fengxb30 commited on
Commit
661b41a
·
verified ·
1 Parent(s): f128c64

Upload trainer_state.json

Browse files
Files changed (1) hide show
  1. trainer_state.json +641 -0
trainer_state.json ADDED
@@ -0,0 +1,641 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 1.9666011787819253,
6
+ "eval_steps": 500,
7
+ "global_step": 500,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "clip_ratio": 0.0,
14
+ "completions/clipped_ratio": 0.0,
15
+ "completions/max_length": 579.5,
16
+ "completions/max_terminated_length": 579.5,
17
+ "completions/mean_length": 445.5,
18
+ "completions/mean_terminated_length": 445.5,
19
+ "completions/min_length": 358.5,
20
+ "completions/min_terminated_length": 358.5,
21
+ "epoch": 0.003929273084479371,
22
+ "grad_norm": 0.46015690099076406,
23
+ "kl": 0.0,
24
+ "learning_rate": 0.0,
25
+ "loss": -0.0157,
26
+ "num_tokens": 18992.0,
27
+ "reward": 9.166666984558105,
28
+ "reward_std": 1.178511381149292,
29
+ "rewards/accuracy_reward/mean": 1.8333333134651184,
30
+ "rewards/accuracy_reward/std": 0.40824830532073975,
31
+ "rewards/format_reward/mean": 0.0,
32
+ "rewards/format_reward/std": 0.0,
33
+ "step": 1
34
+ },
35
+ {
36
+ "clip_ratio": 0.0,
37
+ "completions/clipped_ratio": 0.0,
38
+ "completions/max_length": 554.4736842105264,
39
+ "completions/max_terminated_length": 554.4736842105264,
40
+ "completions/mean_length": 427.37282040244656,
41
+ "completions/mean_terminated_length": 427.37282040244656,
42
+ "completions/min_length": 322.5,
43
+ "completions/min_terminated_length": 322.5,
44
+ "epoch": 0.07858546168958742,
45
+ "grad_norm": 0.37367734361954813,
46
+ "kl": 0.0002414552788985403,
47
+ "learning_rate": 1.14e-06,
48
+ "loss": 0.0021,
49
+ "num_tokens": 366755.0,
50
+ "reward": 9.62719312467073,
51
+ "reward_std": 0.5272287757773149,
52
+ "rewards/accuracy_reward/mean": 1.9254385891713595,
53
+ "rewards/accuracy_reward/std": 0.18263739974875198,
54
+ "rewards/format_reward/mean": 0.0,
55
+ "rewards/format_reward/std": 0.0,
56
+ "step": 20
57
+ },
58
+ {
59
+ "clip_ratio": 0.0,
60
+ "completions/clipped_ratio": 0.0,
61
+ "completions/max_length": 543.875,
62
+ "completions/max_terminated_length": 543.875,
63
+ "completions/mean_length": 423.0708427429199,
64
+ "completions/mean_terminated_length": 423.0708427429199,
65
+ "completions/min_length": 319.9,
66
+ "completions/min_terminated_length": 319.9,
67
+ "epoch": 0.15717092337917485,
68
+ "grad_norm": 0.5322056075322863,
69
+ "kl": 0.0003676414489746094,
70
+ "learning_rate": 2.34e-06,
71
+ "loss": 0.0042,
72
+ "num_tokens": 745300.0,
73
+ "reward": 9.479166793823243,
74
+ "reward_std": 0.5008673369884491,
75
+ "rewards/accuracy_reward/mean": 1.8958333283662796,
76
+ "rewards/accuracy_reward/std": 0.20629913210868836,
77
+ "rewards/format_reward/mean": 0.0,
78
+ "rewards/format_reward/std": 0.0,
79
+ "step": 40
80
+ },
81
+ {
82
+ "clip_ratio": 0.0,
83
+ "completions/clipped_ratio": 0.0,
84
+ "completions/max_length": 571.275,
85
+ "completions/max_terminated_length": 571.275,
86
+ "completions/mean_length": 416.7875122070312,
87
+ "completions/mean_terminated_length": 416.7875122070312,
88
+ "completions/min_length": 318.6,
89
+ "completions/min_terminated_length": 318.6,
90
+ "epoch": 0.2357563850687623,
91
+ "grad_norm": 0.46502934681984226,
92
+ "kl": 0.0017244338989257813,
93
+ "learning_rate": 2.9970400926424076e-06,
94
+ "loss": 0.0059,
95
+ "num_tokens": 1102433.0,
96
+ "reward": 9.479166793823243,
97
+ "reward_std": 0.5597929060459137,
98
+ "rewards/accuracy_reward/mean": 1.8958333283662796,
99
+ "rewards/accuracy_reward/std": 0.2154431849718094,
100
+ "rewards/format_reward/mean": 0.0,
101
+ "rewards/format_reward/std": 0.0,
102
+ "step": 60
103
+ },
104
+ {
105
+ "clip_ratio": 0.0,
106
+ "completions/clipped_ratio": 0.0,
107
+ "completions/max_length": 529.75,
108
+ "completions/max_terminated_length": 529.75,
109
+ "completions/mean_length": 416.1666778564453,
110
+ "completions/mean_terminated_length": 416.1666778564453,
111
+ "completions/min_length": 324.725,
112
+ "completions/min_terminated_length": 324.725,
113
+ "epoch": 0.3143418467583497,
114
+ "grad_norm": 0.03393033103037874,
115
+ "kl": 0.005292510986328125,
116
+ "learning_rate": 2.969362874399016e-06,
117
+ "loss": -0.0009,
118
+ "num_tokens": 1476975.0,
119
+ "reward": 9.583333444595336,
120
+ "reward_std": 0.3535534143447876,
121
+ "rewards/accuracy_reward/mean": 1.9166666626930238,
122
+ "rewards/accuracy_reward/std": 0.17411426901817323,
123
+ "rewards/format_reward/mean": 0.0,
124
+ "rewards/format_reward/std": 0.0,
125
+ "step": 80
126
+ },
127
+ {
128
+ "clip_ratio": 0.0,
129
+ "completions/clipped_ratio": 0.0,
130
+ "completions/max_length": 527.25,
131
+ "completions/max_terminated_length": 527.25,
132
+ "completions/mean_length": 416.6166793823242,
133
+ "completions/mean_terminated_length": 416.6166793823242,
134
+ "completions/min_length": 323.575,
135
+ "completions/min_terminated_length": 323.575,
136
+ "epoch": 0.3929273084479371,
137
+ "grad_norm": 0.030903746658924047,
138
+ "kl": 0.008585357666015625,
139
+ "learning_rate": 2.913086179180945e-06,
140
+ "loss": 0.0066,
141
+ "num_tokens": 1840061.0,
142
+ "reward": 9.604166769981385,
143
+ "reward_std": 0.5597929060459137,
144
+ "rewards/accuracy_reward/mean": 1.920833334326744,
145
+ "rewards/accuracy_reward/std": 0.16390806138515474,
146
+ "rewards/format_reward/mean": 0.0,
147
+ "rewards/format_reward/std": 0.0,
148
+ "step": 100
149
+ },
150
+ {
151
+ "clip_ratio": 0.0,
152
+ "completions/clipped_ratio": 0.0,
153
+ "completions/max_length": 541.625,
154
+ "completions/max_terminated_length": 541.625,
155
+ "completions/mean_length": 409.9083450317383,
156
+ "completions/mean_terminated_length": 409.9083450317383,
157
+ "completions/min_length": 306.25,
158
+ "completions/min_terminated_length": 306.25,
159
+ "epoch": 0.4715127701375246,
160
+ "grad_norm": 0.45989535358245415,
161
+ "kl": 0.012615966796875,
162
+ "learning_rate": 2.829305368846822e-06,
163
+ "loss": -0.0081,
164
+ "num_tokens": 2205137.0,
165
+ "reward": 9.583333444595336,
166
+ "reward_std": 0.5303301155567169,
167
+ "rewards/accuracy_reward/mean": 1.9166666597127915,
168
+ "rewards/accuracy_reward/std": 0.18471990823745726,
169
+ "rewards/format_reward/mean": 0.0,
170
+ "rewards/format_reward/std": 0.0,
171
+ "step": 120
172
+ },
173
+ {
174
+ "clip_ratio": 0.0,
175
+ "completions/clipped_ratio": 0.0,
176
+ "completions/max_length": 566.95,
177
+ "completions/max_terminated_length": 566.95,
178
+ "completions/mean_length": 433.34167709350584,
179
+ "completions/mean_terminated_length": 433.34167709350584,
180
+ "completions/min_length": 330.4,
181
+ "completions/min_terminated_length": 330.4,
182
+ "epoch": 0.550098231827112,
183
+ "grad_norm": 0.037923463367969736,
184
+ "kl": 0.0144561767578125,
185
+ "learning_rate": 2.7196511415705416e-06,
186
+ "loss": 0.0048,
187
+ "num_tokens": 2588775.0,
188
+ "reward": 9.416666793823243,
189
+ "reward_std": 0.4714045524597168,
190
+ "rewards/accuracy_reward/mean": 1.883333331346512,
191
+ "rewards/accuracy_reward/std": 0.2257540464401245,
192
+ "rewards/format_reward/mean": 0.0,
193
+ "rewards/format_reward/std": 0.0,
194
+ "step": 140
195
+ },
196
+ {
197
+ "clip_ratio": 0.0,
198
+ "completions/clipped_ratio": 0.0,
199
+ "completions/max_length": 567.8,
200
+ "completions/max_terminated_length": 567.8,
201
+ "completions/mean_length": 438.3958457946777,
202
+ "completions/mean_terminated_length": 438.3958457946777,
203
+ "completions/min_length": 330.85,
204
+ "completions/min_terminated_length": 330.85,
205
+ "epoch": 0.6286836935166994,
206
+ "grad_norm": 0.03890017799267318,
207
+ "kl": 0.0144439697265625,
208
+ "learning_rate": 2.5862577921562017e-06,
209
+ "loss": 0.0022,
210
+ "num_tokens": 2969880.0,
211
+ "reward": 9.70833342075348,
212
+ "reward_std": 0.294627845287323,
213
+ "rewards/accuracy_reward/mean": 1.941666665673256,
214
+ "rewards/accuracy_reward/std": 0.1278819650411606,
215
+ "rewards/format_reward/mean": 0.0,
216
+ "rewards/format_reward/std": 0.0,
217
+ "step": 160
218
+ },
219
+ {
220
+ "clip_ratio": 0.0,
221
+ "completions/clipped_ratio": 0.0,
222
+ "completions/max_length": 542.925,
223
+ "completions/max_terminated_length": 542.925,
224
+ "completions/mean_length": 432.9458457946777,
225
+ "completions/mean_terminated_length": 432.9458457946777,
226
+ "completions/min_length": 343.425,
227
+ "completions/min_terminated_length": 343.425,
228
+ "epoch": 0.7072691552062869,
229
+ "grad_norm": 0.029836981989722188,
230
+ "kl": 0.01332855224609375,
231
+ "learning_rate": 2.4317216704174657e-06,
232
+ "loss": 0.0032,
233
+ "num_tokens": 3344559.0,
234
+ "reward": 9.666666722297668,
235
+ "reward_std": 0.3535534143447876,
236
+ "rewards/accuracy_reward/mean": 1.933333334326744,
237
+ "rewards/accuracy_reward/std": 0.11828449666500092,
238
+ "rewards/format_reward/mean": 0.0,
239
+ "rewards/format_reward/std": 0.0,
240
+ "step": 180
241
+ },
242
+ {
243
+ "clip_ratio": 0.0,
244
+ "completions/clipped_ratio": 0.0,
245
+ "completions/max_length": 553.825,
246
+ "completions/max_terminated_length": 553.825,
247
+ "completions/mean_length": 432.8000114440918,
248
+ "completions/mean_terminated_length": 432.8000114440918,
249
+ "completions/min_length": 329.975,
250
+ "completions/min_terminated_length": 329.975,
251
+ "epoch": 0.7858546168958742,
252
+ "grad_norm": 0.028665697797427125,
253
+ "kl": 0.01359100341796875,
254
+ "learning_rate": 2.2590506461817453e-06,
255
+ "loss": 0.0011,
256
+ "num_tokens": 3723883.0,
257
+ "reward": 9.666666769981385,
258
+ "reward_std": 0.4714045524597168,
259
+ "rewards/accuracy_reward/mean": 1.9333333283662797,
260
+ "rewards/accuracy_reward/std": 0.14829438030719758,
261
+ "rewards/format_reward/mean": 0.0,
262
+ "rewards/format_reward/std": 0.0,
263
+ "step": 200
264
+ },
265
+ {
266
+ "clip_ratio": 0.0,
267
+ "completions/clipped_ratio": 0.0,
268
+ "completions/max_length": 565.9,
269
+ "completions/max_terminated_length": 565.9,
270
+ "completions/mean_length": 442.46251068115237,
271
+ "completions/mean_terminated_length": 442.46251068115237,
272
+ "completions/min_length": 349.45,
273
+ "completions/min_terminated_length": 349.45,
274
+ "epoch": 0.8644400785854617,
275
+ "grad_norm": 0.03081820386262942,
276
+ "kl": 0.01476898193359375,
277
+ "learning_rate": 2.0716055645254116e-06,
278
+ "loss": -0.0006,
279
+ "num_tokens": 4106760.0,
280
+ "reward": 9.812500023841858,
281
+ "reward_std": 0.2062394917011261,
282
+ "rewards/accuracy_reward/mean": 1.9625000029802322,
283
+ "rewards/accuracy_reward/std": 0.06714880466461182,
284
+ "rewards/format_reward/mean": 0.0,
285
+ "rewards/format_reward/std": 0.0,
286
+ "step": 220
287
+ },
288
+ {
289
+ "clip_ratio": 0.0,
290
+ "completions/clipped_ratio": 0.0,
291
+ "completions/max_length": 570.925,
292
+ "completions/max_terminated_length": 570.925,
293
+ "completions/mean_length": 430.3500122070312,
294
+ "completions/mean_terminated_length": 430.3500122070312,
295
+ "completions/min_length": 335.975,
296
+ "completions/min_terminated_length": 335.975,
297
+ "epoch": 0.9430255402750491,
298
+ "grad_norm": 0.0325454508362931,
299
+ "kl": 0.01363067626953125,
300
+ "learning_rate": 1.8730348307472826e-06,
301
+ "loss": -0.0006,
302
+ "num_tokens": 4498062.0,
303
+ "reward": 9.70833342075348,
304
+ "reward_std": 0.4124789834022522,
305
+ "rewards/accuracy_reward/mean": 1.9416666626930237,
306
+ "rewards/accuracy_reward/std": 0.1278819650411606,
307
+ "rewards/format_reward/mean": 0.0,
308
+ "rewards/format_reward/std": 0.0,
309
+ "step": 240
310
+ },
311
+ {
312
+ "clip_ratio": 0.0,
313
+ "completions/clipped_ratio": 0.0,
314
+ "completions/max_length": 546.625,
315
+ "completions/max_terminated_length": 546.625,
316
+ "completions/mean_length": 429.0875114440918,
317
+ "completions/mean_terminated_length": 429.0875114440918,
318
+ "completions/min_length": 333.85,
319
+ "completions/min_terminated_length": 333.85,
320
+ "epoch": 1.0235756385068762,
321
+ "grad_norm": 0.0281329807884298,
322
+ "kl": 0.01484832763671875,
323
+ "learning_rate": 1.667203398309488e-06,
324
+ "loss": -0.0006,
325
+ "num_tokens": 4844839.0,
326
+ "reward": 9.666666746139526,
327
+ "reward_std": 0.47140454649925234,
328
+ "rewards/accuracy_reward/mean": 1.9333333313465118,
329
+ "rewards/accuracy_reward/std": 0.13859225809574127,
330
+ "rewards/format_reward/mean": 0.0,
331
+ "rewards/format_reward/std": 0.0,
332
+ "step": 260
333
+ },
334
+ {
335
+ "clip_ratio": 0.0,
336
+ "completions/clipped_ratio": 0.0,
337
+ "completions/max_length": 530.05,
338
+ "completions/max_terminated_length": 530.05,
339
+ "completions/mean_length": 412.9875099182129,
340
+ "completions/mean_terminated_length": 412.9875099182129,
341
+ "completions/min_length": 319.675,
342
+ "completions/min_terminated_length": 319.675,
343
+ "epoch": 1.1021611001964637,
344
+ "grad_norm": 0.03324322790874101,
345
+ "kl": 0.0139068603515625,
346
+ "learning_rate": 1.458117541914647e-06,
347
+ "loss": 0.0015,
348
+ "num_tokens": 5222808.0,
349
+ "reward": 9.791666746139526,
350
+ "reward_std": 0.294627845287323,
351
+ "rewards/accuracy_reward/mean": 1.9583333313465119,
352
+ "rewards/accuracy_reward/std": 0.10206207633018494,
353
+ "rewards/format_reward/mean": 0.0,
354
+ "rewards/format_reward/std": 0.0,
355
+ "step": 280
356
+ },
357
+ {
358
+ "clip_ratio": 0.0,
359
+ "completions/clipped_ratio": 0.0,
360
+ "completions/max_length": 524.825,
361
+ "completions/max_terminated_length": 524.825,
362
+ "completions/mean_length": 404.3958435058594,
363
+ "completions/mean_terminated_length": 404.3958435058594,
364
+ "completions/min_length": 308.875,
365
+ "completions/min_terminated_length": 308.875,
366
+ "epoch": 1.180746561886051,
367
+ "grad_norm": 0.03412086623724304,
368
+ "kl": 0.0137176513671875,
369
+ "learning_rate": 1.2498468799258468e-06,
370
+ "loss": 0.0031,
371
+ "num_tokens": 5583225.0,
372
+ "reward": 9.791666746139526,
373
+ "reward_std": 0.294627845287323,
374
+ "rewards/accuracy_reward/mean": 1.9583333283662796,
375
+ "rewards/accuracy_reward/std": 0.10206207633018494,
376
+ "rewards/format_reward/mean": 0.0,
377
+ "rewards/format_reward/std": 0.0,
378
+ "step": 300
379
+ },
380
+ {
381
+ "clip_ratio": 0.0,
382
+ "completions/clipped_ratio": 0.0,
383
+ "completions/max_length": 516.05,
384
+ "completions/max_terminated_length": 516.05,
385
+ "completions/mean_length": 395.33334197998045,
386
+ "completions/mean_terminated_length": 395.33334197998045,
387
+ "completions/min_length": 304.5,
388
+ "completions/min_terminated_length": 304.5,
389
+ "epoch": 1.2593320235756384,
390
+ "grad_norm": 0.03048435461570321,
391
+ "kl": 0.01298065185546875,
392
+ "learning_rate": 1.0464451638743333e-06,
393
+ "loss": 0.0031,
394
+ "num_tokens": 5944449.0,
395
+ "reward": 9.770833396911621,
396
+ "reward_std": 0.2062394917011261,
397
+ "rewards/accuracy_reward/mean": 1.9541666686534882,
398
+ "rewards/accuracy_reward/std": 0.09726334214210511,
399
+ "rewards/format_reward/mean": 0.0,
400
+ "rewards/format_reward/std": 0.0,
401
+ "step": 320
402
+ },
403
+ {
404
+ "clip_ratio": 0.0,
405
+ "completions/clipped_ratio": 0.0,
406
+ "completions/max_length": 553.4,
407
+ "completions/max_terminated_length": 553.4,
408
+ "completions/mean_length": 425.20000915527345,
409
+ "completions/mean_terminated_length": 425.20000915527345,
410
+ "completions/min_length": 323.3,
411
+ "completions/min_terminated_length": 323.3,
412
+ "epoch": 1.3379174852652258,
413
+ "grad_norm": 0.7144758682156833,
414
+ "kl": 0.01336822509765625,
415
+ "learning_rate": 8.518713767970263e-07,
416
+ "loss": 0.0009,
417
+ "num_tokens": 6321679.0,
418
+ "reward": 9.604166793823243,
419
+ "reward_std": 0.4419417679309845,
420
+ "rewards/accuracy_reward/mean": 1.9208333313465118,
421
+ "rewards/accuracy_reward/std": 0.17891300320625306,
422
+ "rewards/format_reward/mean": 0.0,
423
+ "rewards/format_reward/std": 0.0,
424
+ "step": 340
425
+ },
426
+ {
427
+ "clip_ratio": 0.0,
428
+ "completions/clipped_ratio": 0.0,
429
+ "completions/max_length": 574.225,
430
+ "completions/max_terminated_length": 574.225,
431
+ "completions/mean_length": 443.48334350585935,
432
+ "completions/mean_terminated_length": 443.48334350585935,
433
+ "completions/min_length": 340.075,
434
+ "completions/min_terminated_length": 340.075,
435
+ "epoch": 1.4165029469548134,
436
+ "grad_norm": 0.03770084296737852,
437
+ "kl": 0.01368255615234375,
438
+ "learning_rate": 6.69912676134984e-07,
439
+ "loss": 0.0042,
440
+ "num_tokens": 6707325.0,
441
+ "reward": 9.666666746139526,
442
+ "reward_std": 0.35355340838432314,
443
+ "rewards/accuracy_reward/mean": 1.9333333313465118,
444
+ "rewards/accuracy_reward/std": 0.13859225809574127,
445
+ "rewards/format_reward/mean": 0.0,
446
+ "rewards/format_reward/std": 0.0,
447
+ "step": 360
448
+ },
449
+ {
450
+ "clip_ratio": 0.0,
451
+ "completions/clipped_ratio": 0.0,
452
+ "completions/max_length": 562.675,
453
+ "completions/max_terminated_length": 562.675,
454
+ "completions/mean_length": 438.5416793823242,
455
+ "completions/mean_terminated_length": 438.5416793823242,
456
+ "completions/min_length": 346.0,
457
+ "completions/min_terminated_length": 346.0,
458
+ "epoch": 1.4950884086444007,
459
+ "grad_norm": 0.02727792542475105,
460
+ "kl": 0.0121246337890625,
461
+ "learning_rate": 5.041106810216376e-07,
462
+ "loss": 0.0017,
463
+ "num_tokens": 7081575.0,
464
+ "reward": 9.604166769981385,
465
+ "reward_std": 0.4419417679309845,
466
+ "rewards/accuracy_reward/mean": 1.9208333283662795,
467
+ "rewards/accuracy_reward/std": 0.16006682813167572,
468
+ "rewards/format_reward/mean": 0.0,
469
+ "rewards/format_reward/std": 0.0,
470
+ "step": 380
471
+ },
472
+ {
473
+ "clip_ratio": 0.0,
474
+ "completions/clipped_ratio": 0.0,
475
+ "completions/max_length": 549.7,
476
+ "completions/max_terminated_length": 549.7,
477
+ "completions/mean_length": 434.0916748046875,
478
+ "completions/mean_terminated_length": 434.0916748046875,
479
+ "completions/min_length": 341.4,
480
+ "completions/min_terminated_length": 341.4,
481
+ "epoch": 1.5736738703339883,
482
+ "grad_norm": 0.030573782966379155,
483
+ "kl": 0.012384033203125,
484
+ "learning_rate": 3.5769253869489515e-07,
485
+ "loss": -0.0017,
486
+ "num_tokens": 7465665.0,
487
+ "reward": 9.77083342075348,
488
+ "reward_std": 0.3240906298160553,
489
+ "rewards/accuracy_reward/mean": 1.9541666626930236,
490
+ "rewards/accuracy_reward/std": 0.11226828396320343,
491
+ "rewards/format_reward/mean": 0.0,
492
+ "rewards/format_reward/std": 0.0,
493
+ "step": 400
494
+ },
495
+ {
496
+ "clip_ratio": 0.0,
497
+ "completions/clipped_ratio": 0.0,
498
+ "completions/max_length": 529.95,
499
+ "completions/max_terminated_length": 529.95,
500
+ "completions/mean_length": 423.4125099182129,
501
+ "completions/mean_terminated_length": 423.4125099182129,
502
+ "completions/min_length": 337.4,
503
+ "completions/min_terminated_length": 337.4,
504
+ "epoch": 1.6522593320235757,
505
+ "grad_norm": 0.026892810354609234,
506
+ "kl": 0.011712646484375,
507
+ "learning_rate": 2.3350811174697772e-07,
508
+ "loss": 0.0009,
509
+ "num_tokens": 7851292.0,
510
+ "reward": 9.895833349227905,
511
+ "reward_std": 0.1473139226436615,
512
+ "rewards/accuracy_reward/mean": 1.9791666686534881,
513
+ "rewards/accuracy_reward/std": 0.03602609634399414,
514
+ "rewards/format_reward/mean": 0.0,
515
+ "rewards/format_reward/std": 0.0,
516
+ "step": 420
517
+ },
518
+ {
519
+ "clip_ratio": 0.0,
520
+ "completions/clipped_ratio": 0.0,
521
+ "completions/max_length": 548.65,
522
+ "completions/max_terminated_length": 548.65,
523
+ "completions/mean_length": 438.77084121704104,
524
+ "completions/mean_terminated_length": 438.77084121704104,
525
+ "completions/min_length": 344.9,
526
+ "completions/min_terminated_length": 344.9,
527
+ "epoch": 1.730844793713163,
528
+ "grad_norm": 0.7530350302626055,
529
+ "kl": 0.012438201904296875,
530
+ "learning_rate": 1.3397450879073447e-07,
531
+ "loss": 0.0001,
532
+ "num_tokens": 8224807.0,
533
+ "reward": 9.68750011920929,
534
+ "reward_std": 0.4419417679309845,
535
+ "rewards/accuracy_reward/mean": 1.9374999940395354,
536
+ "rewards/accuracy_reward/std": 0.1530931144952774,
537
+ "rewards/format_reward/mean": 0.0,
538
+ "rewards/format_reward/std": 0.0,
539
+ "step": 440
540
+ },
541
+ {
542
+ "clip_ratio": 0.0,
543
+ "completions/clipped_ratio": 0.0,
544
+ "completions/max_length": 531.9,
545
+ "completions/max_terminated_length": 531.9,
546
+ "completions/mean_length": 419.85418167114256,
547
+ "completions/mean_terminated_length": 419.85418167114256,
548
+ "completions/min_length": 322.85,
549
+ "completions/min_terminated_length": 322.85,
550
+ "epoch": 1.8094302554027504,
551
+ "grad_norm": 0.3868391159083008,
552
+ "kl": 0.011206817626953126,
553
+ "learning_rate": 6.102903818991374e-08,
554
+ "loss": 0.0016,
555
+ "num_tokens": 8589626.0,
556
+ "reward": 9.55208342075348,
557
+ "reward_std": 0.5155987203121185,
558
+ "rewards/accuracy_reward/mean": 1.910416665673256,
559
+ "rewards/accuracy_reward/std": 0.1655041679739952,
560
+ "rewards/format_reward/mean": 0.0,
561
+ "rewards/format_reward/std": 0.0,
562
+ "step": 460
563
+ },
564
+ {
565
+ "clip_ratio": 0.0,
566
+ "completions/clipped_ratio": 0.0,
567
+ "completions/max_length": 554.8,
568
+ "completions/max_terminated_length": 554.8,
569
+ "completions/mean_length": 438.46667556762696,
570
+ "completions/mean_terminated_length": 438.46667556762696,
571
+ "completions/min_length": 341.875,
572
+ "completions/min_terminated_length": 341.875,
573
+ "epoch": 1.888015717092338,
574
+ "grad_norm": 0.05259089172915785,
575
+ "kl": 0.011602783203125,
576
+ "learning_rate": 1.60915005555175e-08,
577
+ "loss": -0.0007,
578
+ "num_tokens": 8961090.0,
579
+ "reward": 9.562500095367431,
580
+ "reward_std": 0.3830161988735199,
581
+ "rewards/accuracy_reward/mean": 1.9125,
582
+ "rewards/accuracy_reward/std": 0.16931553483009337,
583
+ "rewards/format_reward/mean": 0.0,
584
+ "rewards/format_reward/std": 0.0,
585
+ "step": 480
586
+ },
587
+ {
588
+ "clip_ratio": 0.0,
589
+ "completions/clipped_ratio": 0.0,
590
+ "completions/max_length": 544.9,
591
+ "completions/max_terminated_length": 544.9,
592
+ "completions/mean_length": 435.1583450317383,
593
+ "completions/mean_terminated_length": 435.1583450317383,
594
+ "completions/min_length": 346.2,
595
+ "completions/min_terminated_length": 346.2,
596
+ "epoch": 1.9666011787819253,
597
+ "grad_norm": 0.02646975626285412,
598
+ "kl": 0.01181793212890625,
599
+ "learning_rate": 3.655394190787975e-11,
600
+ "loss": 0.0015,
601
+ "num_tokens": 9338954.0,
602
+ "reward": 9.750000047683717,
603
+ "reward_std": 0.2357022762298584,
604
+ "rewards/accuracy_reward/mean": 1.9499999970197677,
605
+ "rewards/accuracy_reward/std": 0.08862337470054626,
606
+ "rewards/format_reward/mean": 0.0,
607
+ "rewards/format_reward/std": 0.0,
608
+ "step": 500
609
+ },
610
+ {
611
+ "epoch": 1.9666011787819253,
612
+ "step": 500,
613
+ "total_flos": 0.0,
614
+ "train_loss": 0.0013743292051367462,
615
+ "train_runtime": 7960.7406,
616
+ "train_samples_per_second": 0.754,
617
+ "train_steps_per_second": 0.063
618
+ }
619
+ ],
620
+ "logging_steps": 20,
621
+ "max_steps": 500,
622
+ "num_input_tokens_seen": 9338954,
623
+ "num_train_epochs": 2,
624
+ "save_steps": 500,
625
+ "stateful_callbacks": {
626
+ "TrainerControl": {
627
+ "args": {
628
+ "should_epoch_stop": false,
629
+ "should_evaluate": false,
630
+ "should_log": false,
631
+ "should_save": false,
632
+ "should_training_stop": false
633
+ },
634
+ "attributes": {}
635
+ }
636
+ },
637
+ "total_flos": 0.0,
638
+ "train_batch_size": 2,
639
+ "trial_name": null,
640
+ "trial_params": null
641
+ }