hugodk-sch
commited on
Commit
•
191c1a5
1
Parent(s):
d514169
Model save
Browse files- README.md +13 -16
- adapter_model.safetensors +1 -1
- all_results.json +2 -15
- train_results.json +2 -2
- trainer_state.json +372 -372
README.md
CHANGED
@@ -1,13 +1,10 @@
|
|
1 |
---
|
2 |
library_name: peft
|
3 |
tags:
|
4 |
-
- alignment-handbook
|
5 |
- trl
|
6 |
- dpo
|
7 |
- generated_from_trainer
|
8 |
base_model: NbAiLab/nb-gpt-j-6B-v2
|
9 |
-
datasets:
|
10 |
-
- hugodk-sch/aftonposten_title_prefs
|
11 |
model-index:
|
12 |
- name: aftonposten-6b-align-scan
|
13 |
results: []
|
@@ -18,17 +15,17 @@ should probably proofread and complete it, then remove this comment. -->
|
|
18 |
|
19 |
# aftonposten-6b-align-scan
|
20 |
|
21 |
-
This model is a fine-tuned version of [
|
22 |
It achieves the following results on the evaluation set:
|
23 |
-
- Loss: 0.
|
24 |
-
- Rewards/chosen: 0.
|
25 |
-
- Rewards/rejected: 0.
|
26 |
-
- Rewards/accuracies: 0.
|
27 |
-
- Rewards/margins: 0.
|
28 |
-
- Logps/rejected: -37.
|
29 |
-
- Logps/chosen: -34.
|
30 |
-
- Logits/rejected: -2.
|
31 |
-
- Logits/chosen: -2.
|
32 |
|
33 |
## Model description
|
34 |
|
@@ -63,9 +60,9 @@ The following hyperparameters were used during training:
|
|
63 |
|
64 |
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
65 |
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
66 |
-
| 0.
|
67 |
-
| 0.
|
68 |
-
| 0.
|
69 |
|
70 |
|
71 |
### Framework versions
|
|
|
1 |
---
|
2 |
library_name: peft
|
3 |
tags:
|
|
|
4 |
- trl
|
5 |
- dpo
|
6 |
- generated_from_trainer
|
7 |
base_model: NbAiLab/nb-gpt-j-6B-v2
|
|
|
|
|
8 |
model-index:
|
9 |
- name: aftonposten-6b-align-scan
|
10 |
results: []
|
|
|
15 |
|
16 |
# aftonposten-6b-align-scan
|
17 |
|
18 |
+
This model is a fine-tuned version of [NbAiLab/nb-gpt-j-6B-v2](https://huggingface.co/NbAiLab/nb-gpt-j-6B-v2) on an unknown dataset.
|
19 |
It achieves the following results on the evaluation set:
|
20 |
+
- Loss: 0.6949
|
21 |
+
- Rewards/chosen: -0.0012
|
22 |
+
- Rewards/rejected: 0.0013
|
23 |
+
- Rewards/accuracies: 0.5162
|
24 |
+
- Rewards/margins: -0.0026
|
25 |
+
- Logps/rejected: -37.5122
|
26 |
+
- Logps/chosen: -34.0387
|
27 |
+
- Logits/rejected: -2.2399
|
28 |
+
- Logits/chosen: -2.2448
|
29 |
|
30 |
## Model description
|
31 |
|
|
|
60 |
|
61 |
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
62 |
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
63 |
+
| 0.6886 | 0.26 | 100 | 0.6950 | 0.0021 | 0.0049 | 0.4817 | -0.0028 | -37.5001 | -34.0275 | -2.2393 | -2.2442 |
|
64 |
+
| 0.6909 | 0.52 | 200 | 0.6952 | 0.0005 | 0.0038 | 0.5191 | -0.0033 | -37.5039 | -34.0328 | -2.2391 | -2.2439 |
|
65 |
+
| 0.6869 | 0.78 | 300 | 0.6949 | -0.0012 | 0.0013 | 0.5162 | -0.0026 | -37.5122 | -34.0387 | -2.2399 | -2.2448 |
|
66 |
|
67 |
|
68 |
### Framework versions
|
adapter_model.safetensors
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 176183216
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:a1c41272bfc20de1787ca81e4ccbee9baeb8b772de90ef90ec9ab66107b6cf9f
|
3 |
size 176183216
|
all_results.json
CHANGED
@@ -1,20 +1,7 @@
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
-
"
|
4 |
-
"
|
5 |
-
"eval_logps/chosen": -34.02475357055664,
|
6 |
-
"eval_logps/rejected": -37.51246643066406,
|
7 |
-
"eval_loss": 0.6927995681762695,
|
8 |
-
"eval_rewards/accuracies": 0.5099667906761169,
|
9 |
-
"eval_rewards/chosen": 0.0019583876710385084,
|
10 |
-
"eval_rewards/margins": 0.0011271554976701736,
|
11 |
-
"eval_rewards/rejected": 0.0008312325226143003,
|
12 |
-
"eval_runtime": 145.6648,
|
13 |
-
"eval_samples": 343,
|
14 |
-
"eval_samples_per_second": 2.355,
|
15 |
-
"eval_steps_per_second": 0.295,
|
16 |
-
"train_loss": 0.6919509800997647,
|
17 |
-
"train_runtime": 3253.2141,
|
18 |
"train_samples": 3079,
|
19 |
"train_samples_per_second": 0.946,
|
20 |
"train_steps_per_second": 0.118
|
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
+
"train_loss": 0.691356255791404,
|
4 |
+
"train_runtime": 3254.7157,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
5 |
"train_samples": 3079,
|
6 |
"train_samples_per_second": 0.946,
|
7 |
"train_steps_per_second": 0.118
|
train_results.json
CHANGED
@@ -1,7 +1,7 @@
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
-
"train_loss": 0.
|
4 |
-
"train_runtime":
|
5 |
"train_samples": 3079,
|
6 |
"train_samples_per_second": 0.946,
|
7 |
"train_steps_per_second": 0.118
|
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
+
"train_loss": 0.691356255791404,
|
4 |
+
"train_runtime": 3254.7157,
|
5 |
"train_samples": 3079,
|
6 |
"train_samples_per_second": 0.946,
|
7 |
"train_steps_per_second": 0.118
|
trainer_state.json
CHANGED
@@ -25,589 +25,589 @@
|
|
25 |
{
|
26 |
"epoch": 0.03,
|
27 |
"learning_rate": 1.2820512820512818e-07,
|
28 |
-
"logits/chosen": -1.
|
29 |
-
"logits/rejected": -1.
|
30 |
-
"logps/chosen": -36.
|
31 |
-
"logps/rejected": -33.
|
32 |
-
"loss": 0.
|
33 |
-
"rewards/accuracies": 0.
|
34 |
-
"rewards/chosen": 0.
|
35 |
-
"rewards/margins": 0.
|
36 |
-
"rewards/rejected": -0.
|
37 |
"step": 10
|
38 |
},
|
39 |
{
|
40 |
"epoch": 0.05,
|
41 |
"learning_rate": 2.5641025641025636e-07,
|
42 |
-
"logits/chosen": -1.
|
43 |
-
"logits/rejected": -
|
44 |
-
"logps/chosen": -29.
|
45 |
-
"logps/rejected": -29.
|
46 |
-
"loss": 0.
|
47 |
-
"rewards/accuracies": 0.
|
48 |
-
"rewards/chosen": 0.
|
49 |
-
"rewards/margins": 0.
|
50 |
-
"rewards/rejected": 0.
|
51 |
"step": 20
|
52 |
},
|
53 |
{
|
54 |
"epoch": 0.08,
|
55 |
"learning_rate": 3.8461538461538463e-07,
|
56 |
-
"logits/chosen": -1.
|
57 |
-
"logits/rejected": -1.
|
58 |
-
"logps/chosen": -31.
|
59 |
-
"logps/rejected": -33.
|
60 |
-
"loss": 0.
|
61 |
-
"rewards/accuracies": 0.
|
62 |
-
"rewards/chosen": 0.
|
63 |
-
"rewards/margins":
|
64 |
-
"rewards/rejected": 0.
|
65 |
"step": 30
|
66 |
},
|
67 |
{
|
68 |
"epoch": 0.1,
|
69 |
"learning_rate": 4.999896948438433e-07,
|
70 |
-
"logits/chosen": -2.
|
71 |
-
"logits/rejected": -2.
|
72 |
-
"logps/chosen": -32.
|
73 |
-
"logps/rejected": -32.
|
74 |
-
"loss": 0.
|
75 |
-
"rewards/accuracies": 0.
|
76 |
-
"rewards/chosen": -0.
|
77 |
-
"rewards/margins": -0.
|
78 |
-
"rewards/rejected": 0.
|
79 |
"step": 40
|
80 |
},
|
81 |
{
|
82 |
"epoch": 0.13,
|
83 |
"learning_rate": 4.987541037542186e-07,
|
84 |
-
"logits/chosen": -1.
|
85 |
-
"logits/rejected": -1.
|
86 |
-
"logps/chosen": -33.
|
87 |
-
"logps/rejected": -35.
|
88 |
-
"loss": 0.
|
89 |
-
"rewards/accuracies": 0.
|
90 |
-
"rewards/chosen":
|
91 |
-
"rewards/margins": -0.
|
92 |
-
"rewards/rejected": 0.
|
93 |
"step": 50
|
94 |
},
|
95 |
{
|
96 |
"epoch": 0.16,
|
97 |
"learning_rate": 4.954691471941118e-07,
|
98 |
-
"logits/chosen": -1.
|
99 |
-
"logits/rejected": -1.
|
100 |
-
"logps/chosen": -32.
|
101 |
-
"logps/rejected": -33.
|
102 |
-
"loss": 0.
|
103 |
-
"rewards/accuracies": 0.
|
104 |
-
"rewards/chosen": 0.
|
105 |
-
"rewards/margins": 0.
|
106 |
-
"rewards/rejected": -0.
|
107 |
"step": 60
|
108 |
},
|
109 |
{
|
110 |
"epoch": 0.18,
|
111 |
"learning_rate": 4.901618883413548e-07,
|
112 |
-
"logits/chosen": -2.
|
113 |
-
"logits/rejected": -2.
|
114 |
-
"logps/chosen": -33.
|
115 |
-
"logps/rejected": -36.
|
116 |
-
"loss": 0.
|
117 |
-
"rewards/accuracies": 0.
|
118 |
-
"rewards/chosen":
|
119 |
-
"rewards/margins": 0.
|
120 |
-
"rewards/rejected": -0.
|
121 |
"step": 70
|
122 |
},
|
123 |
{
|
124 |
"epoch": 0.21,
|
125 |
"learning_rate": 4.828760511501322e-07,
|
126 |
-
"logits/chosen": -1.
|
127 |
-
"logits/rejected": -1.
|
128 |
-
"logps/chosen": -34.
|
129 |
-
"logps/rejected": -34.
|
130 |
-
"loss": 0.
|
131 |
-
"rewards/accuracies": 0.
|
132 |
-
"rewards/chosen": 0.
|
133 |
-
"rewards/margins": 0.
|
134 |
-
"rewards/rejected": 0.
|
135 |
"step": 80
|
136 |
},
|
137 |
{
|
138 |
"epoch": 0.23,
|
139 |
"learning_rate": 4.736716601303429e-07,
|
140 |
-
"logits/chosen": -1.
|
141 |
-
"logits/rejected": -1.
|
142 |
-
"logps/chosen": -32.
|
143 |
-
"logps/rejected": -32.
|
144 |
-
"loss": 0.
|
145 |
-
"rewards/accuracies": 0.
|
146 |
-
"rewards/chosen": 0.
|
147 |
-
"rewards/margins": 0.
|
148 |
-
"rewards/rejected": -0.
|
149 |
"step": 90
|
150 |
},
|
151 |
{
|
152 |
"epoch": 0.26,
|
153 |
"learning_rate": 4.62624545834521e-07,
|
154 |
-
"logits/chosen": -2.
|
155 |
-
"logits/rejected": -2.
|
156 |
-
"logps/chosen": -32.
|
157 |
-
"logps/rejected": -31.
|
158 |
-
"loss": 0.
|
159 |
-
"rewards/accuracies": 0.
|
160 |
-
"rewards/chosen": 0.
|
161 |
-
"rewards/margins": 0.
|
162 |
-
"rewards/rejected": -0.
|
163 |
"step": 100
|
164 |
},
|
165 |
{
|
166 |
"epoch": 0.26,
|
167 |
-
"eval_logits/chosen": -2.
|
168 |
-
"eval_logits/rejected": -2.
|
169 |
-
"eval_logps/chosen": -34.
|
170 |
-
"eval_logps/rejected": -37.
|
171 |
-
"eval_loss": 0.
|
172 |
-
"eval_rewards/accuracies": 0.
|
173 |
-
"eval_rewards/chosen":
|
174 |
-
"eval_rewards/margins": -0.
|
175 |
-
"eval_rewards/rejected":
|
176 |
-
"eval_runtime": 146.
|
177 |
"eval_samples_per_second": 2.349,
|
178 |
-
"eval_steps_per_second": 0.
|
179 |
"step": 100
|
180 |
},
|
181 |
{
|
182 |
"epoch": 0.29,
|
183 |
"learning_rate": 4.4982572012636904e-07,
|
184 |
-
"logits/chosen": -2.
|
185 |
-
"logits/rejected": -2.
|
186 |
-
"logps/chosen": -33.
|
187 |
-
"logps/rejected": -34.
|
188 |
-
"loss": 0.
|
189 |
-
"rewards/accuracies": 0.
|
190 |
-
"rewards/chosen": 0.
|
191 |
-
"rewards/margins": 0.
|
192 |
-
"rewards/rejected": 0.
|
193 |
"step": 110
|
194 |
},
|
195 |
{
|
196 |
"epoch": 0.31,
|
197 |
"learning_rate": 4.353806263777677e-07,
|
198 |
-
"logits/chosen": -2.
|
199 |
-
"logits/rejected": -2.
|
200 |
-
"logps/chosen": -32.
|
201 |
-
"logps/rejected": -32.
|
202 |
-
"loss": 0.
|
203 |
-
"rewards/accuracies": 0.
|
204 |
-
"rewards/chosen":
|
205 |
-
"rewards/margins": 0.
|
206 |
-
"rewards/rejected": -0.
|
207 |
"step": 120
|
208 |
},
|
209 |
{
|
210 |
"epoch": 0.34,
|
211 |
"learning_rate": 4.194082707715275e-07,
|
212 |
-
"logits/chosen": -2.
|
213 |
-
"logits/rejected": -2.
|
214 |
-
"logps/chosen": -30.
|
215 |
-
"logps/rejected": -32.
|
216 |
-
"loss": 0.
|
217 |
-
"rewards/accuracies": 0.
|
218 |
-
"rewards/chosen": -0.
|
219 |
-
"rewards/margins": -0.
|
220 |
-
"rewards/rejected": 0.
|
221 |
"step": 130
|
222 |
},
|
223 |
{
|
224 |
"epoch": 0.36,
|
225 |
"learning_rate": 4.020402418666621e-07,
|
226 |
-
"logits/chosen": -1.
|
227 |
-
"logits/rejected": -1.
|
228 |
-
"logps/chosen": -31.
|
229 |
-
"logps/rejected": -32.
|
230 |
-
"loss": 0.
|
231 |
-
"rewards/accuracies": 0.
|
232 |
-
"rewards/chosen": 0.
|
233 |
-
"rewards/margins": 0.
|
234 |
-
"rewards/rejected": -0.
|
235 |
"step": 140
|
236 |
},
|
237 |
{
|
238 |
"epoch": 0.39,
|
239 |
"learning_rate": 3.8341962650351185e-07,
|
240 |
-
"logits/chosen": -1.
|
241 |
-
"logits/rejected": -1.
|
242 |
-
"logps/chosen": -34.
|
243 |
-
"logps/rejected": -34.
|
244 |
-
"loss": 0.
|
245 |
-
"rewards/accuracies": 0.
|
246 |
-
"rewards/chosen": 0.
|
247 |
-
"rewards/margins": 0.
|
248 |
-
"rewards/rejected":
|
249 |
"step": 150
|
250 |
},
|
251 |
{
|
252 |
"epoch": 0.42,
|
253 |
"learning_rate": 3.636998309800572e-07,
|
254 |
-
"logits/chosen": -1.
|
255 |
-
"logits/rejected": -1.
|
256 |
-
"logps/chosen": -36.
|
257 |
-
"logps/rejected": -32.
|
258 |
-
"loss": 0.
|
259 |
-
"rewards/accuracies": 0.
|
260 |
-
"rewards/chosen": 0.
|
261 |
-
"rewards/margins": 0.
|
262 |
-
"rewards/rejected": -0.
|
263 |
"step": 160
|
264 |
},
|
265 |
{
|
266 |
"epoch": 0.44,
|
267 |
"learning_rate": 3.430433172111807e-07,
|
268 |
-
"logits/chosen": -2.
|
269 |
-
"logits/rejected": -2.
|
270 |
-
"logps/chosen": -33.
|
271 |
-
"logps/rejected": -31.
|
272 |
-
"loss": 0.
|
273 |
"rewards/accuracies": 0.5249999761581421,
|
274 |
-
"rewards/chosen": 0.
|
275 |
-
"rewards/margins": 0.
|
276 |
-
"rewards/rejected":
|
277 |
"step": 170
|
278 |
},
|
279 |
{
|
280 |
"epoch": 0.47,
|
281 |
"learning_rate": 3.216202642830543e-07,
|
282 |
-
"logits/chosen": -2.
|
283 |
-
"logits/rejected": -2.
|
284 |
-
"logps/chosen": -32.
|
285 |
-
"logps/rejected": -32.
|
286 |
-
"loss": 0.
|
287 |
-
"rewards/accuracies": 0.
|
288 |
-
"rewards/chosen": 0.
|
289 |
-
"rewards/margins": 0.
|
290 |
-
"rewards/rejected": -0.
|
291 |
"step": 180
|
292 |
},
|
293 |
{
|
294 |
"epoch": 0.49,
|
295 |
"learning_rate": 2.9960716642946403e-07,
|
296 |
-
"logits/chosen": -2.
|
297 |
-
"logits/rejected": -2.
|
298 |
-
"logps/chosen": -31.
|
299 |
-
"logps/rejected": -31.
|
300 |
-
"loss": 0.
|
301 |
-
"rewards/accuracies": 0.
|
302 |
-
"rewards/chosen": -0.
|
303 |
-
"rewards/margins":
|
304 |
-
"rewards/rejected": -0.
|
305 |
"step": 190
|
306 |
},
|
307 |
{
|
308 |
"epoch": 0.52,
|
309 |
"learning_rate": 2.771853789806683e-07,
|
310 |
-
"logits/chosen": -1.
|
311 |
-
"logits/rejected": -1.
|
312 |
-
"logps/chosen": -31.
|
313 |
-
"logps/rejected": -32.
|
314 |
-
"loss": 0.
|
315 |
-
"rewards/accuracies": 0.
|
316 |
-
"rewards/chosen": 0.
|
317 |
-
"rewards/margins": 0.
|
318 |
-
"rewards/rejected": -0.
|
319 |
"step": 200
|
320 |
},
|
321 |
{
|
322 |
"epoch": 0.52,
|
323 |
-
"eval_logits/chosen": -2.
|
324 |
-
"eval_logits/rejected": -2.
|
325 |
-
"eval_logps/chosen": -34.
|
326 |
-
"eval_logps/rejected": -37.
|
327 |
-
"eval_loss": 0.
|
328 |
-
"eval_rewards/accuracies": 0.
|
329 |
-
"eval_rewards/chosen": 0.
|
330 |
-
"eval_rewards/margins": 0.
|
331 |
-
"eval_rewards/rejected": 0.
|
332 |
-
"eval_runtime": 145.
|
333 |
-
"eval_samples_per_second": 2.
|
334 |
"eval_steps_per_second": 0.295,
|
335 |
"step": 200
|
336 |
},
|
337 |
{
|
338 |
"epoch": 0.55,
|
339 |
"learning_rate": 2.5453962426402e-07,
|
340 |
-
"logits/chosen": -2.
|
341 |
-
"logits/rejected": -2.
|
342 |
-
"logps/chosen": -31.
|
343 |
-
"logps/rejected": -33.
|
344 |
-
"loss": 0.
|
345 |
-
"rewards/accuracies": 0.
|
346 |
-
"rewards/chosen": 0.
|
347 |
-
"rewards/margins": 0.
|
348 |
-
"rewards/rejected": -
|
349 |
"step": 210
|
350 |
},
|
351 |
{
|
352 |
"epoch": 0.57,
|
353 |
"learning_rate": 2.318564697655179e-07,
|
354 |
-
"logits/chosen": -1.
|
355 |
-
"logits/rejected": -1.
|
356 |
-
"logps/chosen": -30.
|
357 |
-
"logps/rejected": -31.
|
358 |
-
"loss": 0.
|
359 |
-
"rewards/accuracies": 0.
|
360 |
-
"rewards/chosen": 0.
|
361 |
-
"rewards/margins": 0.
|
362 |
-
"rewards/rejected": -0.
|
363 |
"step": 220
|
364 |
},
|
365 |
{
|
366 |
"epoch": 0.6,
|
367 |
"learning_rate": 2.093227910899832e-07,
|
368 |
-
"logits/chosen": -1.
|
369 |
-
"logits/rejected": -1.
|
370 |
-
"logps/chosen": -33.
|
371 |
-
"logps/rejected": -31.
|
372 |
-
"loss": 0.
|
373 |
-
"rewards/accuracies": 0.
|
374 |
-
"rewards/chosen": 0.
|
375 |
-
"rewards/margins": 0.
|
376 |
-
"rewards/rejected":
|
377 |
"step": 230
|
378 |
},
|
379 |
{
|
380 |
"epoch": 0.62,
|
381 |
"learning_rate": 1.8712423238279356e-07,
|
382 |
-
"logits/chosen": -1.
|
383 |
-
"logits/rejected": -1.
|
384 |
-
"logps/chosen": -34.
|
385 |
-
"logps/rejected": -34.
|
386 |
-
"loss": 0.
|
387 |
"rewards/accuracies": 0.5,
|
388 |
-
"rewards/chosen": -
|
389 |
-
"rewards/margins": -
|
390 |
-
"rewards/rejected":
|
391 |
"step": 240
|
392 |
},
|
393 |
{
|
394 |
"epoch": 0.65,
|
395 |
"learning_rate": 1.654436768970182e-07,
|
396 |
-
"logits/chosen": -2.
|
397 |
-
"logits/rejected": -2.
|
398 |
-
"logps/chosen": -32.
|
399 |
-
"logps/rejected": -36.
|
400 |
-
"loss": 0.
|
401 |
-
"rewards/accuracies": 0.
|
402 |
-
"rewards/chosen": 0.
|
403 |
-
"rewards/margins": 0.
|
404 |
-
"rewards/rejected":
|
405 |
"step": 250
|
406 |
},
|
407 |
{
|
408 |
"epoch": 0.68,
|
409 |
"learning_rate": 1.444597403062196e-07,
|
410 |
-
"logits/chosen": -1.
|
411 |
-
"logits/rejected": -1.
|
412 |
-
"logps/chosen": -34.
|
413 |
-
"logps/rejected": -35.
|
414 |
-
"loss": 0.
|
415 |
"rewards/accuracies": 0.5874999761581421,
|
416 |
-
"rewards/chosen": 0.
|
417 |
-
"rewards/margins": 0.
|
418 |
-
"rewards/rejected": -0.
|
419 |
"step": 260
|
420 |
},
|
421 |
{
|
422 |
"epoch": 0.7,
|
423 |
"learning_rate": 1.2434529917578887e-07,
|
424 |
-
"logits/chosen": -1.
|
425 |
-
"logits/rejected": -1.
|
426 |
-
"logps/chosen": -34.
|
427 |
-
"logps/rejected": -31.
|
428 |
-
"loss": 0.
|
429 |
-
"rewards/accuracies": 0.
|
430 |
-
"rewards/chosen":
|
431 |
-
"rewards/margins":
|
432 |
-
"rewards/rejected": 0.
|
433 |
"step": 270
|
434 |
},
|
435 |
{
|
436 |
"epoch": 0.73,
|
437 |
"learning_rate": 1.0526606671603521e-07,
|
438 |
-
"logits/chosen": -1.
|
439 |
-
"logits/rejected": -1.
|
440 |
-
"logps/chosen": -35.
|
441 |
-
"logps/rejected": -31.
|
442 |
-
"loss": 0.
|
443 |
-
"rewards/accuracies": 0.
|
444 |
-
"rewards/chosen": 0.
|
445 |
-
"rewards/margins": 0.
|
446 |
-
"rewards/rejected": -0.
|
447 |
"step": 280
|
448 |
},
|
449 |
{
|
450 |
"epoch": 0.75,
|
451 |
"learning_rate": 8.737922755071453e-08,
|
452 |
-
"logits/chosen": -2.
|
453 |
-
"logits/rejected": -2.
|
454 |
-
"logps/chosen": -30.
|
455 |
-
"logps/rejected": -32.
|
456 |
-
"loss": 0.
|
457 |
-
"rewards/accuracies": 0.
|
458 |
-
"rewards/chosen":
|
459 |
-
"rewards/margins":
|
460 |
-
"rewards/rejected": 0.
|
461 |
"step": 290
|
462 |
},
|
463 |
{
|
464 |
"epoch": 0.78,
|
465 |
"learning_rate": 7.08321427484816e-08,
|
466 |
-
"logits/chosen": -1.
|
467 |
-
"logits/rejected": -1.
|
468 |
-
"logps/chosen": -32.
|
469 |
-
"logps/rejected": -30.
|
470 |
-
"loss": 0.
|
471 |
-
"rewards/accuracies": 0.
|
472 |
-
"rewards/chosen": 0.
|
473 |
-
"rewards/margins": 0.
|
474 |
-
"rewards/rejected": -0.
|
475 |
"step": 300
|
476 |
},
|
477 |
{
|
478 |
"epoch": 0.78,
|
479 |
-
"eval_logits/chosen": -2.
|
480 |
-
"eval_logits/rejected": -2.
|
481 |
-
"eval_logps/chosen": -34.
|
482 |
-
"eval_logps/rejected": -37.
|
483 |
-
"eval_loss": 0.
|
484 |
-
"eval_rewards/accuracies": 0.
|
485 |
-
"eval_rewards/chosen": 0.
|
486 |
-
"eval_rewards/margins": -0.
|
487 |
-
"eval_rewards/rejected": 0.
|
488 |
-
"eval_runtime": 145.
|
489 |
-
"eval_samples_per_second": 2.
|
490 |
"eval_steps_per_second": 0.295,
|
491 |
"step": 300
|
492 |
},
|
493 |
{
|
494 |
"epoch": 0.81,
|
495 |
"learning_rate": 5.576113578589034e-08,
|
496 |
-
"logits/chosen": -1.
|
497 |
-
"logits/rejected": -1.
|
498 |
-
"logps/chosen": -31.
|
499 |
-
"logps/rejected": -33.
|
500 |
-
"loss": 0.
|
501 |
-
"rewards/accuracies": 0.
|
502 |
-
"rewards/chosen": 0.
|
503 |
-
"rewards/margins": 0.
|
504 |
-
"rewards/rejected":
|
505 |
"step": 310
|
506 |
},
|
507 |
{
|
508 |
"epoch": 0.83,
|
509 |
"learning_rate": 4.229036944380912e-08,
|
510 |
-
"logits/chosen": -1.
|
511 |
-
"logits/rejected": -1.
|
512 |
-
"logps/chosen": -34.
|
513 |
-
"logps/rejected": -33.
|
514 |
-
"loss": 0.
|
515 |
-
"rewards/accuracies": 0.
|
516 |
-
"rewards/chosen": 0.
|
517 |
-
"rewards/margins": 0.
|
518 |
-
"rewards/rejected": -0.
|
519 |
"step": 320
|
520 |
},
|
521 |
{
|
522 |
"epoch": 0.86,
|
523 |
"learning_rate": 3.053082288996112e-08,
|
524 |
-
"logits/chosen": -2.
|
525 |
-
"logits/rejected": -2.
|
526 |
-
"logps/chosen": -33.
|
527 |
-
"logps/rejected": -32.
|
528 |
-
"loss": 0.
|
529 |
"rewards/accuracies": 0.512499988079071,
|
530 |
-
"rewards/chosen": -0.
|
531 |
-
"rewards/margins": 0.
|
532 |
-
"rewards/rejected": -0.
|
533 |
"step": 330
|
534 |
},
|
535 |
{
|
536 |
"epoch": 0.88,
|
537 |
"learning_rate": 2.05793773749158e-08,
|
538 |
-
"logits/chosen": -2.
|
539 |
-
"logits/rejected": -2.
|
540 |
-
"logps/chosen": -34.
|
541 |
-
"logps/rejected": -33.
|
542 |
-
"loss": 0.
|
543 |
-
"rewards/accuracies": 0.
|
544 |
-
"rewards/chosen": 0.
|
545 |
-
"rewards/margins": -0.
|
546 |
-
"rewards/rejected": 0.
|
547 |
"step": 340
|
548 |
},
|
549 |
{
|
550 |
"epoch": 0.91,
|
551 |
"learning_rate": 1.251801807404168e-08,
|
552 |
-
"logits/chosen": -1.
|
553 |
-
"logits/rejected": -1.
|
554 |
-
"logps/chosen": -33.
|
555 |
-
"logps/rejected": -32.
|
556 |
-
"loss": 0.
|
557 |
"rewards/accuracies": 0.5,
|
558 |
-
"rewards/chosen": 0.
|
559 |
-
"rewards/margins": 0.
|
560 |
-
"rewards/rejected": 0.
|
561 |
"step": 350
|
562 |
},
|
563 |
{
|
564 |
"epoch": 0.94,
|
565 |
"learning_rate": 6.41315865106129e-09,
|
566 |
-
"logits/chosen": -1.
|
567 |
-
"logits/rejected": -1.
|
568 |
-
"logps/chosen": -32.
|
569 |
-
"logps/rejected": -35.
|
570 |
-
"loss": 0.
|
571 |
-
"rewards/accuracies": 0.
|
572 |
-
"rewards/chosen": -0.
|
573 |
-
"rewards/margins":
|
574 |
-
"rewards/rejected": 0.
|
575 |
"step": 360
|
576 |
},
|
577 |
{
|
578 |
"epoch": 0.96,
|
579 |
"learning_rate": 2.3150941078050324e-09,
|
580 |
-
"logits/chosen": -2.
|
581 |
-
"logits/rejected": -2.
|
582 |
-
"logps/chosen": -33.
|
583 |
-
"logps/rejected": -29.
|
584 |
-
"loss": 0.
|
585 |
-
"rewards/accuracies": 0.
|
586 |
-
"rewards/chosen": 0.
|
587 |
-
"rewards/margins": 0.
|
588 |
-
"rewards/rejected": 0.
|
589 |
"step": 370
|
590 |
},
|
591 |
{
|
592 |
"epoch": 0.99,
|
593 |
"learning_rate": 2.575864278703266e-10,
|
594 |
-
"logits/chosen": -1.
|
595 |
-
"logits/rejected": -1.
|
596 |
-
"logps/chosen": -34.
|
597 |
-
"logps/rejected": -30.
|
598 |
-
"loss": 0.
|
599 |
-
"rewards/accuracies": 0.
|
600 |
-
"rewards/chosen": 0.
|
601 |
-
"rewards/margins": 0.
|
602 |
-
"rewards/rejected": -0.
|
603 |
"step": 380
|
604 |
},
|
605 |
{
|
606 |
"epoch": 1.0,
|
607 |
"step": 385,
|
608 |
"total_flos": 0.0,
|
609 |
-
"train_loss": 0.
|
610 |
-
"train_runtime":
|
611 |
"train_samples_per_second": 0.946,
|
612 |
"train_steps_per_second": 0.118
|
613 |
}
|
|
|
25 |
{
|
26 |
"epoch": 0.03,
|
27 |
"learning_rate": 1.2820512820512818e-07,
|
28 |
+
"logits/chosen": -1.866300106048584,
|
29 |
+
"logits/rejected": -1.870603084564209,
|
30 |
+
"logps/chosen": -36.973060607910156,
|
31 |
+
"logps/rejected": -33.65780258178711,
|
32 |
+
"loss": 0.6843,
|
33 |
+
"rewards/accuracies": 0.5138888955116272,
|
34 |
+
"rewards/chosen": 0.010061310604214668,
|
35 |
+
"rewards/margins": 0.018380172550678253,
|
36 |
+
"rewards/rejected": -0.008318860083818436,
|
37 |
"step": 10
|
38 |
},
|
39 |
{
|
40 |
"epoch": 0.05,
|
41 |
"learning_rate": 2.5641025641025636e-07,
|
42 |
+
"logits/chosen": -1.9976227283477783,
|
43 |
+
"logits/rejected": -2.0002670288085938,
|
44 |
+
"logps/chosen": -29.652048110961914,
|
45 |
+
"logps/rejected": -29.051151275634766,
|
46 |
+
"loss": 0.6969,
|
47 |
+
"rewards/accuracies": 0.375,
|
48 |
+
"rewards/chosen": -0.0029565312433987856,
|
49 |
+
"rewards/margins": -0.006644421722739935,
|
50 |
+
"rewards/rejected": 0.003687891410663724,
|
51 |
"step": 20
|
52 |
},
|
53 |
{
|
54 |
"epoch": 0.08,
|
55 |
"learning_rate": 3.8461538461538463e-07,
|
56 |
+
"logits/chosen": -1.9202871322631836,
|
57 |
+
"logits/rejected": -1.9175989627838135,
|
58 |
+
"logps/chosen": -31.403018951416016,
|
59 |
+
"logps/rejected": -33.22191619873047,
|
60 |
+
"loss": 0.6911,
|
61 |
+
"rewards/accuracies": 0.5625,
|
62 |
+
"rewards/chosen": 0.0039227609522640705,
|
63 |
+
"rewards/margins": 0.004958181641995907,
|
64 |
+
"rewards/rejected": -0.0010354205733165145,
|
65 |
"step": 30
|
66 |
},
|
67 |
{
|
68 |
"epoch": 0.1,
|
69 |
"learning_rate": 4.999896948438433e-07,
|
70 |
+
"logits/chosen": -2.016842842102051,
|
71 |
+
"logits/rejected": -2.008082151412964,
|
72 |
+
"logps/chosen": -32.58789825439453,
|
73 |
+
"logps/rejected": -32.494510650634766,
|
74 |
+
"loss": 0.6978,
|
75 |
+
"rewards/accuracies": 0.44999998807907104,
|
76 |
+
"rewards/chosen": -0.003327915444970131,
|
77 |
+
"rewards/margins": -0.008775560185313225,
|
78 |
+
"rewards/rejected": 0.005447645206004381,
|
79 |
"step": 40
|
80 |
},
|
81 |
{
|
82 |
"epoch": 0.13,
|
83 |
"learning_rate": 4.987541037542186e-07,
|
84 |
+
"logits/chosen": -1.8639650344848633,
|
85 |
+
"logits/rejected": -1.8531978130340576,
|
86 |
+
"logps/chosen": -33.55598831176758,
|
87 |
+
"logps/rejected": -35.437034606933594,
|
88 |
+
"loss": 0.6956,
|
89 |
+
"rewards/accuracies": 0.4375,
|
90 |
+
"rewards/chosen": 0.0004471995052881539,
|
91 |
+
"rewards/margins": -0.004145150538533926,
|
92 |
+
"rewards/rejected": 0.00459235068410635,
|
93 |
"step": 50
|
94 |
},
|
95 |
{
|
96 |
"epoch": 0.16,
|
97 |
"learning_rate": 4.954691471941118e-07,
|
98 |
+
"logits/chosen": -1.9451839923858643,
|
99 |
+
"logits/rejected": -1.94712233543396,
|
100 |
+
"logps/chosen": -32.599273681640625,
|
101 |
+
"logps/rejected": -33.205684661865234,
|
102 |
+
"loss": 0.6882,
|
103 |
+
"rewards/accuracies": 0.574999988079071,
|
104 |
+
"rewards/chosen": 0.000273245939752087,
|
105 |
+
"rewards/margins": 0.010566400364041328,
|
106 |
+
"rewards/rejected": -0.010293153114616871,
|
107 |
"step": 60
|
108 |
},
|
109 |
{
|
110 |
"epoch": 0.18,
|
111 |
"learning_rate": 4.901618883413548e-07,
|
112 |
+
"logits/chosen": -2.0796267986297607,
|
113 |
+
"logits/rejected": -2.0845961570739746,
|
114 |
+
"logps/chosen": -33.98247528076172,
|
115 |
+
"logps/rejected": -36.595970153808594,
|
116 |
+
"loss": 0.6887,
|
117 |
+
"rewards/accuracies": 0.5874999761581421,
|
118 |
+
"rewards/chosen": 0.0018027916084975004,
|
119 |
+
"rewards/margins": 0.009932766668498516,
|
120 |
+
"rewards/rejected": -0.008129975758492947,
|
121 |
"step": 70
|
122 |
},
|
123 |
{
|
124 |
"epoch": 0.21,
|
125 |
"learning_rate": 4.828760511501322e-07,
|
126 |
+
"logits/chosen": -1.9422763586044312,
|
127 |
+
"logits/rejected": -1.9454386234283447,
|
128 |
+
"logps/chosen": -34.4161491394043,
|
129 |
+
"logps/rejected": -34.59540939331055,
|
130 |
+
"loss": 0.6915,
|
131 |
+
"rewards/accuracies": 0.512499988079071,
|
132 |
+
"rewards/chosen": 0.0034980489872395992,
|
133 |
+
"rewards/margins": 0.004399437457323074,
|
134 |
+
"rewards/rejected": -0.0009013883536681533,
|
135 |
"step": 80
|
136 |
},
|
137 |
{
|
138 |
"epoch": 0.23,
|
139 |
"learning_rate": 4.736716601303429e-07,
|
140 |
+
"logits/chosen": -1.950871467590332,
|
141 |
+
"logits/rejected": -1.9553781747817993,
|
142 |
+
"logps/chosen": -32.455718994140625,
|
143 |
+
"logps/rejected": -32.37923812866211,
|
144 |
+
"loss": 0.6888,
|
145 |
+
"rewards/accuracies": 0.5625,
|
146 |
+
"rewards/chosen": 0.003510843263939023,
|
147 |
+
"rewards/margins": 0.009610405191779137,
|
148 |
+
"rewards/rejected": -0.00609956169500947,
|
149 |
"step": 90
|
150 |
},
|
151 |
{
|
152 |
"epoch": 0.26,
|
153 |
"learning_rate": 4.62624545834521e-07,
|
154 |
+
"logits/chosen": -2.0493884086608887,
|
155 |
+
"logits/rejected": -2.0473878383636475,
|
156 |
+
"logps/chosen": -32.24263381958008,
|
157 |
+
"logps/rejected": -31.301403045654297,
|
158 |
+
"loss": 0.6886,
|
159 |
+
"rewards/accuracies": 0.4625000059604645,
|
160 |
+
"rewards/chosen": -0.0005389736033976078,
|
161 |
+
"rewards/margins": 0.010018542408943176,
|
162 |
+
"rewards/rejected": -0.010557514615356922,
|
163 |
"step": 100
|
164 |
},
|
165 |
{
|
166 |
"epoch": 0.26,
|
167 |
+
"eval_logits/chosen": -2.244236707687378,
|
168 |
+
"eval_logits/rejected": -2.2393476963043213,
|
169 |
+
"eval_logps/chosen": -34.027488708496094,
|
170 |
+
"eval_logps/rejected": -37.500125885009766,
|
171 |
+
"eval_loss": 0.6949825882911682,
|
172 |
+
"eval_rewards/accuracies": 0.48172760009765625,
|
173 |
+
"eval_rewards/chosen": 0.0021188759710639715,
|
174 |
+
"eval_rewards/margins": -0.0028304157312959433,
|
175 |
+
"eval_rewards/rejected": 0.0049492912366986275,
|
176 |
+
"eval_runtime": 146.0369,
|
177 |
"eval_samples_per_second": 2.349,
|
178 |
+
"eval_steps_per_second": 0.294,
|
179 |
"step": 100
|
180 |
},
|
181 |
{
|
182 |
"epoch": 0.29,
|
183 |
"learning_rate": 4.4982572012636904e-07,
|
184 |
+
"logits/chosen": -2.005858898162842,
|
185 |
+
"logits/rejected": -2.0034613609313965,
|
186 |
+
"logps/chosen": -33.22675323486328,
|
187 |
+
"logps/rejected": -34.024375915527344,
|
188 |
+
"loss": 0.6918,
|
189 |
+
"rewards/accuracies": 0.44999998807907104,
|
190 |
+
"rewards/chosen": 0.0051184119656682014,
|
191 |
+
"rewards/margins": 0.0036206122022122145,
|
192 |
+
"rewards/rejected": 0.0014977991813793778,
|
193 |
"step": 110
|
194 |
},
|
195 |
{
|
196 |
"epoch": 0.31,
|
197 |
"learning_rate": 4.353806263777677e-07,
|
198 |
+
"logits/chosen": -2.0165371894836426,
|
199 |
+
"logits/rejected": -2.0081915855407715,
|
200 |
+
"logps/chosen": -32.43837356567383,
|
201 |
+
"logps/rejected": -32.1854362487793,
|
202 |
+
"loss": 0.689,
|
203 |
+
"rewards/accuracies": 0.512499988079071,
|
204 |
+
"rewards/chosen": 0.0019600242376327515,
|
205 |
+
"rewards/margins": 0.009139470756053925,
|
206 |
+
"rewards/rejected": -0.0071794455870985985,
|
207 |
"step": 120
|
208 |
},
|
209 |
{
|
210 |
"epoch": 0.34,
|
211 |
"learning_rate": 4.194082707715275e-07,
|
212 |
+
"logits/chosen": -2.046520709991455,
|
213 |
+
"logits/rejected": -2.0384721755981445,
|
214 |
+
"logps/chosen": -30.485082626342773,
|
215 |
+
"logps/rejected": -32.0345344543457,
|
216 |
+
"loss": 0.6976,
|
217 |
+
"rewards/accuracies": 0.38749998807907104,
|
218 |
+
"rewards/chosen": -0.002810357604175806,
|
219 |
+
"rewards/margins": -0.008352747187018394,
|
220 |
+
"rewards/rejected": 0.005542389117181301,
|
221 |
"step": 130
|
222 |
},
|
223 |
{
|
224 |
"epoch": 0.36,
|
225 |
"learning_rate": 4.020402418666621e-07,
|
226 |
+
"logits/chosen": -1.976843237876892,
|
227 |
+
"logits/rejected": -1.9871089458465576,
|
228 |
+
"logps/chosen": -31.4160099029541,
|
229 |
+
"logps/rejected": -32.535850524902344,
|
230 |
+
"loss": 0.6929,
|
231 |
+
"rewards/accuracies": 0.5249999761581421,
|
232 |
+
"rewards/chosen": 0.0010961029911413789,
|
233 |
+
"rewards/margins": 0.0013871133560314775,
|
234 |
+
"rewards/rejected": -0.0002910103357862681,
|
235 |
"step": 140
|
236 |
},
|
237 |
{
|
238 |
"epoch": 0.39,
|
239 |
"learning_rate": 3.8341962650351185e-07,
|
240 |
+
"logits/chosen": -1.890434980392456,
|
241 |
+
"logits/rejected": -1.8915313482284546,
|
242 |
+
"logps/chosen": -34.165489196777344,
|
243 |
+
"logps/rejected": -34.74839782714844,
|
244 |
+
"loss": 0.691,
|
245 |
+
"rewards/accuracies": 0.5375000238418579,
|
246 |
+
"rewards/chosen": 0.008160250261425972,
|
247 |
+
"rewards/margins": 0.005502746906131506,
|
248 |
+
"rewards/rejected": 0.0026575028896331787,
|
249 |
"step": 150
|
250 |
},
|
251 |
{
|
252 |
"epoch": 0.42,
|
253 |
"learning_rate": 3.636998309800572e-07,
|
254 |
+
"logits/chosen": -1.942751169204712,
|
255 |
+
"logits/rejected": -1.9392738342285156,
|
256 |
+
"logps/chosen": -36.15436553955078,
|
257 |
+
"logps/rejected": -32.734371185302734,
|
258 |
+
"loss": 0.6893,
|
259 |
+
"rewards/accuracies": 0.512499988079071,
|
260 |
+
"rewards/chosen": 0.005194104742258787,
|
261 |
+
"rewards/margins": 0.008680562488734722,
|
262 |
+
"rewards/rejected": -0.0034864575136452913,
|
263 |
"step": 160
|
264 |
},
|
265 |
{
|
266 |
"epoch": 0.44,
|
267 |
"learning_rate": 3.430433172111807e-07,
|
268 |
+
"logits/chosen": -2.042160749435425,
|
269 |
+
"logits/rejected": -2.0347483158111572,
|
270 |
+
"logps/chosen": -33.785484313964844,
|
271 |
+
"logps/rejected": -31.344274520874023,
|
272 |
+
"loss": 0.6928,
|
273 |
"rewards/accuracies": 0.5249999761581421,
|
274 |
+
"rewards/chosen": 0.003714872058480978,
|
275 |
+
"rewards/margins": 0.0014427527785301208,
|
276 |
+
"rewards/rejected": 0.002272119279950857,
|
277 |
"step": 170
|
278 |
},
|
279 |
{
|
280 |
"epoch": 0.47,
|
281 |
"learning_rate": 3.216202642830543e-07,
|
282 |
+
"logits/chosen": -2.0473780632019043,
|
283 |
+
"logits/rejected": -2.0526490211486816,
|
284 |
+
"logps/chosen": -32.52016830444336,
|
285 |
+
"logps/rejected": -32.495323181152344,
|
286 |
+
"loss": 0.6883,
|
287 |
+
"rewards/accuracies": 0.574999988079071,
|
288 |
+
"rewards/chosen": 0.007385374512523413,
|
289 |
+
"rewards/margins": 0.010440578684210777,
|
290 |
+
"rewards/rejected": -0.0030552041716873646,
|
291 |
"step": 180
|
292 |
},
|
293 |
{
|
294 |
"epoch": 0.49,
|
295 |
"learning_rate": 2.9960716642946403e-07,
|
296 |
+
"logits/chosen": -2.0484580993652344,
|
297 |
+
"logits/rejected": -2.045682191848755,
|
298 |
+
"logps/chosen": -31.508682250976562,
|
299 |
+
"logps/rejected": -31.3348388671875,
|
300 |
+
"loss": 0.694,
|
301 |
+
"rewards/accuracies": 0.4749999940395355,
|
302 |
+
"rewards/chosen": -0.005918826907873154,
|
303 |
+
"rewards/margins": -0.0009324293350800872,
|
304 |
+
"rewards/rejected": -0.004986398387700319,
|
305 |
"step": 190
|
306 |
},
|
307 |
{
|
308 |
"epoch": 0.52,
|
309 |
"learning_rate": 2.771853789806683e-07,
|
310 |
+
"logits/chosen": -1.9187612533569336,
|
311 |
+
"logits/rejected": -1.9234310388565063,
|
312 |
+
"logps/chosen": -31.59661865234375,
|
313 |
+
"logps/rejected": -32.798248291015625,
|
314 |
+
"loss": 0.6909,
|
315 |
+
"rewards/accuracies": 0.574999988079071,
|
316 |
+
"rewards/chosen": 0.002899137092754245,
|
317 |
+
"rewards/margins": 0.00509423715993762,
|
318 |
+
"rewards/rejected": -0.002195100300014019,
|
319 |
"step": 200
|
320 |
},
|
321 |
{
|
322 |
"epoch": 0.52,
|
323 |
+
"eval_logits/chosen": -2.2439279556274414,
|
324 |
+
"eval_logits/rejected": -2.239055633544922,
|
325 |
+
"eval_logps/chosen": -34.03278350830078,
|
326 |
+
"eval_logps/rejected": -37.50389099121094,
|
327 |
+
"eval_loss": 0.6952112317085266,
|
328 |
+
"eval_rewards/accuracies": 0.5191029906272888,
|
329 |
+
"eval_rewards/chosen": 0.0005306191742420197,
|
330 |
+
"eval_rewards/margins": -0.0032888990826904774,
|
331 |
+
"eval_rewards/rejected": 0.0038195180241018534,
|
332 |
+
"eval_runtime": 145.8142,
|
333 |
+
"eval_samples_per_second": 2.352,
|
334 |
"eval_steps_per_second": 0.295,
|
335 |
"step": 200
|
336 |
},
|
337 |
{
|
338 |
"epoch": 0.55,
|
339 |
"learning_rate": 2.5453962426402e-07,
|
340 |
+
"logits/chosen": -2.0318386554718018,
|
341 |
+
"logits/rejected": -2.0425145626068115,
|
342 |
+
"logps/chosen": -31.961750030517578,
|
343 |
+
"logps/rejected": -33.86240005493164,
|
344 |
+
"loss": 0.6923,
|
345 |
+
"rewards/accuracies": 0.5249999761581421,
|
346 |
+
"rewards/chosen": 0.0028654958587139845,
|
347 |
+
"rewards/margins": 0.002774373395368457,
|
348 |
+
"rewards/rejected": 9.112273983191699e-05,
|
349 |
"step": 210
|
350 |
},
|
351 |
{
|
352 |
"epoch": 0.57,
|
353 |
"learning_rate": 2.318564697655179e-07,
|
354 |
+
"logits/chosen": -1.9253981113433838,
|
355 |
+
"logits/rejected": -1.9402506351470947,
|
356 |
+
"logps/chosen": -30.093517303466797,
|
357 |
+
"logps/rejected": -31.571548461914062,
|
358 |
+
"loss": 0.6889,
|
359 |
+
"rewards/accuracies": 0.574999988079071,
|
360 |
+
"rewards/chosen": 0.005405673757195473,
|
361 |
+
"rewards/margins": 0.009417025372385979,
|
362 |
+
"rewards/rejected": -0.004011353012174368,
|
363 |
"step": 220
|
364 |
},
|
365 |
{
|
366 |
"epoch": 0.6,
|
367 |
"learning_rate": 2.093227910899832e-07,
|
368 |
+
"logits/chosen": -1.9839330911636353,
|
369 |
+
"logits/rejected": -1.9879001379013062,
|
370 |
+
"logps/chosen": -33.4061279296875,
|
371 |
+
"logps/rejected": -31.564762115478516,
|
372 |
+
"loss": 0.6896,
|
373 |
+
"rewards/accuracies": 0.550000011920929,
|
374 |
+
"rewards/chosen": 0.0034024070482701063,
|
375 |
+
"rewards/margins": 0.008040683344006538,
|
376 |
+
"rewards/rejected": -0.004638276062905788,
|
377 |
"step": 230
|
378 |
},
|
379 |
{
|
380 |
"epoch": 0.62,
|
381 |
"learning_rate": 1.8712423238279356e-07,
|
382 |
+
"logits/chosen": -1.983189344406128,
|
383 |
+
"logits/rejected": -1.9612159729003906,
|
384 |
+
"logps/chosen": -34.15366744995117,
|
385 |
+
"logps/rejected": -34.95070266723633,
|
386 |
+
"loss": 0.6937,
|
387 |
"rewards/accuracies": 0.5,
|
388 |
+
"rewards/chosen": 2.4808850866975263e-05,
|
389 |
+
"rewards/margins": -1.6275793086606427e-06,
|
390 |
+
"rewards/rejected": 2.643577681737952e-05,
|
391 |
"step": 240
|
392 |
},
|
393 |
{
|
394 |
"epoch": 0.65,
|
395 |
"learning_rate": 1.654436768970182e-07,
|
396 |
+
"logits/chosen": -2.0243947505950928,
|
397 |
+
"logits/rejected": -2.021087169647217,
|
398 |
+
"logps/chosen": -32.91602325439453,
|
399 |
+
"logps/rejected": -36.221763610839844,
|
400 |
+
"loss": 0.6928,
|
401 |
+
"rewards/accuracies": 0.4375,
|
402 |
+
"rewards/chosen": 0.002715445589274168,
|
403 |
+
"rewards/margins": 0.001767707639373839,
|
404 |
+
"rewards/rejected": 0.0009477367857471108,
|
405 |
"step": 250
|
406 |
},
|
407 |
{
|
408 |
"epoch": 0.68,
|
409 |
"learning_rate": 1.444597403062196e-07,
|
410 |
+
"logits/chosen": -1.8911654949188232,
|
411 |
+
"logits/rejected": -1.8887317180633545,
|
412 |
+
"logps/chosen": -34.18123245239258,
|
413 |
+
"logps/rejected": -35.515174865722656,
|
414 |
+
"loss": 0.691,
|
415 |
"rewards/accuracies": 0.5874999761581421,
|
416 |
+
"rewards/chosen": 0.0021498873829841614,
|
417 |
+
"rewards/margins": 0.005315869115293026,
|
418 |
+
"rewards/rejected": -0.0031659831292927265,
|
419 |
"step": 260
|
420 |
},
|
421 |
{
|
422 |
"epoch": 0.7,
|
423 |
"learning_rate": 1.2434529917578887e-07,
|
424 |
+
"logits/chosen": -1.875580072402954,
|
425 |
+
"logits/rejected": -1.8730493783950806,
|
426 |
+
"logps/chosen": -34.37896728515625,
|
427 |
+
"logps/rejected": -31.76068687438965,
|
428 |
+
"loss": 0.6925,
|
429 |
+
"rewards/accuracies": 0.5249999761581421,
|
430 |
+
"rewards/chosen": 0.005589136388152838,
|
431 |
+
"rewards/margins": 0.0023957251105457544,
|
432 |
+
"rewards/rejected": 0.0031934112776070833,
|
433 |
"step": 270
|
434 |
},
|
435 |
{
|
436 |
"epoch": 0.73,
|
437 |
"learning_rate": 1.0526606671603521e-07,
|
438 |
+
"logits/chosen": -1.9798663854599,
|
439 |
+
"logits/rejected": -1.9692500829696655,
|
440 |
+
"logps/chosen": -35.32219696044922,
|
441 |
+
"logps/rejected": -31.857141494750977,
|
442 |
+
"loss": 0.6861,
|
443 |
+
"rewards/accuracies": 0.5625,
|
444 |
+
"rewards/chosen": 0.010332418605685234,
|
445 |
+
"rewards/margins": 0.015176350250840187,
|
446 |
+
"rewards/rejected": -0.004843929782509804,
|
447 |
"step": 280
|
448 |
},
|
449 |
{
|
450 |
"epoch": 0.75,
|
451 |
"learning_rate": 8.737922755071453e-08,
|
452 |
+
"logits/chosen": -2.076137065887451,
|
453 |
+
"logits/rejected": -2.061100721359253,
|
454 |
+
"logps/chosen": -30.907434463500977,
|
455 |
+
"logps/rejected": -32.638729095458984,
|
456 |
+
"loss": 0.693,
|
457 |
+
"rewards/accuracies": 0.4749999940395355,
|
458 |
+
"rewards/chosen": 0.00500696524977684,
|
459 |
+
"rewards/margins": 0.0010755129624158144,
|
460 |
+
"rewards/rejected": 0.0039314525201916695,
|
461 |
"step": 290
|
462 |
},
|
463 |
{
|
464 |
"epoch": 0.78,
|
465 |
"learning_rate": 7.08321427484816e-08,
|
466 |
+
"logits/chosen": -1.947422981262207,
|
467 |
+
"logits/rejected": -1.9448877573013306,
|
468 |
+
"logps/chosen": -32.89393997192383,
|
469 |
+
"logps/rejected": -30.828500747680664,
|
470 |
+
"loss": 0.6869,
|
471 |
+
"rewards/accuracies": 0.5375000238418579,
|
472 |
+
"rewards/chosen": 0.009951007552444935,
|
473 |
+
"rewards/margins": 0.013458291999995708,
|
474 |
+
"rewards/rejected": -0.003507285611703992,
|
475 |
"step": 300
|
476 |
},
|
477 |
{
|
478 |
"epoch": 0.78,
|
479 |
+
"eval_logits/chosen": -2.244764566421509,
|
480 |
+
"eval_logits/rejected": -2.2398788928985596,
|
481 |
+
"eval_logps/chosen": -34.038692474365234,
|
482 |
+
"eval_logps/rejected": -37.512168884277344,
|
483 |
+
"eval_loss": 0.6949278712272644,
|
484 |
+
"eval_rewards/accuracies": 0.5161960124969482,
|
485 |
+
"eval_rewards/chosen": -0.0012418677797541022,
|
486 |
+
"eval_rewards/margins": -0.0025783225428313017,
|
487 |
+
"eval_rewards/rejected": 0.0013364545302465558,
|
488 |
+
"eval_runtime": 145.8034,
|
489 |
+
"eval_samples_per_second": 2.352,
|
490 |
"eval_steps_per_second": 0.295,
|
491 |
"step": 300
|
492 |
},
|
493 |
{
|
494 |
"epoch": 0.81,
|
495 |
"learning_rate": 5.576113578589034e-08,
|
496 |
+
"logits/chosen": -1.9299108982086182,
|
497 |
+
"logits/rejected": -1.9266481399536133,
|
498 |
+
"logps/chosen": -31.572988510131836,
|
499 |
+
"logps/rejected": -33.727210998535156,
|
500 |
+
"loss": 0.6912,
|
501 |
+
"rewards/accuracies": 0.5249999761581421,
|
502 |
+
"rewards/chosen": 0.009108476340770721,
|
503 |
+
"rewards/margins": 0.004626707639545202,
|
504 |
+
"rewards/rejected": 0.0044817691668868065,
|
505 |
"step": 310
|
506 |
},
|
507 |
{
|
508 |
"epoch": 0.83,
|
509 |
"learning_rate": 4.229036944380912e-08,
|
510 |
+
"logits/chosen": -1.9814649820327759,
|
511 |
+
"logits/rejected": -1.9691375494003296,
|
512 |
+
"logps/chosen": -34.551979064941406,
|
513 |
+
"logps/rejected": -33.541648864746094,
|
514 |
+
"loss": 0.6866,
|
515 |
+
"rewards/accuracies": 0.550000011920929,
|
516 |
+
"rewards/chosen": 0.013229632750153542,
|
517 |
+
"rewards/margins": 0.014072793535888195,
|
518 |
+
"rewards/rejected": -0.0008431615424342453,
|
519 |
"step": 320
|
520 |
},
|
521 |
{
|
522 |
"epoch": 0.86,
|
523 |
"learning_rate": 3.053082288996112e-08,
|
524 |
+
"logits/chosen": -2.017097234725952,
|
525 |
+
"logits/rejected": -2.0156359672546387,
|
526 |
+
"logps/chosen": -33.49595260620117,
|
527 |
+
"logps/rejected": -32.49913787841797,
|
528 |
+
"loss": 0.6922,
|
529 |
"rewards/accuracies": 0.512499988079071,
|
530 |
+
"rewards/chosen": -0.003048144979402423,
|
531 |
+
"rewards/margins": 0.002819380722939968,
|
532 |
+
"rewards/rejected": -0.00586752500385046,
|
533 |
"step": 330
|
534 |
},
|
535 |
{
|
536 |
"epoch": 0.88,
|
537 |
"learning_rate": 2.05793773749158e-08,
|
538 |
+
"logits/chosen": -2.103210926055908,
|
539 |
+
"logits/rejected": -2.0874040126800537,
|
540 |
+
"logps/chosen": -34.16447830200195,
|
541 |
+
"logps/rejected": -33.07390594482422,
|
542 |
+
"loss": 0.6981,
|
543 |
+
"rewards/accuracies": 0.44999998807907104,
|
544 |
+
"rewards/chosen": 0.004473570734262466,
|
545 |
+
"rewards/margins": -0.008766504004597664,
|
546 |
+
"rewards/rejected": 0.01324007473886013,
|
547 |
"step": 340
|
548 |
},
|
549 |
{
|
550 |
"epoch": 0.91,
|
551 |
"learning_rate": 1.251801807404168e-08,
|
552 |
+
"logits/chosen": -1.9749380350112915,
|
553 |
+
"logits/rejected": -1.9740018844604492,
|
554 |
+
"logps/chosen": -33.25703048706055,
|
555 |
+
"logps/rejected": -32.47133255004883,
|
556 |
+
"loss": 0.689,
|
557 |
"rewards/accuracies": 0.5,
|
558 |
+
"rewards/chosen": 0.00751916179433465,
|
559 |
+
"rewards/margins": 0.009374773129820824,
|
560 |
+
"rewards/rejected": -0.0018556114519014955,
|
561 |
"step": 350
|
562 |
},
|
563 |
{
|
564 |
"epoch": 0.94,
|
565 |
"learning_rate": 6.41315865106129e-09,
|
566 |
+
"logits/chosen": -1.9304084777832031,
|
567 |
+
"logits/rejected": -1.9407765865325928,
|
568 |
+
"logps/chosen": -32.21118927001953,
|
569 |
+
"logps/rejected": -35.31554412841797,
|
570 |
+
"loss": 0.6905,
|
571 |
+
"rewards/accuracies": 0.550000011920929,
|
572 |
+
"rewards/chosen": -0.0019482700154185295,
|
573 |
+
"rewards/margins": 0.0060885134153068066,
|
574 |
+
"rewards/rejected": -0.008036783896386623,
|
575 |
"step": 360
|
576 |
},
|
577 |
{
|
578 |
"epoch": 0.96,
|
579 |
"learning_rate": 2.3150941078050324e-09,
|
580 |
+
"logits/chosen": -2.06982421875,
|
581 |
+
"logits/rejected": -2.0632567405700684,
|
582 |
+
"logps/chosen": -33.641204833984375,
|
583 |
+
"logps/rejected": -29.2253475189209,
|
584 |
+
"loss": 0.6921,
|
585 |
+
"rewards/accuracies": 0.5625,
|
586 |
+
"rewards/chosen": 0.003229249268770218,
|
587 |
+
"rewards/margins": 0.0029065334238111973,
|
588 |
+
"rewards/rejected": 0.000322715932270512,
|
589 |
"step": 370
|
590 |
},
|
591 |
{
|
592 |
"epoch": 0.99,
|
593 |
"learning_rate": 2.575864278703266e-10,
|
594 |
+
"logits/chosen": -1.929220199584961,
|
595 |
+
"logits/rejected": -1.9313905239105225,
|
596 |
+
"logps/chosen": -34.26762008666992,
|
597 |
+
"logps/rejected": -30.9031982421875,
|
598 |
+
"loss": 0.6934,
|
599 |
+
"rewards/accuracies": 0.5249999761581421,
|
600 |
+
"rewards/chosen": -0.00857237633317709,
|
601 |
+
"rewards/margins": 0.0005976069951429963,
|
602 |
+
"rewards/rejected": -0.009169982746243477,
|
603 |
"step": 380
|
604 |
},
|
605 |
{
|
606 |
"epoch": 1.0,
|
607 |
"step": 385,
|
608 |
"total_flos": 0.0,
|
609 |
+
"train_loss": 0.691356255791404,
|
610 |
+
"train_runtime": 3254.7157,
|
611 |
"train_samples_per_second": 0.946,
|
612 |
"train_steps_per_second": 0.118
|
613 |
}
|