hugodk-sch
commited on
Commit
•
20fd664
1
Parent(s):
4a3abcd
Model save
Browse files- README.md +13 -16
- adapter_model.safetensors +1 -1
- all_results.json +2 -15
- train_results.json +2 -2
- trainer_state.json +375 -375
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:
|
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 |
-
| 1.
|
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: 1.0014
|
21 |
+
- Rewards/chosen: 0.0061
|
22 |
+
- Rewards/rejected: 0.0075
|
23 |
+
- Rewards/accuracies: 0.5104
|
24 |
+
- Rewards/margins: -0.0014
|
25 |
+
- Logps/rejected: -37.5016
|
26 |
+
- Logps/chosen: -34.0223
|
27 |
+
- Logits/rejected: -2.2389
|
28 |
+
- Logits/chosen: -2.2438
|
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 |
+
| 1.007 | 0.26 | 100 | 0.9963 | 0.0083 | 0.0045 | 0.5166 | 0.0039 | -37.5077 | -34.0179 | -2.2391 | -2.2440 |
|
64 |
+
| 0.9725 | 0.52 | 200 | 1.0085 | 0.0008 | 0.0092 | 0.4992 | -0.0084 | -37.4982 | -34.0330 | -2.2385 | -2.2434 |
|
65 |
+
| 0.982 | 0.78 | 300 | 1.0014 | 0.0061 | 0.0075 | 0.5104 | -0.0014 | -37.5016 | -34.0223 | -2.2389 | -2.2438 |
|
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:ad191a64493e95cbde4ef60db35429a7160e9cf1c921a1af3f24538d7fe45068
|
3 |
size 176183216
|
all_results.json
CHANGED
@@ -1,20 +1,7 @@
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
-
"
|
4 |
-
"
|
5 |
-
"eval_logps/chosen": -34.0224494934082,
|
6 |
-
"eval_logps/rejected": -37.51002883911133,
|
7 |
-
"eval_loss": 0.9980219602584839,
|
8 |
-
"eval_rewards/accuracies": 0.5141196250915527,
|
9 |
-
"eval_rewards/chosen": 0.0048410315066576,
|
10 |
-
"eval_rewards/margins": 0.002204673830419779,
|
11 |
-
"eval_rewards/rejected": 0.0026363590732216835,
|
12 |
-
"eval_runtime": 145.7351,
|
13 |
-
"eval_samples": 343,
|
14 |
-
"eval_samples_per_second": 2.354,
|
15 |
-
"eval_steps_per_second": 0.295,
|
16 |
-
"train_loss": 0.9964844827528124,
|
17 |
-
"train_runtime": 3255.8213,
|
18 |
"train_samples": 3079,
|
19 |
"train_samples_per_second": 0.946,
|
20 |
"train_steps_per_second": 0.118
|
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
+
"train_loss": 0.9930156608680626,
|
4 |
+
"train_runtime": 3255.7181,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
5 |
"train_samples": 3079,
|
6 |
"train_samples_per_second": 0.946,
|
7 |
"train_steps_per_second": 0.118
|
train_results.json
CHANGED
@@ -1,7 +1,7 @@
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
-
"train_loss": 0.
|
4 |
-
"train_runtime": 3255.
|
5 |
"train_samples": 3079,
|
6 |
"train_samples_per_second": 0.946,
|
7 |
"train_steps_per_second": 0.118
|
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
+
"train_loss": 0.9930156608680626,
|
4 |
+
"train_runtime": 3255.7181,
|
5 |
"train_samples": 3079,
|
6 |
"train_samples_per_second": 0.946,
|
7 |
"train_steps_per_second": 0.118
|
trainer_state.json
CHANGED
@@ -25,589 +25,589 @@
|
|
25 |
{
|
26 |
"epoch": 0.03,
|
27 |
"learning_rate": 1.2820512820512818e-07,
|
28 |
-
"logits/chosen": -1.
|
29 |
-
"logits/rejected": -1.
|
30 |
-
"logps/chosen": -36.
|
31 |
-
"logps/rejected": -33.
|
32 |
-
"loss": 0.
|
33 |
-
"rewards/accuracies": 0.
|
34 |
-
"rewards/chosen": 0.
|
35 |
-
"rewards/margins": 0.
|
36 |
-
"rewards/rejected": -0.
|
37 |
"step": 10
|
38 |
},
|
39 |
{
|
40 |
"epoch": 0.05,
|
41 |
"learning_rate": 2.5641025641025636e-07,
|
42 |
-
"logits/chosen": -1.
|
43 |
-
"logits/rejected": -1.
|
44 |
-
"logps/chosen": -29.
|
45 |
-
"logps/rejected": -29.
|
46 |
-
"loss": 1.
|
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":
|
61 |
-
"rewards/accuracies": 0.
|
62 |
-
"rewards/chosen":
|
63 |
-
"rewards/margins":
|
64 |
-
"rewards/rejected": 0.
|
65 |
"step": 30
|
66 |
},
|
67 |
{
|
68 |
"epoch": 0.1,
|
69 |
"learning_rate": 4.999896948438433e-07,
|
70 |
-
"logits/chosen": -2.
|
71 |
-
"logits/rejected": -2.
|
72 |
-
"logps/chosen": -32.
|
73 |
-
"logps/rejected": -32.
|
74 |
-
"loss": 1.
|
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": 1.
|
89 |
-
"rewards/accuracies": 0.
|
90 |
-
"rewards/chosen":
|
91 |
-
"rewards/margins": -0.
|
92 |
-
"rewards/rejected": 0.
|
93 |
"step": 50
|
94 |
},
|
95 |
{
|
96 |
"epoch": 0.16,
|
97 |
"learning_rate": 4.954691471941118e-07,
|
98 |
-
"logits/chosen": -1.
|
99 |
-
"logits/rejected": -1.
|
100 |
-
"logps/chosen": -32.
|
101 |
-
"logps/rejected": -33.
|
102 |
-
"loss": 0.
|
103 |
-
"rewards/accuracies": 0.
|
104 |
-
"rewards/chosen": 0.
|
105 |
-
"rewards/margins": 0.
|
106 |
-
"rewards/rejected": -0.
|
107 |
"step": 60
|
108 |
},
|
109 |
{
|
110 |
"epoch": 0.18,
|
111 |
"learning_rate": 4.901618883413548e-07,
|
112 |
-
"logits/chosen": -2.
|
113 |
-
"logits/rejected": -2.
|
114 |
-
"logps/chosen": -33.
|
115 |
-
"logps/rejected": -36.
|
116 |
-
"loss": 0.
|
117 |
-
"rewards/accuracies": 0.
|
118 |
-
"rewards/chosen": -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":
|
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":
|
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": 1.
|
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":
|
172 |
-
"eval_rewards/accuracies": 0.
|
173 |
-
"eval_rewards/chosen": 0.
|
174 |
-
"eval_rewards/margins":
|
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": -34.
|
188 |
-
"loss":
|
189 |
-
"rewards/accuracies": 0.
|
190 |
-
"rewards/chosen": 0.
|
191 |
-
"rewards/margins":
|
192 |
-
"rewards/rejected": 0.
|
193 |
"step": 110
|
194 |
},
|
195 |
{
|
196 |
"epoch": 0.31,
|
197 |
"learning_rate": 4.353806263777677e-07,
|
198 |
-
"logits/chosen": -2.
|
199 |
-
"logits/rejected": -2.
|
200 |
-
"logps/chosen": -32.
|
201 |
-
"logps/rejected": -32.
|
202 |
-
"loss": 0.
|
203 |
-
"rewards/accuracies": 0.
|
204 |
-
"rewards/chosen": -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": 1.
|
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":
|
247 |
-
"rewards/margins": 0.
|
248 |
-
"rewards/rejected": -
|
249 |
"step": 150
|
250 |
},
|
251 |
{
|
252 |
"epoch": 0.42,
|
253 |
"learning_rate": 3.636998309800572e-07,
|
254 |
-
"logits/chosen": -1.
|
255 |
-
"logits/rejected": -1.
|
256 |
-
"logps/chosen": -36.
|
257 |
-
"logps/rejected": -32.
|
258 |
-
"loss": 0.
|
259 |
-
"rewards/accuracies": 0.
|
260 |
-
"rewards/chosen": 0.
|
261 |
-
"rewards/margins": 0.
|
262 |
-
"rewards/rejected": -0.
|
263 |
"step": 160
|
264 |
},
|
265 |
{
|
266 |
"epoch": 0.44,
|
267 |
"learning_rate": 3.430433172111807e-07,
|
268 |
-
"logits/chosen": -2.
|
269 |
-
"logits/rejected": -2.
|
270 |
-
"logps/chosen": -33.
|
271 |
-
"logps/rejected": -31.
|
272 |
-
"loss": 0.
|
273 |
-
"rewards/accuracies": 0.
|
274 |
-
"rewards/chosen": 0.
|
275 |
-
"rewards/margins": 0.
|
276 |
-
"rewards/rejected": -0.
|
277 |
"step": 170
|
278 |
},
|
279 |
{
|
280 |
"epoch": 0.47,
|
281 |
"learning_rate": 3.216202642830543e-07,
|
282 |
-
"logits/chosen": -2.
|
283 |
-
"logits/rejected": -2.
|
284 |
-
"logps/chosen": -32.
|
285 |
-
"logps/rejected": -32.
|
286 |
-
"loss": 0.
|
287 |
-
"rewards/accuracies": 0.
|
288 |
-
"rewards/chosen": 0.
|
289 |
-
"rewards/margins": 0.
|
290 |
-
"rewards/rejected": -0.
|
291 |
"step": 180
|
292 |
},
|
293 |
{
|
294 |
"epoch": 0.49,
|
295 |
"learning_rate": 2.9960716642946403e-07,
|
296 |
-
"logits/chosen": -2.
|
297 |
-
"logits/rejected": -2.
|
298 |
-
"logps/chosen": -31.
|
299 |
-
"logps/rejected": -31.
|
300 |
-
"loss":
|
301 |
-
"rewards/accuracies": 0.
|
302 |
-
"rewards/chosen":
|
303 |
-
"rewards/margins":
|
304 |
-
"rewards/rejected": -0.
|
305 |
"step": 190
|
306 |
},
|
307 |
{
|
308 |
"epoch": 0.52,
|
309 |
"learning_rate": 2.771853789806683e-07,
|
310 |
-
"logits/chosen": -1.
|
311 |
-
"logits/rejected": -1.
|
312 |
-
"logps/chosen": -31.
|
313 |
-
"logps/rejected": -32.
|
314 |
-
"loss": 0.
|
315 |
-
"rewards/accuracies": 0.
|
316 |
-
"rewards/chosen": 0.
|
317 |
-
"rewards/margins": 0.
|
318 |
-
"rewards/rejected": 0.
|
319 |
"step": 200
|
320 |
},
|
321 |
{
|
322 |
"epoch": 0.52,
|
323 |
-
"eval_logits/chosen": -2.
|
324 |
-
"eval_logits/rejected": -2.
|
325 |
-
"eval_logps/chosen": -34.
|
326 |
-
"eval_logps/rejected": -37.
|
327 |
-
"eval_loss":
|
328 |
-
"eval_rewards/accuracies": 0.
|
329 |
-
"eval_rewards/chosen": 0.
|
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": 2.5453962426402e-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":
|
403 |
-
"rewards/margins": 0.
|
404 |
-
"rewards/rejected":
|
405 |
"step": 250
|
406 |
},
|
407 |
{
|
408 |
"epoch": 0.68,
|
409 |
"learning_rate": 1.444597403062196e-07,
|
410 |
-
"logits/chosen": -1.
|
411 |
-
"logits/rejected": -1.
|
412 |
-
"logps/chosen": -34.
|
413 |
-
"logps/rejected": -35.
|
414 |
-
"loss": 0.
|
415 |
-
"rewards/accuracies": 0.
|
416 |
-
"rewards/chosen":
|
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": 1.
|
429 |
"rewards/accuracies": 0.4625000059604645,
|
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": -0.
|
447 |
"step": 280
|
448 |
},
|
449 |
{
|
450 |
"epoch": 0.75,
|
451 |
"learning_rate": 8.737922755071453e-08,
|
452 |
-
"logits/chosen": -2.
|
453 |
-
"logits/rejected": -2.
|
454 |
-
"logps/chosen": -30.
|
455 |
-
"logps/rejected": -32.
|
456 |
-
"loss": 0.
|
457 |
-
"rewards/accuracies": 0.
|
458 |
-
"rewards/chosen": 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":
|
484 |
-
"eval_rewards/accuracies": 0.
|
485 |
-
"eval_rewards/chosen": 0.
|
486 |
-
"eval_rewards/margins": 0.
|
487 |
-
"eval_rewards/rejected": 0.
|
488 |
-
"eval_runtime": 145.
|
489 |
-
"eval_samples_per_second": 2.
|
490 |
"eval_steps_per_second": 0.295,
|
491 |
"step": 300
|
492 |
},
|
493 |
{
|
494 |
"epoch": 0.81,
|
495 |
"learning_rate": 5.576113578589034e-08,
|
496 |
-
"logits/chosen": -1.
|
497 |
-
"logits/rejected": -1.
|
498 |
-
"logps/chosen": -31.
|
499 |
-
"logps/rejected": -33.
|
500 |
-
"loss":
|
501 |
-
"rewards/accuracies": 0.
|
502 |
-
"rewards/chosen":
|
503 |
-
"rewards/margins":
|
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":
|
529 |
-
"rewards/accuracies": 0.
|
530 |
-
"rewards/chosen":
|
531 |
-
"rewards/margins":
|
532 |
-
"rewards/rejected":
|
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": 1.
|
543 |
"rewards/accuracies": 0.42500001192092896,
|
544 |
-
"rewards/chosen":
|
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":
|
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":
|
571 |
-
"rewards/accuracies": 0.
|
572 |
-
"rewards/chosen":
|
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":
|
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-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": 3255.
|
611 |
"train_samples_per_second": 0.946,
|
612 |
"train_steps_per_second": 0.118
|
613 |
}
|
|
|
25 |
{
|
26 |
"epoch": 0.03,
|
27 |
"learning_rate": 1.2820512820512818e-07,
|
28 |
+
"logits/chosen": -1.8661351203918457,
|
29 |
+
"logits/rejected": -1.8704447746276855,
|
30 |
+
"logps/chosen": -36.994537353515625,
|
31 |
+
"logps/rejected": -33.65791320800781,
|
32 |
+
"loss": 0.98,
|
33 |
+
"rewards/accuracies": 0.5138888955116272,
|
34 |
+
"rewards/chosen": 0.006032215431332588,
|
35 |
+
"rewards/margins": 0.019952965900301933,
|
36 |
+
"rewards/rejected": -0.013920750468969345,
|
37 |
"step": 10
|
38 |
},
|
39 |
{
|
40 |
"epoch": 0.05,
|
41 |
"learning_rate": 2.5641025641025636e-07,
|
42 |
+
"logits/chosen": -1.9972000122070312,
|
43 |
+
"logits/rejected": -1.9998512268066406,
|
44 |
+
"logps/chosen": -29.651025772094727,
|
45 |
+
"logps/rejected": -29.047237396240234,
|
46 |
+
"loss": 1.0125,
|
47 |
+
"rewards/accuracies": 0.42500001192092896,
|
48 |
+
"rewards/chosen": -0.004414811730384827,
|
49 |
+
"rewards/margins": -0.012516966089606285,
|
50 |
+
"rewards/rejected": 0.008102154359221458,
|
51 |
"step": 20
|
52 |
},
|
53 |
{
|
54 |
"epoch": 0.08,
|
55 |
"learning_rate": 3.8461538461538463e-07,
|
56 |
+
"logits/chosen": -1.919605016708374,
|
57 |
+
"logits/rejected": -1.9169189929962158,
|
58 |
+
"logps/chosen": -31.410541534423828,
|
59 |
+
"logps/rejected": -33.22766876220703,
|
60 |
+
"loss": 0.9926,
|
61 |
+
"rewards/accuracies": 0.550000011920929,
|
62 |
+
"rewards/chosen": 0.002776858163997531,
|
63 |
+
"rewards/margins": 0.007382071111351252,
|
64 |
+
"rewards/rejected": -0.004605212714523077,
|
65 |
"step": 30
|
66 |
},
|
67 |
{
|
68 |
"epoch": 0.1,
|
69 |
"learning_rate": 4.999896948438433e-07,
|
70 |
+
"logits/chosen": -2.016796827316284,
|
71 |
+
"logits/rejected": -2.0080485343933105,
|
72 |
+
"logps/chosen": -32.59518814086914,
|
73 |
+
"logps/rejected": -32.49793243408203,
|
74 |
+
"loss": 1.0166,
|
75 |
+
"rewards/accuracies": 0.4375,
|
76 |
+
"rewards/chosen": -0.00919206440448761,
|
77 |
+
"rewards/margins": -0.016562189906835556,
|
78 |
+
"rewards/rejected": 0.007370126433670521,
|
79 |
"step": 40
|
80 |
},
|
81 |
{
|
82 |
"epoch": 0.13,
|
83 |
"learning_rate": 4.987541037542186e-07,
|
84 |
+
"logits/chosen": -1.8643512725830078,
|
85 |
+
"logits/rejected": -1.853578805923462,
|
86 |
+
"logps/chosen": -33.55515670776367,
|
87 |
+
"logps/rejected": -35.426795959472656,
|
88 |
+
"loss": 1.0116,
|
89 |
+
"rewards/accuracies": 0.4749999940395355,
|
90 |
+
"rewards/chosen": 0.0011615634430199862,
|
91 |
+
"rewards/margins": -0.011613734066486359,
|
92 |
+
"rewards/rejected": 0.012775297276675701,
|
93 |
"step": 50
|
94 |
},
|
95 |
{
|
96 |
"epoch": 0.16,
|
97 |
"learning_rate": 4.954691471941118e-07,
|
98 |
+
"logits/chosen": -1.9452184438705444,
|
99 |
+
"logits/rejected": -1.9471704959869385,
|
100 |
+
"logps/chosen": -32.57328414916992,
|
101 |
+
"logps/rejected": -33.17564010620117,
|
102 |
+
"loss": 0.9844,
|
103 |
+
"rewards/accuracies": 0.5375000238418579,
|
104 |
+
"rewards/chosen": 0.013449115678668022,
|
105 |
+
"rewards/margins": 0.015578977763652802,
|
106 |
+
"rewards/rejected": -0.002129861619323492,
|
107 |
"step": 60
|
108 |
},
|
109 |
{
|
110 |
"epoch": 0.18,
|
111 |
"learning_rate": 4.901618883413548e-07,
|
112 |
+
"logits/chosen": -2.079322338104248,
|
113 |
+
"logits/rejected": -2.0843043327331543,
|
114 |
+
"logps/chosen": -33.998104095458984,
|
115 |
+
"logps/rejected": -36.586326599121094,
|
116 |
+
"loss": 0.9961,
|
117 |
+
"rewards/accuracies": 0.5625,
|
118 |
+
"rewards/chosen": -0.004809260368347168,
|
119 |
+
"rewards/margins": 0.003915563225746155,
|
120 |
+
"rewards/rejected": -0.008724823594093323,
|
121 |
"step": 70
|
122 |
},
|
123 |
{
|
124 |
"epoch": 0.21,
|
125 |
"learning_rate": 4.828760511501322e-07,
|
126 |
+
"logits/chosen": -1.9419012069702148,
|
127 |
+
"logits/rejected": -1.9450546503067017,
|
128 |
+
"logps/chosen": -34.40644836425781,
|
129 |
+
"logps/rejected": -34.56013107299805,
|
130 |
+
"loss": 1.0055,
|
131 |
+
"rewards/accuracies": 0.5625,
|
132 |
+
"rewards/chosen": 0.010679833590984344,
|
133 |
+
"rewards/margins": -0.0054560392163693905,
|
134 |
+
"rewards/rejected": 0.016135873273015022,
|
135 |
"step": 80
|
136 |
},
|
137 |
{
|
138 |
"epoch": 0.23,
|
139 |
"learning_rate": 4.736716601303429e-07,
|
140 |
+
"logits/chosen": -1.9507659673690796,
|
141 |
+
"logits/rejected": -1.9552879333496094,
|
142 |
+
"logps/chosen": -32.45410919189453,
|
143 |
+
"logps/rejected": -32.34144973754883,
|
144 |
+
"loss": 1.0021,
|
145 |
+
"rewards/accuracies": 0.5249999761581421,
|
146 |
+
"rewards/chosen": 0.006657355930656195,
|
147 |
+
"rewards/margins": -0.002072322415187955,
|
148 |
+
"rewards/rejected": 0.008729678578674793,
|
149 |
"step": 90
|
150 |
},
|
151 |
{
|
152 |
"epoch": 0.26,
|
153 |
"learning_rate": 4.62624545834521e-07,
|
154 |
+
"logits/chosen": -2.0492916107177734,
|
155 |
+
"logits/rejected": -2.047295093536377,
|
156 |
+
"logps/chosen": -32.23640441894531,
|
157 |
+
"logps/rejected": -31.247751235961914,
|
158 |
+
"loss": 1.007,
|
159 |
+
"rewards/accuracies": 0.4375,
|
160 |
+
"rewards/chosen": 0.002216553781181574,
|
161 |
+
"rewards/margins": -0.007013450376689434,
|
162 |
+
"rewards/rejected": 0.009230004623532295,
|
163 |
"step": 100
|
164 |
},
|
165 |
{
|
166 |
"epoch": 0.26,
|
167 |
+
"eval_logits/chosen": -2.2440249919891357,
|
168 |
+
"eval_logits/rejected": -2.239149808883667,
|
169 |
+
"eval_logps/chosen": -34.017887115478516,
|
170 |
+
"eval_logps/rejected": -37.50767517089844,
|
171 |
+
"eval_loss": 0.9963483810424805,
|
172 |
+
"eval_rewards/accuracies": 0.5166113376617432,
|
173 |
+
"eval_rewards/chosen": 0.008332207798957825,
|
174 |
+
"eval_rewards/margins": 0.0038595683872699738,
|
175 |
+
"eval_rewards/rejected": 0.004472639411687851,
|
176 |
+
"eval_runtime": 146.1438,
|
177 |
+
"eval_samples_per_second": 2.347,
|
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": -2.005338191986084,
|
185 |
+
"logits/rejected": -2.002920627593994,
|
186 |
+
"logps/chosen": -33.22484588623047,
|
187 |
+
"logps/rejected": -34.01424026489258,
|
188 |
+
"loss": 0.9981,
|
189 |
+
"rewards/accuracies": 0.48750001192092896,
|
190 |
+
"rewards/chosen": 0.009485405869781971,
|
191 |
+
"rewards/margins": 0.001919901347719133,
|
192 |
+
"rewards/rejected": 0.007565504405647516,
|
193 |
"step": 110
|
194 |
},
|
195 |
{
|
196 |
"epoch": 0.31,
|
197 |
"learning_rate": 4.353806263777677e-07,
|
198 |
+
"logits/chosen": -2.0164568424224854,
|
199 |
+
"logits/rejected": -2.008070468902588,
|
200 |
+
"logps/chosen": -32.454193115234375,
|
201 |
+
"logps/rejected": -32.175106048583984,
|
202 |
+
"loss": 0.9978,
|
203 |
+
"rewards/accuracies": 0.4749999940395355,
|
204 |
+
"rewards/chosen": -0.004645171575248241,
|
205 |
+
"rewards/margins": 0.0021557719446718693,
|
206 |
+
"rewards/rejected": -0.006800943519920111,
|
207 |
"step": 120
|
208 |
},
|
209 |
{
|
210 |
"epoch": 0.34,
|
211 |
"learning_rate": 4.194082707715275e-07,
|
212 |
+
"logits/chosen": -2.046295642852783,
|
213 |
+
"logits/rejected": -2.03825044631958,
|
214 |
+
"logps/chosen": -30.491985321044922,
|
215 |
+
"logps/rejected": -32.04901885986328,
|
216 |
+
"loss": 1.0101,
|
217 |
+
"rewards/accuracies": 0.4625000059604645,
|
218 |
+
"rewards/chosen": -0.008134904317557812,
|
219 |
+
"rewards/margins": -0.010127933695912361,
|
220 |
+
"rewards/rejected": 0.001993028912693262,
|
221 |
"step": 130
|
222 |
},
|
223 |
{
|
224 |
"epoch": 0.36,
|
225 |
"learning_rate": 4.020402418666621e-07,
|
226 |
+
"logits/chosen": -1.9772052764892578,
|
227 |
+
"logits/rejected": -1.9874699115753174,
|
228 |
+
"logps/chosen": -31.385913848876953,
|
229 |
+
"logps/rejected": -32.56010818481445,
|
230 |
+
"loss": 0.9705,
|
231 |
+
"rewards/accuracies": 0.5625,
|
232 |
+
"rewards/chosen": 0.016873350366950035,
|
233 |
+
"rewards/margins": 0.02948569692671299,
|
234 |
+
"rewards/rejected": -0.01261234562844038,
|
235 |
"step": 140
|
236 |
},
|
237 |
{
|
238 |
"epoch": 0.39,
|
239 |
"learning_rate": 3.8341962650351185e-07,
|
240 |
+
"logits/chosen": -1.8907018899917603,
|
241 |
+
"logits/rejected": -1.8917919397354126,
|
242 |
+
"logps/chosen": -34.191062927246094,
|
243 |
+
"logps/rejected": -34.757171630859375,
|
244 |
+
"loss": 0.9992,
|
245 |
+
"rewards/accuracies": 0.5,
|
246 |
+
"rewards/chosen": 0.0008138298871926963,
|
247 |
+
"rewards/margins": 0.0007720142602920532,
|
248 |
+
"rewards/rejected": 4.181563781457953e-05,
|
249 |
"step": 150
|
250 |
},
|
251 |
{
|
252 |
"epoch": 0.42,
|
253 |
"learning_rate": 3.636998309800572e-07,
|
254 |
+
"logits/chosen": -1.9424282312393188,
|
255 |
+
"logits/rejected": -1.9389431476593018,
|
256 |
+
"logps/chosen": -36.157745361328125,
|
257 |
+
"logps/rejected": -32.73271942138672,
|
258 |
+
"loss": 0.988,
|
259 |
+
"rewards/accuracies": 0.5874999761581421,
|
260 |
+
"rewards/chosen": 0.0069676609709858894,
|
261 |
+
"rewards/margins": 0.011951583437621593,
|
262 |
+
"rewards/rejected": -0.004983922932296991,
|
263 |
"step": 160
|
264 |
},
|
265 |
{
|
266 |
"epoch": 0.44,
|
267 |
"learning_rate": 3.430433172111807e-07,
|
268 |
+
"logits/chosen": -2.0421340465545654,
|
269 |
+
"logits/rejected": -2.0347371101379395,
|
270 |
+
"logps/chosen": -33.798614501953125,
|
271 |
+
"logps/rejected": -31.370548248291016,
|
272 |
+
"loss": 0.991,
|
273 |
+
"rewards/accuracies": 0.5249999761581421,
|
274 |
+
"rewards/chosen": -0.00037657321081496775,
|
275 |
+
"rewards/margins": 0.008971447125077248,
|
276 |
+
"rewards/rejected": -0.00934801995754242,
|
277 |
"step": 170
|
278 |
},
|
279 |
{
|
280 |
"epoch": 0.47,
|
281 |
"learning_rate": 3.216202642830543e-07,
|
282 |
+
"logits/chosen": -2.0471675395965576,
|
283 |
+
"logits/rejected": -2.0524353981018066,
|
284 |
+
"logps/chosen": -32.519779205322266,
|
285 |
+
"logps/rejected": -32.49123001098633,
|
286 |
+
"loss": 0.9845,
|
287 |
+
"rewards/accuracies": 0.48750001192092896,
|
288 |
+
"rewards/chosen": 0.012504413723945618,
|
289 |
+
"rewards/margins": 0.015548673458397388,
|
290 |
+
"rewards/rejected": -0.003044259501621127,
|
291 |
"step": 180
|
292 |
},
|
293 |
{
|
294 |
"epoch": 0.49,
|
295 |
"learning_rate": 2.9960716642946403e-07,
|
296 |
+
"logits/chosen": -2.0482120513916016,
|
297 |
+
"logits/rejected": -2.0454375743865967,
|
298 |
+
"logps/chosen": -31.485937118530273,
|
299 |
+
"logps/rejected": -31.332433700561523,
|
300 |
+
"loss": 0.9914,
|
301 |
+
"rewards/accuracies": 0.574999988079071,
|
302 |
+
"rewards/chosen": 0.0015064775943756104,
|
303 |
+
"rewards/margins": 0.008616042323410511,
|
304 |
+
"rewards/rejected": -0.007109564729034901,
|
305 |
"step": 190
|
306 |
},
|
307 |
{
|
308 |
"epoch": 0.52,
|
309 |
"learning_rate": 2.771853789806683e-07,
|
310 |
+
"logits/chosen": -1.9187748432159424,
|
311 |
+
"logits/rejected": -1.9234533309936523,
|
312 |
+
"logps/chosen": -31.564071655273438,
|
313 |
+
"logps/rejected": -32.80377960205078,
|
314 |
+
"loss": 0.9725,
|
315 |
+
"rewards/accuracies": 0.625,
|
316 |
+
"rewards/chosen": 0.021104933694005013,
|
317 |
+
"rewards/margins": 0.027528975158929825,
|
318 |
+
"rewards/rejected": -0.006424039602279663,
|
319 |
"step": 200
|
320 |
},
|
321 |
{
|
322 |
"epoch": 0.52,
|
323 |
+
"eval_logits/chosen": -2.24338698387146,
|
324 |
+
"eval_logits/rejected": -2.2385094165802,
|
325 |
+
"eval_logps/chosen": -34.032955169677734,
|
326 |
+
"eval_logps/rejected": -37.49824142456055,
|
327 |
+
"eval_loss": 1.0085184574127197,
|
328 |
+
"eval_rewards/accuracies": 0.4991694688796997,
|
329 |
+
"eval_rewards/chosen": 0.0007996229687705636,
|
330 |
+
"eval_rewards/margins": -0.008389986120164394,
|
331 |
+
"eval_rewards/rejected": 0.009189609438180923,
|
332 |
+
"eval_runtime": 145.9067,
|
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": 2.5453962426402e-07,
|
340 |
+
"logits/chosen": -2.0320358276367188,
|
341 |
+
"logits/rejected": -2.042719602584839,
|
342 |
+
"logps/chosen": -31.940053939819336,
|
343 |
+
"logps/rejected": -33.88956069946289,
|
344 |
+
"loss": 0.9709,
|
345 |
+
"rewards/accuracies": 0.5375000238418579,
|
346 |
+
"rewards/chosen": 0.015625199303030968,
|
347 |
+
"rewards/margins": 0.029055744409561157,
|
348 |
+
"rewards/rejected": -0.01343054324388504,
|
349 |
"step": 210
|
350 |
},
|
351 |
{
|
352 |
"epoch": 0.57,
|
353 |
"learning_rate": 2.318564697655179e-07,
|
354 |
+
"logits/chosen": -1.9255237579345703,
|
355 |
+
"logits/rejected": -1.9403839111328125,
|
356 |
+
"logps/chosen": -30.091089248657227,
|
357 |
+
"logps/rejected": -31.54927635192871,
|
358 |
+
"loss": 0.9942,
|
359 |
+
"rewards/accuracies": 0.5,
|
360 |
+
"rewards/chosen": 0.010226013138890266,
|
361 |
+
"rewards/margins": 0.005776937119662762,
|
362 |
+
"rewards/rejected": 0.004449075553566217,
|
363 |
"step": 220
|
364 |
},
|
365 |
{
|
366 |
"epoch": 0.6,
|
367 |
"learning_rate": 2.093227910899832e-07,
|
368 |
+
"logits/chosen": -1.9833370447158813,
|
369 |
+
"logits/rejected": -1.9873158931732178,
|
370 |
+
"logps/chosen": -33.393131256103516,
|
371 |
+
"logps/rejected": -31.559478759765625,
|
372 |
+
"loss": 0.9827,
|
373 |
+
"rewards/accuracies": 0.512499988079071,
|
374 |
+
"rewards/chosen": 0.012168514542281628,
|
375 |
+
"rewards/margins": 0.017256852239370346,
|
376 |
+
"rewards/rejected": -0.005088338162750006,
|
377 |
"step": 230
|
378 |
},
|
379 |
{
|
380 |
"epoch": 0.62,
|
381 |
"learning_rate": 1.8712423238279356e-07,
|
382 |
+
"logits/chosen": -1.9829031229019165,
|
383 |
+
"logits/rejected": -1.9609248638153076,
|
384 |
+
"logps/chosen": -34.161617279052734,
|
385 |
+
"logps/rejected": -34.973785400390625,
|
386 |
+
"loss": 0.9924,
|
387 |
+
"rewards/accuracies": 0.5,
|
388 |
+
"rewards/chosen": -0.003932595252990723,
|
389 |
+
"rewards/margins": 0.007563352584838867,
|
390 |
+
"rewards/rejected": -0.01149594783782959,
|
391 |
"step": 240
|
392 |
},
|
393 |
{
|
394 |
"epoch": 0.65,
|
395 |
"learning_rate": 1.654436768970182e-07,
|
396 |
+
"logits/chosen": -2.0244410037994385,
|
397 |
+
"logits/rejected": -2.0211281776428223,
|
398 |
+
"logps/chosen": -32.90896987915039,
|
399 |
+
"logps/rejected": -36.21443557739258,
|
400 |
+
"loss": 0.9972,
|
401 |
+
"rewards/accuracies": 0.4749999940395355,
|
402 |
+
"rewards/chosen": 0.008049068041145802,
|
403 |
+
"rewards/margins": 0.0028051852714270353,
|
404 |
+
"rewards/rejected": 0.0052438825368881226,
|
405 |
"step": 250
|
406 |
},
|
407 |
{
|
408 |
"epoch": 0.68,
|
409 |
"learning_rate": 1.444597403062196e-07,
|
410 |
+
"logits/chosen": -1.8918297290802002,
|
411 |
+
"logits/rejected": -1.8893934488296509,
|
412 |
+
"logps/chosen": -34.178001403808594,
|
413 |
+
"logps/rejected": -35.52972412109375,
|
414 |
+
"loss": 0.9823,
|
415 |
+
"rewards/accuracies": 0.5874999761581421,
|
416 |
+
"rewards/chosen": 0.00519949197769165,
|
417 |
+
"rewards/margins": 0.017749812453985214,
|
418 |
+
"rewards/rejected": -0.012550321407616138,
|
419 |
"step": 260
|
420 |
},
|
421 |
{
|
422 |
"epoch": 0.7,
|
423 |
"learning_rate": 1.2434529917578887e-07,
|
424 |
+
"logits/chosen": -1.875998854637146,
|
425 |
+
"logits/rejected": -1.873445749282837,
|
426 |
+
"logps/chosen": -34.39997100830078,
|
427 |
+
"logps/rejected": -31.749908447265625,
|
428 |
+
"loss": 1.0119,
|
429 |
"rewards/accuracies": 0.4625000059604645,
|
430 |
+
"rewards/chosen": -0.0011831462616100907,
|
431 |
+
"rewards/margins": -0.011894017457962036,
|
432 |
+
"rewards/rejected": 0.01071087084710598,
|
433 |
"step": 270
|
434 |
},
|
435 |
{
|
436 |
"epoch": 0.73,
|
437 |
"learning_rate": 1.0526606671603521e-07,
|
438 |
+
"logits/chosen": -1.9797567129135132,
|
439 |
+
"logits/rejected": -1.9691368341445923,
|
440 |
+
"logps/chosen": -35.324066162109375,
|
441 |
+
"logps/rejected": -31.857070922851562,
|
442 |
+
"loss": 0.9757,
|
443 |
+
"rewards/accuracies": 0.5249999761581421,
|
444 |
+
"rewards/chosen": 0.016287770122289658,
|
445 |
+
"rewards/margins": 0.024326175451278687,
|
446 |
+
"rewards/rejected": -0.008038404397666454,
|
447 |
"step": 280
|
448 |
},
|
449 |
{
|
450 |
"epoch": 0.75,
|
451 |
"learning_rate": 8.737922755071453e-08,
|
452 |
+
"logits/chosen": -2.0752615928649902,
|
453 |
+
"logits/rejected": -2.06022310256958,
|
454 |
+
"logps/chosen": -30.904525756835938,
|
455 |
+
"logps/rejected": -32.66613006591797,
|
456 |
+
"loss": 0.9831,
|
457 |
+
"rewards/accuracies": 0.5249999761581421,
|
458 |
+
"rewards/chosen": 0.00980077963322401,
|
459 |
+
"rewards/margins": 0.016948198899626732,
|
460 |
+
"rewards/rejected": -0.007147419266402721,
|
461 |
"step": 290
|
462 |
},
|
463 |
{
|
464 |
"epoch": 0.78,
|
465 |
"learning_rate": 7.08321427484816e-08,
|
466 |
+
"logits/chosen": -1.946509599685669,
|
467 |
+
"logits/rejected": -1.9439618587493896,
|
468 |
+
"logps/chosen": -32.89145278930664,
|
469 |
+
"logps/rejected": -30.81719398498535,
|
470 |
+
"loss": 0.982,
|
471 |
+
"rewards/accuracies": 0.5375000238418579,
|
472 |
+
"rewards/chosen": 0.017829114571213722,
|
473 |
+
"rewards/margins": 0.018021035939455032,
|
474 |
+
"rewards/rejected": -0.00019192098989151418,
|
475 |
"step": 300
|
476 |
},
|
477 |
{
|
478 |
"epoch": 0.78,
|
479 |
+
"eval_logits/chosen": -2.2438137531280518,
|
480 |
+
"eval_logits/rejected": -2.2389373779296875,
|
481 |
+
"eval_logps/chosen": -34.02228546142578,
|
482 |
+
"eval_logps/rejected": -37.50155258178711,
|
483 |
+
"eval_loss": 1.0014283657073975,
|
484 |
+
"eval_rewards/accuracies": 0.5103820562362671,
|
485 |
+
"eval_rewards/chosen": 0.006133119110018015,
|
486 |
+
"eval_rewards/margins": -0.001401092391461134,
|
487 |
+
"eval_rewards/rejected": 0.007534211501479149,
|
488 |
+
"eval_runtime": 145.9391,
|
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": -1.9287340641021729,
|
497 |
+
"logits/rejected": -1.9254703521728516,
|
498 |
+
"logps/chosen": -31.576568603515625,
|
499 |
+
"logps/rejected": -33.73060607910156,
|
500 |
+
"loss": 0.9924,
|
501 |
+
"rewards/accuracies": 0.5249999761581421,
|
502 |
+
"rewards/chosen": 0.013391390442848206,
|
503 |
+
"rewards/margins": 0.00761887151747942,
|
504 |
+
"rewards/rejected": 0.005772518925368786,
|
505 |
"step": 310
|
506 |
},
|
507 |
{
|
508 |
"epoch": 0.83,
|
509 |
"learning_rate": 4.229036944380912e-08,
|
510 |
+
"logits/chosen": -1.980047583580017,
|
511 |
+
"logits/rejected": -1.9677263498306274,
|
512 |
+
"logps/chosen": -34.565635681152344,
|
513 |
+
"logps/rejected": -33.560508728027344,
|
514 |
+
"loss": 0.9739,
|
515 |
+
"rewards/accuracies": 0.5249999761581421,
|
516 |
+
"rewards/chosen": 0.015218675136566162,
|
517 |
+
"rewards/margins": 0.02605234459042549,
|
518 |
+
"rewards/rejected": -0.010833668522536755,
|
519 |
"step": 320
|
520 |
},
|
521 |
{
|
522 |
"epoch": 0.86,
|
523 |
"learning_rate": 3.053082288996112e-08,
|
524 |
+
"logits/chosen": -2.0154004096984863,
|
525 |
+
"logits/rejected": -2.013948917388916,
|
526 |
+
"logps/chosen": -33.46211624145508,
|
527 |
+
"logps/rejected": -32.472103118896484,
|
528 |
+
"loss": 0.9919,
|
529 |
+
"rewards/accuracies": 0.5249999761581421,
|
530 |
+
"rewards/chosen": 0.011838942766189575,
|
531 |
+
"rewards/margins": 0.008100616745650768,
|
532 |
+
"rewards/rejected": 0.0037383257877081633,
|
533 |
"step": 330
|
534 |
},
|
535 |
{
|
536 |
"epoch": 0.88,
|
537 |
"learning_rate": 2.05793773749158e-08,
|
538 |
+
"logits/chosen": -2.1025614738464355,
|
539 |
+
"logits/rejected": -2.086764097213745,
|
540 |
+
"logps/chosen": -34.168495178222656,
|
541 |
+
"logps/rejected": -33.07737731933594,
|
542 |
+
"loss": 1.0149,
|
543 |
"rewards/accuracies": 0.42500001192092896,
|
544 |
+
"rewards/chosen": 0.005447447299957275,
|
545 |
+
"rewards/margins": -0.014881876297295094,
|
546 |
+
"rewards/rejected": 0.020329322665929794,
|
547 |
"step": 340
|
548 |
},
|
549 |
{
|
550 |
"epoch": 0.91,
|
551 |
"learning_rate": 1.251801807404168e-08,
|
552 |
+
"logits/chosen": -1.9740177392959595,
|
553 |
+
"logits/rejected": -1.9730663299560547,
|
554 |
+
"logps/chosen": -33.242271423339844,
|
555 |
+
"logps/rejected": -32.460182189941406,
|
556 |
+
"loss": 0.9826,
|
557 |
+
"rewards/accuracies": 0.550000011920929,
|
558 |
+
"rewards/chosen": 0.019913053140044212,
|
559 |
+
"rewards/margins": 0.017432674765586853,
|
560 |
+
"rewards/rejected": 0.0024803788401186466,
|
561 |
"step": 350
|
562 |
},
|
563 |
{
|
564 |
"epoch": 0.94,
|
565 |
"learning_rate": 6.41315865106129e-09,
|
566 |
+
"logits/chosen": -1.9304364919662476,
|
567 |
+
"logits/rejected": -1.9407978057861328,
|
568 |
+
"logps/chosen": -32.19029998779297,
|
569 |
+
"logps/rejected": -35.303955078125,
|
570 |
+
"loss": 0.9852,
|
571 |
+
"rewards/accuracies": 0.5874999761581421,
|
572 |
+
"rewards/chosen": 0.007193857338279486,
|
573 |
+
"rewards/margins": 0.014792281202971935,
|
574 |
+
"rewards/rejected": -0.00759842386469245,
|
575 |
"step": 360
|
576 |
},
|
577 |
{
|
578 |
"epoch": 0.96,
|
579 |
"learning_rate": 2.3150941078050324e-09,
|
580 |
+
"logits/chosen": -2.069343090057373,
|
581 |
+
"logits/rejected": -2.062788724899292,
|
582 |
+
"logps/chosen": -33.646358489990234,
|
583 |
+
"logps/rejected": -29.213809967041016,
|
584 |
+
"loss": 1.0035,
|
585 |
+
"rewards/accuracies": 0.4625000059604645,
|
586 |
+
"rewards/chosen": 0.0028047324158251286,
|
587 |
+
"rewards/margins": -0.0035038769710808992,
|
588 |
+
"rewards/rejected": 0.006308609154075384,
|
589 |
"step": 370
|
590 |
},
|
591 |
{
|
592 |
"epoch": 0.99,
|
593 |
"learning_rate": 2.575864278703266e-10,
|
594 |
+
"logits/chosen": -1.9290740489959717,
|
595 |
+
"logits/rejected": -1.9312365055084229,
|
596 |
+
"logps/chosen": -34.245155334472656,
|
597 |
+
"logps/rejected": -30.889087677001953,
|
598 |
+
"loss": 0.9948,
|
599 |
+
"rewards/accuracies": 0.574999988079071,
|
600 |
+
"rewards/chosen": -0.003053072141483426,
|
601 |
+
"rewards/margins": 0.005176535341888666,
|
602 |
+
"rewards/rejected": -0.008229607716202736,
|
603 |
"step": 380
|
604 |
},
|
605 |
{
|
606 |
"epoch": 1.0,
|
607 |
"step": 385,
|
608 |
"total_flos": 0.0,
|
609 |
+
"train_loss": 0.9930156608680626,
|
610 |
+
"train_runtime": 3255.7181,
|
611 |
"train_samples_per_second": 0.946,
|
612 |
"train_steps_per_second": 0.118
|
613 |
}
|