aaabiao commited on
Commit
9a5ce37
1 Parent(s): 9985fd5

Add files using large-upload tool

Browse files
Files changed (1) hide show
  1. trainer_state.json +732 -0
trainer_state.json ADDED
@@ -0,0 +1,732 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 1.0,
5
+ "eval_steps": 800000000,
6
+ "global_step": 383,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.0,
13
+ "grad_norm": 1.6875,
14
+ "learning_rate": 1.282051282051282e-07,
15
+ "logits/chosen": -3.3797154426574707,
16
+ "logits/rejected": -3.440782070159912,
17
+ "logps/chosen": -244.57943725585938,
18
+ "logps/rejected": -168.14312744140625,
19
+ "loss": 0.6931,
20
+ "rewards/accuracies": 0.0,
21
+ "rewards/chosen": 0.0,
22
+ "rewards/diff": -0.5416666865348816,
23
+ "rewards/diff_abs": 0.5416666865348816,
24
+ "rewards/rejected": 0.0,
25
+ "rewards/student_margin": 0.0,
26
+ "rewards/teacher_margin": 0.5416666865348816,
27
+ "step": 1
28
+ },
29
+ {
30
+ "epoch": 0.03,
31
+ "grad_norm": 1.7421875,
32
+ "learning_rate": 1.282051282051282e-06,
33
+ "logits/chosen": -3.3581883907318115,
34
+ "logits/rejected": -3.306663990020752,
35
+ "logps/chosen": -323.6011657714844,
36
+ "logps/rejected": -269.5755615234375,
37
+ "loss": 0.6946,
38
+ "rewards/accuracies": 0.48148155212402344,
39
+ "rewards/chosen": -0.002131123561412096,
40
+ "rewards/diff": -2.200160026550293,
41
+ "rewards/diff_abs": 2.200160026550293,
42
+ "rewards/rejected": -0.0022028146777302027,
43
+ "rewards/student_margin": 7.169279706431553e-05,
44
+ "rewards/teacher_margin": 2.2002317905426025,
45
+ "step": 10
46
+ },
47
+ {
48
+ "epoch": 0.05,
49
+ "grad_norm": 1.6875,
50
+ "learning_rate": 2.564102564102564e-06,
51
+ "logits/chosen": -3.5238196849823,
52
+ "logits/rejected": -3.590470552444458,
53
+ "logps/chosen": -277.163818359375,
54
+ "logps/rejected": -192.54022216796875,
55
+ "loss": 0.6932,
56
+ "rewards/accuracies": 0.5333333015441895,
57
+ "rewards/chosen": -0.003379967762157321,
58
+ "rewards/diff": -2.1003687381744385,
59
+ "rewards/diff_abs": 2.1011030673980713,
60
+ "rewards/rejected": -0.007698670029640198,
61
+ "rewards/student_margin": 0.004318701568990946,
62
+ "rewards/teacher_margin": 2.1046876907348633,
63
+ "step": 20
64
+ },
65
+ {
66
+ "epoch": 0.08,
67
+ "grad_norm": 1.6875,
68
+ "learning_rate": 3.846153846153847e-06,
69
+ "logits/chosen": -3.4147961139678955,
70
+ "logits/rejected": -3.542013168334961,
71
+ "logps/chosen": -301.5926208496094,
72
+ "logps/rejected": -231.64608764648438,
73
+ "loss": 0.6909,
74
+ "rewards/accuracies": 0.5666667222976685,
75
+ "rewards/chosen": 0.017984800040721893,
76
+ "rewards/diff": -2.444106101989746,
77
+ "rewards/diff_abs": 2.444106101989746,
78
+ "rewards/rejected": 0.006361725740134716,
79
+ "rewards/student_margin": 0.011623072437942028,
80
+ "rewards/teacher_margin": 2.4557292461395264,
81
+ "step": 30
82
+ },
83
+ {
84
+ "epoch": 0.1,
85
+ "grad_norm": 1.640625,
86
+ "learning_rate": 4.99989574668946e-06,
87
+ "logits/chosen": -3.4892425537109375,
88
+ "logits/rejected": -3.546952486038208,
89
+ "logps/chosen": -249.1132049560547,
90
+ "logps/rejected": -179.81195068359375,
91
+ "loss": 0.6874,
92
+ "rewards/accuracies": 0.36666667461395264,
93
+ "rewards/chosen": 0.002312846016138792,
94
+ "rewards/diff": -2.8649048805236816,
95
+ "rewards/diff_abs": 2.8649048805236816,
96
+ "rewards/rejected": 0.019301289692521095,
97
+ "rewards/student_margin": -0.016988443210721016,
98
+ "rewards/teacher_margin": 2.847916841506958,
99
+ "step": 40
100
+ },
101
+ {
102
+ "epoch": 0.13,
103
+ "grad_norm": 1.609375,
104
+ "learning_rate": 4.987395866955716e-06,
105
+ "logits/chosen": -3.387133836746216,
106
+ "logits/rejected": -3.533109664916992,
107
+ "logps/chosen": -331.1688232421875,
108
+ "logps/rejected": -186.29550170898438,
109
+ "loss": 0.6808,
110
+ "rewards/accuracies": 0.6666666269302368,
111
+ "rewards/chosen": 0.03838258236646652,
112
+ "rewards/diff": -1.9975881576538086,
113
+ "rewards/diff_abs": 2.001307487487793,
114
+ "rewards/rejected": -0.008821181952953339,
115
+ "rewards/student_margin": 0.04720376059412956,
116
+ "rewards/teacher_margin": 2.0447916984558105,
117
+ "step": 50
118
+ },
119
+ {
120
+ "epoch": 0.16,
121
+ "grad_norm": 1.6875,
122
+ "learning_rate": 4.954164717534748e-06,
123
+ "logits/chosen": -3.346745729446411,
124
+ "logits/rejected": -3.3796913623809814,
125
+ "logps/chosen": -327.2835388183594,
126
+ "logps/rejected": -350.4725036621094,
127
+ "loss": 0.679,
128
+ "rewards/accuracies": 0.46666669845581055,
129
+ "rewards/chosen": 0.01055043376982212,
130
+ "rewards/diff": -1.2435307502746582,
131
+ "rewards/diff_abs": 1.2447913885116577,
132
+ "rewards/rejected": 0.049393653869628906,
133
+ "rewards/student_margin": -0.038843221962451935,
134
+ "rewards/teacher_margin": 1.2046875953674316,
135
+ "step": 60
136
+ },
137
+ {
138
+ "epoch": 0.18,
139
+ "grad_norm": 1.6171875,
140
+ "learning_rate": 4.900479264361017e-06,
141
+ "logits/chosen": -3.3996708393096924,
142
+ "logits/rejected": -3.4439964294433594,
143
+ "logps/chosen": -308.00958251953125,
144
+ "logps/rejected": -278.58026123046875,
145
+ "loss": 0.6708,
146
+ "rewards/accuracies": 0.5666666626930237,
147
+ "rewards/chosen": 0.0636972039937973,
148
+ "rewards/diff": -1.3036770820617676,
149
+ "rewards/diff_abs": 1.306718349456787,
150
+ "rewards/rejected": 0.04393672198057175,
151
+ "rewards/student_margin": 0.019760485738515854,
152
+ "rewards/teacher_margin": 1.3234374523162842,
153
+ "step": 70
154
+ },
155
+ {
156
+ "epoch": 0.21,
157
+ "grad_norm": 1.6328125,
158
+ "learning_rate": 4.826786950329646e-06,
159
+ "logits/chosen": -3.520059108734131,
160
+ "logits/rejected": -3.576214551925659,
161
+ "logps/chosen": -283.0509338378906,
162
+ "logps/rejected": -180.75180053710938,
163
+ "loss": 0.6653,
164
+ "rewards/accuracies": 0.6999999284744263,
165
+ "rewards/chosen": 0.07656367868185043,
166
+ "rewards/diff": -1.3558346033096313,
167
+ "rewards/diff_abs": 1.3813632726669312,
168
+ "rewards/rejected": 0.01052325963973999,
169
+ "rewards/student_margin": 0.06604041904211044,
170
+ "rewards/teacher_margin": 1.421875,
171
+ "step": 80
172
+ },
173
+ {
174
+ "epoch": 0.23,
175
+ "grad_norm": 1.6640625,
176
+ "learning_rate": 4.733701966071226e-06,
177
+ "logits/chosen": -3.4589409828186035,
178
+ "logits/rejected": -3.511751890182495,
179
+ "logps/chosen": -335.0478820800781,
180
+ "logps/rejected": -170.95372009277344,
181
+ "loss": 0.665,
182
+ "rewards/accuracies": 0.4000000059604645,
183
+ "rewards/chosen": 0.02129988744854927,
184
+ "rewards/diff": -2.89522647857666,
185
+ "rewards/diff_abs": 2.9056954383850098,
186
+ "rewards/rejected": 0.027463769540190697,
187
+ "rewards/student_margin": -0.006163885351270437,
188
+ "rewards/teacher_margin": 2.8890626430511475,
189
+ "step": 90
190
+ },
191
+ {
192
+ "epoch": 0.26,
193
+ "grad_norm": 1.59375,
194
+ "learning_rate": 4.622000130963015e-06,
195
+ "logits/chosen": -3.4993369579315186,
196
+ "logits/rejected": -3.5635037422180176,
197
+ "logps/chosen": -305.15899658203125,
198
+ "logps/rejected": -202.56192016601562,
199
+ "loss": 0.6583,
200
+ "rewards/accuracies": 0.5,
201
+ "rewards/chosen": 0.018651207908988,
202
+ "rewards/diff": -2.507476329803467,
203
+ "rewards/diff_abs": 2.507476329803467,
204
+ "rewards/rejected": -0.02908078208565712,
205
+ "rewards/student_margin": 0.04773198813199997,
206
+ "rewards/teacher_margin": 2.555208444595337,
207
+ "step": 100
208
+ },
209
+ {
210
+ "epoch": 0.29,
211
+ "grad_norm": 1.5546875,
212
+ "learning_rate": 4.492612427040864e-06,
213
+ "logits/chosen": -3.5523293018341064,
214
+ "logits/rejected": -3.6302642822265625,
215
+ "logps/chosen": -277.225830078125,
216
+ "logps/rejected": -200.26490783691406,
217
+ "loss": 0.6502,
218
+ "rewards/accuracies": 0.5999999642372131,
219
+ "rewards/chosen": 0.024044061079621315,
220
+ "rewards/diff": -1.571934461593628,
221
+ "rewards/diff_abs": 1.571934461593628,
222
+ "rewards/rejected": -0.08318804949522018,
223
+ "rewards/student_margin": 0.10723210871219635,
224
+ "rewards/teacher_margin": 1.6791667938232422,
225
+ "step": 110
226
+ },
227
+ {
228
+ "epoch": 0.31,
229
+ "grad_norm": 1.578125,
230
+ "learning_rate": 4.346617239703676e-06,
231
+ "logits/chosen": -3.480700969696045,
232
+ "logits/rejected": -3.604377031326294,
233
+ "logps/chosen": -304.3082580566406,
234
+ "logps/rejected": -239.17782592773438,
235
+ "loss": 0.6465,
236
+ "rewards/accuracies": 0.7333333492279053,
237
+ "rewards/chosen": 0.10159333795309067,
238
+ "rewards/diff": -1.6648391485214233,
239
+ "rewards/diff_abs": 1.7264082431793213,
240
+ "rewards/rejected": 0.0065366788767278194,
241
+ "rewards/student_margin": 0.09505666792392731,
242
+ "rewards/teacher_margin": 1.7598956823349,
243
+ "step": 120
244
+ },
245
+ {
246
+ "epoch": 0.34,
247
+ "grad_norm": 1.6171875,
248
+ "learning_rate": 4.185231369880461e-06,
249
+ "logits/chosen": -3.216306209564209,
250
+ "logits/rejected": -3.4205565452575684,
251
+ "logps/chosen": -324.16461181640625,
252
+ "logps/rejected": -221.4116668701172,
253
+ "loss": 0.6427,
254
+ "rewards/accuracies": 0.6333333253860474,
255
+ "rewards/chosen": 0.020539376884698868,
256
+ "rewards/diff": -2.5135083198547363,
257
+ "rewards/diff_abs": 2.52120041847229,
258
+ "rewards/rejected": -0.060743771493434906,
259
+ "rewards/student_margin": 0.08128315210342407,
260
+ "rewards/teacher_margin": 2.594791889190674,
261
+ "step": 130
262
+ },
263
+ {
264
+ "epoch": 0.37,
265
+ "grad_norm": 1.640625,
266
+ "learning_rate": 4.009799892569317e-06,
267
+ "logits/chosen": -3.4796624183654785,
268
+ "logits/rejected": -3.4889540672302246,
269
+ "logps/chosen": -294.43646240234375,
270
+ "logps/rejected": -235.06918334960938,
271
+ "loss": 0.637,
272
+ "rewards/accuracies": 0.8333333134651184,
273
+ "rewards/chosen": 0.09381090849637985,
274
+ "rewards/diff": -1.9453909397125244,
275
+ "rewards/diff_abs": 1.9707868099212646,
276
+ "rewards/rejected": -0.13163167238235474,
277
+ "rewards/student_margin": 0.225442573428154,
278
+ "rewards/teacher_margin": 2.1708333492279053,
279
+ "step": 140
280
+ },
281
+ {
282
+ "epoch": 0.39,
283
+ "grad_norm": 1.546875,
284
+ "learning_rate": 3.8217849462726334e-06,
285
+ "logits/chosen": -3.6116485595703125,
286
+ "logits/rejected": -3.564044237136841,
287
+ "logps/chosen": -246.2872314453125,
288
+ "logps/rejected": -221.53182983398438,
289
+ "loss": 0.6364,
290
+ "rewards/accuracies": 0.5333333611488342,
291
+ "rewards/chosen": -0.0010233506327494979,
292
+ "rewards/diff": -1.8935142755508423,
293
+ "rewards/diff_abs": 1.8935142755508423,
294
+ "rewards/rejected": -0.08146756142377853,
295
+ "rewards/student_margin": 0.08044421672821045,
296
+ "rewards/teacher_margin": 1.9739586114883423,
297
+ "step": 150
298
+ },
299
+ {
300
+ "epoch": 0.42,
301
+ "grad_norm": 1.609375,
302
+ "learning_rate": 3.6227535467632873e-06,
303
+ "logits/chosen": -3.4925827980041504,
304
+ "logits/rejected": -3.6650619506835938,
305
+ "logps/chosen": -441.7289123535156,
306
+ "logps/rejected": -258.9549255371094,
307
+ "loss": 0.6285,
308
+ "rewards/accuracies": 0.699999988079071,
309
+ "rewards/chosen": 0.09688085317611694,
310
+ "rewards/diff": -1.887935996055603,
311
+ "rewards/diff_abs": 1.9241327047348022,
312
+ "rewards/rejected": -0.09226653724908829,
313
+ "rewards/student_margin": 0.18914742767810822,
314
+ "rewards/teacher_margin": 2.0770833492279053,
315
+ "step": 160
316
+ },
317
+ {
318
+ "epoch": 0.44,
319
+ "grad_norm": 1.609375,
320
+ "learning_rate": 3.4143645267483144e-06,
321
+ "logits/chosen": -3.485863208770752,
322
+ "logits/rejected": -3.5399489402770996,
323
+ "logps/chosen": -317.8275146484375,
324
+ "logps/rejected": -262.54498291015625,
325
+ "loss": 0.6253,
326
+ "rewards/accuracies": 0.6000000238418579,
327
+ "rewards/chosen": -0.009405359625816345,
328
+ "rewards/diff": -2.4229490756988525,
329
+ "rewards/diff_abs": 2.4852664470672607,
330
+ "rewards/rejected": -0.1302066296339035,
331
+ "rewards/student_margin": 0.12080129235982895,
332
+ "rewards/teacher_margin": 2.543750047683716,
333
+ "step": 170
334
+ },
335
+ {
336
+ "epoch": 0.47,
337
+ "grad_norm": 1.5546875,
338
+ "learning_rate": 3.1983547102818104e-06,
339
+ "logits/chosen": -3.4576945304870605,
340
+ "logits/rejected": -3.5367603302001953,
341
+ "logps/chosen": -356.4649658203125,
342
+ "logps/rejected": -292.81439208984375,
343
+ "loss": 0.6201,
344
+ "rewards/accuracies": 0.6333333253860474,
345
+ "rewards/chosen": -0.02529655024409294,
346
+ "rewards/diff": -1.6506948471069336,
347
+ "rewards/diff_abs": 1.7114416360855103,
348
+ "rewards/rejected": -0.43606019020080566,
349
+ "rewards/student_margin": 0.410763680934906,
350
+ "rewards/teacher_margin": 2.0614585876464844,
351
+ "step": 180
352
+ },
353
+ {
354
+ "epoch": 0.5,
355
+ "grad_norm": 1.5625,
356
+ "learning_rate": 2.9765244371567873e-06,
357
+ "logits/chosen": -3.4763588905334473,
358
+ "logits/rejected": -3.562711715698242,
359
+ "logps/chosen": -280.12835693359375,
360
+ "logps/rejected": -208.5796661376953,
361
+ "loss": 0.6191,
362
+ "rewards/accuracies": 0.5666667222976685,
363
+ "rewards/chosen": 0.045582111924886703,
364
+ "rewards/diff": -2.346590280532837,
365
+ "rewards/diff_abs": 2.3965134620666504,
366
+ "rewards/rejected": -0.23230692744255066,
367
+ "rewards/student_margin": 0.27788907289505005,
368
+ "rewards/teacher_margin": 2.624479055404663,
369
+ "step": 190
370
+ },
371
+ {
372
+ "epoch": 0.52,
373
+ "grad_norm": 1.6484375,
374
+ "learning_rate": 2.7507225579233487e-06,
375
+ "logits/chosen": -3.7000794410705566,
376
+ "logits/rejected": -3.884822368621826,
377
+ "logps/chosen": -268.5820617675781,
378
+ "logps/rejected": -196.71481323242188,
379
+ "loss": 0.6147,
380
+ "rewards/accuracies": 0.6666666269302368,
381
+ "rewards/chosen": 0.0688471719622612,
382
+ "rewards/diff": -2.06192946434021,
383
+ "rewards/diff_abs": 2.06192946434021,
384
+ "rewards/rejected": -0.0957857146859169,
385
+ "rewards/student_margin": 0.1646328866481781,
386
+ "rewards/teacher_margin": 2.2265625,
387
+ "step": 200
388
+ },
389
+ {
390
+ "epoch": 0.55,
391
+ "grad_norm": 1.734375,
392
+ "learning_rate": 2.522831024592615e-06,
393
+ "logits/chosen": -3.5710883140563965,
394
+ "logits/rejected": -3.746605634689331,
395
+ "logps/chosen": -306.7405700683594,
396
+ "logps/rejected": -241.92324829101562,
397
+ "loss": 0.6188,
398
+ "rewards/accuracies": 0.6666666269302368,
399
+ "rewards/chosen": -0.010941224172711372,
400
+ "rewards/diff": -2.189682722091675,
401
+ "rewards/diff_abs": 2.2137255668640137,
402
+ "rewards/rejected": -0.17959186434745789,
403
+ "rewards/student_margin": 0.16865065693855286,
404
+ "rewards/teacher_margin": 2.3583333492279053,
405
+ "step": 210
406
+ },
407
+ {
408
+ "epoch": 0.57,
409
+ "grad_norm": 1.5859375,
410
+ "learning_rate": 2.2947492054556075e-06,
411
+ "logits/chosen": -3.5517051219940186,
412
+ "logits/rejected": -3.7514452934265137,
413
+ "logps/chosen": -323.0804138183594,
414
+ "logps/rejected": -211.4063262939453,
415
+ "loss": 0.6077,
416
+ "rewards/accuracies": 0.6333333849906921,
417
+ "rewards/chosen": -0.08047564327716827,
418
+ "rewards/diff": -1.5008246898651123,
419
+ "rewards/diff_abs": 1.5533117055892944,
420
+ "rewards/rejected": -0.24319279193878174,
421
+ "rewards/student_margin": 0.16271713376045227,
422
+ "rewards/teacher_margin": 1.6635417938232422,
423
+ "step": 220
424
+ },
425
+ {
426
+ "epoch": 0.6,
427
+ "grad_norm": 1.59375,
428
+ "learning_rate": 2.0683780547456666e-06,
429
+ "logits/chosen": -3.480419635772705,
430
+ "logits/rejected": -3.667999744415283,
431
+ "logps/chosen": -314.51739501953125,
432
+ "logps/rejected": -293.828125,
433
+ "loss": 0.6222,
434
+ "rewards/accuracies": 0.7666667103767395,
435
+ "rewards/chosen": 0.07014875113964081,
436
+ "rewards/diff": -1.6531155109405518,
437
+ "rewards/diff_abs": 1.6864144802093506,
438
+ "rewards/rejected": -0.21058997511863708,
439
+ "rewards/student_margin": 0.2807387411594391,
440
+ "rewards/teacher_margin": 1.933854341506958,
441
+ "step": 230
442
+ },
443
+ {
444
+ "epoch": 0.63,
445
+ "grad_norm": 1.546875,
446
+ "learning_rate": 1.845604269082787e-06,
447
+ "logits/chosen": -3.5570831298828125,
448
+ "logits/rejected": -3.799448013305664,
449
+ "logps/chosen": -326.3376159667969,
450
+ "logps/rejected": -229.31527709960938,
451
+ "loss": 0.6077,
452
+ "rewards/accuracies": 0.6000000238418579,
453
+ "rewards/chosen": 0.007120040711015463,
454
+ "rewards/diff": -2.2296009063720703,
455
+ "rewards/diff_abs": 2.2296009063720703,
456
+ "rewards/rejected": -0.10494570434093475,
457
+ "rewards/student_margin": 0.11206575483083725,
458
+ "rewards/teacher_margin": 2.3416669368743896,
459
+ "step": 240
460
+ },
461
+ {
462
+ "epoch": 0.65,
463
+ "grad_norm": 1.578125,
464
+ "learning_rate": 1.628284562748429e-06,
465
+ "logits/chosen": -3.538252592086792,
466
+ "logits/rejected": -3.9445133209228516,
467
+ "logps/chosen": -453.56585693359375,
468
+ "logps/rejected": -191.9573516845703,
469
+ "loss": 0.6036,
470
+ "rewards/accuracies": 0.8333333134651184,
471
+ "rewards/chosen": 0.1685246229171753,
472
+ "rewards/diff": -2.3100411891937256,
473
+ "rewards/diff_abs": 2.372875690460205,
474
+ "rewards/rejected": -0.21674680709838867,
475
+ "rewards/student_margin": 0.38527145981788635,
476
+ "rewards/teacher_margin": 2.695312738418579,
477
+ "step": 250
478
+ },
479
+ {
480
+ "epoch": 0.68,
481
+ "grad_norm": 1.5703125,
482
+ "learning_rate": 1.4182301928489556e-06,
483
+ "logits/chosen": -3.6895079612731934,
484
+ "logits/rejected": -3.920624256134033,
485
+ "logps/chosen": -319.21429443359375,
486
+ "logps/rejected": -181.65463256835938,
487
+ "loss": 0.6049,
488
+ "rewards/accuracies": 0.6666666269302368,
489
+ "rewards/chosen": 0.05849825218319893,
490
+ "rewards/diff": -2.5624961853027344,
491
+ "rewards/diff_abs": 2.587552547454834,
492
+ "rewards/rejected": -0.21494324505329132,
493
+ "rewards/student_margin": 0.27344149351119995,
494
+ "rewards/teacher_margin": 2.835937738418579,
495
+ "step": 260
496
+ },
497
+ {
498
+ "epoch": 0.7,
499
+ "grad_norm": 1.5078125,
500
+ "learning_rate": 1.2171918633431623e-06,
501
+ "logits/chosen": -3.5318374633789062,
502
+ "logits/rejected": -3.4252419471740723,
503
+ "logps/chosen": -341.723388671875,
504
+ "logps/rejected": -344.2121887207031,
505
+ "loss": 0.6112,
506
+ "rewards/accuracies": 0.6333333253860474,
507
+ "rewards/chosen": -0.0180380679666996,
508
+ "rewards/diff": -1.8148130178451538,
509
+ "rewards/diff_abs": 1.9221045970916748,
510
+ "rewards/rejected": -0.31676679849624634,
511
+ "rewards/student_margin": 0.2987287640571594,
512
+ "rewards/teacher_margin": 2.113541603088379,
513
+ "step": 270
514
+ },
515
+ {
516
+ "epoch": 0.73,
517
+ "grad_norm": 1.546875,
518
+ "learning_rate": 1.0268451337516774e-06,
519
+ "logits/chosen": -3.6931426525115967,
520
+ "logits/rejected": -3.896336078643799,
521
+ "logps/chosen": -308.007080078125,
522
+ "logps/rejected": -167.8982696533203,
523
+ "loss": 0.6079,
524
+ "rewards/accuracies": 0.6333333253860474,
525
+ "rewards/chosen": -0.07022675126791,
526
+ "rewards/diff": -2.198145627975464,
527
+ "rewards/diff_abs": 2.226590394973755,
528
+ "rewards/rejected": -0.31322699785232544,
529
+ "rewards/student_margin": 0.24300022423267365,
530
+ "rewards/teacher_margin": 2.441145896911621,
531
+ "step": 280
532
+ },
533
+ {
534
+ "epoch": 0.76,
535
+ "grad_norm": 1.5,
536
+ "learning_rate": 8.487764541597765e-07,
537
+ "logits/chosen": -3.497781753540039,
538
+ "logits/rejected": -3.8565516471862793,
539
+ "logps/chosen": -267.2030944824219,
540
+ "logps/rejected": -168.0758819580078,
541
+ "loss": 0.6053,
542
+ "rewards/accuracies": 0.6999999284744263,
543
+ "rewards/chosen": -0.04164644330739975,
544
+ "rewards/diff": -2.201241970062256,
545
+ "rewards/diff_abs": 2.2781121730804443,
546
+ "rewards/rejected": -0.2831125855445862,
547
+ "rewards/student_margin": 0.24146613478660583,
548
+ "rewards/teacher_margin": 2.4427084922790527,
549
+ "step": 290
550
+ },
551
+ {
552
+ "epoch": 0.78,
553
+ "grad_norm": 1.5625,
554
+ "learning_rate": 6.844699429052377e-07,
555
+ "logits/chosen": -3.385387897491455,
556
+ "logits/rejected": -3.6356453895568848,
557
+ "logps/chosen": -409.85205078125,
558
+ "logps/rejected": -312.1107177734375,
559
+ "loss": 0.6067,
560
+ "rewards/accuracies": 0.699999988079071,
561
+ "rewards/chosen": -0.05188798904418945,
562
+ "rewards/diff": -1.7707669734954834,
563
+ "rewards/diff_abs": 1.8406012058258057,
564
+ "rewards/rejected": -0.28216272592544556,
565
+ "rewards/student_margin": 0.2302747219800949,
566
+ "rewards/teacher_margin": 2.001041889190674,
567
+ "step": 300
568
+ },
569
+ {
570
+ "epoch": 0.81,
571
+ "grad_norm": 1.5703125,
572
+ "learning_rate": 5.352950171529928e-07,
573
+ "logits/chosen": -3.574982166290283,
574
+ "logits/rejected": -3.6512961387634277,
575
+ "logps/chosen": -244.470947265625,
576
+ "logps/rejected": -193.90447998046875,
577
+ "loss": 0.6061,
578
+ "rewards/accuracies": 0.7333332896232605,
579
+ "rewards/chosen": -0.07677438855171204,
580
+ "rewards/diff": -2.3986549377441406,
581
+ "rewards/diff_abs": 2.4272804260253906,
582
+ "rewards/rejected": -0.37082797288894653,
583
+ "rewards/student_margin": 0.2940535545349121,
584
+ "rewards/teacher_margin": 2.6927084922790527,
585
+ "step": 310
586
+ },
587
+ {
588
+ "epoch": 0.84,
589
+ "grad_norm": 1.5703125,
590
+ "learning_rate": 4.024949794498623e-07,
591
+ "logits/chosen": -3.5370934009552,
592
+ "logits/rejected": -3.8875668048858643,
593
+ "logps/chosen": -254.0207977294922,
594
+ "logps/rejected": -177.51011657714844,
595
+ "loss": 0.6019,
596
+ "rewards/accuracies": 0.73333340883255,
597
+ "rewards/chosen": -0.05724753811955452,
598
+ "rewards/diff": -2.3326706886291504,
599
+ "rewards/diff_abs": 2.4501850605010986,
600
+ "rewards/rejected": -0.37457695603370667,
601
+ "rewards/student_margin": 0.31732940673828125,
602
+ "rewards/teacher_margin": 2.6500000953674316,
603
+ "step": 320
604
+ },
605
+ {
606
+ "epoch": 0.86,
607
+ "grad_norm": 1.6171875,
608
+ "learning_rate": 2.8717665538507965e-07,
609
+ "logits/chosen": -3.554170608520508,
610
+ "logits/rejected": -3.562087297439575,
611
+ "logps/chosen": -279.2240905761719,
612
+ "logps/rejected": -198.07553100585938,
613
+ "loss": 0.6066,
614
+ "rewards/accuracies": 0.6333333253860474,
615
+ "rewards/chosen": -0.03678184002637863,
616
+ "rewards/diff": -1.2939647436141968,
617
+ "rewards/diff_abs": 1.5477110147476196,
618
+ "rewards/rejected": -0.4365670084953308,
619
+ "rewards/student_margin": 0.39978522062301636,
620
+ "rewards/teacher_margin": 1.693750023841858,
621
+ "step": 330
622
+ },
623
+ {
624
+ "epoch": 0.89,
625
+ "grad_norm": 1.546875,
626
+ "learning_rate": 1.9030116872178317e-07,
627
+ "logits/chosen": -3.729553699493408,
628
+ "logits/rejected": -3.6963329315185547,
629
+ "logps/chosen": -294.4330139160156,
630
+ "logps/rejected": -217.95785522460938,
631
+ "loss": 0.6061,
632
+ "rewards/accuracies": 0.7333333492279053,
633
+ "rewards/chosen": 0.03005790151655674,
634
+ "rewards/diff": -1.652592420578003,
635
+ "rewards/diff_abs": 1.681958794593811,
636
+ "rewards/rejected": -0.3183913826942444,
637
+ "rewards/student_margin": 0.34844931960105896,
638
+ "rewards/teacher_margin": 2.001041889190674,
639
+ "step": 340
640
+ },
641
+ {
642
+ "epoch": 0.91,
643
+ "grad_norm": 1.5234375,
644
+ "learning_rate": 1.1267593088441886e-07,
645
+ "logits/chosen": -3.5918819904327393,
646
+ "logits/rejected": -3.487471103668213,
647
+ "logps/chosen": -301.32330322265625,
648
+ "logps/rejected": -268.36737060546875,
649
+ "loss": 0.6086,
650
+ "rewards/accuracies": 0.6333333253860474,
651
+ "rewards/chosen": -0.02564469538629055,
652
+ "rewards/diff": -1.5005762577056885,
653
+ "rewards/diff_abs": 1.6122653484344482,
654
+ "rewards/rejected": -0.3542352318763733,
655
+ "rewards/student_margin": 0.3285905420780182,
656
+ "rewards/teacher_margin": 1.8291666507720947,
657
+ "step": 350
658
+ },
659
+ {
660
+ "epoch": 0.94,
661
+ "grad_norm": 1.59375,
662
+ "learning_rate": 5.494791156587686e-08,
663
+ "logits/chosen": -3.7036406993865967,
664
+ "logits/rejected": -3.693377733230591,
665
+ "logps/chosen": -232.45254516601562,
666
+ "logps/rejected": -235.75521850585938,
667
+ "loss": 0.6033,
668
+ "rewards/accuracies": 0.6999999284744263,
669
+ "rewards/chosen": -0.03294830024242401,
670
+ "rewards/diff": -1.6850173473358154,
671
+ "rewards/diff_abs": 1.7215397357940674,
672
+ "rewards/rejected": -0.19897261261940002,
673
+ "rewards/student_margin": 0.166024312376976,
674
+ "rewards/teacher_margin": 1.851041555404663,
675
+ "step": 360
676
+ },
677
+ {
678
+ "epoch": 0.97,
679
+ "grad_norm": 1.5625,
680
+ "learning_rate": 1.7598246540683483e-08,
681
+ "logits/chosen": -3.8011093139648438,
682
+ "logits/rejected": -3.7970664501190186,
683
+ "logps/chosen": -246.9715118408203,
684
+ "logps/rejected": -200.6898956298828,
685
+ "loss": 0.609,
686
+ "rewards/accuracies": 0.7999999523162842,
687
+ "rewards/chosen": -0.005081920884549618,
688
+ "rewards/diff": -1.5501182079315186,
689
+ "rewards/diff_abs": 1.625860571861267,
690
+ "rewards/rejected": -0.3940262198448181,
691
+ "rewards/student_margin": 0.3889443278312683,
692
+ "rewards/teacher_margin": 1.9390627145767212,
693
+ "step": 370
694
+ },
695
+ {
696
+ "epoch": 0.99,
697
+ "grad_norm": 1.5625,
698
+ "learning_rate": 9.382276255742729e-10,
699
+ "logits/chosen": -3.510223388671875,
700
+ "logits/rejected": -3.580479145050049,
701
+ "logps/chosen": -379.2341003417969,
702
+ "logps/rejected": -310.8593444824219,
703
+ "loss": 0.6077,
704
+ "rewards/accuracies": 0.6333333253860474,
705
+ "rewards/chosen": 0.037930965423583984,
706
+ "rewards/diff": -1.598463535308838,
707
+ "rewards/diff_abs": 1.7268564701080322,
708
+ "rewards/rejected": -0.2682930827140808,
709
+ "rewards/student_margin": 0.3062240481376648,
710
+ "rewards/teacher_margin": 1.904687523841858,
711
+ "step": 380
712
+ },
713
+ {
714
+ "epoch": 1.0,
715
+ "step": 383,
716
+ "total_flos": 0.0,
717
+ "train_loss": 0.6328112833182432,
718
+ "train_runtime": 3006.7888,
719
+ "train_samples_per_second": 48.89,
720
+ "train_steps_per_second": 0.127
721
+ }
722
+ ],
723
+ "logging_steps": 10,
724
+ "max_steps": 383,
725
+ "num_input_tokens_seen": 0,
726
+ "num_train_epochs": 1,
727
+ "save_steps": 100000000000000000000000000000000,
728
+ "total_flos": 0.0,
729
+ "train_batch_size": 3,
730
+ "trial_name": null,
731
+ "trial_params": null
732
+ }