hugodk-sch
commited on
Commit
•
599e31f
1
Parent(s):
980d604
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 +376 -376
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: 0.4345
|
21 |
+
- Rewards/chosen: 0.0086
|
22 |
+
- Rewards/rejected: 0.0130
|
23 |
+
- Rewards/accuracies: 0.4809
|
24 |
+
- Rewards/margins: -0.0044
|
25 |
+
- Logps/rejected: -37.5004
|
26 |
+
- Logps/chosen: -34.0238
|
27 |
+
- Logits/rejected: -2.2387
|
28 |
+
- Logits/chosen: -2.2436
|
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 |
+
| 0.3805 | 0.26 | 100 | 0.4061 | 0.0139 | 0.0017 | 0.5395 | 0.0122 | -37.5145 | -34.0171 | -2.2393 | -2.2441 |
|
64 |
+
| 0.371 | 0.52 | 200 | 0.4323 | 0.0091 | 0.0118 | 0.5253 | -0.0027 | -37.5019 | -34.0231 | -2.2390 | -2.2439 |
|
65 |
+
| 0.3839 | 0.78 | 300 | 0.4345 | 0.0086 | 0.0130 | 0.4809 | -0.0044 | -37.5004 | -34.0238 | -2.2387 | -2.2436 |
|
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:dd4d4dffa9b30971c702f1f39924c707a8b1925a5d67f8de841d1128fe03486b
|
3 |
size 176183216
|
all_results.json
CHANGED
@@ -1,21 +1,8 @@
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
-
"
|
4 |
-
"
|
5 |
-
"eval_logps/chosen": -34.03290939331055,
|
6 |
-
"eval_logps/rejected": -37.49677658081055,
|
7 |
-
"eval_loss": 0.5752949118614197,
|
8 |
-
"eval_rewards/accuracies": 0.45764121413230896,
|
9 |
-
"eval_rewards/chosen": 0.001152246491983533,
|
10 |
-
"eval_rewards/margins": -0.012740416452288628,
|
11 |
-
"eval_rewards/rejected": 0.013892664574086666,
|
12 |
-
"eval_runtime": 145.7802,
|
13 |
-
"eval_samples": 343,
|
14 |
-
"eval_samples_per_second": 2.353,
|
15 |
-
"eval_steps_per_second": 0.295,
|
16 |
-
"train_loss": 0.5317366457604743,
|
17 |
-
"train_runtime": 3254.062,
|
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": 0.4092799360101873,
|
4 |
+
"train_runtime": 3256.9261,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
5 |
"train_samples": 3079,
|
6 |
+
"train_samples_per_second": 0.945,
|
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": 0.4092799360101873,
|
4 |
+
"train_runtime": 3256.9261,
|
5 |
"train_samples": 3079,
|
6 |
+
"train_samples_per_second": 0.945,
|
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": 1.2820512820512818e-07,
|
28 |
-
"logits/chosen": -1.
|
29 |
-
"logits/rejected": -1.
|
30 |
-
"logps/chosen": -36.
|
31 |
-
"logps/rejected": -33.
|
32 |
-
"loss": 0.
|
33 |
"rewards/accuracies": 0.5416666865348816,
|
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": 0.
|
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": -
|
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-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":
|
161 |
-
"rewards/margins":
|
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": 146.
|
177 |
-
"eval_samples_per_second": 2.
|
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.
|
185 |
-
"logits/rejected": -2.
|
186 |
-
"logps/chosen": -33.
|
187 |
-
"logps/rejected": -
|
188 |
-
"loss": 0.
|
189 |
-
"rewards/accuracies": 0.
|
190 |
-
"rewards/chosen": 0.
|
191 |
-
"rewards/margins": 0.
|
192 |
-
"rewards/rejected":
|
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": -0.
|
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": 0.
|
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.
|
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":
|
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": 0.
|
304 |
-
"rewards/rejected":
|
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": -0.
|
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": -0.
|
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.
|
388 |
-
"rewards/chosen": 0.
|
389 |
-
"rewards/margins": 0.
|
390 |
-
"rewards/rejected": -0.
|
391 |
"step": 240
|
392 |
},
|
393 |
{
|
394 |
"epoch": 0.65,
|
395 |
"learning_rate": 1.654436768970182e-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": 0.
|
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.
|
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": 0.
|
431 |
-
"rewards/margins": -0.
|
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":
|
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": 0.
|
459 |
-
"rewards/margins": 0.
|
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": 0.4808970093727112,
|
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": -0.
|
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.
|
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.
|
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": 0.48750001192092896,
|
586 |
-
"rewards/chosen":
|
587 |
-
"rewards/margins":
|
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.
|
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": 0.3906,
|
19 |
"rewards/accuracies": 0.0,
|
20 |
"rewards/chosen": 0.0,
|
21 |
"rewards/margins": 0.0,
|
|
|
25 |
{
|
26 |
"epoch": 0.03,
|
27 |
"learning_rate": 1.2820512820512818e-07,
|
28 |
+
"logits/chosen": -1.8662992715835571,
|
29 |
+
"logits/rejected": -1.8706117868423462,
|
30 |
+
"logps/chosen": -36.97681427001953,
|
31 |
+
"logps/rejected": -33.66523361206055,
|
32 |
+
"loss": 0.3447,
|
33 |
"rewards/accuracies": 0.5416666865348816,
|
34 |
+
"rewards/chosen": 0.023829741403460503,
|
35 |
+
"rewards/margins": 0.05195777863264084,
|
36 |
+
"rewards/rejected": -0.028128040954470634,
|
37 |
"step": 10
|
38 |
},
|
39 |
{
|
40 |
"epoch": 0.05,
|
41 |
"learning_rate": 2.5641025641025636e-07,
|
42 |
+
"logits/chosen": -1.997193694114685,
|
43 |
+
"logits/rejected": -1.9998573064804077,
|
44 |
+
"logps/chosen": -29.65359878540039,
|
45 |
+
"logps/rejected": -29.054311752319336,
|
46 |
+
"loss": 0.4534,
|
47 |
+
"rewards/accuracies": 0.4375,
|
48 |
+
"rewards/chosen": -0.009124360978603363,
|
49 |
+
"rewards/margins": -0.016431041061878204,
|
50 |
+
"rewards/rejected": 0.007306680083274841,
|
51 |
"step": 20
|
52 |
},
|
53 |
{
|
54 |
"epoch": 0.08,
|
55 |
"learning_rate": 3.8461538461538463e-07,
|
56 |
+
"logits/chosen": -1.9197280406951904,
|
57 |
+
"logits/rejected": -1.9170429706573486,
|
58 |
+
"logps/chosen": -31.414775848388672,
|
59 |
+
"logps/rejected": -33.24064254760742,
|
60 |
+
"loss": 0.4025,
|
61 |
+
"rewards/accuracies": 0.5625,
|
62 |
+
"rewards/chosen": 0.0010542668169364333,
|
63 |
+
"rewards/margins": 0.018799806013703346,
|
64 |
+
"rewards/rejected": -0.01774553768336773,
|
65 |
"step": 30
|
66 |
},
|
67 |
{
|
68 |
"epoch": 0.1,
|
69 |
"learning_rate": 4.999896948438433e-07,
|
70 |
+
"logits/chosen": -2.0171010494232178,
|
71 |
+
"logits/rejected": -2.008350372314453,
|
72 |
+
"logps/chosen": -32.59648513793945,
|
73 |
+
"logps/rejected": -32.50862121582031,
|
74 |
+
"loss": 0.4532,
|
75 |
+
"rewards/accuracies": 0.44999998807907104,
|
76 |
+
"rewards/chosen": -0.015740757808089256,
|
77 |
+
"rewards/margins": -0.018979396671056747,
|
78 |
+
"rewards/rejected": 0.0032386414241045713,
|
79 |
"step": 40
|
80 |
},
|
81 |
{
|
82 |
"epoch": 0.13,
|
83 |
"learning_rate": 4.987541037542186e-07,
|
84 |
+
"logits/chosen": -1.8645904064178467,
|
85 |
+
"logits/rejected": -1.8538070917129517,
|
86 |
+
"logps/chosen": -33.542823791503906,
|
87 |
+
"logps/rejected": -35.43744659423828,
|
88 |
+
"loss": 0.4304,
|
89 |
+
"rewards/accuracies": 0.5,
|
90 |
+
"rewards/chosen": 0.011724205687642097,
|
91 |
+
"rewards/margins": -0.00019102543592453003,
|
92 |
+
"rewards/rejected": 0.011915231123566628,
|
93 |
"step": 50
|
94 |
},
|
95 |
{
|
96 |
"epoch": 0.16,
|
97 |
"learning_rate": 4.954691471941118e-07,
|
98 |
+
"logits/chosen": -1.9461469650268555,
|
99 |
+
"logits/rejected": -1.948094129562378,
|
100 |
+
"logps/chosen": -32.587215423583984,
|
101 |
+
"logps/rejected": -33.21445846557617,
|
102 |
+
"loss": 0.3539,
|
103 |
+
"rewards/accuracies": 0.637499988079071,
|
104 |
+
"rewards/chosen": 0.010374858975410461,
|
105 |
+
"rewards/margins": 0.04483799636363983,
|
106 |
+
"rewards/rejected": -0.03446313738822937,
|
107 |
"step": 60
|
108 |
},
|
109 |
{
|
110 |
"epoch": 0.18,
|
111 |
"learning_rate": 4.901618883413548e-07,
|
112 |
+
"logits/chosen": -2.079591989517212,
|
113 |
+
"logits/rejected": -2.0845742225646973,
|
114 |
+
"logps/chosen": -34.00908279418945,
|
115 |
+
"logps/rejected": -36.58150863647461,
|
116 |
+
"loss": 0.4506,
|
117 |
+
"rewards/accuracies": 0.512499988079071,
|
118 |
+
"rewards/chosen": -0.01648041605949402,
|
119 |
+
"rewards/margins": -0.006370754446834326,
|
120 |
+
"rewards/rejected": -0.010109663009643555,
|
121 |
"step": 70
|
122 |
},
|
123 |
{
|
124 |
"epoch": 0.21,
|
125 |
"learning_rate": 4.828760511501322e-07,
|
126 |
+
"logits/chosen": -1.9423307180404663,
|
127 |
+
"logits/rejected": -1.9454963207244873,
|
128 |
+
"logps/chosen": -34.39698028564453,
|
129 |
+
"logps/rejected": -34.59761428833008,
|
130 |
+
"loss": 0.3889,
|
131 |
+
"rewards/accuracies": 0.5249999761581421,
|
132 |
+
"rewards/chosen": 0.024663064628839493,
|
133 |
+
"rewards/margins": 0.028829574584960938,
|
134 |
+
"rewards/rejected": -0.00416650902479887,
|
135 |
"step": 80
|
136 |
},
|
137 |
{
|
138 |
"epoch": 0.23,
|
139 |
"learning_rate": 4.736716601303429e-07,
|
140 |
+
"logits/chosen": -1.9508873224258423,
|
141 |
+
"logits/rejected": -1.9553953409194946,
|
142 |
+
"logps/chosen": -32.48583221435547,
|
143 |
+
"logps/rejected": -32.35867691040039,
|
144 |
+
"loss": 0.4445,
|
145 |
+
"rewards/accuracies": 0.4625000059604645,
|
146 |
+
"rewards/chosen": -0.014726865105330944,
|
147 |
+
"rewards/margins": -0.014909917488694191,
|
148 |
+
"rewards/rejected": 0.00018304325931239873,
|
149 |
"step": 90
|
150 |
},
|
151 |
{
|
152 |
"epoch": 0.26,
|
153 |
"learning_rate": 4.62624545834521e-07,
|
154 |
+
"logits/chosen": -2.0491719245910645,
|
155 |
+
"logits/rejected": -2.047179937362671,
|
156 |
+
"logps/chosen": -32.22635269165039,
|
157 |
+
"logps/rejected": -31.287487030029297,
|
158 |
+
"loss": 0.3805,
|
159 |
+
"rewards/accuracies": 0.550000011920929,
|
160 |
+
"rewards/chosen": 0.011584864929318428,
|
161 |
+
"rewards/margins": 0.02860759198665619,
|
162 |
+
"rewards/rejected": -0.01702272891998291,
|
163 |
"step": 100
|
164 |
},
|
165 |
{
|
166 |
"epoch": 0.26,
|
167 |
+
"eval_logits/chosen": -2.244148015975952,
|
168 |
+
"eval_logits/rejected": -2.2392663955688477,
|
169 |
+
"eval_logps/chosen": -34.01713943481445,
|
170 |
+
"eval_logps/rejected": -37.514495849609375,
|
171 |
+
"eval_loss": 0.4061162769794464,
|
172 |
+
"eval_rewards/accuracies": 0.5394518375396729,
|
173 |
+
"eval_rewards/chosen": 0.013928660191595554,
|
174 |
+
"eval_rewards/margins": 0.012230273336172104,
|
175 |
+
"eval_rewards/rejected": 0.0016983875539153814,
|
176 |
+
"eval_runtime": 146.1091,
|
177 |
+
"eval_samples_per_second": 2.348,
|
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": -2.005580186843872,
|
185 |
+
"logits/rejected": -2.0031564235687256,
|
186 |
+
"logps/chosen": -33.24415588378906,
|
187 |
+
"logps/rejected": -33.99993133544922,
|
188 |
+
"loss": 0.4792,
|
189 |
+
"rewards/accuracies": 0.4375,
|
190 |
+
"rewards/chosen": -0.00026987865567207336,
|
191 |
+
"rewards/margins": -0.02382112666964531,
|
192 |
+
"rewards/rejected": 0.023551244288682938,
|
193 |
"step": 110
|
194 |
},
|
195 |
{
|
196 |
"epoch": 0.31,
|
197 |
"learning_rate": 4.353806263777677e-07,
|
198 |
+
"logits/chosen": -2.01680326461792,
|
199 |
+
"logits/rejected": -2.0084121227264404,
|
200 |
+
"logps/chosen": -32.45465850830078,
|
201 |
+
"logps/rejected": -32.1729850769043,
|
202 |
+
"loss": 0.4286,
|
203 |
+
"rewards/accuracies": 0.4375,
|
204 |
+
"rewards/chosen": -0.007803081069141626,
|
205 |
+
"rewards/margins": 0.0013800703454762697,
|
206 |
+
"rewards/rejected": -0.009183152578771114,
|
207 |
"step": 120
|
208 |
},
|
209 |
{
|
210 |
"epoch": 0.34,
|
211 |
"learning_rate": 4.194082707715275e-07,
|
212 |
+
"logits/chosen": -2.0466856956481934,
|
213 |
+
"logits/rejected": -2.0386409759521484,
|
214 |
+
"logps/chosen": -30.494335174560547,
|
215 |
+
"logps/rejected": -32.051963806152344,
|
216 |
+
"loss": 0.4497,
|
217 |
+
"rewards/accuracies": 0.4749999940395355,
|
218 |
+
"rewards/chosen": -0.014895597472786903,
|
219 |
+
"rewards/margins": -0.015729816630482674,
|
220 |
+
"rewards/rejected": 0.0008342192741110921,
|
221 |
"step": 130
|
222 |
},
|
223 |
{
|
224 |
"epoch": 0.36,
|
225 |
"learning_rate": 4.020402418666621e-07,
|
226 |
+
"logits/chosen": -1.977065086364746,
|
227 |
+
"logits/rejected": -1.987343430519104,
|
228 |
+
"logps/chosen": -31.37868309020996,
|
229 |
+
"logps/rejected": -32.54730224609375,
|
230 |
+
"loss": 0.3589,
|
231 |
+
"rewards/accuracies": 0.574999988079071,
|
232 |
+
"rewards/chosen": 0.03278205543756485,
|
233 |
+
"rewards/margins": 0.04271895810961723,
|
234 |
+
"rewards/rejected": -0.009936909191310406,
|
235 |
"step": 140
|
236 |
},
|
237 |
{
|
238 |
"epoch": 0.39,
|
239 |
"learning_rate": 3.8341962650351185e-07,
|
240 |
+
"logits/chosen": -1.8916466236114502,
|
241 |
+
"logits/rejected": -1.8927490711212158,
|
242 |
+
"logps/chosen": -34.209651947021484,
|
243 |
+
"logps/rejected": -34.766143798828125,
|
244 |
+
"loss": 0.4325,
|
245 |
+
"rewards/accuracies": 0.42500001192092896,
|
246 |
+
"rewards/chosen": -0.013569498434662819,
|
247 |
+
"rewards/margins": -0.00645996630191803,
|
248 |
+
"rewards/rejected": -0.007109532598406076,
|
249 |
"step": 150
|
250 |
},
|
251 |
{
|
252 |
"epoch": 0.42,
|
253 |
"learning_rate": 3.636998309800572e-07,
|
254 |
+
"logits/chosen": -1.9427303075790405,
|
255 |
+
"logits/rejected": -1.939252495765686,
|
256 |
+
"logps/chosen": -36.14452362060547,
|
257 |
+
"logps/rejected": -32.73284149169922,
|
258 |
+
"loss": 0.3796,
|
259 |
+
"rewards/accuracies": 0.550000011920929,
|
260 |
+
"rewards/chosen": 0.021723434329032898,
|
261 |
+
"rewards/margins": 0.02979486621916294,
|
262 |
+
"rewards/rejected": -0.008071433752775192,
|
263 |
"step": 160
|
264 |
},
|
265 |
{
|
266 |
"epoch": 0.44,
|
267 |
"learning_rate": 3.430433172111807e-07,
|
268 |
+
"logits/chosen": -2.0424513816833496,
|
269 |
+
"logits/rejected": -2.035060405731201,
|
270 |
+
"logps/chosen": -33.786170959472656,
|
271 |
+
"logps/rejected": -31.34820556640625,
|
272 |
+
"loss": 0.4172,
|
273 |
+
"rewards/accuracies": 0.512499988079071,
|
274 |
+
"rewards/chosen": 0.009350698441267014,
|
275 |
+
"rewards/margins": 0.006434415467083454,
|
276 |
+
"rewards/rejected": 0.0029162843711674213,
|
277 |
"step": 170
|
278 |
},
|
279 |
{
|
280 |
"epoch": 0.47,
|
281 |
"learning_rate": 3.216202642830543e-07,
|
282 |
+
"logits/chosen": -2.0475661754608154,
|
283 |
+
"logits/rejected": -2.052834987640381,
|
284 |
+
"logps/chosen": -32.528114318847656,
|
285 |
+
"logps/rejected": -32.50902557373047,
|
286 |
+
"loss": 0.3766,
|
287 |
+
"rewards/accuracies": 0.4749999940395355,
|
288 |
+
"rewards/chosen": 0.013338183984160423,
|
289 |
+
"rewards/margins": 0.03244578838348389,
|
290 |
+
"rewards/rejected": -0.019107606261968613,
|
291 |
"step": 180
|
292 |
},
|
293 |
{
|
294 |
"epoch": 0.49,
|
295 |
"learning_rate": 2.9960716642946403e-07,
|
296 |
+
"logits/chosen": -2.048661231994629,
|
297 |
+
"logits/rejected": -2.045872926712036,
|
298 |
+
"logps/chosen": -31.496755599975586,
|
299 |
+
"logps/rejected": -31.314464569091797,
|
300 |
+
"loss": 0.4475,
|
301 |
+
"rewards/accuracies": 0.5,
|
302 |
+
"rewards/chosen": -0.00624021515250206,
|
303 |
+
"rewards/margins": -0.009242123924195766,
|
304 |
+
"rewards/rejected": 0.003001909703016281,
|
305 |
"step": 190
|
306 |
},
|
307 |
{
|
308 |
"epoch": 0.52,
|
309 |
"learning_rate": 2.771853789806683e-07,
|
310 |
+
"logits/chosen": -1.9193611145019531,
|
311 |
+
"logits/rejected": -1.9240529537200928,
|
312 |
+
"logps/chosen": -31.579212188720703,
|
313 |
+
"logps/rejected": -32.80603790283203,
|
314 |
+
"loss": 0.371,
|
315 |
+
"rewards/accuracies": 0.6000000238418579,
|
316 |
+
"rewards/chosen": 0.02165621519088745,
|
317 |
+
"rewards/margins": 0.033739686012268066,
|
318 |
+
"rewards/rejected": -0.012083468958735466,
|
319 |
"step": 200
|
320 |
},
|
321 |
{
|
322 |
"epoch": 0.52,
|
323 |
+
"eval_logits/chosen": -2.243856191635132,
|
324 |
+
"eval_logits/rejected": -2.2389791011810303,
|
325 |
+
"eval_logps/chosen": -34.023136138916016,
|
326 |
+
"eval_logps/rejected": -37.501853942871094,
|
327 |
+
"eval_loss": 0.4322855770587921,
|
328 |
+
"eval_rewards/accuracies": 0.5253322720527649,
|
329 |
+
"eval_rewards/chosen": 0.009133166633546352,
|
330 |
+
"eval_rewards/margins": -0.002679171971976757,
|
331 |
+
"eval_rewards/rejected": 0.01181233860552311,
|
332 |
+
"eval_runtime": 145.9618,
|
333 |
+
"eval_samples_per_second": 2.35,
|
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": -2.0320448875427246,
|
341 |
+
"logits/rejected": -2.0427393913269043,
|
342 |
+
"logps/chosen": -31.947372436523438,
|
343 |
+
"logps/rejected": -33.899864196777344,
|
344 |
+
"loss": 0.3622,
|
345 |
+
"rewards/accuracies": 0.6000000238418579,
|
346 |
+
"rewards/chosen": 0.019143681973218918,
|
347 |
+
"rewards/margins": 0.04887578636407852,
|
348 |
+
"rewards/rejected": -0.029732098802924156,
|
349 |
"step": 210
|
350 |
},
|
351 |
{
|
352 |
"epoch": 0.57,
|
353 |
"learning_rate": 2.318564697655179e-07,
|
354 |
+
"logits/chosen": -1.9257709980010986,
|
355 |
+
"logits/rejected": -1.940639853477478,
|
356 |
+
"logps/chosen": -30.079341888427734,
|
357 |
+
"logps/rejected": -31.5616397857666,
|
358 |
+
"loss": 0.3884,
|
359 |
+
"rewards/accuracies": 0.612500011920929,
|
360 |
+
"rewards/chosen": 0.025758206844329834,
|
361 |
+
"rewards/margins": 0.028527002781629562,
|
362 |
+
"rewards/rejected": -0.0027687971014529467,
|
363 |
"step": 220
|
364 |
},
|
365 |
{
|
366 |
"epoch": 0.6,
|
367 |
"learning_rate": 2.093227910899832e-07,
|
368 |
+
"logits/chosen": -1.9829959869384766,
|
369 |
+
"logits/rejected": -1.9869670867919922,
|
370 |
+
"logps/chosen": -33.4053955078125,
|
371 |
+
"logps/rejected": -31.562353134155273,
|
372 |
+
"loss": 0.3918,
|
373 |
+
"rewards/accuracies": 0.5874999761581421,
|
374 |
+
"rewards/chosen": 0.009657363407313824,
|
375 |
+
"rewards/margins": 0.02010050043463707,
|
376 |
+
"rewards/rejected": -0.010443134233355522,
|
377 |
"step": 230
|
378 |
},
|
379 |
{
|
380 |
"epoch": 0.62,
|
381 |
"learning_rate": 1.8712423238279356e-07,
|
382 |
+
"logits/chosen": -1.9836658239364624,
|
383 |
+
"logits/rejected": -1.9616800546646118,
|
384 |
+
"logps/chosen": -34.175201416015625,
|
385 |
+
"logps/rejected": -34.966102600097656,
|
386 |
+
"loss": 0.4346,
|
387 |
+
"rewards/accuracies": 0.5375000238418579,
|
388 |
+
"rewards/chosen": -0.017159918323159218,
|
389 |
+
"rewards/margins": -0.004913450218737125,
|
390 |
+
"rewards/rejected": -0.012246469967067242,
|
391 |
"step": 240
|
392 |
},
|
393 |
{
|
394 |
"epoch": 0.65,
|
395 |
"learning_rate": 1.654436768970182e-07,
|
396 |
+
"logits/chosen": -2.0244596004486084,
|
397 |
+
"logits/rejected": -2.021144390106201,
|
398 |
+
"logps/chosen": -32.932373046875,
|
399 |
+
"logps/rejected": -36.2264289855957,
|
400 |
+
"loss": 0.4397,
|
401 |
+
"rewards/accuracies": 0.512499988079071,
|
402 |
+
"rewards/chosen": -0.005845203995704651,
|
403 |
+
"rewards/margins": -0.004640025552362204,
|
404 |
+
"rewards/rejected": -0.001205177279189229,
|
405 |
"step": 250
|
406 |
},
|
407 |
{
|
408 |
"epoch": 0.68,
|
409 |
"learning_rate": 1.444597403062196e-07,
|
410 |
+
"logits/chosen": -1.891283392906189,
|
411 |
+
"logits/rejected": -1.8888483047485352,
|
412 |
+
"logps/chosen": -34.20048522949219,
|
413 |
+
"logps/rejected": -35.507022857666016,
|
414 |
+
"loss": 0.4441,
|
415 |
+
"rewards/accuracies": 0.5625,
|
416 |
+
"rewards/chosen": -0.009669994935393333,
|
417 |
+
"rewards/margins": -0.0077504729852080345,
|
418 |
+
"rewards/rejected": -0.0019195213681086898,
|
419 |
"step": 260
|
420 |
},
|
421 |
{
|
422 |
"epoch": 0.7,
|
423 |
"learning_rate": 1.2434529917578887e-07,
|
424 |
+
"logits/chosen": -1.8760111331939697,
|
425 |
+
"logits/rejected": -1.8734772205352783,
|
426 |
+
"logps/chosen": -34.39020919799805,
|
427 |
+
"logps/rejected": -31.75579261779785,
|
428 |
+
"loss": 0.4295,
|
429 |
+
"rewards/accuracies": 0.42500001192092896,
|
430 |
+
"rewards/chosen": 0.0059113698080182076,
|
431 |
+
"rewards/margins": -0.006519269198179245,
|
432 |
+
"rewards/rejected": 0.012430639937520027,
|
433 |
"step": 270
|
434 |
},
|
435 |
{
|
436 |
"epoch": 0.73,
|
437 |
"learning_rate": 1.0526606671603521e-07,
|
438 |
+
"logits/chosen": -1.9797089099884033,
|
439 |
+
"logits/rejected": -1.9690707921981812,
|
440 |
+
"logps/chosen": -35.31258773803711,
|
441 |
+
"logps/rejected": -31.837697982788086,
|
442 |
+
"loss": 0.3868,
|
443 |
+
"rewards/accuracies": 0.5249999761581421,
|
444 |
+
"rewards/chosen": 0.03524213656783104,
|
445 |
+
"rewards/margins": 0.0326063297688961,
|
446 |
+
"rewards/rejected": 0.0026358035393059254,
|
447 |
"step": 280
|
448 |
},
|
449 |
{
|
450 |
"epoch": 0.75,
|
451 |
"learning_rate": 8.737922755071453e-08,
|
452 |
+
"logits/chosen": -2.0755209922790527,
|
453 |
+
"logits/rejected": -2.060497760772705,
|
454 |
+
"logps/chosen": -30.90865135192871,
|
455 |
+
"logps/rejected": -32.64521026611328,
|
456 |
+
"loss": 0.41,
|
457 |
+
"rewards/accuracies": 0.44999998807907104,
|
458 |
+
"rewards/chosen": 0.012379061430692673,
|
459 |
+
"rewards/margins": 0.007080032490193844,
|
460 |
+
"rewards/rejected": 0.005299028940498829,
|
461 |
"step": 290
|
462 |
},
|
463 |
{
|
464 |
"epoch": 0.78,
|
465 |
"learning_rate": 7.08321427484816e-08,
|
466 |
+
"logits/chosen": -1.9463651180267334,
|
467 |
+
"logits/rejected": -1.9438308477401733,
|
468 |
+
"logps/chosen": -32.89823532104492,
|
469 |
+
"logps/rejected": -30.81850814819336,
|
470 |
+
"loss": 0.3839,
|
471 |
+
"rewards/accuracies": 0.5874999761581421,
|
472 |
+
"rewards/chosen": 0.02309919334948063,
|
473 |
+
"rewards/margins": 0.024458223953843117,
|
474 |
+
"rewards/rejected": -0.0013590289745479822,
|
475 |
"step": 300
|
476 |
},
|
477 |
{
|
478 |
"epoch": 0.78,
|
479 |
+
"eval_logits/chosen": -2.2435548305511475,
|
480 |
+
"eval_logits/rejected": -2.2386720180511475,
|
481 |
+
"eval_logps/chosen": -34.023826599121094,
|
482 |
+
"eval_logps/rejected": -37.50039291381836,
|
483 |
+
"eval_loss": 0.4344586431980133,
|
484 |
"eval_rewards/accuracies": 0.4808970093727112,
|
485 |
+
"eval_rewards/chosen": 0.00858243927359581,
|
486 |
+
"eval_rewards/margins": -0.004399063065648079,
|
487 |
+
"eval_rewards/rejected": 0.012981505133211613,
|
488 |
+
"eval_runtime": 145.9473,
|
489 |
+
"eval_samples_per_second": 2.35,
|
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": -1.9286285638809204,
|
497 |
+
"logits/rejected": -1.9253568649291992,
|
498 |
+
"logps/chosen": -31.583232879638672,
|
499 |
+
"logps/rejected": -33.75123596191406,
|
500 |
+
"loss": 0.3881,
|
501 |
+
"rewards/accuracies": 0.5375000238418579,
|
502 |
+
"rewards/chosen": 0.016091840341687202,
|
503 |
+
"rewards/margins": 0.023359118029475212,
|
504 |
+
"rewards/rejected": -0.007267280016094446,
|
505 |
"step": 310
|
506 |
},
|
507 |
{
|
508 |
"epoch": 0.83,
|
509 |
"learning_rate": 4.229036944380912e-08,
|
510 |
+
"logits/chosen": -1.9803975820541382,
|
511 |
+
"logits/rejected": -1.9680900573730469,
|
512 |
+
"logps/chosen": -34.58079147338867,
|
513 |
+
"logps/rejected": -33.575767517089844,
|
514 |
+
"loss": 0.3778,
|
515 |
+
"rewards/accuracies": 0.5375000238418579,
|
516 |
+
"rewards/chosen": 0.012225466780364513,
|
517 |
+
"rewards/margins": 0.041770923882722855,
|
518 |
+
"rewards/rejected": -0.029545456171035767,
|
519 |
"step": 320
|
520 |
},
|
521 |
{
|
522 |
"epoch": 0.86,
|
523 |
"learning_rate": 3.053082288996112e-08,
|
524 |
+
"logits/chosen": -2.015634298324585,
|
525 |
+
"logits/rejected": -2.0141713619232178,
|
526 |
+
"logps/chosen": -33.45996856689453,
|
527 |
+
"logps/rejected": -32.48029327392578,
|
528 |
+
"loss": 0.4032,
|
529 |
+
"rewards/accuracies": 0.5625,
|
530 |
+
"rewards/chosen": 0.020660031586885452,
|
531 |
+
"rewards/margins": 0.021225089207291603,
|
532 |
+
"rewards/rejected": -0.0005650619277730584,
|
533 |
"step": 330
|
534 |
},
|
535 |
{
|
536 |
"epoch": 0.88,
|
537 |
"learning_rate": 2.05793773749158e-08,
|
538 |
+
"logits/chosen": -2.1035873889923096,
|
539 |
+
"logits/rejected": -2.0877768993377686,
|
540 |
+
"logps/chosen": -34.162208557128906,
|
541 |
+
"logps/rejected": -33.095733642578125,
|
542 |
+
"loss": 0.439,
|
543 |
+
"rewards/accuracies": 0.4625000059604645,
|
544 |
+
"rewards/chosen": 0.013746557757258415,
|
545 |
+
"rewards/margins": -0.004091509617865086,
|
546 |
+
"rewards/rejected": 0.017838066443800926,
|
547 |
"step": 340
|
548 |
},
|
549 |
{
|
550 |
"epoch": 0.91,
|
551 |
"learning_rate": 1.251801807404168e-08,
|
552 |
+
"logits/chosen": -1.9746344089508057,
|
553 |
+
"logits/rejected": -1.9736888408660889,
|
554 |
+
"logps/chosen": -33.253448486328125,
|
555 |
+
"logps/rejected": -32.458499908447266,
|
556 |
+
"loss": 0.3954,
|
557 |
+
"rewards/accuracies": 0.5249999761581421,
|
558 |
+
"rewards/chosen": 0.02291743829846382,
|
559 |
+
"rewards/margins": 0.017600122839212418,
|
560 |
+
"rewards/rejected": 0.005317316390573978,
|
561 |
"step": 350
|
562 |
},
|
563 |
{
|
564 |
"epoch": 0.94,
|
565 |
"learning_rate": 6.41315865106129e-09,
|
566 |
+
"logits/chosen": -1.9309253692626953,
|
567 |
+
"logits/rejected": -1.941303014755249,
|
568 |
+
"logps/chosen": -32.21305465698242,
|
569 |
+
"logps/rejected": -35.309688568115234,
|
570 |
+
"loss": 0.4095,
|
571 |
+
"rewards/accuracies": 0.42500001192092896,
|
572 |
+
"rewards/chosen": -0.006690672133117914,
|
573 |
+
"rewards/margins": 0.010053041391074657,
|
574 |
+
"rewards/rejected": -0.01674371212720871,
|
575 |
"step": 360
|
576 |
},
|
577 |
{
|
578 |
"epoch": 0.96,
|
579 |
"learning_rate": 2.3150941078050324e-09,
|
580 |
+
"logits/chosen": -2.069108009338379,
|
581 |
+
"logits/rejected": -2.0625574588775635,
|
582 |
+
"logps/chosen": -33.637176513671875,
|
583 |
+
"logps/rejected": -29.221187591552734,
|
584 |
+
"loss": 0.4059,
|
585 |
"rewards/accuracies": 0.48750001192092896,
|
586 |
+
"rewards/chosen": 0.011833530850708485,
|
587 |
+
"rewards/margins": 0.00764369685202837,
|
588 |
+
"rewards/rejected": 0.004189834464341402,
|
589 |
"step": 370
|
590 |
},
|
591 |
{
|
592 |
"epoch": 0.99,
|
593 |
"learning_rate": 2.575864278703266e-10,
|
594 |
+
"logits/chosen": -1.9292182922363281,
|
595 |
+
"logits/rejected": -1.9313886165618896,
|
596 |
+
"logps/chosen": -34.24732971191406,
|
597 |
+
"logps/rejected": -30.893402099609375,
|
598 |
+
"loss": 0.4086,
|
599 |
+
"rewards/accuracies": 0.4749999940395355,
|
600 |
+
"rewards/chosen": -0.006622877903282642,
|
601 |
+
"rewards/margins": 0.009996414184570312,
|
602 |
+
"rewards/rejected": -0.01661929115653038,
|
603 |
"step": 380
|
604 |
},
|
605 |
{
|
606 |
"epoch": 1.0,
|
607 |
"step": 385,
|
608 |
"total_flos": 0.0,
|
609 |
+
"train_loss": 0.4092799360101873,
|
610 |
+
"train_runtime": 3256.9261,
|
611 |
+
"train_samples_per_second": 0.945,
|
612 |
"train_steps_per_second": 0.118
|
613 |
}
|
614 |
],
|