hugodk-sch
commited on
Commit
•
fc4e3e9
1
Parent(s):
e4569e7
Model save
Browse files- README.md +13 -16
- adapter_model.safetensors +1 -1
- all_results.json +3 -16
- train_results.json +3 -3
- trainer_state.json +374 -374
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:
|
25 |
-
- Rewards/rejected:
|
26 |
-
- Rewards/accuracies: 0.
|
27 |
-
- Rewards/margins: 0.
|
28 |
-
- Logps/rejected: -37.
|
29 |
-
- Logps/chosen: -
|
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: 0.4904
|
21 |
+
- Rewards/chosen: 0.2570
|
22 |
+
- Rewards/rejected: 0.2171
|
23 |
+
- Rewards/accuracies: 0.5220
|
24 |
+
- Rewards/margins: 0.0399
|
25 |
+
- Logps/rejected: -37.2754
|
26 |
+
- Logps/chosen: -33.7490
|
27 |
+
- Logits/rejected: -2.2259
|
28 |
+
- Logits/chosen: -2.2308
|
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 |
+
| 0.4639 | 0.26 | 100 | 0.4939 | 0.1483 | 0.1205 | 0.5627 | 0.0278 | -37.3828 | -33.8698 | -2.2305 | -2.2354 |
|
64 |
+
| 0.4308 | 0.52 | 200 | 0.4894 | 0.2606 | 0.2153 | 0.5544 | 0.0453 | -37.2774 | -33.7450 | -2.2290 | -2.2338 |
|
65 |
+
| 0.374 | 0.78 | 300 | 0.4904 | 0.2570 | 0.2171 | 0.5220 | 0.0399 | -37.2754 | -33.7490 | -2.2259 | -2.2308 |
|
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 sha256:18e50d4be9a44117ac2d82a5aa7a6b4ea5cc21668165b759ff991307589f83b4
|
3 |
size 176183216
|
all_results.json
CHANGED
@@ -1,21 +1,8 @@
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
-
"
|
4 |
-
"
|
5 |
-
"eval_logps/chosen": -34.096519470214844,
|
6 |
-
"eval_logps/rejected": -37.601966857910156,
|
7 |
-
"eval_loss": 0.6981170177459717,
|
8 |
-
"eval_rewards/accuracies": 0.545265793800354,
|
9 |
-
"eval_rewards/chosen": -0.0557732917368412,
|
10 |
-
"eval_rewards/margins": 0.021040165796875954,
|
11 |
-
"eval_rewards/rejected": -0.07681345194578171,
|
12 |
-
"eval_runtime": 145.6397,
|
13 |
-
"eval_samples": 343,
|
14 |
-
"eval_samples_per_second": 2.355,
|
15 |
-
"eval_steps_per_second": 0.295,
|
16 |
-
"train_loss": 0.6175476637753573,
|
17 |
-
"train_runtime": 3252.7839,
|
18 |
"train_samples": 3079,
|
19 |
-
"train_samples_per_second": 0.
|
20 |
"train_steps_per_second": 0.118
|
21 |
}
|
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
+
"train_loss": 0.447351616079157,
|
4 |
+
"train_runtime": 3253.4458,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
5 |
"train_samples": 3079,
|
6 |
+
"train_samples_per_second": 0.946,
|
7 |
"train_steps_per_second": 0.118
|
8 |
}
|
train_results.json
CHANGED
@@ -1,8 +1,8 @@
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
-
"train_loss": 0.
|
4 |
-
"train_runtime":
|
5 |
"train_samples": 3079,
|
6 |
-
"train_samples_per_second": 0.
|
7 |
"train_steps_per_second": 0.118
|
8 |
}
|
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
+
"train_loss": 0.447351616079157,
|
4 |
+
"train_runtime": 3253.4458,
|
5 |
"train_samples": 3079,
|
6 |
+
"train_samples_per_second": 0.946,
|
7 |
"train_steps_per_second": 0.118
|
8 |
}
|
trainer_state.json
CHANGED
@@ -15,7 +15,7 @@
|
|
15 |
"logits/rejected": -1.7377450466156006,
|
16 |
"logps/chosen": -29.553977966308594,
|
17 |
"logps/rejected": -42.813133239746094,
|
18 |
-
"loss": 0.
|
19 |
"rewards/accuracies": 0.0,
|
20 |
"rewards/chosen": 0.0,
|
21 |
"rewards/margins": 0.0,
|
@@ -25,590 +25,590 @@
|
|
25 |
{
|
26 |
"epoch": 0.03,
|
27 |
"learning_rate": 1.282051282051282e-06,
|
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": 2.564102564102564e-06,
|
42 |
-
"logits/chosen": -1.
|
43 |
-
"logits/rejected": -2.
|
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":
|
51 |
"step": 20
|
52 |
},
|
53 |
{
|
54 |
"epoch": 0.08,
|
55 |
"learning_rate": 3.846153846153847e-06,
|
56 |
-
"logits/chosen": -1.
|
57 |
-
"logits/rejected": -1.
|
58 |
-
"logps/chosen": -31.
|
59 |
-
"logps/rejected": -33.
|
60 |
-
"loss": 0.
|
61 |
"rewards/accuracies": 0.5249999761581421,
|
62 |
-
"rewards/chosen": 0.
|
63 |
-
"rewards/margins": 0.
|
64 |
-
"rewards/rejected": -0.
|
65 |
"step": 30
|
66 |
},
|
67 |
{
|
68 |
"epoch": 0.1,
|
69 |
"learning_rate": 4.999896948438434e-06,
|
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":
|
79 |
"step": 40
|
80 |
},
|
81 |
{
|
82 |
"epoch": 0.13,
|
83 |
"learning_rate": 4.987541037542187e-06,
|
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": 0.
|
91 |
-
"rewards/margins": 0.
|
92 |
-
"rewards/rejected":
|
93 |
"step": 50
|
94 |
},
|
95 |
{
|
96 |
"epoch": 0.16,
|
97 |
"learning_rate": 4.954691471941119e-06,
|
98 |
-
"logits/chosen": -1.
|
99 |
-
"logits/rejected": -1.
|
100 |
-
"logps/chosen": -32.
|
101 |
-
"logps/rejected": -33.
|
102 |
-
"loss": 0.
|
103 |
"rewards/accuracies": 0.6499999761581421,
|
104 |
-
"rewards/chosen": 0.
|
105 |
-
"rewards/margins": 0.
|
106 |
-
"rewards/rejected":
|
107 |
"step": 60
|
108 |
},
|
109 |
{
|
110 |
"epoch": 0.18,
|
111 |
"learning_rate": 4.901618883413549e-06,
|
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": 0.
|
119 |
-
"rewards/margins": 0.
|
120 |
-
"rewards/rejected":
|
121 |
"step": 70
|
122 |
},
|
123 |
{
|
124 |
"epoch": 0.21,
|
125 |
"learning_rate": 4.828760511501322e-06,
|
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":
|
135 |
"step": 80
|
136 |
},
|
137 |
{
|
138 |
"epoch": 0.23,
|
139 |
"learning_rate": 4.7367166013034295e-06,
|
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":
|
149 |
"step": 90
|
150 |
},
|
151 |
{
|
152 |
"epoch": 0.26,
|
153 |
"learning_rate": 4.626245458345211e-06,
|
154 |
-
"logits/chosen": -2.
|
155 |
-
"logits/rejected": -2.
|
156 |
-
"logps/chosen": -
|
157 |
-
"logps/rejected": -31.
|
158 |
-
"loss": 0.
|
159 |
-
"rewards/accuracies": 0.
|
160 |
-
"rewards/chosen": 0.
|
161 |
-
"rewards/margins": 0.
|
162 |
-
"rewards/rejected":
|
163 |
"step": 100
|
164 |
},
|
165 |
{
|
166 |
"epoch": 0.26,
|
167 |
-
"eval_logits/chosen": -2.
|
168 |
-
"eval_logits/rejected": -2.
|
169 |
-
"eval_logps/chosen": -
|
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":
|
177 |
-
"eval_samples_per_second": 2.
|
178 |
-
"eval_steps_per_second": 0.
|
179 |
"step": 100
|
180 |
},
|
181 |
{
|
182 |
"epoch": 0.29,
|
183 |
"learning_rate": 4.498257201263691e-06,
|
184 |
-
"logits/chosen": -1.
|
185 |
-
"logits/rejected": -1.
|
186 |
-
"logps/chosen": -
|
187 |
-
"logps/rejected": -
|
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": 4.353806263777678e-06,
|
198 |
-
"logits/chosen": -2.
|
199 |
-
"logits/rejected": -1.
|
200 |
-
"logps/chosen": -32.
|
201 |
-
"logps/rejected": -
|
202 |
-
"loss": 0.
|
203 |
-
"rewards/accuracies": 0.
|
204 |
-
"rewards/chosen": 0.
|
205 |
-
"rewards/margins": 0.
|
206 |
-
"rewards/rejected": 0.
|
207 |
"step": 120
|
208 |
},
|
209 |
{
|
210 |
"epoch": 0.34,
|
211 |
"learning_rate": 4.1940827077152755e-06,
|
212 |
-
"logits/chosen": -2.
|
213 |
-
"logits/rejected": -2.
|
214 |
-
"logps/chosen": -30.
|
215 |
-
"logps/rejected": -
|
216 |
-
"loss": 0.
|
217 |
-
"rewards/accuracies": 0.
|
218 |
-
"rewards/chosen": 0.
|
219 |
-
"rewards/margins": 0.
|
220 |
-
"rewards/rejected":
|
221 |
"step": 130
|
222 |
},
|
223 |
{
|
224 |
"epoch": 0.36,
|
225 |
"learning_rate": 4.0204024186666215e-06,
|
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":
|
235 |
"step": 140
|
236 |
},
|
237 |
{
|
238 |
"epoch": 0.39,
|
239 |
"learning_rate": 3.834196265035119e-06,
|
240 |
-
"logits/chosen": -1.
|
241 |
-
"logits/rejected": -1.
|
242 |
-
"logps/chosen": -33.
|
243 |
-
"logps/rejected": -34.
|
244 |
-
"loss": 0.
|
245 |
"rewards/accuracies": 0.612500011920929,
|
246 |
-
"rewards/chosen": 0.
|
247 |
-
"rewards/margins": 0.
|
248 |
-
"rewards/rejected":
|
249 |
"step": 150
|
250 |
},
|
251 |
{
|
252 |
"epoch": 0.42,
|
253 |
"learning_rate": 3.636998309800573e-06,
|
254 |
-
"logits/chosen": -1.
|
255 |
-
"logits/rejected": -1.
|
256 |
-
"logps/chosen": -
|
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": 3.4304331721118078e-06,
|
268 |
-
"logits/chosen": -2.
|
269 |
-
"logits/rejected": -2.
|
270 |
-
"logps/chosen": -33.
|
271 |
-
"logps/rejected": -31.
|
272 |
-
"loss": 0.
|
273 |
-
"rewards/accuracies": 0.
|
274 |
-
"rewards/chosen": 0.
|
275 |
-
"rewards/margins": 0.
|
276 |
-
"rewards/rejected":
|
277 |
"step": 170
|
278 |
},
|
279 |
{
|
280 |
"epoch": 0.47,
|
281 |
"learning_rate": 3.2162026428305436e-06,
|
282 |
-
"logits/chosen": -2.
|
283 |
-
"logits/rejected": -2.
|
284 |
-
"logps/chosen": -
|
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": 2.996071664294641e-06,
|
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": 0.
|
304 |
-
"rewards/rejected":
|
305 |
"step": 190
|
306 |
},
|
307 |
{
|
308 |
"epoch": 0.52,
|
309 |
"learning_rate": 2.7718537898066833e-06,
|
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":
|
319 |
"step": 200
|
320 |
},
|
321 |
{
|
322 |
"epoch": 0.52,
|
323 |
-
"eval_logits/chosen": -2.
|
324 |
-
"eval_logits/rejected": -2.
|
325 |
-
"eval_logps/chosen": -
|
326 |
-
"eval_logps/rejected": -37.
|
327 |
-
"eval_loss": 0.
|
328 |
-
"eval_rewards/accuracies": 0.
|
329 |
-
"eval_rewards/chosen":
|
330 |
-
"eval_rewards/margins": 0.
|
331 |
-
"eval_rewards/rejected":
|
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": 2.5453962426402006e-06,
|
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": 2.3185646976551794e-06,
|
354 |
-
"logits/chosen": -1.
|
355 |
-
"logits/rejected": -1.
|
356 |
-
"logps/chosen": -29.
|
357 |
-
"logps/rejected": -31.
|
358 |
-
"loss": 0.
|
359 |
-
"rewards/accuracies": 0.
|
360 |
-
"rewards/chosen": 0.
|
361 |
-
"rewards/margins": 0.
|
362 |
-
"rewards/rejected":
|
363 |
"step": 220
|
364 |
},
|
365 |
{
|
366 |
"epoch": 0.6,
|
367 |
"learning_rate": 2.0932279108998323e-06,
|
368 |
-
"logits/chosen": -1.
|
369 |
-
"logits/rejected": -1.
|
370 |
-
"logps/chosen": -
|
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": 1.8712423238279358e-06,
|
382 |
-
"logits/chosen": -1.
|
383 |
-
"logits/rejected": -1.
|
384 |
-
"logps/chosen": -33.
|
385 |
-
"logps/rejected": -
|
386 |
-
"loss": 0.
|
387 |
-
"rewards/accuracies": 0.
|
388 |
-
"rewards/chosen": 0.
|
389 |
-
"rewards/margins": 0.
|
390 |
-
"rewards/rejected":
|
391 |
"step": 240
|
392 |
},
|
393 |
{
|
394 |
"epoch": 0.65,
|
395 |
"learning_rate": 1.6544367689701824e-06,
|
396 |
-
"logits/chosen": -2.
|
397 |
-
"logits/rejected": -2.
|
398 |
-
"logps/chosen": -32.
|
399 |
-
"logps/rejected": -
|
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": 1.4445974030621963e-06,
|
410 |
-
"logits/chosen": -1.
|
411 |
-
"logits/rejected": -1.
|
412 |
-
"logps/chosen": -
|
413 |
-
"logps/rejected": -35.
|
414 |
-
"loss": 0.
|
415 |
-
"rewards/accuracies": 0.
|
416 |
-
"rewards/chosen": 0.
|
417 |
-
"rewards/margins": 0.
|
418 |
-
"rewards/rejected":
|
419 |
"step": 260
|
420 |
},
|
421 |
{
|
422 |
"epoch": 0.7,
|
423 |
"learning_rate": 1.243452991757889e-06,
|
424 |
-
"logits/chosen": -1.
|
425 |
-
"logits/rejected": -1.
|
426 |
-
"logps/chosen": -
|
427 |
-
"logps/rejected": -31.
|
428 |
-
"loss": 0.
|
429 |
-
"rewards/accuracies": 0.
|
430 |
-
"rewards/chosen": 0.
|
431 |
-
"rewards/margins": 0.
|
432 |
-
"rewards/rejected":
|
433 |
"step": 270
|
434 |
},
|
435 |
{
|
436 |
"epoch": 0.73,
|
437 |
"learning_rate": 1.0526606671603523e-06,
|
438 |
-
"logits/chosen": -1.
|
439 |
-
"logits/rejected": -1.
|
440 |
-
"logps/chosen": -
|
441 |
-
"logps/rejected": -31.
|
442 |
-
"loss": 0.
|
443 |
"rewards/accuracies": 0.699999988079071,
|
444 |
-
"rewards/chosen": 0.
|
445 |
-
"rewards/margins": 0.
|
446 |
-
"rewards/rejected":
|
447 |
"step": 280
|
448 |
},
|
449 |
{
|
450 |
"epoch": 0.75,
|
451 |
"learning_rate": 8.737922755071455e-07,
|
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": 0.
|
459 |
-
"rewards/margins": 0.
|
460 |
-
"rewards/rejected":
|
461 |
"step": 290
|
462 |
},
|
463 |
{
|
464 |
"epoch": 0.78,
|
465 |
"learning_rate": 7.08321427484816e-07,
|
466 |
-
"logits/chosen": -1.
|
467 |
-
"logits/rejected": -1.
|
468 |
-
"logps/chosen": -32.
|
469 |
-
"logps/rejected": -30.
|
470 |
-
"loss": 0.
|
471 |
"rewards/accuracies": 0.737500011920929,
|
472 |
-
"rewards/chosen": 0.
|
473 |
-
"rewards/margins": 0.
|
474 |
-
"rewards/rejected":
|
475 |
"step": 300
|
476 |
},
|
477 |
{
|
478 |
"epoch": 0.78,
|
479 |
-
"eval_logits/chosen": -2.
|
480 |
-
"eval_logits/rejected": -2.
|
481 |
-
"eval_logps/chosen": -
|
482 |
-
"eval_logps/rejected": -37.
|
483 |
-
"eval_loss": 0.
|
484 |
-
"eval_rewards/accuracies": 0.
|
485 |
-
"eval_rewards/chosen":
|
486 |
-
"eval_rewards/margins": 0.
|
487 |
-
"eval_rewards/rejected":
|
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": 5.576113578589035e-07,
|
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": 4.229036944380913e-07,
|
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":
|
519 |
"step": 320
|
520 |
},
|
521 |
{
|
522 |
"epoch": 0.86,
|
523 |
"learning_rate": 3.053082288996112e-07,
|
524 |
-
"logits/chosen": -2.
|
525 |
-
"logits/rejected": -
|
526 |
-
"logps/chosen": -
|
527 |
-
"logps/rejected": -32.
|
528 |
-
"loss": 0.
|
529 |
-
"rewards/accuracies": 0.
|
530 |
-
"rewards/chosen": 0.
|
531 |
-
"rewards/margins": 0.
|
532 |
-
"rewards/rejected":
|
533 |
"step": 330
|
534 |
},
|
535 |
{
|
536 |
"epoch": 0.88,
|
537 |
"learning_rate": 2.0579377374915805e-07,
|
538 |
-
"logits/chosen": -2.
|
539 |
-
"logits/rejected": -2.
|
540 |
-
"logps/chosen": -33.
|
541 |
-
"logps/rejected": -
|
542 |
-
"loss": 0.
|
543 |
-
"rewards/accuracies": 0.
|
544 |
-
"rewards/chosen": 0.
|
545 |
-
"rewards/margins": 0.
|
546 |
-
"rewards/rejected":
|
547 |
"step": 340
|
548 |
},
|
549 |
{
|
550 |
"epoch": 0.91,
|
551 |
"learning_rate": 1.2518018074041684e-07,
|
552 |
-
"logits/chosen": -1.
|
553 |
-
"logits/rejected": -1.
|
554 |
-
"logps/chosen": -32.
|
555 |
-
"logps/rejected": -32.
|
556 |
-
"loss": 0.
|
557 |
"rewards/accuracies": 0.7124999761581421,
|
558 |
-
"rewards/chosen": 0.
|
559 |
-
"rewards/margins": 0.
|
560 |
-
"rewards/rejected":
|
561 |
"step": 350
|
562 |
},
|
563 |
{
|
564 |
"epoch": 0.94,
|
565 |
"learning_rate": 6.41315865106129e-08,
|
566 |
-
"logits/chosen": -1.
|
567 |
-
"logits/rejected": -1.
|
568 |
-
"logps/chosen": -31.
|
569 |
-
"logps/rejected": -
|
570 |
-
"loss": 0.
|
571 |
-
"rewards/accuracies": 0.
|
572 |
-
"rewards/chosen": 0.
|
573 |
-
"rewards/margins": 0.
|
574 |
-
"rewards/rejected":
|
575 |
"step": 360
|
576 |
},
|
577 |
{
|
578 |
"epoch": 0.96,
|
579 |
"learning_rate": 2.3150941078050325e-08,
|
580 |
-
"logits/chosen": -2.
|
581 |
-
"logits/rejected": -2.
|
582 |
-
"logps/chosen": -33.
|
583 |
-
"logps/rejected": -
|
584 |
-
"loss": 0.
|
585 |
-
"rewards/accuracies": 0.
|
586 |
-
"rewards/chosen": 0.
|
587 |
-
"rewards/margins": 0.
|
588 |
-
"rewards/rejected":
|
589 |
"step": 370
|
590 |
},
|
591 |
{
|
592 |
"epoch": 0.99,
|
593 |
"learning_rate": 2.575864278703266e-09,
|
594 |
-
"logits/chosen": -1.
|
595 |
-
"logits/rejected": -1.
|
596 |
-
"logps/chosen": -33.
|
597 |
-
"logps/rejected": -30.
|
598 |
-
"loss": 0.
|
599 |
-
"rewards/accuracies": 0.
|
600 |
-
"rewards/chosen": 0.
|
601 |
-
"rewards/margins": 0.
|
602 |
-
"rewards/rejected":
|
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.
|
612 |
"train_steps_per_second": 0.118
|
613 |
}
|
614 |
],
|
|
|
15 |
"logits/rejected": -1.7377450466156006,
|
16 |
"logps/chosen": -29.553977966308594,
|
17 |
"logps/rejected": -42.813133239746094,
|
18 |
+
"loss": 0.5,
|
19 |
"rewards/accuracies": 0.0,
|
20 |
"rewards/chosen": 0.0,
|
21 |
"rewards/margins": 0.0,
|
|
|
25 |
{
|
26 |
"epoch": 0.03,
|
27 |
"learning_rate": 1.282051282051282e-06,
|
28 |
+
"logits/chosen": -1.8665987253189087,
|
29 |
+
"logits/rejected": -1.8709272146224976,
|
30 |
+
"logps/chosen": -36.985595703125,
|
31 |
+
"logps/rejected": -33.68160629272461,
|
32 |
+
"loss": 0.4886,
|
33 |
+
"rewards/accuracies": 0.5694444179534912,
|
34 |
+
"rewards/chosen": 0.018904482945799828,
|
35 |
+
"rewards/margins": 0.06528304517269135,
|
36 |
+
"rewards/rejected": -0.04637856408953667,
|
37 |
"step": 10
|
38 |
},
|
39 |
{
|
40 |
"epoch": 0.05,
|
41 |
"learning_rate": 2.564102564102564e-06,
|
42 |
+
"logits/chosen": -1.997780203819275,
|
43 |
+
"logits/rejected": -2.000434398651123,
|
44 |
+
"logps/chosen": -29.643661499023438,
|
45 |
+
"logps/rejected": -29.043325424194336,
|
46 |
+
"loss": 0.5031,
|
47 |
+
"rewards/accuracies": 0.4375,
|
48 |
+
"rewards/chosen": -0.001316396868787706,
|
49 |
+
"rewards/margins": -0.019422104582190514,
|
50 |
+
"rewards/rejected": 0.018105709925293922,
|
51 |
"step": 20
|
52 |
},
|
53 |
{
|
54 |
"epoch": 0.08,
|
55 |
"learning_rate": 3.846153846153847e-06,
|
56 |
+
"logits/chosen": -1.9207446575164795,
|
57 |
+
"logits/rejected": -1.918060064315796,
|
58 |
+
"logps/chosen": -31.41064453125,
|
59 |
+
"logps/rejected": -33.227088928222656,
|
60 |
+
"loss": 0.4976,
|
61 |
"rewards/accuracies": 0.5249999761581421,
|
62 |
+
"rewards/chosen": 0.004905471112579107,
|
63 |
+
"rewards/margins": 0.012669263407588005,
|
64 |
+
"rewards/rejected": -0.007763790898025036,
|
65 |
"step": 30
|
66 |
},
|
67 |
{
|
68 |
"epoch": 0.1,
|
69 |
"learning_rate": 4.999896948438434e-06,
|
70 |
+
"logits/chosen": -2.017446517944336,
|
71 |
+
"logits/rejected": -2.0087125301361084,
|
72 |
+
"logps/chosen": -32.553016662597656,
|
73 |
+
"logps/rejected": -32.50551986694336,
|
74 |
+
"loss": 0.4977,
|
75 |
+
"rewards/accuracies": 0.5,
|
76 |
+
"rewards/chosen": 0.021415216848254204,
|
77 |
+
"rewards/margins": 0.014982220716774464,
|
78 |
+
"rewards/rejected": 0.006432999856770039,
|
79 |
"step": 40
|
80 |
},
|
81 |
{
|
82 |
"epoch": 0.13,
|
83 |
"learning_rate": 4.987541037542187e-06,
|
84 |
+
"logits/chosen": -1.8627235889434814,
|
85 |
+
"logits/rejected": -1.851959228515625,
|
86 |
+
"logps/chosen": -33.5064697265625,
|
87 |
+
"logps/rejected": -35.43267059326172,
|
88 |
+
"loss": 0.4951,
|
89 |
+
"rewards/accuracies": 0.550000011920929,
|
90 |
+
"rewards/chosen": 0.0459100641310215,
|
91 |
+
"rewards/margins": 0.02820250764489174,
|
92 |
+
"rewards/rejected": 0.017707552760839462,
|
93 |
"step": 50
|
94 |
},
|
95 |
{
|
96 |
"epoch": 0.16,
|
97 |
"learning_rate": 4.954691471941119e-06,
|
98 |
+
"logits/chosen": -1.9425691366195679,
|
99 |
+
"logits/rejected": -1.94449782371521,
|
100 |
+
"logps/chosen": -32.46650695800781,
|
101 |
+
"logps/rejected": -33.15652847290039,
|
102 |
+
"loss": 0.4765,
|
103 |
"rewards/accuracies": 0.6499999761581421,
|
104 |
+
"rewards/chosen": 0.12030963599681854,
|
105 |
+
"rewards/margins": 0.10694190114736557,
|
106 |
+
"rewards/rejected": 0.013367725536227226,
|
107 |
"step": 60
|
108 |
},
|
109 |
{
|
110 |
"epoch": 0.18,
|
111 |
"learning_rate": 4.901618883413549e-06,
|
112 |
+
"logits/chosen": -2.073408842086792,
|
113 |
+
"logits/rejected": -2.078367233276367,
|
114 |
+
"logps/chosen": -33.917694091796875,
|
115 |
+
"logps/rejected": -36.547218322753906,
|
116 |
+
"loss": 0.4901,
|
117 |
+
"rewards/accuracies": 0.5375000238418579,
|
118 |
+
"rewards/chosen": 0.06371410191059113,
|
119 |
+
"rewards/margins": 0.04422418028116226,
|
120 |
+
"rewards/rejected": 0.019489921629428864,
|
121 |
"step": 70
|
122 |
},
|
123 |
{
|
124 |
"epoch": 0.21,
|
125 |
"learning_rate": 4.828760511501322e-06,
|
126 |
+
"logits/chosen": -1.9349607229232788,
|
127 |
+
"logits/rejected": -1.9380786418914795,
|
128 |
+
"logps/chosen": -34.223785400390625,
|
129 |
+
"logps/rejected": -34.53069305419922,
|
130 |
+
"loss": 0.4713,
|
131 |
+
"rewards/accuracies": 0.612500011920929,
|
132 |
+
"rewards/chosen": 0.1836203634738922,
|
133 |
+
"rewards/margins": 0.12807974219322205,
|
134 |
+
"rewards/rejected": 0.05554063245654106,
|
135 |
"step": 80
|
136 |
},
|
137 |
{
|
138 |
"epoch": 0.23,
|
139 |
"learning_rate": 4.7367166013034295e-06,
|
140 |
+
"logits/chosen": -1.9439691305160522,
|
141 |
+
"logits/rejected": -1.9484784603118896,
|
142 |
+
"logps/chosen": -32.27050018310547,
|
143 |
+
"logps/rejected": -32.26476287841797,
|
144 |
+
"loss": 0.4779,
|
145 |
+
"rewards/accuracies": 0.637499988079071,
|
146 |
+
"rewards/chosen": 0.1772255003452301,
|
147 |
+
"rewards/margins": 0.09250012785196304,
|
148 |
+
"rewards/rejected": 0.08472537249326706,
|
149 |
"step": 90
|
150 |
},
|
151 |
{
|
152 |
"epoch": 0.26,
|
153 |
"learning_rate": 4.626245458345211e-06,
|
154 |
+
"logits/chosen": -2.0411603450775146,
|
155 |
+
"logits/rejected": -2.039163112640381,
|
156 |
+
"logps/chosen": -31.98573875427246,
|
157 |
+
"logps/rejected": -31.193227767944336,
|
158 |
+
"loss": 0.4639,
|
159 |
+
"rewards/accuracies": 0.6625000238418579,
|
160 |
+
"rewards/chosen": 0.22958631813526154,
|
161 |
+
"rewards/margins": 0.16390272974967957,
|
162 |
+
"rewards/rejected": 0.06568360328674316,
|
163 |
"step": 100
|
164 |
},
|
165 |
{
|
166 |
"epoch": 0.26,
|
167 |
+
"eval_logits/chosen": -2.235391855239868,
|
168 |
+
"eval_logits/rejected": -2.2305493354797363,
|
169 |
+
"eval_logps/chosen": -33.869815826416016,
|
170 |
+
"eval_logps/rejected": -37.382774353027344,
|
171 |
+
"eval_loss": 0.4939241409301758,
|
172 |
+
"eval_rewards/accuracies": 0.5627076625823975,
|
173 |
+
"eval_rewards/chosen": 0.1482628434896469,
|
174 |
+
"eval_rewards/margins": 0.02780282311141491,
|
175 |
+
"eval_rewards/rejected": 0.12046003341674805,
|
176 |
+
"eval_runtime": 145.9747,
|
177 |
+
"eval_samples_per_second": 2.35,
|
178 |
+
"eval_steps_per_second": 0.295,
|
179 |
"step": 100
|
180 |
},
|
181 |
{
|
182 |
"epoch": 0.29,
|
183 |
"learning_rate": 4.498257201263691e-06,
|
184 |
+
"logits/chosen": -1.997287392616272,
|
185 |
+
"logits/rejected": -1.9949369430541992,
|
186 |
+
"logps/chosen": -32.96843719482422,
|
187 |
+
"logps/rejected": -33.866310119628906,
|
188 |
+
"loss": 0.4739,
|
189 |
+
"rewards/accuracies": 0.5249999761581421,
|
190 |
+
"rewards/chosen": 0.24784216284751892,
|
191 |
+
"rewards/margins": 0.10108550637960434,
|
192 |
+
"rewards/rejected": 0.14675670862197876,
|
193 |
"step": 110
|
194 |
},
|
195 |
{
|
196 |
"epoch": 0.31,
|
197 |
"learning_rate": 4.353806263777678e-06,
|
198 |
+
"logits/chosen": -2.008091688156128,
|
199 |
+
"logits/rejected": -1.9997599124908447,
|
200 |
+
"logps/chosen": -32.20352554321289,
|
201 |
+
"logps/rejected": -31.995223999023438,
|
202 |
+
"loss": 0.485,
|
203 |
+
"rewards/accuracies": 0.5375000238418579,
|
204 |
+
"rewards/chosen": 0.2172430008649826,
|
205 |
+
"rewards/margins": 0.06759083271026611,
|
206 |
+
"rewards/rejected": 0.1496521681547165,
|
207 |
"step": 120
|
208 |
},
|
209 |
{
|
210 |
"epoch": 0.34,
|
211 |
"learning_rate": 4.1940827077152755e-06,
|
212 |
+
"logits/chosen": -2.035614490509033,
|
213 |
+
"logits/rejected": -2.027682304382324,
|
214 |
+
"logps/chosen": -30.1588077545166,
|
215 |
+
"logps/rejected": -31.886260986328125,
|
216 |
+
"loss": 0.4717,
|
217 |
+
"rewards/accuracies": 0.550000011920929,
|
218 |
+
"rewards/chosen": 0.2852162718772888,
|
219 |
+
"rewards/margins": 0.1351451873779297,
|
220 |
+
"rewards/rejected": 0.15007111430168152,
|
221 |
"step": 130
|
222 |
},
|
223 |
{
|
224 |
"epoch": 0.36,
|
225 |
"learning_rate": 4.0204024186666215e-06,
|
226 |
+
"logits/chosen": -1.965490698814392,
|
227 |
+
"logits/rejected": -1.9756921529769897,
|
228 |
+
"logps/chosen": -31.065088272094727,
|
229 |
+
"logps/rejected": -32.42934036254883,
|
230 |
+
"loss": 0.4482,
|
231 |
+
"rewards/accuracies": 0.6875,
|
232 |
+
"rewards/chosen": 0.3191176950931549,
|
233 |
+
"rewards/margins": 0.22413134574890137,
|
234 |
+
"rewards/rejected": 0.09498633444309235,
|
235 |
"step": 140
|
236 |
},
|
237 |
{
|
238 |
"epoch": 0.39,
|
239 |
"learning_rate": 3.834196265035119e-06,
|
240 |
+
"logits/chosen": -1.8782259225845337,
|
241 |
+
"logits/rejected": -1.8793823719024658,
|
242 |
+
"logps/chosen": -33.68832778930664,
|
243 |
+
"logps/rejected": -34.58278274536133,
|
244 |
+
"loss": 0.4367,
|
245 |
"rewards/accuracies": 0.612500011920929,
|
246 |
+
"rewards/chosen": 0.45392999053001404,
|
247 |
+
"rewards/margins": 0.2969031035900116,
|
248 |
+
"rewards/rejected": 0.15702682733535767,
|
249 |
"step": 150
|
250 |
},
|
251 |
{
|
252 |
"epoch": 0.42,
|
253 |
"learning_rate": 3.636998309800573e-06,
|
254 |
+
"logits/chosen": -1.9295704364776611,
|
255 |
+
"logits/rejected": -1.9261808395385742,
|
256 |
+
"logps/chosen": -35.74212646484375,
|
257 |
+
"logps/rejected": -32.51028060913086,
|
258 |
+
"loss": 0.4538,
|
259 |
+
"rewards/accuracies": 0.6499999761581421,
|
260 |
+
"rewards/chosen": 0.38659390807151794,
|
261 |
+
"rewards/margins": 0.1953679323196411,
|
262 |
+
"rewards/rejected": 0.1912260353565216,
|
263 |
"step": 160
|
264 |
},
|
265 |
{
|
266 |
"epoch": 0.44,
|
267 |
"learning_rate": 3.4304331721118078e-06,
|
268 |
+
"logits/chosen": -2.031176805496216,
|
269 |
+
"logits/rejected": -2.023855686187744,
|
270 |
+
"logps/chosen": -33.24225616455078,
|
271 |
+
"logps/rejected": -31.193195343017578,
|
272 |
+
"loss": 0.42,
|
273 |
+
"rewards/accuracies": 0.675000011920929,
|
274 |
+
"rewards/chosen": 0.5000473260879517,
|
275 |
+
"rewards/margins": 0.3572581112384796,
|
276 |
+
"rewards/rejected": 0.1427893042564392,
|
277 |
"step": 170
|
278 |
},
|
279 |
{
|
280 |
"epoch": 0.47,
|
281 |
"learning_rate": 3.2162026428305436e-06,
|
282 |
+
"logits/chosen": -2.038222074508667,
|
283 |
+
"logits/rejected": -2.0434067249298096,
|
284 |
+
"logps/chosen": -31.95560646057129,
|
285 |
+
"logps/rejected": -32.17836380004883,
|
286 |
+
"loss": 0.4404,
|
287 |
+
"rewards/accuracies": 0.6875,
|
288 |
+
"rewards/chosen": 0.5302629470825195,
|
289 |
+
"rewards/margins": 0.2541634440422058,
|
290 |
+
"rewards/rejected": 0.2760995924472809,
|
291 |
"step": 180
|
292 |
},
|
293 |
{
|
294 |
"epoch": 0.49,
|
295 |
"learning_rate": 2.996071664294641e-06,
|
296 |
+
"logits/chosen": -2.0387401580810547,
|
297 |
+
"logits/rejected": -2.036006450653076,
|
298 |
+
"logps/chosen": -31.0674991607666,
|
299 |
+
"logps/rejected": -31.083877563476562,
|
300 |
+
"loss": 0.4607,
|
301 |
+
"rewards/accuracies": 0.6499999761581421,
|
302 |
+
"rewards/chosen": 0.3793107569217682,
|
303 |
+
"rewards/margins": 0.16841106116771698,
|
304 |
+
"rewards/rejected": 0.21089968085289001,
|
305 |
"step": 190
|
306 |
},
|
307 |
{
|
308 |
"epoch": 0.52,
|
309 |
"learning_rate": 2.7718537898066833e-06,
|
310 |
+
"logits/chosen": -1.9085681438446045,
|
311 |
+
"logits/rejected": -1.9132543802261353,
|
312 |
+
"logps/chosen": -31.083459854125977,
|
313 |
+
"logps/rejected": -32.602638244628906,
|
314 |
+
"loss": 0.4308,
|
315 |
+
"rewards/accuracies": 0.6875,
|
316 |
+
"rewards/chosen": 0.4705420434474945,
|
317 |
+
"rewards/margins": 0.301074743270874,
|
318 |
+
"rewards/rejected": 0.16946731507778168,
|
319 |
"step": 200
|
320 |
},
|
321 |
{
|
322 |
"epoch": 0.52,
|
323 |
+
"eval_logits/chosen": -2.2338287830352783,
|
324 |
+
"eval_logits/rejected": -2.229020357131958,
|
325 |
+
"eval_logps/chosen": -33.7449951171875,
|
326 |
+
"eval_logps/rejected": -37.27743911743164,
|
327 |
+
"eval_loss": 0.48942965269088745,
|
328 |
+
"eval_rewards/accuracies": 0.5544019937515259,
|
329 |
+
"eval_rewards/chosen": 0.26059985160827637,
|
330 |
+
"eval_rewards/margins": 0.0453372597694397,
|
331 |
+
"eval_rewards/rejected": 0.21526260673999786,
|
332 |
+
"eval_runtime": 145.8953,
|
333 |
+
"eval_samples_per_second": 2.351,
|
334 |
"eval_steps_per_second": 0.295,
|
335 |
"step": 200
|
336 |
},
|
337 |
{
|
338 |
"epoch": 0.55,
|
339 |
"learning_rate": 2.5453962426402006e-06,
|
340 |
+
"logits/chosen": -2.0216596126556396,
|
341 |
+
"logits/rejected": -2.032275915145874,
|
342 |
+
"logps/chosen": -31.500268936157227,
|
343 |
+
"logps/rejected": -33.663352966308594,
|
344 |
+
"loss": 0.4458,
|
345 |
+
"rewards/accuracies": 0.612500011920929,
|
346 |
+
"rewards/chosen": 0.42393389344215393,
|
347 |
+
"rewards/margins": 0.2445230931043625,
|
348 |
+
"rewards/rejected": 0.17941072583198547,
|
349 |
"step": 210
|
350 |
},
|
351 |
{
|
352 |
"epoch": 0.57,
|
353 |
"learning_rate": 2.3185646976551794e-06,
|
354 |
+
"logits/chosen": -1.9136396646499634,
|
355 |
+
"logits/rejected": -1.928344964981079,
|
356 |
+
"logps/chosen": -29.588964462280273,
|
357 |
+
"logps/rejected": -31.396224975585938,
|
358 |
+
"loss": 0.4269,
|
359 |
+
"rewards/accuracies": 0.699999988079071,
|
360 |
+
"rewards/chosen": 0.4703185558319092,
|
361 |
+
"rewards/margins": 0.3245617151260376,
|
362 |
+
"rewards/rejected": 0.1457568258047104,
|
363 |
"step": 220
|
364 |
},
|
365 |
{
|
366 |
"epoch": 0.6,
|
367 |
"learning_rate": 2.0932279108998323e-06,
|
368 |
+
"logits/chosen": -1.970298171043396,
|
369 |
+
"logits/rejected": -1.974283218383789,
|
370 |
+
"logps/chosen": -32.81959915161133,
|
371 |
+
"logps/rejected": -31.408565521240234,
|
372 |
+
"loss": 0.4109,
|
373 |
+
"rewards/accuracies": 0.699999988079071,
|
374 |
+
"rewards/chosen": 0.5380831956863403,
|
375 |
+
"rewards/margins": 0.4114208221435547,
|
376 |
+
"rewards/rejected": 0.12666237354278564,
|
377 |
"step": 230
|
378 |
},
|
379 |
{
|
380 |
"epoch": 0.62,
|
381 |
"learning_rate": 1.8712423238279358e-06,
|
382 |
+
"logits/chosen": -1.9695065021514893,
|
383 |
+
"logits/rejected": -1.9477574825286865,
|
384 |
+
"logps/chosen": -33.58247756958008,
|
385 |
+
"logps/rejected": -34.828121185302734,
|
386 |
+
"loss": 0.4129,
|
387 |
+
"rewards/accuracies": 0.6875,
|
388 |
+
"rewards/chosen": 0.5141419172286987,
|
389 |
+
"rewards/margins": 0.403735876083374,
|
390 |
+
"rewards/rejected": 0.11040612310171127,
|
391 |
"step": 240
|
392 |
},
|
393 |
{
|
394 |
"epoch": 0.65,
|
395 |
"learning_rate": 1.6544367689701824e-06,
|
396 |
+
"logits/chosen": -2.0098202228546143,
|
397 |
+
"logits/rejected": -2.0065340995788574,
|
398 |
+
"logps/chosen": -32.43529510498047,
|
399 |
+
"logps/rejected": -35.97461700439453,
|
400 |
+
"loss": 0.4514,
|
401 |
+
"rewards/accuracies": 0.5625,
|
402 |
+
"rewards/chosen": 0.44079580903053284,
|
403 |
+
"rewards/margins": 0.2155180424451828,
|
404 |
+
"rewards/rejected": 0.22527781128883362,
|
405 |
"step": 250
|
406 |
},
|
407 |
{
|
408 |
"epoch": 0.68,
|
409 |
"learning_rate": 1.4445974030621963e-06,
|
410 |
+
"logits/chosen": -1.8770506381988525,
|
411 |
+
"logits/rejected": -1.8746120929718018,
|
412 |
+
"logps/chosen": -33.7199821472168,
|
413 |
+
"logps/rejected": -35.28092575073242,
|
414 |
+
"loss": 0.4498,
|
415 |
+
"rewards/accuracies": 0.6625000238418579,
|
416 |
+
"rewards/chosen": 0.42157459259033203,
|
417 |
+
"rewards/margins": 0.2202514111995697,
|
418 |
+
"rewards/rejected": 0.2013232260942459,
|
419 |
"step": 260
|
420 |
},
|
421 |
{
|
422 |
"epoch": 0.7,
|
423 |
"learning_rate": 1.243452991757889e-06,
|
424 |
+
"logits/chosen": -1.8618510961532593,
|
425 |
+
"logits/rejected": -1.8593294620513916,
|
426 |
+
"logps/chosen": -33.92017364501953,
|
427 |
+
"logps/rejected": -31.6002197265625,
|
428 |
+
"loss": 0.4397,
|
429 |
+
"rewards/accuracies": 0.6625000238418579,
|
430 |
+
"rewards/chosen": 0.42968273162841797,
|
431 |
+
"rewards/margins": 0.27568089962005615,
|
432 |
+
"rewards/rejected": 0.15400180220603943,
|
433 |
"step": 270
|
434 |
},
|
435 |
{
|
436 |
"epoch": 0.73,
|
437 |
"learning_rate": 1.0526606671603523e-06,
|
438 |
+
"logits/chosen": -1.9657011032104492,
|
439 |
+
"logits/rejected": -1.9552650451660156,
|
440 |
+
"logps/chosen": -34.72232437133789,
|
441 |
+
"logps/rejected": -31.632369995117188,
|
442 |
+
"loss": 0.4114,
|
443 |
"rewards/accuracies": 0.699999988079071,
|
444 |
+
"rewards/chosen": 0.5708868503570557,
|
445 |
+
"rewards/margins": 0.3831265866756439,
|
446 |
+
"rewards/rejected": 0.18776027858257294,
|
447 |
"step": 280
|
448 |
},
|
449 |
{
|
450 |
"epoch": 0.75,
|
451 |
"learning_rate": 8.737922755071455e-07,
|
452 |
+
"logits/chosen": -2.0614376068115234,
|
453 |
+
"logits/rejected": -2.046600341796875,
|
454 |
+
"logps/chosen": -30.400625228881836,
|
455 |
+
"logps/rejected": -32.34136199951172,
|
456 |
+
"loss": 0.456,
|
457 |
+
"rewards/accuracies": 0.625,
|
458 |
+
"rewards/chosen": 0.4711507260799408,
|
459 |
+
"rewards/margins": 0.19172403216362,
|
460 |
+
"rewards/rejected": 0.2794266939163208,
|
461 |
"step": 290
|
462 |
},
|
463 |
{
|
464 |
"epoch": 0.78,
|
465 |
"learning_rate": 7.08321427484816e-07,
|
466 |
+
"logits/chosen": -1.9332094192504883,
|
467 |
+
"logits/rejected": -1.9307467937469482,
|
468 |
+
"logps/chosen": -32.10976028442383,
|
469 |
+
"logps/rejected": -30.661523818969727,
|
470 |
+
"loss": 0.374,
|
471 |
"rewards/accuracies": 0.737500011920929,
|
472 |
+
"rewards/chosen": 0.7356175184249878,
|
473 |
+
"rewards/margins": 0.5958597660064697,
|
474 |
+
"rewards/rejected": 0.13975778222084045,
|
475 |
"step": 300
|
476 |
},
|
477 |
{
|
478 |
"epoch": 0.78,
|
479 |
+
"eval_logits/chosen": -2.2307660579681396,
|
480 |
+
"eval_logits/rejected": -2.22594952583313,
|
481 |
+
"eval_logps/chosen": -33.74896240234375,
|
482 |
+
"eval_logps/rejected": -37.275413513183594,
|
483 |
+
"eval_loss": 0.49038100242614746,
|
484 |
+
"eval_rewards/accuracies": 0.5220099687576294,
|
485 |
+
"eval_rewards/chosen": 0.2570302486419678,
|
486 |
+
"eval_rewards/margins": 0.039943769574165344,
|
487 |
+
"eval_rewards/rejected": 0.21708647906780243,
|
488 |
+
"eval_runtime": 145.8077,
|
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": 5.576113578589035e-07,
|
496 |
+
"logits/chosen": -1.9164836406707764,
|
497 |
+
"logits/rejected": -1.913336992263794,
|
498 |
+
"logps/chosen": -31.014041900634766,
|
499 |
+
"logps/rejected": -33.548377990722656,
|
500 |
+
"loss": 0.4218,
|
501 |
+
"rewards/accuracies": 0.762499988079071,
|
502 |
+
"rewards/chosen": 0.5303782224655151,
|
503 |
+
"rewards/margins": 0.3559793531894684,
|
504 |
+
"rewards/rejected": 0.1743989735841751,
|
505 |
"step": 310
|
506 |
},
|
507 |
{
|
508 |
"epoch": 0.83,
|
509 |
"learning_rate": 4.229036944380913e-07,
|
510 |
+
"logits/chosen": -1.967760682106018,
|
511 |
+
"logits/rejected": -1.955615758895874,
|
512 |
+
"logps/chosen": -34.05602264404297,
|
513 |
+
"logps/rejected": -33.42683410644531,
|
514 |
+
"loss": 0.4147,
|
515 |
+
"rewards/accuracies": 0.737500011920929,
|
516 |
+
"rewards/chosen": 0.48604661226272583,
|
517 |
+
"rewards/margins": 0.385240375995636,
|
518 |
+
"rewards/rejected": 0.10080619156360626,
|
519 |
"step": 320
|
520 |
},
|
521 |
{
|
522 |
"epoch": 0.86,
|
523 |
"learning_rate": 3.053082288996112e-07,
|
524 |
+
"logits/chosen": -2.003138780593872,
|
525 |
+
"logits/rejected": -2.001786470413208,
|
526 |
+
"logps/chosen": -32.86919403076172,
|
527 |
+
"logps/rejected": -32.247493743896484,
|
528 |
+
"loss": 0.4259,
|
529 |
+
"rewards/accuracies": 0.7124999761581421,
|
530 |
+
"rewards/chosen": 0.5549365878105164,
|
531 |
+
"rewards/margins": 0.3460560441017151,
|
532 |
+
"rewards/rejected": 0.20888061821460724,
|
533 |
"step": 330
|
534 |
},
|
535 |
{
|
536 |
"epoch": 0.88,
|
537 |
"learning_rate": 2.0579377374915805e-07,
|
538 |
+
"logits/chosen": -2.0895984172821045,
|
539 |
+
"logits/rejected": -2.073963165283203,
|
540 |
+
"logps/chosen": -33.46659469604492,
|
541 |
+
"logps/rejected": -32.82307815551758,
|
542 |
+
"loss": 0.4171,
|
543 |
+
"rewards/accuracies": 0.6875,
|
544 |
+
"rewards/chosen": 0.6415189504623413,
|
545 |
+
"rewards/margins": 0.37605759501457214,
|
546 |
+
"rewards/rejected": 0.26546135544776917,
|
547 |
"step": 340
|
548 |
},
|
549 |
{
|
550 |
"epoch": 0.91,
|
551 |
"learning_rate": 1.2518018074041684e-07,
|
552 |
+
"logits/chosen": -1.9623206853866577,
|
553 |
+
"logits/rejected": -1.9614824056625366,
|
554 |
+
"logps/chosen": -32.549312591552734,
|
555 |
+
"logps/rejected": -32.24496078491211,
|
556 |
+
"loss": 0.4,
|
557 |
"rewards/accuracies": 0.7124999761581421,
|
558 |
+
"rewards/chosen": 0.6595046520233154,
|
559 |
+
"rewards/margins": 0.46133819222450256,
|
560 |
+
"rewards/rejected": 0.19816650450229645,
|
561 |
"step": 350
|
562 |
},
|
563 |
{
|
564 |
"epoch": 0.94,
|
565 |
"learning_rate": 6.41315865106129e-08,
|
566 |
+
"logits/chosen": -1.9189163446426392,
|
567 |
+
"logits/rejected": -1.9292027950286865,
|
568 |
+
"logps/chosen": -31.60186767578125,
|
569 |
+
"logps/rejected": -34.987525939941406,
|
570 |
+
"loss": 0.4401,
|
571 |
+
"rewards/accuracies": 0.6499999761581421,
|
572 |
+
"rewards/chosen": 0.5425392985343933,
|
573 |
+
"rewards/margins": 0.27143171429634094,
|
574 |
+
"rewards/rejected": 0.27110758423805237,
|
575 |
"step": 360
|
576 |
},
|
577 |
{
|
578 |
"epoch": 0.96,
|
579 |
"learning_rate": 2.3150941078050325e-08,
|
580 |
+
"logits/chosen": -2.0577945709228516,
|
581 |
+
"logits/rejected": -2.051274061203003,
|
582 |
+
"logps/chosen": -33.05121612548828,
|
583 |
+
"logps/rejected": -28.990320205688477,
|
584 |
+
"loss": 0.4233,
|
585 |
+
"rewards/accuracies": 0.699999988079071,
|
586 |
+
"rewards/chosen": 0.5406805276870728,
|
587 |
+
"rewards/margins": 0.3281847834587097,
|
588 |
+
"rewards/rejected": 0.21249575912952423,
|
589 |
"step": 370
|
590 |
},
|
591 |
{
|
592 |
"epoch": 0.99,
|
593 |
"learning_rate": 2.575864278703266e-09,
|
594 |
+
"logits/chosen": -1.917284607887268,
|
595 |
+
"logits/rejected": -1.9194421768188477,
|
596 |
+
"logps/chosen": -33.59749984741211,
|
597 |
+
"logps/rejected": -30.708057403564453,
|
598 |
+
"loss": 0.4061,
|
599 |
+
"rewards/accuracies": 0.762499988079071,
|
600 |
+
"rewards/chosen": 0.5773938298225403,
|
601 |
+
"rewards/margins": 0.429278701543808,
|
602 |
+
"rewards/rejected": 0.14811506867408752,
|
603 |
"step": 380
|
604 |
},
|
605 |
{
|
606 |
"epoch": 1.0,
|
607 |
"step": 385,
|
608 |
"total_flos": 0.0,
|
609 |
+
"train_loss": 0.447351616079157,
|
610 |
+
"train_runtime": 3253.4458,
|
611 |
+
"train_samples_per_second": 0.946,
|
612 |
"train_steps_per_second": 0.118
|
613 |
}
|
614 |
],
|