen/Qwen1.5-7B
tokenizer: Qwen/Qwen1.5-7B-Chat
chat_template:
batch_size: 2
trust_remote_code: False
dpo: True
Qwen/Qwen1.5-14B-Chat:
ref_model: Qwen/Qwen1.5-14B
tokenizer: Qwen/Qwen1.5-14B-Chat
chat_template:
batch_size: 2
num_gpus: 2
trust_remote_code: False
dpo: True
Qwen/Qwen1.5-72B-Chat:
ref_model: Qwen/Qwen1.5-72B
tokenizer: Qwen/Qwen1.5-72B-Chat
chat_template:
batch_size: 1
num_gpus: 4
trust_remote_code: False
dpo: True
mistralai/Mixtral-8x7B-Instruct-v0.1:
ref_model: mistralai/Mixtral-8x7B-v0.1
tokenizer: mistralai/Mixtral-8x7B-Instruct-v0.1
chat_template:
batch_size: 1
num_gpus: 4
trust_remote_code: False
dpo: True
NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO:
ref_model: NousResearch/Nous-Hermes-2-Mixtral-8x7B-SFT
tokenizer: NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO
chat_template:
batch_size: 1
num_gpus: 4
trust_remote_code: True
dpo: True
NousResearch/Nous-Hermes-2-Mistral-7B-DPO:
ref_model: teknium/OpenHermes-2.5-Mistral-7B
tokenizer: NousResearch/Nous-Hermes-2-Mistral-7B-DPO
chat_template:
batch_size: 4
trust_remote_code: False
dpo: True
HuggingFaceH4/zephyr-7b-gemma-v0.1:
ref_model: HuggingFaceH4/zephyr-7b-gemma-sft-v0.1
tokenizer: HuggingFaceH4/zephyr-7b-gemma-v0.1
chat_template:
batch_size: 2
trust_remote_code: False
dpo: True
allenai/tulu-2-dpo-70b:
ref_model: allenai/tulu-2-70b
tokenizer: allenai/tulu-2-dpo-70b
chat_template: tulu
num_gpus: 4
batch_size: 2
trust_remote_code: False
dpo: True
allenai/tulu-2-dpo-13b:
ref_model: allenai/tulu-2-13b
tokenizer: allenai/tulu-2-dpo-13b
chat_template: tulu
num_gpus: 2
batch_size: 2
trust_remote_code: False
dpo: True
allenai/tulu-2-dpo-7b:
ref_model: allenai/tulu-2-7b
tokenizer: allenai/tulu-2-dpo-7b
chat_template: tulu
batch_size: 2
trust_remote_code: False
dpo: True
allenai/OLMo-7B-Instruct:
ref_model: allenai/OLMo-7B-SFT
tokenizer: allenai/OLMo-7B-Instruct
chat_template:
batch_size: 2
trust_remote_code: True
dpo: True
# Added March 21st 2024
weqweasdas/RM-Gemma-2B:
model: weqweasdas/RM-Gemma-2B
tokenizer: weqweasdas/RM-Gemma-2B
chat_template: # empty for tokenizer
batch_size: 16
trust_remote_code: False
dpo: False
weqweasdas/RM-Gemma-7B:
model: weqweasdas/RM-Gemma-7B
tokenizer: weqweasdas/RM-Gemma-7B
chat_template: # empty for tokenizer
batch_size: 16
trust_remote_code: False
dpo: False
weqweasdas/RM-Gemma-7B-4096:
model: weqweasdas/RM-Gemma-7B-4096
tokenizer: weqweasdas/RM-Gemma-7B-4096
chat_template: # empty for tokenizer
batch_size: 16
trust_remote_code: False
dpo: False
Ray2333/reward-model-Mistral-7B-instruct-Unified-Feedback:
model: Ray2333/reward-model-Mistral-7B-instruct-Unified-Feedback
tokenizer: Ray2333/reward-model-Mistral-7B-instruct-Unified-Feedback
chat_template: # empty for tokenizer
batch_size: 16
trust_remote_code: False
dpo: False
hendrydong/Mistral-RM-for-RAFT-GSHF-v0:
model: hendrydong/Mistral-RM-for-RAFT-GSHF-v0
tokenizer: hendrydong/Mistral-RM-for-RAFT-GSHF-v0
chat_template: # empty for tokenizer
batch_size: 16
trust_remote_code: False
dpo: False
weqweasdas/RM-Mistral-7B:
model: weqweasdas/RM-Mistral-7B
tokenizer: weqweasdas/RM-Mistral-7B
chat_template: # empty for tokenizer
batch_size: 16
trust_remote_code: False
dpo: False
# Added March 25th 2024 KTO / Archangel models follow
ContextualAI/archangel_sft-kto_llama7b:
model: ContextualAI/archangel_sft-kto_llama7b
ref_model: ContextualAI/archangel_sft_llama7b
tokenizer: ContextualAI/archangel_sft-kto_llama7b
chat_template: tulu
batch_size: 4
trust_remote_code: False
dpo: True
ContextualAI/archangel_sft-dpo_llama7b:
model: ContextualAI/archangel_sft-dpo_llama7b
ref_model: ContextualAI/archangel_sft_llama7b
tokenizer: ContextualAI/archangel_sft-dpo_llama7b
chat_template: tulu
batch_size: 4
trust_remote_code: False
dpo: True
ContextualAI/archangel_sft-kto_llama13b:
model: ContextualAI/archangel_sft-kto_llama13b
ref_model: ContextualAI/archangel_sft_llama13b
tokenizer: ContextualAI/archangel_sft-kto_llama13b
chat_template: tulu
batch_size: 2
num_gpus: 2
trust_remote_code: False
dpo: True
ContextualAI/archangel_sft-dpo_llama13b:
model: ContextualAI/archangel_sft-dpo_llama13b
ref_model: ContextualAI/archangel_sft_llama13b
tokenizer: ContextualAI/archangel_sft-dpo_llama13b
chat_template: tulu
batch_size: 2
num_gpus: 2
trust_remote_code: False
dpo: True
ContextualAI/archangel_sft-kto_llama30b:
model: ContextualAI/archangel_sft-kto_llama30b
ref_model: ContextualAI/archangel_sft_llama30b
tokenizer: ContextualAI/archangel_sft-kto_llama30b
chat_template: tulu
batch_size: 1
num_gpus: 4
trust_remote_code: False
dpo: True
ContextualAI/archangel_sft-dpo_llama30b:
model: ContextualAI/archangel_sft-dpo_llama30b
ref_model: ContextualAI/archangel_sft_llama30b
tokenizer: ContextualAI/archangel_sft-dpo_llama30b
chat_template: tulu
batch_size: 1
num_gpus: 4
trust_remote_code: False
dpo: True
ContextualAI/archangel_sft-dpo_pythia1-4b:
model: ContextualAI/archangel_sft-dpo_pythia1-4b
ref_model: ContextualAI/archangel_sft_pythia1-4b
tokenizer: ContextualAI/archangel_sft-dpo_pythia1-4b
chat_template: tulu
batch_size: 6
trust_remote_code: False
dpo: True
ContextualAI/archangel_sft-kto_pythia1-4b:
model: ContextualAI/archangel_sft-kto_pythia1-4b
ref_model: ContextualAI/archangel_sft_pythia1-4b
tokenizer: ContextualAI/archangel_sft-kto_pythia1-4b
chat_template: tulu
batch_size: 6
trust_remote_code: False
dpo: True
ContextualAI/archangel_sft-dpo_pythia2-8b:
model: ContextualAI/archangel_sft-dpo_pythia2-8b
ref_model: ContextualAI/archangel_sft_pythia2-8b
tokenizer: ContextualAI/archangel_sft-dpo_pythia2-8b
chat_template: tulu
batch_size: 4
trust_remote_code: False
dpo: True
ContextualAI/archangel_sft-kto_pythia2-8b:
model: ContextualAI/archangel_sft-kto_pythia2-8b
ref_model: ContextualAI/archangel_sft_pythia2-8b
tokenizer: ContextualAI/archangel_sft-kto_pythia2-8b
chat_template: tulu
batch_size: 4
trust_remote_code: False
dpo: True
ContextualAI/archangel_sft-dpo_pythia6-9b:
model: ContextualAI/archangel_sft-dpo_pythia6-9b
ref_model: ContextualAI/archangel_sft_pythia6-9b
tokenizer: ContextualAI/archangel_sft-dpo_pythia6-9b
chat_template: tulu
batch_size: 4
trust_remote_code: False
dpo: True
ContextualAI/archangel_sft-kto_pythia6-9b:
model: ContextualAI/archangel_sft-kto_pythia6-9b
ref_model: ContextualAI/archangel_sft_pythia6-9b
tokenizer: ContextualAI/archangel_sft-kto_pythia6-9b
chat_template: tulu
batch_size: 4
trust_remote_code: False
dpo: True
ContextualAI/archangel_sft-dpo_pythia12-0b:
model: ContextualAI/archangel_sft-dpo_pythia12-0b
ref_model: ContextualAI/archangel_sft_pythia12-0b
tokenizer: ContextualAI/archangel_sft-dpo_pythia12-0b
chat_template: tulu
batch_size: 4
num_gpus: 2
trust_remote_code: False
dpo: True
ContextualAI/archangel_sft-kto_pythia12-0b:
model: ContextualAI/archangel_sft-kto_pythia12-0b
ref_model: ContextualAI/archangel_sft_pythia12-0b
tokenizer: ContextualAI/archangel_sft-kto_pythia12-0b
chat_template: tulu
batch_size: 4
num_gpus: 2
trust_remote_code: False
dpo: True
0-hero/Matter-0.1-7B-DPO-preview:
model: 0-hero/Matter-0.1-7B-DPO-preview
ref_model: 0-hero/Matter-0.1-7B
tokenizer: 0-hero/Matter-0.1-7B-DPO-preview
chat_template: # none for tokenizer
batch_size: 4
trust_remote_code: False
dpo: True
0-hero/Matter-0.1-7B-boost-DPO-preview:
model: 0-hero/Matter-0.1-7B-boost-DPO-preview
ref_model: 0-hero/Matter-0.1-7B-boost
tokenizer: 0-hero/Matter-0.1-7B-boost-DPO-preview
chat_template: # none for tokenizer
batch_size: 4
trust_remote_code: False
dpo: True
openbmb/Eurus-RM-7b:
model: openbmb/Eurus-RM-7b
tokenizer: openbmb/Eurus-RM-7b
chat_template: mistral
batch_size: 16
trust_remote_code: True
dpo: False
openbmb/Eurus-7b-kto:
model: openbmb/Eurus-7b-kto
ref_model: openbmb/Eurus-7b-sft
tokenizer: openbmb/Eurus-7b-kto
chat_template: mistral
batch_size: 4
trust_remote_code: True
dpo: True
Qwen/Qwen1.5-MoE-A2.7B-Chat:
model: Qwen/Qwen1.5-MoE-A2.7B-Chat
ref_model: Qwen/Qwen1.5-MoE-A2.7B
tokenizer: Qwen/Qwen1.5-MoE-A2.7B-Chat
chat_template: # none for tokenizer
num_gpus: 2
batch_size: 3
trust_remote_code: False
dpo: True
stabilityai/stable-code-instruct-3b:
model: stabilityai/stable-code-instruct-3b
ref_model: stabilityai/stable-code-3b
tokenizer: stabilityai/stable-code-instruct-3b
chat_template: # none for tokenizer
batch_size: 4
trust_remote_code: True
dpo: True
HuggingFaceH4/starchat2-15b-v0.1:
model: HuggingFaceH4/starchat2-15b-v0.1
ref_model: HuggingFaceH4/starchat2-15b-sft-v0.1
tokenizer: HuggingFaceH4/starchat2-15b-v0.1
chat_template: # none for tokenizer
batch_size: 4
num_gpus: 2
trust_remote_code: False
dpo: True
stabilityai/stablelm-2-12b-chat:
model: stabilityai/stablelm-2-12b-chat
ref_model: stabilityai/stablelm-2-12b
tokenizer: stabilityai/stablelm-2-12b-chat
chat_template: # none for tokenizer
batch_size: 4
num_gpus: 2
trust_remote_code: True
dpo: True
upstage/SOLAR-10.7B-Instruct-v1.0:
model: upstage/SOLAR-10.7B-Instruct-v1.0
ref_model: upstage/SOLAR-10.7B-v1.0
tokenizer: upstage/SOLAR-10.7B-Instruct-v1.0
chat_template: # none for tokenizer
batch_size: 4
num_gpus: 2
trust_remote_code: False
dpo: True
jondurbin/bagel-dpo-34b-v0.5:
model: jondurbin/bagel-dpo-34b-v0.5
ref_model: jondurbin/bagel-34b-v0.5
tokenizer: jondurbin/bagel-dpo-34b-v0.5
chat_template: # none for tokenizer
batch_size: 2
num_gpus: 4
trust_remote_code: False
dpo: True
openbmb/MiniCPM-2B-dpo-fp32:
model: openbmb/MiniCPM-2B-dpo-fp32
ref_model: openbmb/MiniCPM-2B-sft-fp32
tokenizer: openbmb/MiniCPM-2B-dpo-fp32
chat_template: # none for tokenizer
batch_size: 4
trust_remote_code: True
dpo: True
# Note: way not want to re-run generative models all the time
# meta-llama/Meta-Llama-3-8B-Instruct:
# model: meta-llama/Meta-Llama-3-8B-Instruct
# tokenizer: meta-llama/Meta-Llama-3-8B-Instruct
# chat_template: # none for tokenizer
# trust_remote_code: False
# num_gpus: 1
# generative: True
# dpo: False
# meta-llama/Meta-Llama-3-70B-Instruct:
# model: meta-llama/Meta-Llama-3-70B-Instruct
# tokenizer: meta-llama/Meta-Llama-3-70B-Instruct
# chat_template: # none for tokenizer
# trust_remote_code: False
# num_gpus: 4
# generative: True
# dpo: False
# CohereForAI/c4ai-command-r-plus:
# model: CohereForAI/c4ai-command-r-plus
# tokenizer: CohereForAI/c4ai-command-r-plus
# chat_template: # none for tokenizer
# trust_remote_code: False
# num_gpus: 4
# generative: True
# dpo: False
# End generative reward models
sfairXC/FsfairX-LLaMA3-RM-v0.1:
model: sfairXC/FsfairX-LLaMA3-RM-v0.1
tokenizer: sfairXC/FsfairX-LLaMA3-RM-v0.1
chat_template: # none for tokenizer
batch_size: 4
trust_remote_code: False
dpo: False
torch_dtype: bfloat16
RLHFlow/pair-preference-model-LLaMA3-8B:
model: RLHFlow/pair-preference-model-LLaMA3-8B
tokenizer: RLHFlow/pair-preference-model-LLaMA3-8B
chat_template: # none for tokenizer
batch_size: 4
trust_remote_code: False
dpo: False
RLHFlow/RewardModel-Mistral-7B-for-DPA-v1:
model: RLHFlow/RewardModel-Mistral-7B-for-DPA-v1
tokenizer: RLHFlow/RewardModel-Mistral-7B-for-DPA-v1
chat_template: # none for tokenizer
batch_size: 4
trust_remote_code: True
dpo: False
RLHFlow/LLaMA3-iterative-DPO-final:
model: RLHFlow/LLaMA3-iterative-DPO-final
ref_model: RLHFlow/LLaMA3-SFT
tokenizer: RLHFlow/LLaMA3-iterative-DPO-final
chat_template: # none for tokenizer
batch_size: 4
trust_remote_code: False
num_gpus: 2
dpo: True
RLHFlow/ArmoRM-Llama3-8B-v0.1:
model: RLHFlow/ArmoRM-Llama3-8B-v0.1
tokenizer: RLHFlow/ArmoRM-Llama3-8B-v0.1
chat_template: # none for tokenizer
batch_size: 4
trust_remote_code: True
dpo: False
PKU-Alignment/beaver-7b-v2.0-reward:
model: 'PKU-Alignment/beaver-7b-v2.0-reward'
tokenizer: 'PKU-Alignment/beaver-7b-v2.0-reward'
chat_template: 'pku-align'
batch_size: 16
trust_remote_code: False
dpo: False
PKU-Alignment/beaver-7b-v2.0-cost:
model: 'PKU-Alignment/beaver-7b-v2.0-cost'
tokenizer: 'PKU-Alignment/beaver-7b-v2.0-cost'
chat_template: 'pku-align'
batch_size: 16
trust_remote_code: False
dpo: False
# Tulu 2.5 RMs
allenai/tulu-v2.5-13b-uf-rm:
model: allenai/tulu-v2.5-13b-uf-rm
tokenizer: allenai/tulu-v2.5-13b-uf-rm
chat_template: # none for tokenizer
batch_size: 4
trust_remote_code: False
dpo: False
allenai/tulu-v2.5-13b-preference-mix-rm:
model: allenai/tulu-v2.5-13b-preference-mix-rm
tokenizer: allenai/tulu-v2.5-13b-preference-mix-rm
chat_template: # none for tokenizer
batch_size: 4
trust_remote_code: False
dpo: False
allenai/tulu-v2.5-70b-uf-rm:
model: allenai/tulu-v2.5-70b-uf-rm
tokenizer: allenai/tulu-v2.5-70b-uf-rm
chat_template: # none for tokenizer
batch_size: 2
trust_remote_code: False
dpo: False
allenai/tulu-v2.5-70b-preference-mix-rm:
model: allenai/tulu-v2.5-70b-preference-mix-rm
tokenizer: allenai/tulu-v2.5-70b-preference-mix-rm
chat_template: # none for tokenizer
batch_size: 2
trust_remote_code: False
dpo: False
allenai/llama-3-tulu-2-70b-uf-mean-rm:
model: allenai/llama-3-tulu-2-70b-uf-mean-rm
tokenizer: allenai/llama-3-tulu-2-70b-uf-mean-rm
chat_template: # none for tokenizer
batch_size: 2
num_gpus: 4
trust_remote_code: False
dpo: False
allenai/llama-3-tulu-2-8b-uf-mean-rm:
model: allenai/llama-3-tulu-2-8b-uf-mean-rm
tokenizer: allenai/llama-3-tulu-2-8b-uf-mean-rm
chat_template: # none for tokenizer
batch_size: 4
trust_remote_code: False
dpo: False
Ahjeong/MMPO_Gemma_7b:
model: Ahjeong/MMPO_Gemma_7b
ref_model: kykim0/gemma-7b-ultrachat-sft
tokenizer: Ahjeong/MMPO_Gemma_7b
chat_template: # none for tokenizer
num_gpus: 2
batch_size: 2
trust_remote_code: False
dpo: True
Ahjeong/MMPO_Gemma_7b_gamma1.1_epoch3:
model: Ahjeong/MMPO_Gemma_7b_gamma1.1_epoch3
ref_model: kykim0/gemma-7b-ultrachat-sft
tokenizer: Ahjeong/MMPO_Gemma_7b_gamma1.1_epoch3
chat_template: # none for tokenizer
num_gpus: 2
batch_size: 2
trust_remote_code: False
dpo: True
torch_dtype: bfloat16
wenbopan/Faro-Yi-9B-DPO:
model: wenbopan/Faro-Yi-9B-DPO
ref_model: wenbopan/Faro-Yi-9B
tokenizer: wenbopan/Faro-Yi-9B-DPO
chat_template: # none for tokenizer
batch_size: 2
num_gpus: 2
trust_remote_code: False
dpo: True
allenai/llama-3-tulu-2-dpo-8b:
model: allenai/llama-3-tulu-2-dpo-8b
ref_model: allenai/llama-3-tulu-2-8b
tokenizer: allenai/llama-3-tulu-2-dpo-8b
chat_template: # none for tokenizer
batch_size: 1
num_gpus: 2
trust_remote_code: False
dpo: True
allenai/llama-3-tulu-2-dpo-70b:
model: allenai/llama-3-tulu-2-dpo-70b
ref_model: allenai/llama-3-tulu-2-70b
tokenizer: allenai/llama-3-tulu-2-dpo-70b
chat_template: # none for tokenizer
batch_size: 1
num_gpus: 4
trust_remote_code: False
dpo: True
Ray2333/GRM-llama3-8B-distill:
model: Ray2333/GRM-llama3-8B-distill
tokenizer: Ray2333/GRM-llama3-8B-distill
chat_template: # none for tokenizer
batch_size: 4
trust_remote_code: False
dpo: False
Ray2333/Gemma-2B-rewardmodel-baseline:
model: Ray2333/Gemma-2B-rewardmodel-baseline
tokenizer: Ray2333/Gemma-2B-rewardmodel-baseline
chat_template: # none for tokenizer
batch_size: 8
trust_remote_code: False
dpo: False
Ray2333/GRM-llama3-8B-sftreg:
model: Ray2333/GRM-llama3-8B-sftreg
tokenizer: Ray2333/GRM-llama3-8B-sftreg
chat_template: # none for tokenizer
batch_size: 4
trust_remote_code: False
dpo: False
Ray2333/GRM-Gemma-2B-sftreg:
model: Ray2333/GRM-Gemma-2B-sftreg
tokenizer: Ray2333/GRM-Gemma-2B-sftreg
chat_template: # none for tokenizer
batch_size: 8
trust_remote_code: False
dpo: False
CIR-AMS/BTRM_Qwen2_7b_0613:
model: CIR-AMS/BTRM_Qwen2_7b_0613
tokenizer: CIR-AMS/BTRM_Qwen2_7b_0613
chat_template: # none for tokenizer
batch_size: 4
trust_remote_code: False
dpo: False
torch_dtype: bfloat16
internlm/internlm2-1_8b-reward:
model: internlm/internlm2-1_8b-reward
tokenizer: internlm/internlm2-1_8b-reward
chat_template: # none for tokenizer
batch_size: 16
trust_remote_code: True
dpo: False
internlm/internlm2-7b-reward:
model: internlm/internlm2-7b-reward
tokenizer: internlm/internlm2-7b-reward
chat_template: # none for tokenizer
batch_size: 4
trust_remote_code: True
dpo: False
internlm/internlm2-20b-reward:
model: internlm/internlm2-20b-reward
tokenizer: internlm/internlm2-20b-reward
chat_template: # none for tokenizer
batch_size: 2
trust_remote_code: True
dpo: False
NCSOFT/Llama-3-OffsetBias-RM-8B:
model: NCSOFT/Llama-3-OffsetBias-RM-8B
tokenizer: NCSOFT/Llama-3-OffsetBias-RM-8B
chat_template: # none for tokenizer
batch_size: 4
torch_dtype: bfloat16
dpo: False
trust_remote_code: False
Skywork/Skywork-Reward-Gemma-2-27B:
model: Skywork/Skywork-Reward-Gemma-2-27B
tokenizer: Skywork/Skywork-Reward-Gemma-2-27B
chat_template: # none for tokenizer
batch_size: 2
dpo: False
torch_dtype: bfloat16
trust_remote_code: False
attention_implementation: flash_attention_2
Skywork/Skywork-Reward-Llama-3.1-8B:
model: Skywork/Skywork-Reward-Llama-3.1-8B
tokenizer: Skywork/Skywork-Reward-Llama-3.1-8B
chat_template: # none for tokenizer
batch_size: 8
dpo: False
torch_dtype: bfloat16
trust_remote_code: False
LxzGordon/URM-LLaMa-3.1-8B:
model: LxzGordon/URM-LLaMa-3.1-8B
tokenizer: LxzGordon/URM-LLaMa-3.1-8B
chat_template: # none for tokenizer
batch_size: 4
dpo: False
trust_remote_code: True
quantized: False
LxzGordon/URM-LLaMa-3-8B:
model: LxzGordon/URM-LLaMa-3-8B
tokenizer: LxzGordon/URM-LLaMa-3-8B
chat_template: # none for tokenizer
batch_size: 4
dpo: False
trust_remote_code: True
quantized: False
# Skywork/Skywork-Critic-Llama-3.1-8B:
# model: Skywork/Skywork-Critic-Llama-3.1-8B
# tokenizer: Skywork/Skywork-Critic-Llama-3.1-8B
# chat_template: # none for tokenizer
# batch_size: 4
# dpo: False
# generative: True
# num_gpus: 1
Ray2333/GRM-Gemma-2B-rewardmodel-ft:
model: Ray2333/GRM-Gemma-2B-rewardmodel-ft
tokenizer: Ray2333/GRM-Gemma-2B-rewardmodel-ft
chat_template: # none for tokenizer
batch_size: 16
trust_remote_code: False
dpo: False
quantized: False
Ray2333/Gemma-2B-rewardmodel-ft:
model: Ray2333/Gemma-2B-rewardmodel-ft
tokenizer: Ray2333/Gemma-2B-rewardmodel-ft
chat_template: # none for tokenizer
batch_size: 16
trust_remote_code: False
dpo: False
quantized: False
Ray2333/GRM-Llama3-8B-rewardmodel-ft:
model: Ray2333/GRM-Llama3-8B-rewardmodel-ft
tokenizer: Ray2333/GRM-Llama3-8B-rewardmodel-ft
chat_template: # none for tokenizer
batch_size: 8
trust_remote_code: False
dpo: False
quantized: False
# for QRM models, use export ACCELERATE_MIXED_PRECISION=bf16; for best performance
nicolinho/QRM-Llama3.1-8B:
model: nicolinho/QRM-Llama3.1-8B
tokenizer: nicolinho/QRM-Llama3.1-8B
chat_template: # none for tokenizer
batch_size: 8
dpo: False
trust_remote_code: True
quantized: False
attention_implementation: flash_attention_2
nicolinho/QRM-Llama3-8B:
model: nicolinho/QRM-Llama3-8B
tokenizer: nicolinho/QRM-Llama3-8B
chat_template: # none for tokenizer
batch_size: 8
dpo: False
trust_remote_code: True
quantized: False
attention_implementation: flash_attention_2
Skywork/Skywork-Reward-Gemma-2-27B-v0.2:
model: Skywork/Skywork-Reward-Gemma-2-27B-v0.2
tokenizer: Skywork/Skywork-Reward-Gemma-2-27B-v0.2
chat_template: # none for tokenizer
batch_size: 2
dpo: False
torch_dtype: bfloat16
trust_remote_code: False
attention_implementation: flash_attention_2
quantized: False
Skywork/Skywork-Reward-Llama-3.1-8B-v0.2:
model: Skywork/Skywork-Reward-Llama-3.1-8B-v0.2
tokenizer: Skywork/Skywork-Reward-Llama-3.1-8B-v0.2
chat_template: # none for tokenizer
batch_size: 8
dpo: False
torch_dtype: bfloat16
trust_remote_code: False
quantized: False
Ray2333/GRM-Gemma2-2B-sftreg:
model: Ray2333/GRM-Gemma2-2B-sftreg
tokenizer: Ray2333/GRM-Gemma2-2B-sftreg
chat_template: # none for tokenizer
batch_size: 16
trust_remote_code: False
dpo: False
quantized: False
Ray2333/GRM-llama3.2-3B-sftreg:
model: Ray2333/GRM-llama3.2-3B-sftreg
tokenizer: Ray2333/GRM-llama3.2-3B-sftreg
chat_template: # none for tokenizer
batch_size: 16
trust_remote_code: False
dpo: False
quantized: False
Ray2333/GRM-gemma2-2B-rewardmodel-ft:
model: Ray2333/GRM-gemma2-2B-rewardmodel-ft
tokenizer: Ray2333/GRM-gemma2-2B-rewardmodel-ft
chat_template: # none for tokenizer
batch_size: 16
trust_remote_code: False
dpo: False
quantized: False
Ray2333/GRM-llama3.2-3B-rewardmodel-ft:
model: Ray2333/GRM-llama3.2-3B-rewardmodel-ft
tokenizer: Ray2333/GRM-llama3.2-3B-rewardmodel-ft
chat_template: # none for tokenizer
batch_size: 16
trust_remote_code: False
dpo: False
quantized: False
infly/INF-ORM-Llama3.1-70B:
model: infly/INF-ORM-Llama3.1-70B
tokenizer: infly/INF-ORM-Llama3.1-70B
chat_template: # none for tokenizer
batch_size: 16
torch_dtype: bfloat16
attention_implementation: flash_attention_2
trust_remote_code: False
dpo: False
quantized: False
SultanR/SmolTulu-1.7b-RM:
model: SultanR/SmolTulu-1.7b-RM
tokenizer: SultanR/SmolTulu-1.7b-RM
chat_template: # none for tokenizer
batch_size: 16
trust_remote_code: False
dpo: False
quantized: False
nicolinho/QRM-Gemma-2-27B:
model: nicolinho/QRM-Gemma-2-27B
tokenizer: nicolinho/QRM-Gemma-2-27B
chat_template: # none for tokenizer
batch_size: 1
torch_dtype: bfloat16
attention_implementation: flash_attention_2
max_length: 4096
quantized: False
dpo: False
trust_remote_code: True
nicolinho/QRM-Llama3.1-8B-v2:
model: nicolinho/QRM-Llama3.1-8B-v2
tokenizer: nicolinho/QRM-Llama3.1-8B-v2
chat_template: # none for tokenizer
batch_size: 1
torch_dtype: bfloat16
attention_implementation: flash_attention_2
quantized: False
dpo: False
trust_remote_code: True