yuyijiong commited on
Commit
269fd94
·
verified ·
1 Parent(s): 8d4ccd4

Upload trainer_state.json with huggingface_hub

Browse files
Files changed (1) hide show
  1. trainer_state.json +1293 -0
trainer_state.json ADDED
@@ -0,0 +1,1293 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 1.998381162619573,
5
+ "eval_steps": 1,
6
+ "global_step": 848,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.0235467255334805,
13
+ "grad_norm": 4.103935241699219,
14
+ "learning_rate": 3.060971262053483e-06,
15
+ "logits/chosen": -4.40625,
16
+ "logits/rejected": -4.4375,
17
+ "logps/chosen": -652.0,
18
+ "logps/rejected": -592.0,
19
+ "loss": 0.6951,
20
+ "rewards/accuracies": 0.08124999701976776,
21
+ "rewards/chosen": -0.0006256103515625,
22
+ "rewards/margins": -0.0089111328125,
23
+ "rewards/rejected": 0.00830078125,
24
+ "step": 10
25
+ },
26
+ {
27
+ "epoch": 0.047093451066961,
28
+ "grad_norm": 3.962085247039795,
29
+ "learning_rate": 3.9824154277970135e-06,
30
+ "logits/chosen": -4.5,
31
+ "logits/rejected": -4.5,
32
+ "logps/chosen": -592.0,
33
+ "logps/rejected": -520.0,
34
+ "loss": 0.6868,
35
+ "rewards/accuracies": 0.20000000298023224,
36
+ "rewards/chosen": 0.0006256103515625,
37
+ "rewards/margins": 0.00830078125,
38
+ "rewards/rejected": -0.007659912109375,
39
+ "step": 20
40
+ },
41
+ {
42
+ "epoch": 0.0706401766004415,
43
+ "grad_norm": 4.004490852355957,
44
+ "learning_rate": 4.521425711265269e-06,
45
+ "logits/chosen": -4.46875,
46
+ "logits/rejected": -4.5,
47
+ "logps/chosen": -608.0,
48
+ "logps/rejected": -524.0,
49
+ "loss": 0.6591,
50
+ "rewards/accuracies": 0.3812499940395355,
51
+ "rewards/chosen": 0.053955078125,
52
+ "rewards/margins": 0.076171875,
53
+ "rewards/rejected": -0.0220947265625,
54
+ "step": 30
55
+ },
56
+ {
57
+ "epoch": 0.094186902133922,
58
+ "grad_norm": 3.424196481704712,
59
+ "learning_rate": 4.903859593540544e-06,
60
+ "logits/chosen": -4.4375,
61
+ "logits/rejected": -4.40625,
62
+ "logps/chosen": -584.0,
63
+ "logps/rejected": -504.0,
64
+ "loss": 0.6295,
65
+ "rewards/accuracies": 0.5625,
66
+ "rewards/chosen": 0.09521484375,
67
+ "rewards/margins": 0.1455078125,
68
+ "rewards/rejected": -0.05078125,
69
+ "step": 40
70
+ },
71
+ {
72
+ "epoch": 0.1177336276674025,
73
+ "grad_norm": 3.229916572570801,
74
+ "learning_rate": 4.962732919254658e-06,
75
+ "logits/chosen": -4.5625,
76
+ "logits/rejected": -4.53125,
77
+ "logps/chosen": -548.0,
78
+ "logps/rejected": -502.0,
79
+ "loss": 0.58,
80
+ "rewards/accuracies": 0.6499999761581421,
81
+ "rewards/chosen": 0.15234375,
82
+ "rewards/margins": 0.28125,
83
+ "rewards/rejected": -0.12890625,
84
+ "step": 50
85
+ },
86
+ {
87
+ "epoch": 0.141280353200883,
88
+ "grad_norm": 3.3798811435699463,
89
+ "learning_rate": 4.900621118012423e-06,
90
+ "logits/chosen": -4.40625,
91
+ "logits/rejected": -4.40625,
92
+ "logps/chosen": -580.0,
93
+ "logps/rejected": -454.0,
94
+ "loss": 0.5537,
95
+ "rewards/accuracies": 0.7250000238418579,
96
+ "rewards/chosen": 0.302734375,
97
+ "rewards/margins": 0.423828125,
98
+ "rewards/rejected": -0.12158203125,
99
+ "step": 60
100
+ },
101
+ {
102
+ "epoch": 0.1648270787343635,
103
+ "grad_norm": 3.2893781661987305,
104
+ "learning_rate": 4.8385093167701866e-06,
105
+ "logits/chosen": -4.40625,
106
+ "logits/rejected": -4.40625,
107
+ "logps/chosen": -608.0,
108
+ "logps/rejected": -508.0,
109
+ "loss": 0.5351,
110
+ "rewards/accuracies": 0.7250000238418579,
111
+ "rewards/chosen": 0.251953125,
112
+ "rewards/margins": 0.4609375,
113
+ "rewards/rejected": -0.208984375,
114
+ "step": 70
115
+ },
116
+ {
117
+ "epoch": 0.188373804267844,
118
+ "grad_norm": 3.3645501136779785,
119
+ "learning_rate": 4.776397515527951e-06,
120
+ "logits/chosen": -4.5,
121
+ "logits/rejected": -4.53125,
122
+ "logps/chosen": -588.0,
123
+ "logps/rejected": -468.0,
124
+ "loss": 0.5096,
125
+ "rewards/accuracies": 0.7749999761581421,
126
+ "rewards/chosen": 0.34375,
127
+ "rewards/margins": 0.6796875,
128
+ "rewards/rejected": -0.3359375,
129
+ "step": 80
130
+ },
131
+ {
132
+ "epoch": 0.2119205298013245,
133
+ "grad_norm": 3.14998197555542,
134
+ "learning_rate": 4.714285714285715e-06,
135
+ "logits/chosen": -4.40625,
136
+ "logits/rejected": -4.4375,
137
+ "logps/chosen": -632.0,
138
+ "logps/rejected": -490.0,
139
+ "loss": 0.4901,
140
+ "rewards/accuracies": 0.7749999761581421,
141
+ "rewards/chosen": 0.375,
142
+ "rewards/margins": 0.69140625,
143
+ "rewards/rejected": -0.318359375,
144
+ "step": 90
145
+ },
146
+ {
147
+ "epoch": 0.235467255334805,
148
+ "grad_norm": 3.2294790744781494,
149
+ "learning_rate": 4.652173913043478e-06,
150
+ "logits/chosen": -4.5,
151
+ "logits/rejected": -4.5,
152
+ "logps/chosen": -580.0,
153
+ "logps/rejected": -506.0,
154
+ "loss": 0.4736,
155
+ "rewards/accuracies": 0.762499988079071,
156
+ "rewards/chosen": 0.369140625,
157
+ "rewards/margins": 0.6328125,
158
+ "rewards/rejected": -0.263671875,
159
+ "step": 100
160
+ },
161
+ {
162
+ "epoch": 0.25901398086828553,
163
+ "grad_norm": 2.738917827606201,
164
+ "learning_rate": 4.590062111801243e-06,
165
+ "logits/chosen": -4.5,
166
+ "logits/rejected": -4.59375,
167
+ "logps/chosen": -604.0,
168
+ "logps/rejected": -516.0,
169
+ "loss": 0.4703,
170
+ "rewards/accuracies": 0.762499988079071,
171
+ "rewards/chosen": 0.5,
172
+ "rewards/margins": 0.9296875,
173
+ "rewards/rejected": -0.431640625,
174
+ "step": 110
175
+ },
176
+ {
177
+ "epoch": 0.282560706401766,
178
+ "grad_norm": 3.060237169265747,
179
+ "learning_rate": 4.527950310559007e-06,
180
+ "logits/chosen": -4.40625,
181
+ "logits/rejected": -4.40625,
182
+ "logps/chosen": -632.0,
183
+ "logps/rejected": -552.0,
184
+ "loss": 0.4649,
185
+ "rewards/accuracies": 0.7437499761581421,
186
+ "rewards/chosen": 0.53515625,
187
+ "rewards/margins": 0.99609375,
188
+ "rewards/rejected": -0.462890625,
189
+ "step": 120
190
+ },
191
+ {
192
+ "epoch": 0.30610743193524653,
193
+ "grad_norm": 3.196960926055908,
194
+ "learning_rate": 4.46583850931677e-06,
195
+ "logits/chosen": -4.34375,
196
+ "logits/rejected": -4.34375,
197
+ "logps/chosen": -644.0,
198
+ "logps/rejected": -524.0,
199
+ "loss": 0.4824,
200
+ "rewards/accuracies": 0.762499988079071,
201
+ "rewards/chosen": 0.54296875,
202
+ "rewards/margins": 1.0,
203
+ "rewards/rejected": -0.45703125,
204
+ "step": 130
205
+ },
206
+ {
207
+ "epoch": 0.329654157468727,
208
+ "grad_norm": 3.1361074447631836,
209
+ "learning_rate": 4.4037267080745344e-06,
210
+ "logits/chosen": -4.375,
211
+ "logits/rejected": -4.40625,
212
+ "logps/chosen": -628.0,
213
+ "logps/rejected": -548.0,
214
+ "loss": 0.4539,
215
+ "rewards/accuracies": 0.7749999761581421,
216
+ "rewards/chosen": 0.447265625,
217
+ "rewards/margins": 0.9921875,
218
+ "rewards/rejected": -0.546875,
219
+ "step": 140
220
+ },
221
+ {
222
+ "epoch": 0.35320088300220753,
223
+ "grad_norm": 2.6170785427093506,
224
+ "learning_rate": 4.3416149068322985e-06,
225
+ "logits/chosen": -4.46875,
226
+ "logits/rejected": -4.4375,
227
+ "logps/chosen": -580.0,
228
+ "logps/rejected": -512.0,
229
+ "loss": 0.4478,
230
+ "rewards/accuracies": 0.800000011920929,
231
+ "rewards/chosen": 0.65625,
232
+ "rewards/margins": 1.0390625,
233
+ "rewards/rejected": -0.3828125,
234
+ "step": 150
235
+ },
236
+ {
237
+ "epoch": 0.376747608535688,
238
+ "grad_norm": 2.7828519344329834,
239
+ "learning_rate": 4.279503105590063e-06,
240
+ "logits/chosen": -4.375,
241
+ "logits/rejected": -4.34375,
242
+ "logps/chosen": -608.0,
243
+ "logps/rejected": -506.0,
244
+ "loss": 0.4407,
245
+ "rewards/accuracies": 0.7875000238418579,
246
+ "rewards/chosen": 0.5859375,
247
+ "rewards/margins": 1.09375,
248
+ "rewards/rejected": -0.5078125,
249
+ "step": 160
250
+ },
251
+ {
252
+ "epoch": 0.40029433406916853,
253
+ "grad_norm": 2.906342029571533,
254
+ "learning_rate": 4.217391304347827e-06,
255
+ "logits/chosen": -4.5,
256
+ "logits/rejected": -4.4375,
257
+ "logps/chosen": -584.0,
258
+ "logps/rejected": -520.0,
259
+ "loss": 0.4309,
260
+ "rewards/accuracies": 0.824999988079071,
261
+ "rewards/chosen": 0.6796875,
262
+ "rewards/margins": 1.2578125,
263
+ "rewards/rejected": -0.58203125,
264
+ "step": 170
265
+ },
266
+ {
267
+ "epoch": 0.423841059602649,
268
+ "grad_norm": 2.827556848526001,
269
+ "learning_rate": 4.15527950310559e-06,
270
+ "logits/chosen": -4.5,
271
+ "logits/rejected": -4.53125,
272
+ "logps/chosen": -616.0,
273
+ "logps/rejected": -506.0,
274
+ "loss": 0.4508,
275
+ "rewards/accuracies": 0.78125,
276
+ "rewards/chosen": 0.6875,
277
+ "rewards/margins": 1.140625,
278
+ "rewards/rejected": -0.453125,
279
+ "step": 180
280
+ },
281
+ {
282
+ "epoch": 0.44738778513612953,
283
+ "grad_norm": 2.7354469299316406,
284
+ "learning_rate": 4.093167701863354e-06,
285
+ "logits/chosen": -4.4375,
286
+ "logits/rejected": -4.5,
287
+ "logps/chosen": -616.0,
288
+ "logps/rejected": -510.0,
289
+ "loss": 0.4341,
290
+ "rewards/accuracies": 0.78125,
291
+ "rewards/chosen": 0.78515625,
292
+ "rewards/margins": 1.3046875,
293
+ "rewards/rejected": -0.51953125,
294
+ "step": 190
295
+ },
296
+ {
297
+ "epoch": 0.47093451066961,
298
+ "grad_norm": 2.674173593521118,
299
+ "learning_rate": 4.031055900621118e-06,
300
+ "logits/chosen": -4.4375,
301
+ "logits/rejected": -4.4375,
302
+ "logps/chosen": -600.0,
303
+ "logps/rejected": -474.0,
304
+ "loss": 0.4224,
305
+ "rewards/accuracies": 0.8125,
306
+ "rewards/chosen": 0.72265625,
307
+ "rewards/margins": 1.140625,
308
+ "rewards/rejected": -0.4140625,
309
+ "step": 200
310
+ },
311
+ {
312
+ "epoch": 0.49448123620309054,
313
+ "grad_norm": 3.0752134323120117,
314
+ "learning_rate": 3.968944099378882e-06,
315
+ "logits/chosen": -4.4375,
316
+ "logits/rejected": -4.46875,
317
+ "logps/chosen": -628.0,
318
+ "logps/rejected": -584.0,
319
+ "loss": 0.4401,
320
+ "rewards/accuracies": 0.75,
321
+ "rewards/chosen": 0.71875,
322
+ "rewards/margins": 1.2109375,
323
+ "rewards/rejected": -0.494140625,
324
+ "step": 210
325
+ },
326
+ {
327
+ "epoch": 0.5180279617365711,
328
+ "grad_norm": 3.02925181388855,
329
+ "learning_rate": 3.906832298136646e-06,
330
+ "logits/chosen": -4.34375,
331
+ "logits/rejected": -4.40625,
332
+ "logps/chosen": -644.0,
333
+ "logps/rejected": -532.0,
334
+ "loss": 0.4338,
335
+ "rewards/accuracies": 0.7437499761581421,
336
+ "rewards/chosen": 0.71875,
337
+ "rewards/margins": 1.0078125,
338
+ "rewards/rejected": -0.29296875,
339
+ "step": 220
340
+ },
341
+ {
342
+ "epoch": 0.5415746872700515,
343
+ "grad_norm": 2.621617078781128,
344
+ "learning_rate": 3.8447204968944105e-06,
345
+ "logits/chosen": -4.40625,
346
+ "logits/rejected": -4.4375,
347
+ "logps/chosen": -616.0,
348
+ "logps/rejected": -552.0,
349
+ "loss": 0.4285,
350
+ "rewards/accuracies": 0.7875000238418579,
351
+ "rewards/chosen": 0.81640625,
352
+ "rewards/margins": 1.1640625,
353
+ "rewards/rejected": -0.345703125,
354
+ "step": 230
355
+ },
356
+ {
357
+ "epoch": 0.565121412803532,
358
+ "grad_norm": 2.96905779838562,
359
+ "learning_rate": 3.782608695652174e-06,
360
+ "logits/chosen": -4.46875,
361
+ "logits/rejected": -4.53125,
362
+ "logps/chosen": -560.0,
363
+ "logps/rejected": -486.0,
364
+ "loss": 0.4185,
365
+ "rewards/accuracies": 0.8062499761581421,
366
+ "rewards/chosen": 0.84375,
367
+ "rewards/margins": 1.40625,
368
+ "rewards/rejected": -0.55859375,
369
+ "step": 240
370
+ },
371
+ {
372
+ "epoch": 0.5886681383370125,
373
+ "grad_norm": 3.1430623531341553,
374
+ "learning_rate": 3.7204968944099383e-06,
375
+ "logits/chosen": -4.46875,
376
+ "logits/rejected": -4.53125,
377
+ "logps/chosen": -624.0,
378
+ "logps/rejected": -556.0,
379
+ "loss": 0.4288,
380
+ "rewards/accuracies": 0.768750011920929,
381
+ "rewards/chosen": 0.85546875,
382
+ "rewards/margins": 1.2890625,
383
+ "rewards/rejected": -0.435546875,
384
+ "step": 250
385
+ },
386
+ {
387
+ "epoch": 0.6122148638704931,
388
+ "grad_norm": 2.8485913276672363,
389
+ "learning_rate": 3.6583850931677024e-06,
390
+ "logits/chosen": -4.40625,
391
+ "logits/rejected": -4.4375,
392
+ "logps/chosen": -584.0,
393
+ "logps/rejected": -552.0,
394
+ "loss": 0.4362,
395
+ "rewards/accuracies": 0.800000011920929,
396
+ "rewards/chosen": 0.90625,
397
+ "rewards/margins": 1.3125,
398
+ "rewards/rejected": -0.41015625,
399
+ "step": 260
400
+ },
401
+ {
402
+ "epoch": 0.6357615894039735,
403
+ "grad_norm": 2.8013317584991455,
404
+ "learning_rate": 3.596273291925466e-06,
405
+ "logits/chosen": -4.4375,
406
+ "logits/rejected": -4.4375,
407
+ "logps/chosen": -584.0,
408
+ "logps/rejected": -492.0,
409
+ "loss": 0.4193,
410
+ "rewards/accuracies": 0.7562500238418579,
411
+ "rewards/chosen": 0.75,
412
+ "rewards/margins": 1.25,
413
+ "rewards/rejected": -0.498046875,
414
+ "step": 270
415
+ },
416
+ {
417
+ "epoch": 0.659308314937454,
418
+ "grad_norm": 2.7474491596221924,
419
+ "learning_rate": 3.53416149068323e-06,
420
+ "logits/chosen": -4.4375,
421
+ "logits/rejected": -4.5,
422
+ "logps/chosen": -604.0,
423
+ "logps/rejected": -472.0,
424
+ "loss": 0.4029,
425
+ "rewards/accuracies": 0.793749988079071,
426
+ "rewards/chosen": 0.8046875,
427
+ "rewards/margins": 1.3125,
428
+ "rewards/rejected": -0.5078125,
429
+ "step": 280
430
+ },
431
+ {
432
+ "epoch": 0.6828550404709345,
433
+ "grad_norm": 2.721045732498169,
434
+ "learning_rate": 3.472049689440994e-06,
435
+ "logits/chosen": -4.40625,
436
+ "logits/rejected": -4.46875,
437
+ "logps/chosen": -592.0,
438
+ "logps/rejected": -510.0,
439
+ "loss": 0.4111,
440
+ "rewards/accuracies": 0.8062499761581421,
441
+ "rewards/chosen": 1.09375,
442
+ "rewards/margins": 1.6015625,
443
+ "rewards/rejected": -0.51171875,
444
+ "step": 290
445
+ },
446
+ {
447
+ "epoch": 0.7064017660044151,
448
+ "grad_norm": 2.9466845989227295,
449
+ "learning_rate": 3.4099378881987584e-06,
450
+ "logits/chosen": -4.40625,
451
+ "logits/rejected": -4.40625,
452
+ "logps/chosen": -568.0,
453
+ "logps/rejected": -468.0,
454
+ "loss": 0.4108,
455
+ "rewards/accuracies": 0.7749999761581421,
456
+ "rewards/chosen": 1.1015625,
457
+ "rewards/margins": 1.578125,
458
+ "rewards/rejected": -0.474609375,
459
+ "step": 300
460
+ },
461
+ {
462
+ "epoch": 0.7299484915378955,
463
+ "grad_norm": 2.3606228828430176,
464
+ "learning_rate": 3.347826086956522e-06,
465
+ "logits/chosen": -4.40625,
466
+ "logits/rejected": -4.40625,
467
+ "logps/chosen": -596.0,
468
+ "logps/rejected": -488.0,
469
+ "loss": 0.4041,
470
+ "rewards/accuracies": 0.78125,
471
+ "rewards/chosen": 1.078125,
472
+ "rewards/margins": 1.4453125,
473
+ "rewards/rejected": -0.36328125,
474
+ "step": 310
475
+ },
476
+ {
477
+ "epoch": 0.753495217071376,
478
+ "grad_norm": 3.1279854774475098,
479
+ "learning_rate": 3.285714285714286e-06,
480
+ "logits/chosen": -4.5,
481
+ "logits/rejected": -4.46875,
482
+ "logps/chosen": -564.0,
483
+ "logps/rejected": -510.0,
484
+ "loss": 0.4247,
485
+ "rewards/accuracies": 0.762499988079071,
486
+ "rewards/chosen": 0.9609375,
487
+ "rewards/margins": 1.2265625,
488
+ "rewards/rejected": -0.26171875,
489
+ "step": 320
490
+ },
491
+ {
492
+ "epoch": 0.7770419426048565,
493
+ "grad_norm": 2.6707963943481445,
494
+ "learning_rate": 3.22360248447205e-06,
495
+ "logits/chosen": -4.375,
496
+ "logits/rejected": -4.4375,
497
+ "logps/chosen": -580.0,
498
+ "logps/rejected": -492.0,
499
+ "loss": 0.3787,
500
+ "rewards/accuracies": 0.800000011920929,
501
+ "rewards/chosen": 1.0390625,
502
+ "rewards/margins": 1.4453125,
503
+ "rewards/rejected": -0.400390625,
504
+ "step": 330
505
+ },
506
+ {
507
+ "epoch": 0.8005886681383371,
508
+ "grad_norm": 2.8069331645965576,
509
+ "learning_rate": 3.1614906832298135e-06,
510
+ "logits/chosen": -4.46875,
511
+ "logits/rejected": -4.46875,
512
+ "logps/chosen": -632.0,
513
+ "logps/rejected": -600.0,
514
+ "loss": 0.4048,
515
+ "rewards/accuracies": 0.78125,
516
+ "rewards/chosen": 1.078125,
517
+ "rewards/margins": 1.3125,
518
+ "rewards/rejected": -0.228515625,
519
+ "step": 340
520
+ },
521
+ {
522
+ "epoch": 0.8241353936718175,
523
+ "grad_norm": 2.5299603939056396,
524
+ "learning_rate": 3.099378881987578e-06,
525
+ "logits/chosen": -4.53125,
526
+ "logits/rejected": -4.5,
527
+ "logps/chosen": -600.0,
528
+ "logps/rejected": -532.0,
529
+ "loss": 0.3885,
530
+ "rewards/accuracies": 0.78125,
531
+ "rewards/chosen": 1.1875,
532
+ "rewards/margins": 1.4375,
533
+ "rewards/rejected": -0.251953125,
534
+ "step": 350
535
+ },
536
+ {
537
+ "epoch": 0.847682119205298,
538
+ "grad_norm": 2.7665863037109375,
539
+ "learning_rate": 3.037267080745342e-06,
540
+ "logits/chosen": -4.46875,
541
+ "logits/rejected": -4.4375,
542
+ "logps/chosen": -592.0,
543
+ "logps/rejected": -472.0,
544
+ "loss": 0.3892,
545
+ "rewards/accuracies": 0.862500011920929,
546
+ "rewards/chosen": 1.3515625,
547
+ "rewards/margins": 1.7578125,
548
+ "rewards/rejected": -0.404296875,
549
+ "step": 360
550
+ },
551
+ {
552
+ "epoch": 0.8712288447387785,
553
+ "grad_norm": 2.8109049797058105,
554
+ "learning_rate": 2.975155279503106e-06,
555
+ "logits/chosen": -4.46875,
556
+ "logits/rejected": -4.4375,
557
+ "logps/chosen": -580.0,
558
+ "logps/rejected": -524.0,
559
+ "loss": 0.4115,
560
+ "rewards/accuracies": 0.8062499761581421,
561
+ "rewards/chosen": 1.171875,
562
+ "rewards/margins": 1.5234375,
563
+ "rewards/rejected": -0.349609375,
564
+ "step": 370
565
+ },
566
+ {
567
+ "epoch": 0.8947755702722591,
568
+ "grad_norm": 2.657787799835205,
569
+ "learning_rate": 2.9130434782608695e-06,
570
+ "logits/chosen": -4.40625,
571
+ "logits/rejected": -4.46875,
572
+ "logps/chosen": -568.0,
573
+ "logps/rejected": -532.0,
574
+ "loss": 0.3831,
575
+ "rewards/accuracies": 0.8374999761581421,
576
+ "rewards/chosen": 1.1953125,
577
+ "rewards/margins": 1.578125,
578
+ "rewards/rejected": -0.38671875,
579
+ "step": 380
580
+ },
581
+ {
582
+ "epoch": 0.9183222958057395,
583
+ "grad_norm": 2.7588131427764893,
584
+ "learning_rate": 2.850931677018634e-06,
585
+ "logits/chosen": -4.46875,
586
+ "logits/rejected": -4.46875,
587
+ "logps/chosen": -588.0,
588
+ "logps/rejected": -544.0,
589
+ "loss": 0.3865,
590
+ "rewards/accuracies": 0.8125,
591
+ "rewards/chosen": 1.234375,
592
+ "rewards/margins": 1.4453125,
593
+ "rewards/rejected": -0.216796875,
594
+ "step": 390
595
+ },
596
+ {
597
+ "epoch": 0.94186902133922,
598
+ "grad_norm": 3.0615074634552,
599
+ "learning_rate": 2.788819875776398e-06,
600
+ "logits/chosen": -4.4375,
601
+ "logits/rejected": -4.5,
602
+ "logps/chosen": -576.0,
603
+ "logps/rejected": -502.0,
604
+ "loss": 0.3967,
605
+ "rewards/accuracies": 0.800000011920929,
606
+ "rewards/chosen": 1.1875,
607
+ "rewards/margins": 1.546875,
608
+ "rewards/rejected": -0.359375,
609
+ "step": 400
610
+ },
611
+ {
612
+ "epoch": 0.9654157468727005,
613
+ "grad_norm": 3.119372844696045,
614
+ "learning_rate": 2.7267080745341618e-06,
615
+ "logits/chosen": -4.53125,
616
+ "logits/rejected": -4.5,
617
+ "logps/chosen": -604.0,
618
+ "logps/rejected": -576.0,
619
+ "loss": 0.4029,
620
+ "rewards/accuracies": 0.8125,
621
+ "rewards/chosen": 1.28125,
622
+ "rewards/margins": 1.7265625,
623
+ "rewards/rejected": -0.451171875,
624
+ "step": 410
625
+ },
626
+ {
627
+ "epoch": 0.9889624724061811,
628
+ "grad_norm": 3.135815382003784,
629
+ "learning_rate": 2.6645962732919255e-06,
630
+ "logits/chosen": -4.375,
631
+ "logits/rejected": -4.4375,
632
+ "logps/chosen": -628.0,
633
+ "logps/rejected": -544.0,
634
+ "loss": 0.3687,
635
+ "rewards/accuracies": 0.8500000238418579,
636
+ "rewards/chosen": 1.46875,
637
+ "rewards/margins": 1.671875,
638
+ "rewards/rejected": -0.1982421875,
639
+ "step": 420
640
+ },
641
+ {
642
+ "epoch": 1.0141280353200883,
643
+ "grad_norm": 1.4070848226547241,
644
+ "learning_rate": 2.6024844720496896e-06,
645
+ "logits/chosen": -4.40625,
646
+ "logits/rejected": -4.4375,
647
+ "logps/chosen": -596.0,
648
+ "logps/rejected": -544.0,
649
+ "loss": 0.3031,
650
+ "rewards/accuracies": 0.8695651888847351,
651
+ "rewards/chosen": 1.4140625,
652
+ "rewards/margins": 1.8203125,
653
+ "rewards/rejected": -0.40234375,
654
+ "step": 430
655
+ },
656
+ {
657
+ "epoch": 1.0376747608535688,
658
+ "grad_norm": 1.7027989625930786,
659
+ "learning_rate": 2.5403726708074537e-06,
660
+ "logits/chosen": -4.375,
661
+ "logits/rejected": -4.40625,
662
+ "logps/chosen": -584.0,
663
+ "logps/rejected": -502.0,
664
+ "loss": 0.2702,
665
+ "rewards/accuracies": 0.9312499761581421,
666
+ "rewards/chosen": 1.578125,
667
+ "rewards/margins": 2.21875,
668
+ "rewards/rejected": -0.64453125,
669
+ "step": 440
670
+ },
671
+ {
672
+ "epoch": 1.0612214863870493,
673
+ "grad_norm": 1.9917528629302979,
674
+ "learning_rate": 2.4782608695652178e-06,
675
+ "logits/chosen": -4.4375,
676
+ "logits/rejected": -4.40625,
677
+ "logps/chosen": -584.0,
678
+ "logps/rejected": -508.0,
679
+ "loss": 0.2617,
680
+ "rewards/accuracies": 0.8812500238418579,
681
+ "rewards/chosen": 1.53125,
682
+ "rewards/margins": 2.046875,
683
+ "rewards/rejected": -0.51953125,
684
+ "step": 450
685
+ },
686
+ {
687
+ "epoch": 1.0847682119205297,
688
+ "grad_norm": 1.8857314586639404,
689
+ "learning_rate": 2.4161490683229814e-06,
690
+ "logits/chosen": -4.40625,
691
+ "logits/rejected": -4.46875,
692
+ "logps/chosen": -580.0,
693
+ "logps/rejected": -524.0,
694
+ "loss": 0.2539,
695
+ "rewards/accuracies": 0.925000011920929,
696
+ "rewards/chosen": 1.703125,
697
+ "rewards/margins": 2.171875,
698
+ "rewards/rejected": -0.462890625,
699
+ "step": 460
700
+ },
701
+ {
702
+ "epoch": 1.1083149374540102,
703
+ "grad_norm": 1.778446912765503,
704
+ "learning_rate": 2.3540372670807455e-06,
705
+ "logits/chosen": -4.4375,
706
+ "logits/rejected": -4.53125,
707
+ "logps/chosen": -588.0,
708
+ "logps/rejected": -472.0,
709
+ "loss": 0.2617,
710
+ "rewards/accuracies": 0.8999999761581421,
711
+ "rewards/chosen": 1.5703125,
712
+ "rewards/margins": 2.03125,
713
+ "rewards/rejected": -0.46484375,
714
+ "step": 470
715
+ },
716
+ {
717
+ "epoch": 1.131861662987491,
718
+ "grad_norm": 1.5340267419815063,
719
+ "learning_rate": 2.2919254658385096e-06,
720
+ "logits/chosen": -4.34375,
721
+ "logits/rejected": -4.40625,
722
+ "logps/chosen": -644.0,
723
+ "logps/rejected": -608.0,
724
+ "loss": 0.2414,
725
+ "rewards/accuracies": 0.9375,
726
+ "rewards/chosen": 1.71875,
727
+ "rewards/margins": 2.40625,
728
+ "rewards/rejected": -0.6875,
729
+ "step": 480
730
+ },
731
+ {
732
+ "epoch": 1.1554083885209714,
733
+ "grad_norm": 1.7361700534820557,
734
+ "learning_rate": 2.2298136645962733e-06,
735
+ "logits/chosen": -4.375,
736
+ "logits/rejected": -4.34375,
737
+ "logps/chosen": -612.0,
738
+ "logps/rejected": -564.0,
739
+ "loss": 0.2463,
740
+ "rewards/accuracies": 0.918749988079071,
741
+ "rewards/chosen": 1.5859375,
742
+ "rewards/margins": 2.234375,
743
+ "rewards/rejected": -0.65625,
744
+ "step": 490
745
+ },
746
+ {
747
+ "epoch": 1.1789551140544519,
748
+ "grad_norm": 1.7070685625076294,
749
+ "learning_rate": 2.1677018633540374e-06,
750
+ "logits/chosen": -4.34375,
751
+ "logits/rejected": -4.3125,
752
+ "logps/chosen": -568.0,
753
+ "logps/rejected": -528.0,
754
+ "loss": 0.2385,
755
+ "rewards/accuracies": 0.918749988079071,
756
+ "rewards/chosen": 1.6015625,
757
+ "rewards/margins": 2.09375,
758
+ "rewards/rejected": -0.5,
759
+ "step": 500
760
+ },
761
+ {
762
+ "epoch": 1.2025018395879323,
763
+ "grad_norm": 2.132335901260376,
764
+ "learning_rate": 2.1055900621118015e-06,
765
+ "logits/chosen": -4.4375,
766
+ "logits/rejected": -4.40625,
767
+ "logps/chosen": -636.0,
768
+ "logps/rejected": -548.0,
769
+ "loss": 0.2412,
770
+ "rewards/accuracies": 0.90625,
771
+ "rewards/chosen": 1.734375,
772
+ "rewards/margins": 2.203125,
773
+ "rewards/rejected": -0.46484375,
774
+ "step": 510
775
+ },
776
+ {
777
+ "epoch": 1.2260485651214128,
778
+ "grad_norm": 2.0063838958740234,
779
+ "learning_rate": 2.0434782608695656e-06,
780
+ "logits/chosen": -4.46875,
781
+ "logits/rejected": -4.40625,
782
+ "logps/chosen": -576.0,
783
+ "logps/rejected": -552.0,
784
+ "loss": 0.2437,
785
+ "rewards/accuracies": 0.918749988079071,
786
+ "rewards/chosen": 1.484375,
787
+ "rewards/margins": 2.171875,
788
+ "rewards/rejected": -0.69140625,
789
+ "step": 520
790
+ },
791
+ {
792
+ "epoch": 1.2495952906548933,
793
+ "grad_norm": 1.715623378753662,
794
+ "learning_rate": 1.9813664596273293e-06,
795
+ "logits/chosen": -4.375,
796
+ "logits/rejected": -4.4375,
797
+ "logps/chosen": -620.0,
798
+ "logps/rejected": -556.0,
799
+ "loss": 0.2371,
800
+ "rewards/accuracies": 0.9125000238418579,
801
+ "rewards/chosen": 1.6328125,
802
+ "rewards/margins": 2.09375,
803
+ "rewards/rejected": -0.466796875,
804
+ "step": 530
805
+ },
806
+ {
807
+ "epoch": 1.2731420161883737,
808
+ "grad_norm": 1.8319090604782104,
809
+ "learning_rate": 1.9192546583850934e-06,
810
+ "logits/chosen": -4.375,
811
+ "logits/rejected": -4.34375,
812
+ "logps/chosen": -564.0,
813
+ "logps/rejected": -470.0,
814
+ "loss": 0.2456,
815
+ "rewards/accuracies": 0.949999988079071,
816
+ "rewards/chosen": 1.6015625,
817
+ "rewards/margins": 2.3125,
818
+ "rewards/rejected": -0.70703125,
819
+ "step": 540
820
+ },
821
+ {
822
+ "epoch": 1.2966887417218542,
823
+ "grad_norm": 1.757511854171753,
824
+ "learning_rate": 1.8571428571428573e-06,
825
+ "logits/chosen": -4.4375,
826
+ "logits/rejected": -4.46875,
827
+ "logps/chosen": -604.0,
828
+ "logps/rejected": -528.0,
829
+ "loss": 0.2382,
830
+ "rewards/accuracies": 0.925000011920929,
831
+ "rewards/chosen": 1.7265625,
832
+ "rewards/margins": 2.21875,
833
+ "rewards/rejected": -0.494140625,
834
+ "step": 550
835
+ },
836
+ {
837
+ "epoch": 1.3202354672553347,
838
+ "grad_norm": 1.6837247610092163,
839
+ "learning_rate": 1.7950310559006212e-06,
840
+ "logits/chosen": -4.40625,
841
+ "logits/rejected": -4.40625,
842
+ "logps/chosen": -616.0,
843
+ "logps/rejected": -528.0,
844
+ "loss": 0.2339,
845
+ "rewards/accuracies": 0.887499988079071,
846
+ "rewards/chosen": 1.8125,
847
+ "rewards/margins": 2.203125,
848
+ "rewards/rejected": -0.390625,
849
+ "step": 560
850
+ },
851
+ {
852
+ "epoch": 1.3437821927888154,
853
+ "grad_norm": 1.5681090354919434,
854
+ "learning_rate": 1.7329192546583853e-06,
855
+ "logits/chosen": -4.4375,
856
+ "logits/rejected": -4.4375,
857
+ "logps/chosen": -572.0,
858
+ "logps/rejected": -496.0,
859
+ "loss": 0.2341,
860
+ "rewards/accuracies": 0.9375,
861
+ "rewards/chosen": 1.7890625,
862
+ "rewards/margins": 2.1875,
863
+ "rewards/rejected": -0.40625,
864
+ "step": 570
865
+ },
866
+ {
867
+ "epoch": 1.3673289183222959,
868
+ "grad_norm": 1.8752034902572632,
869
+ "learning_rate": 1.6708074534161492e-06,
870
+ "logits/chosen": -4.5,
871
+ "logits/rejected": -4.5,
872
+ "logps/chosen": -532.0,
873
+ "logps/rejected": -512.0,
874
+ "loss": 0.2257,
875
+ "rewards/accuracies": 0.8999999761581421,
876
+ "rewards/chosen": 1.671875,
877
+ "rewards/margins": 2.390625,
878
+ "rewards/rejected": -0.71875,
879
+ "step": 580
880
+ },
881
+ {
882
+ "epoch": 1.3908756438557763,
883
+ "grad_norm": 1.8631969690322876,
884
+ "learning_rate": 1.608695652173913e-06,
885
+ "logits/chosen": -4.40625,
886
+ "logits/rejected": -4.40625,
887
+ "logps/chosen": -596.0,
888
+ "logps/rejected": -488.0,
889
+ "loss": 0.2405,
890
+ "rewards/accuracies": 0.9125000238418579,
891
+ "rewards/chosen": 1.671875,
892
+ "rewards/margins": 2.28125,
893
+ "rewards/rejected": -0.61328125,
894
+ "step": 590
895
+ },
896
+ {
897
+ "epoch": 1.4144223693892568,
898
+ "grad_norm": 1.8408665657043457,
899
+ "learning_rate": 1.5465838509316772e-06,
900
+ "logits/chosen": -4.4375,
901
+ "logits/rejected": -4.40625,
902
+ "logps/chosen": -568.0,
903
+ "logps/rejected": -508.0,
904
+ "loss": 0.2346,
905
+ "rewards/accuracies": 0.8999999761581421,
906
+ "rewards/chosen": 1.6796875,
907
+ "rewards/margins": 2.34375,
908
+ "rewards/rejected": -0.6640625,
909
+ "step": 600
910
+ },
911
+ {
912
+ "epoch": 1.4379690949227373,
913
+ "grad_norm": 1.8455615043640137,
914
+ "learning_rate": 1.484472049689441e-06,
915
+ "logits/chosen": -4.5,
916
+ "logits/rejected": -4.5,
917
+ "logps/chosen": -632.0,
918
+ "logps/rejected": -548.0,
919
+ "loss": 0.2415,
920
+ "rewards/accuracies": 0.925000011920929,
921
+ "rewards/chosen": 1.8359375,
922
+ "rewards/margins": 2.25,
923
+ "rewards/rejected": -0.41015625,
924
+ "step": 610
925
+ },
926
+ {
927
+ "epoch": 1.4615158204562178,
928
+ "grad_norm": 1.7498400211334229,
929
+ "learning_rate": 1.4223602484472052e-06,
930
+ "logits/chosen": -4.40625,
931
+ "logits/rejected": -4.4375,
932
+ "logps/chosen": -588.0,
933
+ "logps/rejected": -496.0,
934
+ "loss": 0.2428,
935
+ "rewards/accuracies": 0.887499988079071,
936
+ "rewards/chosen": 1.8125,
937
+ "rewards/margins": 2.328125,
938
+ "rewards/rejected": -0.515625,
939
+ "step": 620
940
+ },
941
+ {
942
+ "epoch": 1.4850625459896984,
943
+ "grad_norm": 1.8195767402648926,
944
+ "learning_rate": 1.360248447204969e-06,
945
+ "logits/chosen": -4.40625,
946
+ "logits/rejected": -4.4375,
947
+ "logps/chosen": -616.0,
948
+ "logps/rejected": -520.0,
949
+ "loss": 0.2408,
950
+ "rewards/accuracies": 0.9125000238418579,
951
+ "rewards/chosen": 1.8125,
952
+ "rewards/margins": 2.40625,
953
+ "rewards/rejected": -0.59375,
954
+ "step": 630
955
+ },
956
+ {
957
+ "epoch": 1.508609271523179,
958
+ "grad_norm": 1.84929621219635,
959
+ "learning_rate": 1.2981366459627332e-06,
960
+ "logits/chosen": -4.375,
961
+ "logits/rejected": -4.40625,
962
+ "logps/chosen": -596.0,
963
+ "logps/rejected": -492.0,
964
+ "loss": 0.2462,
965
+ "rewards/accuracies": 0.893750011920929,
966
+ "rewards/chosen": 1.5859375,
967
+ "rewards/margins": 2.328125,
968
+ "rewards/rejected": -0.73046875,
969
+ "step": 640
970
+ },
971
+ {
972
+ "epoch": 1.5321559970566594,
973
+ "grad_norm": 1.943724513053894,
974
+ "learning_rate": 1.236024844720497e-06,
975
+ "logits/chosen": -4.5,
976
+ "logits/rejected": -4.5,
977
+ "logps/chosen": -624.0,
978
+ "logps/rejected": -556.0,
979
+ "loss": 0.2502,
980
+ "rewards/accuracies": 0.918749988079071,
981
+ "rewards/chosen": 1.734375,
982
+ "rewards/margins": 2.296875,
983
+ "rewards/rejected": -0.56640625,
984
+ "step": 650
985
+ },
986
+ {
987
+ "epoch": 1.5557027225901399,
988
+ "grad_norm": 1.7159101963043213,
989
+ "learning_rate": 1.173913043478261e-06,
990
+ "logits/chosen": -4.40625,
991
+ "logits/rejected": -4.4375,
992
+ "logps/chosen": -536.0,
993
+ "logps/rejected": -494.0,
994
+ "loss": 0.2423,
995
+ "rewards/accuracies": 0.90625,
996
+ "rewards/chosen": 1.71875,
997
+ "rewards/margins": 2.34375,
998
+ "rewards/rejected": -0.62109375,
999
+ "step": 660
1000
+ },
1001
+ {
1002
+ "epoch": 1.5792494481236203,
1003
+ "grad_norm": 1.5926669836044312,
1004
+ "learning_rate": 1.111801242236025e-06,
1005
+ "logits/chosen": -4.4375,
1006
+ "logits/rejected": -4.46875,
1007
+ "logps/chosen": -568.0,
1008
+ "logps/rejected": -502.0,
1009
+ "loss": 0.234,
1010
+ "rewards/accuracies": 0.887499988079071,
1011
+ "rewards/chosen": 1.515625,
1012
+ "rewards/margins": 2.03125,
1013
+ "rewards/rejected": -0.515625,
1014
+ "step": 670
1015
+ },
1016
+ {
1017
+ "epoch": 1.6027961736571008,
1018
+ "grad_norm": 1.9127466678619385,
1019
+ "learning_rate": 1.049689440993789e-06,
1020
+ "logits/chosen": -4.46875,
1021
+ "logits/rejected": -4.5,
1022
+ "logps/chosen": -612.0,
1023
+ "logps/rejected": -508.0,
1024
+ "loss": 0.2337,
1025
+ "rewards/accuracies": 0.918749988079071,
1026
+ "rewards/chosen": 1.9375,
1027
+ "rewards/margins": 2.375,
1028
+ "rewards/rejected": -0.4296875,
1029
+ "step": 680
1030
+ },
1031
+ {
1032
+ "epoch": 1.6263428991905813,
1033
+ "grad_norm": 1.569267749786377,
1034
+ "learning_rate": 9.875776397515528e-07,
1035
+ "logits/chosen": -4.4375,
1036
+ "logits/rejected": -4.4375,
1037
+ "logps/chosen": -596.0,
1038
+ "logps/rejected": -548.0,
1039
+ "loss": 0.2294,
1040
+ "rewards/accuracies": 0.90625,
1041
+ "rewards/chosen": 1.8984375,
1042
+ "rewards/margins": 2.28125,
1043
+ "rewards/rejected": -0.375,
1044
+ "step": 690
1045
+ },
1046
+ {
1047
+ "epoch": 1.6498896247240618,
1048
+ "grad_norm": 1.7182821035385132,
1049
+ "learning_rate": 9.254658385093168e-07,
1050
+ "logits/chosen": -4.4375,
1051
+ "logits/rejected": -4.4375,
1052
+ "logps/chosen": -632.0,
1053
+ "logps/rejected": -520.0,
1054
+ "loss": 0.2386,
1055
+ "rewards/accuracies": 0.9437500238418579,
1056
+ "rewards/chosen": 1.8046875,
1057
+ "rewards/margins": 2.546875,
1058
+ "rewards/rejected": -0.734375,
1059
+ "step": 700
1060
+ },
1061
+ {
1062
+ "epoch": 1.6734363502575422,
1063
+ "grad_norm": 1.6479544639587402,
1064
+ "learning_rate": 8.633540372670808e-07,
1065
+ "logits/chosen": -4.40625,
1066
+ "logits/rejected": -4.5,
1067
+ "logps/chosen": -576.0,
1068
+ "logps/rejected": -510.0,
1069
+ "loss": 0.2277,
1070
+ "rewards/accuracies": 0.9375,
1071
+ "rewards/chosen": 1.703125,
1072
+ "rewards/margins": 2.421875,
1073
+ "rewards/rejected": -0.71484375,
1074
+ "step": 710
1075
+ },
1076
+ {
1077
+ "epoch": 1.6969830757910227,
1078
+ "grad_norm": 1.6700725555419922,
1079
+ "learning_rate": 8.012422360248448e-07,
1080
+ "logits/chosen": -4.46875,
1081
+ "logits/rejected": -4.46875,
1082
+ "logps/chosen": -572.0,
1083
+ "logps/rejected": -464.0,
1084
+ "loss": 0.2362,
1085
+ "rewards/accuracies": 0.918749988079071,
1086
+ "rewards/chosen": 1.765625,
1087
+ "rewards/margins": 2.375,
1088
+ "rewards/rejected": -0.609375,
1089
+ "step": 720
1090
+ },
1091
+ {
1092
+ "epoch": 1.7205298013245032,
1093
+ "grad_norm": 1.8397201299667358,
1094
+ "learning_rate": 7.391304347826088e-07,
1095
+ "logits/chosen": -4.4375,
1096
+ "logits/rejected": -4.5,
1097
+ "logps/chosen": -656.0,
1098
+ "logps/rejected": -556.0,
1099
+ "loss": 0.2306,
1100
+ "rewards/accuracies": 0.956250011920929,
1101
+ "rewards/chosen": 1.875,
1102
+ "rewards/margins": 2.65625,
1103
+ "rewards/rejected": -0.77734375,
1104
+ "step": 730
1105
+ },
1106
+ {
1107
+ "epoch": 1.7440765268579839,
1108
+ "grad_norm": 1.8021209239959717,
1109
+ "learning_rate": 6.770186335403728e-07,
1110
+ "logits/chosen": -4.46875,
1111
+ "logits/rejected": -4.46875,
1112
+ "logps/chosen": -600.0,
1113
+ "logps/rejected": -556.0,
1114
+ "loss": 0.2239,
1115
+ "rewards/accuracies": 0.9125000238418579,
1116
+ "rewards/chosen": 1.9296875,
1117
+ "rewards/margins": 2.234375,
1118
+ "rewards/rejected": -0.306640625,
1119
+ "step": 740
1120
+ },
1121
+ {
1122
+ "epoch": 1.7676232523914643,
1123
+ "grad_norm": 1.8263683319091797,
1124
+ "learning_rate": 6.149068322981367e-07,
1125
+ "logits/chosen": -4.4375,
1126
+ "logits/rejected": -4.34375,
1127
+ "logps/chosen": -548.0,
1128
+ "logps/rejected": -516.0,
1129
+ "loss": 0.2322,
1130
+ "rewards/accuracies": 0.8500000238418579,
1131
+ "rewards/chosen": 1.6328125,
1132
+ "rewards/margins": 2.171875,
1133
+ "rewards/rejected": -0.54296875,
1134
+ "step": 750
1135
+ },
1136
+ {
1137
+ "epoch": 1.7911699779249448,
1138
+ "grad_norm": 1.5601938962936401,
1139
+ "learning_rate": 5.527950310559007e-07,
1140
+ "logits/chosen": -4.46875,
1141
+ "logits/rejected": -4.46875,
1142
+ "logps/chosen": -572.0,
1143
+ "logps/rejected": -494.0,
1144
+ "loss": 0.2401,
1145
+ "rewards/accuracies": 0.90625,
1146
+ "rewards/chosen": 1.84375,
1147
+ "rewards/margins": 2.171875,
1148
+ "rewards/rejected": -0.3359375,
1149
+ "step": 760
1150
+ },
1151
+ {
1152
+ "epoch": 1.8147167034584253,
1153
+ "grad_norm": 1.8713370561599731,
1154
+ "learning_rate": 4.906832298136646e-07,
1155
+ "logits/chosen": -4.5,
1156
+ "logits/rejected": -4.4375,
1157
+ "logps/chosen": -544.0,
1158
+ "logps/rejected": -494.0,
1159
+ "loss": 0.2345,
1160
+ "rewards/accuracies": 0.9624999761581421,
1161
+ "rewards/chosen": 1.71875,
1162
+ "rewards/margins": 2.296875,
1163
+ "rewards/rejected": -0.58203125,
1164
+ "step": 770
1165
+ },
1166
+ {
1167
+ "epoch": 1.838263428991906,
1168
+ "grad_norm": 1.3841131925582886,
1169
+ "learning_rate": 4.285714285714286e-07,
1170
+ "logits/chosen": -4.40625,
1171
+ "logits/rejected": -4.375,
1172
+ "logps/chosen": -576.0,
1173
+ "logps/rejected": -500.0,
1174
+ "loss": 0.2201,
1175
+ "rewards/accuracies": 0.887499988079071,
1176
+ "rewards/chosen": 1.6640625,
1177
+ "rewards/margins": 2.3125,
1178
+ "rewards/rejected": -0.640625,
1179
+ "step": 780
1180
+ },
1181
+ {
1182
+ "epoch": 1.8618101545253865,
1183
+ "grad_norm": 1.6164636611938477,
1184
+ "learning_rate": 3.664596273291926e-07,
1185
+ "logits/chosen": -4.46875,
1186
+ "logits/rejected": -4.53125,
1187
+ "logps/chosen": -576.0,
1188
+ "logps/rejected": -520.0,
1189
+ "loss": 0.2429,
1190
+ "rewards/accuracies": 0.90625,
1191
+ "rewards/chosen": 1.6953125,
1192
+ "rewards/margins": 2.1875,
1193
+ "rewards/rejected": -0.4921875,
1194
+ "step": 790
1195
+ },
1196
+ {
1197
+ "epoch": 1.885356880058867,
1198
+ "grad_norm": 1.6874057054519653,
1199
+ "learning_rate": 3.0434782608695656e-07,
1200
+ "logits/chosen": -4.5,
1201
+ "logits/rejected": -4.4375,
1202
+ "logps/chosen": -572.0,
1203
+ "logps/rejected": -524.0,
1204
+ "loss": 0.2396,
1205
+ "rewards/accuracies": 0.9375,
1206
+ "rewards/chosen": 1.7265625,
1207
+ "rewards/margins": 2.234375,
1208
+ "rewards/rejected": -0.51171875,
1209
+ "step": 800
1210
+ },
1211
+ {
1212
+ "epoch": 1.9089036055923474,
1213
+ "grad_norm": 1.7373439073562622,
1214
+ "learning_rate": 2.422360248447205e-07,
1215
+ "logits/chosen": -4.4375,
1216
+ "logits/rejected": -4.46875,
1217
+ "logps/chosen": -588.0,
1218
+ "logps/rejected": -508.0,
1219
+ "loss": 0.2511,
1220
+ "rewards/accuracies": 0.925000011920929,
1221
+ "rewards/chosen": 1.9765625,
1222
+ "rewards/margins": 2.34375,
1223
+ "rewards/rejected": -0.37109375,
1224
+ "step": 810
1225
+ },
1226
+ {
1227
+ "epoch": 1.9324503311258279,
1228
+ "grad_norm": 1.989230990409851,
1229
+ "learning_rate": 1.8012422360248447e-07,
1230
+ "logits/chosen": -4.375,
1231
+ "logits/rejected": -4.34375,
1232
+ "logps/chosen": -592.0,
1233
+ "logps/rejected": -504.0,
1234
+ "loss": 0.2427,
1235
+ "rewards/accuracies": 0.8812500238418579,
1236
+ "rewards/chosen": 1.6640625,
1237
+ "rewards/margins": 2.0625,
1238
+ "rewards/rejected": -0.392578125,
1239
+ "step": 820
1240
+ },
1241
+ {
1242
+ "epoch": 1.9559970566593083,
1243
+ "grad_norm": 1.692975640296936,
1244
+ "learning_rate": 1.1801242236024847e-07,
1245
+ "logits/chosen": -4.40625,
1246
+ "logits/rejected": -4.4375,
1247
+ "logps/chosen": -620.0,
1248
+ "logps/rejected": -512.0,
1249
+ "loss": 0.2267,
1250
+ "rewards/accuracies": 0.9437500238418579,
1251
+ "rewards/chosen": 1.7890625,
1252
+ "rewards/margins": 2.53125,
1253
+ "rewards/rejected": -0.7421875,
1254
+ "step": 830
1255
+ },
1256
+ {
1257
+ "epoch": 1.9795437821927888,
1258
+ "grad_norm": 1.7542818784713745,
1259
+ "learning_rate": 5.590062111801243e-08,
1260
+ "logits/chosen": -4.375,
1261
+ "logits/rejected": -4.46875,
1262
+ "logps/chosen": -604.0,
1263
+ "logps/rejected": -536.0,
1264
+ "loss": 0.2425,
1265
+ "rewards/accuracies": 0.893750011920929,
1266
+ "rewards/chosen": 1.8359375,
1267
+ "rewards/margins": 2.078125,
1268
+ "rewards/rejected": -0.244140625,
1269
+ "step": 840
1270
+ }
1271
+ ],
1272
+ "logging_steps": 10,
1273
+ "max_steps": 848,
1274
+ "num_input_tokens_seen": 0,
1275
+ "num_train_epochs": 2,
1276
+ "save_steps": 250,
1277
+ "stateful_callbacks": {
1278
+ "TrainerControl": {
1279
+ "args": {
1280
+ "should_epoch_stop": false,
1281
+ "should_evaluate": false,
1282
+ "should_log": false,
1283
+ "should_save": true,
1284
+ "should_training_stop": true
1285
+ },
1286
+ "attributes": {}
1287
+ }
1288
+ },
1289
+ "total_flos": 0.0,
1290
+ "train_batch_size": 1,
1291
+ "trial_name": null,
1292
+ "trial_params": null
1293
+ }