End of training
Browse files- all_results.json +5 -5
- model-00001-of-00002.safetensors +1 -1
- model-00002-of-00002.safetensors +1 -1
- train_results.json +5 -5
- trainer_state.json +602 -602
- training_args.bin +1 -1
all_results.json
CHANGED
@@ -1,8 +1,8 @@
|
|
1 |
{
|
2 |
"total_flos": 0.0,
|
3 |
-
"train_loss": 1.
|
4 |
-
"train_runtime":
|
5 |
-
"train_samples":
|
6 |
-
"train_samples_per_second": 0.
|
7 |
-
"train_steps_per_second": 0.
|
8 |
}
|
|
|
1 |
{
|
2 |
"total_flos": 0.0,
|
3 |
+
"train_loss": 1.5718676149845123e-05,
|
4 |
+
"train_runtime": 4262.7362,
|
5 |
+
"train_samples": 48,
|
6 |
+
"train_samples_per_second": 0.375,
|
7 |
+
"train_steps_per_second": 0.023
|
8 |
}
|
model-00001-of-00002.safetensors
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 4996670464
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:7aa4f6bae2562b59ce5b4c50ed5ba43bb6c7a6490d3bdbfd840200784bb86db8
|
3 |
size 4996670464
|
model-00002-of-00002.safetensors
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 1178224960
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:e5d2dcd5763357799ad9114059b0028a01c6683f147409faa0f31aafbd95ece0
|
3 |
size 1178224960
|
train_results.json
CHANGED
@@ -1,8 +1,8 @@
|
|
1 |
{
|
2 |
"total_flos": 0.0,
|
3 |
-
"train_loss": 1.
|
4 |
-
"train_runtime":
|
5 |
-
"train_samples":
|
6 |
-
"train_samples_per_second": 0.
|
7 |
-
"train_steps_per_second": 0.
|
8 |
}
|
|
|
1 |
{
|
2 |
"total_flos": 0.0,
|
3 |
+
"train_loss": 1.5718676149845123e-05,
|
4 |
+
"train_runtime": 4262.7362,
|
5 |
+
"train_samples": 48,
|
6 |
+
"train_samples_per_second": 0.375,
|
7 |
+
"train_steps_per_second": 0.023
|
8 |
}
|
trainer_state.json
CHANGED
@@ -2,7 +2,7 @@
|
|
2 |
"best_global_step": null,
|
3 |
"best_metric": null,
|
4 |
"best_model_checkpoint": null,
|
5 |
-
"epoch":
|
6 |
"eval_steps": 500,
|
7 |
"global_step": 100,
|
8 |
"is_hyper_param_search": false,
|
@@ -10,969 +10,969 @@
|
|
10 |
"is_world_process_zero": true,
|
11 |
"log_history": [
|
12 |
{
|
13 |
-
"completion_length":
|
14 |
-
"epoch": 0.
|
15 |
-
"grad_norm":
|
16 |
"kl": 0.0,
|
17 |
"learning_rate": 1.6666666666666665e-07,
|
18 |
-
"loss": 0.0,
|
19 |
-
"reward":
|
20 |
-
"reward_std":
|
21 |
-
"rewards/concensus_correctness_reward_func":
|
22 |
-
"rewards/consensus_reward_func": 0.
|
23 |
"rewards/cumulative_reward_2": 0.0,
|
24 |
-
"rewards/final_correctness_reward_func": 0.
|
25 |
-
"rewards/question_recreation_reward_func": 0.
|
26 |
"rewards/soft_format_reward_func": 0.0,
|
27 |
-
"rewards/strict_format_reward_func": 0.
|
28 |
-
"rewards/xmlcount_reward_func": 0.
|
29 |
"step": 2
|
30 |
},
|
31 |
{
|
32 |
-
"completion_length":
|
33 |
-
"epoch": 0.
|
34 |
-
"grad_norm": 2.
|
35 |
-
"kl": 0.
|
36 |
"learning_rate": 5e-07,
|
37 |
-
"loss":
|
38 |
-
"reward":
|
39 |
-
"reward_std": 0.
|
40 |
-
"rewards/concensus_correctness_reward_func":
|
41 |
-
"rewards/consensus_reward_func": 0.
|
42 |
"rewards/cumulative_reward_2": 0.0,
|
43 |
-
"rewards/final_correctness_reward_func": 0.
|
44 |
-
"rewards/question_recreation_reward_func": 0.
|
45 |
"rewards/soft_format_reward_func": 0.0,
|
46 |
"rewards/strict_format_reward_func": 0.015625,
|
47 |
-
"rewards/xmlcount_reward_func": 0.
|
48 |
"step": 4
|
49 |
},
|
50 |
{
|
51 |
-
"completion_length":
|
52 |
-
"epoch": 0
|
53 |
-
"grad_norm": 2.
|
54 |
-
"kl": 0.
|
55 |
"learning_rate": 4.994757065594279e-07,
|
56 |
"loss": 0.0,
|
57 |
-
"reward": 1.
|
58 |
-
"reward_std":
|
59 |
-
"rewards/concensus_correctness_reward_func": 0.
|
60 |
-
"rewards/consensus_reward_func": 0.
|
61 |
"rewards/cumulative_reward_2": 0.0,
|
62 |
-
"rewards/final_correctness_reward_func": 0.
|
63 |
-
"rewards/question_recreation_reward_func": 0.
|
64 |
"rewards/soft_format_reward_func": 0.0,
|
65 |
"rewards/strict_format_reward_func": 0.03125,
|
66 |
-
"rewards/xmlcount_reward_func": 0.
|
67 |
"step": 6
|
68 |
},
|
69 |
{
|
70 |
-
"completion_length":
|
71 |
-
"epoch":
|
72 |
-
"grad_norm":
|
73 |
-
"kl": 0.
|
74 |
"learning_rate": 4.979050253066063e-07,
|
75 |
"loss": 0.0,
|
76 |
-
"reward":
|
77 |
-
"reward_std":
|
78 |
-
"rewards/concensus_correctness_reward_func":
|
79 |
-
"rewards/consensus_reward_func": 0.
|
80 |
"rewards/cumulative_reward_2": 0.0,
|
81 |
-
"rewards/final_correctness_reward_func": 0.
|
82 |
-
"rewards/question_recreation_reward_func": 0.
|
83 |
"rewards/soft_format_reward_func": 0.0,
|
84 |
-
"rewards/strict_format_reward_func": 0.
|
85 |
-
"rewards/xmlcount_reward_func": 0.
|
86 |
"step": 8
|
87 |
},
|
88 |
{
|
89 |
-
"completion_length":
|
90 |
-
"epoch":
|
91 |
-
"grad_norm":
|
92 |
-
"kl": 0.
|
93 |
"learning_rate": 4.952945442245597e-07,
|
94 |
-
"loss":
|
95 |
-
"reward":
|
96 |
-
"reward_std": 1.
|
97 |
-
"rewards/concensus_correctness_reward_func":
|
98 |
-
"rewards/consensus_reward_func": 0.
|
99 |
"rewards/cumulative_reward_2": 0.0,
|
100 |
-
"rewards/final_correctness_reward_func": 0.
|
101 |
-
"rewards/question_recreation_reward_func": 0.
|
102 |
"rewards/soft_format_reward_func": 0.0,
|
103 |
"rewards/strict_format_reward_func": 0.015625,
|
104 |
-
"rewards/xmlcount_reward_func": 0.
|
105 |
"step": 10
|
106 |
},
|
107 |
{
|
108 |
-
"completion_length":
|
109 |
-
"epoch": 0
|
110 |
-
"grad_norm":
|
111 |
-
"kl": 0.
|
112 |
"learning_rate": 4.916552125781528e-07,
|
113 |
"loss": 0.0,
|
114 |
-
"reward":
|
115 |
-
"reward_std":
|
116 |
-
"rewards/concensus_correctness_reward_func":
|
117 |
-
"rewards/consensus_reward_func": 0.
|
118 |
"rewards/cumulative_reward_2": 0.0,
|
119 |
-
"rewards/final_correctness_reward_func": 0.
|
120 |
-
"rewards/question_recreation_reward_func": 0.
|
121 |
"rewards/soft_format_reward_func": 0.0,
|
122 |
-
"rewards/strict_format_reward_func": 0.
|
123 |
-
"rewards/xmlcount_reward_func": 0.
|
124 |
"step": 12
|
125 |
},
|
126 |
{
|
127 |
-
"completion_length":
|
128 |
-
"epoch":
|
129 |
-
"grad_norm": 2.
|
130 |
-
"kl": 0.
|
131 |
"learning_rate": 4.870022949890676e-07,
|
132 |
"loss": 0.0,
|
133 |
-
"reward":
|
134 |
-
"reward_std":
|
135 |
-
"rewards/concensus_correctness_reward_func":
|
136 |
-
"rewards/consensus_reward_func": 0.
|
137 |
"rewards/cumulative_reward_2": 0.0,
|
138 |
-
"rewards/final_correctness_reward_func": 0.
|
139 |
-
"rewards/question_recreation_reward_func": 0.
|
140 |
"rewards/soft_format_reward_func": 0.0,
|
141 |
-
"rewards/strict_format_reward_func": 0.
|
142 |
-
"rewards/xmlcount_reward_func": 0.
|
143 |
"step": 14
|
144 |
},
|
145 |
{
|
146 |
-
"completion_length":
|
147 |
-
"epoch":
|
148 |
-
"grad_norm": 3.
|
149 |
-
"kl": 0.
|
150 |
"learning_rate": 4.81355307410676e-07,
|
151 |
"loss": 0.0,
|
152 |
-
"reward":
|
153 |
-
"reward_std":
|
154 |
-
"rewards/concensus_correctness_reward_func": 0.
|
155 |
-
"rewards/consensus_reward_func": 0.
|
156 |
"rewards/cumulative_reward_2": 0.0,
|
157 |
-
"rewards/final_correctness_reward_func": 0.
|
158 |
-
"rewards/question_recreation_reward_func": 0.
|
159 |
"rewards/soft_format_reward_func": 0.0,
|
160 |
-
"rewards/strict_format_reward_func": 0.
|
161 |
-
"rewards/xmlcount_reward_func": 0.
|
162 |
"step": 16
|
163 |
},
|
164 |
{
|
165 |
-
"completion_length":
|
166 |
-
"epoch":
|
167 |
-
"grad_norm":
|
168 |
-
"kl": 0.
|
169 |
"learning_rate": 4.747379352713488e-07,
|
170 |
"loss": 0.0,
|
171 |
-
"reward":
|
172 |
-
"reward_std": 1.
|
173 |
-
"rewards/concensus_correctness_reward_func": 0.
|
174 |
-
"rewards/consensus_reward_func": 0.
|
175 |
"rewards/cumulative_reward_2": 0.0,
|
176 |
-
"rewards/final_correctness_reward_func": 0.
|
177 |
-
"rewards/question_recreation_reward_func": 0.
|
178 |
"rewards/soft_format_reward_func": 0.0,
|
179 |
-
"rewards/strict_format_reward_func": 0.
|
180 |
-
"rewards/xmlcount_reward_func": 0.
|
181 |
"step": 18
|
182 |
},
|
183 |
{
|
184 |
-
"completion_length":
|
185 |
-
"epoch":
|
186 |
-
"grad_norm": 2.
|
187 |
-
"kl": 0.
|
188 |
"learning_rate": 4.6717793412953776e-07,
|
189 |
"loss": 0.0,
|
190 |
-
"reward":
|
191 |
-
"reward_std":
|
192 |
-
"rewards/concensus_correctness_reward_func":
|
193 |
-
"rewards/consensus_reward_func": 0.
|
194 |
"rewards/cumulative_reward_2": 0.0,
|
195 |
-
"rewards/final_correctness_reward_func": 0.
|
196 |
-
"rewards/question_recreation_reward_func": 0.
|
197 |
"rewards/soft_format_reward_func": 0.0,
|
198 |
-
"rewards/strict_format_reward_func": 0.
|
199 |
-
"rewards/xmlcount_reward_func": 0.
|
200 |
"step": 20
|
201 |
},
|
202 |
{
|
203 |
-
"completion_length":
|
204 |
-
"epoch":
|
205 |
-
"grad_norm":
|
206 |
-
"kl": 0.
|
207 |
"learning_rate": 4.5870701325731773e-07,
|
208 |
"loss": 0.0,
|
209 |
-
"reward":
|
210 |
-
"reward_std":
|
211 |
-
"rewards/concensus_correctness_reward_func":
|
212 |
-
"rewards/consensus_reward_func": 0.
|
213 |
"rewards/cumulative_reward_2": 0.0,
|
214 |
-
"rewards/final_correctness_reward_func": 0.
|
215 |
-
"rewards/question_recreation_reward_func": 0.
|
216 |
"rewards/soft_format_reward_func": 0.0,
|
217 |
-
"rewards/strict_format_reward_func": 0.
|
218 |
-
"rewards/xmlcount_reward_func": 0.
|
219 |
"step": 22
|
220 |
},
|
221 |
{
|
222 |
-
"completion_length":
|
223 |
-
"epoch":
|
224 |
-
"grad_norm": 2.
|
225 |
-
"kl": 0.
|
226 |
"learning_rate": 4.4936070264068016e-07,
|
227 |
"loss": 0.0,
|
228 |
-
"reward": 1.
|
229 |
-
"reward_std": 0.
|
230 |
-
"rewards/concensus_correctness_reward_func": 0.
|
231 |
-
"rewards/consensus_reward_func": 0.
|
232 |
"rewards/cumulative_reward_2": 0.0,
|
233 |
-
"rewards/final_correctness_reward_func": 0.
|
234 |
-
"rewards/question_recreation_reward_func": 0.
|
235 |
"rewards/soft_format_reward_func": 0.0,
|
236 |
-
"rewards/strict_format_reward_func": 0.
|
237 |
-
"rewards/xmlcount_reward_func": 0.
|
238 |
"step": 24
|
239 |
},
|
240 |
{
|
241 |
-
"completion_length":
|
242 |
-
"epoch":
|
243 |
-
"grad_norm":
|
244 |
-
"kl": 0.
|
245 |
"learning_rate": 4.391782039544238e-07,
|
246 |
"loss": 0.0,
|
247 |
-
"reward":
|
248 |
-
"reward_std":
|
249 |
-
"rewards/concensus_correctness_reward_func": 0.
|
250 |
-
"rewards/consensus_reward_func": 0.
|
251 |
"rewards/cumulative_reward_2": 0.0,
|
252 |
-
"rewards/final_correctness_reward_func": 0.
|
253 |
-
"rewards/question_recreation_reward_func": 0.
|
254 |
"rewards/soft_format_reward_func": 0.0,
|
255 |
-
"rewards/strict_format_reward_func": 0.
|
256 |
-
"rewards/xmlcount_reward_func": 0.
|
257 |
"step": 26
|
258 |
},
|
259 |
{
|
260 |
-
"completion_length":
|
261 |
-
"epoch":
|
262 |
-
"grad_norm":
|
263 |
-
"kl": 0.
|
264 |
"learning_rate": 4.282022261367073e-07,
|
265 |
"loss": 0.0,
|
266 |
-
"reward": 1.
|
267 |
-
"reward_std": 0.
|
268 |
-
"rewards/concensus_correctness_reward_func": 0.
|
269 |
-
"rewards/consensus_reward_func": 0.
|
270 |
"rewards/cumulative_reward_2": 0.0,
|
271 |
-
"rewards/final_correctness_reward_func": 0.
|
272 |
-
"rewards/question_recreation_reward_func": 0.
|
273 |
-
"rewards/soft_format_reward_func": 0.
|
274 |
-
"rewards/strict_format_reward_func": 0.
|
275 |
-
"rewards/xmlcount_reward_func": 0.
|
276 |
"step": 28
|
277 |
},
|
278 |
{
|
279 |
-
"completion_length":
|
280 |
-
"epoch":
|
281 |
-
"grad_norm":
|
282 |
-
"kl": 0.
|
283 |
"learning_rate": 4.1647880625292027e-07,
|
284 |
"loss": 0.0,
|
285 |
-
"reward":
|
286 |
-
"reward_std":
|
287 |
-
"rewards/concensus_correctness_reward_func":
|
288 |
-
"rewards/consensus_reward_func": 0.
|
289 |
"rewards/cumulative_reward_2": 0.0,
|
290 |
-
"rewards/final_correctness_reward_func": 0.
|
291 |
-
"rewards/question_recreation_reward_func": 0.
|
292 |
"rewards/soft_format_reward_func": 0.0,
|
293 |
-
"rewards/strict_format_reward_func": 0.
|
294 |
-
"rewards/xmlcount_reward_func": 0.
|
295 |
"step": 30
|
296 |
},
|
297 |
{
|
298 |
-
"completion_length":
|
299 |
-
"epoch":
|
300 |
-
"grad_norm":
|
301 |
-
"kl": 0.
|
302 |
"learning_rate": 4.040571164002318e-07,
|
303 |
"loss": 0.0,
|
304 |
-
"reward":
|
305 |
-
"reward_std": 1.
|
306 |
-
"rewards/concensus_correctness_reward_func": 0.
|
307 |
-
"rewards/consensus_reward_func": 0.
|
308 |
"rewards/cumulative_reward_2": 0.0,
|
309 |
-
"rewards/final_correctness_reward_func": 0.
|
310 |
-
"rewards/question_recreation_reward_func": 0.
|
311 |
"rewards/soft_format_reward_func": 0.0,
|
312 |
-
"rewards/strict_format_reward_func": 0.
|
313 |
-
"rewards/xmlcount_reward_func": 0.
|
314 |
"step": 32
|
315 |
},
|
316 |
{
|
317 |
-
"completion_length":
|
318 |
-
"epoch":
|
319 |
-
"grad_norm": 2.
|
320 |
-
"kl": 0.
|
321 |
"learning_rate": 3.909892574627266e-07,
|
322 |
"loss": 0.0,
|
323 |
-
"reward": 2.
|
324 |
-
"reward_std":
|
325 |
-
"rewards/concensus_correctness_reward_func":
|
326 |
-
"rewards/consensus_reward_func": 0.
|
327 |
"rewards/cumulative_reward_2": 0.0,
|
328 |
-
"rewards/final_correctness_reward_func": 0.
|
329 |
-
"rewards/question_recreation_reward_func": 0.
|
330 |
"rewards/soft_format_reward_func": 0.0,
|
331 |
-
"rewards/strict_format_reward_func": 0.
|
332 |
-
"rewards/xmlcount_reward_func": 0.
|
333 |
"step": 34
|
334 |
},
|
335 |
{
|
336 |
-
"completion_length":
|
337 |
-
"epoch":
|
338 |
-
"grad_norm":
|
339 |
-
"kl": 0.
|
340 |
"learning_rate": 3.773300405821908e-07,
|
341 |
"loss": 0.0,
|
342 |
-
"reward":
|
343 |
-
"reward_std":
|
344 |
-
"rewards/concensus_correctness_reward_func":
|
345 |
-
"rewards/consensus_reward_func": 0.
|
346 |
"rewards/cumulative_reward_2": 0.0,
|
347 |
-
"rewards/final_correctness_reward_func": 0.
|
348 |
-
"rewards/question_recreation_reward_func": 0.
|
349 |
"rewards/soft_format_reward_func": 0.0,
|
350 |
-
"rewards/strict_format_reward_func": 0.
|
351 |
-
"rewards/xmlcount_reward_func": 0.
|
352 |
"step": 36
|
353 |
},
|
354 |
{
|
355 |
-
"completion_length":
|
356 |
-
"epoch":
|
357 |
-
"grad_norm":
|
358 |
-
"kl": 0.
|
359 |
"learning_rate": 3.6313675726113475e-07,
|
360 |
"loss": 0.0,
|
361 |
-
"reward": 2.
|
362 |
-
"reward_std": 1.
|
363 |
-
"rewards/concensus_correctness_reward_func": 0.
|
364 |
-
"rewards/consensus_reward_func": 0.
|
365 |
"rewards/cumulative_reward_2": 0.0,
|
366 |
-
"rewards/final_correctness_reward_func": 0.
|
367 |
-
"rewards/question_recreation_reward_func": 0.
|
368 |
"rewards/soft_format_reward_func": 0.0,
|
369 |
-
"rewards/strict_format_reward_func": 0.
|
370 |
-
"rewards/xmlcount_reward_func": 0.
|
371 |
"step": 38
|
372 |
},
|
373 |
{
|
374 |
-
"completion_length":
|
375 |
-
"epoch":
|
376 |
-
"grad_norm":
|
377 |
-
"kl": 0.
|
378 |
"learning_rate": 3.484689390623218e-07,
|
379 |
"loss": 0.0,
|
380 |
-
"reward": 2.
|
381 |
-
"reward_std":
|
382 |
-
"rewards/concensus_correctness_reward_func":
|
383 |
-
"rewards/consensus_reward_func": 0.
|
384 |
"rewards/cumulative_reward_2": 0.0,
|
385 |
-
"rewards/final_correctness_reward_func": 0.
|
386 |
-
"rewards/question_recreation_reward_func": 0.
|
387 |
"rewards/soft_format_reward_func": 0.0,
|
388 |
-
"rewards/strict_format_reward_func": 0.
|
389 |
-
"rewards/xmlcount_reward_func": 0.
|
390 |
"step": 40
|
391 |
},
|
392 |
{
|
393 |
-
"completion_length":
|
394 |
-
"epoch":
|
395 |
-
"grad_norm":
|
396 |
-
"kl": 0.
|
397 |
"learning_rate": 3.3338810791270517e-07,
|
398 |
"loss": 0.0,
|
399 |
-
"reward":
|
400 |
-
"reward_std": 0.
|
401 |
-
"rewards/concensus_correctness_reward_func":
|
402 |
-
"rewards/consensus_reward_func": 0.
|
403 |
"rewards/cumulative_reward_2": 0.0,
|
404 |
-
"rewards/final_correctness_reward_func": 0.
|
405 |
-
"rewards/question_recreation_reward_func": 0.
|
406 |
"rewards/soft_format_reward_func": 0.0,
|
407 |
-
"rewards/strict_format_reward_func": 0.
|
408 |
-
"rewards/xmlcount_reward_func": 0.
|
409 |
"step": 42
|
410 |
},
|
411 |
{
|
412 |
-
"completion_length":
|
413 |
-
"epoch":
|
414 |
-
"grad_norm":
|
415 |
-
"kl": 0.
|
416 |
"learning_rate": 3.179575180590857e-07,
|
417 |
"loss": 0.0,
|
418 |
-
"reward":
|
419 |
-
"reward_std":
|
420 |
-
"rewards/concensus_correctness_reward_func":
|
421 |
-
"rewards/consensus_reward_func": 0.
|
422 |
"rewards/cumulative_reward_2": 0.0,
|
423 |
-
"rewards/final_correctness_reward_func": 0.
|
424 |
-
"rewards/question_recreation_reward_func": 0.
|
425 |
"rewards/soft_format_reward_func": 0.0,
|
426 |
-
"rewards/strict_format_reward_func": 0.
|
427 |
-
"rewards/xmlcount_reward_func": 0.
|
428 |
"step": 44
|
429 |
},
|
430 |
{
|
431 |
-
"completion_length":
|
432 |
-
"epoch":
|
433 |
-
"grad_norm":
|
434 |
-
"kl": 0.
|
435 |
"learning_rate": 3.022418907578188e-07,
|
436 |
"loss": 0.0,
|
437 |
-
"reward": 2.
|
438 |
-
"reward_std":
|
439 |
-
"rewards/concensus_correctness_reward_func":
|
440 |
-
"rewards/consensus_reward_func": 0.
|
441 |
"rewards/cumulative_reward_2": 0.0,
|
442 |
-
"rewards/final_correctness_reward_func": 0.
|
443 |
-
"rewards/question_recreation_reward_func": 0.
|
444 |
"rewards/soft_format_reward_func": 0.0,
|
445 |
-
"rewards/strict_format_reward_func": 0.
|
446 |
-
"rewards/xmlcount_reward_func": 0.
|
447 |
"step": 46
|
448 |
},
|
449 |
{
|
450 |
-
"completion_length":
|
451 |
-
"epoch":
|
452 |
-
"grad_norm":
|
453 |
-
"kl": 0.
|
454 |
"learning_rate": 2.863071428113726e-07,
|
455 |
"loss": 0.0,
|
456 |
-
"reward":
|
457 |
-
"reward_std": 1.
|
458 |
-
"rewards/concensus_correctness_reward_func":
|
459 |
-
"rewards/consensus_reward_func": 0.
|
460 |
"rewards/cumulative_reward_2": 0.0,
|
461 |
-
"rewards/final_correctness_reward_func": 0.
|
462 |
-
"rewards/question_recreation_reward_func": 0.
|
463 |
"rewards/soft_format_reward_func": 0.0,
|
464 |
-
"rewards/strict_format_reward_func": 0.
|
465 |
-
"rewards/xmlcount_reward_func": 0.
|
466 |
"step": 48
|
467 |
},
|
468 |
{
|
469 |
-
"completion_length":
|
470 |
-
"epoch":
|
471 |
-
"grad_norm":
|
472 |
-
"kl": 0.
|
473 |
"learning_rate": 2.7022011009035107e-07,
|
474 |
"loss": 0.0,
|
475 |
-
"reward":
|
476 |
-
"reward_std": 1.
|
477 |
-
"rewards/concensus_correctness_reward_func": 0.
|
478 |
-
"rewards/consensus_reward_func": 0.
|
479 |
"rewards/cumulative_reward_2": 0.0,
|
480 |
-
"rewards/final_correctness_reward_func": 0.
|
481 |
-
"rewards/question_recreation_reward_func": 0.
|
482 |
"rewards/soft_format_reward_func": 0.0,
|
483 |
-
"rewards/strict_format_reward_func": 0.
|
484 |
-
"rewards/xmlcount_reward_func": 0.
|
485 |
"step": 50
|
486 |
},
|
487 |
{
|
488 |
-
"completion_length":
|
489 |
-
"epoch":
|
490 |
-
"grad_norm": 3.
|
491 |
-
"kl": 0.
|
492 |
"learning_rate": 2.540482672006254e-07,
|
493 |
"loss": 0.0,
|
494 |
-
"reward":
|
495 |
-
"reward_std":
|
496 |
-
"rewards/concensus_correctness_reward_func":
|
497 |
-
"rewards/consensus_reward_func": 0.
|
498 |
"rewards/cumulative_reward_2": 0.0,
|
499 |
-
"rewards/final_correctness_reward_func": 0.
|
500 |
-
"rewards/question_recreation_reward_func": 0.
|
501 |
"rewards/soft_format_reward_func": 0.0,
|
502 |
-
"rewards/strict_format_reward_func": 0.
|
503 |
-
"rewards/xmlcount_reward_func": 0.
|
504 |
"step": 52
|
505 |
},
|
506 |
{
|
507 |
-
"completion_length": 257.
|
508 |
-
"epoch":
|
509 |
-
"grad_norm":
|
510 |
-
"kl": 0.
|
511 |
"learning_rate": 2.37859444471388e-07,
|
512 |
"loss": 0.0,
|
513 |
-
"reward":
|
514 |
-
"reward_std":
|
515 |
-
"rewards/concensus_correctness_reward_func": 0.
|
516 |
-
"rewards/consensus_reward_func": 0.
|
517 |
"rewards/cumulative_reward_2": 0.0,
|
518 |
-
"rewards/final_correctness_reward_func": 0.
|
519 |
-
"rewards/question_recreation_reward_func": 0.
|
520 |
"rewards/soft_format_reward_func": 0.0,
|
521 |
-
"rewards/strict_format_reward_func": 0.
|
522 |
-
"rewards/xmlcount_reward_func": 0.
|
523 |
"step": 54
|
524 |
},
|
525 |
{
|
526 |
-
"completion_length":
|
527 |
-
"epoch":
|
528 |
-
"grad_norm":
|
529 |
-
"kl": 0.
|
530 |
"learning_rate": 2.2172154345117894e-07,
|
531 |
"loss": 0.0,
|
532 |
-
"reward":
|
533 |
-
"reward_std": 1.
|
534 |
-
"rewards/concensus_correctness_reward_func":
|
535 |
-
"rewards/consensus_reward_func": 0.
|
536 |
"rewards/cumulative_reward_2": 0.0,
|
537 |
-
"rewards/final_correctness_reward_func": 0.
|
538 |
-
"rewards/question_recreation_reward_func": 0.
|
539 |
"rewards/soft_format_reward_func": 0.0,
|
540 |
-
"rewards/strict_format_reward_func": 0.
|
541 |
-
"rewards/xmlcount_reward_func": 0.
|
542 |
"step": 56
|
543 |
},
|
544 |
{
|
545 |
-
"completion_length":
|
546 |
-
"epoch":
|
547 |
-
"grad_norm": 2.
|
548 |
-
"kl": 0.
|
549 |
"learning_rate": 2.0570225210519433e-07,
|
550 |
"loss": 0.0,
|
551 |
-
"reward":
|
552 |
-
"reward_std": 1.
|
553 |
-
"rewards/concensus_correctness_reward_func": 0.
|
554 |
-
"rewards/consensus_reward_func": 0.
|
555 |
"rewards/cumulative_reward_2": 0.0,
|
556 |
-
"rewards/final_correctness_reward_func": 0.
|
557 |
-
"rewards/question_recreation_reward_func": 0.
|
558 |
"rewards/soft_format_reward_func": 0.0,
|
559 |
"rewards/strict_format_reward_func": 0.015625,
|
560 |
-
"rewards/xmlcount_reward_func": 0.
|
561 |
"step": 58
|
562 |
},
|
563 |
{
|
564 |
-
"completion_length":
|
565 |
-
"epoch":
|
566 |
-
"grad_norm":
|
567 |
-
"kl": 0.
|
568 |
"learning_rate": 1.8986876090843664e-07,
|
569 |
"loss": 0.0,
|
570 |
-
"reward":
|
571 |
-
"reward_std": 1.
|
572 |
-
"rewards/concensus_correctness_reward_func":
|
573 |
-
"rewards/consensus_reward_func": 0.
|
574 |
"rewards/cumulative_reward_2": 0.0,
|
575 |
-
"rewards/final_correctness_reward_func": 0.
|
576 |
-
"rewards/question_recreation_reward_func": 0.
|
577 |
"rewards/soft_format_reward_func": 0.0,
|
578 |
-
"rewards/strict_format_reward_func": 0.
|
579 |
-
"rewards/xmlcount_reward_func": 0.
|
580 |
"step": 60
|
581 |
},
|
582 |
{
|
583 |
-
"completion_length":
|
584 |
-
"epoch":
|
585 |
-
"grad_norm":
|
586 |
-
"kl": 0.
|
587 |
"learning_rate": 1.7428748102551234e-07,
|
588 |
"loss": 0.0,
|
589 |
-
"reward":
|
590 |
-
"reward_std":
|
591 |
-
"rewards/concensus_correctness_reward_func": 0.
|
592 |
-
"rewards/consensus_reward_func": 0.
|
593 |
"rewards/cumulative_reward_2": 0.0,
|
594 |
-
"rewards/final_correctness_reward_func": 0.
|
595 |
-
"rewards/question_recreation_reward_func": 0.
|
596 |
"rewards/soft_format_reward_func": 0.0,
|
597 |
"rewards/strict_format_reward_func": 0.046875,
|
598 |
-
"rewards/xmlcount_reward_func": 0.
|
599 |
"step": 62
|
600 |
},
|
601 |
{
|
602 |
-
"completion_length":
|
603 |
-
"epoch":
|
604 |
-
"grad_norm":
|
605 |
-
"kl": 0.
|
606 |
"learning_rate": 1.5902376575912814e-07,
|
607 |
"loss": 0.0,
|
608 |
-
"reward":
|
609 |
-
"reward_std":
|
610 |
-
"rewards/concensus_correctness_reward_func":
|
611 |
-
"rewards/consensus_reward_func": 0.
|
612 |
"rewards/cumulative_reward_2": 0.0,
|
613 |
-
"rewards/final_correctness_reward_func": 0.
|
614 |
-
"rewards/question_recreation_reward_func": 0.
|
615 |
"rewards/soft_format_reward_func": 0.0,
|
616 |
-
"rewards/strict_format_reward_func": 0.
|
617 |
-
"rewards/xmlcount_reward_func": 0.
|
618 |
"step": 64
|
619 |
},
|
620 |
{
|
621 |
-
"completion_length":
|
622 |
-
"epoch":
|
623 |
-
"grad_norm":
|
624 |
-
"kl": 0.
|
625 |
"learning_rate": 1.4414163643562753e-07,
|
626 |
"loss": 0.0,
|
627 |
-
"reward":
|
628 |
-
"reward_std": 1.
|
629 |
-
"rewards/concensus_correctness_reward_func": 0.
|
630 |
-
"rewards/consensus_reward_func": 0.
|
631 |
"rewards/cumulative_reward_2": 0.0,
|
632 |
-
"rewards/final_correctness_reward_func": 0.
|
633 |
-
"rewards/question_recreation_reward_func": 0.
|
634 |
-
"rewards/soft_format_reward_func": 0.
|
635 |
-
"rewards/strict_format_reward_func": 0.
|
636 |
-
"rewards/xmlcount_reward_func": 0.
|
637 |
"step": 66
|
638 |
},
|
639 |
{
|
640 |
-
"completion_length":
|
641 |
-
"epoch":
|
642 |
-
"grad_norm":
|
643 |
-
"kl": 0.
|
644 |
"learning_rate": 1.2970351387729872e-07,
|
645 |
"loss": 0.0,
|
646 |
-
"reward":
|
647 |
-
"reward_std":
|
648 |
-
"rewards/concensus_correctness_reward_func":
|
649 |
-
"rewards/consensus_reward_func": 0.
|
650 |
"rewards/cumulative_reward_2": 0.0,
|
651 |
-
"rewards/final_correctness_reward_func": 0.
|
652 |
-
"rewards/question_recreation_reward_func": 0.
|
653 |
"rewards/soft_format_reward_func": 0.0,
|
654 |
-
"rewards/strict_format_reward_func": 0.
|
655 |
-
"rewards/xmlcount_reward_func": 0.
|
656 |
"step": 68
|
657 |
},
|
658 |
{
|
659 |
-
"completion_length":
|
660 |
-
"epoch":
|
661 |
-
"grad_norm":
|
662 |
-
"kl": 0.
|
663 |
"learning_rate": 1.1576995658775404e-07,
|
664 |
"loss": 0.0,
|
665 |
-
"reward":
|
666 |
-
"reward_std":
|
667 |
-
"rewards/concensus_correctness_reward_func": 0.
|
668 |
-
"rewards/consensus_reward_func": 0.
|
669 |
"rewards/cumulative_reward_2": 0.0,
|
670 |
-
"rewards/final_correctness_reward_func": 0.
|
671 |
-
"rewards/question_recreation_reward_func": 0.
|
672 |
-
"rewards/soft_format_reward_func": 0.
|
673 |
-
"rewards/strict_format_reward_func": 0.
|
674 |
-
"rewards/xmlcount_reward_func": 0.
|
675 |
"step": 70
|
676 |
},
|
677 |
{
|
678 |
-
"completion_length":
|
679 |
-
"epoch":
|
680 |
-
"grad_norm":
|
681 |
-
"kl": 0.
|
682 |
"learning_rate": 1.0239940674851941e-07,
|
683 |
"loss": 0.0,
|
684 |
-
"reward": 2.
|
685 |
-
"reward_std":
|
686 |
-
"rewards/concensus_correctness_reward_func": 0.
|
687 |
-
"rewards/consensus_reward_func": 0.
|
688 |
"rewards/cumulative_reward_2": 0.0,
|
689 |
-
"rewards/final_correctness_reward_func": 0.
|
690 |
-
"rewards/question_recreation_reward_func": 0.
|
691 |
"rewards/soft_format_reward_func": 0.0,
|
692 |
-
"rewards/strict_format_reward_func": 0.
|
693 |
-
"rewards/xmlcount_reward_func": 0.
|
694 |
"step": 72
|
695 |
},
|
696 |
{
|
697 |
-
"completion_length":
|
698 |
-
"epoch":
|
699 |
-
"grad_norm": 2.
|
700 |
-
"kl": 0.
|
701 |
"learning_rate": 8.964794509221507e-08,
|
702 |
"loss": 0.0,
|
703 |
-
"reward":
|
704 |
-
"reward_std":
|
705 |
-
"rewards/concensus_correctness_reward_func":
|
706 |
-
"rewards/consensus_reward_func": 0.
|
707 |
"rewards/cumulative_reward_2": 0.0,
|
708 |
-
"rewards/final_correctness_reward_func": 0.
|
709 |
-
"rewards/question_recreation_reward_func": 0.
|
710 |
"rewards/soft_format_reward_func": 0.0,
|
711 |
-
"rewards/strict_format_reward_func": 0.
|
712 |
-
"rewards/xmlcount_reward_func": 0.
|
713 |
"step": 74
|
714 |
},
|
715 |
{
|
716 |
-
"completion_length":
|
717 |
-
"epoch":
|
718 |
-
"grad_norm": 2.
|
719 |
-
"kl": 0.
|
720 |
"learning_rate": 7.756905568047392e-08,
|
721 |
"loss": 0.0,
|
722 |
-
"reward":
|
723 |
-
"reward_std":
|
724 |
-
"rewards/concensus_correctness_reward_func":
|
725 |
"rewards/consensus_reward_func": 0.5,
|
726 |
"rewards/cumulative_reward_2": 0.0,
|
727 |
-
"rewards/final_correctness_reward_func": 0.
|
728 |
-
"rewards/question_recreation_reward_func": 0.
|
729 |
"rewards/soft_format_reward_func": 0.0,
|
730 |
"rewards/strict_format_reward_func": 0.015625,
|
731 |
-
"rewards/xmlcount_reward_func": 0.
|
732 |
"step": 76
|
733 |
},
|
734 |
{
|
735 |
-
"completion_length":
|
736 |
-
"epoch":
|
737 |
-
"grad_norm": 2.
|
738 |
-
"kl": 0.
|
739 |
"learning_rate": 6.621340157319996e-08,
|
740 |
"loss": 0.0,
|
741 |
-
"reward":
|
742 |
-
"reward_std":
|
743 |
-
"rewards/concensus_correctness_reward_func":
|
744 |
-
"rewards/consensus_reward_func": 0.
|
745 |
"rewards/cumulative_reward_2": 0.0,
|
746 |
-
"rewards/final_correctness_reward_func": 0.
|
747 |
-
"rewards/question_recreation_reward_func": 0.
|
748 |
"rewards/soft_format_reward_func": 0.0,
|
749 |
-
"rewards/strict_format_reward_func": 0.
|
750 |
-
"rewards/xmlcount_reward_func": 0.
|
751 |
"step": 78
|
752 |
},
|
753 |
{
|
754 |
-
"completion_length":
|
755 |
-
"epoch":
|
756 |
-
"grad_norm": 2.
|
757 |
-
"kl": 0.
|
758 |
"learning_rate": 5.5628612330087724e-08,
|
759 |
"loss": 0.0,
|
760 |
-
"reward":
|
761 |
-
"reward_std": 1.
|
762 |
-
"rewards/concensus_correctness_reward_func": 0.
|
763 |
-
"rewards/consensus_reward_func":
|
764 |
"rewards/cumulative_reward_2": 0.0,
|
765 |
-
"rewards/final_correctness_reward_func": 0.
|
766 |
-
"rewards/question_recreation_reward_func": 0.
|
767 |
"rewards/soft_format_reward_func": 0.0,
|
768 |
"rewards/strict_format_reward_func": 0.0625,
|
769 |
-
"rewards/xmlcount_reward_func": 0.
|
770 |
"step": 80
|
771 |
},
|
772 |
{
|
773 |
-
"completion_length":
|
774 |
-
"epoch":
|
775 |
-
"grad_norm":
|
776 |
-
"kl": 0.
|
777 |
"learning_rate": 4.5859084235697235e-08,
|
778 |
"loss": 0.0,
|
779 |
-
"reward": 1.
|
780 |
-
"reward_std":
|
781 |
-
"rewards/concensus_correctness_reward_func": 0.
|
782 |
"rewards/consensus_reward_func": 0.375,
|
783 |
"rewards/cumulative_reward_2": 0.0,
|
784 |
"rewards/final_correctness_reward_func": 0.3125,
|
785 |
-
"rewards/question_recreation_reward_func": 0.
|
786 |
"rewards/soft_format_reward_func": 0.0,
|
787 |
-
"rewards/strict_format_reward_func": 0.
|
788 |
-
"rewards/xmlcount_reward_func": 0.
|
789 |
"step": 82
|
790 |
},
|
791 |
{
|
792 |
-
"completion_length":
|
793 |
-
"epoch":
|
794 |
-
"grad_norm":
|
795 |
-
"kl": 0.
|
796 |
"learning_rate": 3.6945794086007705e-08,
|
797 |
"loss": 0.0,
|
798 |
-
"reward":
|
799 |
-
"reward_std":
|
800 |
-
"rewards/concensus_correctness_reward_func":
|
801 |
-
"rewards/consensus_reward_func": 0.
|
802 |
"rewards/cumulative_reward_2": 0.0,
|
803 |
-
"rewards/final_correctness_reward_func": 0.
|
804 |
-
"rewards/question_recreation_reward_func": 0.
|
805 |
-
"rewards/soft_format_reward_func": 0.
|
806 |
"rewards/strict_format_reward_func": 0.015625,
|
807 |
-
"rewards/xmlcount_reward_func": 0.
|
808 |
"step": 84
|
809 |
},
|
810 |
{
|
811 |
-
"completion_length":
|
812 |
-
"epoch":
|
813 |
-
"grad_norm":
|
814 |
-
"kl": 0.
|
815 |
"learning_rate": 2.892612731749414e-08,
|
816 |
"loss": 0.0,
|
817 |
-
"reward":
|
818 |
-
"reward_std":
|
819 |
-
"rewards/concensus_correctness_reward_func":
|
820 |
-
"rewards/consensus_reward_func": 0.
|
821 |
"rewards/cumulative_reward_2": 0.0,
|
822 |
-
"rewards/final_correctness_reward_func": 0.
|
823 |
-
"rewards/question_recreation_reward_func": 0.
|
824 |
"rewards/soft_format_reward_func": 0.0,
|
825 |
-
"rewards/strict_format_reward_func": 0.
|
826 |
-
"rewards/xmlcount_reward_func": 0.
|
827 |
"step": 86
|
828 |
},
|
829 |
{
|
830 |
-
"completion_length":
|
831 |
-
"epoch":
|
832 |
-
"grad_norm": 2.
|
833 |
-
"kl": 0.
|
834 |
"learning_rate": 2.183372119961499e-08,
|
835 |
"loss": 0.0,
|
836 |
-
"reward":
|
837 |
-
"reward_std":
|
838 |
-
"rewards/concensus_correctness_reward_func": 0.
|
839 |
-
"rewards/consensus_reward_func": 0.
|
840 |
"rewards/cumulative_reward_2": 0.0,
|
841 |
-
"rewards/final_correctness_reward_func": 0.
|
842 |
-
"rewards/question_recreation_reward_func": 0.
|
843 |
-
"rewards/soft_format_reward_func": 0.
|
844 |
-
"rewards/strict_format_reward_func": 0.
|
845 |
-
"rewards/xmlcount_reward_func": 0.
|
846 |
"step": 88
|
847 |
},
|
848 |
{
|
849 |
-
"completion_length":
|
850 |
-
"epoch":
|
851 |
-
"grad_norm": 2.
|
852 |
-
"kl": 0.
|
853 |
"learning_rate": 1.5698323748414122e-08,
|
854 |
"loss": 0.0,
|
855 |
-
"reward":
|
856 |
-
"reward_std":
|
857 |
-
"rewards/concensus_correctness_reward_func":
|
858 |
-
"rewards/consensus_reward_func": 0.
|
859 |
"rewards/cumulative_reward_2": 0.0,
|
860 |
-
"rewards/final_correctness_reward_func": 0.
|
861 |
-
"rewards/question_recreation_reward_func": 0.
|
862 |
"rewards/soft_format_reward_func": 0.0,
|
863 |
-
"rewards/strict_format_reward_func": 0.
|
864 |
-
"rewards/xmlcount_reward_func": 0.
|
865 |
"step": 90
|
866 |
},
|
867 |
{
|
868 |
-
"completion_length":
|
869 |
-
"epoch":
|
870 |
-
"grad_norm":
|
871 |
-
"kl": 0.
|
872 |
"learning_rate": 1.054566895300324e-08,
|
873 |
"loss": 0.0,
|
874 |
-
"reward":
|
875 |
-
"reward_std":
|
876 |
-
"rewards/concensus_correctness_reward_func":
|
877 |
"rewards/consensus_reward_func": 0.625,
|
878 |
"rewards/cumulative_reward_2": 0.0,
|
879 |
-
"rewards/final_correctness_reward_func": 0.
|
880 |
-
"rewards/question_recreation_reward_func": 0.
|
881 |
"rewards/soft_format_reward_func": 0.0,
|
882 |
-
"rewards/strict_format_reward_func": 0.
|
883 |
-
"rewards/xmlcount_reward_func": 0.
|
884 |
"step": 92
|
885 |
},
|
886 |
{
|
887 |
-
"completion_length":
|
888 |
-
"epoch":
|
889 |
-
"grad_norm":
|
890 |
-
"kl": 0.
|
891 |
"learning_rate": 6.397368838268496e-09,
|
892 |
"loss": 0.0,
|
893 |
-
"reward":
|
894 |
-
"reward_std":
|
895 |
-
"rewards/concensus_correctness_reward_func":
|
896 |
-
"rewards/consensus_reward_func": 0.
|
897 |
"rewards/cumulative_reward_2": 0.0,
|
898 |
-
"rewards/final_correctness_reward_func": 0.
|
899 |
-
"rewards/question_recreation_reward_func": 0.
|
900 |
"rewards/soft_format_reward_func": 0.0,
|
901 |
-
"rewards/strict_format_reward_func": 0.
|
902 |
-
"rewards/xmlcount_reward_func": 0.
|
903 |
"step": 94
|
904 |
},
|
905 |
{
|
906 |
-
"completion_length":
|
907 |
-
"epoch":
|
908 |
-
"grad_norm":
|
909 |
-
"kl": 0.
|
910 |
"learning_rate": 3.2708228165273244e-09,
|
911 |
"loss": 0.0,
|
912 |
-
"reward":
|
913 |
-
"reward_std":
|
914 |
-
"rewards/concensus_correctness_reward_func":
|
915 |
-
"rewards/consensus_reward_func": 0.
|
916 |
"rewards/cumulative_reward_2": 0.0,
|
917 |
-
"rewards/final_correctness_reward_func": 0.
|
918 |
-
"rewards/question_recreation_reward_func": 0.
|
919 |
"rewards/soft_format_reward_func": 0.0,
|
920 |
-
"rewards/strict_format_reward_func": 0.
|
921 |
-
"rewards/xmlcount_reward_func": 0.
|
922 |
"step": 96
|
923 |
},
|
924 |
{
|
925 |
-
"completion_length":
|
926 |
-
"epoch":
|
927 |
-
"grad_norm":
|
928 |
-
"kl": 0.
|
929 |
"learning_rate": 1.1791447083465133e-09,
|
930 |
"loss": 0.0,
|
931 |
-
"reward":
|
932 |
-
"reward_std":
|
933 |
-
"rewards/concensus_correctness_reward_func": 0.
|
934 |
-
"rewards/consensus_reward_func": 0.
|
935 |
"rewards/cumulative_reward_2": 0.0,
|
936 |
-
"rewards/final_correctness_reward_func": 0.
|
937 |
-
"rewards/question_recreation_reward_func": 0.
|
938 |
"rewards/soft_format_reward_func": 0.0,
|
939 |
-
"rewards/strict_format_reward_func": 0.
|
940 |
-
"rewards/xmlcount_reward_func": 0.
|
941 |
"step": 98
|
942 |
},
|
943 |
{
|
944 |
-
"completion_length":
|
945 |
-
"epoch":
|
946 |
-
"grad_norm": 2.
|
947 |
-
"kl": 0.
|
948 |
"learning_rate": 1.3110773862126667e-10,
|
949 |
"loss": 0.0,
|
950 |
-
"reward":
|
951 |
-
"reward_std":
|
952 |
-
"rewards/concensus_correctness_reward_func":
|
953 |
-
"rewards/consensus_reward_func": 0.
|
954 |
"rewards/cumulative_reward_2": 0.0,
|
955 |
-
"rewards/final_correctness_reward_func": 0.
|
956 |
-
"rewards/question_recreation_reward_func": 0.
|
957 |
"rewards/soft_format_reward_func": 0.0,
|
958 |
-
"rewards/strict_format_reward_func": 0.
|
959 |
-
"rewards/xmlcount_reward_func": 0.
|
960 |
"step": 100
|
961 |
},
|
962 |
{
|
963 |
-
"epoch":
|
964 |
"step": 100,
|
965 |
"total_flos": 0.0,
|
966 |
-
"train_loss": 1.
|
967 |
-
"train_runtime":
|
968 |
-
"train_samples_per_second": 0.
|
969 |
-
"train_steps_per_second": 0.
|
970 |
}
|
971 |
],
|
972 |
"logging_steps": 2,
|
973 |
"max_steps": 100,
|
974 |
"num_input_tokens_seen": 0,
|
975 |
-
"num_train_epochs":
|
976 |
"save_steps": 25,
|
977 |
"stateful_callbacks": {
|
978 |
"TrainerControl": {
|
|
|
2 |
"best_global_step": null,
|
3 |
"best_metric": null,
|
4 |
"best_model_checkpoint": null,
|
5 |
+
"epoch": 16.666666666666668,
|
6 |
"eval_steps": 500,
|
7 |
"global_step": 100,
|
8 |
"is_hyper_param_search": false,
|
|
|
10 |
"is_world_process_zero": true,
|
11 |
"log_history": [
|
12 |
{
|
13 |
+
"completion_length": 227.8125,
|
14 |
+
"epoch": 0.3333333333333333,
|
15 |
+
"grad_norm": 6.339727878570557,
|
16 |
"kl": 0.0,
|
17 |
"learning_rate": 1.6666666666666665e-07,
|
18 |
+
"loss": -0.0,
|
19 |
+
"reward": 1.5849424004554749,
|
20 |
+
"reward_std": 1.131471099331975,
|
21 |
+
"rewards/concensus_correctness_reward_func": 0.12262500077486038,
|
22 |
+
"rewards/consensus_reward_func": 0.5625,
|
23 |
"rewards/cumulative_reward_2": 0.0,
|
24 |
+
"rewards/final_correctness_reward_func": 0.0625,
|
25 |
+
"rewards/question_recreation_reward_func": 0.46509863436222076,
|
26 |
"rewards/soft_format_reward_func": 0.0,
|
27 |
+
"rewards/strict_format_reward_func": 0.015625,
|
28 |
+
"rewards/xmlcount_reward_func": 0.35659374902024865,
|
29 |
"step": 2
|
30 |
},
|
31 |
{
|
32 |
+
"completion_length": 287.125,
|
33 |
+
"epoch": 0.6666666666666666,
|
34 |
+
"grad_norm": 2.283590793609619,
|
35 |
+
"kl": 0.000743766751838848,
|
36 |
"learning_rate": 5e-07,
|
37 |
+
"loss": 0.0,
|
38 |
+
"reward": 2.0471834875643253,
|
39 |
+
"reward_std": 0.9305327897891402,
|
40 |
+
"rewards/concensus_correctness_reward_func": 0.0,
|
41 |
+
"rewards/consensus_reward_func": 0.625,
|
42 |
"rewards/cumulative_reward_2": 0.0,
|
43 |
+
"rewards/final_correctness_reward_func": 0.3125,
|
44 |
+
"rewards/question_recreation_reward_func": 0.6308709642617032,
|
45 |
"rewards/soft_format_reward_func": 0.0,
|
46 |
"rewards/strict_format_reward_func": 0.015625,
|
47 |
+
"rewards/xmlcount_reward_func": 0.46318750735372305,
|
48 |
"step": 4
|
49 |
},
|
50 |
{
|
51 |
+
"completion_length": 286.03125,
|
52 |
+
"epoch": 1.0,
|
53 |
+
"grad_norm": 2.7924423217773438,
|
54 |
+
"kl": 0.0008395667373406468,
|
55 |
"learning_rate": 4.994757065594279e-07,
|
56 |
"loss": 0.0,
|
57 |
+
"reward": 1.7317106202244759,
|
58 |
+
"reward_std": 1.3122332114144228,
|
59 |
+
"rewards/concensus_correctness_reward_func": 0.12025000154972076,
|
60 |
+
"rewards/consensus_reward_func": 0.75,
|
61 |
"rewards/cumulative_reward_2": 0.0,
|
62 |
+
"rewards/final_correctness_reward_func": 0.125,
|
63 |
+
"rewards/question_recreation_reward_func": 0.4853668725118041,
|
64 |
"rewards/soft_format_reward_func": 0.0,
|
65 |
"rewards/strict_format_reward_func": 0.03125,
|
66 |
+
"rewards/xmlcount_reward_func": 0.21984374802559614,
|
67 |
"step": 6
|
68 |
},
|
69 |
{
|
70 |
+
"completion_length": 289.90625,
|
71 |
+
"epoch": 1.3333333333333333,
|
72 |
+
"grad_norm": 2.7986364364624023,
|
73 |
+
"kl": 0.0008700692887941841,
|
74 |
"learning_rate": 4.979050253066063e-07,
|
75 |
"loss": 0.0,
|
76 |
+
"reward": 1.865300026256591,
|
77 |
+
"reward_std": 0.8366383262909949,
|
78 |
+
"rewards/concensus_correctness_reward_func": 0.06006250157952309,
|
79 |
+
"rewards/consensus_reward_func": 0.6875,
|
80 |
"rewards/cumulative_reward_2": 0.0,
|
81 |
+
"rewards/final_correctness_reward_func": 0.0625,
|
82 |
+
"rewards/question_recreation_reward_func": 0.6006125509738922,
|
83 |
"rewards/soft_format_reward_func": 0.0,
|
84 |
+
"rewards/strict_format_reward_func": 0.015625,
|
85 |
+
"rewards/xmlcount_reward_func": 0.4389999993145466,
|
86 |
"step": 8
|
87 |
},
|
88 |
{
|
89 |
+
"completion_length": 224.3125,
|
90 |
+
"epoch": 1.6666666666666665,
|
91 |
+
"grad_norm": 3.4794700145721436,
|
92 |
+
"kl": 0.0009347902014269494,
|
93 |
"learning_rate": 4.952945442245597e-07,
|
94 |
+
"loss": 0.0,
|
95 |
+
"reward": 1.8756496086716652,
|
96 |
+
"reward_std": 1.0868838177993894,
|
97 |
+
"rewards/concensus_correctness_reward_func": 0.08568750135600567,
|
98 |
+
"rewards/consensus_reward_func": 0.625,
|
99 |
"rewards/cumulative_reward_2": 0.0,
|
100 |
+
"rewards/final_correctness_reward_func": 0.1875,
|
101 |
+
"rewards/question_recreation_reward_func": 0.4599620746448636,
|
102 |
"rewards/soft_format_reward_func": 0.0,
|
103 |
"rewards/strict_format_reward_func": 0.015625,
|
104 |
+
"rewards/xmlcount_reward_func": 0.5018750056624413,
|
105 |
"step": 10
|
106 |
},
|
107 |
{
|
108 |
+
"completion_length": 296.28125,
|
109 |
+
"epoch": 2.0,
|
110 |
+
"grad_norm": 3.0765440464019775,
|
111 |
+
"kl": 0.0012358422500255983,
|
112 |
"learning_rate": 4.916552125781528e-07,
|
113 |
"loss": 0.0,
|
114 |
+
"reward": 1.9579493142664433,
|
115 |
+
"reward_std": 0.8908882063115016,
|
116 |
+
"rewards/concensus_correctness_reward_func": 0.12025000154972076,
|
117 |
+
"rewards/consensus_reward_func": 0.6875,
|
118 |
"rewards/cumulative_reward_2": 0.0,
|
119 |
+
"rewards/final_correctness_reward_func": 0.1875,
|
120 |
+
"rewards/question_recreation_reward_func": 0.465324345510453,
|
121 |
"rewards/soft_format_reward_func": 0.0,
|
122 |
+
"rewards/strict_format_reward_func": 0.015625,
|
123 |
+
"rewards/xmlcount_reward_func": 0.4817500030621886,
|
124 |
"step": 12
|
125 |
},
|
126 |
{
|
127 |
+
"completion_length": 280.90625,
|
128 |
+
"epoch": 2.3333333333333335,
|
129 |
+
"grad_norm": 2.9266109466552734,
|
130 |
+
"kl": 0.0017369047309330199,
|
131 |
"learning_rate": 4.870022949890676e-07,
|
132 |
"loss": 0.0,
|
133 |
+
"reward": 1.4219924416393042,
|
134 |
+
"reward_std": 0.8281801359262317,
|
135 |
+
"rewards/concensus_correctness_reward_func": 0.0,
|
136 |
+
"rewards/consensus_reward_func": 0.4375,
|
137 |
"rewards/cumulative_reward_2": 0.0,
|
138 |
+
"rewards/final_correctness_reward_func": 0.125,
|
139 |
+
"rewards/question_recreation_reward_func": 0.450836188509129,
|
140 |
"rewards/soft_format_reward_func": 0.0,
|
141 |
+
"rewards/strict_format_reward_func": 0.0625,
|
142 |
+
"rewards/xmlcount_reward_func": 0.34615624509751797,
|
143 |
"step": 14
|
144 |
},
|
145 |
{
|
146 |
+
"completion_length": 243.0625,
|
147 |
+
"epoch": 2.6666666666666665,
|
148 |
+
"grad_norm": 3.0913712978363037,
|
149 |
+
"kl": 0.0019520393034326844,
|
150 |
"learning_rate": 4.81355307410676e-07,
|
151 |
"loss": 0.0,
|
152 |
+
"reward": 2.2805784731172025,
|
153 |
+
"reward_std": 1.2899235817894805,
|
154 |
+
"rewards/concensus_correctness_reward_func": 0.24512499943375587,
|
155 |
+
"rewards/consensus_reward_func": 0.9375,
|
156 |
"rewards/cumulative_reward_2": 0.0,
|
157 |
+
"rewards/final_correctness_reward_func": 0.125,
|
158 |
+
"rewards/question_recreation_reward_func": 0.47473472240380943,
|
159 |
"rewards/soft_format_reward_func": 0.0,
|
160 |
+
"rewards/strict_format_reward_func": 0.046875,
|
161 |
+
"rewards/xmlcount_reward_func": 0.451343753375113,
|
162 |
"step": 16
|
163 |
},
|
164 |
{
|
165 |
+
"completion_length": 267.40625,
|
166 |
+
"epoch": 3.0,
|
167 |
+
"grad_norm": 3.0766594409942627,
|
168 |
+
"kl": 0.0020293756315368228,
|
169 |
"learning_rate": 4.747379352713488e-07,
|
170 |
"loss": 0.0,
|
171 |
+
"reward": 1.9329118076711893,
|
172 |
+
"reward_std": 1.2741229943931103,
|
173 |
+
"rewards/concensus_correctness_reward_func": 0.12256250157952309,
|
174 |
+
"rewards/consensus_reward_func": 0.6875,
|
175 |
"rewards/cumulative_reward_2": 0.0,
|
176 |
+
"rewards/final_correctness_reward_func": 0.25,
|
177 |
+
"rewards/question_recreation_reward_func": 0.619443034986034,
|
178 |
"rewards/soft_format_reward_func": 0.0,
|
179 |
+
"rewards/strict_format_reward_func": 0.015625,
|
180 |
+
"rewards/xmlcount_reward_func": 0.23778124805539846,
|
181 |
"step": 18
|
182 |
},
|
183 |
{
|
184 |
+
"completion_length": 236.09375,
|
185 |
+
"epoch": 3.3333333333333335,
|
186 |
+
"grad_norm": 2.809358596801758,
|
187 |
+
"kl": 0.0026170795026700944,
|
188 |
"learning_rate": 4.6717793412953776e-07,
|
189 |
"loss": 0.0,
|
190 |
+
"reward": 1.8976972922682762,
|
191 |
+
"reward_std": 1.0421031441655941,
|
192 |
+
"rewards/concensus_correctness_reward_func": 0.06012500077486038,
|
193 |
+
"rewards/consensus_reward_func": 0.625,
|
194 |
"rewards/cumulative_reward_2": 0.0,
|
195 |
+
"rewards/final_correctness_reward_func": 0.125,
|
196 |
+
"rewards/question_recreation_reward_func": 0.5437285574153066,
|
197 |
"rewards/soft_format_reward_func": 0.0,
|
198 |
+
"rewards/strict_format_reward_func": 0.03125,
|
199 |
+
"rewards/xmlcount_reward_func": 0.512593756429851,
|
200 |
"step": 20
|
201 |
},
|
202 |
{
|
203 |
+
"completion_length": 265.0625,
|
204 |
+
"epoch": 3.6666666666666665,
|
205 |
+
"grad_norm": 3.7295589447021484,
|
206 |
+
"kl": 0.003362223884323612,
|
207 |
"learning_rate": 4.5870701325731773e-07,
|
208 |
"loss": 0.0,
|
209 |
+
"reward": 1.9837484806776047,
|
210 |
+
"reward_std": 0.753805372864008,
|
211 |
+
"rewards/concensus_correctness_reward_func": 0.11993750184774399,
|
212 |
+
"rewards/consensus_reward_func": 0.8125,
|
213 |
"rewards/cumulative_reward_2": 0.0,
|
214 |
+
"rewards/final_correctness_reward_func": 0.125,
|
215 |
+
"rewards/question_recreation_reward_func": 0.4247172431787476,
|
216 |
"rewards/soft_format_reward_func": 0.0,
|
217 |
+
"rewards/strict_format_reward_func": 0.0,
|
218 |
+
"rewards/xmlcount_reward_func": 0.5015937560237944,
|
219 |
"step": 22
|
220 |
},
|
221 |
{
|
222 |
+
"completion_length": 237.6875,
|
223 |
+
"epoch": 4.0,
|
224 |
+
"grad_norm": 2.7614455223083496,
|
225 |
+
"kl": 0.0042717494507087395,
|
226 |
"learning_rate": 4.4936070264068016e-07,
|
227 |
"loss": 0.0,
|
228 |
+
"reward": 1.828284303745022,
|
229 |
+
"reward_std": 0.7124169690505369,
|
230 |
+
"rewards/concensus_correctness_reward_func": 0.05743750184774399,
|
231 |
+
"rewards/consensus_reward_func": 0.375,
|
232 |
"rewards/cumulative_reward_2": 0.0,
|
233 |
+
"rewards/final_correctness_reward_func": 0.125,
|
234 |
+
"rewards/question_recreation_reward_func": 0.7025655592733528,
|
235 |
"rewards/soft_format_reward_func": 0.0,
|
236 |
+
"rewards/strict_format_reward_func": 0.0625,
|
237 |
+
"rewards/xmlcount_reward_func": 0.5057812501909211,
|
238 |
"step": 24
|
239 |
},
|
240 |
{
|
241 |
+
"completion_length": 296.71875,
|
242 |
+
"epoch": 4.333333333333333,
|
243 |
+
"grad_norm": 3.3926703929901123,
|
244 |
+
"kl": 0.0044053339806851,
|
245 |
"learning_rate": 4.391782039544238e-07,
|
246 |
"loss": 0.0,
|
247 |
+
"reward": 1.3961677476763725,
|
248 |
+
"reward_std": 1.0368912005797029,
|
249 |
+
"rewards/concensus_correctness_reward_func": 0.024687500670552254,
|
250 |
+
"rewards/consensus_reward_func": 0.5625,
|
251 |
"rewards/cumulative_reward_2": 0.0,
|
252 |
+
"rewards/final_correctness_reward_func": 0.0625,
|
253 |
+
"rewards/question_recreation_reward_func": 0.46885524597018957,
|
254 |
"rewards/soft_format_reward_func": 0.0,
|
255 |
+
"rewards/strict_format_reward_func": 0.046875,
|
256 |
+
"rewards/xmlcount_reward_func": 0.230750004760921,
|
257 |
"step": 26
|
258 |
},
|
259 |
{
|
260 |
+
"completion_length": 208.375,
|
261 |
+
"epoch": 4.666666666666667,
|
262 |
+
"grad_norm": 3.8391273021698,
|
263 |
+
"kl": 0.005014390684664249,
|
264 |
"learning_rate": 4.282022261367073e-07,
|
265 |
"loss": 0.0,
|
266 |
+
"reward": 1.6386928837746382,
|
267 |
+
"reward_std": 0.5791225910652429,
|
268 |
+
"rewards/concensus_correctness_reward_func": 0.125,
|
269 |
+
"rewards/consensus_reward_func": 0.5,
|
270 |
"rewards/cumulative_reward_2": 0.0,
|
271 |
+
"rewards/final_correctness_reward_func": 0.0,
|
272 |
+
"rewards/question_recreation_reward_func": 0.5378491813316941,
|
273 |
+
"rewards/soft_format_reward_func": 0.0,
|
274 |
+
"rewards/strict_format_reward_func": 0.03125,
|
275 |
+
"rewards/xmlcount_reward_func": 0.44459374831058085,
|
276 |
"step": 28
|
277 |
},
|
278 |
{
|
279 |
+
"completion_length": 285.625,
|
280 |
+
"epoch": 5.0,
|
281 |
+
"grad_norm": 3.2744433879852295,
|
282 |
+
"kl": 0.004670602691476233,
|
283 |
"learning_rate": 4.1647880625292027e-07,
|
284 |
"loss": 0.0,
|
285 |
+
"reward": 1.5773469675332308,
|
286 |
+
"reward_std": 0.8148748113308102,
|
287 |
+
"rewards/concensus_correctness_reward_func": 0.0,
|
288 |
+
"rewards/consensus_reward_func": 0.5625,
|
289 |
"rewards/cumulative_reward_2": 0.0,
|
290 |
+
"rewards/final_correctness_reward_func": 0.0625,
|
291 |
+
"rewards/question_recreation_reward_func": 0.5454094994347543,
|
292 |
"rewards/soft_format_reward_func": 0.0,
|
293 |
+
"rewards/strict_format_reward_func": 0.03125,
|
294 |
+
"rewards/xmlcount_reward_func": 0.3756875009275973,
|
295 |
"step": 30
|
296 |
},
|
297 |
{
|
298 |
+
"completion_length": 265.375,
|
299 |
+
"epoch": 5.333333333333333,
|
300 |
+
"grad_norm": 3.7803711891174316,
|
301 |
+
"kl": 0.0071189729642355815,
|
302 |
"learning_rate": 4.040571164002318e-07,
|
303 |
"loss": 0.0,
|
304 |
+
"reward": 1.915034051053226,
|
305 |
+
"reward_std": 1.0130134378559887,
|
306 |
+
"rewards/concensus_correctness_reward_func": 0.0,
|
307 |
+
"rewards/consensus_reward_func": 0.75,
|
308 |
"rewards/cumulative_reward_2": 0.0,
|
309 |
+
"rewards/final_correctness_reward_func": 0.125,
|
310 |
+
"rewards/question_recreation_reward_func": 0.5950653096660972,
|
311 |
"rewards/soft_format_reward_func": 0.0,
|
312 |
+
"rewards/strict_format_reward_func": 0.015625,
|
313 |
+
"rewards/xmlcount_reward_func": 0.42934375285403803,
|
314 |
"step": 32
|
315 |
},
|
316 |
{
|
317 |
+
"completion_length": 257.96875,
|
318 |
+
"epoch": 5.666666666666667,
|
319 |
+
"grad_norm": 2.1448609828948975,
|
320 |
+
"kl": 0.006875298960949294,
|
321 |
"learning_rate": 3.909892574627266e-07,
|
322 |
"loss": 0.0,
|
323 |
+
"reward": 2.0118120573461056,
|
324 |
+
"reward_std": 1.1375641755294055,
|
325 |
+
"rewards/concensus_correctness_reward_func": 0.0625,
|
326 |
+
"rewards/consensus_reward_func": 0.5,
|
327 |
"rewards/cumulative_reward_2": 0.0,
|
328 |
+
"rewards/final_correctness_reward_func": 0.4375,
|
329 |
+
"rewards/question_recreation_reward_func": 0.4547183125978336,
|
330 |
"rewards/soft_format_reward_func": 0.0,
|
331 |
+
"rewards/strict_format_reward_func": 0.015625,
|
332 |
+
"rewards/xmlcount_reward_func": 0.5414687562733889,
|
333 |
"step": 34
|
334 |
},
|
335 |
{
|
336 |
+
"completion_length": 228.03125,
|
337 |
+
"epoch": 6.0,
|
338 |
+
"grad_norm": 4.327014923095703,
|
339 |
+
"kl": 0.010515401692828164,
|
340 |
"learning_rate": 3.773300405821908e-07,
|
341 |
"loss": 0.0,
|
342 |
+
"reward": 2.3596832640469074,
|
343 |
+
"reward_std": 0.8677473589777946,
|
344 |
+
"rewards/concensus_correctness_reward_func": 0.29768750071525574,
|
345 |
+
"rewards/consensus_reward_func": 0.625,
|
346 |
"rewards/cumulative_reward_2": 0.0,
|
347 |
+
"rewards/final_correctness_reward_func": 0.125,
|
348 |
+
"rewards/question_recreation_reward_func": 0.5928395111113787,
|
349 |
"rewards/soft_format_reward_func": 0.0,
|
350 |
+
"rewards/strict_format_reward_func": 0.0625,
|
351 |
+
"rewards/xmlcount_reward_func": 0.6566562494263053,
|
352 |
"step": 36
|
353 |
},
|
354 |
{
|
355 |
+
"completion_length": 255.125,
|
356 |
+
"epoch": 6.333333333333333,
|
357 |
+
"grad_norm": 3.2016897201538086,
|
358 |
+
"kl": 0.010390633338829502,
|
359 |
"learning_rate": 3.6313675726113475e-07,
|
360 |
"loss": 0.0,
|
361 |
+
"reward": 2.1920023262500763,
|
362 |
+
"reward_std": 1.4953037183731794,
|
363 |
+
"rewards/concensus_correctness_reward_func": 0.18418750166893005,
|
364 |
+
"rewards/consensus_reward_func": 0.75,
|
365 |
"rewards/cumulative_reward_2": 0.0,
|
366 |
+
"rewards/final_correctness_reward_func": 0.3125,
|
367 |
+
"rewards/question_recreation_reward_func": 0.5606585624627769,
|
368 |
"rewards/soft_format_reward_func": 0.0,
|
369 |
+
"rewards/strict_format_reward_func": 0.015625,
|
370 |
+
"rewards/xmlcount_reward_func": 0.3690312569960952,
|
371 |
"step": 38
|
372 |
},
|
373 |
{
|
374 |
+
"completion_length": 255.21875,
|
375 |
+
"epoch": 6.666666666666667,
|
376 |
+
"grad_norm": 2.821410894393921,
|
377 |
+
"kl": 0.011521625056047924,
|
378 |
"learning_rate": 3.484689390623218e-07,
|
379 |
"loss": 0.0,
|
380 |
+
"reward": 2.26532906293869,
|
381 |
+
"reward_std": 1.2260184331098571,
|
382 |
+
"rewards/concensus_correctness_reward_func": 0.30037499964237213,
|
383 |
+
"rewards/consensus_reward_func": 0.8125,
|
384 |
"rewards/cumulative_reward_2": 0.0,
|
385 |
+
"rewards/final_correctness_reward_func": 0.0625,
|
386 |
+
"rewards/question_recreation_reward_func": 0.6025790590792894,
|
387 |
"rewards/soft_format_reward_func": 0.0,
|
388 |
+
"rewards/strict_format_reward_func": 0.046875,
|
389 |
+
"rewards/xmlcount_reward_func": 0.4405000088736415,
|
390 |
"step": 40
|
391 |
},
|
392 |
{
|
393 |
+
"completion_length": 296.125,
|
394 |
+
"epoch": 7.0,
|
395 |
+
"grad_norm": 2.361708641052246,
|
396 |
+
"kl": 0.009184279042528942,
|
397 |
"learning_rate": 3.3338810791270517e-07,
|
398 |
"loss": 0.0,
|
399 |
+
"reward": 1.8683023676276207,
|
400 |
+
"reward_std": 0.931499857455492,
|
401 |
+
"rewards/concensus_correctness_reward_func": 0.11750000342726707,
|
402 |
+
"rewards/consensus_reward_func": 0.625,
|
403 |
"rewards/cumulative_reward_2": 0.0,
|
404 |
+
"rewards/final_correctness_reward_func": 0.1875,
|
405 |
+
"rewards/question_recreation_reward_func": 0.5745836496353149,
|
406 |
"rewards/soft_format_reward_func": 0.0,
|
407 |
+
"rewards/strict_format_reward_func": 0.046875,
|
408 |
+
"rewards/xmlcount_reward_func": 0.3168437508866191,
|
409 |
"step": 42
|
410 |
},
|
411 |
{
|
412 |
+
"completion_length": 264.34375,
|
413 |
+
"epoch": 7.333333333333333,
|
414 |
+
"grad_norm": 3.3224129676818848,
|
415 |
+
"kl": 0.01854561164509505,
|
416 |
"learning_rate": 3.179575180590857e-07,
|
417 |
"loss": 0.0,
|
418 |
+
"reward": 1.4564557429403067,
|
419 |
+
"reward_std": 0.8835583210457116,
|
420 |
+
"rewards/concensus_correctness_reward_func": 0.0,
|
421 |
+
"rewards/consensus_reward_func": 0.375,
|
422 |
"rewards/cumulative_reward_2": 0.0,
|
423 |
+
"rewards/final_correctness_reward_func": 0.125,
|
424 |
+
"rewards/question_recreation_reward_func": 0.5099557400681078,
|
425 |
"rewards/soft_format_reward_func": 0.0,
|
426 |
+
"rewards/strict_format_reward_func": 0.03125,
|
427 |
+
"rewards/xmlcount_reward_func": 0.41524998657405376,
|
428 |
"step": 44
|
429 |
},
|
430 |
{
|
431 |
+
"completion_length": 302.34375,
|
432 |
+
"epoch": 7.666666666666667,
|
433 |
+
"grad_norm": 3.5825905799865723,
|
434 |
+
"kl": 0.013279616308864206,
|
435 |
"learning_rate": 3.022418907578188e-07,
|
436 |
"loss": 0.0,
|
437 |
+
"reward": 2.223274264484644,
|
438 |
+
"reward_std": 0.9707561411778443,
|
439 |
+
"rewards/concensus_correctness_reward_func": 0.22681250050663948,
|
440 |
+
"rewards/consensus_reward_func": 0.875,
|
441 |
"rewards/cumulative_reward_2": 0.0,
|
442 |
+
"rewards/final_correctness_reward_func": 0.0625,
|
443 |
+
"rewards/question_recreation_reward_func": 0.5176492994651198,
|
444 |
"rewards/soft_format_reward_func": 0.0,
|
445 |
+
"rewards/strict_format_reward_func": 0.015625,
|
446 |
+
"rewards/xmlcount_reward_func": 0.5256875129416585,
|
447 |
"step": 46
|
448 |
},
|
449 |
{
|
450 |
+
"completion_length": 284.5,
|
451 |
+
"epoch": 8.0,
|
452 |
+
"grad_norm": 3.046675682067871,
|
453 |
+
"kl": 0.014580188668332994,
|
454 |
"learning_rate": 2.863071428113726e-07,
|
455 |
"loss": 0.0,
|
456 |
+
"reward": 2.079888518899679,
|
457 |
+
"reward_std": 1.115025261300616,
|
458 |
+
"rewards/concensus_correctness_reward_func": 0.125,
|
459 |
+
"rewards/consensus_reward_func": 0.6875,
|
460 |
"rewards/cumulative_reward_2": 0.0,
|
461 |
+
"rewards/final_correctness_reward_func": 0.25,
|
462 |
+
"rewards/question_recreation_reward_func": 0.6062635215930641,
|
463 |
"rewards/soft_format_reward_func": 0.0,
|
464 |
+
"rewards/strict_format_reward_func": 0.03125,
|
465 |
+
"rewards/xmlcount_reward_func": 0.37987500336021185,
|
466 |
"step": 48
|
467 |
},
|
468 |
{
|
469 |
+
"completion_length": 283.40625,
|
470 |
+
"epoch": 8.333333333333334,
|
471 |
+
"grad_norm": 4.013537883758545,
|
472 |
+
"kl": 0.01412903075106442,
|
473 |
"learning_rate": 2.7022011009035107e-07,
|
474 |
"loss": 0.0,
|
475 |
+
"reward": 1.7887509390711784,
|
476 |
+
"reward_std": 1.0781924333423376,
|
477 |
+
"rewards/concensus_correctness_reward_func": 0.06012500077486038,
|
478 |
+
"rewards/consensus_reward_func": 0.6875,
|
479 |
"rewards/cumulative_reward_2": 0.0,
|
480 |
+
"rewards/final_correctness_reward_func": 0.0625,
|
481 |
+
"rewards/question_recreation_reward_func": 0.4950321572832763,
|
482 |
"rewards/soft_format_reward_func": 0.0,
|
483 |
+
"rewards/strict_format_reward_func": 0.015625,
|
484 |
+
"rewards/xmlcount_reward_func": 0.46796874795109034,
|
485 |
"step": 50
|
486 |
},
|
487 |
{
|
488 |
+
"completion_length": 255.75,
|
489 |
+
"epoch": 8.666666666666666,
|
490 |
+
"grad_norm": 3.4840548038482666,
|
491 |
+
"kl": 0.020143478643149137,
|
492 |
"learning_rate": 2.540482672006254e-07,
|
493 |
"loss": 0.0,
|
494 |
+
"reward": 2.2514243982732296,
|
495 |
+
"reward_std": 0.9967716456158087,
|
496 |
+
"rewards/concensus_correctness_reward_func": 0.14731250144541264,
|
497 |
+
"rewards/consensus_reward_func": 0.6875,
|
498 |
"rewards/cumulative_reward_2": 0.0,
|
499 |
+
"rewards/final_correctness_reward_func": 0.0625,
|
500 |
+
"rewards/question_recreation_reward_func": 0.7038306472823024,
|
501 |
"rewards/soft_format_reward_func": 0.0,
|
502 |
+
"rewards/strict_format_reward_func": 0.09375,
|
503 |
+
"rewards/xmlcount_reward_func": 0.5565312498365529,
|
504 |
"step": 52
|
505 |
},
|
506 |
{
|
507 |
+
"completion_length": 257.25,
|
508 |
+
"epoch": 9.0,
|
509 |
+
"grad_norm": 3.285977602005005,
|
510 |
+
"kl": 0.020005275029689074,
|
511 |
"learning_rate": 2.37859444471388e-07,
|
512 |
"loss": 0.0,
|
513 |
+
"reward": 1.8586036376655102,
|
514 |
+
"reward_std": 0.8674497168976814,
|
515 |
+
"rewards/concensus_correctness_reward_func": 0.23737500235438347,
|
516 |
+
"rewards/consensus_reward_func": 0.5625,
|
517 |
"rewards/cumulative_reward_2": 0.0,
|
518 |
+
"rewards/final_correctness_reward_func": 0.0625,
|
519 |
+
"rewards/question_recreation_reward_func": 0.4530098957475275,
|
520 |
"rewards/soft_format_reward_func": 0.0,
|
521 |
+
"rewards/strict_format_reward_func": 0.0625,
|
522 |
+
"rewards/xmlcount_reward_func": 0.4807187654078007,
|
523 |
"step": 54
|
524 |
},
|
525 |
{
|
526 |
+
"completion_length": 240.28125,
|
527 |
+
"epoch": 9.333333333333334,
|
528 |
+
"grad_norm": 3.7038044929504395,
|
529 |
+
"kl": 0.018931806669570506,
|
530 |
"learning_rate": 2.2172154345117894e-07,
|
531 |
"loss": 0.0,
|
532 |
+
"reward": 1.623201709240675,
|
533 |
+
"reward_std": 1.2911038948222995,
|
534 |
+
"rewards/concensus_correctness_reward_func": 0.06012500077486038,
|
535 |
+
"rewards/consensus_reward_func": 0.625,
|
536 |
"rewards/cumulative_reward_2": 0.0,
|
537 |
+
"rewards/final_correctness_reward_func": 0.0625,
|
538 |
+
"rewards/question_recreation_reward_func": 0.48360797856003046,
|
539 |
"rewards/soft_format_reward_func": 0.0,
|
540 |
+
"rewards/strict_format_reward_func": 0.03125,
|
541 |
+
"rewards/xmlcount_reward_func": 0.3607187531888485,
|
542 |
"step": 56
|
543 |
},
|
544 |
{
|
545 |
+
"completion_length": 269.84375,
|
546 |
+
"epoch": 9.666666666666666,
|
547 |
+
"grad_norm": 2.881014585494995,
|
548 |
+
"kl": 0.017237395339179784,
|
549 |
"learning_rate": 2.0570225210519433e-07,
|
550 |
"loss": 0.0,
|
551 |
+
"reward": 1.9306830489076674,
|
552 |
+
"reward_std": 1.3888351377099752,
|
553 |
+
"rewards/concensus_correctness_reward_func": 0.24249999597668648,
|
554 |
+
"rewards/consensus_reward_func": 0.6875,
|
555 |
"rewards/cumulative_reward_2": 0.0,
|
556 |
+
"rewards/final_correctness_reward_func": 0.0,
|
557 |
+
"rewards/question_recreation_reward_func": 0.5115580353885889,
|
558 |
"rewards/soft_format_reward_func": 0.0,
|
559 |
"rewards/strict_format_reward_func": 0.015625,
|
560 |
+
"rewards/xmlcount_reward_func": 0.4735000031068921,
|
561 |
"step": 58
|
562 |
},
|
563 |
{
|
564 |
+
"completion_length": 249.15625,
|
565 |
+
"epoch": 10.0,
|
566 |
+
"grad_norm": 3.1154470443725586,
|
567 |
+
"kl": 0.02515559794846922,
|
568 |
"learning_rate": 1.8986876090843664e-07,
|
569 |
"loss": 0.0,
|
570 |
+
"reward": 2.0984230153262615,
|
571 |
+
"reward_std": 1.3092318717390299,
|
572 |
+
"rewards/concensus_correctness_reward_func": 0.06012500077486038,
|
573 |
+
"rewards/consensus_reward_func": 0.6875,
|
574 |
"rewards/cumulative_reward_2": 0.0,
|
575 |
+
"rewards/final_correctness_reward_func": 0.1875,
|
576 |
+
"rewards/question_recreation_reward_func": 0.5064230696298182,
|
577 |
"rewards/soft_format_reward_func": 0.0,
|
578 |
+
"rewards/strict_format_reward_func": 0.078125,
|
579 |
+
"rewards/xmlcount_reward_func": 0.5787499956786633,
|
580 |
"step": 60
|
581 |
},
|
582 |
{
|
583 |
+
"completion_length": 253.25,
|
584 |
+
"epoch": 10.333333333333334,
|
585 |
+
"grad_norm": 3.137545347213745,
|
586 |
+
"kl": 0.024528241163352504,
|
587 |
"learning_rate": 1.7428748102551234e-07,
|
588 |
"loss": 0.0,
|
589 |
+
"reward": 1.8578170202672482,
|
590 |
+
"reward_std": 1.1804295498877764,
|
591 |
+
"rewards/concensus_correctness_reward_func": 0.059812501072883606,
|
592 |
+
"rewards/consensus_reward_func": 0.3125,
|
593 |
"rewards/cumulative_reward_2": 0.0,
|
594 |
+
"rewards/final_correctness_reward_func": 0.25,
|
595 |
+
"rewards/question_recreation_reward_func": 0.4470045296475291,
|
596 |
"rewards/soft_format_reward_func": 0.0,
|
597 |
"rewards/strict_format_reward_func": 0.046875,
|
598 |
+
"rewards/xmlcount_reward_func": 0.7416250004898757,
|
599 |
"step": 62
|
600 |
},
|
601 |
{
|
602 |
+
"completion_length": 223.3125,
|
603 |
+
"epoch": 10.666666666666666,
|
604 |
+
"grad_norm": 3.1050543785095215,
|
605 |
+
"kl": 0.02882619173033163,
|
606 |
"learning_rate": 1.5902376575912814e-07,
|
607 |
"loss": 0.0,
|
608 |
+
"reward": 1.743748527020216,
|
609 |
+
"reward_std": 0.6933234713651473,
|
610 |
+
"rewards/concensus_correctness_reward_func": 0.20481249876320362,
|
611 |
+
"rewards/consensus_reward_func": 0.5,
|
612 |
"rewards/cumulative_reward_2": 0.0,
|
613 |
+
"rewards/final_correctness_reward_func": 0.0625,
|
614 |
+
"rewards/question_recreation_reward_func": 0.4549047634936869,
|
615 |
"rewards/soft_format_reward_func": 0.0,
|
616 |
+
"rewards/strict_format_reward_func": 0.03125,
|
617 |
+
"rewards/xmlcount_reward_func": 0.4902812475338578,
|
618 |
"step": 64
|
619 |
},
|
620 |
{
|
621 |
+
"completion_length": 237.8125,
|
622 |
+
"epoch": 11.0,
|
623 |
+
"grad_norm": 3.4488277435302734,
|
624 |
+
"kl": 0.024283534032292664,
|
625 |
"learning_rate": 1.4414163643562753e-07,
|
626 |
"loss": 0.0,
|
627 |
+
"reward": 1.9523196145892143,
|
628 |
+
"reward_std": 1.559489093720913,
|
629 |
+
"rewards/concensus_correctness_reward_func": 0.14818750135600567,
|
630 |
+
"rewards/consensus_reward_func": 0.6875,
|
631 |
"rewards/cumulative_reward_2": 0.0,
|
632 |
+
"rewards/final_correctness_reward_func": 0.0625,
|
633 |
+
"rewards/question_recreation_reward_func": 0.5911321062594652,
|
634 |
+
"rewards/soft_format_reward_func": 0.015625,
|
635 |
+
"rewards/strict_format_reward_func": 0.015625,
|
636 |
+
"rewards/xmlcount_reward_func": 0.43175000394694507,
|
637 |
"step": 66
|
638 |
},
|
639 |
{
|
640 |
+
"completion_length": 285.21875,
|
641 |
+
"epoch": 11.333333333333334,
|
642 |
+
"grad_norm": 3.1919057369232178,
|
643 |
+
"kl": 0.029086316528264433,
|
644 |
"learning_rate": 1.2970351387729872e-07,
|
645 |
"loss": 0.0,
|
646 |
+
"reward": 1.9048342034220695,
|
647 |
+
"reward_std": 1.2717823022976518,
|
648 |
+
"rewards/concensus_correctness_reward_func": 0.06012500077486038,
|
649 |
+
"rewards/consensus_reward_func": 0.75,
|
650 |
"rewards/cumulative_reward_2": 0.0,
|
651 |
+
"rewards/final_correctness_reward_func": 0.0625,
|
652 |
+
"rewards/question_recreation_reward_func": 0.5372404857771471,
|
653 |
"rewards/soft_format_reward_func": 0.0,
|
654 |
+
"rewards/strict_format_reward_func": 0.0625,
|
655 |
+
"rewards/xmlcount_reward_func": 0.4324687570333481,
|
656 |
"step": 68
|
657 |
},
|
658 |
{
|
659 |
+
"completion_length": 274.03125,
|
660 |
+
"epoch": 11.666666666666666,
|
661 |
+
"grad_norm": 4.114351749420166,
|
662 |
+
"kl": 0.028069639985915273,
|
663 |
"learning_rate": 1.1576995658775404e-07,
|
664 |
"loss": 0.0,
|
665 |
+
"reward": 1.8863378204405308,
|
666 |
+
"reward_std": 1.3141623558476567,
|
667 |
+
"rewards/concensus_correctness_reward_func": 0.12256250157952309,
|
668 |
+
"rewards/consensus_reward_func": 0.625,
|
669 |
"rewards/cumulative_reward_2": 0.0,
|
670 |
+
"rewards/final_correctness_reward_func": 0.125,
|
671 |
+
"rewards/question_recreation_reward_func": 0.5985253136605024,
|
672 |
+
"rewards/soft_format_reward_func": 0.015625,
|
673 |
+
"rewards/strict_format_reward_func": 0.078125,
|
674 |
+
"rewards/xmlcount_reward_func": 0.3215000149793923,
|
675 |
"step": 70
|
676 |
},
|
677 |
{
|
678 |
+
"completion_length": 248.75,
|
679 |
+
"epoch": 12.0,
|
680 |
+
"grad_norm": 2.4701993465423584,
|
681 |
+
"kl": 0.020335440058261156,
|
682 |
"learning_rate": 1.0239940674851941e-07,
|
683 |
"loss": 0.0,
|
684 |
+
"reward": 2.2941419184207916,
|
685 |
+
"reward_std": 0.8937435210682452,
|
686 |
+
"rewards/concensus_correctness_reward_func": 0.11487500369548798,
|
687 |
+
"rewards/consensus_reward_func": 0.5625,
|
688 |
"rewards/cumulative_reward_2": 0.0,
|
689 |
+
"rewards/final_correctness_reward_func": 0.4375,
|
690 |
+
"rewards/question_recreation_reward_func": 0.5048919152468443,
|
691 |
"rewards/soft_format_reward_func": 0.0,
|
692 |
+
"rewards/strict_format_reward_func": 0.046875,
|
693 |
+
"rewards/xmlcount_reward_func": 0.627500012749806,
|
694 |
"step": 72
|
695 |
},
|
696 |
{
|
697 |
+
"completion_length": 220.09375,
|
698 |
+
"epoch": 12.333333333333334,
|
699 |
+
"grad_norm": 2.9127869606018066,
|
700 |
+
"kl": 0.022209461370948702,
|
701 |
"learning_rate": 8.964794509221507e-08,
|
702 |
"loss": 0.0,
|
703 |
+
"reward": 2.0042398422956467,
|
704 |
+
"reward_std": 1.136198466643691,
|
705 |
+
"rewards/concensus_correctness_reward_func": 0.24262500181794167,
|
706 |
+
"rewards/consensus_reward_func": 0.5625,
|
707 |
"rewards/cumulative_reward_2": 0.0,
|
708 |
+
"rewards/final_correctness_reward_func": 0.1875,
|
709 |
+
"rewards/question_recreation_reward_func": 0.49439611518755555,
|
710 |
"rewards/soft_format_reward_func": 0.0,
|
711 |
+
"rewards/strict_format_reward_func": 0.015625,
|
712 |
+
"rewards/xmlcount_reward_func": 0.5015937518328428,
|
713 |
"step": 74
|
714 |
},
|
715 |
{
|
716 |
+
"completion_length": 273.3125,
|
717 |
+
"epoch": 12.666666666666666,
|
718 |
+
"grad_norm": 2.498856782913208,
|
719 |
+
"kl": 0.01810361386742443,
|
720 |
"learning_rate": 7.756905568047392e-08,
|
721 |
"loss": 0.0,
|
722 |
+
"reward": 1.7342982944101095,
|
723 |
+
"reward_std": 1.2671478418633342,
|
724 |
+
"rewards/concensus_correctness_reward_func": 0.05999999865889549,
|
725 |
"rewards/consensus_reward_func": 0.5,
|
726 |
"rewards/cumulative_reward_2": 0.0,
|
727 |
+
"rewards/final_correctness_reward_func": 0.1875,
|
728 |
+
"rewards/question_recreation_reward_func": 0.575642024166882,
|
729 |
"rewards/soft_format_reward_func": 0.0,
|
730 |
"rewards/strict_format_reward_func": 0.015625,
|
731 |
+
"rewards/xmlcount_reward_func": 0.39553124993108213,
|
732 |
"step": 76
|
733 |
},
|
734 |
{
|
735 |
+
"completion_length": 239.71875,
|
736 |
+
"epoch": 13.0,
|
737 |
+
"grad_norm": 2.6426241397857666,
|
738 |
+
"kl": 0.027558521192986518,
|
739 |
"learning_rate": 6.621340157319996e-08,
|
740 |
"loss": 0.0,
|
741 |
+
"reward": 2.0297871977090836,
|
742 |
+
"reward_std": 0.9375762529671192,
|
743 |
+
"rewards/concensus_correctness_reward_func": 0.06012500077486038,
|
744 |
+
"rewards/consensus_reward_func": 0.8125,
|
745 |
"rewards/cumulative_reward_2": 0.0,
|
746 |
+
"rewards/final_correctness_reward_func": 0.0,
|
747 |
+
"rewards/question_recreation_reward_func": 0.643943477421999,
|
748 |
"rewards/soft_format_reward_func": 0.0,
|
749 |
+
"rewards/strict_format_reward_func": 0.046875,
|
750 |
+
"rewards/xmlcount_reward_func": 0.4663437455892563,
|
751 |
"step": 78
|
752 |
},
|
753 |
{
|
754 |
+
"completion_length": 217.0,
|
755 |
+
"epoch": 13.333333333333334,
|
756 |
+
"grad_norm": 2.7676427364349365,
|
757 |
+
"kl": 0.03532541490858421,
|
758 |
"learning_rate": 5.5628612330087724e-08,
|
759 |
"loss": 0.0,
|
760 |
+
"reward": 2.5324832424521446,
|
761 |
+
"reward_std": 1.1609387751668692,
|
762 |
+
"rewards/concensus_correctness_reward_func": 0.1401250008493662,
|
763 |
+
"rewards/consensus_reward_func": 1.1875,
|
764 |
"rewards/cumulative_reward_2": 0.0,
|
765 |
+
"rewards/final_correctness_reward_func": 0.125,
|
766 |
+
"rewards/question_recreation_reward_func": 0.4988894378184341,
|
767 |
"rewards/soft_format_reward_func": 0.0,
|
768 |
"rewards/strict_format_reward_func": 0.0625,
|
769 |
+
"rewards/xmlcount_reward_func": 0.5184687459841371,
|
770 |
"step": 80
|
771 |
},
|
772 |
{
|
773 |
+
"completion_length": 263.84375,
|
774 |
+
"epoch": 13.666666666666666,
|
775 |
+
"grad_norm": 3.3539531230926514,
|
776 |
+
"kl": 0.02452456980245188,
|
777 |
"learning_rate": 4.5859084235697235e-08,
|
778 |
"loss": 0.0,
|
779 |
+
"reward": 1.9322984656319022,
|
780 |
+
"reward_std": 1.3757213475182652,
|
781 |
+
"rewards/concensus_correctness_reward_func": 0.12262500077486038,
|
782 |
"rewards/consensus_reward_func": 0.375,
|
783 |
"rewards/cumulative_reward_2": 0.0,
|
784 |
"rewards/final_correctness_reward_func": 0.3125,
|
785 |
+
"rewards/question_recreation_reward_func": 0.5088922204449773,
|
786 |
"rewards/soft_format_reward_func": 0.0,
|
787 |
+
"rewards/strict_format_reward_func": 0.046875,
|
788 |
+
"rewards/xmlcount_reward_func": 0.5664062486030161,
|
789 |
"step": 82
|
790 |
},
|
791 |
{
|
792 |
+
"completion_length": 258.9375,
|
793 |
+
"epoch": 14.0,
|
794 |
+
"grad_norm": 3.50706148147583,
|
795 |
+
"kl": 0.02300187383661978,
|
796 |
"learning_rate": 3.6945794086007705e-08,
|
797 |
"loss": 0.0,
|
798 |
+
"reward": 2.368156984448433,
|
799 |
+
"reward_std": 1.27269869716838,
|
800 |
+
"rewards/concensus_correctness_reward_func": 0.23487500101327896,
|
801 |
+
"rewards/consensus_reward_func": 0.9375,
|
802 |
"rewards/cumulative_reward_2": 0.0,
|
803 |
+
"rewards/final_correctness_reward_func": 0.125,
|
804 |
+
"rewards/question_recreation_reward_func": 0.6125319767743349,
|
805 |
+
"rewards/soft_format_reward_func": 0.015625,
|
806 |
"rewards/strict_format_reward_func": 0.015625,
|
807 |
+
"rewards/xmlcount_reward_func": 0.42700000666081905,
|
808 |
"step": 84
|
809 |
},
|
810 |
{
|
811 |
+
"completion_length": 238.8125,
|
812 |
+
"epoch": 14.333333333333334,
|
813 |
+
"grad_norm": 2.6380765438079834,
|
814 |
+
"kl": 0.027566006290726364,
|
815 |
"learning_rate": 2.892612731749414e-08,
|
816 |
"loss": 0.0,
|
817 |
+
"reward": 2.055258920416236,
|
818 |
+
"reward_std": 0.6858638301491737,
|
819 |
+
"rewards/concensus_correctness_reward_func": 0.18012499809265137,
|
820 |
+
"rewards/consensus_reward_func": 0.75,
|
821 |
"rewards/cumulative_reward_2": 0.0,
|
822 |
+
"rewards/final_correctness_reward_func": 0.0,
|
823 |
+
"rewards/question_recreation_reward_func": 0.5494776804698631,
|
824 |
"rewards/soft_format_reward_func": 0.0,
|
825 |
+
"rewards/strict_format_reward_func": 0.015625,
|
826 |
+
"rewards/xmlcount_reward_func": 0.5600312501192093,
|
827 |
"step": 86
|
828 |
},
|
829 |
{
|
830 |
+
"completion_length": 283.0,
|
831 |
+
"epoch": 14.666666666666666,
|
832 |
+
"grad_norm": 2.740476369857788,
|
833 |
+
"kl": 0.02960860973689705,
|
834 |
"learning_rate": 2.183372119961499e-08,
|
835 |
"loss": 0.0,
|
836 |
+
"reward": 1.8798525519669056,
|
837 |
+
"reward_std": 1.1110668628825806,
|
838 |
+
"rewards/concensus_correctness_reward_func": 0.06012500077486038,
|
839 |
+
"rewards/consensus_reward_func": 0.6875,
|
840 |
"rewards/cumulative_reward_2": 0.0,
|
841 |
+
"rewards/final_correctness_reward_func": 0.125,
|
842 |
+
"rewards/question_recreation_reward_func": 0.5739150438457727,
|
843 |
+
"rewards/soft_format_reward_func": 0.015625,
|
844 |
+
"rewards/strict_format_reward_func": 0.046875,
|
845 |
+
"rewards/xmlcount_reward_func": 0.37081251526251435,
|
846 |
"step": 88
|
847 |
},
|
848 |
{
|
849 |
+
"completion_length": 242.3125,
|
850 |
+
"epoch": 15.0,
|
851 |
+
"grad_norm": 2.513399124145508,
|
852 |
+
"kl": 0.026559468533378094,
|
853 |
"learning_rate": 1.5698323748414122e-08,
|
854 |
"loss": 0.0,
|
855 |
+
"reward": 2.03849926404655,
|
856 |
+
"reward_std": 0.964166424702853,
|
857 |
+
"rewards/concensus_correctness_reward_func": 0.14824999682605267,
|
858 |
+
"rewards/consensus_reward_func": 0.625,
|
859 |
"rewards/cumulative_reward_2": 0.0,
|
860 |
+
"rewards/final_correctness_reward_func": 0.1875,
|
861 |
+
"rewards/question_recreation_reward_func": 0.5135617647320032,
|
862 |
"rewards/soft_format_reward_func": 0.0,
|
863 |
+
"rewards/strict_format_reward_func": 0.0625,
|
864 |
+
"rewards/xmlcount_reward_func": 0.5016875043511391,
|
865 |
"step": 90
|
866 |
},
|
867 |
{
|
868 |
+
"completion_length": 205.65625,
|
869 |
+
"epoch": 15.333333333333334,
|
870 |
+
"grad_norm": 3.1790051460266113,
|
871 |
+
"kl": 0.036513498693238944,
|
872 |
"learning_rate": 1.054566895300324e-08,
|
873 |
"loss": 0.0,
|
874 |
+
"reward": 2.030654199421406,
|
875 |
+
"reward_std": 1.0498718353919685,
|
876 |
+
"rewards/concensus_correctness_reward_func": 0.08824999816715717,
|
877 |
"rewards/consensus_reward_func": 0.625,
|
878 |
"rewards/cumulative_reward_2": 0.0,
|
879 |
+
"rewards/final_correctness_reward_func": 0.0625,
|
880 |
+
"rewards/question_recreation_reward_func": 0.6112791877239943,
|
881 |
"rewards/soft_format_reward_func": 0.0,
|
882 |
+
"rewards/strict_format_reward_func": 0.046875,
|
883 |
+
"rewards/xmlcount_reward_func": 0.596749996766448,
|
884 |
"step": 92
|
885 |
},
|
886 |
{
|
887 |
+
"completion_length": 263.0,
|
888 |
+
"epoch": 15.666666666666666,
|
889 |
+
"grad_norm": 3.719067096710205,
|
890 |
+
"kl": 0.021271194782457314,
|
891 |
"learning_rate": 6.397368838268496e-09,
|
892 |
"loss": 0.0,
|
893 |
+
"reward": 2.2683150228112936,
|
894 |
+
"reward_std": 1.4738470809534192,
|
895 |
+
"rewards/concensus_correctness_reward_func": 0.36025000363588333,
|
896 |
+
"rewards/consensus_reward_func": 0.5,
|
897 |
"rewards/cumulative_reward_2": 0.0,
|
898 |
+
"rewards/final_correctness_reward_func": 0.1875,
|
899 |
+
"rewards/question_recreation_reward_func": 0.5795337841846049,
|
900 |
"rewards/soft_format_reward_func": 0.0,
|
901 |
+
"rewards/strict_format_reward_func": 0.046875,
|
902 |
+
"rewards/xmlcount_reward_func": 0.5941562601365149,
|
903 |
"step": 94
|
904 |
},
|
905 |
{
|
906 |
+
"completion_length": 257.71875,
|
907 |
+
"epoch": 16.0,
|
908 |
+
"grad_norm": 5.480099201202393,
|
909 |
+
"kl": 0.028246068861335516,
|
910 |
"learning_rate": 3.2708228165273244e-09,
|
911 |
"loss": 0.0,
|
912 |
+
"reward": 2.2440423257648945,
|
913 |
+
"reward_std": 0.9810051110107452,
|
914 |
+
"rewards/concensus_correctness_reward_func": 0.18512500077486038,
|
915 |
+
"rewards/consensus_reward_func": 0.8125,
|
916 |
"rewards/cumulative_reward_2": 0.0,
|
917 |
+
"rewards/final_correctness_reward_func": 0.1875,
|
918 |
+
"rewards/question_recreation_reward_func": 0.4930110676214099,
|
919 |
"rewards/soft_format_reward_func": 0.0,
|
920 |
+
"rewards/strict_format_reward_func": 0.046875,
|
921 |
+
"rewards/xmlcount_reward_func": 0.5190312538761646,
|
922 |
"step": 96
|
923 |
},
|
924 |
{
|
925 |
+
"completion_length": 239.34375,
|
926 |
+
"epoch": 16.333333333333332,
|
927 |
+
"grad_norm": 3.926326274871826,
|
928 |
+
"kl": 0.03365567361470312,
|
929 |
"learning_rate": 1.1791447083465133e-09,
|
930 |
"loss": 0.0,
|
931 |
+
"reward": 1.9388289339840412,
|
932 |
+
"reward_std": 0.958217971608974,
|
933 |
+
"rewards/concensus_correctness_reward_func": 0.11999999731779099,
|
934 |
+
"rewards/consensus_reward_func": 0.625,
|
935 |
"rewards/cumulative_reward_2": 0.0,
|
936 |
+
"rewards/final_correctness_reward_func": 0.0625,
|
937 |
+
"rewards/question_recreation_reward_func": 0.4502664606552571,
|
938 |
"rewards/soft_format_reward_func": 0.0,
|
939 |
+
"rewards/strict_format_reward_func": 0.0625,
|
940 |
+
"rewards/xmlcount_reward_func": 0.6185624990612268,
|
941 |
"step": 98
|
942 |
},
|
943 |
{
|
944 |
+
"completion_length": 271.6875,
|
945 |
+
"epoch": 16.666666666666668,
|
946 |
+
"grad_norm": 2.2361977100372314,
|
947 |
+
"kl": 0.02408603549702093,
|
948 |
"learning_rate": 1.3110773862126667e-10,
|
949 |
"loss": 0.0,
|
950 |
+
"reward": 1.8076389655470848,
|
951 |
+
"reward_std": 1.001751037707436,
|
952 |
+
"rewards/concensus_correctness_reward_func": 0.2303125038743019,
|
953 |
+
"rewards/consensus_reward_func": 0.5,
|
954 |
"rewards/cumulative_reward_2": 0.0,
|
955 |
+
"rewards/final_correctness_reward_func": 0.0,
|
956 |
+
"rewards/question_recreation_reward_func": 0.5585764544084668,
|
957 |
"rewards/soft_format_reward_func": 0.0,
|
958 |
+
"rewards/strict_format_reward_func": 0.046875,
|
959 |
+
"rewards/xmlcount_reward_func": 0.4718750088359229,
|
960 |
"step": 100
|
961 |
},
|
962 |
{
|
963 |
+
"epoch": 16.666666666666668,
|
964 |
"step": 100,
|
965 |
"total_flos": 0.0,
|
966 |
+
"train_loss": 1.5718676149845123e-05,
|
967 |
+
"train_runtime": 4262.7362,
|
968 |
+
"train_samples_per_second": 0.375,
|
969 |
+
"train_steps_per_second": 0.023
|
970 |
}
|
971 |
],
|
972 |
"logging_steps": 2,
|
973 |
"max_steps": 100,
|
974 |
"num_input_tokens_seen": 0,
|
975 |
+
"num_train_epochs": 17,
|
976 |
"save_steps": 25,
|
977 |
"stateful_callbacks": {
|
978 |
"TrainerControl": {
|
training_args.bin
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 5944
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:f771f4e41b930eadf03c7a0abf7ec0d72660f696af7340eebd672e2ccbbd1025
|
3 |
size 5944
|