hugodk-sch
commited on
Commit
•
cd1ee07
1
Parent(s):
9e15481
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 +381 -381
- 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: 0.
|
25 |
-
- Rewards/rejected: 0.
|
26 |
-
- Rewards/accuracies: 0.
|
27 |
-
- Rewards/margins:
|
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: 0.5000
|
21 |
+
- Rewards/chosen: 0.0003
|
22 |
+
- Rewards/rejected: 0.0003
|
23 |
+
- Rewards/accuracies: 0.5365
|
24 |
+
- Rewards/margins: 0.0000
|
25 |
+
- Logps/rejected: -37.4870
|
26 |
+
- Logps/chosen: -34.0017
|
27 |
+
- Logits/rejected: -2.2396
|
28 |
+
- Logits/chosen: -2.2445
|
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 |
+
| 0.5 | 0.26 | 100 | 0.5000 | 0.0001 | 0.0002 | 0.5137 | -0.0002 | -37.4933 | -34.0268 | -2.2390 | -2.2439 |
|
64 |
+
| 0.5 | 0.52 | 200 | 0.5000 | 0.0003 | 0.0002 | 0.5299 | 0.0001 | -37.4969 | -34.0030 | -2.2396 | -2.2444 |
|
65 |
+
| 0.4999 | 0.78 | 300 | 0.5000 | 0.0003 | 0.0003 | 0.5365 | 0.0000 | -37.4870 | -34.0017 | -2.2396 | -2.2445 |
|
66 |
|
67 |
|
68 |
### Framework versions
|
adapter_config.json
CHANGED
@@ -19,13 +19,13 @@
|
|
19 |
"rank_pattern": {},
|
20 |
"revision": null,
|
21 |
"target_modules": [
|
22 |
-
"up_proj",
|
23 |
"q_proj",
|
24 |
"gate_proj",
|
25 |
-
"
|
26 |
-
"
|
27 |
"v_proj",
|
28 |
-
"
|
|
|
29 |
],
|
30 |
"task_type": "CAUSAL_LM",
|
31 |
"use_rslora": false
|
|
|
19 |
"rank_pattern": {},
|
20 |
"revision": null,
|
21 |
"target_modules": [
|
|
|
22 |
"q_proj",
|
23 |
"gate_proj",
|
24 |
+
"up_proj",
|
25 |
+
"k_proj",
|
26 |
"v_proj",
|
27 |
+
"o_proj",
|
28 |
+
"down_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 sha256:3eca000156eaf07e8ac034ca10d58da5ad65e1f240e08fe64a18d57afb6fb4c4
|
3 |
size 176183216
|
all_results.json
CHANGED
@@ -1,21 +1,8 @@
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
-
"
|
4 |
-
"
|
5 |
-
"eval_logps/chosen": -34.02364730834961,
|
6 |
-
"eval_logps/rejected": -37.501853942871094,
|
7 |
-
"eval_loss": 0.6989675760269165,
|
8 |
-
"eval_rewards/accuracies": 0.49958473443984985,
|
9 |
-
"eval_rewards/chosen": 0.009816422127187252,
|
10 |
-
"eval_rewards/margins": -0.0034731123596429825,
|
11 |
-
"eval_rewards/rejected": 0.01328953169286251,
|
12 |
-
"eval_runtime": 145.7819,
|
13 |
-
"eval_samples": 343,
|
14 |
-
"eval_samples_per_second": 2.353,
|
15 |
-
"eval_steps_per_second": 0.295,
|
16 |
-
"train_loss": 0.6916276120520257,
|
17 |
-
"train_runtime": 3257.6062,
|
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": 0.4999783658362054,
|
4 |
+
"train_runtime": 3254.8055,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
5 |
"train_samples": 3079,
|
6 |
+
"train_samples_per_second": 0.946,
|
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": 0.4999783658362054,
|
4 |
+
"train_runtime": 3254.8055,
|
5 |
"train_samples": 3079,
|
6 |
+
"train_samples_per_second": 0.946,
|
7 |
"train_steps_per_second": 0.118
|
8 |
}
|
trainer_state.json
CHANGED
@@ -15,7 +15,7 @@
|
|
15 |
"logits/rejected": -1.7377450466156006,
|
16 |
"logps/chosen": -29.553977966308594,
|
17 |
"logps/rejected": -42.813133239746094,
|
18 |
-
"loss": 0.
|
19 |
"rewards/accuracies": 0.0,
|
20 |
"rewards/chosen": 0.0,
|
21 |
"rewards/margins": 0.0,
|
@@ -25,590 +25,590 @@
|
|
25 |
{
|
26 |
"epoch": 0.03,
|
27 |
"learning_rate": 1.2820512820512818e-07,
|
28 |
-
"logits/chosen": -1.
|
29 |
-
"logits/rejected": -1.
|
30 |
-
"logps/chosen": -
|
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": -
|
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":
|
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":
|
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.
|
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":
|
119 |
-
"rewards/margins":
|
120 |
-
"rewards/rejected": -
|
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.
|
146 |
-
"rewards/chosen": 0.
|
147 |
-
"rewards/margins":
|
148 |
-
"rewards/rejected": 0.
|
149 |
"step": 90
|
150 |
},
|
151 |
{
|
152 |
"epoch": 0.26,
|
153 |
"learning_rate": 4.62624545834521e-07,
|
154 |
-
"logits/chosen": -2.
|
155 |
-
"logits/rejected": -2.
|
156 |
-
"logps/chosen": -32.
|
157 |
-
"logps/rejected": -31.
|
158 |
-
"loss": 0.
|
159 |
-
"rewards/accuracies": 0.
|
160 |
-
"rewards/chosen":
|
161 |
-
"rewards/margins": 0.
|
162 |
-
"rewards/rejected": 0.
|
163 |
"step": 100
|
164 |
},
|
165 |
{
|
166 |
"epoch": 0.26,
|
167 |
-
"eval_logits/chosen": -2.
|
168 |
-
"eval_logits/rejected": -2.
|
169 |
-
"eval_logps/chosen": -34.
|
170 |
-
"eval_logps/rejected": -37.
|
171 |
-
"eval_loss": 0.
|
172 |
-
"eval_rewards/accuracies": 0.
|
173 |
-
"eval_rewards/chosen":
|
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":
|
191 |
-
"rewards/margins": -
|
192 |
-
"rewards/rejected": 0.
|
193 |
"step": 110
|
194 |
},
|
195 |
{
|
196 |
"epoch": 0.31,
|
197 |
"learning_rate": 4.353806263777677e-07,
|
198 |
-
"logits/chosen": -2.
|
199 |
-
"logits/rejected": -2.
|
200 |
-
"logps/chosen": -32.
|
201 |
-
"logps/rejected": -32.
|
202 |
-
"loss": 0.
|
203 |
-
"rewards/accuracies": 0.
|
204 |
-
"rewards/chosen": -
|
205 |
-
"rewards/margins":
|
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": -
|
219 |
-
"rewards/margins": -0.
|
220 |
-
"rewards/rejected":
|
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.
|
246 |
-
"rewards/chosen": 0.
|
247 |
-
"rewards/margins": 0.
|
248 |
-
"rewards/rejected": -
|
249 |
"step": 150
|
250 |
},
|
251 |
{
|
252 |
"epoch": 0.42,
|
253 |
"learning_rate": 3.636998309800572e-07,
|
254 |
-
"logits/chosen": -1.
|
255 |
-
"logits/rejected": -1.
|
256 |
-
"logps/chosen": -36.
|
257 |
-
"logps/rejected": -32.
|
258 |
-
"loss": 0.
|
259 |
-
"rewards/accuracies": 0.
|
260 |
-
"rewards/chosen": 0.
|
261 |
-
"rewards/margins":
|
262 |
-
"rewards/rejected":
|
263 |
"step": 160
|
264 |
},
|
265 |
{
|
266 |
"epoch": 0.44,
|
267 |
"learning_rate": 3.430433172111807e-07,
|
268 |
-
"logits/chosen": -2.
|
269 |
-
"logits/rejected": -2.
|
270 |
-
"logps/chosen": -33.
|
271 |
-
"logps/rejected": -31.
|
272 |
-
"loss": 0.
|
273 |
-
"rewards/accuracies": 0.
|
274 |
-
"rewards/chosen": 0.
|
275 |
-
"rewards/margins": 0.
|
276 |
-
"rewards/rejected": -
|
277 |
"step": 170
|
278 |
},
|
279 |
{
|
280 |
"epoch": 0.47,
|
281 |
"learning_rate": 3.216202642830543e-07,
|
282 |
-
"logits/chosen": -2.
|
283 |
-
"logits/rejected": -2.
|
284 |
-
"logps/chosen": -32.
|
285 |
-
"logps/rejected": -32.
|
286 |
-
"loss": 0.
|
287 |
-
"rewards/accuracies": 0.
|
288 |
-
"rewards/chosen": 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": 0.
|
303 |
-
"rewards/margins": 0.
|
304 |
-
"rewards/rejected": -
|
305 |
"step": 190
|
306 |
},
|
307 |
{
|
308 |
"epoch": 0.52,
|
309 |
"learning_rate": 2.771853789806683e-07,
|
310 |
-
"logits/chosen": -1.
|
311 |
-
"logits/rejected": -1.
|
312 |
-
"logps/chosen": -31.
|
313 |
-
"logps/rejected": -32.
|
314 |
-
"loss": 0.
|
315 |
-
"rewards/accuracies": 0.
|
316 |
-
"rewards/chosen":
|
317 |
-
"rewards/margins":
|
318 |
-
"rewards/rejected": 0.
|
319 |
"step": 200
|
320 |
},
|
321 |
{
|
322 |
"epoch": 0.52,
|
323 |
-
"eval_logits/chosen": -2.
|
324 |
-
"eval_logits/rejected": -2.
|
325 |
-
"eval_logps/chosen": -34.
|
326 |
-
"eval_logps/rejected": -37.
|
327 |
-
"eval_loss": 0.
|
328 |
-
"eval_rewards/accuracies": 0.
|
329 |
-
"eval_rewards/chosen": 0.
|
330 |
-
"eval_rewards/margins": 0.
|
331 |
-
"eval_rewards/rejected": 0.
|
332 |
-
"eval_runtime":
|
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":
|
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": -
|
389 |
-
"rewards/margins":
|
390 |
-
"rewards/rejected": 0.
|
391 |
"step": 240
|
392 |
},
|
393 |
{
|
394 |
"epoch": 0.65,
|
395 |
"learning_rate": 1.654436768970182e-07,
|
396 |
-
"logits/chosen": -2.
|
397 |
-
"logits/rejected": -2.
|
398 |
-
"logps/chosen": -32.
|
399 |
-
"logps/rejected": -36.
|
400 |
-
"loss": 0.
|
401 |
-
"rewards/accuracies": 0.
|
402 |
-
"rewards/chosen": 0.
|
403 |
-
"rewards/margins": 0.
|
404 |
-
"rewards/rejected": 0.
|
405 |
"step": 250
|
406 |
},
|
407 |
{
|
408 |
"epoch": 0.68,
|
409 |
"learning_rate": 1.444597403062196e-07,
|
410 |
-
"logits/chosen": -1.
|
411 |
-
"logits/rejected": -1.
|
412 |
-
"logps/chosen": -34.
|
413 |
-
"logps/rejected": -35.
|
414 |
-
"loss": 0.
|
415 |
-
"rewards/accuracies": 0.
|
416 |
-
"rewards/chosen":
|
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":
|
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": -
|
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.
|
485 |
-
"eval_rewards/chosen": 0.
|
486 |
-
"eval_rewards/margins": -
|
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.
|
502 |
-
"rewards/chosen": 0.
|
503 |
-
"rewards/margins":
|
504 |
-
"rewards/rejected":
|
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": -
|
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":
|
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": -
|
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": 0.
|
575 |
"step": 360
|
576 |
},
|
577 |
{
|
578 |
"epoch": 0.96,
|
579 |
"learning_rate": 2.3150941078050324e-09,
|
580 |
-
"logits/chosen": -2.
|
581 |
-
"logits/rejected": -2.
|
582 |
-
"logps/chosen": -33.
|
583 |
-
"logps/rejected": -29.
|
584 |
-
"loss": 0.
|
585 |
-
"rewards/accuracies": 0.
|
586 |
-
"rewards/chosen":
|
587 |
-
"rewards/margins": -0.
|
588 |
-
"rewards/rejected": 0.
|
589 |
"step": 370
|
590 |
},
|
591 |
{
|
592 |
"epoch": 0.99,
|
593 |
"learning_rate": 2.575864278703266e-10,
|
594 |
-
"logits/chosen": -1.
|
595 |
-
"logits/rejected": -1.
|
596 |
-
"logps/chosen": -34.
|
597 |
-
"logps/rejected": -30.
|
598 |
-
"loss": 0.
|
599 |
-
"rewards/accuracies": 0.
|
600 |
-
"rewards/chosen": -
|
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": 0.5,
|
19 |
"rewards/accuracies": 0.0,
|
20 |
"rewards/chosen": 0.0,
|
21 |
"rewards/margins": 0.0,
|
|
|
25 |
{
|
26 |
"epoch": 0.03,
|
27 |
"learning_rate": 1.2820512820512818e-07,
|
28 |
+
"logits/chosen": -1.8662651777267456,
|
29 |
+
"logits/rejected": -1.8705823421478271,
|
30 |
+
"logps/chosen": -37.01448440551758,
|
31 |
+
"logps/rejected": -33.66516876220703,
|
32 |
+
"loss": 0.5,
|
33 |
+
"rewards/accuracies": 0.5,
|
34 |
+
"rewards/chosen": -7.885291415732354e-05,
|
35 |
+
"rewards/margins": 0.0002720804768614471,
|
36 |
+
"rewards/rejected": -0.00035093340557068586,
|
37 |
"step": 10
|
38 |
},
|
39 |
{
|
40 |
"epoch": 0.05,
|
41 |
"learning_rate": 2.5641025641025636e-07,
|
42 |
+
"logits/chosen": -1.9972978830337524,
|
43 |
+
"logits/rejected": -1.9999454021453857,
|
44 |
+
"logps/chosen": -29.626657485961914,
|
45 |
+
"logps/rejected": -29.065475463867188,
|
46 |
+
"loss": 0.5,
|
47 |
+
"rewards/accuracies": 0.550000011920929,
|
48 |
+
"rewards/chosen": 0.00015535543207079172,
|
49 |
+
"rewards/margins": 0.00017565980670042336,
|
50 |
+
"rewards/rejected": -2.0304325516917743e-05,
|
51 |
"step": 20
|
52 |
},
|
53 |
{
|
54 |
"epoch": 0.08,
|
55 |
"learning_rate": 3.8461538461538463e-07,
|
56 |
+
"logits/chosen": -1.9197067022323608,
|
57 |
+
"logits/rejected": -1.917022466659546,
|
58 |
+
"logps/chosen": -31.42661476135254,
|
59 |
+
"logps/rejected": -33.22706985473633,
|
60 |
+
"loss": 0.5,
|
61 |
+
"rewards/accuracies": 0.5375000238418579,
|
62 |
+
"rewards/chosen": -0.0001051858052960597,
|
63 |
+
"rewards/margins": -1.9103928934782743e-05,
|
64 |
+
"rewards/rejected": -8.608187636127695e-05,
|
65 |
"step": 30
|
66 |
},
|
67 |
{
|
68 |
"epoch": 0.1,
|
69 |
"learning_rate": 4.999896948438433e-07,
|
70 |
+
"logits/chosen": -2.0163283348083496,
|
71 |
+
"logits/rejected": -2.0075771808624268,
|
72 |
+
"logps/chosen": -32.59123992919922,
|
73 |
+
"logps/rejected": -32.48072052001953,
|
74 |
+
"loss": 0.5001,
|
75 |
+
"rewards/accuracies": 0.4625000059604645,
|
76 |
+
"rewards/chosen": -0.00014431352610699832,
|
77 |
+
"rewards/margins": -0.00046383924200199544,
|
78 |
+
"rewards/rejected": 0.0003195256576873362,
|
79 |
"step": 40
|
80 |
},
|
81 |
{
|
82 |
"epoch": 0.13,
|
83 |
"learning_rate": 4.987541037542186e-07,
|
84 |
+
"logits/chosen": -1.8642585277557373,
|
85 |
+
"logits/rejected": -1.8534786701202393,
|
86 |
+
"logps/chosen": -33.55145263671875,
|
87 |
+
"logps/rejected": -35.415096282958984,
|
88 |
+
"loss": 0.5001,
|
89 |
+
"rewards/accuracies": 0.375,
|
90 |
+
"rewards/chosen": 6.026435949024744e-05,
|
91 |
+
"rewards/margins": -0.0003122163761872798,
|
92 |
+
"rewards/rejected": 0.00037248071748763323,
|
93 |
"step": 50
|
94 |
},
|
95 |
{
|
96 |
"epoch": 0.16,
|
97 |
"learning_rate": 4.954691471941118e-07,
|
98 |
+
"logits/chosen": -1.9458900690078735,
|
99 |
+
"logits/rejected": -1.9478263854980469,
|
100 |
+
"logps/chosen": -32.57926559448242,
|
101 |
+
"logps/rejected": -33.194175720214844,
|
102 |
+
"loss": 0.4999,
|
103 |
+
"rewards/accuracies": 0.5625,
|
104 |
+
"rewards/chosen": 0.00020916396169923246,
|
105 |
+
"rewards/margins": 0.00043710897443816066,
|
106 |
+
"rewards/rejected": -0.00022794504184275866,
|
107 |
"step": 60
|
108 |
},
|
109 |
{
|
110 |
"epoch": 0.18,
|
111 |
"learning_rate": 4.901618883413548e-07,
|
112 |
+
"logits/chosen": -2.0795435905456543,
|
113 |
+
"logits/rejected": -2.084535837173462,
|
114 |
+
"logps/chosen": -33.986812591552734,
|
115 |
+
"logps/rejected": -36.573219299316406,
|
116 |
+
"loss": 0.5,
|
117 |
+
"rewards/accuracies": 0.48750001192092896,
|
118 |
+
"rewards/chosen": 1.6729489288991317e-05,
|
119 |
+
"rewards/margins": 6.016326369717717e-05,
|
120 |
+
"rewards/rejected": -4.343378896010108e-05,
|
121 |
"step": 70
|
122 |
},
|
123 |
{
|
124 |
"epoch": 0.21,
|
125 |
"learning_rate": 4.828760511501322e-07,
|
126 |
+
"logits/chosen": -1.9420961141586304,
|
127 |
+
"logits/rejected": -1.9452673196792603,
|
128 |
+
"logps/chosen": -34.41278839111328,
|
129 |
+
"logps/rejected": -34.570655822753906,
|
130 |
+
"loss": 0.5,
|
131 |
+
"rewards/accuracies": 0.512499988079071,
|
132 |
+
"rewards/chosen": 0.00015023796004243195,
|
133 |
+
"rewards/margins": -6.727596337441355e-05,
|
134 |
+
"rewards/rejected": 0.0002175139234168455,
|
135 |
"step": 80
|
136 |
},
|
137 |
{
|
138 |
"epoch": 0.23,
|
139 |
"learning_rate": 4.736716601303429e-07,
|
140 |
+
"logits/chosen": -1.9506244659423828,
|
141 |
+
"logits/rejected": -1.9551340341567993,
|
142 |
+
"logps/chosen": -32.45224380493164,
|
143 |
+
"logps/rejected": -32.344200134277344,
|
144 |
+
"loss": 0.5,
|
145 |
+
"rewards/accuracies": 0.512499988079071,
|
146 |
+
"rewards/chosen": 0.0001517904020147398,
|
147 |
+
"rewards/margins": 4.694242761615897e-06,
|
148 |
+
"rewards/rejected": 0.00014709614333696663,
|
149 |
"step": 90
|
150 |
},
|
151 |
{
|
152 |
"epoch": 0.26,
|
153 |
"learning_rate": 4.62624545834521e-07,
|
154 |
+
"logits/chosen": -2.048786163330078,
|
155 |
+
"logits/rejected": -2.046797275543213,
|
156 |
+
"logps/chosen": -32.23818588256836,
|
157 |
+
"logps/rejected": -31.283069610595703,
|
158 |
+
"loss": 0.5,
|
159 |
+
"rewards/accuracies": 0.5874999761581421,
|
160 |
+
"rewards/chosen": 2.6492711185710505e-05,
|
161 |
+
"rewards/margins": 0.0001950610603671521,
|
162 |
+
"rewards/rejected": -0.00016856836737133563,
|
163 |
"step": 100
|
164 |
},
|
165 |
{
|
166 |
"epoch": 0.26,
|
167 |
+
"eval_logits/chosen": -2.243879556655884,
|
168 |
+
"eval_logits/rejected": -2.2390010356903076,
|
169 |
+
"eval_logps/chosen": -34.026790618896484,
|
170 |
+
"eval_logps/rejected": -37.49333572387695,
|
171 |
+
"eval_loss": 0.5000349879264832,
|
172 |
+
"eval_rewards/accuracies": 0.5137043595314026,
|
173 |
+
"eval_rewards/chosen": 7.762637687847018e-05,
|
174 |
+
"eval_rewards/margins": -0.00015523387992288917,
|
175 |
+
"eval_rewards/rejected": 0.00023286025680135936,
|
176 |
+
"eval_runtime": 146.2199,
|
177 |
+
"eval_samples_per_second": 2.346,
|
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": -2.005380392074585,
|
185 |
+
"logits/rejected": -2.0029690265655518,
|
186 |
+
"logps/chosen": -33.23069381713867,
|
187 |
+
"logps/rejected": -34.01119613647461,
|
188 |
+
"loss": 0.5,
|
189 |
+
"rewards/accuracies": 0.5249999761581421,
|
190 |
+
"rewards/chosen": 0.0001312425301875919,
|
191 |
+
"rewards/margins": -5.051253901910968e-05,
|
192 |
+
"rewards/rejected": 0.00018175505101680756,
|
193 |
"step": 110
|
194 |
},
|
195 |
{
|
196 |
"epoch": 0.31,
|
197 |
"learning_rate": 4.353806263777677e-07,
|
198 |
+
"logits/chosen": -2.016681432723999,
|
199 |
+
"logits/rejected": -2.0083086490631104,
|
200 |
+
"logps/chosen": -32.43777847290039,
|
201 |
+
"logps/rejected": -32.1855354309082,
|
202 |
+
"loss": 0.5,
|
203 |
+
"rewards/accuracies": 0.5625,
|
204 |
+
"rewards/chosen": 7.132566679501906e-05,
|
205 |
+
"rewards/margins": 0.00031160234357230365,
|
206 |
+
"rewards/rejected": -0.0002402766840532422,
|
207 |
"step": 120
|
208 |
},
|
209 |
{
|
210 |
"epoch": 0.34,
|
211 |
"learning_rate": 4.194082707715275e-07,
|
212 |
+
"logits/chosen": -2.04573392868042,
|
213 |
+
"logits/rejected": -2.0376956462860107,
|
214 |
+
"logps/chosen": -30.474706649780273,
|
215 |
+
"logps/rejected": -32.02512741088867,
|
216 |
+
"loss": 0.5001,
|
217 |
+
"rewards/accuracies": 0.44999998807907104,
|
218 |
+
"rewards/chosen": 1.0108102287631482e-05,
|
219 |
+
"rewards/margins": -0.0002687241358216852,
|
220 |
+
"rewards/rejected": 0.00027883221628144383,
|
221 |
"step": 130
|
222 |
},
|
223 |
{
|
224 |
"epoch": 0.36,
|
225 |
"learning_rate": 4.020402418666621e-07,
|
226 |
+
"logits/chosen": -1.9768989086151123,
|
227 |
+
"logits/rejected": -1.9871835708618164,
|
228 |
+
"logps/chosen": -31.406789779663086,
|
229 |
+
"logps/rejected": -32.54368209838867,
|
230 |
+
"loss": 0.5,
|
231 |
+
"rewards/accuracies": 0.550000011920929,
|
232 |
+
"rewards/chosen": 0.00012870995851699263,
|
233 |
+
"rewards/margins": 0.00021670905698556453,
|
234 |
+
"rewards/rejected": -8.799910574452952e-05,
|
235 |
"step": 140
|
236 |
},
|
237 |
{
|
238 |
"epoch": 0.39,
|
239 |
"learning_rate": 3.8341962650351185e-07,
|
240 |
+
"logits/chosen": -1.8907657861709595,
|
241 |
+
"logits/rejected": -1.8918464183807373,
|
242 |
+
"logps/chosen": -34.17363739013672,
|
243 |
+
"logps/rejected": -34.758506774902344,
|
244 |
+
"loss": 0.5,
|
245 |
+
"rewards/accuracies": 0.5375000238418579,
|
246 |
+
"rewards/chosen": 0.00019054378208238631,
|
247 |
+
"rewards/margins": 0.00020306630176492035,
|
248 |
+
"rewards/rejected": -1.2522534234449267e-05,
|
249 |
"step": 150
|
250 |
},
|
251 |
{
|
252 |
"epoch": 0.42,
|
253 |
"learning_rate": 3.636998309800572e-07,
|
254 |
+
"logits/chosen": -1.9421974420547485,
|
255 |
+
"logits/rejected": -1.9387152194976807,
|
256 |
+
"logps/chosen": -36.14430618286133,
|
257 |
+
"logps/rejected": -32.699371337890625,
|
258 |
+
"loss": 0.5,
|
259 |
+
"rewards/accuracies": 0.6000000238418579,
|
260 |
+
"rewards/chosen": 0.0002737110189627856,
|
261 |
+
"rewards/margins": 3.994655708083883e-05,
|
262 |
+
"rewards/rejected": 0.00023376448370981961,
|
263 |
"step": 160
|
264 |
},
|
265 |
{
|
266 |
"epoch": 0.44,
|
267 |
"learning_rate": 3.430433172111807e-07,
|
268 |
+
"logits/chosen": -2.042232036590576,
|
269 |
+
"logits/rejected": -2.034842014312744,
|
270 |
+
"logps/chosen": -33.76560974121094,
|
271 |
+
"logps/rejected": -31.359146118164062,
|
272 |
+
"loss": 0.4999,
|
273 |
+
"rewards/accuracies": 0.550000011920929,
|
274 |
+
"rewards/chosen": 0.0003225207910872996,
|
275 |
+
"rewards/margins": 0.00039547347114421427,
|
276 |
+
"rewards/rejected": -7.295267278095707e-05,
|
277 |
"step": 170
|
278 |
},
|
279 |
{
|
280 |
"epoch": 0.47,
|
281 |
"learning_rate": 3.216202642830543e-07,
|
282 |
+
"logits/chosen": -2.047433376312256,
|
283 |
+
"logits/rejected": -2.052696704864502,
|
284 |
+
"logps/chosen": -32.51101303100586,
|
285 |
+
"logps/rejected": -32.49535369873047,
|
286 |
+
"loss": 0.4999,
|
287 |
+
"rewards/accuracies": 0.550000011920929,
|
288 |
+
"rewards/chosen": 0.00033775376505218446,
|
289 |
+
"rewards/margins": 0.00043991307029500604,
|
290 |
+
"rewards/rejected": -0.0001021592179313302,
|
291 |
"step": 180
|
292 |
},
|
293 |
{
|
294 |
"epoch": 0.49,
|
295 |
"learning_rate": 2.9960716642946403e-07,
|
296 |
+
"logits/chosen": -2.048191785812378,
|
297 |
+
"logits/rejected": -2.0454134941101074,
|
298 |
+
"logps/chosen": -31.475040435791016,
|
299 |
+
"logps/rejected": -31.315017700195312,
|
300 |
+
"loss": 0.5,
|
301 |
+
"rewards/accuracies": 0.5249999761581421,
|
302 |
+
"rewards/chosen": 0.00013913081784266979,
|
303 |
+
"rewards/margins": 0.00010714305244619027,
|
304 |
+
"rewards/rejected": 3.1987779948394746e-05,
|
305 |
"step": 190
|
306 |
},
|
307 |
{
|
308 |
"epoch": 0.52,
|
309 |
"learning_rate": 2.771853789806683e-07,
|
310 |
+
"logits/chosen": -1.9189927577972412,
|
311 |
+
"logits/rejected": -1.9236657619476318,
|
312 |
+
"logps/chosen": -31.597900390625,
|
313 |
+
"logps/rejected": -32.774513244628906,
|
314 |
+
"loss": 0.5,
|
315 |
+
"rewards/accuracies": 0.5,
|
316 |
+
"rewards/chosen": 8.383009844692424e-05,
|
317 |
+
"rewards/margins": -8.035473001655191e-05,
|
318 |
+
"rewards/rejected": 0.0001641848066356033,
|
319 |
"step": 200
|
320 |
},
|
321 |
{
|
322 |
"epoch": 0.52,
|
323 |
+
"eval_logits/chosen": -2.244427442550659,
|
324 |
+
"eval_logits/rejected": -2.2395591735839844,
|
325 |
+
"eval_logps/chosen": -34.002986907958984,
|
326 |
+
"eval_logps/rejected": -37.49692916870117,
|
327 |
+
"eval_loss": 0.4999743700027466,
|
328 |
+
"eval_rewards/accuracies": 0.529900312423706,
|
329 |
+
"eval_rewards/chosen": 0.00031563788070343435,
|
330 |
+
"eval_rewards/margins": 0.0001187350062537007,
|
331 |
+
"eval_rewards/rejected": 0.00019690288172569126,
|
332 |
+
"eval_runtime": 145.9131,
|
333 |
+
"eval_samples_per_second": 2.351,
|
334 |
+
"eval_steps_per_second": 0.295,
|
335 |
"step": 200
|
336 |
},
|
337 |
{
|
338 |
"epoch": 0.55,
|
339 |
"learning_rate": 2.5453962426402e-07,
|
340 |
+
"logits/chosen": -2.031794786453247,
|
341 |
+
"logits/rejected": -2.0424611568450928,
|
342 |
+
"logps/chosen": -31.942611694335938,
|
343 |
+
"logps/rejected": -33.87322998046875,
|
344 |
+
"loss": 0.4999,
|
345 |
+
"rewards/accuracies": 0.512499988079071,
|
346 |
+
"rewards/chosen": 0.0002869277377612889,
|
347 |
+
"rewards/margins": 0.0003922684118151665,
|
348 |
+
"rewards/rejected": -0.00010534075408941135,
|
349 |
"step": 210
|
350 |
},
|
351 |
{
|
352 |
"epoch": 0.57,
|
353 |
"learning_rate": 2.318564697655179e-07,
|
354 |
+
"logits/chosen": -1.925797462463379,
|
355 |
+
"logits/rejected": -1.9406683444976807,
|
356 |
+
"logps/chosen": -30.079837799072266,
|
357 |
+
"logps/rejected": -31.56089210510254,
|
358 |
+
"loss": 0.4999,
|
359 |
+
"rewards/accuracies": 0.5874999761581421,
|
360 |
+
"rewards/chosen": 0.00031700366525910795,
|
361 |
+
"rewards/margins": 0.0003441544540692121,
|
362 |
+
"rewards/rejected": -2.715078335313592e-05,
|
363 |
"step": 220
|
364 |
},
|
365 |
{
|
366 |
"epoch": 0.6,
|
367 |
"learning_rate": 2.093227910899832e-07,
|
368 |
+
"logits/chosen": -1.9836803674697876,
|
369 |
+
"logits/rejected": -1.9876391887664795,
|
370 |
+
"logps/chosen": -33.396968841552734,
|
371 |
+
"logps/rejected": -31.5497989654541,
|
372 |
+
"loss": 0.5,
|
373 |
+
"rewards/accuracies": 0.550000011920929,
|
374 |
+
"rewards/chosen": 0.00020496857177931815,
|
375 |
+
"rewards/margins": 0.0002099033008562401,
|
376 |
+
"rewards/rejected": -4.934729986416642e-06,
|
377 |
"step": 230
|
378 |
},
|
379 |
{
|
380 |
"epoch": 0.62,
|
381 |
"learning_rate": 1.8712423238279356e-07,
|
382 |
+
"logits/chosen": -1.9834476709365845,
|
383 |
+
"logits/rejected": -1.9614884853363037,
|
384 |
+
"logps/chosen": -34.1593017578125,
|
385 |
+
"logps/rejected": -34.974971771240234,
|
386 |
+
"loss": 0.5,
|
387 |
+
"rewards/accuracies": 0.550000011920929,
|
388 |
+
"rewards/chosen": -5.55333899683319e-05,
|
389 |
+
"rewards/margins": 0.00018626711971592158,
|
390 |
+
"rewards/rejected": -0.00024180053151212633,
|
391 |
"step": 240
|
392 |
},
|
393 |
{
|
394 |
"epoch": 0.65,
|
395 |
"learning_rate": 1.654436768970182e-07,
|
396 |
+
"logits/chosen": -2.0252513885498047,
|
397 |
+
"logits/rejected": -2.021942615509033,
|
398 |
+
"logps/chosen": -32.891014099121094,
|
399 |
+
"logps/rejected": -36.209808349609375,
|
400 |
+
"loss": 0.4999,
|
401 |
+
"rewards/accuracies": 0.4749999940395355,
|
402 |
+
"rewards/chosen": 0.00034049636451527476,
|
403 |
+
"rewards/margins": 0.00018933510000351816,
|
404 |
+
"rewards/rejected": 0.00015116124995984137,
|
405 |
"step": 250
|
406 |
},
|
407 |
{
|
408 |
"epoch": 0.68,
|
409 |
"learning_rate": 1.444597403062196e-07,
|
410 |
+
"logits/chosen": -1.891709327697754,
|
411 |
+
"logits/rejected": -1.889259934425354,
|
412 |
+
"logps/chosen": -34.162227630615234,
|
413 |
+
"logps/rejected": -35.52161407470703,
|
414 |
+
"loss": 0.4999,
|
415 |
+
"rewards/accuracies": 0.612500011920929,
|
416 |
+
"rewards/chosen": 0.0002616822312120348,
|
417 |
+
"rewards/margins": 0.0004316047707106918,
|
418 |
+
"rewards/rejected": -0.0001699226413620636,
|
419 |
"step": 260
|
420 |
},
|
421 |
{
|
422 |
"epoch": 0.7,
|
423 |
"learning_rate": 1.2434529917578887e-07,
|
424 |
+
"logits/chosen": -1.8765720129013062,
|
425 |
+
"logits/rejected": -1.8740289211273193,
|
426 |
+
"logps/chosen": -34.38764190673828,
|
427 |
+
"logps/rejected": -31.73568344116211,
|
428 |
+
"loss": 0.5,
|
429 |
+
"rewards/accuracies": 0.44999998807907104,
|
430 |
+
"rewards/chosen": 9.958025475498289e-05,
|
431 |
+
"rewards/margins": -0.00025685896980576217,
|
432 |
+
"rewards/rejected": 0.00035643921000882983,
|
433 |
"step": 270
|
434 |
},
|
435 |
{
|
436 |
"epoch": 0.73,
|
437 |
"learning_rate": 1.0526606671603521e-07,
|
438 |
+
"logits/chosen": -1.980292558670044,
|
439 |
+
"logits/rejected": -1.9696547985076904,
|
440 |
+
"logps/chosen": -35.3267936706543,
|
441 |
+
"logps/rejected": -31.829299926757812,
|
442 |
+
"loss": 0.5,
|
443 |
+
"rewards/accuracies": 0.4749999940395355,
|
444 |
+
"rewards/chosen": 0.00029847928090021014,
|
445 |
+
"rewards/margins": 0.00018156312580686063,
|
446 |
+
"rewards/rejected": 0.00011691616236930713,
|
447 |
"step": 280
|
448 |
},
|
449 |
{
|
450 |
"epoch": 0.75,
|
451 |
"learning_rate": 8.737922755071453e-08,
|
452 |
+
"logits/chosen": -2.075955390930176,
|
453 |
+
"logits/rejected": -2.0609400272369385,
|
454 |
+
"logps/chosen": -30.922338485717773,
|
455 |
+
"logps/rejected": -32.651031494140625,
|
456 |
+
"loss": 0.5,
|
457 |
+
"rewards/accuracies": 0.4625000059604645,
|
458 |
+
"rewards/chosen": 1.7866250345832668e-05,
|
459 |
+
"rewards/margins": 9.798279279493727e-06,
|
460 |
+
"rewards/rejected": 8.067952876444906e-06,
|
461 |
"step": 290
|
462 |
},
|
463 |
{
|
464 |
"epoch": 0.78,
|
465 |
"learning_rate": 7.08321427484816e-08,
|
466 |
+
"logits/chosen": -1.9476518630981445,
|
467 |
+
"logits/rejected": -1.945123314857483,
|
468 |
+
"logps/chosen": -32.88804626464844,
|
469 |
+
"logps/rejected": -30.81081771850586,
|
470 |
+
"loss": 0.4999,
|
471 |
+
"rewards/accuracies": 0.4625000059604645,
|
472 |
+
"rewards/chosen": 0.00039066909812390804,
|
473 |
+
"rewards/margins": 0.0003307634615339339,
|
474 |
+
"rewards/rejected": 5.9905636589974165e-05,
|
475 |
"step": 300
|
476 |
},
|
477 |
{
|
478 |
"epoch": 0.78,
|
479 |
+
"eval_logits/chosen": -2.2444517612457275,
|
480 |
+
"eval_logits/rejected": -2.239577054977417,
|
481 |
+
"eval_logps/chosen": -34.00169372558594,
|
482 |
+
"eval_logps/rejected": -37.48704528808594,
|
483 |
+
"eval_loss": 0.49999141693115234,
|
484 |
+
"eval_rewards/accuracies": 0.5365448594093323,
|
485 |
+
"eval_rewards/chosen": 0.0003285549464635551,
|
486 |
+
"eval_rewards/margins": 3.2781601476017386e-05,
|
487 |
+
"eval_rewards/rejected": 0.00029577332315966487,
|
488 |
+
"eval_runtime": 145.927,
|
489 |
+
"eval_samples_per_second": 2.35,
|
490 |
+
"eval_steps_per_second": 0.295,
|
491 |
"step": 300
|
492 |
},
|
493 |
{
|
494 |
"epoch": 0.81,
|
495 |
"learning_rate": 5.576113578589034e-08,
|
496 |
+
"logits/chosen": -1.929610013961792,
|
497 |
+
"logits/rejected": -1.9263417720794678,
|
498 |
+
"logps/chosen": -31.55474281311035,
|
499 |
+
"logps/rejected": -33.74074935913086,
|
500 |
+
"loss": 0.4999,
|
501 |
+
"rewards/accuracies": 0.612500011920929,
|
502 |
+
"rewards/chosen": 0.0004860997141804546,
|
503 |
+
"rewards/margins": 0.0004720650613307953,
|
504 |
+
"rewards/rejected": 1.4034636478754692e-05,
|
505 |
"step": 310
|
506 |
},
|
507 |
{
|
508 |
"epoch": 0.83,
|
509 |
"learning_rate": 4.229036944380912e-08,
|
510 |
+
"logits/chosen": -1.980975866317749,
|
511 |
+
"logits/rejected": -1.9686771631240845,
|
512 |
+
"logps/chosen": -34.552978515625,
|
513 |
+
"logps/rejected": -33.54259490966797,
|
514 |
+
"loss": 0.4999,
|
515 |
+
"rewards/accuracies": 0.550000011920929,
|
516 |
+
"rewards/chosen": 0.0004309539217501879,
|
517 |
+
"rewards/margins": 0.00046852690866217017,
|
518 |
+
"rewards/rejected": -3.757303420570679e-05,
|
519 |
"step": 320
|
520 |
},
|
521 |
{
|
522 |
"epoch": 0.86,
|
523 |
"learning_rate": 3.053082288996112e-08,
|
524 |
+
"logits/chosen": -2.016475200653076,
|
525 |
+
"logits/rejected": -2.015024185180664,
|
526 |
+
"logps/chosen": -33.44532012939453,
|
527 |
+
"logps/rejected": -32.47471237182617,
|
528 |
+
"loss": 0.4999,
|
529 |
+
"rewards/accuracies": 0.512499988079071,
|
530 |
+
"rewards/chosen": 0.00040476821595802903,
|
531 |
+
"rewards/margins": 0.0003560716868378222,
|
532 |
+
"rewards/rejected": 4.869652912020683e-05,
|
533 |
"step": 330
|
534 |
},
|
535 |
{
|
536 |
"epoch": 0.88,
|
537 |
"learning_rate": 2.05793773749158e-08,
|
538 |
+
"logits/chosen": -2.10372257232666,
|
539 |
+
"logits/rejected": -2.0879228115081787,
|
540 |
+
"logps/chosen": -34.1622200012207,
|
541 |
+
"logps/rejected": -33.08952331542969,
|
542 |
+
"loss": 0.5,
|
543 |
+
"rewards/accuracies": 0.4124999940395355,
|
544 |
+
"rewards/chosen": 0.00017170429055113345,
|
545 |
+
"rewards/margins": -0.00011341056961100549,
|
546 |
+
"rewards/rejected": 0.00028511486016213894,
|
547 |
"step": 340
|
548 |
},
|
549 |
{
|
550 |
"epoch": 0.91,
|
551 |
"learning_rate": 1.251801807404168e-08,
|
552 |
+
"logits/chosen": -1.9753808975219727,
|
553 |
+
"logits/rejected": -1.9744256734848022,
|
554 |
+
"logps/chosen": -33.2431755065918,
|
555 |
+
"logps/rejected": -32.457496643066406,
|
556 |
+
"loss": 0.4999,
|
557 |
+
"rewards/accuracies": 0.44999998807907104,
|
558 |
+
"rewards/chosen": 0.0003892013046424836,
|
559 |
+
"rewards/margins": 0.00031266576843336225,
|
560 |
+
"rewards/rejected": 7.653549255337566e-05,
|
561 |
"step": 350
|
562 |
},
|
563 |
{
|
564 |
"epoch": 0.94,
|
565 |
"learning_rate": 6.41315865106129e-09,
|
566 |
+
"logits/chosen": -1.9311378002166748,
|
567 |
+
"logits/rejected": -1.9414989948272705,
|
568 |
+
"logps/chosen": -32.21795654296875,
|
569 |
+
"logps/rejected": -35.26204299926758,
|
570 |
+
"loss": 0.5001,
|
571 |
+
"rewards/accuracies": 0.44999998807907104,
|
572 |
+
"rewards/chosen": -0.00013266984024085104,
|
573 |
+
"rewards/margins": -0.0003998771426267922,
|
574 |
+
"rewards/rejected": 0.00026720724417828023,
|
575 |
"step": 360
|
576 |
},
|
577 |
{
|
578 |
"epoch": 0.96,
|
579 |
"learning_rate": 2.3150941078050324e-09,
|
580 |
+
"logits/chosen": -2.0701680183410645,
|
581 |
+
"logits/rejected": -2.063605546951294,
|
582 |
+
"logps/chosen": -33.66097640991211,
|
583 |
+
"logps/rejected": -29.19500732421875,
|
584 |
+
"loss": 0.5001,
|
585 |
+
"rewards/accuracies": 0.4000000059604645,
|
586 |
+
"rewards/chosen": -9.009057976072654e-05,
|
587 |
+
"rewards/margins": -0.0004042644868604839,
|
588 |
+
"rewards/rejected": 0.0003141739289276302,
|
589 |
"step": 370
|
590 |
},
|
591 |
{
|
592 |
"epoch": 0.99,
|
593 |
"learning_rate": 2.575864278703266e-10,
|
594 |
+
"logits/chosen": -1.9299176931381226,
|
595 |
+
"logits/rejected": -1.9320815801620483,
|
596 |
+
"logps/chosen": -34.23712921142578,
|
597 |
+
"logps/rejected": -30.907180786132812,
|
598 |
+
"loss": 0.4999,
|
599 |
+
"rewards/accuracies": 0.550000011920929,
|
600 |
+
"rewards/chosen": 1.9172584870830178e-05,
|
601 |
+
"rewards/margins": 0.00036473083309829235,
|
602 |
+
"rewards/rejected": -0.0003455582191236317,
|
603 |
"step": 380
|
604 |
},
|
605 |
{
|
606 |
"epoch": 1.0,
|
607 |
"step": 385,
|
608 |
"total_flos": 0.0,
|
609 |
+
"train_loss": 0.4999783658362054,
|
610 |
+
"train_runtime": 3254.8055,
|
611 |
+
"train_samples_per_second": 0.946,
|
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 sha256:1b322081d1b31ac18d7549fcc597dd984caab9247d9431dcc16128652c9c9ae5
|
3 |
size 4984
|