hugodk-sch
commited on
Commit
•
ae8768e
1
Parent(s):
4a6d08d
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 +377 -377
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: 0.3845
|
21 |
+
- Rewards/chosen: 0.0182
|
22 |
+
- Rewards/rejected: 0.0177
|
23 |
+
- Rewards/accuracies: 0.4934
|
24 |
+
- Rewards/margins: 0.0005
|
25 |
+
- Logps/rejected: -37.4970
|
26 |
+
- Logps/chosen: -34.0143
|
27 |
+
- Logits/rejected: -2.2316
|
28 |
+
- Logits/chosen: -2.2364
|
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 |
+
| 0.3038 | 0.26 | 100 | 0.3512 | 0.0196 | 0.0046 | 0.5424 | 0.0150 | -37.5115 | -34.0128 | -2.2324 | -2.2372 |
|
64 |
+
| 0.3157 | 0.52 | 200 | 0.3716 | 0.0148 | -0.0016 | 0.5245 | 0.0164 | -37.5184 | -34.0181 | -2.2322 | -2.2371 |
|
65 |
+
| 0.2156 | 0.78 | 300 | 0.3845 | 0.0182 | 0.0177 | 0.4934 | 0.0005 | -37.4970 | -34.0143 | -2.2316 | -2.2364 |
|
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 sha256:1697658a6a16b5b6fa72dbed3433705c0a58db99550d85b605badd7b5de5b313
|
3 |
size 176183216
|
all_results.json
CHANGED
@@ -1,21 +1,8 @@
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
-
"
|
4 |
-
"
|
5 |
-
"eval_logps/chosen": -33.75144958496094,
|
6 |
-
"eval_logps/rejected": -37.262733459472656,
|
7 |
-
"eval_loss": 0.49370110034942627,
|
8 |
-
"eval_rewards/accuracies": 0.5274086594581604,
|
9 |
-
"eval_rewards/chosen": 0.25479215383529663,
|
10 |
-
"eval_rewards/margins": 0.02629559487104416,
|
11 |
-
"eval_rewards/rejected": 0.22849655151367188,
|
12 |
-
"eval_runtime": 145.6018,
|
13 |
-
"eval_samples": 343,
|
14 |
-
"eval_samples_per_second": 2.356,
|
15 |
-
"eval_steps_per_second": 0.295,
|
16 |
-
"train_loss": 0.447351616079157,
|
17 |
-
"train_runtime": 3253.4458,
|
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": 0.31703355428460356,
|
4 |
+
"train_runtime": 3251.5033,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
5 |
"train_samples": 3079,
|
6 |
+
"train_samples_per_second": 0.947,
|
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": 0.31703355428460356,
|
4 |
+
"train_runtime": 3251.5033,
|
5 |
"train_samples": 3079,
|
6 |
+
"train_samples_per_second": 0.947,
|
7 |
"train_steps_per_second": 0.118
|
8 |
}
|
trainer_state.json
CHANGED
@@ -15,7 +15,7 @@
|
|
15 |
"logits/rejected": -1.7377450466156006,
|
16 |
"logps/chosen": -29.553977966308594,
|
17 |
"logps/rejected": -42.813133239746094,
|
18 |
-
"loss": 0.
|
19 |
"rewards/accuracies": 0.0,
|
20 |
"rewards/chosen": 0.0,
|
21 |
"rewards/margins": 0.0,
|
@@ -25,590 +25,590 @@
|
|
25 |
{
|
26 |
"epoch": 0.03,
|
27 |
"learning_rate": 1.282051282051282e-06,
|
28 |
-
"logits/chosen": -1.
|
29 |
-
"logits/rejected": -1.
|
30 |
-
"logps/chosen": -36.
|
31 |
-
"logps/rejected": -33.
|
32 |
-
"loss": 0.
|
33 |
-
"rewards/accuracies": 0.
|
34 |
-
"rewards/chosen": 0.
|
35 |
-
"rewards/margins": 0.
|
36 |
-
"rewards/rejected": -0.
|
37 |
"step": 10
|
38 |
},
|
39 |
{
|
40 |
"epoch": 0.05,
|
41 |
"learning_rate": 2.564102564102564e-06,
|
42 |
-
"logits/chosen": -1.
|
43 |
-
"logits/rejected": -2.
|
44 |
-
"logps/chosen": -29.
|
45 |
-
"logps/rejected": -29.
|
46 |
-
"loss": 0.
|
47 |
-
"rewards/accuracies": 0.
|
48 |
-
"rewards/chosen": -0.
|
49 |
-
"rewards/margins": -0.
|
50 |
-
"rewards/rejected": 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": 0.
|
63 |
-
"rewards/margins": 0.
|
64 |
-
"rewards/rejected": -0.
|
65 |
"step": 30
|
66 |
},
|
67 |
{
|
68 |
"epoch": 0.1,
|
69 |
"learning_rate": 4.999896948438434e-06,
|
70 |
-
"logits/chosen": -2.
|
71 |
-
"logits/rejected": -2.
|
72 |
-
"logps/chosen": -32.
|
73 |
-
"logps/rejected": -32.
|
74 |
-
"loss": 0.
|
75 |
"rewards/accuracies": 0.5,
|
76 |
-
"rewards/chosen": 0.
|
77 |
-
"rewards/margins": 0.
|
78 |
-
"rewards/rejected": 0.
|
79 |
"step": 40
|
80 |
},
|
81 |
{
|
82 |
"epoch": 0.13,
|
83 |
"learning_rate": 4.987541037542187e-06,
|
84 |
-
"logits/chosen": -1.
|
85 |
-
"logits/rejected": -1.
|
86 |
-
"logps/chosen": -33.
|
87 |
-
"logps/rejected": -35.
|
88 |
-
"loss": 0.
|
89 |
-
"rewards/accuracies": 0.
|
90 |
-
"rewards/chosen": 0.
|
91 |
-
"rewards/margins": 0.
|
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": -
|
170 |
-
"eval_logps/rejected": -37.
|
171 |
-
"eval_loss": 0.
|
172 |
-
"eval_rewards/accuracies": 0.
|
173 |
-
"eval_rewards/chosen": 0.
|
174 |
-
"eval_rewards/margins": 0.
|
175 |
-
"eval_rewards/rejected": 0.
|
176 |
-
"eval_runtime": 145.
|
177 |
-
"eval_samples_per_second": 2.
|
178 |
"eval_steps_per_second": 0.295,
|
179 |
"step": 100
|
180 |
},
|
181 |
{
|
182 |
"epoch": 0.29,
|
183 |
"learning_rate": 4.498257201263691e-06,
|
184 |
-
"logits/chosen": -1.
|
185 |
-
"logits/rejected": -1.
|
186 |
-
"logps/chosen": -
|
187 |
-
"logps/rejected": -
|
188 |
-
"loss": 0.
|
189 |
-
"rewards/accuracies": 0.
|
190 |
-
"rewards/chosen": 0.
|
191 |
-
"rewards/margins": 0.
|
192 |
-
"rewards/rejected": 0.
|
193 |
"step": 110
|
194 |
},
|
195 |
{
|
196 |
"epoch": 0.31,
|
197 |
"learning_rate": 4.353806263777678e-06,
|
198 |
-
"logits/chosen": -2.
|
199 |
-
"logits/rejected": -
|
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": 0.6875,
|
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": 0.612500011920929,
|
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.
|
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.
|
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.
|
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.
|
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": -
|
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.
|
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": 0.3086,
|
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": -1.8665881156921387,
|
29 |
+
"logits/rejected": -1.8709055185317993,
|
30 |
+
"logps/chosen": -36.99662399291992,
|
31 |
+
"logps/rejected": -33.65571594238281,
|
32 |
+
"loss": 0.3052,
|
33 |
+
"rewards/accuracies": 0.5416666865348816,
|
34 |
+
"rewards/chosen": 0.008980684913694859,
|
35 |
+
"rewards/margins": 0.032059140503406525,
|
36 |
+
"rewards/rejected": -0.02307845838367939,
|
37 |
"step": 10
|
38 |
},
|
39 |
{
|
40 |
"epoch": 0.05,
|
41 |
"learning_rate": 2.564102564102564e-06,
|
42 |
+
"logits/chosen": -1.9977455139160156,
|
43 |
+
"logits/rejected": -2.000382423400879,
|
44 |
+
"logps/chosen": -29.642925262451172,
|
45 |
+
"logps/rejected": -29.056737899780273,
|
46 |
+
"loss": 0.3525,
|
47 |
+
"rewards/accuracies": 0.42500001192092896,
|
48 |
+
"rewards/chosen": -0.0006573178106918931,
|
49 |
+
"rewards/margins": -0.00669272243976593,
|
50 |
+
"rewards/rejected": 0.006035405211150646,
|
51 |
"step": 20
|
52 |
},
|
53 |
{
|
54 |
"epoch": 0.08,
|
55 |
"learning_rate": 3.846153846153847e-06,
|
56 |
+
"logits/chosen": -1.9204607009887695,
|
57 |
+
"logits/rejected": -1.9177772998809814,
|
58 |
+
"logps/chosen": -31.42336654663086,
|
59 |
+
"logps/rejected": -33.22785568237305,
|
60 |
+
"loss": 0.3603,
|
61 |
+
"rewards/accuracies": 0.48750001192092896,
|
62 |
+
"rewards/chosen": -0.006546213291585445,
|
63 |
+
"rewards/margins": 0.0019128695130348206,
|
64 |
+
"rewards/rejected": -0.00845908559858799,
|
65 |
"step": 30
|
66 |
},
|
67 |
{
|
68 |
"epoch": 0.1,
|
69 |
"learning_rate": 4.999896948438434e-06,
|
70 |
+
"logits/chosen": -2.0172362327575684,
|
71 |
+
"logits/rejected": -2.008507251739502,
|
72 |
+
"logps/chosen": -32.56964874267578,
|
73 |
+
"logps/rejected": -32.50572967529297,
|
74 |
+
"loss": 0.3558,
|
75 |
"rewards/accuracies": 0.5,
|
76 |
+
"rewards/chosen": 0.006439635064452887,
|
77 |
+
"rewards/margins": 0.00019515231542754918,
|
78 |
+
"rewards/rejected": 0.006244482938200235,
|
79 |
"step": 40
|
80 |
},
|
81 |
{
|
82 |
"epoch": 0.13,
|
83 |
"learning_rate": 4.987541037542187e-06,
|
84 |
+
"logits/chosen": -1.8618619441986084,
|
85 |
+
"logits/rejected": -1.8510783910751343,
|
86 |
+
"logps/chosen": -33.56026077270508,
|
87 |
+
"logps/rejected": -35.45254898071289,
|
88 |
+
"loss": 0.3691,
|
89 |
+
"rewards/accuracies": 0.4000000059604645,
|
90 |
+
"rewards/chosen": -0.0025013976264744997,
|
91 |
+
"rewards/margins": -0.002315213903784752,
|
92 |
+
"rewards/rejected": -0.00018618404283188283,
|
93 |
"step": 50
|
94 |
},
|
95 |
{
|
96 |
"epoch": 0.16,
|
97 |
"learning_rate": 4.954691471941119e-06,
|
98 |
+
"logits/chosen": -1.9393202066421509,
|
99 |
+
"logits/rejected": -1.9412600994110107,
|
100 |
+
"logps/chosen": -32.57838439941406,
|
101 |
+
"logps/rejected": -33.215576171875,
|
102 |
+
"loss": 0.3063,
|
103 |
+
"rewards/accuracies": 0.612500011920929,
|
104 |
+
"rewards/chosen": 0.01962057128548622,
|
105 |
+
"rewards/margins": 0.05940054729580879,
|
106 |
+
"rewards/rejected": -0.03977997973561287,
|
107 |
"step": 60
|
108 |
},
|
109 |
{
|
110 |
"epoch": 0.18,
|
111 |
"learning_rate": 4.901618883413549e-06,
|
112 |
+
"logits/chosen": -2.0718436241149902,
|
113 |
+
"logits/rejected": -2.0768017768859863,
|
114 |
+
"logps/chosen": -33.97806167602539,
|
115 |
+
"logps/rejected": -36.63082504272461,
|
116 |
+
"loss": 0.4257,
|
117 |
+
"rewards/accuracies": 0.5,
|
118 |
+
"rewards/chosen": 0.00938049890100956,
|
119 |
+
"rewards/margins": 0.06513925641775131,
|
120 |
+
"rewards/rejected": -0.0557587556540966,
|
121 |
"step": 70
|
122 |
},
|
123 |
{
|
124 |
"epoch": 0.21,
|
125 |
"learning_rate": 4.828760511501322e-06,
|
126 |
+
"logits/chosen": -1.9351739883422852,
|
127 |
+
"logits/rejected": -1.9383188486099243,
|
128 |
+
"logps/chosen": -34.33073043823242,
|
129 |
+
"logps/rejected": -34.61904525756836,
|
130 |
+
"loss": 0.2902,
|
131 |
+
"rewards/accuracies": 0.625,
|
132 |
+
"rewards/chosen": 0.08736880123615265,
|
133 |
+
"rewards/margins": 0.11134655773639679,
|
134 |
+
"rewards/rejected": -0.023977745324373245,
|
135 |
"step": 80
|
136 |
},
|
137 |
{
|
138 |
"epoch": 0.23,
|
139 |
"learning_rate": 4.7367166013034295e-06,
|
140 |
+
"logits/chosen": -1.944392204284668,
|
141 |
+
"logits/rejected": -1.9489190578460693,
|
142 |
+
"logps/chosen": -32.419586181640625,
|
143 |
+
"logps/rejected": -32.3698844909668,
|
144 |
+
"loss": 0.3611,
|
145 |
+
"rewards/accuracies": 0.574999988079071,
|
146 |
+
"rewards/chosen": 0.043051257729530334,
|
147 |
+
"rewards/margins": 0.052930813282728195,
|
148 |
+
"rewards/rejected": -0.009879561141133308,
|
149 |
"step": 90
|
150 |
},
|
151 |
{
|
152 |
"epoch": 0.26,
|
153 |
"learning_rate": 4.626245458345211e-06,
|
154 |
+
"logits/chosen": -2.0419769287109375,
|
155 |
+
"logits/rejected": -2.0399627685546875,
|
156 |
+
"logps/chosen": -32.174407958984375,
|
157 |
+
"logps/rejected": -31.26608657836914,
|
158 |
+
"loss": 0.3038,
|
159 |
+
"rewards/accuracies": 0.5874999761581421,
|
160 |
+
"rewards/chosen": 0.059785228222608566,
|
161 |
+
"rewards/margins": 0.05967242643237114,
|
162 |
+
"rewards/rejected": 0.00011279433965682983,
|
163 |
"step": 100
|
164 |
},
|
165 |
{
|
166 |
"epoch": 0.26,
|
167 |
+
"eval_logits/chosen": -2.2372143268585205,
|
168 |
+
"eval_logits/rejected": -2.2323503494262695,
|
169 |
+
"eval_logps/chosen": -34.01276779174805,
|
170 |
+
"eval_logps/rejected": -37.51152420043945,
|
171 |
+
"eval_loss": 0.3511974811553955,
|
172 |
+
"eval_rewards/accuracies": 0.5423588156700134,
|
173 |
+
"eval_rewards/chosen": 0.01960929110646248,
|
174 |
+
"eval_rewards/margins": 0.01502405758947134,
|
175 |
+
"eval_rewards/rejected": 0.004585230257362127,
|
176 |
+
"eval_runtime": 145.9032,
|
177 |
+
"eval_samples_per_second": 2.351,
|
178 |
"eval_steps_per_second": 0.295,
|
179 |
"step": 100
|
180 |
},
|
181 |
{
|
182 |
"epoch": 0.29,
|
183 |
"learning_rate": 4.498257201263691e-06,
|
184 |
+
"logits/chosen": -1.998891830444336,
|
185 |
+
"logits/rejected": -1.996492624282837,
|
186 |
+
"logps/chosen": -33.14598083496094,
|
187 |
+
"logps/rejected": -34.020729064941406,
|
188 |
+
"loss": 0.4689,
|
189 |
+
"rewards/accuracies": 0.6000000238418579,
|
190 |
+
"rewards/chosen": 0.08805312216281891,
|
191 |
+
"rewards/margins": 0.08027410507202148,
|
192 |
+
"rewards/rejected": 0.007779018487781286,
|
193 |
"step": 110
|
194 |
},
|
195 |
{
|
196 |
"epoch": 0.31,
|
197 |
"learning_rate": 4.353806263777678e-06,
|
198 |
+
"logits/chosen": -2.010932445526123,
|
199 |
+
"logits/rejected": -2.0025696754455566,
|
200 |
+
"logps/chosen": -32.37172317504883,
|
201 |
+
"logps/rejected": -32.118797302246094,
|
202 |
+
"loss": 0.4465,
|
203 |
+
"rewards/accuracies": 0.574999988079071,
|
204 |
+
"rewards/chosen": 0.06586603820323944,
|
205 |
+
"rewards/margins": 0.02742874063551426,
|
206 |
+
"rewards/rejected": 0.03843729570508003,
|
207 |
"step": 120
|
208 |
},
|
209 |
{
|
210 |
"epoch": 0.34,
|
211 |
"learning_rate": 4.1940827077152755e-06,
|
212 |
+
"logits/chosen": -2.0387539863586426,
|
213 |
+
"logits/rejected": -2.030724287033081,
|
214 |
+
"logps/chosen": -30.41655921936035,
|
215 |
+
"logps/rejected": -32.060333251953125,
|
216 |
+
"loss": 0.3844,
|
217 |
+
"rewards/accuracies": 0.5375000238418579,
|
218 |
+
"rewards/chosen": 0.05324209854006767,
|
219 |
+
"rewards/margins": 0.05983499437570572,
|
220 |
+
"rewards/rejected": -0.00659290561452508,
|
221 |
"step": 130
|
222 |
},
|
223 |
{
|
224 |
"epoch": 0.36,
|
225 |
"learning_rate": 4.0204024186666215e-06,
|
226 |
+
"logits/chosen": -1.967858076095581,
|
227 |
+
"logits/rejected": -1.9781148433685303,
|
228 |
+
"logps/chosen": -31.223413467407227,
|
229 |
+
"logps/rejected": -32.55517578125,
|
230 |
+
"loss": 0.3904,
|
231 |
"rewards/accuracies": 0.6875,
|
232 |
+
"rewards/chosen": 0.1766217201948166,
|
233 |
+
"rewards/margins": 0.1948881596326828,
|
234 |
+
"rewards/rejected": -0.01826643943786621,
|
235 |
"step": 140
|
236 |
},
|
237 |
{
|
238 |
"epoch": 0.39,
|
239 |
"learning_rate": 3.834196265035119e-06,
|
240 |
+
"logits/chosen": -1.880910873413086,
|
241 |
+
"logits/rejected": -1.8820507526397705,
|
242 |
+
"logps/chosen": -34.01464080810547,
|
243 |
+
"logps/rejected": -34.783546447753906,
|
244 |
+
"loss": 0.3871,
|
245 |
"rewards/accuracies": 0.612500011920929,
|
246 |
+
"rewards/chosen": 0.160243421792984,
|
247 |
+
"rewards/margins": 0.1839032918214798,
|
248 |
+
"rewards/rejected": -0.023659853264689445,
|
249 |
"step": 150
|
250 |
},
|
251 |
{
|
252 |
"epoch": 0.42,
|
253 |
"learning_rate": 3.636998309800573e-06,
|
254 |
+
"logits/chosen": -1.933895468711853,
|
255 |
+
"logits/rejected": -1.9304730892181396,
|
256 |
+
"logps/chosen": -36.02853775024414,
|
257 |
+
"logps/rejected": -32.699058532714844,
|
258 |
+
"loss": 0.2729,
|
259 |
+
"rewards/accuracies": 0.637499988079071,
|
260 |
+
"rewards/chosen": 0.1288261115550995,
|
261 |
+
"rewards/margins": 0.10750452429056168,
|
262 |
+
"rewards/rejected": 0.021321602165699005,
|
263 |
"step": 160
|
264 |
},
|
265 |
{
|
266 |
"epoch": 0.44,
|
267 |
"learning_rate": 3.4304331721118078e-06,
|
268 |
+
"logits/chosen": -2.0341715812683105,
|
269 |
+
"logits/rejected": -2.0267820358276367,
|
270 |
+
"logps/chosen": -33.55347442626953,
|
271 |
+
"logps/rejected": -31.3526554107666,
|
272 |
+
"loss": 0.294,
|
273 |
+
"rewards/accuracies": 0.6625000238418579,
|
274 |
+
"rewards/chosen": 0.21994857490062714,
|
275 |
+
"rewards/margins": 0.22067300975322723,
|
276 |
+
"rewards/rejected": -0.0007244400912895799,
|
277 |
"step": 170
|
278 |
},
|
279 |
{
|
280 |
"epoch": 0.47,
|
281 |
"learning_rate": 3.2162026428305436e-06,
|
282 |
+
"logits/chosen": -2.0400891304016113,
|
283 |
+
"logits/rejected": -2.045360565185547,
|
284 |
+
"logps/chosen": -32.370338439941406,
|
285 |
+
"logps/rejected": -32.4719123840332,
|
286 |
+
"loss": 0.2829,
|
287 |
+
"rewards/accuracies": 0.6625000238418579,
|
288 |
+
"rewards/chosen": 0.15700635313987732,
|
289 |
+
"rewards/margins": 0.145101398229599,
|
290 |
+
"rewards/rejected": 0.011904975399374962,
|
291 |
"step": 180
|
292 |
},
|
293 |
{
|
294 |
"epoch": 0.49,
|
295 |
"learning_rate": 2.996071664294641e-06,
|
296 |
+
"logits/chosen": -2.041393280029297,
|
297 |
+
"logits/rejected": -2.038623809814453,
|
298 |
+
"logps/chosen": -31.328174591064453,
|
299 |
+
"logps/rejected": -31.316492080688477,
|
300 |
+
"loss": 0.3044,
|
301 |
+
"rewards/accuracies": 0.699999988079071,
|
302 |
+
"rewards/chosen": 0.14470075070858002,
|
303 |
+
"rewards/margins": 0.1431477963924408,
|
304 |
+
"rewards/rejected": 0.0015529401134699583,
|
305 |
"step": 190
|
306 |
},
|
307 |
{
|
308 |
"epoch": 0.52,
|
309 |
"learning_rate": 2.7718537898066833e-06,
|
310 |
+
"logits/chosen": -1.9122215509414673,
|
311 |
+
"logits/rejected": -1.9168663024902344,
|
312 |
+
"logps/chosen": -31.424020767211914,
|
313 |
+
"logps/rejected": -32.784080505371094,
|
314 |
+
"loss": 0.3157,
|
315 |
+
"rewards/accuracies": 0.699999988079071,
|
316 |
+
"rewards/chosen": 0.16403506696224213,
|
317 |
+
"rewards/margins": 0.1578713059425354,
|
318 |
+
"rewards/rejected": 0.0061637843027710915,
|
319 |
"step": 200
|
320 |
},
|
321 |
{
|
322 |
"epoch": 0.52,
|
323 |
+
"eval_logits/chosen": -2.237051010131836,
|
324 |
+
"eval_logits/rejected": -2.2321863174438477,
|
325 |
+
"eval_logps/chosen": -34.018070220947266,
|
326 |
+
"eval_logps/rejected": -37.51838684082031,
|
327 |
+
"eval_loss": 0.3716273605823517,
|
328 |
+
"eval_rewards/accuracies": 0.5245016813278198,
|
329 |
+
"eval_rewards/chosen": 0.014834923669695854,
|
330 |
+
"eval_rewards/margins": 0.016425320878624916,
|
331 |
+
"eval_rewards/rejected": -0.0015903981402516365,
|
332 |
+
"eval_runtime": 145.5151,
|
333 |
+
"eval_samples_per_second": 2.357,
|
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": -2.023789882659912,
|
341 |
+
"logits/rejected": -2.034484386444092,
|
342 |
+
"logps/chosen": -31.767370223999023,
|
343 |
+
"logps/rejected": -33.890621185302734,
|
344 |
+
"loss": 0.275,
|
345 |
+
"rewards/accuracies": 0.675000011920929,
|
346 |
+
"rewards/chosen": 0.18354059755802155,
|
347 |
+
"rewards/margins": 0.20867136120796204,
|
348 |
+
"rewards/rejected": -0.025130782276391983,
|
349 |
"step": 210
|
350 |
},
|
351 |
{
|
352 |
"epoch": 0.57,
|
353 |
"learning_rate": 2.3185646976551794e-06,
|
354 |
+
"logits/chosen": -1.916685700416565,
|
355 |
+
"logits/rejected": -1.9314892292022705,
|
356 |
+
"logps/chosen": -29.956628799438477,
|
357 |
+
"logps/rejected": -31.564035415649414,
|
358 |
+
"loss": 0.2958,
|
359 |
+
"rewards/accuracies": 0.6625000238418579,
|
360 |
+
"rewards/chosen": 0.13942097127437592,
|
361 |
+
"rewards/margins": 0.14469322562217712,
|
362 |
+
"rewards/rejected": -0.005272268317639828,
|
363 |
"step": 220
|
364 |
},
|
365 |
{
|
366 |
"epoch": 0.6,
|
367 |
"learning_rate": 2.0932279108998323e-06,
|
368 |
+
"logits/chosen": -1.9737945795059204,
|
369 |
+
"logits/rejected": -1.9777710437774658,
|
370 |
+
"logps/chosen": -33.19129180908203,
|
371 |
+
"logps/rejected": -31.5566463470459,
|
372 |
+
"loss": 0.287,
|
373 |
+
"rewards/accuracies": 0.6625000238418579,
|
374 |
+
"rewards/chosen": 0.20355579257011414,
|
375 |
+
"rewards/margins": 0.21016716957092285,
|
376 |
+
"rewards/rejected": -0.0066113718785345554,
|
377 |
"step": 230
|
378 |
},
|
379 |
{
|
380 |
"epoch": 0.62,
|
381 |
"learning_rate": 1.8712423238279358e-06,
|
382 |
+
"logits/chosen": -1.9743419885635376,
|
383 |
+
"logits/rejected": -1.9523779153823853,
|
384 |
+
"logps/chosen": -33.9401969909668,
|
385 |
+
"logps/rejected": -35.008758544921875,
|
386 |
+
"loss": 0.3185,
|
387 |
+
"rewards/accuracies": 0.6625000238418579,
|
388 |
+
"rewards/chosen": 0.1921955645084381,
|
389 |
+
"rewards/margins": 0.24436470866203308,
|
390 |
+
"rewards/rejected": -0.052169155329465866,
|
391 |
"step": 240
|
392 |
},
|
393 |
{
|
394 |
"epoch": 0.65,
|
395 |
"learning_rate": 1.6544367689701824e-06,
|
396 |
+
"logits/chosen": -2.0161709785461426,
|
397 |
+
"logits/rejected": -2.0128414630889893,
|
398 |
+
"logps/chosen": -32.74829864501953,
|
399 |
+
"logps/rejected": -36.236392974853516,
|
400 |
+
"loss": 0.2847,
|
401 |
+
"rewards/accuracies": 0.7250000238418579,
|
402 |
+
"rewards/chosen": 0.15909257531166077,
|
403 |
+
"rewards/margins": 0.1694144755601883,
|
404 |
+
"rewards/rejected": -0.010321905836462975,
|
405 |
"step": 250
|
406 |
},
|
407 |
{
|
408 |
"epoch": 0.68,
|
409 |
"learning_rate": 1.4445974030621963e-06,
|
410 |
+
"logits/chosen": -1.8847742080688477,
|
411 |
+
"logits/rejected": -1.8823268413543701,
|
412 |
+
"logps/chosen": -34.01182174682617,
|
413 |
+
"logps/rejected": -35.481346130371094,
|
414 |
+
"loss": 0.3073,
|
415 |
+
"rewards/accuracies": 0.612500011920929,
|
416 |
+
"rewards/chosen": 0.1589193046092987,
|
417 |
+
"rewards/margins": 0.13797220587730408,
|
418 |
+
"rewards/rejected": 0.020947108045220375,
|
419 |
"step": 260
|
420 |
},
|
421 |
{
|
422 |
"epoch": 0.7,
|
423 |
"learning_rate": 1.243452991757889e-06,
|
424 |
+
"logits/chosen": -1.8693536520004272,
|
425 |
+
"logits/rejected": -1.866838812828064,
|
426 |
+
"logps/chosen": -34.20549392700195,
|
427 |
+
"logps/rejected": -31.726673126220703,
|
428 |
+
"loss": 0.2733,
|
429 |
+
"rewards/accuracies": 0.6875,
|
430 |
+
"rewards/chosen": 0.17289286851882935,
|
431 |
+
"rewards/margins": 0.1327010840177536,
|
432 |
+
"rewards/rejected": 0.040191780775785446,
|
433 |
"step": 270
|
434 |
},
|
435 |
{
|
436 |
"epoch": 0.73,
|
437 |
"learning_rate": 1.0526606671603523e-06,
|
438 |
+
"logits/chosen": -1.9730831384658813,
|
439 |
+
"logits/rejected": -1.962480902671814,
|
440 |
+
"logps/chosen": -35.06049728393555,
|
441 |
+
"logps/rejected": -31.792781829833984,
|
442 |
+
"loss": 0.2523,
|
443 |
+
"rewards/accuracies": 0.75,
|
444 |
+
"rewards/chosen": 0.2665289342403412,
|
445 |
+
"rewards/margins": 0.22314274311065674,
|
446 |
+
"rewards/rejected": 0.04338619112968445,
|
447 |
"step": 280
|
448 |
},
|
449 |
{
|
450 |
"epoch": 0.75,
|
451 |
"learning_rate": 8.737922755071455e-07,
|
452 |
+
"logits/chosen": -2.0680813789367676,
|
453 |
+
"logits/rejected": -2.053079128265381,
|
454 |
+
"logps/chosen": -30.738479614257812,
|
455 |
+
"logps/rejected": -32.61243438720703,
|
456 |
+
"loss": 0.3194,
|
457 |
+
"rewards/accuracies": 0.6000000238418579,
|
458 |
+
"rewards/chosen": 0.16707859933376312,
|
459 |
+
"rewards/margins": 0.13161785900592804,
|
460 |
+
"rewards/rejected": 0.035460732877254486,
|
461 |
"step": 290
|
462 |
},
|
463 |
{
|
464 |
"epoch": 0.78,
|
465 |
"learning_rate": 7.08321427484816e-07,
|
466 |
+
"logits/chosen": -1.9398882389068604,
|
467 |
+
"logits/rejected": -1.9373395442962646,
|
468 |
+
"logps/chosen": -32.6181640625,
|
469 |
+
"logps/rejected": -30.843700408935547,
|
470 |
+
"loss": 0.2156,
|
471 |
+
"rewards/accuracies": 0.75,
|
472 |
+
"rewards/chosen": 0.2780519425868988,
|
473 |
+
"rewards/margins": 0.3022567331790924,
|
474 |
+
"rewards/rejected": -0.024204757064580917,
|
475 |
"step": 300
|
476 |
},
|
477 |
{
|
478 |
"epoch": 0.78,
|
479 |
+
"eval_logits/chosen": -2.2364187240600586,
|
480 |
+
"eval_logits/rejected": -2.2315518856048584,
|
481 |
+
"eval_logps/chosen": -34.01428985595703,
|
482 |
+
"eval_logps/rejected": -37.496952056884766,
|
483 |
+
"eval_loss": 0.38450533151626587,
|
484 |
+
"eval_rewards/accuracies": 0.49335551261901855,
|
485 |
+
"eval_rewards/chosen": 0.018236981704831123,
|
486 |
+
"eval_rewards/margins": 0.0005384809919632971,
|
487 |
+
"eval_rewards/rejected": 0.01769850216805935,
|
488 |
+
"eval_runtime": 145.7485,
|
489 |
+
"eval_samples_per_second": 2.353,
|
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": -1.922579050064087,
|
497 |
+
"logits/rejected": -1.9193273782730103,
|
498 |
+
"logps/chosen": -31.345911026000977,
|
499 |
+
"logps/rejected": -33.72126007080078,
|
500 |
+
"loss": 0.2917,
|
501 |
+
"rewards/accuracies": 0.699999988079071,
|
502 |
+
"rewards/chosen": 0.23169513046741486,
|
503 |
+
"rewards/margins": 0.21289470791816711,
|
504 |
+
"rewards/rejected": 0.018800420686602592,
|
505 |
"step": 310
|
506 |
},
|
507 |
{
|
508 |
"epoch": 0.83,
|
509 |
"learning_rate": 4.229036944380913e-07,
|
510 |
+
"logits/chosen": -1.9754743576049805,
|
511 |
+
"logits/rejected": -1.9631853103637695,
|
512 |
+
"logps/chosen": -34.408077239990234,
|
513 |
+
"logps/rejected": -33.58232879638672,
|
514 |
+
"loss": 0.2394,
|
515 |
+
"rewards/accuracies": 0.675000011920929,
|
516 |
+
"rewards/chosen": 0.1691979616880417,
|
517 |
+
"rewards/margins": 0.20833876729011536,
|
518 |
+
"rewards/rejected": -0.03914082050323486,
|
519 |
"step": 320
|
520 |
},
|
521 |
{
|
522 |
"epoch": 0.86,
|
523 |
"learning_rate": 3.053082288996112e-07,
|
524 |
+
"logits/chosen": -2.0105607509613037,
|
525 |
+
"logits/rejected": -2.009115219116211,
|
526 |
+
"logps/chosen": -33.31591033935547,
|
527 |
+
"logps/rejected": -32.47368621826172,
|
528 |
+
"loss": 0.2769,
|
529 |
+
"rewards/accuracies": 0.6499999761581421,
|
530 |
+
"rewards/chosen": 0.15289874374866486,
|
531 |
+
"rewards/margins": 0.14759239554405212,
|
532 |
+
"rewards/rejected": 0.005306343547999859,
|
533 |
"step": 330
|
534 |
},
|
535 |
{
|
536 |
"epoch": 0.88,
|
537 |
"learning_rate": 2.0579377374915805e-07,
|
538 |
+
"logits/chosen": -2.096872091293335,
|
539 |
+
"logits/rejected": -2.0811073780059814,
|
540 |
+
"logps/chosen": -33.87510681152344,
|
541 |
+
"logps/rejected": -33.06427764892578,
|
542 |
+
"loss": 0.2792,
|
543 |
+
"rewards/accuracies": 0.737500011920929,
|
544 |
+
"rewards/chosen": 0.27385497093200684,
|
545 |
+
"rewards/margins": 0.2254684418439865,
|
546 |
+
"rewards/rejected": 0.04838654398918152,
|
547 |
"step": 340
|
548 |
},
|
549 |
{
|
550 |
"epoch": 0.91,
|
551 |
"learning_rate": 1.2518018074041684e-07,
|
552 |
+
"logits/chosen": -1.969496488571167,
|
553 |
+
"logits/rejected": -1.9685735702514648,
|
554 |
+
"logps/chosen": -32.98945999145508,
|
555 |
+
"logps/rejected": -32.4643440246582,
|
556 |
+
"loss": 0.2958,
|
557 |
+
"rewards/accuracies": 0.737500011920929,
|
558 |
+
"rewards/chosen": 0.26337358355522156,
|
559 |
+
"rewards/margins": 0.26264840364456177,
|
560 |
+
"rewards/rejected": 0.0007252089562825859,
|
561 |
"step": 350
|
562 |
},
|
563 |
{
|
564 |
"epoch": 0.94,
|
565 |
"learning_rate": 6.41315865106129e-08,
|
566 |
+
"logits/chosen": -1.9255645275115967,
|
567 |
+
"logits/rejected": -1.9359004497528076,
|
568 |
+
"logps/chosen": -32.013362884521484,
|
569 |
+
"logps/rejected": -35.26326370239258,
|
570 |
+
"loss": 0.3425,
|
571 |
+
"rewards/accuracies": 0.637499988079071,
|
572 |
+
"rewards/chosen": 0.1721937507390976,
|
573 |
+
"rewards/margins": 0.14924712479114532,
|
574 |
+
"rewards/rejected": 0.022946633398532867,
|
575 |
"step": 360
|
576 |
},
|
577 |
{
|
578 |
"epoch": 0.96,
|
579 |
"learning_rate": 2.3150941078050325e-08,
|
580 |
+
"logits/chosen": -2.0643956661224365,
|
581 |
+
"logits/rejected": -2.057886838912964,
|
582 |
+
"logps/chosen": -33.48772430419922,
|
583 |
+
"logps/rejected": -29.191638946533203,
|
584 |
+
"loss": 0.298,
|
585 |
+
"rewards/accuracies": 0.625,
|
586 |
+
"rewards/chosen": 0.14782151579856873,
|
587 |
+
"rewards/margins": 0.11651048809289932,
|
588 |
+
"rewards/rejected": 0.03131101652979851,
|
589 |
"step": 370
|
590 |
},
|
591 |
{
|
592 |
"epoch": 0.99,
|
593 |
"learning_rate": 2.575864278703266e-09,
|
594 |
+
"logits/chosen": -1.9235725402832031,
|
595 |
+
"logits/rejected": -1.9257465600967407,
|
596 |
+
"logps/chosen": -33.965919494628906,
|
597 |
+
"logps/rejected": -30.839218139648438,
|
598 |
+
"loss": 0.2616,
|
599 |
+
"rewards/accuracies": 0.699999988079071,
|
600 |
+
"rewards/chosen": 0.24581687152385712,
|
601 |
+
"rewards/margins": 0.2157471626996994,
|
602 |
+
"rewards/rejected": 0.03006969951093197,
|
603 |
"step": 380
|
604 |
},
|
605 |
{
|
606 |
"epoch": 1.0,
|
607 |
"step": 385,
|
608 |
"total_flos": 0.0,
|
609 |
+
"train_loss": 0.31703355428460356,
|
610 |
+
"train_runtime": 3251.5033,
|
611 |
+
"train_samples_per_second": 0.947,
|
612 |
"train_steps_per_second": 0.118
|
613 |
}
|
614 |
],
|