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 +658 -658
- training_args.bin +1 -1
all_results.json
CHANGED
@@ -1,8 +1,8 @@
|
|
1 |
{
|
2 |
"total_flos": 0.0,
|
3 |
-
"train_loss":
|
4 |
-
"train_runtime":
|
5 |
-
"train_samples":
|
6 |
-
"train_samples_per_second":
|
7 |
-
"train_steps_per_second": 0.
|
8 |
}
|
|
|
1 |
{
|
2 |
"total_flos": 0.0,
|
3 |
+
"train_loss": 5.246203290880657e-06,
|
4 |
+
"train_runtime": 3714.3127,
|
5 |
+
"train_samples": 48,
|
6 |
+
"train_samples_per_second": 0.431,
|
7 |
+
"train_steps_per_second": 0.027
|
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:da4263e7a35562370a51d0fa373be9cc7d793b22ac7adabe400565ca4c12e9b4
|
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 sha256:820a3848e15044629f249d7c069f6631abe495b0c8c46d64e65d1d8e91b39678
|
3 |
size 1178224960
|
train_results.json
CHANGED
@@ -1,8 +1,8 @@
|
|
1 |
{
|
2 |
"total_flos": 0.0,
|
3 |
-
"train_loss":
|
4 |
-
"train_runtime":
|
5 |
-
"train_samples":
|
6 |
-
"train_samples_per_second":
|
7 |
-
"train_steps_per_second": 0.
|
8 |
}
|
|
|
1 |
{
|
2 |
"total_flos": 0.0,
|
3 |
+
"train_loss": 5.246203290880657e-06,
|
4 |
+
"train_runtime": 3714.3127,
|
5 |
+
"train_samples": 48,
|
6 |
+
"train_samples_per_second": 0.431,
|
7 |
+
"train_steps_per_second": 0.027
|
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":
|
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":
|
35 |
-
"kl": 0.
|
36 |
"learning_rate": 5e-07,
|
37 |
"loss": 0.0,
|
38 |
-
"reward":
|
39 |
-
"reward_std":
|
40 |
-
"rewards/concensus_correctness_reward_func":
|
41 |
-
"rewards/consensus_reward_func":
|
42 |
"rewards/cumulative_reward_2": 0.0,
|
43 |
-
"rewards/final_correctness_reward_func":
|
44 |
-
"rewards/question_recreation_reward_func": 0.
|
45 |
"rewards/soft_format_reward_func": 0.0,
|
46 |
-
"rewards/strict_format_reward_func": 0.
|
47 |
-
"rewards/xmlcount_reward_func": 0.
|
48 |
"step": 4
|
49 |
},
|
50 |
{
|
51 |
-
"completion_length":
|
52 |
-
"epoch": 0
|
53 |
-
"grad_norm":
|
54 |
-
"kl":
|
55 |
"learning_rate": 4.994757065594279e-07,
|
56 |
-
"loss": 0.
|
57 |
-
"reward":
|
58 |
-
"reward_std":
|
59 |
-
"rewards/concensus_correctness_reward_func":
|
60 |
-
"rewards/consensus_reward_func":
|
61 |
"rewards/cumulative_reward_2": 0.0,
|
62 |
-
"rewards/final_correctness_reward_func":
|
63 |
-
"rewards/question_recreation_reward_func": 0.
|
64 |
"rewards/soft_format_reward_func": 0.0,
|
65 |
-
"rewards/strict_format_reward_func": 0.
|
66 |
-
"rewards/xmlcount_reward_func":
|
67 |
"step": 6
|
68 |
},
|
69 |
{
|
70 |
-
"completion_length":
|
71 |
-
"epoch":
|
72 |
-
"grad_norm":
|
73 |
-
"kl":
|
74 |
"learning_rate": 4.979050253066063e-07,
|
75 |
-
"loss": 0.
|
76 |
-
"reward":
|
77 |
-
"reward_std":
|
78 |
-
"rewards/concensus_correctness_reward_func":
|
79 |
-
"rewards/consensus_reward_func":
|
80 |
"rewards/cumulative_reward_2": 0.0,
|
81 |
-
"rewards/final_correctness_reward_func":
|
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":
|
93 |
"learning_rate": 4.952945442245597e-07,
|
94 |
-
"loss": 0.
|
95 |
-
"reward":
|
96 |
-
"reward_std":
|
97 |
-
"rewards/concensus_correctness_reward_func":
|
98 |
-
"rewards/consensus_reward_func":
|
99 |
-
"rewards/cumulative_reward_2": 0.0,
|
100 |
-
"rewards/final_correctness_reward_func":
|
101 |
-
"rewards/question_recreation_reward_func": 0.
|
102 |
-
"rewards/soft_format_reward_func": 0.
|
103 |
-
"rewards/strict_format_reward_func": 0.
|
104 |
-
"rewards/xmlcount_reward_func":
|
105 |
"step": 10
|
106 |
},
|
107 |
{
|
108 |
-
"completion_length":
|
109 |
-
"epoch": 0
|
110 |
-
"grad_norm":
|
111 |
-
"kl":
|
112 |
"learning_rate": 4.916552125781528e-07,
|
113 |
-
"loss": 0.
|
114 |
-
"reward":
|
115 |
-
"reward_std":
|
116 |
-
"rewards/concensus_correctness_reward_func":
|
117 |
-
"rewards/consensus_reward_func":
|
118 |
"rewards/cumulative_reward_2": 0.0,
|
119 |
-
"rewards/final_correctness_reward_func":
|
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":
|
124 |
"step": 12
|
125 |
},
|
126 |
{
|
127 |
-
"completion_length":
|
128 |
-
"epoch":
|
129 |
-
"grad_norm":
|
130 |
-
"kl":
|
131 |
"learning_rate": 4.870022949890676e-07,
|
132 |
-
"loss": 0.
|
133 |
-
"reward":
|
134 |
-
"reward_std":
|
135 |
-
"rewards/concensus_correctness_reward_func":
|
136 |
-
"rewards/consensus_reward_func":
|
137 |
"rewards/cumulative_reward_2": 0.0,
|
138 |
-
"rewards/final_correctness_reward_func":
|
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":
|
149 |
-
"kl":
|
150 |
"learning_rate": 4.81355307410676e-07,
|
151 |
-
"loss": 0.
|
152 |
-
"reward":
|
153 |
-
"reward_std":
|
154 |
-
"rewards/concensus_correctness_reward_func":
|
155 |
-
"rewards/consensus_reward_func":
|
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":
|
162 |
"step": 16
|
163 |
},
|
164 |
{
|
165 |
-
"completion_length":
|
166 |
-
"epoch":
|
167 |
-
"grad_norm":
|
168 |
-
"kl":
|
169 |
"learning_rate": 4.747379352713488e-07,
|
170 |
-
"loss": 0.
|
171 |
-
"reward":
|
172 |
-
"reward_std":
|
173 |
-
"rewards/concensus_correctness_reward_func":
|
174 |
-
"rewards/consensus_reward_func":
|
175 |
"rewards/cumulative_reward_2": 0.0,
|
176 |
-
"rewards/final_correctness_reward_func":
|
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":
|
181 |
"step": 18
|
182 |
},
|
183 |
{
|
184 |
-
"completion_length":
|
185 |
-
"epoch":
|
186 |
-
"grad_norm":
|
187 |
-
"kl":
|
188 |
"learning_rate": 4.6717793412953776e-07,
|
189 |
-
"loss": 0.
|
190 |
-
"reward":
|
191 |
-
"reward_std":
|
192 |
-
"rewards/concensus_correctness_reward_func":
|
193 |
-
"rewards/consensus_reward_func":
|
194 |
"rewards/cumulative_reward_2": 0.0,
|
195 |
-
"rewards/final_correctness_reward_func":
|
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":
|
200 |
"step": 20
|
201 |
},
|
202 |
{
|
203 |
-
"completion_length":
|
204 |
-
"epoch":
|
205 |
-
"grad_norm":
|
206 |
-
"kl":
|
207 |
"learning_rate": 4.5870701325731773e-07,
|
208 |
-
"loss": 0.
|
209 |
-
"reward":
|
210 |
-
"reward_std":
|
211 |
-
"rewards/concensus_correctness_reward_func":
|
212 |
-
"rewards/consensus_reward_func":
|
213 |
"rewards/cumulative_reward_2": 0.0,
|
214 |
-
"rewards/final_correctness_reward_func":
|
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":
|
219 |
"step": 22
|
220 |
},
|
221 |
{
|
222 |
-
"completion_length":
|
223 |
-
"epoch":
|
224 |
-
"grad_norm":
|
225 |
-
"kl":
|
226 |
"learning_rate": 4.4936070264068016e-07,
|
227 |
-
"loss": 0.
|
228 |
-
"reward":
|
229 |
-
"reward_std":
|
230 |
-
"rewards/concensus_correctness_reward_func":
|
231 |
-
"rewards/consensus_reward_func":
|
232 |
"rewards/cumulative_reward_2": 0.0,
|
233 |
-
"rewards/final_correctness_reward_func":
|
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":
|
238 |
"step": 24
|
239 |
},
|
240 |
{
|
241 |
-
"completion_length":
|
242 |
-
"epoch":
|
243 |
-
"grad_norm":
|
244 |
-
"kl":
|
245 |
"learning_rate": 4.391782039544238e-07,
|
246 |
-
"loss": 0.
|
247 |
-
"reward":
|
248 |
-
"reward_std":
|
249 |
-
"rewards/concensus_correctness_reward_func":
|
250 |
-
"rewards/consensus_reward_func":
|
251 |
"rewards/cumulative_reward_2": 0.0,
|
252 |
-
"rewards/final_correctness_reward_func":
|
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":
|
257 |
"step": 26
|
258 |
},
|
259 |
{
|
260 |
-
"completion_length":
|
261 |
-
"epoch":
|
262 |
-
"grad_norm":
|
263 |
-
"kl":
|
264 |
"learning_rate": 4.282022261367073e-07,
|
265 |
-
"loss": 0.
|
266 |
-
"reward":
|
267 |
-
"reward_std":
|
268 |
-
"rewards/concensus_correctness_reward_func":
|
269 |
-
"rewards/consensus_reward_func": 0.
|
270 |
"rewards/cumulative_reward_2": 0.0,
|
271 |
-
"rewards/final_correctness_reward_func":
|
272 |
-
"rewards/question_recreation_reward_func": 0.
|
273 |
"rewards/soft_format_reward_func": 0.0,
|
274 |
-
"rewards/strict_format_reward_func": 0.
|
275 |
-
"rewards/xmlcount_reward_func":
|
276 |
"step": 28
|
277 |
},
|
278 |
{
|
279 |
-
"completion_length":
|
280 |
-
"epoch":
|
281 |
-
"grad_norm":
|
282 |
-
"kl":
|
283 |
"learning_rate": 4.1647880625292027e-07,
|
284 |
-
"loss": 0.
|
285 |
-
"reward":
|
286 |
-
"reward_std":
|
287 |
-
"rewards/concensus_correctness_reward_func":
|
288 |
-
"rewards/consensus_reward_func":
|
289 |
"rewards/cumulative_reward_2": 0.0,
|
290 |
-
"rewards/final_correctness_reward_func":
|
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":
|
295 |
"step": 30
|
296 |
},
|
297 |
{
|
298 |
-
"completion_length":
|
299 |
-
"epoch":
|
300 |
-
"grad_norm":
|
301 |
-
"kl":
|
302 |
"learning_rate": 4.040571164002318e-07,
|
303 |
-
"loss": 0.
|
304 |
-
"reward":
|
305 |
-
"reward_std":
|
306 |
-
"rewards/concensus_correctness_reward_func":
|
307 |
-
"rewards/consensus_reward_func":
|
308 |
"rewards/cumulative_reward_2": 0.0,
|
309 |
-
"rewards/final_correctness_reward_func":
|
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":
|
314 |
"step": 32
|
315 |
},
|
316 |
{
|
317 |
-
"completion_length":
|
318 |
-
"epoch":
|
319 |
-
"grad_norm":
|
320 |
-
"kl":
|
321 |
"learning_rate": 3.909892574627266e-07,
|
322 |
-
"loss": 0.
|
323 |
-
"reward":
|
324 |
-
"reward_std":
|
325 |
-
"rewards/concensus_correctness_reward_func":
|
326 |
-
"rewards/consensus_reward_func":
|
327 |
"rewards/cumulative_reward_2": 0.0,
|
328 |
-
"rewards/final_correctness_reward_func":
|
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":
|
333 |
"step": 34
|
334 |
},
|
335 |
{
|
336 |
-
"completion_length":
|
337 |
-
"epoch":
|
338 |
-
"grad_norm":
|
339 |
-
"kl":
|
340 |
"learning_rate": 3.773300405821908e-07,
|
341 |
-
"loss": 0.
|
342 |
-
"reward":
|
343 |
-
"reward_std":
|
344 |
-
"rewards/concensus_correctness_reward_func":
|
345 |
-
"rewards/consensus_reward_func": 1.
|
346 |
"rewards/cumulative_reward_2": 0.0,
|
347 |
-
"rewards/final_correctness_reward_func":
|
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":
|
359 |
"learning_rate": 3.6313675726113475e-07,
|
360 |
-
"loss": 0.
|
361 |
-
"reward":
|
362 |
-
"reward_std":
|
363 |
-
"rewards/concensus_correctness_reward_func":
|
364 |
-
"rewards/consensus_reward_func": 0.
|
365 |
"rewards/cumulative_reward_2": 0.0,
|
366 |
-
"rewards/final_correctness_reward_func":
|
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":
|
371 |
"step": 38
|
372 |
},
|
373 |
{
|
374 |
-
"completion_length":
|
375 |
-
"epoch":
|
376 |
-
"grad_norm":
|
377 |
-
"kl":
|
378 |
"learning_rate": 3.484689390623218e-07,
|
379 |
-
"loss": 0.
|
380 |
-
"reward":
|
381 |
-
"reward_std":
|
382 |
-
"rewards/concensus_correctness_reward_func":
|
383 |
-
"rewards/consensus_reward_func":
|
384 |
"rewards/cumulative_reward_2": 0.0,
|
385 |
-
"rewards/final_correctness_reward_func":
|
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":
|
390 |
"step": 40
|
391 |
},
|
392 |
{
|
393 |
-
"completion_length":
|
394 |
-
"epoch":
|
395 |
-
"grad_norm":
|
396 |
-
"kl":
|
397 |
"learning_rate": 3.3338810791270517e-07,
|
398 |
-
"loss": 0.
|
399 |
-
"reward":
|
400 |
-
"reward_std":
|
401 |
-
"rewards/concensus_correctness_reward_func":
|
402 |
-
"rewards/consensus_reward_func":
|
403 |
"rewards/cumulative_reward_2": 0.0,
|
404 |
-
"rewards/final_correctness_reward_func":
|
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":
|
416 |
"learning_rate": 3.179575180590857e-07,
|
417 |
-
"loss": 0.
|
418 |
-
"reward":
|
419 |
-
"reward_std":
|
420 |
-
"rewards/concensus_correctness_reward_func":
|
421 |
-
"rewards/consensus_reward_func":
|
422 |
"rewards/cumulative_reward_2": 0.0,
|
423 |
-
"rewards/final_correctness_reward_func":
|
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":
|
428 |
"step": 44
|
429 |
},
|
430 |
{
|
431 |
-
"completion_length":
|
432 |
-
"epoch":
|
433 |
-
"grad_norm":
|
434 |
-
"kl":
|
435 |
"learning_rate": 3.022418907578188e-07,
|
436 |
-
"loss": 0.
|
437 |
-
"reward":
|
438 |
-
"reward_std":
|
439 |
-
"rewards/concensus_correctness_reward_func":
|
440 |
-
"rewards/consensus_reward_func":
|
441 |
"rewards/cumulative_reward_2": 0.0,
|
442 |
-
"rewards/final_correctness_reward_func":
|
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":
|
447 |
"step": 46
|
448 |
},
|
449 |
{
|
450 |
-
"completion_length":
|
451 |
-
"epoch":
|
452 |
-
"grad_norm":
|
453 |
-
"kl":
|
454 |
"learning_rate": 2.863071428113726e-07,
|
455 |
-
"loss": 0.
|
456 |
-
"reward":
|
457 |
-
"reward_std":
|
458 |
-
"rewards/concensus_correctness_reward_func":
|
459 |
-
"rewards/consensus_reward_func":
|
460 |
"rewards/cumulative_reward_2": 0.0,
|
461 |
-
"rewards/final_correctness_reward_func":
|
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":
|
473 |
"learning_rate": 2.7022011009035107e-07,
|
474 |
-
"loss": 0.
|
475 |
-
"reward":
|
476 |
-
"reward_std":
|
477 |
-
"rewards/concensus_correctness_reward_func":
|
478 |
-
"rewards/consensus_reward_func":
|
479 |
"rewards/cumulative_reward_2": 0.0,
|
480 |
-
"rewards/final_correctness_reward_func":
|
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":
|
485 |
"step": 50
|
486 |
},
|
487 |
{
|
488 |
-
"completion_length":
|
489 |
-
"epoch":
|
490 |
-
"grad_norm":
|
491 |
-
"kl":
|
492 |
"learning_rate": 2.540482672006254e-07,
|
493 |
-
"loss": 0.
|
494 |
-
"reward":
|
495 |
-
"reward_std":
|
496 |
-
"rewards/concensus_correctness_reward_func":
|
497 |
-
"rewards/consensus_reward_func":
|
498 |
"rewards/cumulative_reward_2": 0.0,
|
499 |
-
"rewards/final_correctness_reward_func":
|
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":
|
504 |
"step": 52
|
505 |
},
|
506 |
{
|
507 |
-
"completion_length":
|
508 |
-
"epoch":
|
509 |
-
"grad_norm":
|
510 |
-
"kl":
|
511 |
"learning_rate": 2.37859444471388e-07,
|
512 |
-
"loss": 0.
|
513 |
-
"reward":
|
514 |
-
"reward_std":
|
515 |
-
"rewards/concensus_correctness_reward_func":
|
516 |
-
"rewards/consensus_reward_func": 0.
|
517 |
"rewards/cumulative_reward_2": 0.0,
|
518 |
-
"rewards/final_correctness_reward_func":
|
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":
|
523 |
"step": 54
|
524 |
},
|
525 |
{
|
526 |
-
"completion_length":
|
527 |
-
"epoch":
|
528 |
-
"grad_norm":
|
529 |
-
"kl":
|
530 |
"learning_rate": 2.2172154345117894e-07,
|
531 |
-
"loss": 0.
|
532 |
-
"reward":
|
533 |
-
"reward_std":
|
534 |
-
"rewards/concensus_correctness_reward_func":
|
535 |
-
"rewards/consensus_reward_func": 1.
|
536 |
"rewards/cumulative_reward_2": 0.0,
|
537 |
-
"rewards/final_correctness_reward_func":
|
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":
|
548 |
-
"kl":
|
549 |
"learning_rate": 2.0570225210519433e-07,
|
550 |
-
"loss": 0.
|
551 |
-
"reward":
|
552 |
-
"reward_std":
|
553 |
-
"rewards/concensus_correctness_reward_func":
|
554 |
-
"rewards/consensus_reward_func":
|
555 |
"rewards/cumulative_reward_2": 0.0,
|
556 |
-
"rewards/final_correctness_reward_func":
|
557 |
-
"rewards/question_recreation_reward_func": 0.
|
558 |
"rewards/soft_format_reward_func": 0.0,
|
559 |
-
"rewards/strict_format_reward_func": 0.
|
560 |
-
"rewards/xmlcount_reward_func":
|
561 |
"step": 58
|
562 |
},
|
563 |
{
|
564 |
-
"completion_length":
|
565 |
-
"epoch":
|
566 |
-
"grad_norm":
|
567 |
-
"kl":
|
568 |
"learning_rate": 1.8986876090843664e-07,
|
569 |
-
"loss": 0.
|
570 |
-
"reward":
|
571 |
-
"reward_std":
|
572 |
-
"rewards/concensus_correctness_reward_func":
|
573 |
-
"rewards/consensus_reward_func":
|
574 |
"rewards/cumulative_reward_2": 0.0,
|
575 |
-
"rewards/final_correctness_reward_func":
|
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":
|
580 |
"step": 60
|
581 |
},
|
582 |
{
|
583 |
-
"completion_length":
|
584 |
-
"epoch":
|
585 |
-
"grad_norm":
|
586 |
-
"kl":
|
587 |
"learning_rate": 1.7428748102551234e-07,
|
588 |
-
"loss": 0.
|
589 |
-
"reward":
|
590 |
-
"reward_std":
|
591 |
-
"rewards/concensus_correctness_reward_func":
|
592 |
-
"rewards/consensus_reward_func":
|
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.
|
598 |
-
"rewards/xmlcount_reward_func":
|
599 |
"step": 62
|
600 |
},
|
601 |
{
|
602 |
-
"completion_length":
|
603 |
-
"epoch":
|
604 |
-
"grad_norm":
|
605 |
-
"kl":
|
606 |
"learning_rate": 1.5902376575912814e-07,
|
607 |
-
"loss": 0.
|
608 |
-
"reward":
|
609 |
-
"reward_std":
|
610 |
-
"rewards/concensus_correctness_reward_func":
|
611 |
-
"rewards/consensus_reward_func":
|
612 |
"rewards/cumulative_reward_2": 0.0,
|
613 |
-
"rewards/final_correctness_reward_func":
|
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":
|
618 |
"step": 64
|
619 |
},
|
620 |
{
|
621 |
-
"completion_length":
|
622 |
-
"epoch":
|
623 |
-
"grad_norm":
|
624 |
-
"kl":
|
625 |
"learning_rate": 1.4414163643562753e-07,
|
626 |
-
"loss": 0.
|
627 |
-
"reward":
|
628 |
-
"reward_std":
|
629 |
-
"rewards/concensus_correctness_reward_func":
|
630 |
-
"rewards/consensus_reward_func":
|
631 |
"rewards/cumulative_reward_2": 0.0,
|
632 |
-
"rewards/final_correctness_reward_func":
|
633 |
-
"rewards/question_recreation_reward_func": 0.
|
634 |
"rewards/soft_format_reward_func": 0.0,
|
635 |
-
"rewards/strict_format_reward_func": 0.
|
636 |
-
"rewards/xmlcount_reward_func":
|
637 |
"step": 66
|
638 |
},
|
639 |
{
|
640 |
-
"completion_length":
|
641 |
-
"epoch":
|
642 |
-
"grad_norm":
|
643 |
-
"kl":
|
644 |
"learning_rate": 1.2970351387729872e-07,
|
645 |
-
"loss": 0.
|
646 |
-
"reward":
|
647 |
-
"reward_std":
|
648 |
-
"rewards/concensus_correctness_reward_func":
|
649 |
-
"rewards/consensus_reward_func":
|
650 |
"rewards/cumulative_reward_2": 0.0,
|
651 |
-
"rewards/final_correctness_reward_func":
|
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":
|
656 |
"step": 68
|
657 |
},
|
658 |
{
|
659 |
-
"completion_length":
|
660 |
-
"epoch":
|
661 |
-
"grad_norm":
|
662 |
-
"kl":
|
663 |
"learning_rate": 1.1576995658775404e-07,
|
664 |
-
"loss": 0.
|
665 |
-
"reward":
|
666 |
-
"reward_std":
|
667 |
-
"rewards/concensus_correctness_reward_func":
|
668 |
-
"rewards/consensus_reward_func":
|
669 |
-
"rewards/cumulative_reward_2": 0.0,
|
670 |
-
"rewards/final_correctness_reward_func":
|
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":
|
675 |
"step": 70
|
676 |
},
|
677 |
{
|
678 |
-
"completion_length":
|
679 |
-
"epoch":
|
680 |
-
"grad_norm":
|
681 |
-
"kl":
|
682 |
"learning_rate": 1.0239940674851941e-07,
|
683 |
-
"loss": 0.
|
684 |
-
"reward":
|
685 |
-
"reward_std":
|
686 |
-
"rewards/concensus_correctness_reward_func":
|
687 |
-
"rewards/consensus_reward_func":
|
688 |
"rewards/cumulative_reward_2": 0.0,
|
689 |
-
"rewards/final_correctness_reward_func":
|
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":
|
700 |
-
"kl":
|
701 |
"learning_rate": 8.964794509221507e-08,
|
702 |
-
"loss": 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":
|
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":
|
713 |
"step": 74
|
714 |
},
|
715 |
{
|
716 |
-
"completion_length":
|
717 |
-
"epoch":
|
718 |
-
"grad_norm":
|
719 |
-
"kl":
|
720 |
"learning_rate": 7.756905568047392e-08,
|
721 |
-
"loss": 0.
|
722 |
-
"reward":
|
723 |
-
"reward_std":
|
724 |
-
"rewards/concensus_correctness_reward_func":
|
725 |
-
"rewards/consensus_reward_func":
|
726 |
"rewards/cumulative_reward_2": 0.0,
|
727 |
-
"rewards/final_correctness_reward_func":
|
728 |
-
"rewards/question_recreation_reward_func": 0.
|
729 |
"rewards/soft_format_reward_func": 0.0,
|
730 |
-
"rewards/strict_format_reward_func": 0.
|
731 |
-
"rewards/xmlcount_reward_func":
|
732 |
"step": 76
|
733 |
},
|
734 |
{
|
735 |
-
"completion_length":
|
736 |
-
"epoch":
|
737 |
-
"grad_norm":
|
738 |
-
"kl":
|
739 |
"learning_rate": 6.621340157319996e-08,
|
740 |
-
"loss": 0.
|
741 |
-
"reward":
|
742 |
-
"reward_std":
|
743 |
-
"rewards/concensus_correctness_reward_func":
|
744 |
-
"rewards/consensus_reward_func":
|
745 |
"rewards/cumulative_reward_2": 0.0,
|
746 |
-
"rewards/final_correctness_reward_func":
|
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":
|
751 |
"step": 78
|
752 |
},
|
753 |
{
|
754 |
-
"completion_length":
|
755 |
-
"epoch":
|
756 |
-
"grad_norm":
|
757 |
-
"kl":
|
758 |
"learning_rate": 5.5628612330087724e-08,
|
759 |
-
"loss": 0.
|
760 |
-
"reward":
|
761 |
-
"reward_std":
|
762 |
-
"rewards/concensus_correctness_reward_func":
|
763 |
-
"rewards/consensus_reward_func": 0.
|
764 |
"rewards/cumulative_reward_2": 0.0,
|
765 |
-
"rewards/final_correctness_reward_func":
|
766 |
-
"rewards/question_recreation_reward_func": 0.
|
767 |
"rewards/soft_format_reward_func": 0.0,
|
768 |
-
"rewards/strict_format_reward_func": 0.
|
769 |
-
"rewards/xmlcount_reward_func":
|
770 |
"step": 80
|
771 |
},
|
772 |
{
|
773 |
-
"completion_length":
|
774 |
-
"epoch":
|
775 |
-
"grad_norm":
|
776 |
-
"kl":
|
777 |
"learning_rate": 4.5859084235697235e-08,
|
778 |
-
"loss": 0.
|
779 |
-
"reward":
|
780 |
-
"reward_std":
|
781 |
-
"rewards/concensus_correctness_reward_func":
|
782 |
-
"rewards/consensus_reward_func":
|
783 |
"rewards/cumulative_reward_2": 0.0,
|
784 |
-
"rewards/final_correctness_reward_func":
|
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": 263.
|
793 |
-
"epoch":
|
794 |
-
"grad_norm":
|
795 |
-
"kl":
|
796 |
"learning_rate": 3.6945794086007705e-08,
|
797 |
-
"loss": 0.
|
798 |
-
"reward":
|
799 |
-
"reward_std":
|
800 |
-
"rewards/concensus_correctness_reward_func":
|
801 |
-
"rewards/consensus_reward_func":
|
802 |
"rewards/cumulative_reward_2": 0.0,
|
803 |
-
"rewards/final_correctness_reward_func":
|
804 |
-
"rewards/question_recreation_reward_func": 0.
|
805 |
"rewards/soft_format_reward_func": 0.0,
|
806 |
-
"rewards/strict_format_reward_func": 0.
|
807 |
-
"rewards/xmlcount_reward_func":
|
808 |
"step": 84
|
809 |
},
|
810 |
{
|
811 |
-
"completion_length":
|
812 |
-
"epoch":
|
813 |
-
"grad_norm":
|
814 |
-
"kl":
|
815 |
"learning_rate": 2.892612731749414e-08,
|
816 |
-
"loss": 0.
|
817 |
-
"reward":
|
818 |
-
"reward_std":
|
819 |
-
"rewards/concensus_correctness_reward_func":
|
820 |
-
"rewards/consensus_reward_func":
|
821 |
"rewards/cumulative_reward_2": 0.0,
|
822 |
-
"rewards/final_correctness_reward_func":
|
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":
|
833 |
-
"kl":
|
834 |
"learning_rate": 2.183372119961499e-08,
|
835 |
-
"loss": 0.
|
836 |
-
"reward":
|
837 |
-
"reward_std":
|
838 |
-
"rewards/concensus_correctness_reward_func":
|
839 |
-
"rewards/consensus_reward_func":
|
840 |
"rewards/cumulative_reward_2": 0.0,
|
841 |
-
"rewards/final_correctness_reward_func":
|
842 |
-
"rewards/question_recreation_reward_func": 0.
|
843 |
"rewards/soft_format_reward_func": 0.0,
|
844 |
-
"rewards/strict_format_reward_func": 0.
|
845 |
-
"rewards/xmlcount_reward_func":
|
846 |
"step": 88
|
847 |
},
|
848 |
{
|
849 |
-
"completion_length":
|
850 |
-
"epoch":
|
851 |
-
"grad_norm":
|
852 |
-
"kl":
|
853 |
"learning_rate": 1.5698323748414122e-08,
|
854 |
-
"loss": 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":
|
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":
|
872 |
"learning_rate": 1.054566895300324e-08,
|
873 |
-
"loss": 0.
|
874 |
-
"reward":
|
875 |
-
"reward_std":
|
876 |
-
"rewards/concensus_correctness_reward_func":
|
877 |
-
"rewards/consensus_reward_func":
|
878 |
"rewards/cumulative_reward_2": 0.0,
|
879 |
-
"rewards/final_correctness_reward_func":
|
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":
|
891 |
"learning_rate": 6.397368838268496e-09,
|
892 |
-
"loss": 0.
|
893 |
-
"reward":
|
894 |
-
"reward_std":
|
895 |
-
"rewards/concensus_correctness_reward_func":
|
896 |
-
"rewards/consensus_reward_func":
|
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":
|
910 |
"learning_rate": 3.2708228165273244e-09,
|
911 |
-
"loss": 0.
|
912 |
-
"reward":
|
913 |
-
"reward_std":
|
914 |
-
"rewards/concensus_correctness_reward_func":
|
915 |
-
"rewards/consensus_reward_func": 1.
|
916 |
"rewards/cumulative_reward_2": 0.0,
|
917 |
-
"rewards/final_correctness_reward_func":
|
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":
|
922 |
"step": 96
|
923 |
},
|
924 |
{
|
925 |
-
"completion_length":
|
926 |
-
"epoch":
|
927 |
-
"grad_norm":
|
928 |
-
"kl":
|
929 |
"learning_rate": 1.1791447083465133e-09,
|
930 |
-
"loss": 0.
|
931 |
-
"reward":
|
932 |
-
"reward_std":
|
933 |
-
"rewards/concensus_correctness_reward_func":
|
934 |
-
"rewards/consensus_reward_func": 0.
|
935 |
"rewards/cumulative_reward_2": 0.0,
|
936 |
-
"rewards/final_correctness_reward_func":
|
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":
|
941 |
"step": 98
|
942 |
},
|
943 |
{
|
944 |
-
"completion_length":
|
945 |
-
"epoch":
|
946 |
-
"grad_norm":
|
947 |
-
"kl":
|
948 |
"learning_rate": 1.3110773862126667e-10,
|
949 |
-
"loss": 0.
|
950 |
-
"reward":
|
951 |
-
"reward_std":
|
952 |
-
"rewards/concensus_correctness_reward_func":
|
953 |
-
"rewards/consensus_reward_func":
|
954 |
"rewards/cumulative_reward_2": 0.0,
|
955 |
-
"rewards/final_correctness_reward_func":
|
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":
|
960 |
"step": 100
|
961 |
},
|
962 |
{
|
963 |
-
"epoch":
|
964 |
"step": 100,
|
965 |
"total_flos": 0.0,
|
966 |
-
"train_loss":
|
967 |
-
"train_runtime":
|
968 |
-
"train_samples_per_second":
|
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": 248.21875,
|
14 |
+
"epoch": 0.3333333333333333,
|
15 |
+
"grad_norm": 3.4614758491516113,
|
16 |
"kl": 0.0,
|
17 |
"learning_rate": 1.6666666666666665e-07,
|
18 |
"loss": -0.0,
|
19 |
+
"reward": 1.7851608581840992,
|
20 |
+
"reward_std": 1.0083980893250555,
|
21 |
+
"rewards/concensus_correctness_reward_func": 0.12012499943375587,
|
22 |
+
"rewards/consensus_reward_func": 0.4375,
|
23 |
"rewards/cumulative_reward_2": 0.0,
|
24 |
+
"rewards/final_correctness_reward_func": 0.25,
|
25 |
+
"rewards/question_recreation_reward_func": 0.518067104741931,
|
26 |
"rewards/soft_format_reward_func": 0.0,
|
27 |
+
"rewards/strict_format_reward_func": 0.046875,
|
28 |
+
"rewards/xmlcount_reward_func": 0.4125937600620091,
|
29 |
"step": 2
|
30 |
},
|
31 |
{
|
32 |
+
"completion_length": 273.46875,
|
33 |
+
"epoch": 0.6666666666666666,
|
34 |
+
"grad_norm": 2.2614688873291016,
|
35 |
+
"kl": 0.000583945713515277,
|
36 |
"learning_rate": 5e-07,
|
37 |
"loss": 0.0,
|
38 |
+
"reward": 2.26899578794837,
|
39 |
+
"reward_std": 0.8781831180676818,
|
40 |
+
"rewards/concensus_correctness_reward_func": 0.18174999952316284,
|
41 |
+
"rewards/consensus_reward_func": 0.625,
|
42 |
"rewards/cumulative_reward_2": 0.0,
|
43 |
+
"rewards/final_correctness_reward_func": 0.125,
|
44 |
+
"rewards/question_recreation_reward_func": 0.6493395259603858,
|
45 |
"rewards/soft_format_reward_func": 0.0,
|
46 |
+
"rewards/strict_format_reward_func": 0.046875,
|
47 |
+
"rewards/xmlcount_reward_func": 0.6410312540829182,
|
48 |
"step": 4
|
49 |
},
|
50 |
{
|
51 |
+
"completion_length": 230.65625,
|
52 |
+
"epoch": 1.0,
|
53 |
+
"grad_norm": 2.344890594482422,
|
54 |
+
"kl": 0.0007161081903177546,
|
55 |
"learning_rate": 4.994757065594279e-07,
|
56 |
+
"loss": 0.0,
|
57 |
+
"reward": 1.7817438933998346,
|
58 |
+
"reward_std": 0.8162978617474437,
|
59 |
+
"rewards/concensus_correctness_reward_func": 0.09124999865889549,
|
60 |
+
"rewards/consensus_reward_func": 0.6875,
|
61 |
"rewards/cumulative_reward_2": 0.0,
|
62 |
+
"rewards/final_correctness_reward_func": 0.0,
|
63 |
+
"rewards/question_recreation_reward_func": 0.4062751979799941,
|
64 |
"rewards/soft_format_reward_func": 0.0,
|
65 |
+
"rewards/strict_format_reward_func": 0.046875,
|
66 |
+
"rewards/xmlcount_reward_func": 0.5498437532223761,
|
67 |
"step": 6
|
68 |
},
|
69 |
{
|
70 |
+
"completion_length": 219.8125,
|
71 |
+
"epoch": 1.3333333333333333,
|
72 |
+
"grad_norm": 26.660634994506836,
|
73 |
+
"kl": 0.0012114797100366559,
|
74 |
"learning_rate": 4.979050253066063e-07,
|
75 |
+
"loss": -0.0,
|
76 |
+
"reward": 1.8910054825246334,
|
77 |
+
"reward_std": 1.1408569859668205,
|
78 |
+
"rewards/concensus_correctness_reward_func": 0.12249999865889549,
|
79 |
+
"rewards/consensus_reward_func": 0.4375,
|
80 |
"rewards/cumulative_reward_2": 0.0,
|
81 |
+
"rewards/final_correctness_reward_func": 0.125,
|
82 |
+
"rewards/question_recreation_reward_func": 0.536755473469384,
|
83 |
"rewards/soft_format_reward_func": 0.0,
|
84 |
+
"rewards/strict_format_reward_func": 0.078125,
|
85 |
+
"rewards/xmlcount_reward_func": 0.5911250002682209,
|
86 |
"step": 8
|
87 |
},
|
88 |
{
|
89 |
+
"completion_length": 202.46875,
|
90 |
+
"epoch": 1.6666666666666665,
|
91 |
+
"grad_norm": 7.760759353637695,
|
92 |
+
"kl": 0.0012618956752703525,
|
93 |
"learning_rate": 4.952945442245597e-07,
|
94 |
+
"loss": 0.0,
|
95 |
+
"reward": 1.9051898624747992,
|
96 |
+
"reward_std": 0.958464287687093,
|
97 |
+
"rewards/concensus_correctness_reward_func": 0.12012499943375587,
|
98 |
+
"rewards/consensus_reward_func": 0.4375,
|
99 |
+
"rewards/cumulative_reward_2": 0.0,
|
100 |
+
"rewards/final_correctness_reward_func": 0.1875,
|
101 |
+
"rewards/question_recreation_reward_func": 0.3846898500341922,
|
102 |
+
"rewards/soft_format_reward_func": 0.015625,
|
103 |
+
"rewards/strict_format_reward_func": 0.09375,
|
104 |
+
"rewards/xmlcount_reward_func": 0.6660000011324883,
|
105 |
"step": 10
|
106 |
},
|
107 |
{
|
108 |
+
"completion_length": 336.28125,
|
109 |
+
"epoch": 2.0,
|
110 |
+
"grad_norm": 2.4356420040130615,
|
111 |
+
"kl": 0.0008406055276282132,
|
112 |
"learning_rate": 4.916552125781528e-07,
|
113 |
+
"loss": 0.0,
|
114 |
+
"reward": 1.8280094247311354,
|
115 |
+
"reward_std": 1.094328472390771,
|
116 |
+
"rewards/concensus_correctness_reward_func": 0.06012500077486038,
|
117 |
+
"rewards/consensus_reward_func": 0.5625,
|
118 |
"rewards/cumulative_reward_2": 0.0,
|
119 |
+
"rewards/final_correctness_reward_func": 0.3125,
|
120 |
+
"rewards/question_recreation_reward_func": 0.5948844254016876,
|
121 |
"rewards/soft_format_reward_func": 0.0,
|
122 |
+
"rewards/strict_format_reward_func": 0.015625,
|
123 |
+
"rewards/xmlcount_reward_func": 0.28237500321120024,
|
124 |
"step": 12
|
125 |
},
|
126 |
{
|
127 |
+
"completion_length": 282.8125,
|
128 |
+
"epoch": 2.3333333333333335,
|
129 |
+
"grad_norm": 2.0542495250701904,
|
130 |
+
"kl": 0.0009237846843461739,
|
131 |
"learning_rate": 4.870022949890676e-07,
|
132 |
+
"loss": 0.0,
|
133 |
+
"reward": 1.8283372893929482,
|
134 |
+
"reward_std": 1.189666131976992,
|
135 |
+
"rewards/concensus_correctness_reward_func": 0.23618749901652336,
|
136 |
+
"rewards/consensus_reward_func": 0.625,
|
137 |
"rewards/cumulative_reward_2": 0.0,
|
138 |
+
"rewards/final_correctness_reward_func": 0.125,
|
139 |
+
"rewards/question_recreation_reward_func": 0.6880872882902622,
|
140 |
"rewards/soft_format_reward_func": 0.0,
|
141 |
+
"rewards/strict_format_reward_func": 0.0,
|
142 |
+
"rewards/xmlcount_reward_func": 0.15406249510124326,
|
143 |
"step": 14
|
144 |
},
|
145 |
{
|
146 |
+
"completion_length": 266.84375,
|
147 |
+
"epoch": 2.6666666666666665,
|
148 |
+
"grad_norm": 3.6135895252227783,
|
149 |
+
"kl": 0.0011918679920199793,
|
150 |
"learning_rate": 4.81355307410676e-07,
|
151 |
+
"loss": 0.0,
|
152 |
+
"reward": 1.9164189714938402,
|
153 |
+
"reward_std": 1.160076865926385,
|
154 |
+
"rewards/concensus_correctness_reward_func": 0.11756250262260437,
|
155 |
+
"rewards/consensus_reward_func": 0.625,
|
156 |
"rewards/cumulative_reward_2": 0.0,
|
157 |
+
"rewards/final_correctness_reward_func": 0.125,
|
158 |
+
"rewards/question_recreation_reward_func": 0.48829398211091757,
|
159 |
"rewards/soft_format_reward_func": 0.0,
|
160 |
+
"rewards/strict_format_reward_func": 0.046875,
|
161 |
+
"rewards/xmlcount_reward_func": 0.513687499333173,
|
162 |
"step": 16
|
163 |
},
|
164 |
{
|
165 |
+
"completion_length": 238.6875,
|
166 |
+
"epoch": 3.0,
|
167 |
+
"grad_norm": 2.7586710453033447,
|
168 |
+
"kl": 0.0012445114989532158,
|
169 |
"learning_rate": 4.747379352713488e-07,
|
170 |
+
"loss": 0.0,
|
171 |
+
"reward": 1.9129376076161861,
|
172 |
+
"reward_std": 0.9141782382503152,
|
173 |
+
"rewards/concensus_correctness_reward_func": 0.05999999865889549,
|
174 |
+
"rewards/consensus_reward_func": 0.5625,
|
175 |
"rewards/cumulative_reward_2": 0.0,
|
176 |
+
"rewards/final_correctness_reward_func": 0.125,
|
177 |
+
"rewards/question_recreation_reward_func": 0.5800001341849566,
|
178 |
"rewards/soft_format_reward_func": 0.0,
|
179 |
+
"rewards/strict_format_reward_func": 0.015625,
|
180 |
+
"rewards/xmlcount_reward_func": 0.5698124971240759,
|
181 |
"step": 18
|
182 |
},
|
183 |
{
|
184 |
+
"completion_length": 243.90625,
|
185 |
+
"epoch": 3.3333333333333335,
|
186 |
+
"grad_norm": 3.191770553588867,
|
187 |
+
"kl": 0.0015616379932907876,
|
188 |
"learning_rate": 4.6717793412953776e-07,
|
189 |
+
"loss": 0.0,
|
190 |
+
"reward": 1.7438320461660624,
|
191 |
+
"reward_std": 1.1812182366847992,
|
192 |
+
"rewards/concensus_correctness_reward_func": 0.1197500005364418,
|
193 |
+
"rewards/consensus_reward_func": 0.75,
|
194 |
"rewards/cumulative_reward_2": 0.0,
|
195 |
+
"rewards/final_correctness_reward_func": 0.0625,
|
196 |
+
"rewards/question_recreation_reward_func": 0.40089456969872117,
|
197 |
"rewards/soft_format_reward_func": 0.0,
|
198 |
+
"rewards/strict_format_reward_func": 0.015625,
|
199 |
+
"rewards/xmlcount_reward_func": 0.3950625052675605,
|
200 |
"step": 20
|
201 |
},
|
202 |
{
|
203 |
+
"completion_length": 295.375,
|
204 |
+
"epoch": 3.6666666666666665,
|
205 |
+
"grad_norm": 2.1044507026672363,
|
206 |
+
"kl": 0.002216623950516805,
|
207 |
"learning_rate": 4.5870701325731773e-07,
|
208 |
+
"loss": 0.0,
|
209 |
+
"reward": 2.3217686265707016,
|
210 |
+
"reward_std": 0.9629865860333666,
|
211 |
+
"rewards/concensus_correctness_reward_func": 0.12012499943375587,
|
212 |
+
"rewards/consensus_reward_func": 0.8125,
|
213 |
"rewards/cumulative_reward_2": 0.0,
|
214 |
+
"rewards/final_correctness_reward_func": 0.3125,
|
215 |
+
"rewards/question_recreation_reward_func": 0.5311123521532863,
|
216 |
"rewards/soft_format_reward_func": 0.0,
|
217 |
+
"rewards/strict_format_reward_func": 0.03125,
|
218 |
+
"rewards/xmlcount_reward_func": 0.5142812496051192,
|
219 |
"step": 22
|
220 |
},
|
221 |
{
|
222 |
+
"completion_length": 250.9375,
|
223 |
+
"epoch": 4.0,
|
224 |
+
"grad_norm": 2.842360496520996,
|
225 |
+
"kl": 0.001575836147821974,
|
226 |
"learning_rate": 4.4936070264068016e-07,
|
227 |
+
"loss": 0.0,
|
228 |
+
"reward": 1.8550556376576424,
|
229 |
+
"reward_std": 1.1311865192838013,
|
230 |
+
"rewards/concensus_correctness_reward_func": 0.11993750184774399,
|
231 |
+
"rewards/consensus_reward_func": 0.5,
|
232 |
"rewards/cumulative_reward_2": 0.0,
|
233 |
+
"rewards/final_correctness_reward_func": 0.25,
|
234 |
+
"rewards/question_recreation_reward_func": 0.6037118844687939,
|
235 |
"rewards/soft_format_reward_func": 0.0,
|
236 |
+
"rewards/strict_format_reward_func": 0.015625,
|
237 |
+
"rewards/xmlcount_reward_func": 0.36578124947845936,
|
238 |
"step": 24
|
239 |
},
|
240 |
{
|
241 |
+
"completion_length": 236.375,
|
242 |
+
"epoch": 4.333333333333333,
|
243 |
+
"grad_norm": 3.5260348320007324,
|
244 |
+
"kl": 0.0018173266253143083,
|
245 |
"learning_rate": 4.391782039544238e-07,
|
246 |
+
"loss": 0.0,
|
247 |
+
"reward": 2.022034127265215,
|
248 |
+
"reward_std": 0.9448904548771679,
|
249 |
+
"rewards/concensus_correctness_reward_func": 0.26981250010430813,
|
250 |
+
"rewards/consensus_reward_func": 0.8125,
|
251 |
"rewards/cumulative_reward_2": 0.0,
|
252 |
+
"rewards/final_correctness_reward_func": 0.1875,
|
253 |
+
"rewards/question_recreation_reward_func": 0.410690407268703,
|
254 |
"rewards/soft_format_reward_func": 0.0,
|
255 |
+
"rewards/strict_format_reward_func": 0.0,
|
256 |
+
"rewards/xmlcount_reward_func": 0.34153125202283263,
|
257 |
"step": 26
|
258 |
},
|
259 |
{
|
260 |
+
"completion_length": 222.1875,
|
261 |
+
"epoch": 4.666666666666667,
|
262 |
+
"grad_norm": 2.8555054664611816,
|
263 |
+
"kl": 0.0025639020168455318,
|
264 |
"learning_rate": 4.282022261367073e-07,
|
265 |
+
"loss": 0.0,
|
266 |
+
"reward": 1.850969212129712,
|
267 |
+
"reward_std": 0.8895461307838559,
|
268 |
+
"rewards/concensus_correctness_reward_func": 0.05999999865889549,
|
269 |
+
"rewards/consensus_reward_func": 0.625,
|
270 |
"rewards/cumulative_reward_2": 0.0,
|
271 |
+
"rewards/final_correctness_reward_func": 0.0625,
|
272 |
+
"rewards/question_recreation_reward_func": 0.5744379386305809,
|
273 |
"rewards/soft_format_reward_func": 0.0,
|
274 |
+
"rewards/strict_format_reward_func": 0.015625,
|
275 |
+
"rewards/xmlcount_reward_func": 0.5134062475990504,
|
276 |
"step": 28
|
277 |
},
|
278 |
{
|
279 |
+
"completion_length": 312.1875,
|
280 |
+
"epoch": 5.0,
|
281 |
+
"grad_norm": 2.6812455654144287,
|
282 |
+
"kl": 0.002034817836829461,
|
283 |
"learning_rate": 4.1647880625292027e-07,
|
284 |
+
"loss": 0.0,
|
285 |
+
"reward": 2.235242122784257,
|
286 |
+
"reward_std": 0.903150615748018,
|
287 |
+
"rewards/concensus_correctness_reward_func": 0.302437499165535,
|
288 |
+
"rewards/consensus_reward_func": 0.625,
|
289 |
"rewards/cumulative_reward_2": 0.0,
|
290 |
+
"rewards/final_correctness_reward_func": 0.0625,
|
291 |
+
"rewards/question_recreation_reward_func": 0.701929610222578,
|
292 |
"rewards/soft_format_reward_func": 0.0,
|
293 |
+
"rewards/strict_format_reward_func": 0.0625,
|
294 |
+
"rewards/xmlcount_reward_func": 0.48087499663233757,
|
295 |
"step": 30
|
296 |
},
|
297 |
{
|
298 |
+
"completion_length": 222.03125,
|
299 |
+
"epoch": 5.333333333333333,
|
300 |
+
"grad_norm": 3.4555599689483643,
|
301 |
+
"kl": 0.0031993754600989632,
|
302 |
"learning_rate": 4.040571164002318e-07,
|
303 |
+
"loss": 0.0,
|
304 |
+
"reward": 2.742512159049511,
|
305 |
+
"reward_std": 1.1281160314101726,
|
306 |
+
"rewards/concensus_correctness_reward_func": 0.42487499490380287,
|
307 |
+
"rewards/consensus_reward_func": 0.625,
|
308 |
"rewards/cumulative_reward_2": 0.0,
|
309 |
+
"rewards/final_correctness_reward_func": 0.375,
|
310 |
+
"rewards/question_recreation_reward_func": 0.6586684007197618,
|
311 |
"rewards/soft_format_reward_func": 0.0,
|
312 |
+
"rewards/strict_format_reward_func": 0.0625,
|
313 |
+
"rewards/xmlcount_reward_func": 0.5964687503874302,
|
314 |
"step": 32
|
315 |
},
|
316 |
{
|
317 |
+
"completion_length": 289.875,
|
318 |
+
"epoch": 5.666666666666667,
|
319 |
+
"grad_norm": 2.337428092956543,
|
320 |
+
"kl": 0.0029264501317811664,
|
321 |
"learning_rate": 3.909892574627266e-07,
|
322 |
+
"loss": 0.0,
|
323 |
+
"reward": 1.9419155293144286,
|
324 |
+
"reward_std": 1.168605322483927,
|
325 |
+
"rewards/concensus_correctness_reward_func": 0.30256250128149986,
|
326 |
+
"rewards/consensus_reward_func": 0.75,
|
327 |
"rewards/cumulative_reward_2": 0.0,
|
328 |
+
"rewards/final_correctness_reward_func": 0.0625,
|
329 |
+
"rewards/question_recreation_reward_func": 0.4774467754177749,
|
330 |
"rewards/soft_format_reward_func": 0.0,
|
331 |
+
"rewards/strict_format_reward_func": 0.03125,
|
332 |
+
"rewards/xmlcount_reward_func": 0.31815625075250864,
|
333 |
"step": 34
|
334 |
},
|
335 |
{
|
336 |
+
"completion_length": 271.75,
|
337 |
+
"epoch": 6.0,
|
338 |
+
"grad_norm": 3.06026291847229,
|
339 |
+
"kl": 0.0034197392051282804,
|
340 |
"learning_rate": 3.773300405821908e-07,
|
341 |
+
"loss": 0.0,
|
342 |
+
"reward": 2.5837525203824043,
|
343 |
+
"reward_std": 1.1307801827788353,
|
344 |
+
"rewards/concensus_correctness_reward_func": 0.29768750071525574,
|
345 |
+
"rewards/consensus_reward_func": 1.0,
|
346 |
"rewards/cumulative_reward_2": 0.0,
|
347 |
+
"rewards/final_correctness_reward_func": 0.3125,
|
348 |
+
"rewards/question_recreation_reward_func": 0.54522127751261,
|
349 |
"rewards/soft_format_reward_func": 0.0,
|
350 |
+
"rewards/strict_format_reward_func": 0.03125,
|
351 |
+
"rewards/xmlcount_reward_func": 0.3970937414560467,
|
352 |
"step": 36
|
353 |
},
|
354 |
{
|
355 |
+
"completion_length": 273.96875,
|
356 |
+
"epoch": 6.333333333333333,
|
357 |
+
"grad_norm": 6.918361186981201,
|
358 |
+
"kl": 0.0025484783691354096,
|
359 |
"learning_rate": 3.6313675726113475e-07,
|
360 |
+
"loss": 0.0,
|
361 |
+
"reward": 1.684893824160099,
|
362 |
+
"reward_std": 1.0457911072298884,
|
363 |
+
"rewards/concensus_correctness_reward_func": 0.1797499991953373,
|
364 |
+
"rewards/consensus_reward_func": 0.5625,
|
365 |
"rewards/cumulative_reward_2": 0.0,
|
366 |
+
"rewards/final_correctness_reward_func": 0.0,
|
367 |
+
"rewards/question_recreation_reward_func": 0.5752063244581223,
|
368 |
"rewards/soft_format_reward_func": 0.0,
|
369 |
+
"rewards/strict_format_reward_func": 0.0,
|
370 |
+
"rewards/xmlcount_reward_func": 0.3674375033006072,
|
371 |
"step": 38
|
372 |
},
|
373 |
{
|
374 |
+
"completion_length": 268.4375,
|
375 |
+
"epoch": 6.666666666666667,
|
376 |
+
"grad_norm": 2.9002139568328857,
|
377 |
+
"kl": 0.002866584218281787,
|
378 |
"learning_rate": 3.484689390623218e-07,
|
379 |
+
"loss": 0.0,
|
380 |
+
"reward": 1.900494983419776,
|
381 |
+
"reward_std": 1.2571525508537889,
|
382 |
+
"rewards/concensus_correctness_reward_func": 0.24512499943375587,
|
383 |
+
"rewards/consensus_reward_func": 0.625,
|
384 |
"rewards/cumulative_reward_2": 0.0,
|
385 |
+
"rewards/final_correctness_reward_func": 0.125,
|
386 |
+
"rewards/question_recreation_reward_func": 0.6020261840894818,
|
387 |
"rewards/soft_format_reward_func": 0.0,
|
388 |
+
"rewards/strict_format_reward_func": 0.0,
|
389 |
+
"rewards/xmlcount_reward_func": 0.3033437514677644,
|
390 |
"step": 40
|
391 |
},
|
392 |
{
|
393 |
+
"completion_length": 292.96875,
|
394 |
+
"epoch": 7.0,
|
395 |
+
"grad_norm": 2.595947742462158,
|
396 |
+
"kl": 0.0034651760361157358,
|
397 |
"learning_rate": 3.3338810791270517e-07,
|
398 |
+
"loss": 0.0,
|
399 |
+
"reward": 2.582839421927929,
|
400 |
+
"reward_std": 0.8570395507849753,
|
401 |
+
"rewards/concensus_correctness_reward_func": 0.42537499964237213,
|
402 |
+
"rewards/consensus_reward_func": 0.6875,
|
403 |
"rewards/cumulative_reward_2": 0.0,
|
404 |
+
"rewards/final_correctness_reward_func": 0.375,
|
405 |
+
"rewards/question_recreation_reward_func": 0.5414331587962806,
|
406 |
"rewards/soft_format_reward_func": 0.0,
|
407 |
+
"rewards/strict_format_reward_func": 0.03125,
|
408 |
+
"rewards/xmlcount_reward_func": 0.5222812481224537,
|
409 |
"step": 42
|
410 |
},
|
411 |
{
|
412 |
+
"completion_length": 277.78125,
|
413 |
+
"epoch": 7.333333333333333,
|
414 |
+
"grad_norm": 3.5444931983947754,
|
415 |
+
"kl": 0.005151846205990296,
|
416 |
"learning_rate": 3.179575180590857e-07,
|
417 |
+
"loss": 0.0,
|
418 |
+
"reward": 2.2039641477167606,
|
419 |
+
"reward_std": 1.085899718105793,
|
420 |
+
"rewards/concensus_correctness_reward_func": 0.1823749952018261,
|
421 |
+
"rewards/consensus_reward_func": 0.75,
|
422 |
"rewards/cumulative_reward_2": 0.0,
|
423 |
+
"rewards/final_correctness_reward_func": 0.1875,
|
424 |
+
"rewards/question_recreation_reward_func": 0.4583704025717452,
|
425 |
"rewards/soft_format_reward_func": 0.0,
|
426 |
+
"rewards/strict_format_reward_func": 0.015625,
|
427 |
+
"rewards/xmlcount_reward_func": 0.6100937575101852,
|
428 |
"step": 44
|
429 |
},
|
430 |
{
|
431 |
+
"completion_length": 255.9375,
|
432 |
+
"epoch": 7.666666666666667,
|
433 |
+
"grad_norm": 2.611354112625122,
|
434 |
+
"kl": 0.004227373341564089,
|
435 |
"learning_rate": 3.022418907578188e-07,
|
436 |
+
"loss": 0.0,
|
437 |
+
"reward": 1.8945876602083445,
|
438 |
+
"reward_std": 0.9345793006941676,
|
439 |
+
"rewards/concensus_correctness_reward_func": 0.18268750235438347,
|
440 |
+
"rewards/consensus_reward_func": 0.5,
|
441 |
"rewards/cumulative_reward_2": 0.0,
|
442 |
+
"rewards/final_correctness_reward_func": 0.1875,
|
443 |
+
"rewards/question_recreation_reward_func": 0.56027518119663,
|
444 |
"rewards/soft_format_reward_func": 0.0,
|
445 |
+
"rewards/strict_format_reward_func": 0.015625,
|
446 |
+
"rewards/xmlcount_reward_func": 0.4484999952837825,
|
447 |
"step": 46
|
448 |
},
|
449 |
{
|
450 |
+
"completion_length": 211.78125,
|
451 |
+
"epoch": 8.0,
|
452 |
+
"grad_norm": 7.4239501953125,
|
453 |
+
"kl": 0.005186592134123202,
|
454 |
"learning_rate": 2.863071428113726e-07,
|
455 |
+
"loss": 0.0,
|
456 |
+
"reward": 2.3263955637812614,
|
457 |
+
"reward_std": 0.9710670886561275,
|
458 |
+
"rewards/concensus_correctness_reward_func": 0.05962499976158142,
|
459 |
+
"rewards/consensus_reward_func": 0.75,
|
460 |
"rewards/cumulative_reward_2": 0.0,
|
461 |
+
"rewards/final_correctness_reward_func": 0.25,
|
462 |
+
"rewards/question_recreation_reward_func": 0.6223018001765013,
|
463 |
"rewards/soft_format_reward_func": 0.0,
|
464 |
+
"rewards/strict_format_reward_func": 0.046875,
|
465 |
+
"rewards/xmlcount_reward_func": 0.5975937452167273,
|
466 |
"step": 48
|
467 |
},
|
468 |
{
|
469 |
+
"completion_length": 237.625,
|
470 |
+
"epoch": 8.333333333333334,
|
471 |
+
"grad_norm": 3.110980272293091,
|
472 |
+
"kl": 0.005397260436438955,
|
473 |
"learning_rate": 2.7022011009035107e-07,
|
474 |
+
"loss": 0.0,
|
475 |
+
"reward": 2.3635387681424618,
|
476 |
+
"reward_std": 1.1881022350862622,
|
477 |
+
"rewards/concensus_correctness_reward_func": 0.1797499991953373,
|
478 |
+
"rewards/consensus_reward_func": 0.875,
|
479 |
"rewards/cumulative_reward_2": 0.0,
|
480 |
+
"rewards/final_correctness_reward_func": 0.1875,
|
481 |
+
"rewards/question_recreation_reward_func": 0.4961012676358223,
|
482 |
"rewards/soft_format_reward_func": 0.0,
|
483 |
+
"rewards/strict_format_reward_func": 0.015625,
|
484 |
+
"rewards/xmlcount_reward_func": 0.6095625087618828,
|
485 |
"step": 50
|
486 |
},
|
487 |
{
|
488 |
+
"completion_length": 302.375,
|
489 |
+
"epoch": 8.666666666666666,
|
490 |
+
"grad_norm": 3.359147787094116,
|
491 |
+
"kl": 0.00526040667318739,
|
492 |
"learning_rate": 2.540482672006254e-07,
|
493 |
+
"loss": 0.0,
|
494 |
+
"reward": 1.951256174594164,
|
495 |
+
"reward_std": 1.4674215724226087,
|
496 |
+
"rewards/concensus_correctness_reward_func": 0.14731250144541264,
|
497 |
+
"rewards/consensus_reward_func": 0.5625,
|
498 |
"rewards/cumulative_reward_2": 0.0,
|
499 |
+
"rewards/final_correctness_reward_func": 0.3125,
|
500 |
+
"rewards/question_recreation_reward_func": 0.5990999331697822,
|
501 |
"rewards/soft_format_reward_func": 0.0,
|
502 |
+
"rewards/strict_format_reward_func": 0.03125,
|
503 |
+
"rewards/xmlcount_reward_func": 0.29859375115484,
|
504 |
"step": 52
|
505 |
},
|
506 |
{
|
507 |
+
"completion_length": 271.34375,
|
508 |
+
"epoch": 9.0,
|
509 |
+
"grad_norm": 2.3752264976501465,
|
510 |
+
"kl": 0.006538017390994355,
|
511 |
"learning_rate": 2.37859444471388e-07,
|
512 |
+
"loss": 0.0,
|
513 |
+
"reward": 2.0346078351140022,
|
514 |
+
"reward_std": 0.7186543988646008,
|
515 |
+
"rewards/concensus_correctness_reward_func": 0.24512499943375587,
|
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": 0.5727328350767493,
|
520 |
"rewards/soft_format_reward_func": 0.0,
|
521 |
+
"rewards/strict_format_reward_func": 0.015625,
|
522 |
+
"rewards/xmlcount_reward_func": 0.5761250024661422,
|
523 |
"step": 54
|
524 |
},
|
525 |
{
|
526 |
+
"completion_length": 258.5625,
|
527 |
+
"epoch": 9.333333333333334,
|
528 |
+
"grad_norm": 2.415621519088745,
|
529 |
+
"kl": 0.0068657751398859546,
|
530 |
"learning_rate": 2.2172154345117894e-07,
|
531 |
+
"loss": 0.0,
|
532 |
+
"reward": 2.7013516426086426,
|
533 |
+
"reward_std": 1.1713198693469167,
|
534 |
+
"rewards/concensus_correctness_reward_func": 0.17987500131130219,
|
535 |
+
"rewards/consensus_reward_func": 1.0625,
|
536 |
"rewards/cumulative_reward_2": 0.0,
|
537 |
+
"rewards/final_correctness_reward_func": 0.25,
|
538 |
+
"rewards/question_recreation_reward_func": 0.5597578389570117,
|
539 |
"rewards/soft_format_reward_func": 0.0,
|
540 |
+
"rewards/strict_format_reward_func": 0.0625,
|
541 |
+
"rewards/xmlcount_reward_func": 0.5867187539115548,
|
542 |
"step": 56
|
543 |
},
|
544 |
{
|
545 |
+
"completion_length": 250.4375,
|
546 |
+
"epoch": 9.666666666666666,
|
547 |
+
"grad_norm": 3.3267598152160645,
|
548 |
+
"kl": 0.007265938082127832,
|
549 |
"learning_rate": 2.0570225210519433e-07,
|
550 |
+
"loss": 0.0,
|
551 |
+
"reward": 1.7129965675994754,
|
552 |
+
"reward_std": 0.9574196808971465,
|
553 |
+
"rewards/concensus_correctness_reward_func": 0.24262499809265137,
|
554 |
+
"rewards/consensus_reward_func": 0.3125,
|
555 |
"rewards/cumulative_reward_2": 0.0,
|
556 |
+
"rewards/final_correctness_reward_func": 0.125,
|
557 |
+
"rewards/question_recreation_reward_func": 0.5204653043183498,
|
558 |
"rewards/soft_format_reward_func": 0.0,
|
559 |
+
"rewards/strict_format_reward_func": 0.03125,
|
560 |
+
"rewards/xmlcount_reward_func": 0.4811562584945932,
|
561 |
"step": 58
|
562 |
},
|
563 |
{
|
564 |
+
"completion_length": 294.5625,
|
565 |
+
"epoch": 10.0,
|
566 |
+
"grad_norm": 3.0382344722747803,
|
567 |
+
"kl": 0.007241527950100135,
|
568 |
"learning_rate": 1.8986876090843664e-07,
|
569 |
+
"loss": 0.0,
|
570 |
+
"reward": 1.9392794030718505,
|
571 |
+
"reward_std": 1.0110920977240312,
|
572 |
+
"rewards/concensus_correctness_reward_func": 0.11993750184774399,
|
573 |
+
"rewards/consensus_reward_func": 0.8125,
|
574 |
"rewards/cumulative_reward_2": 0.0,
|
575 |
+
"rewards/final_correctness_reward_func": 0.0625,
|
576 |
+
"rewards/question_recreation_reward_func": 0.500998193398118,
|
577 |
"rewards/soft_format_reward_func": 0.0,
|
578 |
+
"rewards/strict_format_reward_func": 0.015625,
|
579 |
+
"rewards/xmlcount_reward_func": 0.42771875113248825,
|
580 |
"step": 60
|
581 |
},
|
582 |
{
|
583 |
+
"completion_length": 222.125,
|
584 |
+
"epoch": 10.333333333333334,
|
585 |
+
"grad_norm": 3.4102041721343994,
|
586 |
+
"kl": 0.00738686349359341,
|
587 |
"learning_rate": 1.7428748102551234e-07,
|
588 |
+
"loss": 0.0,
|
589 |
+
"reward": 1.7328537330031395,
|
590 |
+
"reward_std": 0.9404821525095031,
|
591 |
+
"rewards/concensus_correctness_reward_func": 0.1223749965429306,
|
592 |
+
"rewards/consensus_reward_func": 0.5625,
|
593 |
"rewards/cumulative_reward_2": 0.0,
|
594 |
+
"rewards/final_correctness_reward_func": 0.0,
|
595 |
+
"rewards/question_recreation_reward_func": 0.45538498694077134,
|
596 |
"rewards/soft_format_reward_func": 0.0,
|
597 |
+
"rewards/strict_format_reward_func": 0.03125,
|
598 |
+
"rewards/xmlcount_reward_func": 0.5613437537103891,
|
599 |
"step": 62
|
600 |
},
|
601 |
{
|
602 |
+
"completion_length": 322.46875,
|
603 |
+
"epoch": 10.666666666666666,
|
604 |
+
"grad_norm": 2.5114853382110596,
|
605 |
+
"kl": 0.005600389951723628,
|
606 |
"learning_rate": 1.5902376575912814e-07,
|
607 |
+
"loss": 0.0,
|
608 |
+
"reward": 2.345261871814728,
|
609 |
+
"reward_std": 1.4083554986864328,
|
610 |
+
"rewards/concensus_correctness_reward_func": 0.35993750020861626,
|
611 |
+
"rewards/consensus_reward_func": 0.6875,
|
612 |
"rewards/cumulative_reward_2": 0.0,
|
613 |
+
"rewards/final_correctness_reward_func": 0.25,
|
614 |
+
"rewards/question_recreation_reward_func": 0.621855610050261,
|
615 |
"rewards/soft_format_reward_func": 0.0,
|
616 |
+
"rewards/strict_format_reward_func": 0.015625,
|
617 |
+
"rewards/xmlcount_reward_func": 0.4103437541052699,
|
618 |
"step": 64
|
619 |
},
|
620 |
{
|
621 |
+
"completion_length": 238.65625,
|
622 |
+
"epoch": 11.0,
|
623 |
+
"grad_norm": 2.935357093811035,
|
624 |
+
"kl": 0.00783520688128192,
|
625 |
"learning_rate": 1.4414163643562753e-07,
|
626 |
+
"loss": 0.0,
|
627 |
+
"reward": 2.1366525441408157,
|
628 |
+
"reward_std": 1.0672186479205266,
|
629 |
+
"rewards/concensus_correctness_reward_func": 0.11743750050663948,
|
630 |
+
"rewards/consensus_reward_func": 0.5625,
|
631 |
"rewards/cumulative_reward_2": 0.0,
|
632 |
+
"rewards/final_correctness_reward_func": 0.1875,
|
633 |
+
"rewards/question_recreation_reward_func": 0.588996303267777,
|
634 |
"rewards/soft_format_reward_func": 0.0,
|
635 |
+
"rewards/strict_format_reward_func": 0.0,
|
636 |
+
"rewards/xmlcount_reward_func": 0.6802187506109476,
|
637 |
"step": 66
|
638 |
},
|
639 |
{
|
640 |
+
"completion_length": 266.90625,
|
641 |
+
"epoch": 11.333333333333334,
|
642 |
+
"grad_norm": 2.8767664432525635,
|
643 |
+
"kl": 0.006433373549953103,
|
644 |
"learning_rate": 1.2970351387729872e-07,
|
645 |
+
"loss": 0.0,
|
646 |
+
"reward": 1.6148663735948503,
|
647 |
+
"reward_std": 1.3225993164815009,
|
648 |
+
"rewards/concensus_correctness_reward_func": 0.20812500081956387,
|
649 |
+
"rewards/consensus_reward_func": 0.4375,
|
650 |
"rewards/cumulative_reward_2": 0.0,
|
651 |
+
"rewards/final_correctness_reward_func": 0.0625,
|
652 |
+
"rewards/question_recreation_reward_func": 0.5250538997352123,
|
653 |
"rewards/soft_format_reward_func": 0.0,
|
654 |
+
"rewards/strict_format_reward_func": 0.03125,
|
655 |
+
"rewards/xmlcount_reward_func": 0.3504374986514449,
|
656 |
"step": 68
|
657 |
},
|
658 |
{
|
659 |
+
"completion_length": 250.21875,
|
660 |
+
"epoch": 11.666666666666666,
|
661 |
+
"grad_norm": 4.7036638259887695,
|
662 |
+
"kl": 0.007804472814314067,
|
663 |
"learning_rate": 1.1576995658775404e-07,
|
664 |
+
"loss": 0.0,
|
665 |
+
"reward": 2.2799468226730824,
|
666 |
+
"reward_std": 1.166436342522502,
|
667 |
+
"rewards/concensus_correctness_reward_func": 0.30262500047683716,
|
668 |
+
"rewards/consensus_reward_func": 0.6875,
|
669 |
+
"rewards/cumulative_reward_2": 0.0,
|
670 |
+
"rewards/final_correctness_reward_func": 0.1875,
|
671 |
+
"rewards/question_recreation_reward_func": 0.5988530963659286,
|
672 |
+
"rewards/soft_format_reward_func": 0.015625,
|
673 |
+
"rewards/strict_format_reward_func": 0.015625,
|
674 |
+
"rewards/xmlcount_reward_func": 0.47221875097602606,
|
675 |
"step": 70
|
676 |
},
|
677 |
{
|
678 |
+
"completion_length": 248.9375,
|
679 |
+
"epoch": 12.0,
|
680 |
+
"grad_norm": 2.7038285732269287,
|
681 |
+
"kl": 0.00650077304453589,
|
682 |
"learning_rate": 1.0239940674851941e-07,
|
683 |
+
"loss": 0.0,
|
684 |
+
"reward": 1.7189936628565192,
|
685 |
+
"reward_std": 1.1028610868379474,
|
686 |
+
"rewards/concensus_correctness_reward_func": 0.11999999731779099,
|
687 |
+
"rewards/consensus_reward_func": 0.5625,
|
688 |
"rewards/cumulative_reward_2": 0.0,
|
689 |
+
"rewards/final_correctness_reward_func": 0.125,
|
690 |
+
"rewards/question_recreation_reward_func": 0.4415561552159488,
|
691 |
"rewards/soft_format_reward_func": 0.0,
|
692 |
+
"rewards/strict_format_reward_func": 0.015625,
|
693 |
+
"rewards/xmlcount_reward_func": 0.4543124968186021,
|
694 |
"step": 72
|
695 |
},
|
696 |
{
|
697 |
+
"completion_length": 241.9375,
|
698 |
+
"epoch": 12.333333333333334,
|
699 |
+
"grad_norm": 3.6059765815734863,
|
700 |
+
"kl": 0.007852518698200583,
|
701 |
"learning_rate": 8.964794509221507e-08,
|
702 |
+
"loss": 0.0,
|
703 |
+
"reward": 1.7060329876840115,
|
704 |
+
"reward_std": 1.2234434876590967,
|
705 |
+
"rewards/concensus_correctness_reward_func": 0.05999999865889549,
|
706 |
+
"rewards/consensus_reward_func": 0.6875,
|
707 |
"rewards/cumulative_reward_2": 0.0,
|
708 |
+
"rewards/final_correctness_reward_func": 0.0625,
|
709 |
+
"rewards/question_recreation_reward_func": 0.4156580241397023,
|
710 |
"rewards/soft_format_reward_func": 0.0,
|
711 |
+
"rewards/strict_format_reward_func": 0.03125,
|
712 |
+
"rewards/xmlcount_reward_func": 0.44912500213831663,
|
713 |
"step": 74
|
714 |
},
|
715 |
{
|
716 |
+
"completion_length": 286.28125,
|
717 |
+
"epoch": 12.666666666666666,
|
718 |
+
"grad_norm": 2.444171190261841,
|
719 |
+
"kl": 0.00814603897742927,
|
720 |
"learning_rate": 7.756905568047392e-08,
|
721 |
+
"loss": 0.0,
|
722 |
+
"reward": 2.122521236538887,
|
723 |
+
"reward_std": 0.9578379703452811,
|
724 |
+
"rewards/concensus_correctness_reward_func": 0.42274999618530273,
|
725 |
+
"rewards/consensus_reward_func": 0.3125,
|
726 |
"rewards/cumulative_reward_2": 0.0,
|
727 |
+
"rewards/final_correctness_reward_func": 0.25,
|
728 |
+
"rewards/question_recreation_reward_func": 0.504114959621802,
|
729 |
"rewards/soft_format_reward_func": 0.0,
|
730 |
+
"rewards/strict_format_reward_func": 0.015625,
|
731 |
+
"rewards/xmlcount_reward_func": 0.6175312478444539,
|
732 |
"step": 76
|
733 |
},
|
734 |
{
|
735 |
+
"completion_length": 234.6875,
|
736 |
+
"epoch": 13.0,
|
737 |
+
"grad_norm": 3.0294716358184814,
|
738 |
+
"kl": 0.008441783153102733,
|
739 |
"learning_rate": 6.621340157319996e-08,
|
740 |
+
"loss": 0.0,
|
741 |
+
"reward": 2.081810735166073,
|
742 |
+
"reward_std": 0.9143577099894173,
|
743 |
+
"rewards/concensus_correctness_reward_func": 0.125,
|
744 |
+
"rewards/consensus_reward_func": 0.8125,
|
745 |
"rewards/cumulative_reward_2": 0.0,
|
746 |
+
"rewards/final_correctness_reward_func": 0.25,
|
747 |
+
"rewards/question_recreation_reward_func": 0.5170919904485345,
|
748 |
"rewards/soft_format_reward_func": 0.0,
|
749 |
+
"rewards/strict_format_reward_func": 0.0,
|
750 |
+
"rewards/xmlcount_reward_func": 0.37721875379793346,
|
751 |
"step": 78
|
752 |
},
|
753 |
{
|
754 |
+
"completion_length": 224.40625,
|
755 |
+
"epoch": 13.333333333333334,
|
756 |
+
"grad_norm": 3.4747705459594727,
|
757 |
+
"kl": 0.01025624800240621,
|
758 |
"learning_rate": 5.5628612330087724e-08,
|
759 |
+
"loss": 0.0,
|
760 |
+
"reward": 1.9557546898722649,
|
761 |
+
"reward_std": 0.9802739145234227,
|
762 |
+
"rewards/concensus_correctness_reward_func": 0.24450000002980232,
|
763 |
+
"rewards/consensus_reward_func": 0.5625,
|
764 |
"rewards/cumulative_reward_2": 0.0,
|
765 |
+
"rewards/final_correctness_reward_func": 0.0625,
|
766 |
+
"rewards/question_recreation_reward_func": 0.562535947188735,
|
767 |
"rewards/soft_format_reward_func": 0.0,
|
768 |
+
"rewards/strict_format_reward_func": 0.03125,
|
769 |
+
"rewards/xmlcount_reward_func": 0.4924687468446791,
|
770 |
"step": 80
|
771 |
},
|
772 |
{
|
773 |
+
"completion_length": 234.75,
|
774 |
+
"epoch": 13.666666666666666,
|
775 |
+
"grad_norm": 2.7190427780151367,
|
776 |
+
"kl": 0.008365696892724372,
|
777 |
"learning_rate": 4.5859084235697235e-08,
|
778 |
+
"loss": 0.0,
|
779 |
+
"reward": 1.6996353343129158,
|
780 |
+
"reward_std": 0.6642087557120249,
|
781 |
+
"rewards/concensus_correctness_reward_func": 0.05999999865889549,
|
782 |
+
"rewards/consensus_reward_func": 0.5625,
|
783 |
"rewards/cumulative_reward_2": 0.0,
|
784 |
+
"rewards/final_correctness_reward_func": 0.125,
|
785 |
+
"rewards/question_recreation_reward_func": 0.395666572265327,
|
786 |
"rewards/soft_format_reward_func": 0.0,
|
787 |
+
"rewards/strict_format_reward_func": 0.046875,
|
788 |
+
"rewards/xmlcount_reward_func": 0.5095937536098063,
|
789 |
"step": 82
|
790 |
},
|
791 |
{
|
792 |
+
"completion_length": 263.96875,
|
793 |
+
"epoch": 14.0,
|
794 |
+
"grad_norm": 2.663147449493408,
|
795 |
+
"kl": 0.010309035773389041,
|
796 |
"learning_rate": 3.6945794086007705e-08,
|
797 |
+
"loss": 0.0,
|
798 |
+
"reward": 2.0628331936895847,
|
799 |
+
"reward_std": 1.0619396911934018,
|
800 |
+
"rewards/concensus_correctness_reward_func": 0.17756250128149986,
|
801 |
+
"rewards/consensus_reward_func": 0.6875,
|
802 |
"rewards/cumulative_reward_2": 0.0,
|
803 |
+
"rewards/final_correctness_reward_func": 0.125,
|
804 |
+
"rewards/question_recreation_reward_func": 0.5063332146964967,
|
805 |
"rewards/soft_format_reward_func": 0.0,
|
806 |
+
"rewards/strict_format_reward_func": 0.046875,
|
807 |
+
"rewards/xmlcount_reward_func": 0.5195625014603138,
|
808 |
"step": 84
|
809 |
},
|
810 |
{
|
811 |
+
"completion_length": 231.1875,
|
812 |
+
"epoch": 14.333333333333334,
|
813 |
+
"grad_norm": 3.154367685317993,
|
814 |
+
"kl": 0.01306345833290834,
|
815 |
"learning_rate": 2.892612731749414e-08,
|
816 |
+
"loss": 0.0,
|
817 |
+
"reward": 2.1949321180582047,
|
818 |
+
"reward_std": 1.4137339515145868,
|
819 |
+
"rewards/concensus_correctness_reward_func": 0.18025000020861626,
|
820 |
+
"rewards/consensus_reward_func": 0.8125,
|
821 |
"rewards/cumulative_reward_2": 0.0,
|
822 |
+
"rewards/final_correctness_reward_func": 0.125,
|
823 |
+
"rewards/question_recreation_reward_func": 0.49090082419570535,
|
824 |
"rewards/soft_format_reward_func": 0.0,
|
825 |
+
"rewards/strict_format_reward_func": 0.03125,
|
826 |
+
"rewards/xmlcount_reward_func": 0.5550312586128712,
|
827 |
"step": 86
|
828 |
},
|
829 |
{
|
830 |
+
"completion_length": 315.9375,
|
831 |
+
"epoch": 14.666666666666666,
|
832 |
+
"grad_norm": 2.249851703643799,
|
833 |
+
"kl": 0.006411108406609856,
|
834 |
"learning_rate": 2.183372119961499e-08,
|
835 |
+
"loss": 0.0,
|
836 |
+
"reward": 2.2250705547630787,
|
837 |
+
"reward_std": 1.5259972442872822,
|
838 |
+
"rewards/concensus_correctness_reward_func": 0.39806249737739563,
|
839 |
+
"rewards/consensus_reward_func": 0.625,
|
840 |
"rewards/cumulative_reward_2": 0.0,
|
841 |
+
"rewards/final_correctness_reward_func": 0.3125,
|
842 |
+
"rewards/question_recreation_reward_func": 0.5338830836117268,
|
843 |
"rewards/soft_format_reward_func": 0.0,
|
844 |
+
"rewards/strict_format_reward_func": 0.046875,
|
845 |
+
"rewards/xmlcount_reward_func": 0.3087499989196658,
|
846 |
"step": 88
|
847 |
},
|
848 |
{
|
849 |
+
"completion_length": 261.09375,
|
850 |
+
"epoch": 15.0,
|
851 |
+
"grad_norm": 2.743177652359009,
|
852 |
+
"kl": 0.00881043096887879,
|
853 |
"learning_rate": 1.5698323748414122e-08,
|
854 |
+
"loss": 0.0,
|
855 |
+
"reward": 1.895367180928588,
|
856 |
+
"reward_std": 1.2597154856775887,
|
857 |
+
"rewards/concensus_correctness_reward_func": 0.18012499809265137,
|
858 |
+
"rewards/consensus_reward_func": 0.625,
|
859 |
"rewards/cumulative_reward_2": 0.0,
|
860 |
+
"rewards/final_correctness_reward_func": 0.125,
|
861 |
+
"rewards/question_recreation_reward_func": 0.5660547083243728,
|
862 |
"rewards/soft_format_reward_func": 0.0,
|
863 |
+
"rewards/strict_format_reward_func": 0.046875,
|
864 |
+
"rewards/xmlcount_reward_func": 0.35231249686330557,
|
865 |
"step": 90
|
866 |
},
|
867 |
{
|
868 |
+
"completion_length": 277.90625,
|
869 |
+
"epoch": 15.333333333333334,
|
870 |
+
"grad_norm": 3.3928024768829346,
|
871 |
+
"kl": 0.006757803043001331,
|
872 |
"learning_rate": 1.054566895300324e-08,
|
873 |
+
"loss": 0.0,
|
874 |
+
"reward": 1.7474502064287663,
|
875 |
+
"reward_std": 0.7529050658922642,
|
876 |
+
"rewards/concensus_correctness_reward_func": 0.1798749975860119,
|
877 |
+
"rewards/consensus_reward_func": 0.5,
|
878 |
"rewards/cumulative_reward_2": 0.0,
|
879 |
+
"rewards/final_correctness_reward_func": 0.1875,
|
880 |
+
"rewards/question_recreation_reward_func": 0.5421689683571458,
|
881 |
"rewards/soft_format_reward_func": 0.0,
|
882 |
+
"rewards/strict_format_reward_func": 0.0,
|
883 |
+
"rewards/xmlcount_reward_func": 0.337906246073544,
|
884 |
"step": 92
|
885 |
},
|
886 |
{
|
887 |
+
"completion_length": 304.09375,
|
888 |
+
"epoch": 15.666666666666666,
|
889 |
+
"grad_norm": 2.197697162628174,
|
890 |
+
"kl": 0.006906057009473443,
|
891 |
"learning_rate": 6.397368838268496e-09,
|
892 |
+
"loss": 0.0,
|
893 |
+
"reward": 1.6495681330561638,
|
894 |
+
"reward_std": 0.6983778811700176,
|
895 |
+
"rewards/concensus_correctness_reward_func": 0.06012500077486038,
|
896 |
+
"rewards/consensus_reward_func": 0.4375,
|
897 |
"rewards/cumulative_reward_2": 0.0,
|
898 |
+
"rewards/final_correctness_reward_func": 0.125,
|
899 |
+
"rewards/question_recreation_reward_func": 0.5284118615090847,
|
900 |
"rewards/soft_format_reward_func": 0.0,
|
901 |
+
"rewards/strict_format_reward_func": 0.015625,
|
902 |
+
"rewards/xmlcount_reward_func": 0.48290624911896884,
|
903 |
"step": 94
|
904 |
},
|
905 |
{
|
906 |
+
"completion_length": 269.59375,
|
907 |
+
"epoch": 16.0,
|
908 |
+
"grad_norm": 2.6010444164276123,
|
909 |
+
"kl": 0.011520398882566951,
|
910 |
"learning_rate": 3.2708228165273244e-09,
|
911 |
+
"loss": 0.0,
|
912 |
+
"reward": 2.7417864352464676,
|
913 |
+
"reward_std": 1.6308739269152284,
|
914 |
+
"rewards/concensus_correctness_reward_func": 0.30268750339746475,
|
915 |
+
"rewards/consensus_reward_func": 1.0625,
|
916 |
"rewards/cumulative_reward_2": 0.0,
|
917 |
+
"rewards/final_correctness_reward_func": 0.3125,
|
918 |
+
"rewards/question_recreation_reward_func": 0.5818177200853825,
|
919 |
"rewards/soft_format_reward_func": 0.0,
|
920 |
+
"rewards/strict_format_reward_func": 0.046875,
|
921 |
+
"rewards/xmlcount_reward_func": 0.4354062539059669,
|
922 |
"step": 96
|
923 |
},
|
924 |
{
|
925 |
+
"completion_length": 256.625,
|
926 |
+
"epoch": 16.333333333333332,
|
927 |
+
"grad_norm": 3.064824104309082,
|
928 |
+
"kl": 0.013104435958666727,
|
929 |
"learning_rate": 1.1791447083465133e-09,
|
930 |
+
"loss": 0.0,
|
931 |
+
"reward": 2.3778811804950237,
|
932 |
+
"reward_std": 1.2875230926729273,
|
933 |
+
"rewards/concensus_correctness_reward_func": 0.18262499943375587,
|
934 |
+
"rewards/consensus_reward_func": 0.8125,
|
935 |
"rewards/cumulative_reward_2": 0.0,
|
936 |
+
"rewards/final_correctness_reward_func": 0.1875,
|
937 |
+
"rewards/question_recreation_reward_func": 0.44909992604516447,
|
938 |
"rewards/soft_format_reward_func": 0.0,
|
939 |
+
"rewards/strict_format_reward_func": 0.0625,
|
940 |
+
"rewards/xmlcount_reward_func": 0.6836562557145953,
|
941 |
"step": 98
|
942 |
},
|
943 |
{
|
944 |
+
"completion_length": 251.875,
|
945 |
+
"epoch": 16.666666666666668,
|
946 |
+
"grad_norm": 5.960361957550049,
|
947 |
+
"kl": 0.014634610328357667,
|
948 |
"learning_rate": 1.3110773862126667e-10,
|
949 |
+
"loss": 0.0,
|
950 |
+
"reward": 1.9424380734562874,
|
951 |
+
"reward_std": 1.420625472906977,
|
952 |
+
"rewards/concensus_correctness_reward_func": 0.24249999597668648,
|
953 |
+
"rewards/consensus_reward_func": 0.625,
|
954 |
"rewards/cumulative_reward_2": 0.0,
|
955 |
+
"rewards/final_correctness_reward_func": 0.125,
|
956 |
+
"rewards/question_recreation_reward_func": 0.38812559354119003,
|
957 |
"rewards/soft_format_reward_func": 0.0,
|
958 |
+
"rewards/strict_format_reward_func": 0.046875,
|
959 |
+
"rewards/xmlcount_reward_func": 0.5149374911561608,
|
960 |
"step": 100
|
961 |
},
|
962 |
{
|
963 |
+
"epoch": 16.666666666666668,
|
964 |
"step": 100,
|
965 |
"total_flos": 0.0,
|
966 |
+
"train_loss": 5.246203290880657e-06,
|
967 |
+
"train_runtime": 3714.3127,
|
968 |
+
"train_samples_per_second": 0.431,
|
969 |
+
"train_steps_per_second": 0.027
|
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 sha256:929dae0ec6f304f24eac715a1449fdddd810343375eb432e5ee668c04ef01240
|
3 |
size 5944
|