hugodk-sch
commited on
Commit
•
d75ba52
1
Parent(s):
1050cfb
Model save
Browse files- README.md +13 -16
- adapter_config.json +4 -4
- adapter_model.safetensors +1 -1
- all_results.json +3 -16
- train_results.json +3 -3
- trainer_state.json +374 -374
- training_args.bin +1 -1
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:
|
25 |
-
- Rewards/rejected: 0.
|
26 |
-
- Rewards/accuracies: 0.
|
27 |
-
- Rewards/margins: -0.
|
28 |
-
- Logps/rejected: -37.
|
29 |
-
- Logps/chosen: -34.
|
30 |
-
- Logits/rejected: -2.
|
31 |
-
- Logits/chosen: -2.
|
32 |
|
33 |
## Model description
|
34 |
|
@@ -63,9 +60,9 @@ The following hyperparameters were used during training:
|
|
63 |
|
64 |
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
65 |
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
66 |
-
| 0.
|
67 |
-
| 0.
|
68 |
-
| 0.
|
69 |
|
70 |
|
71 |
### Framework versions
|
|
|
1 |
---
|
2 |
library_name: peft
|
3 |
tags:
|
|
|
4 |
- trl
|
5 |
- dpo
|
6 |
- generated_from_trainer
|
7 |
base_model: NbAiLab/nb-gpt-j-6B-v2
|
|
|
|
|
8 |
model-index:
|
9 |
- name: aftonposten-6b-align-scan
|
10 |
results: []
|
|
|
15 |
|
16 |
# aftonposten-6b-align-scan
|
17 |
|
18 |
+
This model is a fine-tuned version of [NbAiLab/nb-gpt-j-6B-v2](https://huggingface.co/NbAiLab/nb-gpt-j-6B-v2) on an unknown dataset.
|
19 |
It achieves the following results on the evaluation set:
|
20 |
+
- Loss: 0.6982
|
21 |
+
- Rewards/chosen: 0.0156
|
22 |
+
- Rewards/rejected: 0.0171
|
23 |
+
- Rewards/accuracies: 0.4983
|
24 |
+
- Rewards/margins: -0.0015
|
25 |
+
- Logps/rejected: -37.4976
|
26 |
+
- Logps/chosen: -34.0172
|
27 |
+
- Logits/rejected: -2.2388
|
28 |
+
- Logits/chosen: -2.2437
|
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 |
+
| 0.6951 | 0.26 | 100 | 0.6966 | 0.0084 | 0.0081 | 0.4954 | 0.0003 | -37.5076 | -34.0252 | -2.2389 | -2.2438 |
|
64 |
+
| 0.6891 | 0.52 | 200 | 0.6947 | 0.0092 | 0.0046 | 0.4934 | 0.0046 | -37.5115 | -34.0243 | -2.2384 | -2.2432 |
|
65 |
+
| 0.693 | 0.78 | 300 | 0.6982 | 0.0156 | 0.0171 | 0.4983 | -0.0015 | -37.4976 | -34.0172 | -2.2388 | -2.2437 |
|
66 |
|
67 |
|
68 |
### Framework versions
|
adapter_config.json
CHANGED
@@ -19,13 +19,13 @@
|
|
19 |
"rank_pattern": {},
|
20 |
"revision": null,
|
21 |
"target_modules": [
|
|
|
|
|
22 |
"gate_proj",
|
|
|
23 |
"down_proj",
|
24 |
"v_proj",
|
25 |
-
"
|
26 |
-
"q_proj",
|
27 |
-
"k_proj",
|
28 |
-
"up_proj"
|
29 |
],
|
30 |
"task_type": "CAUSAL_LM",
|
31 |
"use_rslora": false
|
|
|
19 |
"rank_pattern": {},
|
20 |
"revision": null,
|
21 |
"target_modules": [
|
22 |
+
"up_proj",
|
23 |
+
"q_proj",
|
24 |
"gate_proj",
|
25 |
+
"o_proj",
|
26 |
"down_proj",
|
27 |
"v_proj",
|
28 |
+
"k_proj"
|
|
|
|
|
|
|
29 |
],
|
30 |
"task_type": "CAUSAL_LM",
|
31 |
"use_rslora": false
|
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:b30a3614e90564741bd0b06f4b5d211ac9ab0d5d4651e82dec2b571e49a407d7
|
3 |
size 176183216
|
all_results.json
CHANGED
@@ -1,21 +1,8 @@
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
-
"
|
4 |
-
"
|
5 |
-
"eval_logps/chosen": -34.03584671020508,
|
6 |
-
"eval_logps/rejected": -37.496219635009766,
|
7 |
-
"eval_loss": 0.7054064273834229,
|
8 |
-
"eval_rewards/accuracies": 0.472591370344162,
|
9 |
-
"eval_rewards/chosen": -0.001035186112858355,
|
10 |
-
"eval_rewards/margins": -0.017356980592012405,
|
11 |
-
"eval_rewards/rejected": 0.01632179319858551,
|
12 |
-
"eval_runtime": 145.8229,
|
13 |
-
"eval_samples": 343,
|
14 |
-
"eval_samples_per_second": 2.352,
|
15 |
-
"eval_steps_per_second": 0.295,
|
16 |
-
"train_loss": 0.6928803004227675,
|
17 |
-
"train_runtime": 3255.7403,
|
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": 0.6916276120520257,
|
4 |
+
"train_runtime": 3257.6062,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
5 |
"train_samples": 3079,
|
6 |
+
"train_samples_per_second": 0.945,
|
7 |
"train_steps_per_second": 0.118
|
8 |
}
|
train_results.json
CHANGED
@@ -1,8 +1,8 @@
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
-
"train_loss": 0.
|
4 |
-
"train_runtime":
|
5 |
"train_samples": 3079,
|
6 |
-
"train_samples_per_second": 0.
|
7 |
"train_steps_per_second": 0.118
|
8 |
}
|
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
+
"train_loss": 0.6916276120520257,
|
4 |
+
"train_runtime": 3257.6062,
|
5 |
"train_samples": 3079,
|
6 |
+
"train_samples_per_second": 0.945,
|
7 |
"train_steps_per_second": 0.118
|
8 |
}
|
trainer_state.json
CHANGED
@@ -25,590 +25,590 @@
|
|
25 |
{
|
26 |
"epoch": 0.03,
|
27 |
"learning_rate": 1.2820512820512818e-07,
|
28 |
-
"logits/chosen": -1.
|
29 |
-
"logits/rejected": -1.
|
30 |
-
"logps/chosen": -
|
31 |
-
"logps/rejected": -33.
|
32 |
-
"loss": 0.
|
33 |
-
"rewards/accuracies": 0.
|
34 |
-
"rewards/chosen":
|
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": -2.
|
44 |
-
"logps/chosen": -29.
|
45 |
-
"logps/rejected": -29.
|
46 |
-
"loss": 0.
|
47 |
-
"rewards/accuracies": 0.
|
48 |
-
"rewards/chosen":
|
49 |
-
"rewards/margins":
|
50 |
-
"rewards/rejected": 0.
|
51 |
"step": 20
|
52 |
},
|
53 |
{
|
54 |
"epoch": 0.08,
|
55 |
"learning_rate": 3.8461538461538463e-07,
|
56 |
-
"logits/chosen": -1.
|
57 |
-
"logits/rejected": -1.
|
58 |
-
"logps/chosen": -31.
|
59 |
-
"logps/rejected": -33.
|
60 |
-
"loss": 0.
|
61 |
-
"rewards/accuracies": 0.
|
62 |
-
"rewards/chosen": 0.
|
63 |
-
"rewards/margins": 0.
|
64 |
-
"rewards/rejected":
|
65 |
"step": 30
|
66 |
},
|
67 |
{
|
68 |
"epoch": 0.1,
|
69 |
"learning_rate": 4.999896948438433e-07,
|
70 |
-
"logits/chosen": -2.
|
71 |
-
"logits/rejected": -2.
|
72 |
-
"logps/chosen": -32.
|
73 |
-
"logps/rejected": -32.
|
74 |
-
"loss": 0.
|
75 |
"rewards/accuracies": 0.48750001192092896,
|
76 |
-
"rewards/chosen": -0.
|
77 |
-
"rewards/margins": -0.
|
78 |
-
"rewards/rejected": 0.
|
79 |
"step": 40
|
80 |
},
|
81 |
{
|
82 |
"epoch": 0.13,
|
83 |
"learning_rate": 4.987541037542186e-07,
|
84 |
-
"logits/chosen": -1.
|
85 |
-
"logits/rejected": -1.
|
86 |
-
"logps/chosen": -33.
|
87 |
-
"logps/rejected": -35.
|
88 |
-
"loss": 0.
|
89 |
-
"rewards/accuracies": 0.
|
90 |
-
"rewards/chosen":
|
91 |
-
"rewards/margins": -0.
|
92 |
-
"rewards/rejected": 0.
|
93 |
"step": 50
|
94 |
},
|
95 |
{
|
96 |
"epoch": 0.16,
|
97 |
"learning_rate": 4.954691471941118e-07,
|
98 |
-
"logits/chosen": -1.
|
99 |
-
"logits/rejected": -1.
|
100 |
-
"logps/chosen": -32.
|
101 |
-
"logps/rejected": -33.
|
102 |
-
"loss": 0.
|
103 |
-
"rewards/accuracies": 0.
|
104 |
-
"rewards/chosen": 0.
|
105 |
-
"rewards/margins": 0.
|
106 |
-
"rewards/rejected": -0.
|
107 |
"step": 60
|
108 |
},
|
109 |
{
|
110 |
"epoch": 0.18,
|
111 |
"learning_rate": 4.901618883413548e-07,
|
112 |
-
"logits/chosen": -2.
|
113 |
-
"logits/rejected": -2.
|
114 |
-
"logps/chosen": -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": 0.
|
131 |
-
"rewards/accuracies": 0.
|
132 |
-
"rewards/chosen": 0.
|
133 |
-
"rewards/margins":
|
134 |
-
"rewards/rejected": 0.
|
135 |
"step": 80
|
136 |
},
|
137 |
{
|
138 |
"epoch": 0.23,
|
139 |
"learning_rate": 4.736716601303429e-07,
|
140 |
-
"logits/chosen": -1.
|
141 |
-
"logits/rejected": -1.
|
142 |
-
"logps/chosen": -32.
|
143 |
-
"logps/rejected": -32.
|
144 |
-
"loss": 0.
|
145 |
"rewards/accuracies": 0.5249999761581421,
|
146 |
-
"rewards/chosen": 0.
|
147 |
-
"rewards/margins": 0.
|
148 |
-
"rewards/rejected":
|
149 |
"step": 90
|
150 |
},
|
151 |
{
|
152 |
"epoch": 0.26,
|
153 |
"learning_rate": 4.62624545834521e-07,
|
154 |
-
"logits/chosen": -2.
|
155 |
-
"logits/rejected": -2.
|
156 |
-
"logps/chosen": -32.
|
157 |
-
"logps/rejected": -31.
|
158 |
-
"loss": 0.
|
159 |
"rewards/accuracies": 0.5,
|
160 |
-
"rewards/chosen": 0.
|
161 |
-
"rewards/margins": 0.
|
162 |
-
"rewards/rejected":
|
163 |
"step": 100
|
164 |
},
|
165 |
{
|
166 |
"epoch": 0.26,
|
167 |
-
"eval_logits/chosen": -2.
|
168 |
-
"eval_logits/rejected": -2.
|
169 |
-
"eval_logps/chosen": -34.
|
170 |
-
"eval_logps/rejected": -37.
|
171 |
-
"eval_loss": 0.
|
172 |
-
"eval_rewards/accuracies": 0.
|
173 |
-
"eval_rewards/chosen": 0.
|
174 |
-
"eval_rewards/margins": 0.
|
175 |
-
"eval_rewards/rejected": 0.
|
176 |
-
"eval_runtime": 146.
|
177 |
-
"eval_samples_per_second": 2.
|
178 |
"eval_steps_per_second": 0.294,
|
179 |
"step": 100
|
180 |
},
|
181 |
{
|
182 |
"epoch": 0.29,
|
183 |
"learning_rate": 4.4982572012636904e-07,
|
184 |
-
"logits/chosen": -2.
|
185 |
-
"logits/rejected": -2.
|
186 |
-
"logps/chosen": -33.
|
187 |
-
"logps/rejected": -34.
|
188 |
-
"loss": 0.
|
189 |
-
"rewards/accuracies": 0.
|
190 |
-
"rewards/chosen": 0.
|
191 |
-
"rewards/margins": 0.
|
192 |
-
"rewards/rejected":
|
193 |
"step": 110
|
194 |
},
|
195 |
{
|
196 |
"epoch": 0.31,
|
197 |
"learning_rate": 4.353806263777677e-07,
|
198 |
-
"logits/chosen": -2.
|
199 |
-
"logits/rejected": -2.
|
200 |
-
"logps/chosen": -32.
|
201 |
-
"logps/rejected": -32.
|
202 |
-
"loss": 0.
|
203 |
-
"rewards/accuracies": 0.
|
204 |
-
"rewards/chosen": -0.
|
205 |
-
"rewards/margins": -0.
|
206 |
-
"rewards/rejected": -0.
|
207 |
"step": 120
|
208 |
},
|
209 |
{
|
210 |
"epoch": 0.34,
|
211 |
"learning_rate": 4.194082707715275e-07,
|
212 |
-
"logits/chosen": -2.
|
213 |
-
"logits/rejected": -2.
|
214 |
-
"logps/chosen": -30.
|
215 |
-
"logps/rejected": -32.
|
216 |
-
"loss": 0.
|
217 |
-
"rewards/accuracies": 0.
|
218 |
-
"rewards/chosen": -0.
|
219 |
-
"rewards/margins": -0.
|
220 |
-
"rewards/rejected": 0.
|
221 |
"step": 130
|
222 |
},
|
223 |
{
|
224 |
"epoch": 0.36,
|
225 |
"learning_rate": 4.020402418666621e-07,
|
226 |
-
"logits/chosen": -1.
|
227 |
-
"logits/rejected": -1.
|
228 |
-
"logps/chosen": -31.
|
229 |
-
"logps/rejected": -32.
|
230 |
-
"loss": 0.
|
231 |
-
"rewards/accuracies": 0.
|
232 |
-
"rewards/chosen": 0.
|
233 |
-
"rewards/margins": 0.
|
234 |
-
"rewards/rejected":
|
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": 0.4749999940395355,
|
246 |
-
"rewards/chosen": 0.
|
247 |
-
"rewards/margins": 0.
|
248 |
-
"rewards/rejected": -0.
|
249 |
"step": 150
|
250 |
},
|
251 |
{
|
252 |
"epoch": 0.42,
|
253 |
"learning_rate": 3.636998309800572e-07,
|
254 |
-
"logits/chosen": -1.
|
255 |
-
"logits/rejected": -1.
|
256 |
-
"logps/chosen": -36.
|
257 |
-
"logps/rejected": -32.
|
258 |
-
"loss": 0.
|
259 |
-
"rewards/accuracies": 0.
|
260 |
-
"rewards/chosen": 0.
|
261 |
-
"rewards/margins": 0.
|
262 |
-
"rewards/rejected": 0.
|
263 |
"step": 160
|
264 |
},
|
265 |
{
|
266 |
"epoch": 0.44,
|
267 |
"learning_rate": 3.430433172111807e-07,
|
268 |
-
"logits/chosen": -2.
|
269 |
-
"logits/rejected": -2.
|
270 |
-
"logps/chosen": -33.
|
271 |
-
"logps/rejected": -31.
|
272 |
-
"loss": 0.
|
273 |
-
"rewards/accuracies": 0.
|
274 |
-
"rewards/chosen": 0.
|
275 |
-
"rewards/margins": 0.
|
276 |
-
"rewards/rejected": 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": 0.
|
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":
|
319 |
"step": 200
|
320 |
},
|
321 |
{
|
322 |
"epoch": 0.52,
|
323 |
-
"eval_logits/chosen": -2.
|
324 |
-
"eval_logits/rejected": -2.
|
325 |
-
"eval_logps/chosen": -34.
|
326 |
-
"eval_logps/rejected": -37.
|
327 |
-
"eval_loss": 0.
|
328 |
-
"eval_rewards/accuracies": 0.
|
329 |
-
"eval_rewards/chosen": 0.
|
330 |
-
"eval_rewards/margins": 0.
|
331 |
-
"eval_rewards/rejected": 0.
|
332 |
-
"eval_runtime":
|
333 |
-
"eval_samples_per_second": 2.
|
334 |
-
"eval_steps_per_second": 0.
|
335 |
"step": 200
|
336 |
},
|
337 |
{
|
338 |
"epoch": 0.55,
|
339 |
"learning_rate": 2.5453962426402e-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":
|
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":
|
391 |
"step": 240
|
392 |
},
|
393 |
{
|
394 |
"epoch": 0.65,
|
395 |
"learning_rate": 1.654436768970182e-07,
|
396 |
-
"logits/chosen": -2.
|
397 |
-
"logits/rejected": -2.
|
398 |
-
"logps/chosen": -32.
|
399 |
-
"logps/rejected": -36.
|
400 |
-
"loss": 0.
|
401 |
-
"rewards/accuracies": 0.
|
402 |
-
"rewards/chosen": 0.
|
403 |
-
"rewards/margins": 0.
|
404 |
-
"rewards/rejected":
|
405 |
"step": 250
|
406 |
},
|
407 |
{
|
408 |
"epoch": 0.68,
|
409 |
"learning_rate": 1.444597403062196e-07,
|
410 |
-
"logits/chosen": -1.
|
411 |
-
"logits/rejected": -1.
|
412 |
-
"logps/chosen": -34.
|
413 |
-
"logps/rejected": -35.
|
414 |
-
"loss": 0.
|
415 |
-
"rewards/accuracies": 0.
|
416 |
-
"rewards/chosen": 0.
|
417 |
-
"rewards/margins": 0.
|
418 |
-
"rewards/rejected": 0.
|
419 |
"step": 260
|
420 |
},
|
421 |
{
|
422 |
"epoch": 0.7,
|
423 |
"learning_rate": 1.2434529917578887e-07,
|
424 |
-
"logits/chosen": -1.
|
425 |
-
"logits/rejected": -1.
|
426 |
-
"logps/chosen": -34.
|
427 |
-
"logps/rejected": -31.
|
428 |
-
"loss": 0.
|
429 |
-
"rewards/accuracies": 0.
|
430 |
-
"rewards/chosen":
|
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":
|
459 |
-
"rewards/margins":
|
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":
|
475 |
"step": 300
|
476 |
},
|
477 |
{
|
478 |
"epoch": 0.78,
|
479 |
-
"eval_logits/chosen": -2.
|
480 |
-
"eval_logits/rejected": -2.
|
481 |
-
"eval_logps/chosen": -34.
|
482 |
-
"eval_logps/rejected": -37.
|
483 |
-
"eval_loss": 0.
|
484 |
"eval_rewards/accuracies": 0.49833887815475464,
|
485 |
-
"eval_rewards/chosen": 0.
|
486 |
-
"eval_rewards/margins": -0.
|
487 |
-
"eval_rewards/rejected": 0.
|
488 |
-
"eval_runtime":
|
489 |
-
"eval_samples_per_second": 2.
|
490 |
-
"eval_steps_per_second": 0.
|
491 |
"step": 300
|
492 |
},
|
493 |
{
|
494 |
"epoch": 0.81,
|
495 |
"learning_rate": 5.576113578589034e-08,
|
496 |
-
"logits/chosen": -1.
|
497 |
-
"logits/rejected": -1.
|
498 |
-
"logps/chosen": -31.
|
499 |
-
"logps/rejected": -33.
|
500 |
-
"loss": 0.
|
501 |
"rewards/accuracies": 0.5375000238418579,
|
502 |
-
"rewards/chosen": 0.
|
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": 0.
|
529 |
-
"rewards/accuracies": 0.
|
530 |
-
"rewards/chosen": 0.
|
531 |
-
"rewards/margins": 0.
|
532 |
-
"rewards/rejected": 0.
|
533 |
"step": 330
|
534 |
},
|
535 |
{
|
536 |
"epoch": 0.88,
|
537 |
"learning_rate": 2.05793773749158e-08,
|
538 |
-
"logits/chosen": -2.
|
539 |
-
"logits/rejected": -2.
|
540 |
-
"logps/chosen": -34.
|
541 |
-
"logps/rejected": -33.
|
542 |
-
"loss": 0.
|
543 |
-
"rewards/accuracies": 0.
|
544 |
-
"rewards/chosen": 0.
|
545 |
-
"rewards/margins": -0.
|
546 |
-
"rewards/rejected": 0.
|
547 |
"step": 340
|
548 |
},
|
549 |
{
|
550 |
"epoch": 0.91,
|
551 |
"learning_rate": 1.251801807404168e-08,
|
552 |
-
"logits/chosen": -1.
|
553 |
-
"logits/rejected": -1.
|
554 |
-
"logps/chosen": -33.
|
555 |
-
"logps/rejected": -32.
|
556 |
-
"loss": 0.
|
557 |
-
"rewards/accuracies": 0.
|
558 |
-
"rewards/chosen": 0.
|
559 |
-
"rewards/margins": 0.
|
560 |
-
"rewards/rejected": 0.
|
561 |
"step": 350
|
562 |
},
|
563 |
{
|
564 |
"epoch": 0.94,
|
565 |
"learning_rate": 6.41315865106129e-09,
|
566 |
-
"logits/chosen": -1.
|
567 |
-
"logits/rejected": -1.
|
568 |
-
"logps/chosen": -32.
|
569 |
-
"logps/rejected": -35.
|
570 |
-
"loss": 0.
|
571 |
-
"rewards/accuracies": 0.
|
572 |
-
"rewards/chosen": 0.
|
573 |
-
"rewards/margins": 0.
|
574 |
-
"rewards/rejected":
|
575 |
"step": 360
|
576 |
},
|
577 |
{
|
578 |
"epoch": 0.96,
|
579 |
"learning_rate": 2.3150941078050324e-09,
|
580 |
-
"logits/chosen": -2.
|
581 |
-
"logits/rejected": -2.
|
582 |
-
"logps/chosen": -33.
|
583 |
-
"logps/rejected": -29.
|
584 |
-
"loss": 0.
|
585 |
-
"rewards/accuracies": 0.
|
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":
|
611 |
-
"train_samples_per_second": 0.
|
612 |
"train_steps_per_second": 0.118
|
613 |
}
|
614 |
],
|
|
|
25 |
{
|
26 |
"epoch": 0.03,
|
27 |
"learning_rate": 1.2820512820512818e-07,
|
28 |
+
"logits/chosen": -1.866059422492981,
|
29 |
+
"logits/rejected": -1.870389699935913,
|
30 |
+
"logps/chosen": -36.98250198364258,
|
31 |
+
"logps/rejected": -33.64270782470703,
|
32 |
+
"loss": 0.6804,
|
33 |
+
"rewards/accuracies": 0.4027777910232544,
|
34 |
+
"rewards/chosen": 0.021689780056476593,
|
35 |
+
"rewards/margins": 0.033059459179639816,
|
36 |
+
"rewards/rejected": -0.011369682848453522,
|
37 |
"step": 10
|
38 |
},
|
39 |
{
|
40 |
"epoch": 0.05,
|
41 |
"learning_rate": 2.5641025641025636e-07,
|
42 |
+
"logits/chosen": -1.9973828792572021,
|
43 |
+
"logits/rejected": -2.0000317096710205,
|
44 |
+
"logps/chosen": -29.631702423095703,
|
45 |
+
"logps/rejected": -29.05877113342285,
|
46 |
+
"loss": 0.6937,
|
47 |
+
"rewards/accuracies": 0.512499988079071,
|
48 |
+
"rewards/chosen": 0.009445475414395332,
|
49 |
+
"rewards/margins": 0.005240145605057478,
|
50 |
+
"rewards/rejected": 0.004205327946692705,
|
51 |
"step": 20
|
52 |
},
|
53 |
{
|
54 |
"epoch": 0.08,
|
55 |
"learning_rate": 3.8461538461538463e-07,
|
56 |
+
"logits/chosen": -1.9198739528656006,
|
57 |
+
"logits/rejected": -1.9171788692474365,
|
58 |
+
"logps/chosen": -31.4038028717041,
|
59 |
+
"logps/rejected": -33.20899963378906,
|
60 |
+
"loss": 0.696,
|
61 |
+
"rewards/accuracies": 0.5,
|
62 |
+
"rewards/chosen": 0.01106296293437481,
|
63 |
+
"rewards/margins": 0.0025471807457506657,
|
64 |
+
"rewards/rejected": 0.008515783585608006,
|
65 |
"step": 30
|
66 |
},
|
67 |
{
|
68 |
"epoch": 0.1,
|
69 |
"learning_rate": 4.999896948438433e-07,
|
70 |
+
"logits/chosen": -2.0166726112365723,
|
71 |
+
"logits/rejected": -2.0079195499420166,
|
72 |
+
"logps/chosen": -32.57840347290039,
|
73 |
+
"logps/rejected": -32.493370056152344,
|
74 |
+
"loss": 0.7066,
|
75 |
"rewards/accuracies": 0.48750001192092896,
|
76 |
+
"rewards/chosen": -0.00143451988697052,
|
77 |
+
"rewards/margins": -0.018809262663125992,
|
78 |
+
"rewards/rejected": 0.017374742776155472,
|
79 |
"step": 40
|
80 |
},
|
81 |
{
|
82 |
"epoch": 0.13,
|
83 |
"learning_rate": 4.987541037542186e-07,
|
84 |
+
"logits/chosen": -1.8648544549942017,
|
85 |
+
"logits/rejected": -1.8540741205215454,
|
86 |
+
"logps/chosen": -33.54566955566406,
|
87 |
+
"logps/rejected": -35.43162155151367,
|
88 |
+
"loss": 0.7004,
|
89 |
+
"rewards/accuracies": 0.4749999940395355,
|
90 |
+
"rewards/chosen": 0.01062955055385828,
|
91 |
+
"rewards/margins": -0.00802132673561573,
|
92 |
+
"rewards/rejected": 0.018650878220796585,
|
93 |
"step": 50
|
94 |
},
|
95 |
{
|
96 |
"epoch": 0.16,
|
97 |
"learning_rate": 4.954691471941118e-07,
|
98 |
+
"logits/chosen": -1.9450620412826538,
|
99 |
+
"logits/rejected": -1.947021245956421,
|
100 |
+
"logps/chosen": -32.5836296081543,
|
101 |
+
"logps/rejected": -33.20051193237305,
|
102 |
+
"loss": 0.6761,
|
103 |
+
"rewards/accuracies": 0.5375000238418579,
|
104 |
+
"rewards/chosen": 0.014901289716362953,
|
105 |
+
"rewards/margins": 0.04112350940704346,
|
106 |
+
"rewards/rejected": -0.026222219690680504,
|
107 |
"step": 60
|
108 |
},
|
109 |
{
|
110 |
"epoch": 0.18,
|
111 |
"learning_rate": 4.901618883413548e-07,
|
112 |
+
"logits/chosen": -2.079589366912842,
|
113 |
+
"logits/rejected": -2.0845787525177,
|
114 |
+
"logps/chosen": -33.9754524230957,
|
115 |
+
"logps/rejected": -36.5793342590332,
|
116 |
+
"loss": 0.6876,
|
117 |
+
"rewards/accuracies": 0.512499988079071,
|
118 |
+
"rewards/chosen": 0.01173111516982317,
|
119 |
+
"rewards/margins": 0.021146392449736595,
|
120 |
+
"rewards/rejected": -0.00941527634859085,
|
121 |
"step": 70
|
122 |
},
|
123 |
{
|
124 |
"epoch": 0.21,
|
125 |
"learning_rate": 4.828760511501322e-07,
|
126 |
+
"logits/chosen": -1.9424660205841064,
|
127 |
+
"logits/rejected": -1.945634126663208,
|
128 |
+
"logps/chosen": -34.394561767578125,
|
129 |
+
"logps/rejected": -34.57395553588867,
|
130 |
+
"loss": 0.6921,
|
131 |
+
"rewards/accuracies": 0.5625,
|
132 |
+
"rewards/chosen": 0.029921507462859154,
|
133 |
+
"rewards/margins": 0.013314949348568916,
|
134 |
+
"rewards/rejected": 0.016606558114290237,
|
135 |
"step": 80
|
136 |
},
|
137 |
{
|
138 |
"epoch": 0.23,
|
139 |
"learning_rate": 4.736716601303429e-07,
|
140 |
+
"logits/chosen": -1.950564980506897,
|
141 |
+
"logits/rejected": -1.955082893371582,
|
142 |
+
"logps/chosen": -32.44951248168945,
|
143 |
+
"logps/rejected": -32.353668212890625,
|
144 |
+
"loss": 0.6909,
|
145 |
"rewards/accuracies": 0.5249999761581421,
|
146 |
+
"rewards/chosen": 0.0161209125071764,
|
147 |
+
"rewards/margins": 0.011405264027416706,
|
148 |
+
"rewards/rejected": 0.004715651273727417,
|
149 |
"step": 90
|
150 |
},
|
151 |
{
|
152 |
"epoch": 0.26,
|
153 |
"learning_rate": 4.62624545834521e-07,
|
154 |
+
"logits/chosen": -2.0489234924316406,
|
155 |
+
"logits/rejected": -2.046917200088501,
|
156 |
+
"logps/chosen": -32.236671447753906,
|
157 |
+
"logps/rejected": -31.263763427734375,
|
158 |
+
"loss": 0.6951,
|
159 |
"rewards/accuracies": 0.5,
|
160 |
+
"rewards/chosen": 0.0037471160758286715,
|
161 |
+
"rewards/margins": 0.0015437646070495248,
|
162 |
+
"rewards/rejected": 0.002203352050855756,
|
163 |
"step": 100
|
164 |
},
|
165 |
{
|
166 |
"epoch": 0.26,
|
167 |
+
"eval_logits/chosen": -2.243756055831909,
|
168 |
+
"eval_logits/rejected": -2.238872766494751,
|
169 |
+
"eval_logps/chosen": -34.025177001953125,
|
170 |
+
"eval_logps/rejected": -37.50759506225586,
|
171 |
+
"eval_loss": 0.696557343006134,
|
172 |
+
"eval_rewards/accuracies": 0.49543190002441406,
|
173 |
+
"eval_rewards/chosen": 0.008434689603745937,
|
174 |
+
"eval_rewards/margins": 0.0003138432221021503,
|
175 |
+
"eval_rewards/rejected": 0.008120844140648842,
|
176 |
+
"eval_runtime": 146.3373,
|
177 |
+
"eval_samples_per_second": 2.344,
|
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": -2.0052237510681152,
|
185 |
+
"logits/rejected": -2.002803325653076,
|
186 |
+
"logps/chosen": -33.24525833129883,
|
187 |
+
"logps/rejected": -34.01074981689453,
|
188 |
+
"loss": 0.7064,
|
189 |
+
"rewards/accuracies": 0.48750001192092896,
|
190 |
+
"rewards/chosen": -0.0013012334238737822,
|
191 |
+
"rewards/margins": -0.018059223890304565,
|
192 |
+
"rewards/rejected": 0.016757991164922714,
|
193 |
"step": 110
|
194 |
},
|
195 |
{
|
196 |
"epoch": 0.31,
|
197 |
"learning_rate": 4.353806263777677e-07,
|
198 |
+
"logits/chosen": -2.016326665878296,
|
199 |
+
"logits/rejected": -2.007951498031616,
|
200 |
+
"logps/chosen": -32.478302001953125,
|
201 |
+
"logps/rejected": -32.186073303222656,
|
202 |
+
"loss": 0.7012,
|
203 |
+
"rewards/accuracies": 0.48750001192092896,
|
204 |
+
"rewards/chosen": -0.030056962743401527,
|
205 |
+
"rewards/margins": -0.007947373203933239,
|
206 |
+
"rewards/rejected": -0.022109590470790863,
|
207 |
"step": 120
|
208 |
},
|
209 |
{
|
210 |
"epoch": 0.34,
|
211 |
"learning_rate": 4.194082707715275e-07,
|
212 |
+
"logits/chosen": -2.045950412750244,
|
213 |
+
"logits/rejected": -2.037904739379883,
|
214 |
+
"logps/chosen": -30.491313934326172,
|
215 |
+
"logps/rejected": -32.0540771484375,
|
216 |
+
"loss": 0.7025,
|
217 |
+
"rewards/accuracies": 0.4375,
|
218 |
+
"rewards/chosen": -0.01403898186981678,
|
219 |
+
"rewards/margins": -0.013075167313218117,
|
220 |
+
"rewards/rejected": -0.0009638145565986633,
|
221 |
"step": 130
|
222 |
},
|
223 |
{
|
224 |
"epoch": 0.36,
|
225 |
"learning_rate": 4.020402418666621e-07,
|
226 |
+
"logits/chosen": -1.9765291213989258,
|
227 |
+
"logits/rejected": -1.986802101135254,
|
228 |
+
"logps/chosen": -31.39646339416504,
|
229 |
+
"logps/rejected": -32.522178649902344,
|
230 |
+
"loss": 0.6923,
|
231 |
+
"rewards/accuracies": 0.512499988079071,
|
232 |
+
"rewards/chosen": 0.02087983302772045,
|
233 |
+
"rewards/margins": 0.00944516807794571,
|
234 |
+
"rewards/rejected": 0.011434664018452168,
|
235 |
"step": 140
|
236 |
},
|
237 |
{
|
238 |
"epoch": 0.39,
|
239 |
"learning_rate": 3.8341962650351185e-07,
|
240 |
+
"logits/chosen": -1.8905407190322876,
|
241 |
+
"logits/rejected": -1.8916336297988892,
|
242 |
+
"logps/chosen": -34.17607116699219,
|
243 |
+
"logps/rejected": -34.76817321777344,
|
244 |
+
"loss": 0.6851,
|
245 |
"rewards/accuracies": 0.4749999940395355,
|
246 |
+
"rewards/chosen": 0.01495879702270031,
|
247 |
+
"rewards/margins": 0.024784717708826065,
|
248 |
+
"rewards/rejected": -0.00982591975480318,
|
249 |
"step": 150
|
250 |
},
|
251 |
{
|
252 |
"epoch": 0.42,
|
253 |
"learning_rate": 3.636998309800572e-07,
|
254 |
+
"logits/chosen": -1.942565679550171,
|
255 |
+
"logits/rejected": -1.9390919208526611,
|
256 |
+
"logps/chosen": -36.1614875793457,
|
257 |
+
"logps/rejected": -32.72848129272461,
|
258 |
+
"loss": 0.6907,
|
259 |
+
"rewards/accuracies": 0.4625000059604645,
|
260 |
+
"rewards/chosen": 0.00917207170277834,
|
261 |
+
"rewards/margins": 0.01433003693819046,
|
262 |
+
"rewards/rejected": -0.0051579661667346954,
|
263 |
"step": 160
|
264 |
},
|
265 |
{
|
266 |
"epoch": 0.44,
|
267 |
"learning_rate": 3.430433172111807e-07,
|
268 |
+
"logits/chosen": -2.0418710708618164,
|
269 |
+
"logits/rejected": -2.034475564956665,
|
270 |
+
"logps/chosen": -33.76544952392578,
|
271 |
+
"logps/rejected": -31.358556747436523,
|
272 |
+
"loss": 0.6791,
|
273 |
+
"rewards/accuracies": 0.574999988079071,
|
274 |
+
"rewards/chosen": 0.02917439118027687,
|
275 |
+
"rewards/margins": 0.03521214798092842,
|
276 |
+
"rewards/rejected": -0.0060377540066838264,
|
277 |
"step": 170
|
278 |
},
|
279 |
{
|
280 |
"epoch": 0.47,
|
281 |
"learning_rate": 3.216202642830543e-07,
|
282 |
+
"logits/chosen": -2.0476231575012207,
|
283 |
+
"logits/rejected": -2.0529019832611084,
|
284 |
+
"logps/chosen": -32.533023834228516,
|
285 |
+
"logps/rejected": -32.506805419921875,
|
286 |
+
"loss": 0.6814,
|
287 |
+
"rewards/accuracies": 0.5625,
|
288 |
+
"rewards/chosen": 0.010587882250547409,
|
289 |
+
"rewards/margins": 0.03008626028895378,
|
290 |
+
"rewards/rejected": -0.019498378038406372,
|
291 |
"step": 180
|
292 |
},
|
293 |
{
|
294 |
"epoch": 0.49,
|
295 |
"learning_rate": 2.9960716642946403e-07,
|
296 |
+
"logits/chosen": -2.048576593399048,
|
297 |
+
"logits/rejected": -2.0457961559295654,
|
298 |
+
"logps/chosen": -31.470510482788086,
|
299 |
+
"logps/rejected": -31.322284698486328,
|
300 |
+
"loss": 0.687,
|
301 |
+
"rewards/accuracies": 0.5625,
|
302 |
+
"rewards/chosen": 0.016599375754594803,
|
303 |
+
"rewards/margins": 0.020261235535144806,
|
304 |
+
"rewards/rejected": -0.0036618602462112904,
|
305 |
"step": 190
|
306 |
},
|
307 |
{
|
308 |
"epoch": 0.52,
|
309 |
"learning_rate": 2.771853789806683e-07,
|
310 |
+
"logits/chosen": -1.9184268712997437,
|
311 |
+
"logits/rejected": -1.9230947494506836,
|
312 |
+
"logps/chosen": -31.57404136657715,
|
313 |
+
"logps/rejected": -32.775550842285156,
|
314 |
+
"loss": 0.6891,
|
315 |
+
"rewards/accuracies": 0.48750001192092896,
|
316 |
+
"rewards/chosen": 0.02901501953601837,
|
317 |
+
"rewards/margins": 0.015173261985182762,
|
318 |
+
"rewards/rejected": 0.013841753825545311,
|
319 |
"step": 200
|
320 |
},
|
321 |
{
|
322 |
"epoch": 0.52,
|
323 |
+
"eval_logits/chosen": -2.243248462677002,
|
324 |
+
"eval_logits/rejected": -2.238374948501587,
|
325 |
+
"eval_logps/chosen": -34.02431869506836,
|
326 |
+
"eval_logps/rejected": -37.511497497558594,
|
327 |
+
"eval_loss": 0.6946919560432434,
|
328 |
+
"eval_rewards/accuracies": 0.49335551261901855,
|
329 |
+
"eval_rewards/chosen": 0.009210066869854927,
|
330 |
+
"eval_rewards/margins": 0.0045979218557477,
|
331 |
+
"eval_rewards/rejected": 0.004612144082784653,
|
332 |
+
"eval_runtime": 146.0514,
|
333 |
+
"eval_samples_per_second": 2.348,
|
334 |
+
"eval_steps_per_second": 0.294,
|
335 |
"step": 200
|
336 |
},
|
337 |
{
|
338 |
"epoch": 0.55,
|
339 |
"learning_rate": 2.5453962426402e-07,
|
340 |
+
"logits/chosen": -2.0314748287200928,
|
341 |
+
"logits/rejected": -2.0421440601348877,
|
342 |
+
"logps/chosen": -31.954639434814453,
|
343 |
+
"logps/rejected": -33.87318801879883,
|
344 |
+
"loss": 0.6864,
|
345 |
+
"rewards/accuracies": 0.550000011920929,
|
346 |
+
"rewards/chosen": 0.014999927952885628,
|
347 |
+
"rewards/margins": 0.024442464113235474,
|
348 |
+
"rewards/rejected": -0.009442536160349846,
|
349 |
"step": 210
|
350 |
},
|
351 |
{
|
352 |
"epoch": 0.57,
|
353 |
"learning_rate": 2.318564697655179e-07,
|
354 |
+
"logits/chosen": -1.9249544143676758,
|
355 |
+
"logits/rejected": -1.9398053884506226,
|
356 |
+
"logps/chosen": -30.072036743164062,
|
357 |
+
"logps/rejected": -31.55165672302246,
|
358 |
+
"loss": 0.6827,
|
359 |
+
"rewards/accuracies": 0.6000000238418579,
|
360 |
+
"rewards/chosen": 0.035550959408283234,
|
361 |
+
"rewards/margins": 0.029681822285056114,
|
362 |
+
"rewards/rejected": 0.005869132932275534,
|
363 |
"step": 220
|
364 |
},
|
365 |
{
|
366 |
"epoch": 0.6,
|
367 |
"learning_rate": 2.093227910899832e-07,
|
368 |
+
"logits/chosen": -1.9830297231674194,
|
369 |
+
"logits/rejected": -1.9869968891143799,
|
370 |
+
"logps/chosen": -33.3900146484375,
|
371 |
+
"logps/rejected": -31.545156478881836,
|
372 |
+
"loss": 0.6872,
|
373 |
+
"rewards/accuracies": 0.5375000238418579,
|
374 |
+
"rewards/chosen": 0.024708259850740433,
|
375 |
+
"rewards/margins": 0.02097567543387413,
|
376 |
+
"rewards/rejected": 0.00373258744366467,
|
377 |
"step": 230
|
378 |
},
|
379 |
{
|
380 |
"epoch": 0.62,
|
381 |
"learning_rate": 1.8712423238279356e-07,
|
382 |
+
"logits/chosen": -1.9825174808502197,
|
383 |
+
"logits/rejected": -1.9605509042739868,
|
384 |
+
"logps/chosen": -34.16549301147461,
|
385 |
+
"logps/rejected": -34.945499420166016,
|
386 |
+
"loss": 0.7065,
|
387 |
+
"rewards/accuracies": 0.5,
|
388 |
+
"rewards/chosen": -0.010572168976068497,
|
389 |
+
"rewards/margins": -0.015337007120251656,
|
390 |
+
"rewards/rejected": 0.004764837212860584,
|
391 |
"step": 240
|
392 |
},
|
393 |
{
|
394 |
"epoch": 0.65,
|
395 |
"learning_rate": 1.654436768970182e-07,
|
396 |
+
"logits/chosen": -2.0242011547088623,
|
397 |
+
"logits/rejected": -2.02089262008667,
|
398 |
+
"logps/chosen": -32.91257858276367,
|
399 |
+
"logps/rejected": -36.22357177734375,
|
400 |
+
"loss": 0.6913,
|
401 |
+
"rewards/accuracies": 0.5,
|
402 |
+
"rewards/chosen": 0.01124012004584074,
|
403 |
+
"rewards/margins": 0.010020162910223007,
|
404 |
+
"rewards/rejected": 0.0012199539924040437,
|
405 |
"step": 250
|
406 |
},
|
407 |
{
|
408 |
"epoch": 0.68,
|
409 |
"learning_rate": 1.444597403062196e-07,
|
410 |
+
"logits/chosen": -1.8910300731658936,
|
411 |
+
"logits/rejected": -1.888593316078186,
|
412 |
+
"logps/chosen": -34.191619873046875,
|
413 |
+
"logps/rejected": -35.50851058959961,
|
414 |
+
"loss": 0.6975,
|
415 |
+
"rewards/accuracies": 0.6000000238418579,
|
416 |
+
"rewards/chosen": -0.0028982784133404493,
|
417 |
+
"rewards/margins": 0.0006028197822161019,
|
418 |
+
"rewards/rejected": -0.003501094877719879,
|
419 |
"step": 260
|
420 |
},
|
421 |
{
|
422 |
"epoch": 0.7,
|
423 |
"learning_rate": 1.2434529917578887e-07,
|
424 |
+
"logits/chosen": -1.8757693767547607,
|
425 |
+
"logits/rejected": -1.8732143640518188,
|
426 |
+
"logps/chosen": -34.38648223876953,
|
427 |
+
"logps/rejected": -31.753692626953125,
|
428 |
+
"loss": 0.6995,
|
429 |
+
"rewards/accuracies": 0.512499988079071,
|
430 |
+
"rewards/chosen": 0.010007266886532307,
|
431 |
+
"rewards/margins": -0.005866709630936384,
|
432 |
+
"rewards/rejected": 0.015873977914452553,
|
433 |
"step": 270
|
434 |
},
|
435 |
{
|
436 |
"epoch": 0.73,
|
437 |
"learning_rate": 1.0526606671603521e-07,
|
438 |
+
"logits/chosen": -1.9795001745224,
|
439 |
+
"logits/rejected": -1.9688692092895508,
|
440 |
+
"logps/chosen": -35.32746887207031,
|
441 |
+
"logps/rejected": -31.84292221069336,
|
442 |
+
"loss": 0.6843,
|
443 |
+
"rewards/accuracies": 0.550000011920929,
|
444 |
+
"rewards/chosen": 0.026254409924149513,
|
445 |
+
"rewards/margins": 0.027990642935037613,
|
446 |
+
"rewards/rejected": -0.001736226724460721,
|
447 |
"step": 280
|
448 |
},
|
449 |
{
|
450 |
"epoch": 0.75,
|
451 |
"learning_rate": 8.737922755071453e-08,
|
452 |
+
"logits/chosen": -2.0753533840179443,
|
453 |
+
"logits/rejected": -2.0603318214416504,
|
454 |
+
"logps/chosen": -30.902263641357422,
|
455 |
+
"logps/rejected": -32.65242385864258,
|
456 |
+
"loss": 0.6872,
|
457 |
+
"rewards/accuracies": 0.512499988079071,
|
458 |
+
"rewards/chosen": 0.019674303010106087,
|
459 |
+
"rewards/margins": 0.020201902836561203,
|
460 |
+
"rewards/rejected": -0.0005276011070236564,
|
461 |
"step": 290
|
462 |
},
|
463 |
{
|
464 |
"epoch": 0.78,
|
465 |
"learning_rate": 7.08321427484816e-08,
|
466 |
+
"logits/chosen": -1.946671724319458,
|
467 |
+
"logits/rejected": -1.9441486597061157,
|
468 |
+
"logps/chosen": -32.906158447265625,
|
469 |
+
"logps/rejected": -30.805999755859375,
|
470 |
+
"loss": 0.693,
|
471 |
+
"rewards/accuracies": 0.4749999940395355,
|
472 |
+
"rewards/chosen": 0.018858108669519424,
|
473 |
+
"rewards/margins": 0.009128611534833908,
|
474 |
+
"rewards/rejected": 0.009729497134685516,
|
475 |
"step": 300
|
476 |
},
|
477 |
{
|
478 |
"epoch": 0.78,
|
479 |
+
"eval_logits/chosen": -2.243694305419922,
|
480 |
+
"eval_logits/rejected": -2.2388274669647217,
|
481 |
+
"eval_logps/chosen": -34.017173767089844,
|
482 |
+
"eval_logps/rejected": -37.497615814208984,
|
483 |
+
"eval_loss": 0.6981561183929443,
|
484 |
"eval_rewards/accuracies": 0.49833887815475464,
|
485 |
+
"eval_rewards/chosen": 0.015641551464796066,
|
486 |
+
"eval_rewards/margins": -0.0014640000881627202,
|
487 |
+
"eval_rewards/rejected": 0.017105549573898315,
|
488 |
+
"eval_runtime": 146.0253,
|
489 |
+
"eval_samples_per_second": 2.349,
|
490 |
+
"eval_steps_per_second": 0.294,
|
491 |
"step": 300
|
492 |
},
|
493 |
{
|
494 |
"epoch": 0.81,
|
495 |
"learning_rate": 5.576113578589034e-08,
|
496 |
+
"logits/chosen": -1.9287757873535156,
|
497 |
+
"logits/rejected": -1.9255115985870361,
|
498 |
+
"logps/chosen": -31.573471069335938,
|
499 |
+
"logps/rejected": -33.71125030517578,
|
500 |
+
"loss": 0.698,
|
501 |
"rewards/accuracies": 0.5375000238418579,
|
502 |
+
"rewards/chosen": 0.026891669258475304,
|
503 |
+
"rewards/margins": -0.000918733305297792,
|
504 |
+
"rewards/rejected": 0.02781040407717228,
|
505 |
"step": 310
|
506 |
},
|
507 |
{
|
508 |
"epoch": 0.83,
|
509 |
"learning_rate": 4.229036944380912e-08,
|
510 |
+
"logits/chosen": -1.9807313680648804,
|
511 |
+
"logits/rejected": -1.968425989151001,
|
512 |
+
"logps/chosen": -34.54930877685547,
|
513 |
+
"logps/rejected": -33.57306671142578,
|
514 |
+
"loss": 0.6615,
|
515 |
+
"rewards/accuracies": 0.637499988079071,
|
516 |
+
"rewards/chosen": 0.04208679869771004,
|
517 |
+
"rewards/margins": 0.07289497554302216,
|
518 |
+
"rewards/rejected": -0.030808180570602417,
|
519 |
"step": 320
|
520 |
},
|
521 |
{
|
522 |
"epoch": 0.86,
|
523 |
"learning_rate": 3.053082288996112e-08,
|
524 |
+
"logits/chosen": -2.0161099433898926,
|
525 |
+
"logits/rejected": -2.0146496295928955,
|
526 |
+
"logps/chosen": -33.46127700805664,
|
527 |
+
"logps/rejected": -32.4744873046875,
|
528 |
+
"loss": 0.6883,
|
529 |
+
"rewards/accuracies": 0.5375000238418579,
|
530 |
+
"rewards/chosen": 0.022065162658691406,
|
531 |
+
"rewards/margins": 0.01747960038483143,
|
532 |
+
"rewards/rejected": 0.004585559479892254,
|
533 |
"step": 330
|
534 |
},
|
535 |
{
|
536 |
"epoch": 0.88,
|
537 |
"learning_rate": 2.05793773749158e-08,
|
538 |
+
"logits/chosen": -2.102802276611328,
|
539 |
+
"logits/rejected": -2.0869994163513184,
|
540 |
+
"logps/chosen": -34.16718292236328,
|
541 |
+
"logps/rejected": -33.08827590942383,
|
542 |
+
"loss": 0.706,
|
543 |
+
"rewards/accuracies": 0.48750001192092896,
|
544 |
+
"rewards/chosen": 0.010987209156155586,
|
545 |
+
"rewards/margins": -0.01579815149307251,
|
546 |
+
"rewards/rejected": 0.026785362511873245,
|
547 |
"step": 340
|
548 |
},
|
549 |
{
|
550 |
"epoch": 0.91,
|
551 |
"learning_rate": 1.251801807404168e-08,
|
552 |
+
"logits/chosen": -1.97471022605896,
|
553 |
+
"logits/rejected": -1.9737637042999268,
|
554 |
+
"logps/chosen": -33.25645065307617,
|
555 |
+
"logps/rejected": -32.46539306640625,
|
556 |
+
"loss": 0.6861,
|
557 |
+
"rewards/accuracies": 0.512499988079071,
|
558 |
+
"rewards/chosen": 0.023080622777342796,
|
559 |
+
"rewards/margins": 0.023302335292100906,
|
560 |
+
"rewards/rejected": -0.00022171511955093592,
|
561 |
"step": 350
|
562 |
},
|
563 |
{
|
564 |
"epoch": 0.94,
|
565 |
"learning_rate": 6.41315865106129e-09,
|
566 |
+
"logits/chosen": -1.9304895401000977,
|
567 |
+
"logits/rejected": -1.9408460855484009,
|
568 |
+
"logps/chosen": -32.22315979003906,
|
569 |
+
"logps/rejected": -35.28199005126953,
|
570 |
+
"loss": 0.7085,
|
571 |
+
"rewards/accuracies": 0.42500001192092896,
|
572 |
+
"rewards/chosen": -0.01662321388721466,
|
573 |
+
"rewards/margins": -0.022716889157891273,
|
574 |
+
"rewards/rejected": 0.006093672942370176,
|
575 |
"step": 360
|
576 |
},
|
577 |
{
|
578 |
"epoch": 0.96,
|
579 |
"learning_rate": 2.3150941078050324e-09,
|
580 |
+
"logits/chosen": -2.0692858695983887,
|
581 |
+
"logits/rejected": -2.0627408027648926,
|
582 |
+
"logps/chosen": -33.64545440673828,
|
583 |
+
"logps/rejected": -29.203441619873047,
|
584 |
+
"loss": 0.704,
|
585 |
+
"rewards/accuracies": 0.42500001192092896,
|
586 |
+
"rewards/chosen": 0.00586346909403801,
|
587 |
+
"rewards/margins": -0.014821496792137623,
|
588 |
+
"rewards/rejected": 0.020684964954853058,
|
589 |
"step": 370
|
590 |
},
|
591 |
{
|
592 |
"epoch": 0.99,
|
593 |
"learning_rate": 2.575864278703266e-10,
|
594 |
+
"logits/chosen": -1.928911805152893,
|
595 |
+
"logits/rejected": -1.9310725927352905,
|
596 |
+
"logps/chosen": -34.24489212036133,
|
597 |
+
"logps/rejected": -30.92240333557129,
|
598 |
+
"loss": 0.6763,
|
599 |
+
"rewards/accuracies": 0.5249999761581421,
|
600 |
+
"rewards/chosen": -0.005258283112198114,
|
601 |
+
"rewards/margins": 0.03954260051250458,
|
602 |
+
"rewards/rejected": -0.044800881296396255,
|
603 |
"step": 380
|
604 |
},
|
605 |
{
|
606 |
"epoch": 1.0,
|
607 |
"step": 385,
|
608 |
"total_flos": 0.0,
|
609 |
+
"train_loss": 0.6916276120520257,
|
610 |
+
"train_runtime": 3257.6062,
|
611 |
+
"train_samples_per_second": 0.945,
|
612 |
"train_steps_per_second": 0.118
|
613 |
}
|
614 |
],
|
training_args.bin
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 4984
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:914e32cb321285a370aa5440b23d05260ae348cf6fed9440d5a1e634ebac11d9
|
3 |
size 4984
|