yuyijiong commited on
Commit
4bc3655
·
verified ·
1 Parent(s): 4e030e2

Upload trainer_state.json with huggingface_hub

Browse files
Files changed (1) hide show
  1. trainer_state.json +1308 -0
trainer_state.json ADDED
@@ -0,0 +1,1308 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 1.997207114508305,
5
+ "eval_steps": 1,
6
+ "global_step": 850,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.023519035719535498,
13
+ "grad_norm": 4.1415791511535645,
14
+ "learning_rate": 3.060971262053483e-06,
15
+ "logits/chosen": -4.5,
16
+ "logits/rejected": -4.5,
17
+ "logps/chosen": -624.0,
18
+ "logps/rejected": -452.0,
19
+ "loss": 0.6933,
20
+ "rewards/accuracies": 0.08749999850988388,
21
+ "rewards/chosen": -0.007080078125,
22
+ "rewards/margins": -0.0089111328125,
23
+ "rewards/rejected": 0.00183868408203125,
24
+ "step": 10
25
+ },
26
+ {
27
+ "epoch": 0.047038071439070996,
28
+ "grad_norm": 3.7295737266540527,
29
+ "learning_rate": 3.9824154277970135e-06,
30
+ "logits/chosen": -4.53125,
31
+ "logits/rejected": -4.46875,
32
+ "logps/chosen": -552.0,
33
+ "logps/rejected": -422.0,
34
+ "loss": 0.6766,
35
+ "rewards/accuracies": 0.296875,
36
+ "rewards/chosen": 0.0263671875,
37
+ "rewards/margins": 0.044677734375,
38
+ "rewards/rejected": -0.018310546875,
39
+ "step": 20
40
+ },
41
+ {
42
+ "epoch": 0.0705571071586065,
43
+ "grad_norm": 3.65460205078125,
44
+ "learning_rate": 4.521425711265269e-06,
45
+ "logits/chosen": -4.46875,
46
+ "logits/rejected": -4.40625,
47
+ "logps/chosen": -592.0,
48
+ "logps/rejected": -436.0,
49
+ "loss": 0.6273,
50
+ "rewards/accuracies": 0.5718749761581421,
51
+ "rewards/chosen": 0.09423828125,
52
+ "rewards/margins": 0.158203125,
53
+ "rewards/rejected": -0.0634765625,
54
+ "step": 30
55
+ },
56
+ {
57
+ "epoch": 0.09407614287814199,
58
+ "grad_norm": 3.148322343826294,
59
+ "learning_rate": 4.903859593540544e-06,
60
+ "logits/chosen": -4.4375,
61
+ "logits/rejected": -4.40625,
62
+ "logps/chosen": -596.0,
63
+ "logps/rejected": -456.0,
64
+ "loss": 0.5544,
65
+ "rewards/accuracies": 0.75,
66
+ "rewards/chosen": 0.185546875,
67
+ "rewards/margins": 0.35546875,
68
+ "rewards/rejected": -0.169921875,
69
+ "step": 40
70
+ },
71
+ {
72
+ "epoch": 0.1175951785976775,
73
+ "grad_norm": 4.775664806365967,
74
+ "learning_rate": 4.9628252788104095e-06,
75
+ "logits/chosen": -4.4375,
76
+ "logits/rejected": -4.4375,
77
+ "logps/chosen": -576.0,
78
+ "logps/rejected": -444.0,
79
+ "loss": 0.4791,
80
+ "rewards/accuracies": 0.793749988079071,
81
+ "rewards/chosen": 0.2578125,
82
+ "rewards/margins": 0.55859375,
83
+ "rewards/rejected": -0.30078125,
84
+ "step": 50
85
+ },
86
+ {
87
+ "epoch": 0.141114214317213,
88
+ "grad_norm": 2.693018913269043,
89
+ "learning_rate": 4.9008674101610905e-06,
90
+ "logits/chosen": -4.5,
91
+ "logits/rejected": -4.5,
92
+ "logps/chosen": -584.0,
93
+ "logps/rejected": -444.0,
94
+ "loss": 0.4168,
95
+ "rewards/accuracies": 0.887499988079071,
96
+ "rewards/chosen": 0.39453125,
97
+ "rewards/margins": 0.87890625,
98
+ "rewards/rejected": -0.484375,
99
+ "step": 60
100
+ },
101
+ {
102
+ "epoch": 0.1646332500367485,
103
+ "grad_norm": 2.0642476081848145,
104
+ "learning_rate": 4.838909541511772e-06,
105
+ "logits/chosen": -4.4375,
106
+ "logits/rejected": -4.40625,
107
+ "logps/chosen": -596.0,
108
+ "logps/rejected": -454.0,
109
+ "loss": 0.3891,
110
+ "rewards/accuracies": 0.862500011920929,
111
+ "rewards/chosen": 0.388671875,
112
+ "rewards/margins": 1.0078125,
113
+ "rewards/rejected": -0.625,
114
+ "step": 70
115
+ },
116
+ {
117
+ "epoch": 0.18815228575628398,
118
+ "grad_norm": 2.0007693767547607,
119
+ "learning_rate": 4.776951672862453e-06,
120
+ "logits/chosen": -4.53125,
121
+ "logits/rejected": -4.5,
122
+ "logps/chosen": -600.0,
123
+ "logps/rejected": -492.0,
124
+ "loss": 0.3605,
125
+ "rewards/accuracies": 0.875,
126
+ "rewards/chosen": 0.49609375,
127
+ "rewards/margins": 1.3515625,
128
+ "rewards/rejected": -0.8515625,
129
+ "step": 80
130
+ },
131
+ {
132
+ "epoch": 0.2116713214758195,
133
+ "grad_norm": 2.3289802074432373,
134
+ "learning_rate": 4.714993804213135e-06,
135
+ "logits/chosen": -4.4375,
136
+ "logits/rejected": -4.4375,
137
+ "logps/chosen": -584.0,
138
+ "logps/rejected": -468.0,
139
+ "loss": 0.3619,
140
+ "rewards/accuracies": 0.84375,
141
+ "rewards/chosen": 0.396484375,
142
+ "rewards/margins": 1.34375,
143
+ "rewards/rejected": -0.9453125,
144
+ "step": 90
145
+ },
146
+ {
147
+ "epoch": 0.235190357195355,
148
+ "grad_norm": 2.2528398036956787,
149
+ "learning_rate": 4.653035935563817e-06,
150
+ "logits/chosen": -4.40625,
151
+ "logits/rejected": -4.46875,
152
+ "logps/chosen": -620.0,
153
+ "logps/rejected": -468.0,
154
+ "loss": 0.3334,
155
+ "rewards/accuracies": 0.831250011920929,
156
+ "rewards/chosen": 0.59765625,
157
+ "rewards/margins": 1.609375,
158
+ "rewards/rejected": -1.0078125,
159
+ "step": 100
160
+ },
161
+ {
162
+ "epoch": 0.2587093929148905,
163
+ "grad_norm": 1.8650789260864258,
164
+ "learning_rate": 4.591078066914498e-06,
165
+ "logits/chosen": -4.40625,
166
+ "logits/rejected": -4.375,
167
+ "logps/chosen": -596.0,
168
+ "logps/rejected": -416.0,
169
+ "loss": 0.2843,
170
+ "rewards/accuracies": 0.8999999761581421,
171
+ "rewards/chosen": 0.55859375,
172
+ "rewards/margins": 1.8125,
173
+ "rewards/rejected": -1.25,
174
+ "step": 110
175
+ },
176
+ {
177
+ "epoch": 0.282228428634426,
178
+ "grad_norm": 2.2233972549438477,
179
+ "learning_rate": 4.52912019826518e-06,
180
+ "logits/chosen": -4.40625,
181
+ "logits/rejected": -4.4375,
182
+ "logps/chosen": -596.0,
183
+ "logps/rejected": -476.0,
184
+ "loss": 0.2863,
185
+ "rewards/accuracies": 0.8812500238418579,
186
+ "rewards/chosen": 0.63671875,
187
+ "rewards/margins": 1.8984375,
188
+ "rewards/rejected": -1.2578125,
189
+ "step": 120
190
+ },
191
+ {
192
+ "epoch": 0.3057474643539615,
193
+ "grad_norm": 2.017451286315918,
194
+ "learning_rate": 4.467162329615862e-06,
195
+ "logits/chosen": -4.46875,
196
+ "logits/rejected": -4.5,
197
+ "logps/chosen": -612.0,
198
+ "logps/rejected": -448.0,
199
+ "loss": 0.2741,
200
+ "rewards/accuracies": 0.887499988079071,
201
+ "rewards/chosen": 0.64453125,
202
+ "rewards/margins": 1.9453125,
203
+ "rewards/rejected": -1.3046875,
204
+ "step": 130
205
+ },
206
+ {
207
+ "epoch": 0.329266500073497,
208
+ "grad_norm": 2.127619743347168,
209
+ "learning_rate": 4.405204460966543e-06,
210
+ "logits/chosen": -4.40625,
211
+ "logits/rejected": -4.40625,
212
+ "logps/chosen": -616.0,
213
+ "logps/rejected": -454.0,
214
+ "loss": 0.2705,
215
+ "rewards/accuracies": 0.8843749761581421,
216
+ "rewards/chosen": 0.8828125,
217
+ "rewards/margins": 2.265625,
218
+ "rewards/rejected": -1.375,
219
+ "step": 140
220
+ },
221
+ {
222
+ "epoch": 0.3527855357930325,
223
+ "grad_norm": 1.8466318845748901,
224
+ "learning_rate": 4.343246592317225e-06,
225
+ "logits/chosen": -4.46875,
226
+ "logits/rejected": -4.40625,
227
+ "logps/chosen": -624.0,
228
+ "logps/rejected": -492.0,
229
+ "loss": 0.2757,
230
+ "rewards/accuracies": 0.887499988079071,
231
+ "rewards/chosen": 0.84375,
232
+ "rewards/margins": 2.28125,
233
+ "rewards/rejected": -1.4375,
234
+ "step": 150
235
+ },
236
+ {
237
+ "epoch": 0.37630457151256796,
238
+ "grad_norm": 1.8905202150344849,
239
+ "learning_rate": 4.2812887236679065e-06,
240
+ "logits/chosen": -4.40625,
241
+ "logits/rejected": -4.375,
242
+ "logps/chosen": -556.0,
243
+ "logps/rejected": -450.0,
244
+ "loss": 0.2506,
245
+ "rewards/accuracies": 0.8999999761581421,
246
+ "rewards/chosen": 0.94140625,
247
+ "rewards/margins": 2.328125,
248
+ "rewards/rejected": -1.390625,
249
+ "step": 160
250
+ },
251
+ {
252
+ "epoch": 0.3998236072321035,
253
+ "grad_norm": 1.59074866771698,
254
+ "learning_rate": 4.219330855018588e-06,
255
+ "logits/chosen": -4.53125,
256
+ "logits/rejected": -4.4375,
257
+ "logps/chosen": -576.0,
258
+ "logps/rejected": -452.0,
259
+ "loss": 0.2633,
260
+ "rewards/accuracies": 0.887499988079071,
261
+ "rewards/chosen": 0.9453125,
262
+ "rewards/margins": 2.328125,
263
+ "rewards/rejected": -1.390625,
264
+ "step": 170
265
+ },
266
+ {
267
+ "epoch": 0.423342642951639,
268
+ "grad_norm": 1.7788141965866089,
269
+ "learning_rate": 4.157372986369269e-06,
270
+ "logits/chosen": -4.4375,
271
+ "logits/rejected": -4.4375,
272
+ "logps/chosen": -600.0,
273
+ "logps/rejected": -500.0,
274
+ "loss": 0.2619,
275
+ "rewards/accuracies": 0.859375,
276
+ "rewards/chosen": 0.9765625,
277
+ "rewards/margins": 2.375,
278
+ "rewards/rejected": -1.3984375,
279
+ "step": 180
280
+ },
281
+ {
282
+ "epoch": 0.4468616786711745,
283
+ "grad_norm": 2.1362152099609375,
284
+ "learning_rate": 4.095415117719951e-06,
285
+ "logits/chosen": -4.4375,
286
+ "logits/rejected": -4.46875,
287
+ "logps/chosen": -592.0,
288
+ "logps/rejected": -472.0,
289
+ "loss": 0.2743,
290
+ "rewards/accuracies": 0.918749988079071,
291
+ "rewards/chosen": 1.09375,
292
+ "rewards/margins": 2.578125,
293
+ "rewards/rejected": -1.4921875,
294
+ "step": 190
295
+ },
296
+ {
297
+ "epoch": 0.47038071439071,
298
+ "grad_norm": 1.8583593368530273,
299
+ "learning_rate": 4.033457249070632e-06,
300
+ "logits/chosen": -4.46875,
301
+ "logits/rejected": -4.40625,
302
+ "logps/chosen": -588.0,
303
+ "logps/rejected": -460.0,
304
+ "loss": 0.2359,
305
+ "rewards/accuracies": 0.9125000238418579,
306
+ "rewards/chosen": 1.2265625,
307
+ "rewards/margins": 2.640625,
308
+ "rewards/rejected": -1.4140625,
309
+ "step": 200
310
+ },
311
+ {
312
+ "epoch": 0.4938997501102455,
313
+ "grad_norm": 2.6427791118621826,
314
+ "learning_rate": 3.971499380421314e-06,
315
+ "logits/chosen": -4.46875,
316
+ "logits/rejected": -4.5,
317
+ "logps/chosen": -568.0,
318
+ "logps/rejected": -456.0,
319
+ "loss": 0.2418,
320
+ "rewards/accuracies": 0.8968750238418579,
321
+ "rewards/chosen": 1.2578125,
322
+ "rewards/margins": 2.4375,
323
+ "rewards/rejected": -1.1796875,
324
+ "step": 210
325
+ },
326
+ {
327
+ "epoch": 0.517418785829781,
328
+ "grad_norm": 1.6700748205184937,
329
+ "learning_rate": 3.909541511771995e-06,
330
+ "logits/chosen": -4.4375,
331
+ "logits/rejected": -4.4375,
332
+ "logps/chosen": -580.0,
333
+ "logps/rejected": -442.0,
334
+ "loss": 0.2332,
335
+ "rewards/accuracies": 0.8968750238418579,
336
+ "rewards/chosen": 1.140625,
337
+ "rewards/margins": 2.8125,
338
+ "rewards/rejected": -1.671875,
339
+ "step": 220
340
+ },
341
+ {
342
+ "epoch": 0.5409378215493165,
343
+ "grad_norm": 2.4413912296295166,
344
+ "learning_rate": 3.847583643122677e-06,
345
+ "logits/chosen": -4.5,
346
+ "logits/rejected": -4.5,
347
+ "logps/chosen": -612.0,
348
+ "logps/rejected": -484.0,
349
+ "loss": 0.2338,
350
+ "rewards/accuracies": 0.9125000238418579,
351
+ "rewards/chosen": 1.3125,
352
+ "rewards/margins": 2.765625,
353
+ "rewards/rejected": -1.453125,
354
+ "step": 230
355
+ },
356
+ {
357
+ "epoch": 0.564456857268852,
358
+ "grad_norm": 1.718529462814331,
359
+ "learning_rate": 3.7856257744733583e-06,
360
+ "logits/chosen": -4.46875,
361
+ "logits/rejected": -4.46875,
362
+ "logps/chosen": -592.0,
363
+ "logps/rejected": -432.0,
364
+ "loss": 0.2044,
365
+ "rewards/accuracies": 0.925000011920929,
366
+ "rewards/chosen": 1.5546875,
367
+ "rewards/margins": 3.09375,
368
+ "rewards/rejected": -1.5390625,
369
+ "step": 240
370
+ },
371
+ {
372
+ "epoch": 0.5879758929883875,
373
+ "grad_norm": 2.0037829875946045,
374
+ "learning_rate": 3.7236679058240398e-06,
375
+ "logits/chosen": -4.4375,
376
+ "logits/rejected": -4.40625,
377
+ "logps/chosen": -584.0,
378
+ "logps/rejected": -458.0,
379
+ "loss": 0.221,
380
+ "rewards/accuracies": 0.8999999761581421,
381
+ "rewards/chosen": 1.375,
382
+ "rewards/margins": 3.03125,
383
+ "rewards/rejected": -1.6484375,
384
+ "step": 250
385
+ },
386
+ {
387
+ "epoch": 0.611494928707923,
388
+ "grad_norm": 1.673599362373352,
389
+ "learning_rate": 3.6617100371747216e-06,
390
+ "logits/chosen": -4.53125,
391
+ "logits/rejected": -4.5,
392
+ "logps/chosen": -564.0,
393
+ "logps/rejected": -468.0,
394
+ "loss": 0.2305,
395
+ "rewards/accuracies": 0.9125000238418579,
396
+ "rewards/chosen": 1.453125,
397
+ "rewards/margins": 2.96875,
398
+ "rewards/rejected": -1.515625,
399
+ "step": 260
400
+ },
401
+ {
402
+ "epoch": 0.6350139644274585,
403
+ "grad_norm": 1.8723180294036865,
404
+ "learning_rate": 3.5997521685254026e-06,
405
+ "logits/chosen": -4.46875,
406
+ "logits/rejected": -4.46875,
407
+ "logps/chosen": -564.0,
408
+ "logps/rejected": -430.0,
409
+ "loss": 0.2233,
410
+ "rewards/accuracies": 0.921875,
411
+ "rewards/chosen": 1.4296875,
412
+ "rewards/margins": 3.0625,
413
+ "rewards/rejected": -1.6328125,
414
+ "step": 270
415
+ },
416
+ {
417
+ "epoch": 0.658533000146994,
418
+ "grad_norm": 1.8895870447158813,
419
+ "learning_rate": 3.5377942998760845e-06,
420
+ "logits/chosen": -4.4375,
421
+ "logits/rejected": -4.4375,
422
+ "logps/chosen": -584.0,
423
+ "logps/rejected": -468.0,
424
+ "loss": 0.2191,
425
+ "rewards/accuracies": 0.909375011920929,
426
+ "rewards/chosen": 1.5234375,
427
+ "rewards/margins": 3.21875,
428
+ "rewards/rejected": -1.6953125,
429
+ "step": 280
430
+ },
431
+ {
432
+ "epoch": 0.6820520358665295,
433
+ "grad_norm": 1.5127581357955933,
434
+ "learning_rate": 3.4758364312267663e-06,
435
+ "logits/chosen": -4.40625,
436
+ "logits/rejected": -4.40625,
437
+ "logps/chosen": -596.0,
438
+ "logps/rejected": -450.0,
439
+ "loss": 0.213,
440
+ "rewards/accuracies": 0.9125000238418579,
441
+ "rewards/chosen": 1.4609375,
442
+ "rewards/margins": 2.9375,
443
+ "rewards/rejected": -1.46875,
444
+ "step": 290
445
+ },
446
+ {
447
+ "epoch": 0.705571071586065,
448
+ "grad_norm": 2.202705144882202,
449
+ "learning_rate": 3.4138785625774478e-06,
450
+ "logits/chosen": -4.46875,
451
+ "logits/rejected": -4.40625,
452
+ "logps/chosen": -576.0,
453
+ "logps/rejected": -468.0,
454
+ "loss": 0.2195,
455
+ "rewards/accuracies": 0.8999999761581421,
456
+ "rewards/chosen": 1.578125,
457
+ "rewards/margins": 3.0,
458
+ "rewards/rejected": -1.421875,
459
+ "step": 300
460
+ },
461
+ {
462
+ "epoch": 0.7290901073056004,
463
+ "grad_norm": 1.8141247034072876,
464
+ "learning_rate": 3.351920693928129e-06,
465
+ "logits/chosen": -4.40625,
466
+ "logits/rejected": -4.4375,
467
+ "logps/chosen": -560.0,
468
+ "logps/rejected": -436.0,
469
+ "loss": 0.2075,
470
+ "rewards/accuracies": 0.9156249761581421,
471
+ "rewards/chosen": 1.578125,
472
+ "rewards/margins": 3.1875,
473
+ "rewards/rejected": -1.609375,
474
+ "step": 310
475
+ },
476
+ {
477
+ "epoch": 0.7526091430251359,
478
+ "grad_norm": 1.7854108810424805,
479
+ "learning_rate": 3.2899628252788106e-06,
480
+ "logits/chosen": -4.34375,
481
+ "logits/rejected": -4.375,
482
+ "logps/chosen": -616.0,
483
+ "logps/rejected": -462.0,
484
+ "loss": 0.2124,
485
+ "rewards/accuracies": 0.8999999761581421,
486
+ "rewards/chosen": 1.5859375,
487
+ "rewards/margins": 3.203125,
488
+ "rewards/rejected": -1.6171875,
489
+ "step": 320
490
+ },
491
+ {
492
+ "epoch": 0.7761281787446714,
493
+ "grad_norm": 1.5797982215881348,
494
+ "learning_rate": 3.2280049566294925e-06,
495
+ "logits/chosen": -4.4375,
496
+ "logits/rejected": -4.46875,
497
+ "logps/chosen": -620.0,
498
+ "logps/rejected": -496.0,
499
+ "loss": 0.2276,
500
+ "rewards/accuracies": 0.9312499761581421,
501
+ "rewards/chosen": 1.7734375,
502
+ "rewards/margins": 3.515625,
503
+ "rewards/rejected": -1.7421875,
504
+ "step": 330
505
+ },
506
+ {
507
+ "epoch": 0.799647214464207,
508
+ "grad_norm": 1.9662494659423828,
509
+ "learning_rate": 3.1660470879801735e-06,
510
+ "logits/chosen": -4.46875,
511
+ "logits/rejected": -4.4375,
512
+ "logps/chosen": -568.0,
513
+ "logps/rejected": -456.0,
514
+ "loss": 0.2161,
515
+ "rewards/accuracies": 0.903124988079071,
516
+ "rewards/chosen": 1.65625,
517
+ "rewards/margins": 3.0625,
518
+ "rewards/rejected": -1.3984375,
519
+ "step": 340
520
+ },
521
+ {
522
+ "epoch": 0.8231662501837425,
523
+ "grad_norm": 1.7154417037963867,
524
+ "learning_rate": 3.1040892193308553e-06,
525
+ "logits/chosen": -4.4375,
526
+ "logits/rejected": -4.4375,
527
+ "logps/chosen": -572.0,
528
+ "logps/rejected": -462.0,
529
+ "loss": 0.1925,
530
+ "rewards/accuracies": 0.909375011920929,
531
+ "rewards/chosen": 1.6640625,
532
+ "rewards/margins": 3.15625,
533
+ "rewards/rejected": -1.4921875,
534
+ "step": 350
535
+ },
536
+ {
537
+ "epoch": 0.846685285903278,
538
+ "grad_norm": 2.1363542079925537,
539
+ "learning_rate": 3.042131350681537e-06,
540
+ "logits/chosen": -4.40625,
541
+ "logits/rejected": -4.375,
542
+ "logps/chosen": -520.0,
543
+ "logps/rejected": -410.0,
544
+ "loss": 0.225,
545
+ "rewards/accuracies": 0.9156249761581421,
546
+ "rewards/chosen": 1.4609375,
547
+ "rewards/margins": 3.171875,
548
+ "rewards/rejected": -1.703125,
549
+ "step": 360
550
+ },
551
+ {
552
+ "epoch": 0.8702043216228135,
553
+ "grad_norm": 1.9305567741394043,
554
+ "learning_rate": 2.980173482032218e-06,
555
+ "logits/chosen": -4.5,
556
+ "logits/rejected": -4.5,
557
+ "logps/chosen": -572.0,
558
+ "logps/rejected": -452.0,
559
+ "loss": 0.1921,
560
+ "rewards/accuracies": 0.918749988079071,
561
+ "rewards/chosen": 1.671875,
562
+ "rewards/margins": 3.453125,
563
+ "rewards/rejected": -1.7890625,
564
+ "step": 370
565
+ },
566
+ {
567
+ "epoch": 0.893723357342349,
568
+ "grad_norm": 1.7296861410140991,
569
+ "learning_rate": 2.9182156133829e-06,
570
+ "logits/chosen": -4.46875,
571
+ "logits/rejected": -4.5,
572
+ "logps/chosen": -592.0,
573
+ "logps/rejected": -450.0,
574
+ "loss": 0.2142,
575
+ "rewards/accuracies": 0.918749988079071,
576
+ "rewards/chosen": 1.59375,
577
+ "rewards/margins": 3.296875,
578
+ "rewards/rejected": -1.703125,
579
+ "step": 380
580
+ },
581
+ {
582
+ "epoch": 0.9172423930618845,
583
+ "grad_norm": 1.9843934774398804,
584
+ "learning_rate": 2.8562577447335815e-06,
585
+ "logits/chosen": -4.4375,
586
+ "logits/rejected": -4.46875,
587
+ "logps/chosen": -576.0,
588
+ "logps/rejected": -426.0,
589
+ "loss": 0.213,
590
+ "rewards/accuracies": 0.9281250238418579,
591
+ "rewards/chosen": 1.625,
592
+ "rewards/margins": 3.390625,
593
+ "rewards/rejected": -1.7578125,
594
+ "step": 390
595
+ },
596
+ {
597
+ "epoch": 0.94076142878142,
598
+ "grad_norm": 2.4459681510925293,
599
+ "learning_rate": 2.7942998760842625e-06,
600
+ "logits/chosen": -4.375,
601
+ "logits/rejected": -4.34375,
602
+ "logps/chosen": -576.0,
603
+ "logps/rejected": -496.0,
604
+ "loss": 0.1966,
605
+ "rewards/accuracies": 0.9312499761581421,
606
+ "rewards/chosen": 1.75,
607
+ "rewards/margins": 3.515625,
608
+ "rewards/rejected": -1.765625,
609
+ "step": 400
610
+ },
611
+ {
612
+ "epoch": 0.9642804645009555,
613
+ "grad_norm": 2.3723385334014893,
614
+ "learning_rate": 2.7323420074349443e-06,
615
+ "logits/chosen": -4.40625,
616
+ "logits/rejected": -4.40625,
617
+ "logps/chosen": -556.0,
618
+ "logps/rejected": -452.0,
619
+ "loss": 0.2096,
620
+ "rewards/accuracies": 0.887499988079071,
621
+ "rewards/chosen": 1.5703125,
622
+ "rewards/margins": 3.234375,
623
+ "rewards/rejected": -1.671875,
624
+ "step": 410
625
+ },
626
+ {
627
+ "epoch": 0.987799500220491,
628
+ "grad_norm": 2.0075254440307617,
629
+ "learning_rate": 2.670384138785626e-06,
630
+ "logits/chosen": -4.375,
631
+ "logits/rejected": -4.40625,
632
+ "logps/chosen": -588.0,
633
+ "logps/rejected": -492.0,
634
+ "loss": 0.219,
635
+ "rewards/accuracies": 0.890625,
636
+ "rewards/chosen": 1.7265625,
637
+ "rewards/margins": 3.4375,
638
+ "rewards/rejected": -1.7109375,
639
+ "step": 420
640
+ },
641
+ {
642
+ "epoch": 1.0094076142878141,
643
+ "grad_norm": 0.9782338738441467,
644
+ "learning_rate": 2.6146220570012394e-06,
645
+ "logits/chosen": -4.4375,
646
+ "logits/rejected": -4.40625,
647
+ "logps/chosen": -576.0,
648
+ "logps/rejected": -480.0,
649
+ "loss": 0.1502,
650
+ "rewards/accuracies": 0.9319728016853333,
651
+ "rewards/chosen": 1.78125,
652
+ "rewards/margins": 3.625,
653
+ "rewards/rejected": -1.84375,
654
+ "step": 430
655
+ },
656
+ {
657
+ "epoch": 1.0329266500073497,
658
+ "grad_norm": 0.8791906833648682,
659
+ "learning_rate": 2.552664188351921e-06,
660
+ "logits/chosen": -4.4375,
661
+ "logits/rejected": -4.5,
662
+ "logps/chosen": -568.0,
663
+ "logps/rejected": -436.0,
664
+ "loss": 0.1162,
665
+ "rewards/accuracies": 0.971875011920929,
666
+ "rewards/chosen": 1.890625,
667
+ "rewards/margins": 4.125,
668
+ "rewards/rejected": -2.25,
669
+ "step": 440
670
+ },
671
+ {
672
+ "epoch": 1.0564456857268851,
673
+ "grad_norm": 0.9551910161972046,
674
+ "learning_rate": 2.4907063197026023e-06,
675
+ "logits/chosen": -4.5,
676
+ "logits/rejected": -4.5,
677
+ "logps/chosen": -568.0,
678
+ "logps/rejected": -434.0,
679
+ "loss": 0.1221,
680
+ "rewards/accuracies": 0.956250011920929,
681
+ "rewards/chosen": 1.8359375,
682
+ "rewards/margins": 3.84375,
683
+ "rewards/rejected": -2.015625,
684
+ "step": 450
685
+ },
686
+ {
687
+ "epoch": 1.0799647214464208,
688
+ "grad_norm": 1.0174229145050049,
689
+ "learning_rate": 2.4287484510532837e-06,
690
+ "logits/chosen": -4.5,
691
+ "logits/rejected": -4.53125,
692
+ "logps/chosen": -576.0,
693
+ "logps/rejected": -446.0,
694
+ "loss": 0.122,
695
+ "rewards/accuracies": 0.9624999761581421,
696
+ "rewards/chosen": 1.953125,
697
+ "rewards/margins": 4.0,
698
+ "rewards/rejected": -2.046875,
699
+ "step": 460
700
+ },
701
+ {
702
+ "epoch": 1.1034837571659561,
703
+ "grad_norm": 1.0029239654541016,
704
+ "learning_rate": 2.3667905824039656e-06,
705
+ "logits/chosen": -4.4375,
706
+ "logits/rejected": -4.46875,
707
+ "logps/chosen": -592.0,
708
+ "logps/rejected": -458.0,
709
+ "loss": 0.1126,
710
+ "rewards/accuracies": 0.971875011920929,
711
+ "rewards/chosen": 1.921875,
712
+ "rewards/margins": 4.21875,
713
+ "rewards/rejected": -2.296875,
714
+ "step": 470
715
+ },
716
+ {
717
+ "epoch": 1.1270027928854918,
718
+ "grad_norm": 1.4849553108215332,
719
+ "learning_rate": 2.304832713754647e-06,
720
+ "logits/chosen": -4.4375,
721
+ "logits/rejected": -4.375,
722
+ "logps/chosen": -600.0,
723
+ "logps/rejected": -504.0,
724
+ "loss": 0.1161,
725
+ "rewards/accuracies": 0.9781249761581421,
726
+ "rewards/chosen": 2.09375,
727
+ "rewards/margins": 4.40625,
728
+ "rewards/rejected": -2.3125,
729
+ "step": 480
730
+ },
731
+ {
732
+ "epoch": 1.1505218286050272,
733
+ "grad_norm": 0.9113921523094177,
734
+ "learning_rate": 2.242874845105329e-06,
735
+ "logits/chosen": -4.46875,
736
+ "logits/rejected": -4.40625,
737
+ "logps/chosen": -584.0,
738
+ "logps/rejected": -474.0,
739
+ "loss": 0.1067,
740
+ "rewards/accuracies": 0.965624988079071,
741
+ "rewards/chosen": 1.921875,
742
+ "rewards/margins": 4.125,
743
+ "rewards/rejected": -2.203125,
744
+ "step": 490
745
+ },
746
+ {
747
+ "epoch": 1.1740408643245628,
748
+ "grad_norm": 1.4038139581680298,
749
+ "learning_rate": 2.1809169764560103e-06,
750
+ "logits/chosen": -4.4375,
751
+ "logits/rejected": -4.40625,
752
+ "logps/chosen": -596.0,
753
+ "logps/rejected": -450.0,
754
+ "loss": 0.1208,
755
+ "rewards/accuracies": 0.96875,
756
+ "rewards/chosen": 1.9375,
757
+ "rewards/margins": 4.21875,
758
+ "rewards/rejected": -2.265625,
759
+ "step": 500
760
+ },
761
+ {
762
+ "epoch": 1.1975599000440982,
763
+ "grad_norm": 0.9754643440246582,
764
+ "learning_rate": 2.1189591078066917e-06,
765
+ "logits/chosen": -4.40625,
766
+ "logits/rejected": -4.375,
767
+ "logps/chosen": -584.0,
768
+ "logps/rejected": -456.0,
769
+ "loss": 0.1145,
770
+ "rewards/accuracies": 0.9624999761581421,
771
+ "rewards/chosen": 1.8515625,
772
+ "rewards/margins": 4.0625,
773
+ "rewards/rejected": -2.203125,
774
+ "step": 510
775
+ },
776
+ {
777
+ "epoch": 1.2210789357636336,
778
+ "grad_norm": 0.7279557585716248,
779
+ "learning_rate": 2.057001239157373e-06,
780
+ "logits/chosen": -4.40625,
781
+ "logits/rejected": -4.375,
782
+ "logps/chosen": -588.0,
783
+ "logps/rejected": -460.0,
784
+ "loss": 0.111,
785
+ "rewards/accuracies": 0.984375,
786
+ "rewards/chosen": 1.765625,
787
+ "rewards/margins": 4.21875,
788
+ "rewards/rejected": -2.46875,
789
+ "step": 520
790
+ },
791
+ {
792
+ "epoch": 1.2445979714831692,
793
+ "grad_norm": 1.1293213367462158,
794
+ "learning_rate": 1.9950433705080545e-06,
795
+ "logits/chosen": -4.4375,
796
+ "logits/rejected": -4.46875,
797
+ "logps/chosen": -572.0,
798
+ "logps/rejected": -428.0,
799
+ "loss": 0.1192,
800
+ "rewards/accuracies": 0.981249988079071,
801
+ "rewards/chosen": 1.9921875,
802
+ "rewards/margins": 4.28125,
803
+ "rewards/rejected": -2.296875,
804
+ "step": 530
805
+ },
806
+ {
807
+ "epoch": 1.2681170072027048,
808
+ "grad_norm": 0.7407346963882446,
809
+ "learning_rate": 1.9330855018587364e-06,
810
+ "logits/chosen": -4.46875,
811
+ "logits/rejected": -4.40625,
812
+ "logps/chosen": -596.0,
813
+ "logps/rejected": -456.0,
814
+ "loss": 0.1107,
815
+ "rewards/accuracies": 0.965624988079071,
816
+ "rewards/chosen": 2.015625,
817
+ "rewards/margins": 4.25,
818
+ "rewards/rejected": -2.25,
819
+ "step": 540
820
+ },
821
+ {
822
+ "epoch": 1.2916360429222402,
823
+ "grad_norm": 1.1689890623092651,
824
+ "learning_rate": 1.8711276332094178e-06,
825
+ "logits/chosen": -4.46875,
826
+ "logits/rejected": -4.4375,
827
+ "logps/chosen": -600.0,
828
+ "logps/rejected": -472.0,
829
+ "loss": 0.116,
830
+ "rewards/accuracies": 0.9468749761581421,
831
+ "rewards/chosen": 1.828125,
832
+ "rewards/margins": 4.21875,
833
+ "rewards/rejected": -2.375,
834
+ "step": 550
835
+ },
836
+ {
837
+ "epoch": 1.3151550786417756,
838
+ "grad_norm": 1.0410897731781006,
839
+ "learning_rate": 1.8091697645600993e-06,
840
+ "logits/chosen": -4.46875,
841
+ "logits/rejected": -4.46875,
842
+ "logps/chosen": -584.0,
843
+ "logps/rejected": -436.0,
844
+ "loss": 0.1166,
845
+ "rewards/accuracies": 0.965624988079071,
846
+ "rewards/chosen": 1.9375,
847
+ "rewards/margins": 4.34375,
848
+ "rewards/rejected": -2.40625,
849
+ "step": 560
850
+ },
851
+ {
852
+ "epoch": 1.3386741143613112,
853
+ "grad_norm": 0.8815547227859497,
854
+ "learning_rate": 1.7472118959107809e-06,
855
+ "logits/chosen": -4.4375,
856
+ "logits/rejected": -4.40625,
857
+ "logps/chosen": -580.0,
858
+ "logps/rejected": -482.0,
859
+ "loss": 0.1003,
860
+ "rewards/accuracies": 0.953125,
861
+ "rewards/chosen": 1.96875,
862
+ "rewards/margins": 4.25,
863
+ "rewards/rejected": -2.28125,
864
+ "step": 570
865
+ },
866
+ {
867
+ "epoch": 1.3621931500808466,
868
+ "grad_norm": 1.0066012144088745,
869
+ "learning_rate": 1.6852540272614623e-06,
870
+ "logits/chosen": -4.4375,
871
+ "logits/rejected": -4.40625,
872
+ "logps/chosen": -576.0,
873
+ "logps/rejected": -516.0,
874
+ "loss": 0.1079,
875
+ "rewards/accuracies": 0.956250011920929,
876
+ "rewards/chosen": 1.8671875,
877
+ "rewards/margins": 4.03125,
878
+ "rewards/rejected": -2.15625,
879
+ "step": 580
880
+ },
881
+ {
882
+ "epoch": 1.3857121858003822,
883
+ "grad_norm": 1.2409735918045044,
884
+ "learning_rate": 1.623296158612144e-06,
885
+ "logits/chosen": -4.4375,
886
+ "logits/rejected": -4.46875,
887
+ "logps/chosen": -600.0,
888
+ "logps/rejected": -464.0,
889
+ "loss": 0.1281,
890
+ "rewards/accuracies": 0.9750000238418579,
891
+ "rewards/chosen": 1.890625,
892
+ "rewards/margins": 4.71875,
893
+ "rewards/rejected": -2.828125,
894
+ "step": 590
895
+ },
896
+ {
897
+ "epoch": 1.4092312215199176,
898
+ "grad_norm": 1.059398889541626,
899
+ "learning_rate": 1.5613382899628254e-06,
900
+ "logits/chosen": -4.5,
901
+ "logits/rejected": -4.5,
902
+ "logps/chosen": -568.0,
903
+ "logps/rejected": -452.0,
904
+ "loss": 0.1181,
905
+ "rewards/accuracies": 0.953125,
906
+ "rewards/chosen": 1.921875,
907
+ "rewards/margins": 4.34375,
908
+ "rewards/rejected": -2.421875,
909
+ "step": 600
910
+ },
911
+ {
912
+ "epoch": 1.4327502572394533,
913
+ "grad_norm": 0.8506196141242981,
914
+ "learning_rate": 1.4993804213135068e-06,
915
+ "logits/chosen": -4.4375,
916
+ "logits/rejected": -4.4375,
917
+ "logps/chosen": -576.0,
918
+ "logps/rejected": -450.0,
919
+ "loss": 0.1021,
920
+ "rewards/accuracies": 0.96875,
921
+ "rewards/chosen": 1.921875,
922
+ "rewards/margins": 4.40625,
923
+ "rewards/rejected": -2.484375,
924
+ "step": 610
925
+ },
926
+ {
927
+ "epoch": 1.4562692929589887,
928
+ "grad_norm": 1.0724049806594849,
929
+ "learning_rate": 1.4374225526641887e-06,
930
+ "logits/chosen": -4.46875,
931
+ "logits/rejected": -4.5,
932
+ "logps/chosen": -576.0,
933
+ "logps/rejected": -462.0,
934
+ "loss": 0.1089,
935
+ "rewards/accuracies": 0.9624999761581421,
936
+ "rewards/chosen": 1.796875,
937
+ "rewards/margins": 4.21875,
938
+ "rewards/rejected": -2.40625,
939
+ "step": 620
940
+ },
941
+ {
942
+ "epoch": 1.479788328678524,
943
+ "grad_norm": 0.8412142992019653,
944
+ "learning_rate": 1.37546468401487e-06,
945
+ "logits/chosen": -4.40625,
946
+ "logits/rejected": -4.4375,
947
+ "logps/chosen": -576.0,
948
+ "logps/rejected": -440.0,
949
+ "loss": 0.1178,
950
+ "rewards/accuracies": 0.953125,
951
+ "rewards/chosen": 1.9140625,
952
+ "rewards/margins": 4.09375,
953
+ "rewards/rejected": -2.1875,
954
+ "step": 630
955
+ },
956
+ {
957
+ "epoch": 1.5033073643980597,
958
+ "grad_norm": 1.080023169517517,
959
+ "learning_rate": 1.3135068153655513e-06,
960
+ "logits/chosen": -4.40625,
961
+ "logits/rejected": -4.4375,
962
+ "logps/chosen": -600.0,
963
+ "logps/rejected": -452.0,
964
+ "loss": 0.1097,
965
+ "rewards/accuracies": 0.9624999761581421,
966
+ "rewards/chosen": 2.0,
967
+ "rewards/margins": 4.5625,
968
+ "rewards/rejected": -2.546875,
969
+ "step": 640
970
+ },
971
+ {
972
+ "epoch": 1.5268264001175953,
973
+ "grad_norm": 1.2255281209945679,
974
+ "learning_rate": 1.2515489467162332e-06,
975
+ "logits/chosen": -4.375,
976
+ "logits/rejected": -4.40625,
977
+ "logps/chosen": -572.0,
978
+ "logps/rejected": -456.0,
979
+ "loss": 0.1078,
980
+ "rewards/accuracies": 0.971875011920929,
981
+ "rewards/chosen": 1.96875,
982
+ "rewards/margins": 4.40625,
983
+ "rewards/rejected": -2.421875,
984
+ "step": 650
985
+ },
986
+ {
987
+ "epoch": 1.5503454358371307,
988
+ "grad_norm": 0.9179627299308777,
989
+ "learning_rate": 1.1895910780669146e-06,
990
+ "logits/chosen": -4.40625,
991
+ "logits/rejected": -4.4375,
992
+ "logps/chosen": -568.0,
993
+ "logps/rejected": -444.0,
994
+ "loss": 0.1105,
995
+ "rewards/accuracies": 0.9750000238418579,
996
+ "rewards/chosen": 1.890625,
997
+ "rewards/margins": 4.28125,
998
+ "rewards/rejected": -2.40625,
999
+ "step": 660
1000
+ },
1001
+ {
1002
+ "epoch": 1.573864471556666,
1003
+ "grad_norm": 1.0199528932571411,
1004
+ "learning_rate": 1.127633209417596e-06,
1005
+ "logits/chosen": -4.46875,
1006
+ "logits/rejected": -4.46875,
1007
+ "logps/chosen": -548.0,
1008
+ "logps/rejected": -450.0,
1009
+ "loss": 0.1132,
1010
+ "rewards/accuracies": 0.965624988079071,
1011
+ "rewards/chosen": 1.8984375,
1012
+ "rewards/margins": 4.21875,
1013
+ "rewards/rejected": -2.3125,
1014
+ "step": 670
1015
+ },
1016
+ {
1017
+ "epoch": 1.5973835072762017,
1018
+ "grad_norm": 1.190191626548767,
1019
+ "learning_rate": 1.0656753407682777e-06,
1020
+ "logits/chosen": -4.5,
1021
+ "logits/rejected": -4.53125,
1022
+ "logps/chosen": -596.0,
1023
+ "logps/rejected": -460.0,
1024
+ "loss": 0.1178,
1025
+ "rewards/accuracies": 0.965624988079071,
1026
+ "rewards/chosen": 1.9375,
1027
+ "rewards/margins": 4.3125,
1028
+ "rewards/rejected": -2.359375,
1029
+ "step": 680
1030
+ },
1031
+ {
1032
+ "epoch": 1.6209025429957373,
1033
+ "grad_norm": 1.029616355895996,
1034
+ "learning_rate": 1.0037174721189593e-06,
1035
+ "logits/chosen": -4.5625,
1036
+ "logits/rejected": -4.59375,
1037
+ "logps/chosen": -552.0,
1038
+ "logps/rejected": -450.0,
1039
+ "loss": 0.1087,
1040
+ "rewards/accuracies": 0.9593750238418579,
1041
+ "rewards/chosen": 1.8515625,
1042
+ "rewards/margins": 4.375,
1043
+ "rewards/rejected": -2.515625,
1044
+ "step": 690
1045
+ },
1046
+ {
1047
+ "epoch": 1.6444215787152727,
1048
+ "grad_norm": 1.2294584512710571,
1049
+ "learning_rate": 9.417596034696406e-07,
1050
+ "logits/chosen": -4.40625,
1051
+ "logits/rejected": -4.40625,
1052
+ "logps/chosen": -576.0,
1053
+ "logps/rejected": -444.0,
1054
+ "loss": 0.1049,
1055
+ "rewards/accuracies": 0.984375,
1056
+ "rewards/chosen": 1.8671875,
1057
+ "rewards/margins": 4.21875,
1058
+ "rewards/rejected": -2.34375,
1059
+ "step": 700
1060
+ },
1061
+ {
1062
+ "epoch": 1.6679406144348081,
1063
+ "grad_norm": 1.0543959140777588,
1064
+ "learning_rate": 8.798017348203223e-07,
1065
+ "logits/chosen": -4.4375,
1066
+ "logits/rejected": -4.5,
1067
+ "logps/chosen": -580.0,
1068
+ "logps/rejected": -458.0,
1069
+ "loss": 0.1063,
1070
+ "rewards/accuracies": 0.965624988079071,
1071
+ "rewards/chosen": 1.9609375,
1072
+ "rewards/margins": 4.40625,
1073
+ "rewards/rejected": -2.46875,
1074
+ "step": 710
1075
+ },
1076
+ {
1077
+ "epoch": 1.6914596501543437,
1078
+ "grad_norm": 1.0166019201278687,
1079
+ "learning_rate": 8.178438661710038e-07,
1080
+ "logits/chosen": -4.4375,
1081
+ "logits/rejected": -4.5,
1082
+ "logps/chosen": -560.0,
1083
+ "logps/rejected": -448.0,
1084
+ "loss": 0.1014,
1085
+ "rewards/accuracies": 0.9624999761581421,
1086
+ "rewards/chosen": 1.8828125,
1087
+ "rewards/margins": 4.4375,
1088
+ "rewards/rejected": -2.5625,
1089
+ "step": 720
1090
+ },
1091
+ {
1092
+ "epoch": 1.7149786858738791,
1093
+ "grad_norm": 1.0257729291915894,
1094
+ "learning_rate": 7.558859975216853e-07,
1095
+ "logits/chosen": -4.40625,
1096
+ "logits/rejected": -4.375,
1097
+ "logps/chosen": -604.0,
1098
+ "logps/rejected": -524.0,
1099
+ "loss": 0.1245,
1100
+ "rewards/accuracies": 0.9593750238418579,
1101
+ "rewards/chosen": 1.921875,
1102
+ "rewards/margins": 4.46875,
1103
+ "rewards/rejected": -2.546875,
1104
+ "step": 730
1105
+ },
1106
+ {
1107
+ "epoch": 1.7384977215934145,
1108
+ "grad_norm": 1.058556318283081,
1109
+ "learning_rate": 6.939281288723669e-07,
1110
+ "logits/chosen": -4.40625,
1111
+ "logits/rejected": -4.40625,
1112
+ "logps/chosen": -588.0,
1113
+ "logps/rejected": -468.0,
1114
+ "loss": 0.1131,
1115
+ "rewards/accuracies": 0.953125,
1116
+ "rewards/chosen": 1.90625,
1117
+ "rewards/margins": 4.09375,
1118
+ "rewards/rejected": -2.1875,
1119
+ "step": 740
1120
+ },
1121
+ {
1122
+ "epoch": 1.7620167573129502,
1123
+ "grad_norm": 1.1007747650146484,
1124
+ "learning_rate": 6.319702602230483e-07,
1125
+ "logits/chosen": -4.40625,
1126
+ "logits/rejected": -4.375,
1127
+ "logps/chosen": -620.0,
1128
+ "logps/rejected": -510.0,
1129
+ "loss": 0.0978,
1130
+ "rewards/accuracies": 0.971875011920929,
1131
+ "rewards/chosen": 1.9453125,
1132
+ "rewards/margins": 4.71875,
1133
+ "rewards/rejected": -2.765625,
1134
+ "step": 750
1135
+ },
1136
+ {
1137
+ "epoch": 1.7855357930324858,
1138
+ "grad_norm": 1.0114929676055908,
1139
+ "learning_rate": 5.700123915737299e-07,
1140
+ "logits/chosen": -4.46875,
1141
+ "logits/rejected": -4.4375,
1142
+ "logps/chosen": -584.0,
1143
+ "logps/rejected": -472.0,
1144
+ "loss": 0.1013,
1145
+ "rewards/accuracies": 0.981249988079071,
1146
+ "rewards/chosen": 2.015625,
1147
+ "rewards/margins": 4.53125,
1148
+ "rewards/rejected": -2.53125,
1149
+ "step": 760
1150
+ },
1151
+ {
1152
+ "epoch": 1.8090548287520212,
1153
+ "grad_norm": 0.9178210496902466,
1154
+ "learning_rate": 5.080545229244115e-07,
1155
+ "logits/chosen": -4.40625,
1156
+ "logits/rejected": -4.4375,
1157
+ "logps/chosen": -600.0,
1158
+ "logps/rejected": -470.0,
1159
+ "loss": 0.1082,
1160
+ "rewards/accuracies": 0.965624988079071,
1161
+ "rewards/chosen": 1.9375,
1162
+ "rewards/margins": 4.5,
1163
+ "rewards/rejected": -2.5625,
1164
+ "step": 770
1165
+ },
1166
+ {
1167
+ "epoch": 1.8325738644715566,
1168
+ "grad_norm": 0.7771146893501282,
1169
+ "learning_rate": 4.4609665427509294e-07,
1170
+ "logits/chosen": -4.375,
1171
+ "logits/rejected": -4.4375,
1172
+ "logps/chosen": -600.0,
1173
+ "logps/rejected": -456.0,
1174
+ "loss": 0.1043,
1175
+ "rewards/accuracies": 0.9750000238418579,
1176
+ "rewards/chosen": 1.765625,
1177
+ "rewards/margins": 4.5625,
1178
+ "rewards/rejected": -2.796875,
1179
+ "step": 780
1180
+ },
1181
+ {
1182
+ "epoch": 1.8560929001910922,
1183
+ "grad_norm": 0.9903047680854797,
1184
+ "learning_rate": 3.8413878562577453e-07,
1185
+ "logits/chosen": -4.40625,
1186
+ "logits/rejected": -4.4375,
1187
+ "logps/chosen": -560.0,
1188
+ "logps/rejected": -462.0,
1189
+ "loss": 0.1029,
1190
+ "rewards/accuracies": 0.9750000238418579,
1191
+ "rewards/chosen": 1.875,
1192
+ "rewards/margins": 4.4375,
1193
+ "rewards/rejected": -2.5625,
1194
+ "step": 790
1195
+ },
1196
+ {
1197
+ "epoch": 1.8796119359106278,
1198
+ "grad_norm": 1.1646795272827148,
1199
+ "learning_rate": 3.22180916976456e-07,
1200
+ "logits/chosen": -4.5,
1201
+ "logits/rejected": -4.46875,
1202
+ "logps/chosen": -572.0,
1203
+ "logps/rejected": -444.0,
1204
+ "loss": 0.1103,
1205
+ "rewards/accuracies": 0.965624988079071,
1206
+ "rewards/chosen": 2.078125,
1207
+ "rewards/margins": 4.53125,
1208
+ "rewards/rejected": -2.4375,
1209
+ "step": 800
1210
+ },
1211
+ {
1212
+ "epoch": 1.9031309716301632,
1213
+ "grad_norm": 0.7632185220718384,
1214
+ "learning_rate": 2.6022304832713755e-07,
1215
+ "logits/chosen": -4.4375,
1216
+ "logits/rejected": -4.40625,
1217
+ "logps/chosen": -584.0,
1218
+ "logps/rejected": -438.0,
1219
+ "loss": 0.0989,
1220
+ "rewards/accuracies": 0.9624999761581421,
1221
+ "rewards/chosen": 1.9140625,
1222
+ "rewards/margins": 4.53125,
1223
+ "rewards/rejected": -2.625,
1224
+ "step": 810
1225
+ },
1226
+ {
1227
+ "epoch": 1.9266500073496986,
1228
+ "grad_norm": 0.9221560955047607,
1229
+ "learning_rate": 1.982651796778191e-07,
1230
+ "logits/chosen": -4.53125,
1231
+ "logits/rejected": -4.53125,
1232
+ "logps/chosen": -592.0,
1233
+ "logps/rejected": -462.0,
1234
+ "loss": 0.1141,
1235
+ "rewards/accuracies": 0.9593750238418579,
1236
+ "rewards/chosen": 1.921875,
1237
+ "rewards/margins": 4.40625,
1238
+ "rewards/rejected": -2.484375,
1239
+ "step": 820
1240
+ },
1241
+ {
1242
+ "epoch": 1.9501690430692342,
1243
+ "grad_norm": 1.339996576309204,
1244
+ "learning_rate": 1.3630731102850064e-07,
1245
+ "logits/chosen": -4.4375,
1246
+ "logits/rejected": -4.4375,
1247
+ "logps/chosen": -596.0,
1248
+ "logps/rejected": -454.0,
1249
+ "loss": 0.1099,
1250
+ "rewards/accuracies": 0.9750000238418579,
1251
+ "rewards/chosen": 1.8203125,
1252
+ "rewards/margins": 4.46875,
1253
+ "rewards/rejected": -2.640625,
1254
+ "step": 830
1255
+ },
1256
+ {
1257
+ "epoch": 1.9736880787887696,
1258
+ "grad_norm": 1.047699213027954,
1259
+ "learning_rate": 7.434944237918216e-08,
1260
+ "logits/chosen": -4.40625,
1261
+ "logits/rejected": -4.4375,
1262
+ "logps/chosen": -592.0,
1263
+ "logps/rejected": -446.0,
1264
+ "loss": 0.1054,
1265
+ "rewards/accuracies": 0.96875,
1266
+ "rewards/chosen": 1.890625,
1267
+ "rewards/margins": 4.75,
1268
+ "rewards/rejected": -2.875,
1269
+ "step": 840
1270
+ },
1271
+ {
1272
+ "epoch": 1.997207114508305,
1273
+ "grad_norm": 0.8556466698646545,
1274
+ "learning_rate": 1.2391573729863694e-08,
1275
+ "logits/chosen": -4.46875,
1276
+ "logits/rejected": -4.4375,
1277
+ "logps/chosen": -600.0,
1278
+ "logps/rejected": -516.0,
1279
+ "loss": 0.1072,
1280
+ "rewards/accuracies": 0.965624988079071,
1281
+ "rewards/chosen": 1.90625,
1282
+ "rewards/margins": 4.53125,
1283
+ "rewards/rejected": -2.625,
1284
+ "step": 850
1285
+ }
1286
+ ],
1287
+ "logging_steps": 10,
1288
+ "max_steps": 850,
1289
+ "num_input_tokens_seen": 0,
1290
+ "num_train_epochs": 2,
1291
+ "save_steps": 250,
1292
+ "stateful_callbacks": {
1293
+ "TrainerControl": {
1294
+ "args": {
1295
+ "should_epoch_stop": false,
1296
+ "should_evaluate": false,
1297
+ "should_log": false,
1298
+ "should_save": true,
1299
+ "should_training_stop": true
1300
+ },
1301
+ "attributes": {}
1302
+ }
1303
+ },
1304
+ "total_flos": 0.0,
1305
+ "train_batch_size": 1,
1306
+ "trial_name": null,
1307
+ "trial_params": null
1308
+ }