RikkiXu commited on
Commit
b622b6f
1 Parent(s): c2002f4

Model save

Browse files
README.md CHANGED
@@ -32,7 +32,7 @@ More information needed
32
  ### Training hyperparameters
33
 
34
  The following hyperparameters were used during training:
35
- - learning_rate: 1e-07
36
  - train_batch_size: 4
37
  - eval_batch_size: 4
38
  - seed: 42
 
32
  ### Training hyperparameters
33
 
34
  The following hyperparameters were used during training:
35
+ - learning_rate: 4e-07
36
  - train_batch_size: 4
37
  - eval_batch_size: 4
38
  - seed: 42
all_results.json CHANGED
@@ -1,8 +1,8 @@
1
  {
2
  "epoch": 1.0,
3
- "train_loss": 0.5277958785073232,
4
- "train_runtime": 6372.8883,
5
  "train_samples": 52922,
6
- "train_samples_per_second": 8.304,
7
  "train_steps_per_second": 0.065
8
  }
 
1
  {
2
  "epoch": 1.0,
3
+ "train_loss": 3.8421780889894426,
4
+ "train_runtime": 6381.4933,
5
  "train_samples": 52922,
6
+ "train_samples_per_second": 8.293,
7
  "train_steps_per_second": 0.065
8
  }
model-00001-of-00003.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:ba15de61ecf1fab38692965bea81fb4fbc15bdea9730cfaafee7a941062fd99a
3
  size 4943162336
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4e88c6ac9b4127cd43c95069301c9c933118b1525797db81fc6d80f544cf7cec
3
  size 4943162336
model-00002-of-00003.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:a86b9d593b597cc2ecc5f43476a07ebe0e278ba57f94d4f7643b216560e89d16
3
  size 4999819336
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bd64f1edcaa32df3d1b32cd35af92f119cd07ea2ba7ebf00b6828eed52da98f0
3
  size 4999819336
model-00003-of-00003.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:b9c8ca0b8509b351f3013c55b40a6f892b4fec7ad730321d44ee3a7d0c8d4bf4
3
  size 4540516344
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:87c91c34961d0019bf71d071d06cfd82742211fd4d3379b61e5bc9cad550096c
3
  size 4540516344
runs/Jun30_16-42-14_n136-100-194/events.out.tfevents.1719737283.n136-100-194.302982.0 CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:824e28b5cb20671cfc59b0993e4e6babc44a148db8e00053f715d99ea18124f3
3
- size 32943
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ae7c4645315dc9db21552bd2b3b636781f6717efc0255ff69d2ebf3333897606
3
+ size 33985
train_results.json CHANGED
@@ -1,8 +1,8 @@
1
  {
2
  "epoch": 1.0,
3
- "train_loss": 0.5277958785073232,
4
- "train_runtime": 6372.8883,
5
  "train_samples": 52922,
6
- "train_samples_per_second": 8.304,
7
  "train_steps_per_second": 0.065
8
  }
 
1
  {
2
  "epoch": 1.0,
3
+ "train_loss": 3.8421780889894426,
4
+ "train_runtime": 6381.4933,
5
  "train_samples": 52922,
6
+ "train_samples_per_second": 8.293,
7
  "train_steps_per_second": 0.065
8
  }
trainer_state.json CHANGED
@@ -10,13 +10,13 @@
10
  "log_history": [
11
  {
12
  "epoch": 0.0,
13
- "grad_norm": 1308.2122296641476,
14
- "learning_rate": 2.3809523809523806e-09,
15
  "logits/chosen": -2.7005977630615234,
16
  "logits/rejected": -2.6288318634033203,
17
  "logps/chosen": -1.1158788204193115,
18
  "logps/rejected": -1.1333446502685547,
19
- "loss": 0.6931,
20
  "rewards/accuracies": 0.0,
21
  "rewards/chosen": 0.0,
22
  "rewards/margins": 0.0,
@@ -25,626 +25,626 @@
25
  },
26
  {
27
  "epoch": 0.02,
28
- "grad_norm": 1049.0503356036236,
29
- "learning_rate": 2.3809523809523807e-08,
30
- "logits/chosen": -2.762399435043335,
31
- "logits/rejected": -2.6968984603881836,
32
- "logps/chosen": -0.8372963070869446,
33
- "logps/rejected": -0.8176102638244629,
34
- "loss": 0.7125,
35
- "rewards/accuracies": 0.4236111044883728,
36
- "rewards/chosen": 0.03224152699112892,
37
- "rewards/margins": 0.04410284012556076,
38
- "rewards/rejected": -0.011861314065754414,
39
  "step": 10
40
  },
41
  {
42
  "epoch": 0.05,
43
- "grad_norm": 1236.016019303981,
44
- "learning_rate": 4.7619047619047613e-08,
45
- "logits/chosen": -2.689128875732422,
46
- "logits/rejected": -2.64937686920166,
47
- "logps/chosen": -0.9927361607551575,
48
- "logps/rejected": -1.03745436668396,
49
- "loss": 0.7127,
50
- "rewards/accuracies": 0.5062500238418579,
51
- "rewards/chosen": -0.09930654615163803,
52
- "rewards/margins": 0.010919012129306793,
53
- "rewards/rejected": -0.11022555828094482,
54
  "step": 20
55
  },
56
  {
57
  "epoch": 0.07,
58
- "grad_norm": 1468.1565781012905,
59
- "learning_rate": 7.142857142857142e-08,
60
- "logits/chosen": -2.7302658557891846,
61
- "logits/rejected": -2.6768224239349365,
62
- "logps/chosen": -0.967939555644989,
63
- "logps/rejected": -0.9760215878486633,
64
- "loss": 0.6942,
65
- "rewards/accuracies": 0.59375,
66
- "rewards/chosen": -0.1426212042570114,
67
- "rewards/margins": 0.09910523146390915,
68
- "rewards/rejected": -0.24172644317150116,
69
  "step": 30
70
  },
71
  {
72
  "epoch": 0.1,
73
- "grad_norm": 1718.8019373304546,
74
- "learning_rate": 9.523809523809523e-08,
75
- "logits/chosen": -2.6734580993652344,
76
- "logits/rejected": -2.6297881603240967,
77
- "logps/chosen": -0.9952117204666138,
78
- "logps/rejected": -0.920923113822937,
79
- "loss": 0.6688,
80
- "rewards/accuracies": 0.675000011920929,
81
- "rewards/chosen": -0.6706060171127319,
82
- "rewards/margins": 0.4042028486728668,
83
- "rewards/rejected": -1.0748088359832764,
84
  "step": 40
85
  },
86
  {
87
  "epoch": 0.12,
88
- "grad_norm": 886.1855666160817,
89
- "learning_rate": 9.988531521210217e-08,
90
- "logits/chosen": -2.739663600921631,
91
- "logits/rejected": -2.6859798431396484,
92
- "logps/chosen": -0.9715211987495422,
93
- "logps/rejected": -0.9160677790641785,
94
- "loss": 0.6298,
95
- "rewards/accuracies": 0.75,
96
- "rewards/chosen": -0.4754611551761627,
97
- "rewards/margins": 0.44252967834472656,
98
- "rewards/rejected": -0.9179908633232117,
99
  "step": 50
100
  },
101
  {
102
  "epoch": 0.15,
103
- "grad_norm": 1659.7392402989283,
104
- "learning_rate": 9.94203097871474e-08,
105
- "logits/chosen": -2.743115186691284,
106
- "logits/rejected": -2.676964521408081,
107
- "logps/chosen": -0.9205001592636108,
108
- "logps/rejected": -0.9147119522094727,
109
- "loss": 0.6278,
110
- "rewards/accuracies": 0.7749999761581421,
111
- "rewards/chosen": -0.4151291251182556,
112
- "rewards/margins": 0.4165617823600769,
113
- "rewards/rejected": -0.8316909074783325,
114
  "step": 60
115
  },
116
  {
117
  "epoch": 0.17,
118
- "grad_norm": 1314.7754621298307,
119
- "learning_rate": 9.860114570402053e-08,
120
- "logits/chosen": -2.744157314300537,
121
- "logits/rejected": -2.7204043865203857,
122
- "logps/chosen": -0.9306680560112,
123
- "logps/rejected": -0.8743250966072083,
124
- "loss": 0.5734,
125
- "rewards/accuracies": 0.793749988079071,
126
- "rewards/chosen": -0.3868916630744934,
127
- "rewards/margins": 0.6596783399581909,
128
- "rewards/rejected": -1.046570062637329,
129
  "step": 70
130
  },
131
  {
132
  "epoch": 0.19,
133
- "grad_norm": 1220.0578782834486,
134
- "learning_rate": 9.743369330335185e-08,
135
- "logits/chosen": -2.6871607303619385,
136
- "logits/rejected": -2.6331558227539062,
137
- "logps/chosen": -0.901233971118927,
138
- "logps/rejected": -0.9974308013916016,
139
- "loss": 0.5749,
140
- "rewards/accuracies": 0.768750011920929,
141
- "rewards/chosen": -0.6774497032165527,
142
- "rewards/margins": 0.5885565876960754,
143
- "rewards/rejected": -1.266006350517273,
144
  "step": 80
145
  },
146
  {
147
  "epoch": 0.22,
148
- "grad_norm": 906.8663763416491,
149
- "learning_rate": 9.592631884948653e-08,
150
- "logits/chosen": -2.6903910636901855,
151
- "logits/rejected": -2.6411349773406982,
152
- "logps/chosen": -1.0175859928131104,
153
- "logps/rejected": -1.0327794551849365,
154
- "loss": 0.5569,
155
- "rewards/accuracies": 0.71875,
156
- "rewards/chosen": -0.40117961168289185,
157
- "rewards/margins": 0.8878555297851562,
158
- "rewards/rejected": -1.2890350818634033,
159
  "step": 90
160
  },
161
  {
162
  "epoch": 0.24,
163
- "grad_norm": 866.6073604256766,
164
- "learning_rate": 9.408982457568138e-08,
165
- "logits/chosen": -2.7176403999328613,
166
- "logits/rejected": -2.671095371246338,
167
- "logps/chosen": -0.8981878161430359,
168
- "logps/rejected": -0.9157611131668091,
169
- "loss": 0.5818,
170
  "rewards/accuracies": 0.8062499761581421,
171
- "rewards/chosen": -0.6816717386245728,
172
- "rewards/margins": 0.9627196192741394,
173
- "rewards/rejected": -1.6443912982940674,
174
  "step": 100
175
  },
176
  {
177
  "epoch": 0.27,
178
- "grad_norm": 1160.110812326109,
179
- "learning_rate": 9.193737127252132e-08,
180
- "logits/chosen": -2.7276268005371094,
181
- "logits/rejected": -2.6847987174987793,
182
- "logps/chosen": -0.9827289581298828,
183
- "logps/rejected": -0.94146728515625,
184
- "loss": 0.5535,
185
- "rewards/accuracies": 0.762499988079071,
186
- "rewards/chosen": -0.4507879614830017,
187
- "rewards/margins": 0.8122564554214478,
188
- "rewards/rejected": -1.2630443572998047,
189
  "step": 110
190
  },
191
  {
192
  "epoch": 0.29,
193
- "grad_norm": 1118.8296262026665,
194
- "learning_rate": 8.94843839743072e-08,
195
- "logits/chosen": -2.684906244277954,
196
- "logits/rejected": -2.6362223625183105,
197
- "logps/chosen": -0.8857740163803101,
198
- "logps/rejected": -0.9783684611320496,
199
- "loss": 0.5057,
200
- "rewards/accuracies": 0.831250011920929,
201
- "rewards/chosen": -0.4952009320259094,
202
- "rewards/margins": 1.4160592555999756,
203
- "rewards/rejected": -1.9112603664398193,
204
  "step": 120
205
  },
206
  {
207
  "epoch": 0.31,
208
- "grad_norm": 816.4913832702339,
209
- "learning_rate": 8.674844141929039e-08,
210
- "logits/chosen": -2.7162153720855713,
211
- "logits/rejected": -2.680973529815674,
212
- "logps/chosen": -0.8731144666671753,
213
- "logps/rejected": -0.8701594471931458,
214
- "loss": 0.5054,
215
- "rewards/accuracies": 0.831250011920929,
216
- "rewards/chosen": -0.34728819131851196,
217
- "rewards/margins": 1.2909767627716064,
218
- "rewards/rejected": -1.6382650136947632,
219
  "step": 130
220
  },
221
  {
222
  "epoch": 0.34,
223
- "grad_norm": 1469.7710854903712,
224
- "learning_rate": 8.374915007591053e-08,
225
- "logits/chosen": -2.691988945007324,
226
- "logits/rejected": -2.652631998062134,
227
- "logps/chosen": -0.9311792254447937,
228
- "logps/rejected": -0.9303830862045288,
229
- "loss": 0.534,
230
- "rewards/accuracies": 0.7875000238418579,
231
- "rewards/chosen": -0.5356858968734741,
232
- "rewards/margins": 1.0560630559921265,
233
- "rewards/rejected": -1.5917489528656006,
234
  "step": 140
235
  },
236
  {
237
  "epoch": 0.36,
238
- "grad_norm": 1688.4287131310427,
239
- "learning_rate": 8.05080036377971e-08,
240
- "logits/chosen": -2.7248551845550537,
241
- "logits/rejected": -2.6649651527404785,
242
- "logps/chosen": -0.9584500193595886,
243
- "logps/rejected": -0.9874745607376099,
244
- "loss": 0.5552,
245
- "rewards/accuracies": 0.768750011920929,
246
- "rewards/chosen": -0.24187500774860382,
247
- "rewards/margins": 1.0612785816192627,
248
- "rewards/rejected": -1.303153395652771,
249
  "step": 150
250
  },
251
  {
252
  "epoch": 0.39,
253
- "grad_norm": 883.2476388726114,
254
- "learning_rate": 7.704822899442949e-08,
255
- "logits/chosen": -2.717927932739258,
256
- "logits/rejected": -2.6812281608581543,
257
- "logps/chosen": -0.9134844541549683,
258
- "logps/rejected": -0.9353663325309753,
259
- "loss": 0.5424,
260
- "rewards/accuracies": 0.824999988079071,
261
- "rewards/chosen": -0.3312217593193054,
262
- "rewards/margins": 1.2677682638168335,
263
- "rewards/rejected": -1.5989899635314941,
264
  "step": 160
265
  },
266
  {
267
  "epoch": 0.41,
268
- "grad_norm": 1741.9681865719251,
269
- "learning_rate": 7.339461978126947e-08,
270
- "logits/chosen": -2.696554660797119,
271
- "logits/rejected": -2.628129482269287,
272
- "logps/chosen": -0.896826446056366,
273
- "logps/rejected": -0.8772991895675659,
274
- "loss": 0.5503,
275
  "rewards/accuracies": 0.800000011920929,
276
- "rewards/chosen": -0.48312321305274963,
277
- "rewards/margins": 1.3627839088439941,
278
- "rewards/rejected": -1.845907211303711,
279
  "step": 170
280
  },
281
  {
282
  "epoch": 0.44,
283
- "grad_norm": 1946.5851641629447,
284
- "learning_rate": 6.957335870218904e-08,
285
- "logits/chosen": -2.673733949661255,
286
- "logits/rejected": -2.6041712760925293,
287
- "logps/chosen": -0.939963698387146,
288
- "logps/rejected": -0.956584095954895,
289
- "loss": 0.4957,
290
- "rewards/accuracies": 0.8500000238418579,
291
- "rewards/chosen": 0.2670658528804779,
292
- "rewards/margins": 1.5085264444351196,
293
- "rewards/rejected": -1.2414608001708984,
294
  "step": 180
295
  },
296
  {
297
  "epoch": 0.46,
298
- "grad_norm": 2276.5142714905005,
299
- "learning_rate": 6.56118298974763e-08,
300
- "logits/chosen": -2.732027292251587,
301
- "logits/rejected": -2.7048563957214355,
302
- "logps/chosen": -0.9618963003158569,
303
- "logps/rejected": -1.0246574878692627,
304
- "loss": 0.5708,
305
- "rewards/accuracies": 0.7749999761581421,
306
- "rewards/chosen": -0.570631742477417,
307
- "rewards/margins": 1.4420315027236938,
308
- "rewards/rejected": -2.012662887573242,
309
  "step": 190
310
  },
311
  {
312
  "epoch": 0.48,
313
- "grad_norm": 1178.2661546943384,
314
- "learning_rate": 6.153842270203887e-08,
315
- "logits/chosen": -2.6935534477233887,
316
- "logits/rejected": -2.6572394371032715,
317
- "logps/chosen": -0.8579891920089722,
318
- "logps/rejected": -0.931664764881134,
319
- "loss": 0.5316,
320
- "rewards/accuracies": 0.793749988079071,
321
- "rewards/chosen": -0.0793720930814743,
322
- "rewards/margins": 1.138346552848816,
323
- "rewards/rejected": -1.217718482017517,
324
  "step": 200
325
  },
326
  {
327
  "epoch": 0.51,
328
- "grad_norm": 913.5130412259963,
329
- "learning_rate": 5.738232820012406e-08,
330
- "logits/chosen": -2.7250068187713623,
331
- "logits/rejected": -2.6632461547851562,
332
- "logps/chosen": -1.0246375799179077,
333
- "logps/rejected": -0.9633069038391113,
334
- "loss": 0.5393,
335
- "rewards/accuracies": 0.768750011920929,
336
- "rewards/chosen": -0.25627315044403076,
337
- "rewards/margins": 1.4161301851272583,
338
- "rewards/rejected": -1.672403335571289,
339
  "step": 210
340
  },
341
  {
342
  "epoch": 0.53,
343
- "grad_norm": 1025.54215312441,
344
- "learning_rate": 5.317333003449687e-08,
345
- "logits/chosen": -2.733330011367798,
346
- "logits/rejected": -2.715299606323242,
347
- "logps/chosen": -0.8904998898506165,
348
- "logps/rejected": -0.8720332980155945,
349
- "loss": 0.5228,
350
- "rewards/accuracies": 0.8062499761581421,
351
- "rewards/chosen": 0.2184334546327591,
352
- "rewards/margins": 1.4306997060775757,
353
- "rewards/rejected": -1.212266206741333,
354
  "step": 220
355
  },
356
  {
357
  "epoch": 0.56,
358
- "grad_norm": 932.0109937677821,
359
- "learning_rate": 4.894159096919109e-08,
360
- "logits/chosen": -2.6878042221069336,
361
- "logits/rejected": -2.65002179145813,
362
- "logps/chosen": -0.9498124122619629,
363
- "logps/rejected": -0.9380944967269897,
364
- "loss": 0.4868,
365
- "rewards/accuracies": 0.75,
366
- "rewards/chosen": -0.5880553722381592,
367
- "rewards/margins": 1.365509271621704,
368
- "rewards/rejected": -1.9535646438598633,
369
  "step": 230
370
  },
371
  {
372
  "epoch": 0.58,
373
- "grad_norm": 936.3493959909762,
374
- "learning_rate": 4.471743673537994e-08,
375
- "logits/chosen": -2.7069993019104004,
376
- "logits/rejected": -2.6865086555480957,
377
- "logps/chosen": -0.9886058568954468,
378
- "logps/rejected": -0.9898191690444946,
379
- "loss": 0.5347,
380
- "rewards/accuracies": 0.7875000238418579,
381
- "rewards/chosen": 0.04917572811245918,
382
- "rewards/margins": 1.6429868936538696,
383
- "rewards/rejected": -1.5938111543655396,
384
  "step": 240
385
  },
386
  {
387
  "epoch": 0.6,
388
- "grad_norm": 984.7410594483609,
389
- "learning_rate": 4.053113870938224e-08,
390
- "logits/chosen": -2.758028507232666,
391
- "logits/rejected": -2.679352283477783,
392
- "logps/chosen": -0.8968666791915894,
393
- "logps/rejected": -0.9234074354171753,
394
- "loss": 0.5007,
395
- "rewards/accuracies": 0.8374999761581421,
396
- "rewards/chosen": 0.0006614074227400124,
397
- "rewards/margins": 1.3911110162734985,
398
- "rewards/rejected": -1.3904496431350708,
399
  "step": 250
400
  },
401
  {
402
  "epoch": 0.63,
403
- "grad_norm": 2539.435048877582,
404
- "learning_rate": 3.641269698018933e-08,
405
- "logits/chosen": -2.713007926940918,
406
- "logits/rejected": -2.6481966972351074,
407
- "logps/chosen": -1.007727861404419,
408
- "logps/rejected": -0.9894694089889526,
409
- "loss": 0.497,
410
- "rewards/accuracies": 0.824999988079071,
411
- "rewards/chosen": 0.06799235194921494,
412
- "rewards/margins": 1.8265388011932373,
413
- "rewards/rejected": -1.7585465908050537,
414
  "step": 260
415
  },
416
  {
417
  "epoch": 0.65,
418
- "grad_norm": 1438.0528773750598,
419
- "learning_rate": 3.2391625361107026e-08,
420
- "logits/chosen": -2.7407853603363037,
421
- "logits/rejected": -2.6641056537628174,
422
- "logps/chosen": -0.9492539167404175,
423
- "logps/rejected": -0.9392199516296387,
424
- "loss": 0.5391,
425
- "rewards/accuracies": 0.862500011920929,
426
- "rewards/chosen": -0.02730640210211277,
427
- "rewards/margins": 1.7186416387557983,
428
- "rewards/rejected": -1.745948076248169,
429
  "step": 270
430
  },
431
  {
432
  "epoch": 0.68,
433
- "grad_norm": 1220.7276779461947,
434
- "learning_rate": 2.8496739886173992e-08,
435
- "logits/chosen": -2.697509288787842,
436
- "logits/rejected": -2.640756607055664,
437
- "logps/chosen": -0.9139662981033325,
438
- "logps/rejected": -0.9015041589736938,
439
- "loss": 0.4723,
440
- "rewards/accuracies": 0.84375,
441
- "rewards/chosen": 0.023266727104783058,
442
- "rewards/margins": 1.4177985191345215,
443
- "rewards/rejected": -1.3945319652557373,
444
  "step": 280
445
  },
446
  {
447
  "epoch": 0.7,
448
- "grad_norm": 1542.4514800663226,
449
- "learning_rate": 2.4755952307046063e-08,
450
- "logits/chosen": -2.742187023162842,
451
- "logits/rejected": -2.699744701385498,
452
- "logps/chosen": -1.0006037950515747,
453
- "logps/rejected": -0.9843395352363586,
454
- "loss": 0.448,
455
- "rewards/accuracies": 0.831250011920929,
456
- "rewards/chosen": 0.14072290062904358,
457
- "rewards/margins": 1.5138235092163086,
458
- "rewards/rejected": -1.3731005191802979,
459
  "step": 290
460
  },
461
  {
462
  "epoch": 0.73,
463
- "grad_norm": 1452.2302527315665,
464
- "learning_rate": 2.1196070070200995e-08,
465
- "logits/chosen": -2.7377541065216064,
466
- "logits/rejected": -2.6858716011047363,
467
- "logps/chosen": -0.900943398475647,
468
- "logps/rejected": -0.9453694224357605,
469
- "loss": 0.484,
470
- "rewards/accuracies": 0.824999988079071,
471
- "rewards/chosen": -0.028856370598077774,
472
- "rewards/margins": 1.6425704956054688,
473
- "rewards/rejected": -1.671426773071289,
474
  "step": 300
475
  },
476
  {
477
  "epoch": 0.75,
478
- "grad_norm": 966.038328895732,
479
- "learning_rate": 1.7842604207878005e-08,
480
- "logits/chosen": -2.7250287532806396,
481
- "logits/rejected": -2.6620233058929443,
482
- "logps/chosen": -1.0749655961990356,
483
- "logps/rejected": -1.015625238418579,
484
- "loss": 0.4403,
485
  "rewards/accuracies": 0.8062499761581421,
486
- "rewards/chosen": -0.16425299644470215,
487
- "rewards/margins": 1.6934306621551514,
488
- "rewards/rejected": -1.8576834201812744,
489
  "step": 310
490
  },
491
  {
492
  "epoch": 0.77,
493
- "grad_norm": 1686.580221181831,
494
- "learning_rate": 1.4719586519455534e-08,
495
- "logits/chosen": -2.728663444519043,
496
- "logits/rejected": -2.64876389503479,
497
- "logps/chosen": -0.9033769369125366,
498
- "logps/rejected": -0.9375128746032715,
499
- "loss": 0.4301,
500
- "rewards/accuracies": 0.9125000238418579,
501
- "rewards/chosen": 0.11283926665782928,
502
- "rewards/margins": 2.1201956272125244,
503
- "rewards/rejected": -2.0073564052581787,
504
  "step": 320
505
  },
506
  {
507
  "epoch": 0.8,
508
- "grad_norm": 1227.1950699118374,
509
- "learning_rate": 1.18493973533924e-08,
510
- "logits/chosen": -2.679917097091675,
511
- "logits/rejected": -2.611525297164917,
512
- "logps/chosen": -0.9286376237869263,
513
- "logps/rejected": -0.9542851448059082,
514
- "loss": 0.4554,
515
- "rewards/accuracies": 0.800000011920929,
516
- "rewards/chosen": -0.18441525101661682,
517
- "rewards/margins": 1.7495098114013672,
518
- "rewards/rejected": -1.9339250326156616,
519
  "step": 330
520
  },
521
  {
522
  "epoch": 0.82,
523
- "grad_norm": 1294.7256357947958,
524
- "learning_rate": 9.252605223891208e-09,
525
- "logits/chosen": -2.759120464324951,
526
- "logits/rejected": -2.6892759799957275,
527
- "logps/chosen": -0.8945956230163574,
528
- "logps/rejected": -0.9347489476203918,
529
- "loss": 0.5031,
530
- "rewards/accuracies": 0.8125,
531
- "rewards/chosen": -0.42182081937789917,
532
- "rewards/margins": 1.480148196220398,
533
- "rewards/rejected": -1.9019691944122314,
534
  "step": 340
535
  },
536
  {
537
  "epoch": 0.85,
538
- "grad_norm": 914.832611398372,
539
- "learning_rate": 6.947819411632222e-09,
540
- "logits/chosen": -2.6895413398742676,
541
- "logits/rejected": -2.635143995285034,
542
- "logps/chosen": -0.9759384393692017,
543
- "logps/rejected": -0.96312016248703,
544
- "loss": 0.4451,
545
- "rewards/accuracies": 0.831250011920929,
546
- "rewards/chosen": -0.2653306722640991,
547
- "rewards/margins": 1.628769874572754,
548
- "rewards/rejected": -1.894100546836853,
549
  "step": 350
550
  },
551
  {
552
  "epoch": 0.87,
553
- "grad_norm": 1785.1282881305524,
554
- "learning_rate": 4.951556604879048e-09,
555
- "logits/chosen": -2.7073302268981934,
556
- "logits/rejected": -2.661701202392578,
557
- "logps/chosen": -0.9704357385635376,
558
- "logps/rejected": -1.0066497325897217,
559
- "loss": 0.4535,
560
  "rewards/accuracies": 0.7749999761581421,
561
- "rewards/chosen": -0.30233365297317505,
562
- "rewards/margins": 1.4092557430267334,
563
- "rewards/rejected": -1.7115894556045532,
564
  "step": 360
565
  },
566
  {
567
  "epoch": 0.89,
568
- "grad_norm": 821.6518799466353,
569
- "learning_rate": 3.278122536639888e-09,
570
- "logits/chosen": -2.7481675148010254,
571
- "logits/rejected": -2.692375898361206,
572
- "logps/chosen": -0.8963810205459595,
573
- "logps/rejected": -0.8914927244186401,
574
- "loss": 0.4351,
575
- "rewards/accuracies": 0.8374999761581421,
576
- "rewards/chosen": 0.013069706968963146,
577
- "rewards/margins": 1.6356351375579834,
578
- "rewards/rejected": -1.622565507888794,
579
  "step": 370
580
  },
581
  {
582
  "epoch": 0.92,
583
- "grad_norm": 820.7523726208289,
584
- "learning_rate": 1.9395094661033118e-09,
585
- "logits/chosen": -2.717128276824951,
586
- "logits/rejected": -2.693026542663574,
587
- "logps/chosen": -0.9069439172744751,
588
- "logps/rejected": -0.9684022068977356,
589
- "loss": 0.4381,
590
- "rewards/accuracies": 0.831250011920929,
591
- "rewards/chosen": -0.17060108482837677,
592
- "rewards/margins": 1.5492388010025024,
593
- "rewards/rejected": -1.7198398113250732,
594
  "step": 380
595
  },
596
  {
597
  "epoch": 0.94,
598
- "grad_norm": 791.5724729924565,
599
- "learning_rate": 9.453102390227175e-10,
600
- "logits/chosen": -2.694126605987549,
601
- "logits/rejected": -2.655355453491211,
602
- "logps/chosen": -0.9297587275505066,
603
- "logps/rejected": -0.9920668601989746,
604
- "loss": 0.4482,
605
- "rewards/accuracies": 0.856249988079071,
606
- "rewards/chosen": -0.0569925419986248,
607
- "rewards/margins": 1.6660388708114624,
608
- "rewards/rejected": -1.7230314016342163,
609
  "step": 390
610
  },
611
  {
612
  "epoch": 0.97,
613
- "grad_norm": 1054.605495809102,
614
- "learning_rate": 3.0264954291494007e-10,
615
- "logits/chosen": -2.7315666675567627,
616
- "logits/rejected": -2.6392362117767334,
617
- "logps/chosen": -0.970044732093811,
618
- "logps/rejected": -0.9285378456115723,
619
- "loss": 0.4633,
620
- "rewards/accuracies": 0.84375,
621
- "rewards/chosen": -0.0074133919551968575,
622
- "rewards/margins": 1.79477858543396,
623
- "rewards/rejected": -1.802191972732544,
624
  "step": 400
625
  },
626
  {
627
  "epoch": 0.99,
628
- "grad_norm": 1541.6219713183218,
629
- "learning_rate": 1.6132849715988494e-11,
630
- "logits/chosen": -2.7301852703094482,
631
- "logits/rejected": -2.6915435791015625,
632
- "logps/chosen": -0.9321584701538086,
633
- "logps/rejected": -0.9507132768630981,
634
- "loss": 0.4578,
635
- "rewards/accuracies": 0.8687499761581421,
636
- "rewards/chosen": -0.2989009618759155,
637
- "rewards/margins": 1.854230523109436,
638
- "rewards/rejected": -2.1531314849853516,
639
  "step": 410
640
  },
641
  {
642
  "epoch": 1.0,
643
  "step": 413,
644
  "total_flos": 0.0,
645
- "train_loss": 0.5277958785073232,
646
- "train_runtime": 6372.8883,
647
- "train_samples_per_second": 8.304,
648
  "train_steps_per_second": 0.065
649
  }
650
  ],
 
10
  "log_history": [
11
  {
12
  "epoch": 0.0,
13
+ "grad_norm": 6401.270603874373,
14
+ "learning_rate": 9.523809523809522e-09,
15
  "logits/chosen": -2.7005977630615234,
16
  "logits/rejected": -2.6288318634033203,
17
  "logps/chosen": -1.1158788204193115,
18
  "logps/rejected": -1.1333446502685547,
19
+ "loss": 0.7544,
20
  "rewards/accuracies": 0.0,
21
  "rewards/chosen": 0.0,
22
  "rewards/margins": 0.0,
 
25
  },
26
  {
27
  "epoch": 0.02,
28
+ "grad_norm": 9369.590990783972,
29
+ "learning_rate": 9.523809523809523e-08,
30
+ "logits/chosen": -2.76228666305542,
31
+ "logits/rejected": -2.6970374584198,
32
+ "logps/chosen": -0.837486743927002,
33
+ "logps/rejected": -0.8182350993156433,
34
+ "loss": 0.9695,
35
+ "rewards/accuracies": 0.4305555522441864,
36
+ "rewards/chosen": 0.06597563624382019,
37
+ "rewards/margins": 0.437710702419281,
38
+ "rewards/rejected": -0.3717350959777832,
39
  "step": 10
40
  },
41
  {
42
  "epoch": 0.05,
43
+ "grad_norm": 5966.657402243146,
44
+ "learning_rate": 1.9047619047619045e-07,
45
+ "logits/chosen": -2.6901049613952637,
46
+ "logits/rejected": -2.6502909660339355,
47
+ "logps/chosen": -0.9933319091796875,
48
+ "logps/rejected": -1.0394352674484253,
49
+ "loss": 1.0318,
50
+ "rewards/accuracies": 0.6499999761581421,
51
+ "rewards/chosen": -0.794396698474884,
52
+ "rewards/margins": 0.7471516728401184,
53
+ "rewards/rejected": -1.5415483713150024,
54
  "step": 20
55
  },
56
  {
57
  "epoch": 0.07,
58
+ "grad_norm": 8820.198504372876,
59
+ "learning_rate": 2.857142857142857e-07,
60
+ "logits/chosen": -2.7333264350891113,
61
+ "logits/rejected": -2.6793360710144043,
62
+ "logps/chosen": -0.9710652232170105,
63
+ "logps/rejected": -0.9799602627754211,
64
+ "loss": 1.3198,
65
+ "rewards/accuracies": 0.706250011920929,
66
+ "rewards/chosen": -2.275942325592041,
67
+ "rewards/margins": 0.9020620584487915,
68
+ "rewards/rejected": -3.178004264831543,
69
  "step": 30
70
  },
71
  {
72
  "epoch": 0.1,
73
+ "grad_norm": 8453.783513094899,
74
+ "learning_rate": 3.809523809523809e-07,
75
+ "logits/chosen": -2.6771621704101562,
76
+ "logits/rejected": -2.6321842670440674,
77
+ "logps/chosen": -0.989823043346405,
78
+ "logps/rejected": -0.9216930270195007,
79
+ "loss": 2.0555,
80
+ "rewards/accuracies": 0.762499988079071,
81
+ "rewards/chosen": -0.6586966514587402,
82
+ "rewards/margins": 5.100310325622559,
83
+ "rewards/rejected": -5.759006500244141,
84
  "step": 40
85
  },
86
  {
87
  "epoch": 0.12,
88
+ "grad_norm": 4192.139232222726,
89
+ "learning_rate": 3.995412608484087e-07,
90
+ "logits/chosen": -2.743403911590576,
91
+ "logits/rejected": -2.6878693103790283,
92
+ "logps/chosen": -0.9671042561531067,
93
+ "logps/rejected": -0.917597770690918,
94
+ "loss": 2.6495,
95
+ "rewards/accuracies": 0.793749988079071,
96
+ "rewards/chosen": -0.16885781288146973,
97
+ "rewards/margins": 5.186079978942871,
98
+ "rewards/rejected": -5.35493803024292,
99
  "step": 50
100
  },
101
  {
102
  "epoch": 0.15,
103
+ "grad_norm": 5643.860863524967,
104
+ "learning_rate": 3.976812391485896e-07,
105
+ "logits/chosen": -2.7438769340515137,
106
+ "logits/rejected": -2.676765203475952,
107
+ "logps/chosen": -0.911353588104248,
108
+ "logps/rejected": -0.9122518301010132,
109
+ "loss": 3.8047,
110
+ "rewards/accuracies": 0.8062499761581421,
111
+ "rewards/chosen": 2.4976494312286377,
112
+ "rewards/margins": 5.426072120666504,
113
+ "rewards/rejected": -2.928422212600708,
114
  "step": 60
115
  },
116
  {
117
  "epoch": 0.17,
118
+ "grad_norm": 4497.230754903385,
119
+ "learning_rate": 3.9440458281608213e-07,
120
+ "logits/chosen": -2.740940570831299,
121
+ "logits/rejected": -2.7162723541259766,
122
+ "logps/chosen": -0.9154363870620728,
123
+ "logps/rejected": -0.868497371673584,
124
+ "loss": 3.6432,
125
+ "rewards/accuracies": 0.824999988079071,
126
+ "rewards/chosen": 5.681364059448242,
127
+ "rewards/margins": 8.000432968139648,
128
+ "rewards/rejected": -2.319068431854248,
129
  "step": 70
130
  },
131
  {
132
  "epoch": 0.19,
133
+ "grad_norm": 9843.974275847575,
134
+ "learning_rate": 3.897347732134074e-07,
135
+ "logits/chosen": -2.679215908050537,
136
+ "logits/rejected": -2.625516891479492,
137
+ "logps/chosen": -0.9146322011947632,
138
+ "logps/rejected": -1.0181081295013428,
139
+ "loss": 5.767,
140
+ "rewards/accuracies": 0.800000011920929,
141
+ "rewards/chosen": -10.08639907836914,
142
+ "rewards/margins": 6.582289695739746,
143
+ "rewards/rejected": -16.668689727783203,
144
  "step": 80
145
  },
146
  {
147
  "epoch": 0.22,
148
+ "grad_norm": 4773.013380320505,
149
+ "learning_rate": 3.8370527539794614e-07,
150
+ "logits/chosen": -2.6771388053894043,
151
+ "logits/rejected": -2.6291418075561523,
152
+ "logps/chosen": -1.003847360610962,
153
+ "logps/rejected": -1.0297266244888306,
154
+ "loss": 4.6354,
155
+ "rewards/accuracies": 0.737500011920929,
156
+ "rewards/chosen": 4.863407611846924,
157
+ "rewards/margins": 9.78220272064209,
158
+ "rewards/rejected": -4.918795585632324,
159
  "step": 90
160
  },
161
  {
162
  "epoch": 0.24,
163
+ "grad_norm": 3074.8663144850243,
164
+ "learning_rate": 3.763592983027255e-07,
165
+ "logits/chosen": -2.705735683441162,
166
+ "logits/rejected": -2.6605448722839355,
167
+ "logps/chosen": -0.9163268804550171,
168
+ "logps/rejected": -0.9396775960922241,
169
+ "loss": 5.8585,
170
  "rewards/accuracies": 0.8062499761581421,
171
+ "rewards/chosen": -12.477940559387207,
172
+ "rewards/margins": 7.702305793762207,
173
+ "rewards/rejected": -20.180248260498047,
174
  "step": 100
175
  },
176
  {
177
  "epoch": 0.27,
178
+ "grad_norm": 4099.610429119441,
179
+ "learning_rate": 3.6774948509008527e-07,
180
+ "logits/chosen": -2.714970111846924,
181
+ "logits/rejected": -2.6705470085144043,
182
+ "logps/chosen": -0.9598251581192017,
183
+ "logps/rejected": -0.9319995641708374,
184
+ "loss": 5.1529,
185
+ "rewards/accuracies": 0.7875000238418579,
186
+ "rewards/chosen": 9.19798755645752,
187
+ "rewards/margins": 10.779365539550781,
188
+ "rewards/rejected": -1.5813770294189453,
189
  "step": 110
190
  },
191
  {
192
  "epoch": 0.29,
193
+ "grad_norm": 4599.711217449366,
194
+ "learning_rate": 3.579375358972288e-07,
195
+ "logits/chosen": -2.678779125213623,
196
+ "logits/rejected": -2.6315762996673584,
197
+ "logps/chosen": -0.9081487655639648,
198
+ "logps/rejected": -1.0060938596725464,
199
+ "loss": 4.0915,
200
+ "rewards/accuracies": 0.768750011920929,
201
+ "rewards/chosen": -13.663342475891113,
202
+ "rewards/margins": 9.755656242370605,
203
+ "rewards/rejected": -23.418994903564453,
204
  "step": 120
205
  },
206
  {
207
  "epoch": 0.31,
208
+ "grad_norm": 4010.334966061441,
209
+ "learning_rate": 3.4699376567716156e-07,
210
+ "logits/chosen": -2.7230353355407715,
211
+ "logits/rejected": -2.684389591217041,
212
+ "logps/chosen": -0.8652521967887878,
213
+ "logps/rejected": -0.8799147605895996,
214
+ "loss": 4.4027,
215
+ "rewards/accuracies": 0.762499988079071,
216
+ "rewards/chosen": 2.19469952583313,
217
+ "rewards/margins": 15.263641357421875,
218
+ "rewards/rejected": -13.068939208984375,
219
  "step": 130
220
  },
221
  {
222
  "epoch": 0.34,
223
+ "grad_norm": 5239.11146834966,
224
+ "learning_rate": 3.349966003036421e-07,
225
+ "logits/chosen": -2.689558506011963,
226
+ "logits/rejected": -2.649766445159912,
227
+ "logps/chosen": -0.9352903366088867,
228
+ "logps/rejected": -0.9416161775588989,
229
+ "loss": 4.7953,
230
+ "rewards/accuracies": 0.7749999761581421,
231
+ "rewards/chosen": -4.734063148498535,
232
+ "rewards/margins": 8.841203689575195,
233
+ "rewards/rejected": -13.575268745422363,
234
  "step": 140
235
  },
236
  {
237
  "epoch": 0.36,
238
+ "grad_norm": 5394.35498681908,
239
+ "learning_rate": 3.220320145511884e-07,
240
+ "logits/chosen": -2.7070841789245605,
241
+ "logits/rejected": -2.647737979888916,
242
+ "logps/chosen": -0.9441506266593933,
243
+ "logps/rejected": -0.9885166883468628,
244
+ "loss": 4.2219,
245
+ "rewards/accuracies": 0.793749988079071,
246
+ "rewards/chosen": 5.9402689933776855,
247
+ "rewards/margins": 12.97706413269043,
248
+ "rewards/rejected": -7.036795139312744,
249
  "step": 150
250
  },
251
  {
252
  "epoch": 0.39,
253
+ "grad_norm": 5022.189692479379,
254
+ "learning_rate": 3.0819291597771795e-07,
255
+ "logits/chosen": -2.7051825523376465,
256
+ "logits/rejected": -2.667494297027588,
257
+ "logps/chosen": -0.911395251750946,
258
+ "logps/rejected": -0.939487099647522,
259
+ "loss": 4.7963,
260
+ "rewards/accuracies": 0.856249988079071,
261
+ "rewards/chosen": -0.6114660501480103,
262
+ "rewards/margins": 9.443866729736328,
263
+ "rewards/rejected": -10.055331230163574,
264
  "step": 160
265
  },
266
  {
267
  "epoch": 0.41,
268
+ "grad_norm": 5428.944545727042,
269
+ "learning_rate": 2.9357847912507786e-07,
270
+ "logits/chosen": -2.6787288188934326,
271
+ "logits/rejected": -2.609421968460083,
272
+ "logps/chosen": -0.8976411819458008,
273
+ "logps/rejected": -0.8857674598693848,
274
+ "loss": 4.6262,
275
  "rewards/accuracies": 0.800000011920929,
276
+ "rewards/chosen": -2.82297945022583,
277
+ "rewards/margins": 10.640687942504883,
278
+ "rewards/rejected": -13.463666915893555,
279
  "step": 170
280
  },
281
  {
282
  "epoch": 0.44,
283
+ "grad_norm": 7317.882582449178,
284
+ "learning_rate": 2.7829343480875617e-07,
285
+ "logits/chosen": -2.6716930866241455,
286
+ "logits/rejected": -2.6018152236938477,
287
+ "logps/chosen": -0.9342878460884094,
288
+ "logps/rejected": -0.9536906480789185,
289
+ "loss": 4.5209,
290
+ "rewards/accuracies": 0.800000011920929,
291
+ "rewards/chosen": 4.173262596130371,
292
+ "rewards/margins": 8.933877944946289,
293
+ "rewards/rejected": -4.760615348815918,
294
  "step": 180
295
  },
296
  {
297
  "epoch": 0.46,
298
+ "grad_norm": 5046.2946182405685,
299
+ "learning_rate": 2.624473195899052e-07,
300
+ "logits/chosen": -2.737992763519287,
301
+ "logits/rejected": -2.7089955806732178,
302
+ "logps/chosen": -0.9629039764404297,
303
+ "logps/rejected": -1.039236307144165,
304
+ "loss": 4.5521,
305
+ "rewards/accuracies": 0.78125,
306
+ "rewards/chosen": -3.3569388389587402,
307
+ "rewards/margins": 13.995905876159668,
308
+ "rewards/rejected": -17.352848052978516,
309
  "step": 190
310
  },
311
  {
312
  "epoch": 0.48,
313
+ "grad_norm": 4268.8163809344915,
314
+ "learning_rate": 2.4615369080815547e-07,
315
+ "logits/chosen": -2.6982626914978027,
316
+ "logits/rejected": -2.6629488468170166,
317
+ "logps/chosen": -0.8523995280265808,
318
+ "logps/rejected": -0.9246847033500671,
319
+ "loss": 3.8184,
320
+ "rewards/accuracies": 0.78125,
321
+ "rewards/chosen": 2.3979854583740234,
322
+ "rewards/margins": 4.996596336364746,
323
+ "rewards/rejected": -2.5986106395721436,
324
  "step": 200
325
  },
326
  {
327
  "epoch": 0.51,
328
+ "grad_norm": 4211.242306423206,
329
+ "learning_rate": 2.2952931280049625e-07,
330
+ "logits/chosen": -2.7346115112304688,
331
+ "logits/rejected": -2.6734609603881836,
332
+ "logps/chosen": -1.0063531398773193,
333
+ "logps/rejected": -0.9570119976997375,
334
+ "loss": 4.9954,
335
+ "rewards/accuracies": 0.762499988079071,
336
+ "rewards/chosen": 7.86081600189209,
337
+ "rewards/margins": 13.075413703918457,
338
+ "rewards/rejected": -5.214597225189209,
339
  "step": 210
340
  },
341
  {
342
  "epoch": 0.53,
343
+ "grad_norm": 4906.2448320907815,
344
+ "learning_rate": 2.1269332013798747e-07,
345
+ "logits/chosen": -2.7431142330169678,
346
+ "logits/rejected": -2.7241249084472656,
347
+ "logps/chosen": -0.8835189938545227,
348
+ "logps/rejected": -0.8670462369918823,
349
+ "loss": 4.3795,
350
+ "rewards/accuracies": 0.78125,
351
+ "rewards/chosen": 4.582579135894775,
352
+ "rewards/margins": 8.150335311889648,
353
+ "rewards/rejected": -3.567755937576294,
354
  "step": 220
355
  },
356
  {
357
  "epoch": 0.56,
358
+ "grad_norm": 2995.5119741253625,
359
+ "learning_rate": 1.9576636387676436e-07,
360
+ "logits/chosen": -2.690732955932617,
361
+ "logits/rejected": -2.653067111968994,
362
+ "logps/chosen": -0.960831344127655,
363
+ "logps/rejected": -0.9556485414505005,
364
+ "loss": 4.0487,
365
+ "rewards/accuracies": 0.7562500238418579,
366
+ "rewards/chosen": -8.449748039245605,
367
+ "rewards/margins": 10.095115661621094,
368
+ "rewards/rejected": -18.544864654541016,
369
  "step": 230
370
  },
371
  {
372
  "epoch": 0.58,
373
+ "grad_norm": 3504.414329050279,
374
+ "learning_rate": 1.7886974694151976e-07,
375
+ "logits/chosen": -2.7119805812835693,
376
+ "logits/rejected": -2.6879172325134277,
377
+ "logps/chosen": -0.990290641784668,
378
+ "logps/rejected": -0.9934972524642944,
379
+ "loss": 4.3644,
380
+ "rewards/accuracies": 0.793749988079071,
381
+ "rewards/chosen": -0.5965616106987,
382
+ "rewards/margins": 9.211602210998535,
383
+ "rewards/rejected": -9.808164596557617,
384
  "step": 240
385
  },
386
  {
387
  "epoch": 0.6,
388
+ "grad_norm": 3400.5848210057216,
389
+ "learning_rate": 1.6212455483752895e-07,
390
+ "logits/chosen": -2.756906270980835,
391
+ "logits/rejected": -2.6796135902404785,
392
+ "logps/chosen": -0.8838168978691101,
393
+ "logps/rejected": -0.9137406349182129,
394
+ "loss": 4.5034,
395
+ "rewards/accuracies": 0.7875000238418579,
396
+ "rewards/chosen": 6.5281982421875,
397
+ "rewards/margins": 8.64702033996582,
398
+ "rewards/rejected": -2.1188230514526367,
399
  "step": 250
400
  },
401
  {
402
  "epoch": 0.63,
403
+ "grad_norm": 6194.117841583386,
404
+ "learning_rate": 1.4565078792075733e-07,
405
+ "logits/chosen": -2.7132773399353027,
406
+ "logits/rejected": -2.6494650840759277,
407
+ "logps/chosen": -1.002362847328186,
408
+ "logps/rejected": -0.9982520341873169,
409
+ "loss": 4.8134,
410
+ "rewards/accuracies": 0.84375,
411
+ "rewards/chosen": 3.0224878787994385,
412
+ "rewards/margins": 16.206506729125977,
413
+ "rewards/rejected": -13.1840181350708,
414
  "step": 260
415
  },
416
  {
417
  "epoch": 0.65,
418
+ "grad_norm": 4565.495892627232,
419
+ "learning_rate": 1.295665014444281e-07,
420
+ "logits/chosen": -2.7381529808044434,
421
+ "logits/rejected": -2.6608738899230957,
422
+ "logps/chosen": -0.9501218795776367,
423
+ "logps/rejected": -0.9476363062858582,
424
+ "loss": 5.3754,
425
+ "rewards/accuracies": 0.8500000238418579,
426
+ "rewards/chosen": -0.570526123046875,
427
+ "rewards/margins": 12.367398262023926,
428
+ "rewards/rejected": -12.9379243850708,
429
  "step": 270
430
  },
431
  {
432
  "epoch": 0.68,
433
+ "grad_norm": 5337.153187944306,
434
+ "learning_rate": 1.1398695954469597e-07,
435
+ "logits/chosen": -2.6872425079345703,
436
+ "logits/rejected": -2.630267381668091,
437
+ "logps/chosen": -0.9056104421615601,
438
+ "logps/rejected": -0.8939152956008911,
439
+ "loss": 4.1053,
440
+ "rewards/accuracies": 0.793749988079071,
441
+ "rewards/chosen": 4.294297218322754,
442
+ "rewards/margins": 7.472552299499512,
443
+ "rewards/rejected": -3.1782548427581787,
444
  "step": 280
445
  },
446
  {
447
  "epoch": 0.7,
448
+ "grad_norm": 3582.07962645892,
449
+ "learning_rate": 9.902380922818425e-08,
450
+ "logits/chosen": -2.7334370613098145,
451
+ "logits/rejected": -2.6919913291931152,
452
+ "logps/chosen": -0.9840775728225708,
453
+ "logps/rejected": -0.9756690263748169,
454
+ "loss": 3.2759,
455
+ "rewards/accuracies": 0.762499988079071,
456
+ "rewards/chosen": 8.966680526733398,
457
+ "rewards/margins": 11.496904373168945,
458
+ "rewards/rejected": -2.5302233695983887,
459
  "step": 290
460
  },
461
  {
462
  "epoch": 0.73,
463
+ "grad_norm": 4767.591882910886,
464
+ "learning_rate": 8.478428028080398e-08,
465
+ "logits/chosen": -2.7305169105529785,
466
+ "logits/rejected": -2.6773815155029297,
467
+ "logps/chosen": -0.8988749384880066,
468
+ "logps/rejected": -0.9437707662582397,
469
+ "loss": 4.3175,
470
+ "rewards/accuracies": 0.7875000238418579,
471
+ "rewards/chosen": 0.8898951411247253,
472
+ "rewards/margins": 8.447718620300293,
473
+ "rewards/rejected": -7.55782413482666,
474
  "step": 300
475
  },
476
  {
477
  "epoch": 0.75,
478
+ "grad_norm": 4819.380329592898,
479
+ "learning_rate": 7.137041683151202e-08,
480
+ "logits/chosen": -2.7228643894195557,
481
+ "logits/rejected": -2.6581058502197266,
482
+ "logps/chosen": -1.0781683921813965,
483
+ "logps/rejected": -1.028840184211731,
484
+ "loss": 2.9744,
485
  "rewards/accuracies": 0.8062499761581421,
486
+ "rewards/chosen": -2.4226202964782715,
487
+ "rewards/margins": 13.473236083984375,
488
+ "rewards/rejected": -15.895855903625488,
489
  "step": 310
490
  },
491
  {
492
  "epoch": 0.77,
493
+ "grad_norm": 7840.551721640683,
494
+ "learning_rate": 5.8878346077822135e-08,
495
+ "logits/chosen": -2.7280871868133545,
496
+ "logits/rejected": -2.649958848953247,
497
+ "logps/chosen": -0.9020591974258423,
498
+ "logps/rejected": -0.9361578822135925,
499
+ "loss": 2.7082,
500
+ "rewards/accuracies": 0.8812500238418579,
501
+ "rewards/chosen": 1.223115086555481,
502
+ "rewards/margins": 10.582406997680664,
503
+ "rewards/rejected": -9.35929012298584,
504
  "step": 320
505
  },
506
  {
507
  "epoch": 0.8,
508
+ "grad_norm": 4662.77535052248,
509
+ "learning_rate": 4.73975894135696e-08,
510
+ "logits/chosen": -2.6770853996276855,
511
+ "logits/rejected": -2.6099040508270264,
512
+ "logps/chosen": -0.9263202548027039,
513
+ "logps/rejected": -0.9608638882637024,
514
+ "loss": 3.1985,
515
+ "rewards/accuracies": 0.8125,
516
+ "rewards/chosen": 0.2365754395723343,
517
+ "rewards/margins": 13.195585250854492,
518
+ "rewards/rejected": -12.959010124206543,
519
  "step": 330
520
  },
521
  {
522
  "epoch": 0.82,
523
+ "grad_norm": 4550.588002339864,
524
+ "learning_rate": 3.701042089556483e-08,
525
+ "logits/chosen": -2.756493330001831,
526
+ "logits/rejected": -2.687851667404175,
527
+ "logps/chosen": -0.8901381492614746,
528
+ "logps/rejected": -0.9301478266716003,
529
+ "loss": 3.841,
530
+ "rewards/accuracies": 0.78125,
531
+ "rewards/chosen": 0.11963929980993271,
532
+ "rewards/margins": 7.3289618492126465,
533
+ "rewards/rejected": -7.209322929382324,
534
  "step": 340
535
  },
536
  {
537
  "epoch": 0.85,
538
+ "grad_norm": 5464.471487236709,
539
+ "learning_rate": 2.779127764652889e-08,
540
+ "logits/chosen": -2.689107656478882,
541
+ "logits/rejected": -2.6330015659332275,
542
+ "logps/chosen": -0.9756801724433899,
543
+ "logps/rejected": -0.9646003842353821,
544
+ "loss": 3.6421,
545
+ "rewards/accuracies": 0.8062499761581421,
546
+ "rewards/chosen": -1.1974527835845947,
547
+ "rewards/margins": 9.013090133666992,
548
+ "rewards/rejected": -10.210542678833008,
549
  "step": 350
550
  },
551
  {
552
  "epoch": 0.87,
553
+ "grad_norm": 5949.708940984834,
554
+ "learning_rate": 1.9806226419516193e-08,
555
+ "logits/chosen": -2.704460620880127,
556
+ "logits/rejected": -2.656071186065674,
557
+ "logps/chosen": -0.9623576402664185,
558
+ "logps/rejected": -1.0082406997680664,
559
+ "loss": 3.5231,
560
  "rewards/accuracies": 0.7749999761581421,
561
+ "rewards/chosen": 2.5273587703704834,
562
+ "rewards/margins": 11.88086223602295,
563
+ "rewards/rejected": -9.35350227355957,
564
  "step": 360
565
  },
566
  {
567
  "epoch": 0.89,
568
+ "grad_norm": 4320.933402478669,
569
+ "learning_rate": 1.3112490146559552e-08,
570
+ "logits/chosen": -2.7451281547546387,
571
+ "logits/rejected": -2.686728000640869,
572
+ "logps/chosen": -0.8951610326766968,
573
+ "logps/rejected": -0.89850914478302,
574
+ "loss": 3.0053,
575
+ "rewards/accuracies": 0.856249988079071,
576
+ "rewards/chosen": 0.6753175258636475,
577
+ "rewards/margins": 12.29626750946045,
578
+ "rewards/rejected": -11.620949745178223,
579
  "step": 370
580
  },
581
  {
582
  "epoch": 0.92,
583
+ "grad_norm": 2514.940389992379,
584
+ "learning_rate": 7.758037864413247e-09,
585
+ "logits/chosen": -2.7158432006835938,
586
+ "logits/rejected": -2.6906635761260986,
587
+ "logps/chosen": -0.9033122062683105,
588
+ "logps/rejected": -0.9709407091140747,
589
+ "loss": 2.8751,
590
+ "rewards/accuracies": 0.856249988079071,
591
+ "rewards/chosen": 0.9628832936286926,
592
+ "rewards/margins": 10.83133316040039,
593
+ "rewards/rejected": -9.868449211120605,
594
  "step": 380
595
  },
596
  {
597
  "epoch": 0.94,
598
+ "grad_norm": 3504.225752431698,
599
+ "learning_rate": 3.78124095609087e-09,
600
+ "logits/chosen": -2.6947999000549316,
601
+ "logits/rejected": -2.6553878784179688,
602
+ "logps/chosen": -0.9263744354248047,
603
+ "logps/rejected": -0.9935058355331421,
604
+ "loss": 3.019,
605
+ "rewards/accuracies": 0.862500011920929,
606
+ "rewards/chosen": 1.4072116613388062,
607
+ "rewards/margins": 10.741894721984863,
608
+ "rewards/rejected": -9.334683418273926,
609
  "step": 390
610
  },
611
  {
612
  "epoch": 0.97,
613
+ "grad_norm": 4987.634749508018,
614
+ "learning_rate": 1.2105981716597603e-09,
615
+ "logits/chosen": -2.7300946712493896,
616
+ "logits/rejected": -2.6389007568359375,
617
+ "logps/chosen": -0.9686774015426636,
618
+ "logps/rejected": -0.9328421354293823,
619
+ "loss": 3.7864,
620
+ "rewards/accuracies": 0.831250011920929,
621
+ "rewards/chosen": 0.6465551853179932,
622
+ "rewards/margins": 11.809611320495605,
623
+ "rewards/rejected": -11.163057327270508,
624
  "step": 400
625
  },
626
  {
627
  "epoch": 0.99,
628
+ "grad_norm": 5473.226219590305,
629
+ "learning_rate": 6.453139886395398e-11,
630
+ "logits/chosen": -2.7284317016601562,
631
+ "logits/rejected": -2.6886637210845947,
632
+ "logps/chosen": -0.9334842562675476,
633
+ "logps/rejected": -0.9600637555122375,
634
+ "loss": 3.6391,
635
+ "rewards/accuracies": 0.875,
636
+ "rewards/chosen": -2.157397747039795,
637
+ "rewards/margins": 13.2835054397583,
638
+ "rewards/rejected": -15.440902709960938,
639
  "step": 410
640
  },
641
  {
642
  "epoch": 1.0,
643
  "step": 413,
644
  "total_flos": 0.0,
645
+ "train_loss": 3.8421780889894426,
646
+ "train_runtime": 6381.4933,
647
+ "train_samples_per_second": 8.293,
648
  "train_steps_per_second": 0.065
649
  }
650
  ],
training_args.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:93c9cdc5d47cf48041862ca767083718bee40c0c9d421521bddf246b76bcabfa
3
  size 6264
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d13f3fc8759aca2ff57e3efc81099fc07a2315dc63120d75fb63db001a764cfe
3
  size 6264