hugodk-sch
commited on
Commit
•
1950075
1
Parent(s):
31c48fa
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 +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: -
|
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: 0.4464
|
21 |
+
- Rewards/chosen: 0.0179
|
22 |
+
- Rewards/rejected: -0.0124
|
23 |
+
- Rewards/accuracies: 0.5428
|
24 |
+
- Rewards/margins: 0.0304
|
25 |
+
- Logps/rejected: -37.5321
|
26 |
+
- Logps/chosen: -34.0121
|
27 |
+
- Logits/rejected: -2.2303
|
28 |
+
- Logits/chosen: -2.2352
|
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 |
+
| 0.3817 | 0.26 | 100 | 0.4287 | 0.0507 | 0.0294 | 0.5577 | 0.0213 | -37.4799 | -33.9712 | -2.2329 | -2.2378 |
|
64 |
+
| 0.3991 | 0.52 | 200 | 0.4536 | -0.0010 | -0.0197 | 0.5374 | 0.0187 | -37.5412 | -34.0358 | -2.2300 | -2.2349 |
|
65 |
+
| 0.2816 | 0.78 | 300 | 0.4464 | 0.0179 | -0.0124 | 0.5428 | 0.0304 | -37.5321 | -34.0121 | -2.2303 | -2.2352 |
|
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 sha256:9323562170691969f33b21814d35e8b1eb7c2c35b3293abb03e7e74ca2e2815d
|
3 |
size 176183216
|
all_results.json
CHANGED
@@ -1,20 +1,7 @@
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
-
"
|
4 |
-
"
|
5 |
-
"eval_logps/chosen": -33.767173767089844,
|
6 |
-
"eval_logps/rejected": -37.282623291015625,
|
7 |
-
"eval_loss": 0.49339157342910767,
|
8 |
-
"eval_rewards/accuracies": 0.5456810593605042,
|
9 |
-
"eval_rewards/chosen": 0.21390400826931,
|
10 |
-
"eval_rewards/margins": 0.026708098128437996,
|
11 |
-
"eval_rewards/rejected": 0.18719588220119476,
|
12 |
-
"eval_runtime": 145.5786,
|
13 |
-
"eval_samples": 343,
|
14 |
-
"eval_samples_per_second": 2.356,
|
15 |
-
"eval_steps_per_second": 0.295,
|
16 |
-
"train_loss": 0.4538224170734356,
|
17 |
-
"train_runtime": 3252.427,
|
18 |
"train_samples": 3079,
|
19 |
"train_samples_per_second": 0.947,
|
20 |
"train_steps_per_second": 0.118
|
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
+
"train_loss": 0.3756082092012678,
|
4 |
+
"train_runtime": 3251.4223,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
5 |
"train_samples": 3079,
|
6 |
"train_samples_per_second": 0.947,
|
7 |
"train_steps_per_second": 0.118
|
train_results.json
CHANGED
@@ -1,7 +1,7 @@
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
-
"train_loss": 0.
|
4 |
-
"train_runtime":
|
5 |
"train_samples": 3079,
|
6 |
"train_samples_per_second": 0.947,
|
7 |
"train_steps_per_second": 0.118
|
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
+
"train_loss": 0.3756082092012678,
|
4 |
+
"train_runtime": 3251.4223,
|
5 |
"train_samples": 3079,
|
6 |
"train_samples_per_second": 0.947,
|
7 |
"train_steps_per_second": 0.118
|
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,
|
@@ -25,589 +25,589 @@
|
|
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": -
|
44 |
-
"logps/chosen": -29.
|
45 |
-
"logps/rejected": -29.
|
46 |
-
"loss": 0.
|
47 |
"rewards/accuracies": 0.48750001192092896,
|
48 |
-
"rewards/chosen": 0.
|
49 |
-
"rewards/margins": -0.
|
50 |
-
"rewards/rejected": 0.
|
51 |
"step": 20
|
52 |
},
|
53 |
{
|
54 |
"epoch": 0.08,
|
55 |
"learning_rate": 3.846153846153847e-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":
|
65 |
"step": 30
|
66 |
},
|
67 |
{
|
68 |
"epoch": 0.1,
|
69 |
"learning_rate": 4.999896948438434e-06,
|
70 |
-
"logits/chosen": -2.
|
71 |
-
"logits/rejected": -2.
|
72 |
-
"logps/chosen": -32.
|
73 |
-
"logps/rejected": -32.
|
74 |
-
"loss": 0.
|
75 |
-
"rewards/accuracies": 0.
|
76 |
-
"rewards/chosen": 0.
|
77 |
-
"rewards/margins": 0.
|
78 |
-
"rewards/rejected":
|
79 |
"step": 40
|
80 |
},
|
81 |
{
|
82 |
"epoch": 0.13,
|
83 |
"learning_rate": 4.987541037542187e-06,
|
84 |
-
"logits/chosen": -1.
|
85 |
-
"logits/rejected": -1.
|
86 |
-
"logps/chosen": -33.
|
87 |
-
"logps/rejected": -35.
|
88 |
-
"loss": 0.
|
89 |
-
"rewards/accuracies": 0.
|
90 |
-
"rewards/chosen": 0.
|
91 |
-
"rewards/margins":
|
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": -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-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.
|
146 |
-
"rewards/chosen": 0.
|
147 |
-
"rewards/margins": 0.
|
148 |
-
"rewards/rejected": 0.
|
149 |
"step": 90
|
150 |
},
|
151 |
{
|
152 |
"epoch": 0.26,
|
153 |
"learning_rate": 4.626245458345211e-06,
|
154 |
-
"logits/chosen": -2.
|
155 |
-
"logits/rejected": -2.
|
156 |
-
"logps/chosen": -
|
157 |
-
"logps/rejected": -31.
|
158 |
-
"loss": 0.
|
159 |
-
"rewards/accuracies": 0.
|
160 |
-
"rewards/chosen": 0.
|
161 |
-
"rewards/margins": 0.
|
162 |
-
"rewards/rejected": 0.
|
163 |
"step": 100
|
164 |
},
|
165 |
{
|
166 |
"epoch": 0.26,
|
167 |
-
"eval_logits/chosen": -2.
|
168 |
-
"eval_logits/rejected": -2.
|
169 |
-
"eval_logps/chosen": -33.
|
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":
|
177 |
-
"eval_samples_per_second": 2.
|
178 |
-
"eval_steps_per_second": 0.
|
179 |
"step": 100
|
180 |
},
|
181 |
{
|
182 |
"epoch": 0.29,
|
183 |
"learning_rate": 4.498257201263691e-06,
|
184 |
-
"logits/chosen": -1.
|
185 |
-
"logits/rejected": -1.
|
186 |
-
"logps/chosen": -
|
187 |
-
"logps/rejected": -33.
|
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": -2.
|
200 |
-
"logps/chosen": -32.
|
201 |
-
"logps/rejected": -
|
202 |
-
"loss": 0.
|
203 |
-
"rewards/accuracies": 0.
|
204 |
-
"rewards/chosen": 0.
|
205 |
-
"rewards/margins": 0.
|
206 |
-
"rewards/rejected": 0.
|
207 |
"step": 120
|
208 |
},
|
209 |
{
|
210 |
"epoch": 0.34,
|
211 |
"learning_rate": 4.1940827077152755e-06,
|
212 |
-
"logits/chosen": -2.
|
213 |
-
"logits/rejected": -2.
|
214 |
-
"logps/chosen": -30.
|
215 |
-
"logps/rejected": -
|
216 |
-
"loss": 0.
|
217 |
-
"rewards/accuracies": 0.
|
218 |
-
"rewards/chosen": 0.
|
219 |
-
"rewards/margins": 0.
|
220 |
-
"rewards/rejected": 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": 0.625,
|
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": -
|
257 |
-
"logps/rejected": -32.
|
258 |
-
"loss": 0.
|
259 |
-
"rewards/accuracies": 0.
|
260 |
-
"rewards/chosen": 0.
|
261 |
-
"rewards/margins": 0.
|
262 |
-
"rewards/rejected": 0.
|
263 |
"step": 160
|
264 |
},
|
265 |
{
|
266 |
"epoch": 0.44,
|
267 |
"learning_rate": 3.4304331721118078e-06,
|
268 |
-
"logits/chosen": -2.
|
269 |
-
"logits/rejected": -2.
|
270 |
-
"logps/chosen": -33.
|
271 |
-
"logps/rejected": -31.
|
272 |
-
"loss": 0.
|
273 |
"rewards/accuracies": 0.6875,
|
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": -
|
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": 0.6875,
|
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": -
|
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.
|
335 |
"step": 200
|
336 |
},
|
337 |
{
|
338 |
"epoch": 0.55,
|
339 |
"learning_rate": 2.5453962426402006e-06,
|
340 |
-
"logits/chosen": -2.
|
341 |
-
"logits/rejected": -2.
|
342 |
-
"logps/chosen": -31.
|
343 |
-
"logps/rejected": -33.
|
344 |
-
"loss": 0.
|
345 |
-
"rewards/accuracies": 0.
|
346 |
-
"rewards/chosen": 0.
|
347 |
-
"rewards/margins": 0.
|
348 |
-
"rewards/rejected": 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": -
|
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": -
|
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": -2.
|
398 |
-
"logps/chosen": -32.
|
399 |
-
"logps/rejected": -
|
400 |
-
"loss": 0.
|
401 |
-
"rewards/accuracies": 0.
|
402 |
-
"rewards/chosen": 0.
|
403 |
-
"rewards/margins": 0.
|
404 |
-
"rewards/rejected": 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": -
|
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": -
|
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": 0.75,
|
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": -
|
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": -2.
|
525 |
-
"logits/rejected": -2.
|
526 |
-
"logps/chosen": -
|
527 |
-
"logps/rejected": -32.
|
528 |
-
"loss": 0.
|
529 |
"rewards/accuracies": 0.6499999761581421,
|
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": -
|
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": -
|
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": 0.947,
|
612 |
"train_steps_per_second": 0.118
|
613 |
}
|
|
|
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": 1.282051282051282e-06,
|
28 |
+
"logits/chosen": -1.8662821054458618,
|
29 |
+
"logits/rejected": -1.8706018924713135,
|
30 |
+
"logps/chosen": -36.98260498046875,
|
31 |
+
"logps/rejected": -33.66376876831055,
|
32 |
+
"loss": 0.3501,
|
33 |
+
"rewards/accuracies": 0.5694444179534912,
|
34 |
+
"rewards/chosen": 0.01919599249958992,
|
35 |
+
"rewards/margins": 0.04614981636404991,
|
36 |
+
"rewards/rejected": -0.02695382386445999,
|
37 |
"step": 10
|
38 |
},
|
39 |
{
|
40 |
"epoch": 0.05,
|
41 |
"learning_rate": 2.564102564102564e-06,
|
42 |
+
"logits/chosen": -1.9981708526611328,
|
43 |
+
"logits/rejected": -2.000824451446533,
|
44 |
+
"logps/chosen": -29.6453914642334,
|
45 |
+
"logps/rejected": -29.056461334228516,
|
46 |
+
"loss": 0.4314,
|
47 |
"rewards/accuracies": 0.48750001192092896,
|
48 |
+
"rewards/chosen": -0.0025594178587198257,
|
49 |
+
"rewards/margins": -0.008148794062435627,
|
50 |
+
"rewards/rejected": 0.005589376203715801,
|
51 |
"step": 20
|
52 |
},
|
53 |
{
|
54 |
"epoch": 0.08,
|
55 |
"learning_rate": 3.846153846153847e-06,
|
56 |
+
"logits/chosen": -1.9205211400985718,
|
57 |
+
"logits/rejected": -1.917824149131775,
|
58 |
+
"logps/chosen": -31.402332305908203,
|
59 |
+
"logps/rejected": -33.20569610595703,
|
60 |
+
"loss": 0.4366,
|
61 |
+
"rewards/accuracies": 0.574999988079071,
|
62 |
+
"rewards/chosen": 0.011007689870893955,
|
63 |
+
"rewards/margins": 0.0007972270250320435,
|
64 |
+
"rewards/rejected": 0.010210464708507061,
|
65 |
"step": 30
|
66 |
},
|
67 |
{
|
68 |
"epoch": 0.1,
|
69 |
"learning_rate": 4.999896948438434e-06,
|
70 |
+
"logits/chosen": -2.0170958042144775,
|
71 |
+
"logits/rejected": -2.0083703994750977,
|
72 |
+
"logps/chosen": -32.547943115234375,
|
73 |
+
"logps/rejected": -32.49810791015625,
|
74 |
+
"loss": 0.4286,
|
75 |
+
"rewards/accuracies": 0.5375000238418579,
|
76 |
+
"rewards/chosen": 0.023093106225132942,
|
77 |
+
"rewards/margins": 0.011441526934504509,
|
78 |
+
"rewards/rejected": 0.011651577427983284,
|
79 |
"step": 40
|
80 |
},
|
81 |
{
|
82 |
"epoch": 0.13,
|
83 |
"learning_rate": 4.987541037542187e-06,
|
84 |
+
"logits/chosen": -1.8613898754119873,
|
85 |
+
"logits/rejected": -1.8506320714950562,
|
86 |
+
"logps/chosen": -33.56399154663086,
|
87 |
+
"logps/rejected": -35.46310043334961,
|
88 |
+
"loss": 0.4498,
|
89 |
+
"rewards/accuracies": 0.4625000059604645,
|
90 |
+
"rewards/chosen": -0.005208463408052921,
|
91 |
+
"rewards/margins": 0.0033974028192460537,
|
92 |
+
"rewards/rejected": -0.008605867624282837,
|
93 |
"step": 50
|
94 |
},
|
95 |
{
|
96 |
"epoch": 0.16,
|
97 |
"learning_rate": 4.954691471941119e-06,
|
98 |
+
"logits/chosen": -1.9389715194702148,
|
99 |
+
"logits/rejected": -1.9409143924713135,
|
100 |
+
"logps/chosen": -32.57710266113281,
|
101 |
+
"logps/rejected": -33.21430206298828,
|
102 |
+
"loss": 0.3914,
|
103 |
+
"rewards/accuracies": 0.612500011920929,
|
104 |
+
"rewards/chosen": 0.018466468900442123,
|
105 |
+
"rewards/margins": 0.05280427262187004,
|
106 |
+
"rewards/rejected": -0.03433779999613762,
|
107 |
"step": 60
|
108 |
},
|
109 |
{
|
110 |
"epoch": 0.18,
|
111 |
"learning_rate": 4.901618883413549e-06,
|
112 |
+
"logits/chosen": -2.0723679065704346,
|
113 |
+
"logits/rejected": -2.0773472785949707,
|
114 |
+
"logps/chosen": -33.946128845214844,
|
115 |
+
"logps/rejected": -36.60929489135742,
|
116 |
+
"loss": 0.4672,
|
117 |
+
"rewards/accuracies": 0.5,
|
118 |
+
"rewards/chosen": 0.033884815871715546,
|
119 |
+
"rewards/margins": 0.0662197396159172,
|
120 |
+
"rewards/rejected": -0.03233493119478226,
|
121 |
"step": 70
|
122 |
},
|
123 |
{
|
124 |
"epoch": 0.21,
|
125 |
"learning_rate": 4.828760511501322e-06,
|
126 |
+
"logits/chosen": -1.9366763830184937,
|
127 |
+
"logits/rejected": -1.9398372173309326,
|
128 |
+
"logps/chosen": -34.32460403442383,
|
129 |
+
"logps/rejected": -34.59550094604492,
|
130 |
+
"loss": 0.3932,
|
131 |
+
"rewards/accuracies": 0.5625,
|
132 |
+
"rewards/chosen": 0.0825641006231308,
|
133 |
+
"rewards/margins": 0.08503931760787964,
|
134 |
+
"rewards/rejected": -0.002475212560966611,
|
135 |
"step": 80
|
136 |
},
|
137 |
{
|
138 |
"epoch": 0.23,
|
139 |
"learning_rate": 4.7367166013034295e-06,
|
140 |
+
"logits/chosen": -1.9469963312149048,
|
141 |
+
"logits/rejected": -1.9515202045440674,
|
142 |
+
"logps/chosen": -32.401084899902344,
|
143 |
+
"logps/rejected": -32.32483673095703,
|
144 |
+
"loss": 0.4402,
|
145 |
+
"rewards/accuracies": 0.574999988079071,
|
146 |
+
"rewards/chosen": 0.05306895822286606,
|
147 |
+
"rewards/margins": 0.02581069990992546,
|
148 |
+
"rewards/rejected": 0.027258265763521194,
|
149 |
"step": 90
|
150 |
},
|
151 |
{
|
152 |
"epoch": 0.26,
|
153 |
"learning_rate": 4.626245458345211e-06,
|
154 |
+
"logits/chosen": -2.0446877479553223,
|
155 |
+
"logits/rejected": -2.0427000522613525,
|
156 |
+
"logps/chosen": -32.168331146240234,
|
157 |
+
"logps/rejected": -31.263940811157227,
|
158 |
+
"loss": 0.3817,
|
159 |
+
"rewards/accuracies": 0.5375000238418579,
|
160 |
+
"rewards/chosen": 0.05800473690032959,
|
161 |
+
"rewards/margins": 0.056188035756349564,
|
162 |
+
"rewards/rejected": 0.0018166989320889115,
|
163 |
"step": 100
|
164 |
},
|
165 |
{
|
166 |
"epoch": 0.26,
|
167 |
+
"eval_logits/chosen": -2.2377800941467285,
|
168 |
+
"eval_logits/rejected": -2.2329328060150146,
|
169 |
+
"eval_logps/chosen": -33.97116470336914,
|
170 |
+
"eval_logps/rejected": -37.47990036010742,
|
171 |
+
"eval_loss": 0.42868927121162415,
|
172 |
+
"eval_rewards/accuracies": 0.5577242374420166,
|
173 |
+
"eval_rewards/chosen": 0.05071057379245758,
|
174 |
+
"eval_rewards/margins": 0.021336428821086884,
|
175 |
+
"eval_rewards/rejected": 0.029374146834015846,
|
176 |
+
"eval_runtime": 146.0732,
|
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": 4.498257201263691e-06,
|
184 |
+
"logits/chosen": -1.9993603229522705,
|
185 |
+
"logits/rejected": -1.996971845626831,
|
186 |
+
"logps/chosen": -33.112144470214844,
|
187 |
+
"logps/rejected": -33.986610412597656,
|
188 |
+
"loss": 0.5297,
|
189 |
+
"rewards/accuracies": 0.574999988079071,
|
190 |
+
"rewards/chosen": 0.105336032807827,
|
191 |
+
"rewards/margins": 0.07112538814544678,
|
192 |
+
"rewards/rejected": 0.03421063348650932,
|
193 |
"step": 110
|
194 |
},
|
195 |
{
|
196 |
"epoch": 0.31,
|
197 |
"learning_rate": 4.353806263777678e-06,
|
198 |
+
"logits/chosen": -2.009766101837158,
|
199 |
+
"logits/rejected": -2.0014333724975586,
|
200 |
+
"logps/chosen": -32.351280212402344,
|
201 |
+
"logps/rejected": -32.099830627441406,
|
202 |
+
"loss": 0.5288,
|
203 |
+
"rewards/accuracies": 0.550000011920929,
|
204 |
+
"rewards/chosen": 0.07490243762731552,
|
205 |
+
"rewards/margins": 0.025563379749655724,
|
206 |
+
"rewards/rejected": 0.04933905601501465,
|
207 |
"step": 120
|
208 |
},
|
209 |
{
|
210 |
"epoch": 0.34,
|
211 |
"learning_rate": 4.1940827077152755e-06,
|
212 |
+
"logits/chosen": -2.0372567176818848,
|
213 |
+
"logits/rejected": -2.029240369796753,
|
214 |
+
"logps/chosen": -30.34758949279785,
|
215 |
+
"logps/rejected": -32.03752899169922,
|
216 |
+
"loss": 0.4171,
|
217 |
+
"rewards/accuracies": 0.550000011920929,
|
218 |
+
"rewards/chosen": 0.10250232368707657,
|
219 |
+
"rewards/margins": 0.09011942148208618,
|
220 |
+
"rewards/rejected": 0.01238289289176464,
|
221 |
"step": 130
|
222 |
},
|
223 |
{
|
224 |
"epoch": 0.36,
|
225 |
"learning_rate": 4.0204024186666215e-06,
|
226 |
+
"logits/chosen": -1.966684341430664,
|
227 |
+
"logits/rejected": -1.9769500494003296,
|
228 |
+
"logps/chosen": -31.219501495361328,
|
229 |
+
"logps/rejected": -32.558082580566406,
|
230 |
+
"loss": 0.4241,
|
231 |
+
"rewards/accuracies": 0.675000011920929,
|
232 |
+
"rewards/chosen": 0.1601306051015854,
|
233 |
+
"rewards/margins": 0.178691104054451,
|
234 |
+
"rewards/rejected": -0.018560513854026794,
|
235 |
"step": 140
|
236 |
},
|
237 |
{
|
238 |
"epoch": 0.39,
|
239 |
"learning_rate": 3.834196265035119e-06,
|
240 |
+
"logits/chosen": -1.8799800872802734,
|
241 |
+
"logits/rejected": -1.8811286687850952,
|
242 |
+
"logps/chosen": -34.00492858886719,
|
243 |
+
"logps/rejected": -34.78579330444336,
|
244 |
+
"loss": 0.3742,
|
245 |
"rewards/accuracies": 0.625,
|
246 |
+
"rewards/chosen": 0.15021029114723206,
|
247 |
+
"rewards/margins": 0.17304366827011108,
|
248 |
+
"rewards/rejected": -0.02283337712287903,
|
249 |
"step": 150
|
250 |
},
|
251 |
{
|
252 |
"epoch": 0.42,
|
253 |
"learning_rate": 3.636998309800573e-06,
|
254 |
+
"logits/chosen": -1.932443618774414,
|
255 |
+
"logits/rejected": -1.9290469884872437,
|
256 |
+
"logps/chosen": -36.061256408691406,
|
257 |
+
"logps/rejected": -32.7225456237793,
|
258 |
+
"loss": 0.377,
|
259 |
+
"rewards/accuracies": 0.625,
|
260 |
+
"rewards/chosen": 0.0883389413356781,
|
261 |
+
"rewards/margins": 0.0881747230887413,
|
262 |
+
"rewards/rejected": 0.00016421750478912145,
|
263 |
"step": 160
|
264 |
},
|
265 |
{
|
266 |
"epoch": 0.44,
|
267 |
"learning_rate": 3.4304331721118078e-06,
|
268 |
+
"logits/chosen": -2.0334668159484863,
|
269 |
+
"logits/rejected": -2.026076078414917,
|
270 |
+
"logps/chosen": -33.546730041503906,
|
271 |
+
"logps/rejected": -31.355152130126953,
|
272 |
+
"loss": 0.3457,
|
273 |
"rewards/accuracies": 0.6875,
|
274 |
+
"rewards/chosen": 0.2009061872959137,
|
275 |
+
"rewards/margins": 0.2035447061061859,
|
276 |
+
"rewards/rejected": -0.0026385621167719364,
|
277 |
"step": 170
|
278 |
},
|
279 |
{
|
280 |
"epoch": 0.47,
|
281 |
"learning_rate": 3.2162026428305436e-06,
|
282 |
+
"logits/chosen": -2.0394930839538574,
|
283 |
+
"logits/rejected": -2.044752359390259,
|
284 |
+
"logps/chosen": -32.33013153076172,
|
285 |
+
"logps/rejected": -32.470909118652344,
|
286 |
+
"loss": 0.2919,
|
287 |
+
"rewards/accuracies": 0.6875,
|
288 |
+
"rewards/chosen": 0.17172439396381378,
|
289 |
+
"rewards/margins": 0.1603413224220276,
|
290 |
+
"rewards/rejected": 0.01138305850327015,
|
291 |
"step": 180
|
292 |
},
|
293 |
{
|
294 |
"epoch": 0.49,
|
295 |
"learning_rate": 2.996071664294641e-06,
|
296 |
+
"logits/chosen": -2.0403592586517334,
|
297 |
+
"logits/rejected": -2.0375704765319824,
|
298 |
+
"logps/chosen": -31.314708709716797,
|
299 |
+
"logps/rejected": -31.29607582092285,
|
300 |
+
"loss": 0.4191,
|
301 |
+
"rewards/accuracies": 0.612500011920929,
|
302 |
+
"rewards/chosen": 0.13939552009105682,
|
303 |
+
"rewards/margins": 0.12168798595666885,
|
304 |
+
"rewards/rejected": 0.017707547172904015,
|
305 |
"step": 190
|
306 |
},
|
307 |
{
|
308 |
"epoch": 0.52,
|
309 |
"learning_rate": 2.7718537898066833e-06,
|
310 |
+
"logits/chosen": -1.9107002019882202,
|
311 |
+
"logits/rejected": -1.915356993675232,
|
312 |
+
"logps/chosen": -31.39776611328125,
|
313 |
+
"logps/rejected": -32.7841682434082,
|
314 |
+
"loss": 0.3991,
|
315 |
"rewards/accuracies": 0.6875,
|
316 |
+
"rewards/chosen": 0.1668148785829544,
|
317 |
+
"rewards/margins": 0.16140125691890717,
|
318 |
+
"rewards/rejected": 0.005413623061031103,
|
319 |
"step": 200
|
320 |
},
|
321 |
{
|
322 |
"epoch": 0.52,
|
323 |
+
"eval_logits/chosen": -2.234880208969116,
|
324 |
+
"eval_logits/rejected": -2.2300188541412354,
|
325 |
+
"eval_logps/chosen": -34.03581619262695,
|
326 |
+
"eval_logps/rejected": -37.54121017456055,
|
327 |
+
"eval_loss": 0.4536176025867462,
|
328 |
+
"eval_rewards/accuracies": 0.5373754501342773,
|
329 |
+
"eval_rewards/chosen": -0.00101160176564008,
|
330 |
+
"eval_rewards/margins": 0.01865854486823082,
|
331 |
+
"eval_rewards/rejected": -0.019670147448778152,
|
332 |
+
"eval_runtime": 145.4825,
|
333 |
+
"eval_samples_per_second": 2.358,
|
334 |
+
"eval_steps_per_second": 0.296,
|
335 |
"step": 200
|
336 |
},
|
337 |
{
|
338 |
"epoch": 0.55,
|
339 |
"learning_rate": 2.5453962426402006e-06,
|
340 |
+
"logits/chosen": -2.022460460662842,
|
341 |
+
"logits/rejected": -2.03314471244812,
|
342 |
+
"logps/chosen": -31.764179229736328,
|
343 |
+
"logps/rejected": -33.924598693847656,
|
344 |
+
"loss": 0.2827,
|
345 |
+
"rewards/accuracies": 0.75,
|
346 |
+
"rewards/chosen": 0.1656998097896576,
|
347 |
+
"rewards/margins": 0.21521887183189392,
|
348 |
+
"rewards/rejected": -0.049519073218107224,
|
349 |
"step": 210
|
350 |
},
|
351 |
{
|
352 |
"epoch": 0.57,
|
353 |
"learning_rate": 2.3185646976551794e-06,
|
354 |
+
"logits/chosen": -1.9155023097991943,
|
355 |
+
"logits/rejected": -1.9303117990493774,
|
356 |
+
"logps/chosen": -29.97748374938965,
|
357 |
+
"logps/rejected": -31.563989639282227,
|
358 |
+
"loss": 0.3625,
|
359 |
+
"rewards/accuracies": 0.6499999761581421,
|
360 |
+
"rewards/chosen": 0.10724345594644547,
|
361 |
+
"rewards/margins": 0.11189230531454086,
|
362 |
+
"rewards/rejected": -0.004648865200579166,
|
363 |
"step": 220
|
364 |
},
|
365 |
{
|
366 |
"epoch": 0.6,
|
367 |
"learning_rate": 2.0932279108998323e-06,
|
368 |
+
"logits/chosen": -1.9727697372436523,
|
369 |
+
"logits/rejected": -1.9767513275146484,
|
370 |
+
"logps/chosen": -33.23460006713867,
|
371 |
+
"logps/rejected": -31.581985473632812,
|
372 |
+
"loss": 0.3234,
|
373 |
+
"rewards/accuracies": 0.7250000238418579,
|
374 |
+
"rewards/chosen": 0.1462947130203247,
|
375 |
+
"rewards/margins": 0.17244157195091248,
|
376 |
+
"rewards/rejected": -0.026146870106458664,
|
377 |
"step": 230
|
378 |
},
|
379 |
{
|
380 |
"epoch": 0.62,
|
381 |
"learning_rate": 1.8712423238279358e-06,
|
382 |
+
"logits/chosen": -1.971836805343628,
|
383 |
+
"logits/rejected": -1.9498882293701172,
|
384 |
+
"logps/chosen": -33.9459228515625,
|
385 |
+
"logps/rejected": -35.02843475341797,
|
386 |
+
"loss": 0.3771,
|
387 |
+
"rewards/accuracies": 0.699999988079071,
|
388 |
+
"rewards/chosen": 0.1662607342004776,
|
389 |
+
"rewards/margins": 0.2283717393875122,
|
390 |
+
"rewards/rejected": -0.0621110200881958,
|
391 |
"step": 240
|
392 |
},
|
393 |
{
|
394 |
"epoch": 0.65,
|
395 |
"learning_rate": 1.6544367689701824e-06,
|
396 |
+
"logits/chosen": -2.0139522552490234,
|
397 |
+
"logits/rejected": -2.010632038116455,
|
398 |
+
"logps/chosen": -32.76514434814453,
|
399 |
+
"logps/rejected": -36.231590270996094,
|
400 |
+
"loss": 0.3601,
|
401 |
+
"rewards/accuracies": 0.637499988079071,
|
402 |
+
"rewards/chosen": 0.1279398649930954,
|
403 |
+
"rewards/margins": 0.1332733929157257,
|
404 |
+
"rewards/rejected": -0.005333536770194769,
|
405 |
"step": 250
|
406 |
},
|
407 |
{
|
408 |
"epoch": 0.68,
|
409 |
"learning_rate": 1.4445974030621963e-06,
|
410 |
+
"logits/chosen": -1.8817164897918701,
|
411 |
+
"logits/rejected": -1.8792623281478882,
|
412 |
+
"logps/chosen": -34.03675842285156,
|
413 |
+
"logps/rejected": -35.52077102661133,
|
414 |
+
"loss": 0.3873,
|
415 |
+
"rewards/accuracies": 0.699999988079071,
|
416 |
+
"rewards/chosen": 0.12131254374980927,
|
417 |
+
"rewards/margins": 0.13423141837120056,
|
418 |
+
"rewards/rejected": -0.012918862514197826,
|
419 |
"step": 260
|
420 |
},
|
421 |
{
|
422 |
"epoch": 0.7,
|
423 |
"learning_rate": 1.243452991757889e-06,
|
424 |
+
"logits/chosen": -1.8678665161132812,
|
425 |
+
"logits/rejected": -1.8653860092163086,
|
426 |
+
"logps/chosen": -34.24772262573242,
|
427 |
+
"logps/rejected": -31.75213050842285,
|
428 |
+
"loss": 0.3738,
|
429 |
+
"rewards/accuracies": 0.6499999761581421,
|
430 |
+
"rewards/chosen": 0.11989933252334595,
|
431 |
+
"rewards/margins": 0.1045403853058815,
|
432 |
+
"rewards/rejected": 0.015358942560851574,
|
433 |
"step": 270
|
434 |
},
|
435 |
{
|
436 |
"epoch": 0.73,
|
437 |
"learning_rate": 1.0526606671603523e-06,
|
438 |
+
"logits/chosen": -1.9715077877044678,
|
439 |
+
"logits/rejected": -1.9609073400497437,
|
440 |
+
"logps/chosen": -35.06243133544922,
|
441 |
+
"logps/rejected": -31.837047576904297,
|
442 |
+
"loss": 0.2652,
|
443 |
+
"rewards/accuracies": 0.7749999761581421,
|
444 |
+
"rewards/chosen": 0.2353697121143341,
|
445 |
+
"rewards/margins": 0.2322143018245697,
|
446 |
+
"rewards/rejected": 0.003155359299853444,
|
447 |
"step": 280
|
448 |
},
|
449 |
{
|
450 |
"epoch": 0.75,
|
451 |
"learning_rate": 8.737922755071455e-07,
|
452 |
+
"logits/chosen": -2.0671467781066895,
|
453 |
+
"logits/rejected": -2.052150011062622,
|
454 |
+
"logps/chosen": -30.7277774810791,
|
455 |
+
"logps/rejected": -32.63950729370117,
|
456 |
+
"loss": 0.3926,
|
457 |
+
"rewards/accuracies": 0.625,
|
458 |
+
"rewards/chosen": 0.15707775950431824,
|
459 |
+
"rewards/margins": 0.14721640944480896,
|
460 |
+
"rewards/rejected": 0.009861335158348083,
|
461 |
"step": 290
|
462 |
},
|
463 |
{
|
464 |
"epoch": 0.78,
|
465 |
"learning_rate": 7.08321427484816e-07,
|
466 |
+
"logits/chosen": -1.9383538961410522,
|
467 |
+
"logits/rejected": -1.935782790184021,
|
468 |
+
"logps/chosen": -32.601158142089844,
|
469 |
+
"logps/rejected": -30.8317813873291,
|
470 |
+
"loss": 0.2816,
|
471 |
"rewards/accuracies": 0.75,
|
472 |
+
"rewards/chosen": 0.26075875759124756,
|
473 |
+
"rewards/margins": 0.2727365493774414,
|
474 |
+
"rewards/rejected": -0.011977789923548698,
|
475 |
"step": 300
|
476 |
},
|
477 |
{
|
478 |
"epoch": 0.78,
|
479 |
+
"eval_logits/chosen": -2.2351884841918945,
|
480 |
+
"eval_logits/rejected": -2.230333089828491,
|
481 |
+
"eval_logps/chosen": -34.01213836669922,
|
482 |
+
"eval_logps/rejected": -37.532142639160156,
|
483 |
+
"eval_loss": 0.4463596045970917,
|
484 |
+
"eval_rewards/accuracies": 0.5427741408348083,
|
485 |
+
"eval_rewards/chosen": 0.017930733039975166,
|
486 |
+
"eval_rewards/margins": 0.03035038523375988,
|
487 |
+
"eval_rewards/rejected": -0.012419654987752438,
|
488 |
+
"eval_runtime": 145.7115,
|
489 |
+
"eval_samples_per_second": 2.354,
|
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": -1.9204727411270142,
|
497 |
+
"logits/rejected": -1.9172074794769287,
|
498 |
+
"logps/chosen": -31.360204696655273,
|
499 |
+
"logps/rejected": -33.759212493896484,
|
500 |
+
"loss": 0.3011,
|
501 |
+
"rewards/accuracies": 0.75,
|
502 |
+
"rewards/chosen": 0.19451384246349335,
|
503 |
+
"rewards/margins": 0.20816302299499512,
|
504 |
+
"rewards/rejected": -0.01364919263869524,
|
505 |
"step": 310
|
506 |
},
|
507 |
{
|
508 |
"epoch": 0.83,
|
509 |
"learning_rate": 4.229036944380913e-07,
|
510 |
+
"logits/chosen": -1.9719558954238892,
|
511 |
+
"logits/rejected": -1.959651231765747,
|
512 |
+
"logps/chosen": -34.44158935546875,
|
513 |
+
"logps/rejected": -33.596771240234375,
|
514 |
+
"loss": 0.2835,
|
515 |
+
"rewards/accuracies": 0.675000011920929,
|
516 |
+
"rewards/chosen": 0.12358621507883072,
|
517 |
+
"rewards/margins": 0.1699295938014984,
|
518 |
+
"rewards/rejected": -0.0463433675467968,
|
519 |
"step": 320
|
520 |
},
|
521 |
{
|
522 |
"epoch": 0.86,
|
523 |
"learning_rate": 3.053082288996112e-07,
|
524 |
+
"logits/chosen": -2.0068559646606445,
|
525 |
+
"logits/rejected": -2.00539231300354,
|
526 |
+
"logps/chosen": -33.331153869628906,
|
527 |
+
"logps/rejected": -32.49522018432617,
|
528 |
+
"loss": 0.3707,
|
529 |
"rewards/accuracies": 0.6499999761581421,
|
530 |
+
"rewards/chosen": 0.12371524423360825,
|
531 |
+
"rewards/margins": 0.13622507452964783,
|
532 |
+
"rewards/rejected": -0.012509837746620178,
|
533 |
"step": 330
|
534 |
},
|
535 |
{
|
536 |
"epoch": 0.88,
|
537 |
"learning_rate": 2.0579377374915805e-07,
|
538 |
+
"logits/chosen": -2.0939385890960693,
|
539 |
+
"logits/rejected": -2.0781521797180176,
|
540 |
+
"logps/chosen": -33.88405227661133,
|
541 |
+
"logps/rejected": -33.090179443359375,
|
542 |
+
"loss": 0.36,
|
543 |
+
"rewards/accuracies": 0.699999988079071,
|
544 |
+
"rewards/chosen": 0.23627004027366638,
|
545 |
+
"rewards/margins": 0.21398480236530304,
|
546 |
+
"rewards/rejected": 0.02228522300720215,
|
547 |
"step": 340
|
548 |
},
|
549 |
{
|
550 |
"epoch": 0.91,
|
551 |
"learning_rate": 1.2518018074041684e-07,
|
552 |
+
"logits/chosen": -1.9664535522460938,
|
553 |
+
"logits/rejected": -1.9655389785766602,
|
554 |
+
"logps/chosen": -32.96236038208008,
|
555 |
+
"logps/rejected": -32.478477478027344,
|
556 |
+
"loss": 0.3,
|
557 |
+
"rewards/accuracies": 0.737500011920929,
|
558 |
+
"rewards/chosen": 0.2557886242866516,
|
559 |
+
"rewards/margins": 0.2664529085159302,
|
560 |
+
"rewards/rejected": -0.0106642572209239,
|
561 |
"step": 350
|
562 |
},
|
563 |
{
|
564 |
"epoch": 0.94,
|
565 |
"learning_rate": 6.41315865106129e-08,
|
566 |
+
"logits/chosen": -1.921805739402771,
|
567 |
+
"logits/rejected": -1.9321292638778687,
|
568 |
+
"logps/chosen": -32.022247314453125,
|
569 |
+
"logps/rejected": -35.31328582763672,
|
570 |
+
"loss": 0.3671,
|
571 |
+
"rewards/accuracies": 0.7124999761581421,
|
572 |
+
"rewards/chosen": 0.14595167338848114,
|
573 |
+
"rewards/margins": 0.16557307541370392,
|
574 |
+
"rewards/rejected": -0.01962139829993248,
|
575 |
"step": 360
|
576 |
},
|
577 |
{
|
578 |
"epoch": 0.96,
|
579 |
"learning_rate": 2.3150941078050325e-08,
|
580 |
+
"logits/chosen": -2.060645580291748,
|
581 |
+
"logits/rejected": -2.0541434288024902,
|
582 |
+
"logps/chosen": -33.4498405456543,
|
583 |
+
"logps/rejected": -29.211456298828125,
|
584 |
+
"loss": 0.3226,
|
585 |
+
"rewards/accuracies": 0.6875,
|
586 |
+
"rewards/chosen": 0.16169998049736023,
|
587 |
+
"rewards/margins": 0.1497238427400589,
|
588 |
+
"rewards/rejected": 0.011976108886301517,
|
589 |
"step": 370
|
590 |
},
|
591 |
{
|
592 |
"epoch": 0.99,
|
593 |
"learning_rate": 2.575864278703266e-09,
|
594 |
+
"logits/chosen": -1.920902967453003,
|
595 |
+
"logits/rejected": -1.9230811595916748,
|
596 |
+
"logps/chosen": -33.97188186645508,
|
597 |
+
"logps/rejected": -30.862768173217773,
|
598 |
+
"loss": 0.3117,
|
599 |
+
"rewards/accuracies": 0.7124999761581421,
|
600 |
+
"rewards/chosen": 0.21373608708381653,
|
601 |
+
"rewards/margins": 0.20584869384765625,
|
602 |
+
"rewards/rejected": 0.00788736529648304,
|
603 |
"step": 380
|
604 |
},
|
605 |
{
|
606 |
"epoch": 1.0,
|
607 |
"step": 385,
|
608 |
"total_flos": 0.0,
|
609 |
+
"train_loss": 0.3756082092012678,
|
610 |
+
"train_runtime": 3251.4223,
|
611 |
"train_samples_per_second": 0.947,
|
612 |
"train_steps_per_second": 0.118
|
613 |
}
|