hugodk-sch
commited on
Commit
•
4470de4
1
Parent(s):
93f15d0
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: 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: 0.9713
|
21 |
+
- Rewards/chosen: -0.0468
|
22 |
+
- Rewards/rejected: -0.0779
|
23 |
+
- Rewards/accuracies: 0.5282
|
24 |
+
- Rewards/margins: 0.0310
|
25 |
+
- Logps/rejected: -37.6032
|
26 |
+
- Logps/chosen: -34.0866
|
27 |
+
- Logits/rejected: -2.2201
|
28 |
+
- Logits/chosen: -2.2249
|
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 |
+
| 0.8913 | 0.26 | 100 | 0.9845 | -0.0055 | -0.0215 | 0.5195 | 0.0159 | -37.5405 | -34.0407 | -2.2273 | -2.2322 |
|
64 |
+
| 0.7293 | 0.52 | 200 | 0.9602 | -0.0172 | -0.0580 | 0.5714 | 0.0408 | -37.5811 | -34.0537 | -2.2238 | -2.2286 |
|
65 |
+
| 0.6144 | 0.78 | 300 | 0.9713 | -0.0468 | -0.0779 | 0.5282 | 0.0310 | -37.6032 | -34.0866 | -2.2201 | -2.2249 |
|
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 sha256:561acdc4b0687e023339344ce10f08031f158883ab4007975eb3440d742cb9de
|
3 |
size 176183216
|
all_results.json
CHANGED
@@ -1,21 +1,8 @@
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
-
"
|
4 |
-
"
|
5 |
-
"eval_logps/chosen": -34.009239196777344,
|
6 |
-
"eval_logps/rejected": -37.498390197753906,
|
7 |
-
"eval_loss": 0.3797042965888977,
|
8 |
-
"eval_rewards/accuracies": 0.5336378812789917,
|
9 |
-
"eval_rewards/chosen": 0.02277994342148304,
|
10 |
-
"eval_rewards/margins": 0.006376237142831087,
|
11 |
-
"eval_rewards/rejected": 0.01640370488166809,
|
12 |
-
"eval_runtime": 145.6289,
|
13 |
-
"eval_samples": 343,
|
14 |
-
"eval_samples_per_second": 2.355,
|
15 |
-
"eval_steps_per_second": 0.295,
|
16 |
-
"train_loss": 0.31703355428460356,
|
17 |
-
"train_runtime": 3251.5033,
|
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": 0.8125879040012112,
|
4 |
+
"train_runtime": 3249.3137,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
5 |
"train_samples": 3079,
|
6 |
+
"train_samples_per_second": 0.948,
|
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": 0.8125879040012112,
|
4 |
+
"train_runtime": 3249.3137,
|
5 |
"train_samples": 3079,
|
6 |
+
"train_samples_per_second": 0.948,
|
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,
|
@@ -25,155 +25,155 @@
|
|
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":
|
49 |
-
"rewards/margins":
|
50 |
-
"rewards/rejected": 0.
|
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.
|
62 |
-
"rewards/chosen":
|
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":
|
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": 4.987541037542187e-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":
|
91 |
-
"rewards/margins":
|
92 |
-
"rewards/rejected": -0.
|
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.
|
104 |
-
"rewards/chosen": 0.
|
105 |
-
"rewards/margins": 0.
|
106 |
-
"rewards/rejected": -0.
|
107 |
"step": 60
|
108 |
},
|
109 |
{
|
110 |
"epoch": 0.18,
|
111 |
"learning_rate": 4.901618883413549e-06,
|
112 |
-
"logits/chosen": -2.
|
113 |
-
"logits/rejected": -2.
|
114 |
-
"logps/chosen": -
|
115 |
-
"logps/rejected": -36.
|
116 |
-
"loss": 0.
|
117 |
-
"rewards/accuracies": 0.
|
118 |
-
"rewards/chosen": 0.
|
119 |
-
"rewards/margins": 0.
|
120 |
-
"rewards/rejected": -0.
|
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": -0.
|
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": 0.574999988079071,
|
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": -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": 0.
|
174 |
-
"eval_rewards/margins": 0.
|
175 |
-
"eval_rewards/rejected": 0.
|
176 |
-
"eval_runtime": 145.
|
177 |
"eval_samples_per_second": 2.351,
|
178 |
"eval_steps_per_second": 0.295,
|
179 |
"step": 100
|
@@ -181,434 +181,434 @@
|
|
181 |
{
|
182 |
"epoch": 0.29,
|
183 |
"learning_rate": 4.498257201263691e-06,
|
184 |
-
"logits/chosen": -1.
|
185 |
-
"logits/rejected": -1.
|
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": 4.353806263777678e-06,
|
198 |
-
"logits/chosen": -2.
|
199 |
-
"logits/rejected": -
|
200 |
-
"logps/chosen": -32.
|
201 |
-
"logps/rejected": -32.
|
202 |
-
"loss": 0.
|
203 |
"rewards/accuracies": 0.574999988079071,
|
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": -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": 4.0204024186666215e-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": -0.
|
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": -
|
243 |
-
"logps/rejected": -34.
|
244 |
-
"loss": 0.
|
245 |
-
"rewards/accuracies": 0.
|
246 |
-
"rewards/chosen": 0.
|
247 |
-
"rewards/margins": 0.
|
248 |
-
"rewards/rejected": -0.
|
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": -36.
|
257 |
-
"logps/rejected": -32.
|
258 |
-
"loss": 0.
|
259 |
"rewards/accuracies": 0.637499988079071,
|
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": -0.
|
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": -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": 2.996071664294641e-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": 0.
|
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": 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": 2.357,
|
334 |
-
"eval_steps_per_second": 0.
|
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": 0.675000011920929,
|
346 |
-
"rewards/chosen": 0.
|
347 |
-
"rewards/margins": 0.
|
348 |
-
"rewards/rejected": -0.
|
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": -0.
|
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": -33.
|
371 |
-
"logps/rejected": -31.
|
372 |
-
"loss": 0.
|
373 |
-
"rewards/accuracies": 0.
|
374 |
-
"rewards/chosen": 0.
|
375 |
-
"rewards/margins": 0.
|
376 |
-
"rewards/rejected": -0.
|
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": -35.
|
386 |
-
"loss": 0.
|
387 |
-
"rewards/accuracies": 0.
|
388 |
-
"rewards/chosen": 0.
|
389 |
-
"rewards/margins": 0.
|
390 |
-
"rewards/rejected": -0.
|
391 |
"step": 240
|
392 |
},
|
393 |
{
|
394 |
"epoch": 0.65,
|
395 |
"learning_rate": 1.6544367689701824e-06,
|
396 |
-
"logits/chosen": -2.
|
397 |
-
"logits/rejected": -
|
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": -0.
|
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": 0.
|
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": -34.
|
427 |
-
"logps/rejected": -31.
|
428 |
-
"loss": 0.
|
429 |
-
"rewards/accuracies": 0.
|
430 |
-
"rewards/chosen": 0.
|
431 |
-
"rewards/margins": 0.
|
432 |
-
"rewards/rejected": 0.
|
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": -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": 8.737922755071455e-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": 0.
|
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.
|
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": 5.576113578589035e-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": 0.
|
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": -0.
|
519 |
"step": 320
|
520 |
},
|
521 |
{
|
522 |
"epoch": 0.86,
|
523 |
"learning_rate": 3.053082288996112e-07,
|
524 |
-
"logits/chosen": -
|
525 |
-
"logits/rejected": -
|
526 |
-
"logps/chosen": -33.
|
527 |
-
"logps/rejected": -32.
|
528 |
-
"loss": 0.
|
529 |
-
"rewards/accuracies": 0.
|
530 |
-
"rewards/chosen": 0.
|
531 |
-
"rewards/margins": 0.
|
532 |
-
"rewards/rejected": 0.
|
533 |
"step": 330
|
534 |
},
|
535 |
{
|
536 |
"epoch": 0.88,
|
537 |
"learning_rate": 2.0579377374915805e-07,
|
538 |
-
"logits/chosen": -2.
|
539 |
-
"logits/rejected": -2.
|
540 |
-
"logps/chosen": -33.
|
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": 1.2518018074041684e-07,
|
552 |
-
"logits/chosen": -1.
|
553 |
-
"logits/rejected": -1.
|
554 |
-
"logps/chosen": -32.
|
555 |
-
"logps/rejected": -32.
|
556 |
-
"loss": 0.
|
557 |
-
"rewards/accuracies": 0.
|
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-08,
|
566 |
-
"logits/chosen": -1.
|
567 |
-
"logits/rejected": -1.
|
568 |
-
"logps/chosen": -
|
569 |
-
"logps/rejected": -35.
|
570 |
-
"loss": 0.
|
571 |
-
"rewards/accuracies": 0.
|
572 |
-
"rewards/chosen": 0.
|
573 |
-
"rewards/margins": 0.
|
574 |
-
"rewards/rejected": 0.
|
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": -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-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": 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.
|
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": 1.0,
|
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": -1.8667490482330322,
|
29 |
+
"logits/rejected": -1.8710733652114868,
|
30 |
+
"logps/chosen": -36.97007369995117,
|
31 |
+
"logps/rejected": -33.66944885253906,
|
32 |
+
"loss": 0.9317,
|
33 |
+
"rewards/accuracies": 0.5694444179534912,
|
34 |
+
"rewards/chosen": 0.03287407010793686,
|
35 |
+
"rewards/margins": 0.06830974668264389,
|
36 |
+
"rewards/rejected": -0.03543568402528763,
|
37 |
"step": 10
|
38 |
},
|
39 |
{
|
40 |
"epoch": 0.05,
|
41 |
"learning_rate": 2.564102564102564e-06,
|
42 |
+
"logits/chosen": -1.9981460571289062,
|
43 |
+
"logits/rejected": -2.000789165496826,
|
44 |
+
"logps/chosen": -29.641231536865234,
|
45 |
+
"logps/rejected": -29.06744384765625,
|
46 |
+
"loss": 0.9955,
|
47 |
+
"rewards/accuracies": 0.44999998807907104,
|
48 |
+
"rewards/chosen": 0.0008672710391692817,
|
49 |
+
"rewards/margins": 0.004467610269784927,
|
50 |
+
"rewards/rejected": -0.0036003391724079847,
|
51 |
"step": 20
|
52 |
},
|
53 |
{
|
54 |
"epoch": 0.08,
|
55 |
"learning_rate": 3.846153846153847e-06,
|
56 |
+
"logits/chosen": -1.920600175857544,
|
57 |
+
"logits/rejected": -1.917925238609314,
|
58 |
+
"logps/chosen": -31.395061492919922,
|
59 |
+
"logps/rejected": -33.240909576416016,
|
60 |
+
"loss": 0.9609,
|
61 |
+
"rewards/accuracies": 0.6000000238418579,
|
62 |
+
"rewards/chosen": 0.01893135905265808,
|
63 |
+
"rewards/margins": 0.03913776949048042,
|
64 |
+
"rewards/rejected": -0.020206410437822342,
|
65 |
"step": 30
|
66 |
},
|
67 |
{
|
68 |
"epoch": 0.1,
|
69 |
"learning_rate": 4.999896948438434e-06,
|
70 |
+
"logits/chosen": -2.017815113067627,
|
71 |
+
"logits/rejected": -2.0090720653533936,
|
72 |
+
"logps/chosen": -32.5806884765625,
|
73 |
+
"logps/rejected": -32.515098571777344,
|
74 |
+
"loss": 1.0013,
|
75 |
+
"rewards/accuracies": 0.44999998807907104,
|
76 |
+
"rewards/chosen": -0.003494268748909235,
|
77 |
+
"rewards/margins": -0.0013132141903042793,
|
78 |
+
"rewards/rejected": -0.002181055024266243,
|
79 |
"step": 40
|
80 |
},
|
81 |
{
|
82 |
"epoch": 0.13,
|
83 |
"learning_rate": 4.987541037542187e-06,
|
84 |
+
"logits/chosen": -1.8630876541137695,
|
85 |
+
"logits/rejected": -1.8523353338241577,
|
86 |
+
"logps/chosen": -33.549766540527344,
|
87 |
+
"logps/rejected": -35.46318435668945,
|
88 |
+
"loss": 0.9833,
|
89 |
+
"rewards/accuracies": 0.550000011920929,
|
90 |
+
"rewards/chosen": 0.006940312683582306,
|
91 |
+
"rewards/margins": 0.016694897785782814,
|
92 |
+
"rewards/rejected": -0.009754580445587635,
|
93 |
"step": 50
|
94 |
},
|
95 |
{
|
96 |
"epoch": 0.16,
|
97 |
"learning_rate": 4.954691471941119e-06,
|
98 |
+
"logits/chosen": -1.9419746398925781,
|
99 |
+
"logits/rejected": -1.943914771080017,
|
100 |
+
"logps/chosen": -32.527896881103516,
|
101 |
+
"logps/rejected": -33.21547317504883,
|
102 |
+
"loss": 0.9153,
|
103 |
+
"rewards/accuracies": 0.625,
|
104 |
+
"rewards/chosen": 0.06505907326936722,
|
105 |
+
"rewards/margins": 0.10474522411823273,
|
106 |
+
"rewards/rejected": -0.03968615084886551,
|
107 |
"step": 60
|
108 |
},
|
109 |
{
|
110 |
"epoch": 0.18,
|
111 |
"learning_rate": 4.901618883413549e-06,
|
112 |
+
"logits/chosen": -2.07257080078125,
|
113 |
+
"logits/rejected": -2.0775399208068848,
|
114 |
+
"logps/chosen": -34.00202560424805,
|
115 |
+
"logps/rejected": -36.622886657714844,
|
116 |
+
"loss": 0.9636,
|
117 |
+
"rewards/accuracies": 0.512499988079071,
|
118 |
+
"rewards/chosen": -0.012188142165541649,
|
119 |
+
"rewards/margins": 0.03642461448907852,
|
120 |
+
"rewards/rejected": -0.04861275106668472,
|
121 |
"step": 70
|
122 |
},
|
123 |
{
|
124 |
"epoch": 0.21,
|
125 |
"learning_rate": 4.828760511501322e-06,
|
126 |
+
"logits/chosen": -1.9333629608154297,
|
127 |
+
"logits/rejected": -1.9364970922470093,
|
128 |
+
"logps/chosen": -34.302101135253906,
|
129 |
+
"logps/rejected": -34.63160705566406,
|
130 |
+
"loss": 0.862,
|
131 |
+
"rewards/accuracies": 0.5625,
|
132 |
+
"rewards/chosen": 0.1131378561258316,
|
133 |
+
"rewards/margins": 0.1484164148569107,
|
134 |
+
"rewards/rejected": -0.0352785661816597,
|
135 |
"step": 80
|
136 |
},
|
137 |
{
|
138 |
"epoch": 0.23,
|
139 |
"learning_rate": 4.7367166013034295e-06,
|
140 |
+
"logits/chosen": -1.9408857822418213,
|
141 |
+
"logits/rejected": -1.945412039756775,
|
142 |
+
"logps/chosen": -32.36528015136719,
|
143 |
+
"logps/rejected": -32.34526824951172,
|
144 |
+
"loss": 0.9225,
|
145 |
"rewards/accuracies": 0.574999988079071,
|
146 |
+
"rewards/chosen": 0.09192506223917007,
|
147 |
+
"rewards/margins": 0.07964853197336197,
|
148 |
+
"rewards/rejected": 0.012276534922420979,
|
149 |
"step": 90
|
150 |
},
|
151 |
{
|
152 |
"epoch": 0.26,
|
153 |
"learning_rate": 4.626245458345211e-06,
|
154 |
+
"logits/chosen": -2.037550210952759,
|
155 |
+
"logits/rejected": -2.0355725288391113,
|
156 |
+
"logps/chosen": -32.142730712890625,
|
157 |
+
"logps/rejected": -31.29366683959961,
|
158 |
+
"loss": 0.8913,
|
159 |
+
"rewards/accuracies": 0.574999988079071,
|
160 |
+
"rewards/chosen": 0.08829358220100403,
|
161 |
+
"rewards/margins": 0.1130049005150795,
|
162 |
+
"rewards/rejected": -0.02471131458878517,
|
163 |
"step": 100
|
164 |
},
|
165 |
{
|
166 |
"epoch": 0.26,
|
167 |
+
"eval_logits/chosen": -2.232161283493042,
|
168 |
+
"eval_logits/rejected": -2.2273108959198,
|
169 |
+
"eval_logps/chosen": -34.040714263916016,
|
170 |
+
"eval_logps/rejected": -37.54047775268555,
|
171 |
+
"eval_loss": 0.9844526052474976,
|
172 |
+
"eval_rewards/accuracies": 0.5195183157920837,
|
173 |
+
"eval_rewards/chosen": -0.005542654078453779,
|
174 |
+
"eval_rewards/margins": 0.015924591571092606,
|
175 |
+
"eval_rewards/rejected": -0.021467244252562523,
|
176 |
+
"eval_runtime": 145.9018,
|
177 |
"eval_samples_per_second": 2.351,
|
178 |
"eval_steps_per_second": 0.295,
|
179 |
"step": 100
|
|
|
181 |
{
|
182 |
"epoch": 0.29,
|
183 |
"learning_rate": 4.498257201263691e-06,
|
184 |
+
"logits/chosen": -1.9926517009735107,
|
185 |
+
"logits/rejected": -1.9902753829956055,
|
186 |
+
"logps/chosen": -33.12412643432617,
|
187 |
+
"logps/rejected": -34.011417388916016,
|
188 |
+
"loss": 0.9361,
|
189 |
+
"rewards/accuracies": 0.625,
|
190 |
+
"rewards/chosen": 0.10772128403186798,
|
191 |
+
"rewards/margins": 0.09156213700771332,
|
192 |
+
"rewards/rejected": 0.016159160062670708,
|
193 |
"step": 110
|
194 |
},
|
195 |
{
|
196 |
"epoch": 0.31,
|
197 |
"learning_rate": 4.353806263777678e-06,
|
198 |
+
"logits/chosen": -2.00441312789917,
|
199 |
+
"logits/rejected": -1.996093988418579,
|
200 |
+
"logps/chosen": -32.33955383300781,
|
201 |
+
"logps/rejected": -32.13432312011719,
|
202 |
+
"loss": 0.9401,
|
203 |
"rewards/accuracies": 0.574999988079071,
|
204 |
+
"rewards/chosen": 0.09481850266456604,
|
205 |
+
"rewards/margins": 0.07035262137651443,
|
206 |
+
"rewards/rejected": 0.024465877562761307,
|
207 |
"step": 120
|
208 |
},
|
209 |
{
|
210 |
"epoch": 0.34,
|
211 |
"learning_rate": 4.1940827077152755e-06,
|
212 |
+
"logits/chosen": -2.0318965911865234,
|
213 |
+
"logits/rejected": -2.023927688598633,
|
214 |
+
"logps/chosen": -30.336984634399414,
|
215 |
+
"logps/rejected": -32.0634765625,
|
216 |
+
"loss": 0.9061,
|
217 |
+
"rewards/accuracies": 0.550000011920929,
|
218 |
+
"rewards/chosen": 0.12486012279987335,
|
219 |
+
"rewards/margins": 0.1342838853597641,
|
220 |
+
"rewards/rejected": -0.009423775598406792,
|
221 |
"step": 130
|
222 |
},
|
223 |
{
|
224 |
"epoch": 0.36,
|
225 |
"learning_rate": 4.0204024186666215e-06,
|
226 |
+
"logits/chosen": -1.9620994329452515,
|
227 |
+
"logits/rejected": -1.9723354578018188,
|
228 |
+
"logps/chosen": -31.222240447998047,
|
229 |
+
"logps/rejected": -32.57916259765625,
|
230 |
+
"loss": 0.795,
|
231 |
+
"rewards/accuracies": 0.6499999761581421,
|
232 |
+
"rewards/chosen": 0.17767605185508728,
|
233 |
+
"rewards/margins": 0.21753115952014923,
|
234 |
+
"rewards/rejected": -0.03985511139035225,
|
235 |
"step": 140
|
236 |
},
|
237 |
{
|
238 |
"epoch": 0.39,
|
239 |
"learning_rate": 3.834196265035119e-06,
|
240 |
+
"logits/chosen": -1.8727748394012451,
|
241 |
+
"logits/rejected": -1.87395441532135,
|
242 |
+
"logps/chosen": -33.931861877441406,
|
243 |
+
"logps/rejected": -34.79869842529297,
|
244 |
+
"loss": 0.7946,
|
245 |
+
"rewards/accuracies": 0.6000000238418579,
|
246 |
+
"rewards/chosen": 0.23474939167499542,
|
247 |
+
"rewards/margins": 0.272051066160202,
|
248 |
+
"rewards/rejected": -0.03730170056223869,
|
249 |
"step": 150
|
250 |
},
|
251 |
{
|
252 |
"epoch": 0.42,
|
253 |
"learning_rate": 3.636998309800573e-06,
|
254 |
+
"logits/chosen": -1.9248745441436768,
|
255 |
+
"logits/rejected": -1.9214649200439453,
|
256 |
+
"logps/chosen": -36.014469146728516,
|
257 |
+
"logps/rejected": -32.73783493041992,
|
258 |
+
"loss": 0.8532,
|
259 |
"rewards/accuracies": 0.637499988079071,
|
260 |
+
"rewards/chosen": 0.14148668944835663,
|
261 |
+
"rewards/margins": 0.15506146848201752,
|
262 |
+
"rewards/rejected": -0.013574766926467419,
|
263 |
"step": 160
|
264 |
},
|
265 |
{
|
266 |
"epoch": 0.44,
|
267 |
"learning_rate": 3.4304331721118078e-06,
|
268 |
+
"logits/chosen": -2.025555372238159,
|
269 |
+
"logits/rejected": -2.01819109916687,
|
270 |
+
"logps/chosen": -33.50218200683594,
|
271 |
+
"logps/rejected": -31.41971206665039,
|
272 |
+
"loss": 0.7292,
|
273 |
+
"rewards/accuracies": 0.699999988079071,
|
274 |
+
"rewards/chosen": 0.26611366868019104,
|
275 |
+
"rewards/margins": 0.3271873891353607,
|
276 |
+
"rewards/rejected": -0.06107370927929878,
|
277 |
"step": 170
|
278 |
},
|
279 |
{
|
280 |
"epoch": 0.47,
|
281 |
"learning_rate": 3.2162026428305436e-06,
|
282 |
+
"logits/chosen": -2.0320167541503906,
|
283 |
+
"logits/rejected": -2.037261486053467,
|
284 |
+
"logps/chosen": -32.24850845336914,
|
285 |
+
"logps/rejected": -32.45344924926758,
|
286 |
+
"loss": 0.7865,
|
287 |
+
"rewards/accuracies": 0.637499988079071,
|
288 |
+
"rewards/chosen": 0.2666531801223755,
|
289 |
+
"rewards/margins": 0.23813048005104065,
|
290 |
+
"rewards/rejected": 0.028522688895463943,
|
291 |
"step": 180
|
292 |
},
|
293 |
{
|
294 |
"epoch": 0.49,
|
295 |
"learning_rate": 2.996071664294641e-06,
|
296 |
+
"logits/chosen": -2.032525062561035,
|
297 |
+
"logits/rejected": -2.0297436714172363,
|
298 |
+
"logps/chosen": -31.313217163085938,
|
299 |
+
"logps/rejected": -31.349472045898438,
|
300 |
+
"loss": 0.8387,
|
301 |
+
"rewards/accuracies": 0.637499988079071,
|
302 |
+
"rewards/chosen": 0.1581628918647766,
|
303 |
+
"rewards/margins": 0.18629543483257294,
|
304 |
+
"rewards/rejected": -0.02813255414366722,
|
305 |
"step": 190
|
306 |
},
|
307 |
{
|
308 |
"epoch": 0.52,
|
309 |
"learning_rate": 2.7718537898066833e-06,
|
310 |
+
"logits/chosen": -1.902632713317871,
|
311 |
+
"logits/rejected": -1.907284140586853,
|
312 |
+
"logps/chosen": -31.320043563842773,
|
313 |
+
"logps/rejected": -32.85698699951172,
|
314 |
+
"loss": 0.7293,
|
315 |
+
"rewards/accuracies": 0.7124999761581421,
|
316 |
+
"rewards/chosen": 0.2576131224632263,
|
317 |
+
"rewards/margins": 0.31706100702285767,
|
318 |
+
"rewards/rejected": -0.05944784730672836,
|
319 |
"step": 200
|
320 |
},
|
321 |
{
|
322 |
"epoch": 0.52,
|
323 |
+
"eval_logits/chosen": -2.228637933731079,
|
324 |
+
"eval_logits/rejected": -2.2238004207611084,
|
325 |
+
"eval_logps/chosen": -34.053680419921875,
|
326 |
+
"eval_logps/rejected": -37.581058502197266,
|
327 |
+
"eval_loss": 0.9601577520370483,
|
328 |
+
"eval_rewards/accuracies": 0.5714285373687744,
|
329 |
+
"eval_rewards/chosen": -0.01721162348985672,
|
330 |
+
"eval_rewards/margins": 0.040783192962408066,
|
331 |
+
"eval_rewards/rejected": -0.057994820177555084,
|
332 |
+
"eval_runtime": 145.5388,
|
333 |
"eval_samples_per_second": 2.357,
|
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": -2.0149245262145996,
|
341 |
+
"logits/rejected": -2.025560140609741,
|
342 |
+
"logps/chosen": -31.77438735961914,
|
343 |
+
"logps/rejected": -33.95419692993164,
|
344 |
+
"loss": 0.7666,
|
345 |
"rewards/accuracies": 0.675000011920929,
|
346 |
+
"rewards/chosen": 0.17722666263580322,
|
347 |
+
"rewards/margins": 0.25957340002059937,
|
348 |
+
"rewards/rejected": -0.08234670013189316,
|
349 |
"step": 210
|
350 |
},
|
351 |
{
|
352 |
"epoch": 0.57,
|
353 |
"learning_rate": 2.3185646976551794e-06,
|
354 |
+
"logits/chosen": -1.906951904296875,
|
355 |
+
"logits/rejected": -1.9217418432235718,
|
356 |
+
"logps/chosen": -29.83829116821289,
|
357 |
+
"logps/rejected": -31.636096954345703,
|
358 |
+
"loss": 0.7204,
|
359 |
+
"rewards/accuracies": 0.737500011920929,
|
360 |
+
"rewards/chosen": 0.2459266185760498,
|
361 |
+
"rewards/margins": 0.3160557448863983,
|
362 |
+
"rewards/rejected": -0.07012919336557388,
|
363 |
"step": 220
|
364 |
},
|
365 |
{
|
366 |
"epoch": 0.6,
|
367 |
"learning_rate": 2.0932279108998323e-06,
|
368 |
+
"logits/chosen": -1.9629713296890259,
|
369 |
+
"logits/rejected": -1.9669532775878906,
|
370 |
+
"logps/chosen": -33.124656677246094,
|
371 |
+
"logps/rejected": -31.630443572998047,
|
372 |
+
"loss": 0.7348,
|
373 |
+
"rewards/accuracies": 0.699999988079071,
|
374 |
+
"rewards/chosen": 0.2635299265384674,
|
375 |
+
"rewards/margins": 0.3365571200847626,
|
376 |
+
"rewards/rejected": -0.07302714884281158,
|
377 |
"step": 230
|
378 |
},
|
379 |
{
|
380 |
"epoch": 0.62,
|
381 |
"learning_rate": 1.8712423238279358e-06,
|
382 |
+
"logits/chosen": -1.9614177942276,
|
383 |
+
"logits/rejected": -1.9395818710327148,
|
384 |
+
"logps/chosen": -33.87095260620117,
|
385 |
+
"logps/rejected": -35.10104751586914,
|
386 |
+
"loss": 0.6865,
|
387 |
+
"rewards/accuracies": 0.7250000238418579,
|
388 |
+
"rewards/chosen": 0.254514217376709,
|
389 |
+
"rewards/margins": 0.38974156975746155,
|
390 |
+
"rewards/rejected": -0.13522735238075256,
|
391 |
"step": 240
|
392 |
},
|
393 |
{
|
394 |
"epoch": 0.65,
|
395 |
"learning_rate": 1.6544367689701824e-06,
|
396 |
+
"logits/chosen": -2.0029492378234863,
|
397 |
+
"logits/rejected": -1.999629020690918,
|
398 |
+
"logps/chosen": -32.730865478515625,
|
399 |
+
"logps/rejected": -36.28009796142578,
|
400 |
+
"loss": 0.8055,
|
401 |
+
"rewards/accuracies": 0.6625000238418579,
|
402 |
+
"rewards/chosen": 0.17478153109550476,
|
403 |
+
"rewards/margins": 0.2244400531053543,
|
404 |
+
"rewards/rejected": -0.04965851828455925,
|
405 |
"step": 250
|
406 |
},
|
407 |
{
|
408 |
"epoch": 0.68,
|
409 |
"learning_rate": 1.4445974030621963e-06,
|
410 |
+
"logits/chosen": -1.8703190088272095,
|
411 |
+
"logits/rejected": -1.8679043054580688,
|
412 |
+
"logps/chosen": -33.98231887817383,
|
413 |
+
"logps/rejected": -35.54644775390625,
|
414 |
+
"loss": 0.8042,
|
415 |
+
"rewards/accuracies": 0.7250000238418579,
|
416 |
+
"rewards/chosen": 0.18547315895557404,
|
417 |
+
"rewards/margins": 0.22311437129974365,
|
418 |
+
"rewards/rejected": -0.03764120861887932,
|
419 |
"step": 260
|
420 |
},
|
421 |
{
|
422 |
"epoch": 0.7,
|
423 |
"learning_rate": 1.243452991757889e-06,
|
424 |
+
"logits/chosen": -1.8551464080810547,
|
425 |
+
"logits/rejected": -1.852746605873108,
|
426 |
+
"logps/chosen": -34.20850372314453,
|
427 |
+
"logps/rejected": -31.803356170654297,
|
428 |
+
"loss": 0.8096,
|
429 |
+
"rewards/accuracies": 0.675000011920929,
|
430 |
+
"rewards/chosen": 0.170187309384346,
|
431 |
+
"rewards/margins": 0.19900819659233093,
|
432 |
+
"rewards/rejected": -0.02882089652121067,
|
433 |
"step": 270
|
434 |
},
|
435 |
{
|
436 |
"epoch": 0.73,
|
437 |
"learning_rate": 1.0526606671603523e-06,
|
438 |
+
"logits/chosen": -1.9582526683807373,
|
439 |
+
"logits/rejected": -1.947749376296997,
|
440 |
+
"logps/chosen": -35.0114631652832,
|
441 |
+
"logps/rejected": -31.88564682006836,
|
442 |
+
"loss": 0.692,
|
443 |
+
"rewards/accuracies": 0.737500011920929,
|
444 |
+
"rewards/chosen": 0.3106640875339508,
|
445 |
+
"rewards/margins": 0.3508565425872803,
|
446 |
+
"rewards/rejected": -0.040192440152168274,
|
447 |
"step": 280
|
448 |
},
|
449 |
{
|
450 |
"epoch": 0.75,
|
451 |
"learning_rate": 8.737922755071455e-07,
|
452 |
+
"logits/chosen": -2.053699493408203,
|
453 |
+
"logits/rejected": -2.038789749145508,
|
454 |
+
"logps/chosen": -30.727243423461914,
|
455 |
+
"logps/rejected": -32.641685485839844,
|
456 |
+
"loss": 0.8827,
|
457 |
+
"rewards/accuracies": 0.574999988079071,
|
458 |
+
"rewards/chosen": 0.17719359695911407,
|
459 |
+
"rewards/margins": 0.16805905103683472,
|
460 |
+
"rewards/rejected": 0.009134533815085888,
|
461 |
"step": 290
|
462 |
},
|
463 |
{
|
464 |
"epoch": 0.78,
|
465 |
"learning_rate": 7.08321427484816e-07,
|
466 |
+
"logits/chosen": -1.9243850708007812,
|
467 |
+
"logits/rejected": -1.9218356609344482,
|
468 |
+
"logps/chosen": -32.43050003051758,
|
469 |
+
"logps/rejected": -30.8950138092041,
|
470 |
+
"loss": 0.6144,
|
471 |
+
"rewards/accuracies": 0.7749999761581421,
|
472 |
+
"rewards/chosen": 0.446951299905777,
|
473 |
+
"rewards/margins": 0.5173346400260925,
|
474 |
+
"rewards/rejected": -0.07038338482379913,
|
475 |
"step": 300
|
476 |
},
|
477 |
{
|
478 |
"epoch": 0.78,
|
479 |
+
"eval_logits/chosen": -2.2248916625976562,
|
480 |
+
"eval_logits/rejected": -2.2200686931610107,
|
481 |
+
"eval_logps/chosen": -34.08660125732422,
|
482 |
+
"eval_logps/rejected": -37.603153228759766,
|
483 |
+
"eval_loss": 0.9712583422660828,
|
484 |
+
"eval_rewards/accuracies": 0.5282392501831055,
|
485 |
+
"eval_rewards/chosen": -0.046843186020851135,
|
486 |
+
"eval_rewards/margins": 0.031036507338285446,
|
487 |
+
"eval_rewards/rejected": -0.07787969708442688,
|
488 |
+
"eval_runtime": 145.8399,
|
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": -1.9094560146331787,
|
497 |
+
"logits/rejected": -1.9062097072601318,
|
498 |
+
"logps/chosen": -31.33370018005371,
|
499 |
+
"logps/rejected": -33.83475112915039,
|
500 |
+
"loss": 0.728,
|
501 |
+
"rewards/accuracies": 0.737500011920929,
|
502 |
+
"rewards/chosen": 0.24268333613872528,
|
503 |
+
"rewards/margins": 0.32601919770240784,
|
504 |
+
"rewards/rejected": -0.08333584666252136,
|
505 |
"step": 310
|
506 |
},
|
507 |
{
|
508 |
"epoch": 0.83,
|
509 |
"learning_rate": 4.229036944380913e-07,
|
510 |
+
"logits/chosen": -1.9593706130981445,
|
511 |
+
"logits/rejected": -1.9471466541290283,
|
512 |
+
"logps/chosen": -34.3392333984375,
|
513 |
+
"logps/rejected": -33.68544387817383,
|
514 |
+
"loss": 0.6954,
|
515 |
+
"rewards/accuracies": 0.6875,
|
516 |
+
"rewards/chosen": 0.2311573475599289,
|
517 |
+
"rewards/margins": 0.36310091614723206,
|
518 |
+
"rewards/rejected": -0.13194358348846436,
|
519 |
"step": 320
|
520 |
},
|
521 |
{
|
522 |
"epoch": 0.86,
|
523 |
"learning_rate": 3.053082288996112e-07,
|
524 |
+
"logits/chosen": -1.9940307140350342,
|
525 |
+
"logits/rejected": -1.9925845861434937,
|
526 |
+
"logps/chosen": -33.20854568481445,
|
527 |
+
"logps/rejected": -32.552764892578125,
|
528 |
+
"loss": 0.7359,
|
529 |
+
"rewards/accuracies": 0.762499988079071,
|
530 |
+
"rewards/chosen": 0.24952416121959686,
|
531 |
+
"rewards/margins": 0.31538745760917664,
|
532 |
+
"rewards/rejected": -0.06586329638957977,
|
533 |
"step": 330
|
534 |
},
|
535 |
{
|
536 |
"epoch": 0.88,
|
537 |
"learning_rate": 2.0579377374915805e-07,
|
538 |
+
"logits/chosen": -2.080763578414917,
|
539 |
+
"logits/rejected": -2.065063953399658,
|
540 |
+
"logps/chosen": -33.82006072998047,
|
541 |
+
"logps/rejected": -33.105167388916016,
|
542 |
+
"loss": 0.7407,
|
543 |
+
"rewards/accuracies": 0.6875,
|
544 |
+
"rewards/chosen": 0.32339948415756226,
|
545 |
+
"rewards/margins": 0.31181785464286804,
|
546 |
+
"rewards/rejected": 0.011581619270145893,
|
547 |
"step": 340
|
548 |
},
|
549 |
{
|
550 |
"epoch": 0.91,
|
551 |
"learning_rate": 1.2518018074041684e-07,
|
552 |
+
"logits/chosen": -1.9535402059555054,
|
553 |
+
"logits/rejected": -1.9526821374893188,
|
554 |
+
"logps/chosen": -32.8734130859375,
|
555 |
+
"logps/rejected": -32.565185546875,
|
556 |
+
"loss": 0.645,
|
557 |
+
"rewards/accuracies": 0.7124999761581421,
|
558 |
+
"rewards/chosen": 0.36781299114227295,
|
559 |
+
"rewards/margins": 0.4578477442264557,
|
560 |
+
"rewards/rejected": -0.09003473073244095,
|
561 |
"step": 350
|
562 |
},
|
563 |
{
|
564 |
"epoch": 0.94,
|
565 |
"learning_rate": 6.41315865106129e-08,
|
566 |
+
"logits/chosen": -1.909014344215393,
|
567 |
+
"logits/rejected": -1.919298768043518,
|
568 |
+
"logps/chosen": -31.879894256591797,
|
569 |
+
"logps/rejected": -35.34550857543945,
|
570 |
+
"loss": 0.6942,
|
571 |
+
"rewards/accuracies": 0.7250000238418579,
|
572 |
+
"rewards/chosen": 0.29231759905815125,
|
573 |
+
"rewards/margins": 0.34339430928230286,
|
574 |
+
"rewards/rejected": -0.051076728850603104,
|
575 |
"step": 360
|
576 |
},
|
577 |
{
|
578 |
"epoch": 0.96,
|
579 |
"learning_rate": 2.3150941078050325e-08,
|
580 |
+
"logits/chosen": -2.0480055809020996,
|
581 |
+
"logits/rejected": -2.0415189266204834,
|
582 |
+
"logps/chosen": -33.37665939331055,
|
583 |
+
"logps/rejected": -29.24251937866211,
|
584 |
+
"loss": 0.7583,
|
585 |
+
"rewards/accuracies": 0.7124999761581421,
|
586 |
+
"rewards/chosen": 0.24777868390083313,
|
587 |
+
"rewards/margins": 0.262265145778656,
|
588 |
+
"rewards/rejected": -0.014486486092209816,
|
589 |
"step": 370
|
590 |
},
|
591 |
{
|
592 |
"epoch": 0.99,
|
593 |
"learning_rate": 2.575864278703266e-09,
|
594 |
+
"logits/chosen": -1.9075886011123657,
|
595 |
+
"logits/rejected": -1.9097877740859985,
|
596 |
+
"logps/chosen": -33.8558464050293,
|
597 |
+
"logps/rejected": -30.982568740844727,
|
598 |
+
"loss": 0.6567,
|
599 |
+
"rewards/accuracies": 0.737500011920929,
|
600 |
+
"rewards/chosen": 0.3448841869831085,
|
601 |
+
"rewards/margins": 0.44383174180984497,
|
602 |
+
"rewards/rejected": -0.09894753992557526,
|
603 |
"step": 380
|
604 |
},
|
605 |
{
|
606 |
"epoch": 1.0,
|
607 |
"step": 385,
|
608 |
"total_flos": 0.0,
|
609 |
+
"train_loss": 0.8125879040012112,
|
610 |
+
"train_runtime": 3249.3137,
|
611 |
+
"train_samples_per_second": 0.948,
|
612 |
"train_steps_per_second": 0.118
|
613 |
}
|
614 |
],
|