Model save
Browse files- README.md +2 -2
- all_results.json +4 -4
- model-00001-of-00003.safetensors +1 -1
- model-00002-of-00003.safetensors +1 -1
- model-00003-of-00003.safetensors +1 -1
- runs/Jun21_01-05-25_n136-112-146/events.out.tfevents.1718904669.n136-112-146.2741958.0 +2 -2
- train_results.json +4 -4
- trainer_state.json +543 -288
- training_args.bin +1 -1
README.md
CHANGED
@@ -14,7 +14,7 @@ should probably proofread and complete it, then remove this comment. -->
|
|
14 |
|
15 |
# zephyr-7b-dpo-full
|
16 |
|
17 |
-
This model is a fine-tuned version of [princeton-nlp/Mistral-7B-Base-SFT-DPO](https://huggingface.co/princeton-nlp/Mistral-7B-Base-SFT-DPO) on
|
18 |
|
19 |
## Model description
|
20 |
|
@@ -55,5 +55,5 @@ The following hyperparameters were used during training:
|
|
55 |
|
56 |
- Transformers 4.39.3
|
57 |
- Pytorch 2.1.2+cu118
|
58 |
-
- Datasets 2.
|
59 |
- Tokenizers 0.15.2
|
|
|
14 |
|
15 |
# zephyr-7b-dpo-full
|
16 |
|
17 |
+
This model is a fine-tuned version of [princeton-nlp/Mistral-7B-Base-SFT-DPO](https://huggingface.co/princeton-nlp/Mistral-7B-Base-SFT-DPO) on an unknown dataset.
|
18 |
|
19 |
## Model description
|
20 |
|
|
|
55 |
|
56 |
- Transformers 4.39.3
|
57 |
- Pytorch 2.1.2+cu118
|
58 |
+
- Datasets 2.19.1
|
59 |
- Tokenizers 0.15.2
|
all_results.json
CHANGED
@@ -1,8 +1,8 @@
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
-
"train_loss": 0.
|
4 |
-
"train_runtime":
|
5 |
-
"train_samples":
|
6 |
-
"train_samples_per_second": 8.
|
7 |
"train_steps_per_second": 0.034
|
8 |
}
|
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
+
"train_loss": 0.2186826115846634,
|
4 |
+
"train_runtime": 11940.2176,
|
5 |
+
"train_samples": 102360,
|
6 |
+
"train_samples_per_second": 8.573,
|
7 |
"train_steps_per_second": 0.034
|
8 |
}
|
model-00001-of-00003.safetensors
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 4943162336
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:1a118ac948197a3bbfd249f1956387edf3828fdc2a6cc414bf8ce3c3e7a6e2e3
|
3 |
size 4943162336
|
model-00002-of-00003.safetensors
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 4999819336
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:08a70e605bdd6f68144e056258758c5577009cb50b28c0583f8d451060cb6684
|
3 |
size 4999819336
|
model-00003-of-00003.safetensors
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 4540516344
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:7fc90fce317ef012990a74b3c58b0b4875f221d5b4f150cec69f3ed0b9db0134
|
3 |
size 4540516344
|
runs/Jun21_01-05-25_n136-112-146/events.out.tfevents.1718904669.n136-112-146.2741958.0
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:f043138c85c0913a83007d89e74a3e8aa588674d22499745c700ebc6897955f7
|
3 |
+
size 33336
|
train_results.json
CHANGED
@@ -1,8 +1,8 @@
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
-
"train_loss": 0.
|
4 |
-
"train_runtime":
|
5 |
-
"train_samples":
|
6 |
-
"train_samples_per_second": 8.
|
7 |
"train_steps_per_second": 0.034
|
8 |
}
|
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
+
"train_loss": 0.2186826115846634,
|
4 |
+
"train_runtime": 11940.2176,
|
5 |
+
"train_samples": 102360,
|
6 |
+
"train_samples_per_second": 8.573,
|
7 |
"train_steps_per_second": 0.034
|
8 |
}
|
trainer_state.json
CHANGED
@@ -3,19 +3,19 @@
|
|
3 |
"best_model_checkpoint": null,
|
4 |
"epoch": 1.0,
|
5 |
"eval_steps": 500,
|
6 |
-
"global_step":
|
7 |
"is_hyper_param_search": false,
|
8 |
"is_local_process_zero": true,
|
9 |
"is_world_process_zero": true,
|
10 |
"log_history": [
|
11 |
{
|
12 |
"epoch": 0.0,
|
13 |
-
"grad_norm":
|
14 |
-
"learning_rate":
|
15 |
-
"logits/chosen": -0.
|
16 |
-
"logits/rejected": -0.
|
17 |
-
"logps/chosen": -
|
18 |
-
"logps/rejected": -
|
19 |
"loss": 0.6931,
|
20 |
"rewards/accuracies": 0.0,
|
21 |
"rewards/chosen": 0.0,
|
@@ -24,362 +24,617 @@
|
|
24 |
"step": 1
|
25 |
},
|
26 |
{
|
27 |
-
"epoch": 0.
|
28 |
-
"grad_norm":
|
29 |
-
"learning_rate":
|
30 |
-
"logits/chosen":
|
31 |
-
"logits/rejected": 0.
|
32 |
-
"logps/chosen": -
|
33 |
-
"logps/rejected": -
|
34 |
-
"loss": 0.
|
35 |
-
"rewards/accuracies": 0.
|
36 |
-
"rewards/chosen":
|
37 |
-
"rewards/margins": 0.
|
38 |
-
"rewards/rejected": -0.
|
39 |
"step": 10
|
40 |
},
|
41 |
{
|
42 |
-
"epoch": 0.
|
43 |
-
"grad_norm":
|
44 |
-
"learning_rate":
|
45 |
-
"logits/chosen":
|
46 |
-
"logits/rejected": 0.
|
47 |
-
"logps/chosen": -
|
48 |
-
"logps/rejected": -
|
49 |
-
"loss": 0.
|
50 |
-
"rewards/accuracies": 0.
|
51 |
-
"rewards/chosen":
|
52 |
-
"rewards/margins":
|
53 |
-
"rewards/rejected": -
|
54 |
"step": 20
|
55 |
},
|
56 |
{
|
57 |
-
"epoch": 0.
|
58 |
-
"grad_norm":
|
59 |
-
"learning_rate":
|
60 |
-
"logits/chosen":
|
61 |
-
"logits/rejected": 0.
|
62 |
-
"logps/chosen": -
|
63 |
-
"logps/rejected": -
|
64 |
-
"loss": 0.
|
65 |
-
"rewards/accuracies": 0.
|
66 |
-
"rewards/chosen":
|
67 |
-
"rewards/margins":
|
68 |
-
"rewards/rejected": -
|
69 |
"step": 30
|
70 |
},
|
71 |
{
|
72 |
-
"epoch": 0.
|
73 |
-
"grad_norm":
|
74 |
-
"learning_rate":
|
75 |
-
"logits/chosen": -0.
|
76 |
-
"logits/rejected": 0.
|
77 |
-
"logps/chosen": -
|
78 |
-
"logps/rejected": -
|
79 |
-
"loss": 0.
|
80 |
-
"rewards/accuracies": 0.
|
81 |
-
"rewards/chosen":
|
82 |
-
"rewards/margins": 3.
|
83 |
-
"rewards/rejected": -
|
84 |
"step": 40
|
85 |
},
|
86 |
{
|
87 |
-
"epoch": 0.
|
88 |
-
"grad_norm":
|
89 |
-
"learning_rate": 4.
|
90 |
-
"logits/chosen":
|
91 |
-
"logits/rejected": 0.
|
92 |
-
"logps/chosen": -
|
93 |
-
"logps/rejected": -
|
94 |
-
"loss": 0.
|
95 |
-
"rewards/accuracies": 0.
|
96 |
-
"rewards/chosen":
|
97 |
-
"rewards/margins":
|
98 |
-
"rewards/rejected": -
|
99 |
"step": 50
|
100 |
},
|
101 |
{
|
102 |
-
"epoch": 0.
|
103 |
-
"grad_norm":
|
104 |
-
"learning_rate": 4.
|
105 |
-
"logits/chosen":
|
106 |
-
"logits/rejected": 0.
|
107 |
-
"logps/chosen": -
|
108 |
-
"logps/rejected": -
|
109 |
-
"loss": 0.
|
110 |
-
"rewards/accuracies": 0.
|
111 |
-
"rewards/chosen":
|
112 |
-
"rewards/margins":
|
113 |
-
"rewards/rejected": -4.
|
114 |
"step": 60
|
115 |
},
|
116 |
{
|
117 |
-
"epoch": 0.
|
118 |
-
"grad_norm":
|
119 |
-
"learning_rate": 4.
|
120 |
-
"logits/chosen":
|
121 |
-
"logits/rejected": 0.
|
122 |
-
"logps/chosen": -
|
123 |
-
"logps/rejected": -
|
124 |
-
"loss": 0.
|
125 |
-
"rewards/accuracies": 0.
|
126 |
-
"rewards/chosen":
|
127 |
-
"rewards/margins":
|
128 |
-
"rewards/rejected": -
|
129 |
"step": 70
|
130 |
},
|
131 |
{
|
132 |
-
"epoch": 0.
|
133 |
-
"grad_norm":
|
134 |
-
"learning_rate": 4.
|
135 |
-
"logits/chosen":
|
136 |
-
"logits/rejected": 0.
|
137 |
-
"logps/chosen": -
|
138 |
-
"logps/rejected": -
|
139 |
-
"loss": 0.
|
140 |
-
"rewards/accuracies": 0.
|
141 |
-
"rewards/chosen": -
|
142 |
-
"rewards/margins":
|
143 |
-
"rewards/rejected": -
|
144 |
"step": 80
|
145 |
},
|
146 |
{
|
147 |
-
"epoch": 0.
|
148 |
-
"grad_norm":
|
149 |
-
"learning_rate":
|
150 |
-
"logits/chosen":
|
151 |
-
"logits/rejected": 0.
|
152 |
-
"logps/chosen": -
|
153 |
-
"logps/rejected": -
|
154 |
-
"loss": 0.
|
155 |
-
"rewards/accuracies": 0.
|
156 |
-
"rewards/chosen":
|
157 |
-
"rewards/margins":
|
158 |
-
"rewards/rejected": -
|
159 |
"step": 90
|
160 |
},
|
161 |
{
|
162 |
-
"epoch": 0.
|
163 |
-
"grad_norm":
|
164 |
-
"learning_rate":
|
165 |
-
"logits/chosen":
|
166 |
-
"logits/rejected": 0.
|
167 |
-
"logps/chosen": -
|
168 |
-
"logps/rejected": -
|
169 |
-
"loss": 0.
|
170 |
-
"rewards/accuracies": 0.
|
171 |
-
"rewards/chosen": -1.
|
172 |
-
"rewards/margins":
|
173 |
-
"rewards/rejected": -
|
174 |
"step": 100
|
175 |
},
|
176 |
{
|
177 |
-
"epoch": 0.
|
178 |
-
"grad_norm":
|
179 |
-
"learning_rate":
|
180 |
-
"logits/chosen":
|
181 |
-
"logits/rejected": 0.
|
182 |
-
"logps/chosen": -
|
183 |
-
"logps/rejected": -
|
184 |
-
"loss": 0.
|
185 |
-
"rewards/accuracies": 0.
|
186 |
-
"rewards/chosen": -
|
187 |
-
"rewards/margins":
|
188 |
-
"rewards/rejected": -
|
189 |
"step": 110
|
190 |
},
|
191 |
{
|
192 |
-
"epoch": 0.
|
193 |
-
"grad_norm":
|
194 |
-
"learning_rate":
|
195 |
-
"logits/chosen":
|
196 |
-
"logits/rejected": 0.
|
197 |
-
"logps/chosen": -
|
198 |
-
"logps/rejected": -
|
199 |
-
"loss": 0.
|
200 |
-
"rewards/accuracies": 0.
|
201 |
-
"rewards/chosen": -
|
202 |
-
"rewards/margins":
|
203 |
-
"rewards/rejected": -
|
204 |
"step": 120
|
205 |
},
|
206 |
{
|
207 |
-
"epoch": 0.
|
208 |
-
"grad_norm":
|
209 |
-
"learning_rate":
|
210 |
-
"logits/chosen":
|
211 |
-
"logits/rejected": 0.
|
212 |
-
"logps/chosen": -
|
213 |
-
"logps/rejected": -
|
214 |
-
"loss": 0.
|
215 |
-
"rewards/accuracies": 0.
|
216 |
-
"rewards/chosen": -
|
217 |
-
"rewards/margins":
|
218 |
-
"rewards/rejected": -5.
|
219 |
"step": 130
|
220 |
},
|
221 |
{
|
222 |
-
"epoch": 0.
|
223 |
-
"grad_norm":
|
224 |
-
"learning_rate":
|
225 |
-
"logits/chosen":
|
226 |
-
"logits/rejected": 0.
|
227 |
-
"logps/chosen": -
|
228 |
-
"logps/rejected": -
|
229 |
-
"loss": 0.
|
230 |
-
"rewards/accuracies": 0.
|
231 |
-
"rewards/chosen": -
|
232 |
-
"rewards/margins":
|
233 |
-
"rewards/rejected": -
|
234 |
"step": 140
|
235 |
},
|
236 |
{
|
237 |
-
"epoch": 0.
|
238 |
-
"grad_norm":
|
239 |
-
"learning_rate":
|
240 |
-
"logits/chosen": -0.
|
241 |
-
"logits/rejected": 0.
|
242 |
-
"logps/chosen": -
|
243 |
-
"logps/rejected": -
|
244 |
-
"loss": 0.
|
245 |
-
"rewards/accuracies": 0.
|
246 |
-
"rewards/chosen": -
|
247 |
-
"rewards/margins":
|
248 |
-
"rewards/rejected": -
|
249 |
"step": 150
|
250 |
},
|
251 |
{
|
252 |
-
"epoch": 0.
|
253 |
-
"grad_norm":
|
254 |
-
"learning_rate":
|
255 |
-
"logits/chosen":
|
256 |
-
"logits/rejected": 0.
|
257 |
-
"logps/chosen": -
|
258 |
-
"logps/rejected": -
|
259 |
-
"loss": 0.
|
260 |
-
"rewards/accuracies": 0.
|
261 |
-
"rewards/chosen": -
|
262 |
-
"rewards/margins":
|
263 |
-
"rewards/rejected": -
|
264 |
"step": 160
|
265 |
},
|
266 |
{
|
267 |
-
"epoch": 0.
|
268 |
-
"grad_norm":
|
269 |
-
"learning_rate":
|
270 |
-
"logits/chosen": -0.
|
271 |
-
"logits/rejected": 0.
|
272 |
-
"logps/chosen": -
|
273 |
-
"logps/rejected": -
|
274 |
-
"loss": 0.
|
275 |
-
"rewards/accuracies": 0.
|
276 |
-
"rewards/chosen": -
|
277 |
-
"rewards/margins":
|
278 |
-
"rewards/rejected": -
|
279 |
"step": 170
|
280 |
},
|
281 |
{
|
282 |
-
"epoch": 0.
|
283 |
-
"grad_norm":
|
284 |
-
"learning_rate":
|
285 |
-
"logits/chosen": -0.
|
286 |
-
"logits/rejected": 0.
|
287 |
-
"logps/chosen": -
|
288 |
-
"logps/rejected": -
|
289 |
-
"loss": 0.
|
290 |
-
"rewards/accuracies": 0.
|
291 |
-
"rewards/chosen":
|
292 |
-
"rewards/margins":
|
293 |
-
"rewards/rejected": -
|
294 |
"step": 180
|
295 |
},
|
296 |
{
|
297 |
-
"epoch": 0.
|
298 |
-
"grad_norm":
|
299 |
-
"learning_rate":
|
300 |
-
"logits/chosen":
|
301 |
-
"logits/rejected": 0.
|
302 |
-
"logps/chosen": -341.
|
303 |
-
"logps/rejected": -
|
304 |
-
"loss": 0.
|
305 |
-
"rewards/accuracies": 0.
|
306 |
-
"rewards/chosen": -
|
307 |
-
"rewards/margins":
|
308 |
-
"rewards/rejected": -
|
309 |
"step": 190
|
310 |
},
|
311 |
{
|
312 |
-
"epoch": 0.
|
313 |
-
"grad_norm":
|
314 |
-
"learning_rate":
|
315 |
-
"logits/chosen":
|
316 |
-
"logits/rejected": 0.
|
317 |
-
"logps/chosen": -
|
318 |
-
"logps/rejected": -
|
319 |
-
"loss": 0.
|
320 |
-
"rewards/accuracies": 0.
|
321 |
-
"rewards/chosen": -
|
322 |
-
"rewards/margins":
|
323 |
-
"rewards/rejected": -6.
|
324 |
"step": 200
|
325 |
},
|
326 |
{
|
327 |
-
"epoch": 0.
|
328 |
-
"grad_norm":
|
329 |
-
"learning_rate": 2.
|
330 |
-
"logits/chosen": -0.
|
331 |
-
"logits/rejected": 0.
|
332 |
-
"logps/chosen": -
|
333 |
-
"logps/rejected": -
|
334 |
-
"loss": 0.
|
335 |
-
"rewards/accuracies": 0.
|
336 |
-
"rewards/chosen":
|
337 |
-
"rewards/margins":
|
338 |
-
"rewards/rejected": -
|
339 |
"step": 210
|
340 |
},
|
341 |
{
|
342 |
-
"epoch": 0.
|
343 |
-
"grad_norm":
|
344 |
-
"learning_rate":
|
345 |
-
"logits/chosen":
|
346 |
-
"logits/rejected": 0.
|
347 |
-
"logps/chosen": -
|
348 |
-
"logps/rejected": -
|
349 |
-
"loss": 0.
|
350 |
-
"rewards/accuracies": 0.
|
351 |
-
"rewards/chosen": -
|
352 |
-
"rewards/margins":
|
353 |
-
"rewards/rejected": -5.
|
354 |
"step": 220
|
355 |
},
|
356 |
{
|
357 |
-
"epoch": 0.
|
358 |
-
"grad_norm":
|
359 |
-
"learning_rate": 2.
|
360 |
-
"logits/chosen":
|
361 |
-
"logits/rejected": 0.
|
362 |
-
"logps/chosen": -
|
363 |
-
"logps/rejected": -
|
364 |
-
"loss": 0.
|
365 |
-
"rewards/accuracies": 0.
|
366 |
-
"rewards/chosen": -
|
367 |
-
"rewards/margins":
|
368 |
-
"rewards/rejected": -
|
369 |
"step": 230
|
370 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
371 |
{
|
372 |
"epoch": 1.0,
|
373 |
-
"step":
|
374 |
"total_flos": 0.0,
|
375 |
-
"train_loss": 0.
|
376 |
-
"train_runtime":
|
377 |
-
"train_samples_per_second": 8.
|
378 |
"train_steps_per_second": 0.034
|
379 |
}
|
380 |
],
|
381 |
"logging_steps": 10,
|
382 |
-
"max_steps":
|
383 |
"num_input_tokens_seen": 0,
|
384 |
"num_train_epochs": 1,
|
385 |
"save_steps": 100,
|
|
|
3 |
"best_model_checkpoint": null,
|
4 |
"epoch": 1.0,
|
5 |
"eval_steps": 500,
|
6 |
+
"global_step": 400,
|
7 |
"is_hyper_param_search": false,
|
8 |
"is_local_process_zero": true,
|
9 |
"is_world_process_zero": true,
|
10 |
"log_history": [
|
11 |
{
|
12 |
"epoch": 0.0,
|
13 |
+
"grad_norm": 248.672711818404,
|
14 |
+
"learning_rate": 1.25e-08,
|
15 |
+
"logits/chosen": -0.5811702013015747,
|
16 |
+
"logits/rejected": -0.11655431985855103,
|
17 |
+
"logps/chosen": -351.5902099609375,
|
18 |
+
"logps/rejected": -240.969970703125,
|
19 |
"loss": 0.6931,
|
20 |
"rewards/accuracies": 0.0,
|
21 |
"rewards/chosen": 0.0,
|
|
|
24 |
"step": 1
|
25 |
},
|
26 |
{
|
27 |
+
"epoch": 0.03,
|
28 |
+
"grad_norm": 222.11982816621227,
|
29 |
+
"learning_rate": 1.25e-07,
|
30 |
+
"logits/chosen": 0.2612163722515106,
|
31 |
+
"logits/rejected": 0.2365657538175583,
|
32 |
+
"logps/chosen": -333.2138366699219,
|
33 |
+
"logps/rejected": -244.68914794921875,
|
34 |
+
"loss": 0.6848,
|
35 |
+
"rewards/accuracies": 0.4965277910232544,
|
36 |
+
"rewards/chosen": 0.018286287784576416,
|
37 |
+
"rewards/margins": 0.01924367994070053,
|
38 |
+
"rewards/rejected": -0.0009573940187692642,
|
39 |
"step": 10
|
40 |
},
|
41 |
{
|
42 |
+
"epoch": 0.05,
|
43 |
+
"grad_norm": 131.0722901505231,
|
44 |
+
"learning_rate": 2.5e-07,
|
45 |
+
"logits/chosen": 0.00122042594011873,
|
46 |
+
"logits/rejected": 0.21793019771575928,
|
47 |
+
"logps/chosen": -320.78399658203125,
|
48 |
+
"logps/rejected": -234.4599609375,
|
49 |
+
"loss": 0.5397,
|
50 |
+
"rewards/accuracies": 0.778124988079071,
|
51 |
+
"rewards/chosen": 0.3007456660270691,
|
52 |
+
"rewards/margins": 0.44470709562301636,
|
53 |
+
"rewards/rejected": -0.14396145939826965,
|
54 |
"step": 20
|
55 |
},
|
56 |
{
|
57 |
+
"epoch": 0.07,
|
58 |
+
"grad_norm": 80.67754344694362,
|
59 |
+
"learning_rate": 3.75e-07,
|
60 |
+
"logits/chosen": 0.11814385652542114,
|
61 |
+
"logits/rejected": -0.13333633542060852,
|
62 |
+
"logps/chosen": -305.92999267578125,
|
63 |
+
"logps/rejected": -253.8010711669922,
|
64 |
+
"loss": 0.3279,
|
65 |
+
"rewards/accuracies": 0.8343750238418579,
|
66 |
+
"rewards/chosen": 0.9548959732055664,
|
67 |
+
"rewards/margins": 2.203428268432617,
|
68 |
+
"rewards/rejected": -1.2485322952270508,
|
69 |
"step": 30
|
70 |
},
|
71 |
{
|
72 |
+
"epoch": 0.1,
|
73 |
+
"grad_norm": 66.3293163499791,
|
74 |
+
"learning_rate": 5e-07,
|
75 |
+
"logits/chosen": -0.011821460910141468,
|
76 |
+
"logits/rejected": -0.11223969608545303,
|
77 |
+
"logps/chosen": -307.73626708984375,
|
78 |
+
"logps/rejected": -266.0770263671875,
|
79 |
+
"loss": 0.2627,
|
80 |
+
"rewards/accuracies": 0.8812500238418579,
|
81 |
+
"rewards/chosen": 1.6875461339950562,
|
82 |
+
"rewards/margins": 3.763516664505005,
|
83 |
+
"rewards/rejected": -2.075970411300659,
|
84 |
"step": 40
|
85 |
},
|
86 |
{
|
87 |
+
"epoch": 0.12,
|
88 |
+
"grad_norm": 72.19599673796307,
|
89 |
+
"learning_rate": 4.990486745229364e-07,
|
90 |
+
"logits/chosen": 0.3851412832736969,
|
91 |
+
"logits/rejected": 0.5114809274673462,
|
92 |
+
"logps/chosen": -316.8995056152344,
|
93 |
+
"logps/rejected": -269.6571350097656,
|
94 |
+
"loss": 0.229,
|
95 |
+
"rewards/accuracies": 0.8843749761581421,
|
96 |
+
"rewards/chosen": 1.225061058998108,
|
97 |
+
"rewards/margins": 4.17338228225708,
|
98 |
+
"rewards/rejected": -2.9483208656311035,
|
99 |
"step": 50
|
100 |
},
|
101 |
{
|
102 |
+
"epoch": 0.15,
|
103 |
+
"grad_norm": 53.904395061776626,
|
104 |
+
"learning_rate": 4.96201938253052e-07,
|
105 |
+
"logits/chosen": 0.19928967952728271,
|
106 |
+
"logits/rejected": 0.3621904253959656,
|
107 |
+
"logps/chosen": -323.8656921386719,
|
108 |
+
"logps/rejected": -287.19903564453125,
|
109 |
+
"loss": 0.2394,
|
110 |
+
"rewards/accuracies": 0.8812500238418579,
|
111 |
+
"rewards/chosen": 0.07580803334712982,
|
112 |
+
"rewards/margins": 4.283209800720215,
|
113 |
+
"rewards/rejected": -4.207401752471924,
|
114 |
"step": 60
|
115 |
},
|
116 |
{
|
117 |
+
"epoch": 0.17,
|
118 |
+
"grad_norm": 70.90163796602725,
|
119 |
+
"learning_rate": 4.91481456572267e-07,
|
120 |
+
"logits/chosen": 0.33822208642959595,
|
121 |
+
"logits/rejected": 0.5376901626586914,
|
122 |
+
"logps/chosen": -337.15509033203125,
|
123 |
+
"logps/rejected": -272.99346923828125,
|
124 |
+
"loss": 0.2201,
|
125 |
+
"rewards/accuracies": 0.9156249761581421,
|
126 |
+
"rewards/chosen": 0.19220082461833954,
|
127 |
+
"rewards/margins": 4.760615348815918,
|
128 |
+
"rewards/rejected": -4.56841516494751,
|
129 |
"step": 70
|
130 |
},
|
131 |
{
|
132 |
+
"epoch": 0.2,
|
133 |
+
"grad_norm": 52.67912669957963,
|
134 |
+
"learning_rate": 4.849231551964771e-07,
|
135 |
+
"logits/chosen": 0.9565310478210449,
|
136 |
+
"logits/rejected": 0.9506447911262512,
|
137 |
+
"logps/chosen": -327.2256774902344,
|
138 |
+
"logps/rejected": -293.74700927734375,
|
139 |
+
"loss": 0.2138,
|
140 |
+
"rewards/accuracies": 0.9375,
|
141 |
+
"rewards/chosen": -1.5700352191925049,
|
142 |
+
"rewards/margins": 5.063401222229004,
|
143 |
+
"rewards/rejected": -6.633436679840088,
|
144 |
"step": 80
|
145 |
},
|
146 |
{
|
147 |
+
"epoch": 0.23,
|
148 |
+
"grad_norm": 65.22420033118601,
|
149 |
+
"learning_rate": 4.7657694675916247e-07,
|
150 |
+
"logits/chosen": 0.3500242233276367,
|
151 |
+
"logits/rejected": 0.21482405066490173,
|
152 |
+
"logps/chosen": -315.8086242675781,
|
153 |
+
"logps/rejected": -293.5943908691406,
|
154 |
+
"loss": 0.2317,
|
155 |
+
"rewards/accuracies": 0.90625,
|
156 |
+
"rewards/chosen": 0.491361141204834,
|
157 |
+
"rewards/margins": 5.032988548278809,
|
158 |
+
"rewards/rejected": -4.541626930236816,
|
159 |
"step": 90
|
160 |
},
|
161 |
{
|
162 |
+
"epoch": 0.25,
|
163 |
+
"grad_norm": 54.96225429454786,
|
164 |
+
"learning_rate": 4.6650635094610966e-07,
|
165 |
+
"logits/chosen": 0.10862906277179718,
|
166 |
+
"logits/rejected": 0.24883398413658142,
|
167 |
+
"logps/chosen": -349.65911865234375,
|
168 |
+
"logps/rejected": -326.064697265625,
|
169 |
+
"loss": 0.1952,
|
170 |
+
"rewards/accuracies": 0.893750011920929,
|
171 |
+
"rewards/chosen": -1.6933103799819946,
|
172 |
+
"rewards/margins": 4.797018051147461,
|
173 |
+
"rewards/rejected": -6.490328788757324,
|
174 |
"step": 100
|
175 |
},
|
176 |
{
|
177 |
+
"epoch": 0.28,
|
178 |
+
"grad_norm": 52.005066827882644,
|
179 |
+
"learning_rate": 4.5478801107224794e-07,
|
180 |
+
"logits/chosen": 0.11174388229846954,
|
181 |
+
"logits/rejected": 0.27005332708358765,
|
182 |
+
"logps/chosen": -326.76959228515625,
|
183 |
+
"logps/rejected": -285.99151611328125,
|
184 |
+
"loss": 0.1986,
|
185 |
+
"rewards/accuracies": 0.934374988079071,
|
186 |
+
"rewards/chosen": -0.6088568568229675,
|
187 |
+
"rewards/margins": 5.2384033203125,
|
188 |
+
"rewards/rejected": -5.847259521484375,
|
189 |
"step": 110
|
190 |
},
|
191 |
{
|
192 |
+
"epoch": 0.3,
|
193 |
+
"grad_norm": 80.2735393844808,
|
194 |
+
"learning_rate": 4.415111107797445e-07,
|
195 |
+
"logits/chosen": 0.12390895932912827,
|
196 |
+
"logits/rejected": 0.2671768069267273,
|
197 |
+
"logps/chosen": -314.864013671875,
|
198 |
+
"logps/rejected": -296.4173278808594,
|
199 |
+
"loss": 0.2022,
|
200 |
+
"rewards/accuracies": 0.890625,
|
201 |
+
"rewards/chosen": -0.602489709854126,
|
202 |
+
"rewards/margins": 5.041877746582031,
|
203 |
+
"rewards/rejected": -5.644367694854736,
|
204 |
"step": 120
|
205 |
},
|
206 |
{
|
207 |
+
"epoch": 0.33,
|
208 |
+
"grad_norm": 53.04554797674892,
|
209 |
+
"learning_rate": 4.2677669529663686e-07,
|
210 |
+
"logits/chosen": 0.2591269612312317,
|
211 |
+
"logits/rejected": 0.31469181180000305,
|
212 |
+
"logps/chosen": -337.55279541015625,
|
213 |
+
"logps/rejected": -303.430908203125,
|
214 |
+
"loss": 0.1929,
|
215 |
+
"rewards/accuracies": 0.8999999761581421,
|
216 |
+
"rewards/chosen": -0.5142609477043152,
|
217 |
+
"rewards/margins": 5.346030235290527,
|
218 |
+
"rewards/rejected": -5.86029052734375,
|
219 |
"step": 130
|
220 |
},
|
221 |
{
|
222 |
+
"epoch": 0.35,
|
223 |
+
"grad_norm": 46.55373068779065,
|
224 |
+
"learning_rate": 4.106969024216348e-07,
|
225 |
+
"logits/chosen": 0.5539799928665161,
|
226 |
+
"logits/rejected": 0.3750854432582855,
|
227 |
+
"logps/chosen": -323.2265625,
|
228 |
+
"logps/rejected": -283.9201354980469,
|
229 |
+
"loss": 0.1948,
|
230 |
+
"rewards/accuracies": 0.934374988079071,
|
231 |
+
"rewards/chosen": -0.46524372696876526,
|
232 |
+
"rewards/margins": 5.173353672027588,
|
233 |
+
"rewards/rejected": -5.63859748840332,
|
234 |
"step": 140
|
235 |
},
|
236 |
{
|
237 |
+
"epoch": 0.38,
|
238 |
+
"grad_norm": 57.84664559821586,
|
239 |
+
"learning_rate": 3.933941090877615e-07,
|
240 |
+
"logits/chosen": -0.04622136428952217,
|
241 |
+
"logits/rejected": 0.10014678537845612,
|
242 |
+
"logps/chosen": -330.5162353515625,
|
243 |
+
"logps/rejected": -301.1103515625,
|
244 |
+
"loss": 0.2027,
|
245 |
+
"rewards/accuracies": 0.8999999761581421,
|
246 |
+
"rewards/chosen": -0.5420152544975281,
|
247 |
+
"rewards/margins": 4.931308746337891,
|
248 |
+
"rewards/rejected": -5.473323822021484,
|
249 |
"step": 150
|
250 |
},
|
251 |
{
|
252 |
+
"epoch": 0.4,
|
253 |
+
"grad_norm": 90.75814646308999,
|
254 |
+
"learning_rate": 3.75e-07,
|
255 |
+
"logits/chosen": 0.11180607229471207,
|
256 |
+
"logits/rejected": -0.048911936581134796,
|
257 |
+
"logps/chosen": -295.4419860839844,
|
258 |
+
"logps/rejected": -279.9349670410156,
|
259 |
+
"loss": 0.1841,
|
260 |
+
"rewards/accuracies": 0.921875,
|
261 |
+
"rewards/chosen": -0.33810311555862427,
|
262 |
+
"rewards/margins": 4.924574375152588,
|
263 |
+
"rewards/rejected": -5.262677192687988,
|
264 |
"step": 160
|
265 |
},
|
266 |
{
|
267 |
+
"epoch": 0.42,
|
268 |
+
"grad_norm": 48.25868450687608,
|
269 |
+
"learning_rate": 3.5565456543517485e-07,
|
270 |
+
"logits/chosen": -0.2750697731971741,
|
271 |
+
"logits/rejected": 0.026401836425065994,
|
272 |
+
"logps/chosen": -324.1236572265625,
|
273 |
+
"logps/rejected": -290.73785400390625,
|
274 |
+
"loss": 0.1938,
|
275 |
+
"rewards/accuracies": 0.887499988079071,
|
276 |
+
"rewards/chosen": -0.8598020672798157,
|
277 |
+
"rewards/margins": 5.191195487976074,
|
278 |
+
"rewards/rejected": -6.050997257232666,
|
279 |
"step": 170
|
280 |
},
|
281 |
{
|
282 |
+
"epoch": 0.45,
|
283 |
+
"grad_norm": 55.57273036697954,
|
284 |
+
"learning_rate": 3.355050358314172e-07,
|
285 |
+
"logits/chosen": -0.17330403625965118,
|
286 |
+
"logits/rejected": -0.37704771757125854,
|
287 |
+
"logps/chosen": -335.1918029785156,
|
288 |
+
"logps/rejected": -303.3643493652344,
|
289 |
+
"loss": 0.1893,
|
290 |
+
"rewards/accuracies": 0.940625011920929,
|
291 |
+
"rewards/chosen": 0.3236238360404968,
|
292 |
+
"rewards/margins": 5.686396598815918,
|
293 |
+
"rewards/rejected": -5.3627729415893555,
|
294 |
"step": 180
|
295 |
},
|
296 |
{
|
297 |
+
"epoch": 0.47,
|
298 |
+
"grad_norm": 34.4619211960604,
|
299 |
+
"learning_rate": 3.147047612756302e-07,
|
300 |
+
"logits/chosen": 0.051471106708049774,
|
301 |
+
"logits/rejected": 0.08085541427135468,
|
302 |
+
"logps/chosen": -341.36932373046875,
|
303 |
+
"logps/rejected": -306.84906005859375,
|
304 |
+
"loss": 0.1714,
|
305 |
+
"rewards/accuracies": 0.9312499761581421,
|
306 |
+
"rewards/chosen": -0.4273417592048645,
|
307 |
+
"rewards/margins": 5.435866355895996,
|
308 |
+
"rewards/rejected": -5.863208293914795,
|
309 |
"step": 190
|
310 |
},
|
311 |
{
|
312 |
+
"epoch": 0.5,
|
313 |
+
"grad_norm": 56.25034316469711,
|
314 |
+
"learning_rate": 2.934120444167326e-07,
|
315 |
+
"logits/chosen": 0.2834271490573883,
|
316 |
+
"logits/rejected": 0.4803285002708435,
|
317 |
+
"logps/chosen": -320.8459777832031,
|
318 |
+
"logps/rejected": -293.8531799316406,
|
319 |
+
"loss": 0.1797,
|
320 |
+
"rewards/accuracies": 0.9125000238418579,
|
321 |
+
"rewards/chosen": -1.590319275856018,
|
322 |
+
"rewards/margins": 5.1105732917785645,
|
323 |
+
"rewards/rejected": -6.700892448425293,
|
324 |
"step": 200
|
325 |
},
|
326 |
{
|
327 |
+
"epoch": 0.53,
|
328 |
+
"grad_norm": 42.616613118118565,
|
329 |
+
"learning_rate": 2.717889356869146e-07,
|
330 |
+
"logits/chosen": -0.1653163880109787,
|
331 |
+
"logits/rejected": -0.19997279345989227,
|
332 |
+
"logps/chosen": -297.6966857910156,
|
333 |
+
"logps/rejected": -281.0061950683594,
|
334 |
+
"loss": 0.1905,
|
335 |
+
"rewards/accuracies": 0.921875,
|
336 |
+
"rewards/chosen": 0.32219845056533813,
|
337 |
+
"rewards/margins": 5.068012237548828,
|
338 |
+
"rewards/rejected": -4.745813846588135,
|
339 |
"step": 210
|
340 |
},
|
341 |
{
|
342 |
+
"epoch": 0.55,
|
343 |
+
"grad_norm": 54.94298294558815,
|
344 |
+
"learning_rate": 2.5e-07,
|
345 |
+
"logits/chosen": 0.11906982958316803,
|
346 |
+
"logits/rejected": 0.16367843747138977,
|
347 |
+
"logps/chosen": -322.9801330566406,
|
348 |
+
"logps/rejected": -289.7718811035156,
|
349 |
+
"loss": 0.2379,
|
350 |
+
"rewards/accuracies": 0.9312499761581421,
|
351 |
+
"rewards/chosen": -0.8957729339599609,
|
352 |
+
"rewards/margins": 4.5600104331970215,
|
353 |
+
"rewards/rejected": -5.455783367156982,
|
354 |
"step": 220
|
355 |
},
|
356 |
{
|
357 |
+
"epoch": 0.57,
|
358 |
+
"grad_norm": 45.798677109719435,
|
359 |
+
"learning_rate": 2.2821106431308543e-07,
|
360 |
+
"logits/chosen": 0.26459816098213196,
|
361 |
+
"logits/rejected": 0.37918907403945923,
|
362 |
+
"logps/chosen": -308.4639892578125,
|
363 |
+
"logps/rejected": -284.7757873535156,
|
364 |
+
"loss": 0.1974,
|
365 |
+
"rewards/accuracies": 0.90625,
|
366 |
+
"rewards/chosen": -0.7637487649917603,
|
367 |
+
"rewards/margins": 5.488121032714844,
|
368 |
+
"rewards/rejected": -6.251870155334473,
|
369 |
"step": 230
|
370 |
},
|
371 |
+
{
|
372 |
+
"epoch": 0.6,
|
373 |
+
"grad_norm": 71.171077628863,
|
374 |
+
"learning_rate": 2.065879555832674e-07,
|
375 |
+
"logits/chosen": 0.11944101750850677,
|
376 |
+
"logits/rejected": 0.10174547135829926,
|
377 |
+
"logps/chosen": -317.2291564941406,
|
378 |
+
"logps/rejected": -301.72357177734375,
|
379 |
+
"loss": 0.1761,
|
380 |
+
"rewards/accuracies": 0.9281250238418579,
|
381 |
+
"rewards/chosen": -0.4725034832954407,
|
382 |
+
"rewards/margins": 5.045371055603027,
|
383 |
+
"rewards/rejected": -5.517874717712402,
|
384 |
+
"step": 240
|
385 |
+
},
|
386 |
+
{
|
387 |
+
"epoch": 0.62,
|
388 |
+
"grad_norm": 58.77603788283333,
|
389 |
+
"learning_rate": 1.8529523872436977e-07,
|
390 |
+
"logits/chosen": -0.0030237496830523014,
|
391 |
+
"logits/rejected": 0.24394559860229492,
|
392 |
+
"logps/chosen": -337.0401306152344,
|
393 |
+
"logps/rejected": -304.41485595703125,
|
394 |
+
"loss": 0.2199,
|
395 |
+
"rewards/accuracies": 0.90625,
|
396 |
+
"rewards/chosen": -1.1832717657089233,
|
397 |
+
"rewards/margins": 4.3705525398254395,
|
398 |
+
"rewards/rejected": -5.553823947906494,
|
399 |
+
"step": 250
|
400 |
+
},
|
401 |
+
{
|
402 |
+
"epoch": 0.65,
|
403 |
+
"grad_norm": 58.549851578658085,
|
404 |
+
"learning_rate": 1.6449496416858282e-07,
|
405 |
+
"logits/chosen": 0.2534635663032532,
|
406 |
+
"logits/rejected": 0.33454760909080505,
|
407 |
+
"logps/chosen": -322.7724609375,
|
408 |
+
"logps/rejected": -291.37420654296875,
|
409 |
+
"loss": 0.1828,
|
410 |
+
"rewards/accuracies": 0.9437500238418579,
|
411 |
+
"rewards/chosen": -1.3282363414764404,
|
412 |
+
"rewards/margins": 5.129273414611816,
|
413 |
+
"rewards/rejected": -6.457509517669678,
|
414 |
+
"step": 260
|
415 |
+
},
|
416 |
+
{
|
417 |
+
"epoch": 0.68,
|
418 |
+
"grad_norm": 61.69689329910415,
|
419 |
+
"learning_rate": 1.4434543456482518e-07,
|
420 |
+
"logits/chosen": 0.16627629101276398,
|
421 |
+
"logits/rejected": 0.17112889885902405,
|
422 |
+
"logps/chosen": -328.3280944824219,
|
423 |
+
"logps/rejected": -302.532958984375,
|
424 |
+
"loss": 0.1701,
|
425 |
+
"rewards/accuracies": 0.918749988079071,
|
426 |
+
"rewards/chosen": -0.8465593457221985,
|
427 |
+
"rewards/margins": 5.068365573883057,
|
428 |
+
"rewards/rejected": -5.914924621582031,
|
429 |
+
"step": 270
|
430 |
+
},
|
431 |
+
{
|
432 |
+
"epoch": 0.7,
|
433 |
+
"grad_norm": 82.28572950186891,
|
434 |
+
"learning_rate": 1.2500000000000005e-07,
|
435 |
+
"logits/chosen": 0.07045526802539825,
|
436 |
+
"logits/rejected": 0.055424489080905914,
|
437 |
+
"logps/chosen": -314.9088134765625,
|
438 |
+
"logps/rejected": -283.484130859375,
|
439 |
+
"loss": 0.1988,
|
440 |
+
"rewards/accuracies": 0.934374988079071,
|
441 |
+
"rewards/chosen": -0.05264568328857422,
|
442 |
+
"rewards/margins": 5.250518798828125,
|
443 |
+
"rewards/rejected": -5.303164958953857,
|
444 |
+
"step": 280
|
445 |
+
},
|
446 |
+
{
|
447 |
+
"epoch": 0.72,
|
448 |
+
"grad_norm": 73.90506249227809,
|
449 |
+
"learning_rate": 1.0660589091223854e-07,
|
450 |
+
"logits/chosen": 0.08644680678844452,
|
451 |
+
"logits/rejected": 0.3770992159843445,
|
452 |
+
"logps/chosen": -334.16839599609375,
|
453 |
+
"logps/rejected": -295.61138916015625,
|
454 |
+
"loss": 0.1816,
|
455 |
+
"rewards/accuracies": 0.9312499761581421,
|
456 |
+
"rewards/chosen": -0.6267386674880981,
|
457 |
+
"rewards/margins": 5.2511420249938965,
|
458 |
+
"rewards/rejected": -5.877881050109863,
|
459 |
+
"step": 290
|
460 |
+
},
|
461 |
+
{
|
462 |
+
"epoch": 0.75,
|
463 |
+
"grad_norm": 55.979754279546064,
|
464 |
+
"learning_rate": 8.930309757836516e-08,
|
465 |
+
"logits/chosen": 0.0685787945985794,
|
466 |
+
"logits/rejected": 0.4188007414340973,
|
467 |
+
"logps/chosen": -330.8103332519531,
|
468 |
+
"logps/rejected": -297.2096252441406,
|
469 |
+
"loss": 0.1831,
|
470 |
+
"rewards/accuracies": 0.8999999761581421,
|
471 |
+
"rewards/chosen": -0.8838691711425781,
|
472 |
+
"rewards/margins": 4.987154960632324,
|
473 |
+
"rewards/rejected": -5.871024131774902,
|
474 |
+
"step": 300
|
475 |
+
},
|
476 |
+
{
|
477 |
+
"epoch": 0.78,
|
478 |
+
"grad_norm": 50.80261213037778,
|
479 |
+
"learning_rate": 7.322330470336313e-08,
|
480 |
+
"logits/chosen": 0.22002212703227997,
|
481 |
+
"logits/rejected": 0.2094310224056244,
|
482 |
+
"logps/chosen": -321.89739990234375,
|
483 |
+
"logps/rejected": -294.4070739746094,
|
484 |
+
"loss": 0.1741,
|
485 |
+
"rewards/accuracies": 0.918749988079071,
|
486 |
+
"rewards/chosen": -0.5131534337997437,
|
487 |
+
"rewards/margins": 5.067374229431152,
|
488 |
+
"rewards/rejected": -5.580528259277344,
|
489 |
+
"step": 310
|
490 |
+
},
|
491 |
+
{
|
492 |
+
"epoch": 0.8,
|
493 |
+
"grad_norm": 62.71796095386463,
|
494 |
+
"learning_rate": 5.848888922025552e-08,
|
495 |
+
"logits/chosen": 0.0758807510137558,
|
496 |
+
"logits/rejected": 0.19501206278800964,
|
497 |
+
"logps/chosen": -337.13995361328125,
|
498 |
+
"logps/rejected": -296.71954345703125,
|
499 |
+
"loss": 0.1841,
|
500 |
+
"rewards/accuracies": 0.953125,
|
501 |
+
"rewards/chosen": -0.4793934226036072,
|
502 |
+
"rewards/margins": 5.182127952575684,
|
503 |
+
"rewards/rejected": -5.661521911621094,
|
504 |
+
"step": 320
|
505 |
+
},
|
506 |
+
{
|
507 |
+
"epoch": 0.82,
|
508 |
+
"grad_norm": 55.862146356475115,
|
509 |
+
"learning_rate": 4.521198892775202e-08,
|
510 |
+
"logits/chosen": 0.08274734020233154,
|
511 |
+
"logits/rejected": -0.0012732266914099455,
|
512 |
+
"logps/chosen": -333.7369079589844,
|
513 |
+
"logps/rejected": -296.9526062011719,
|
514 |
+
"loss": 0.1768,
|
515 |
+
"rewards/accuracies": 0.925000011920929,
|
516 |
+
"rewards/chosen": -0.4573189616203308,
|
517 |
+
"rewards/margins": 5.402385711669922,
|
518 |
+
"rewards/rejected": -5.859705448150635,
|
519 |
+
"step": 330
|
520 |
+
},
|
521 |
+
{
|
522 |
+
"epoch": 0.85,
|
523 |
+
"grad_norm": 78.6552872883194,
|
524 |
+
"learning_rate": 3.349364905389032e-08,
|
525 |
+
"logits/chosen": -0.16344629228115082,
|
526 |
+
"logits/rejected": 0.0033723146189004183,
|
527 |
+
"logps/chosen": -322.8833312988281,
|
528 |
+
"logps/rejected": -298.59588623046875,
|
529 |
+
"loss": 0.1682,
|
530 |
+
"rewards/accuracies": 0.8999999761581421,
|
531 |
+
"rewards/chosen": -0.6776003837585449,
|
532 |
+
"rewards/margins": 5.152725696563721,
|
533 |
+
"rewards/rejected": -5.830325126647949,
|
534 |
+
"step": 340
|
535 |
+
},
|
536 |
+
{
|
537 |
+
"epoch": 0.88,
|
538 |
+
"grad_norm": 57.94825784387087,
|
539 |
+
"learning_rate": 2.3423053240837514e-08,
|
540 |
+
"logits/chosen": 0.06374481320381165,
|
541 |
+
"logits/rejected": 0.13793043792247772,
|
542 |
+
"logps/chosen": -319.5912780761719,
|
543 |
+
"logps/rejected": -297.6089172363281,
|
544 |
+
"loss": 0.1912,
|
545 |
+
"rewards/accuracies": 0.903124988079071,
|
546 |
+
"rewards/chosen": -0.5590634346008301,
|
547 |
+
"rewards/margins": 5.25864315032959,
|
548 |
+
"rewards/rejected": -5.81770658493042,
|
549 |
+
"step": 350
|
550 |
+
},
|
551 |
+
{
|
552 |
+
"epoch": 0.9,
|
553 |
+
"grad_norm": 48.20871737390797,
|
554 |
+
"learning_rate": 1.507684480352292e-08,
|
555 |
+
"logits/chosen": 0.1408473700284958,
|
556 |
+
"logits/rejected": 0.24622826278209686,
|
557 |
+
"logps/chosen": -323.8467712402344,
|
558 |
+
"logps/rejected": -301.0394592285156,
|
559 |
+
"loss": 0.1714,
|
560 |
+
"rewards/accuracies": 0.9312499761581421,
|
561 |
+
"rewards/chosen": -0.5114877820014954,
|
562 |
+
"rewards/margins": 5.479299068450928,
|
563 |
+
"rewards/rejected": -5.990786552429199,
|
564 |
+
"step": 360
|
565 |
+
},
|
566 |
+
{
|
567 |
+
"epoch": 0.93,
|
568 |
+
"grad_norm": 66.71686397462808,
|
569 |
+
"learning_rate": 8.518543427732949e-09,
|
570 |
+
"logits/chosen": -0.14612798392772675,
|
571 |
+
"logits/rejected": -0.002347910311073065,
|
572 |
+
"logps/chosen": -305.1243896484375,
|
573 |
+
"logps/rejected": -282.12835693359375,
|
574 |
+
"loss": 0.1673,
|
575 |
+
"rewards/accuracies": 0.921875,
|
576 |
+
"rewards/chosen": -0.6524958610534668,
|
577 |
+
"rewards/margins": 5.196988105773926,
|
578 |
+
"rewards/rejected": -5.849484443664551,
|
579 |
+
"step": 370
|
580 |
+
},
|
581 |
+
{
|
582 |
+
"epoch": 0.95,
|
583 |
+
"grad_norm": 61.291842343672215,
|
584 |
+
"learning_rate": 3.798061746947995e-09,
|
585 |
+
"logits/chosen": 0.044129520654678345,
|
586 |
+
"logits/rejected": 0.14980553090572357,
|
587 |
+
"logps/chosen": -319.0976257324219,
|
588 |
+
"logps/rejected": -283.139892578125,
|
589 |
+
"loss": 0.1714,
|
590 |
+
"rewards/accuracies": 0.940625011920929,
|
591 |
+
"rewards/chosen": -0.5317240953445435,
|
592 |
+
"rewards/margins": 5.177973747253418,
|
593 |
+
"rewards/rejected": -5.70969820022583,
|
594 |
+
"step": 380
|
595 |
+
},
|
596 |
+
{
|
597 |
+
"epoch": 0.97,
|
598 |
+
"grad_norm": 56.153629064135075,
|
599 |
+
"learning_rate": 9.513254770636137e-10,
|
600 |
+
"logits/chosen": -0.012590537779033184,
|
601 |
+
"logits/rejected": 0.09102629870176315,
|
602 |
+
"logps/chosen": -334.525634765625,
|
603 |
+
"logps/rejected": -287.55987548828125,
|
604 |
+
"loss": 0.1797,
|
605 |
+
"rewards/accuracies": 0.925000011920929,
|
606 |
+
"rewards/chosen": -0.5071713328361511,
|
607 |
+
"rewards/margins": 5.1908040046691895,
|
608 |
+
"rewards/rejected": -5.697975158691406,
|
609 |
+
"step": 390
|
610 |
+
},
|
611 |
+
{
|
612 |
+
"epoch": 1.0,
|
613 |
+
"grad_norm": 40.78979544232789,
|
614 |
+
"learning_rate": 0.0,
|
615 |
+
"logits/chosen": 0.030362462624907494,
|
616 |
+
"logits/rejected": 0.237002894282341,
|
617 |
+
"logps/chosen": -320.5105895996094,
|
618 |
+
"logps/rejected": -285.7564697265625,
|
619 |
+
"loss": 0.1703,
|
620 |
+
"rewards/accuracies": 0.9156249761581421,
|
621 |
+
"rewards/chosen": -0.8242006301879883,
|
622 |
+
"rewards/margins": 5.20082950592041,
|
623 |
+
"rewards/rejected": -6.025030136108398,
|
624 |
+
"step": 400
|
625 |
+
},
|
626 |
{
|
627 |
"epoch": 1.0,
|
628 |
+
"step": 400,
|
629 |
"total_flos": 0.0,
|
630 |
+
"train_loss": 0.2186826115846634,
|
631 |
+
"train_runtime": 11940.2176,
|
632 |
+
"train_samples_per_second": 8.573,
|
633 |
"train_steps_per_second": 0.034
|
634 |
}
|
635 |
],
|
636 |
"logging_steps": 10,
|
637 |
+
"max_steps": 400,
|
638 |
"num_input_tokens_seen": 0,
|
639 |
"num_train_epochs": 1,
|
640 |
"save_steps": 100,
|
training_args.bin
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 6328
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:b6f555e8b191e5dee3bafecffb3a88f44ea1db78545e88e62c9e7b6e0969daad
|
3 |
size 6328
|