Model save
Browse files- README.md +2 -3
- all_results.json +5 -5
- config.json +1 -1
- model-00001-of-00003.safetensors +1 -1
- model-00002-of-00003.safetensors +1 -1
- model-00003-of-00003.safetensors +1 -1
- runs/Jul03_20-00-02_n136-129-074/events.out.tfevents.1720008022.n136-129-074.3083115.0 +2 -2
- tokenizer.json +0 -1
- train_results.json +5 -5
- trainer_state.json +362 -572
- training_args.bin +2 -2
README.md
CHANGED
@@ -1,5 +1,4 @@
|
|
1 |
---
|
2 |
-
base_model: princeton-nlp/Mistral-7B-Base-SFT-SimPO
|
3 |
tags:
|
4 |
- trl
|
5 |
- dpo
|
@@ -14,7 +13,7 @@ should probably proofread and complete it, then remove this comment. -->
|
|
14 |
|
15 |
# zephyr-7b-dpo-full
|
16 |
|
17 |
-
This model
|
18 |
|
19 |
## Model description
|
20 |
|
@@ -33,7 +32,7 @@ More information needed
|
|
33 |
### Training hyperparameters
|
34 |
|
35 |
The following hyperparameters were used during training:
|
36 |
-
- learning_rate: 1e-
|
37 |
- train_batch_size: 4
|
38 |
- eval_batch_size: 4
|
39 |
- seed: 42
|
|
|
1 |
---
|
|
|
2 |
tags:
|
3 |
- trl
|
4 |
- dpo
|
|
|
13 |
|
14 |
# zephyr-7b-dpo-full
|
15 |
|
16 |
+
This model was trained from scratch on the None dataset.
|
17 |
|
18 |
## Model description
|
19 |
|
|
|
32 |
### Training hyperparameters
|
33 |
|
34 |
The following hyperparameters were used during training:
|
35 |
+
- learning_rate: 1e-08
|
36 |
- train_batch_size: 4
|
37 |
- eval_batch_size: 4
|
38 |
- seed: 42
|
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.
|
8 |
}
|
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
+
"train_loss": 0.6880149315033469,
|
4 |
+
"train_runtime": 4603.368,
|
5 |
+
"train_samples": 38288,
|
6 |
+
"train_samples_per_second": 8.317,
|
7 |
+
"train_steps_per_second": 0.065
|
8 |
}
|
config.json
CHANGED
@@ -20,7 +20,7 @@
|
|
20 |
"sliding_window": 4096,
|
21 |
"tie_word_embeddings": false,
|
22 |
"torch_dtype": "bfloat16",
|
23 |
-
"transformers_version": "4.
|
24 |
"use_cache": false,
|
25 |
"vocab_size": 32000
|
26 |
}
|
|
|
20 |
"sliding_window": 4096,
|
21 |
"tie_word_embeddings": false,
|
22 |
"torch_dtype": "bfloat16",
|
23 |
+
"transformers_version": "4.39.3",
|
24 |
"use_cache": false,
|
25 |
"vocab_size": 32000
|
26 |
}
|
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 sha256:25a506977023960c048f2933f9aec9fc01ac4cef4294a89c59fb94311d62d0ba
|
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:f36b418c6f0204d63a482c4d40cf9a099d1c71913006174de004f3468b2600ac
|
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 sha256:0098c8b3746d7ba8d54ad9a79ed49c5d66268bbf89e84d00df21498d52d9cb42
|
3 |
size 4540516344
|
runs/Jul03_20-00-02_n136-129-074/events.out.tfevents.1720008022.n136-129-074.3083115.0
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 sha256:5c476ad368f70bb551a6766cd2d3c0fa0a42b8470e8b61a505a8c02a979fe20e
|
3 |
+
size 25843
|
tokenizer.json
CHANGED
@@ -134,7 +134,6 @@
|
|
134 |
"end_of_word_suffix": null,
|
135 |
"fuse_unk": true,
|
136 |
"byte_fallback": true,
|
137 |
-
"ignore_merges": false,
|
138 |
"vocab": {
|
139 |
"<unk>": 0,
|
140 |
"<s>": 1,
|
|
|
134 |
"end_of_word_suffix": null,
|
135 |
"fuse_unk": true,
|
136 |
"byte_fallback": true,
|
|
|
137 |
"vocab": {
|
138 |
"<unk>": 0,
|
139 |
"<s>": 1,
|
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.
|
8 |
}
|
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
+
"train_loss": 0.6880149315033469,
|
4 |
+
"train_runtime": 4603.368,
|
5 |
+
"train_samples": 38288,
|
6 |
+
"train_samples_per_second": 8.317,
|
7 |
+
"train_steps_per_second": 0.065
|
8 |
}
|
trainer_state.json
CHANGED
@@ -1,22 +1,22 @@
|
|
1 |
{
|
2 |
"best_metric": null,
|
3 |
"best_model_checkpoint": null,
|
4 |
-
"epoch": 0.
|
5 |
"eval_steps": 10000000,
|
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": -1.
|
16 |
-
"logits/rejected": -1.
|
17 |
-
"logps/chosen": -1.
|
18 |
-
"logps/rejected": -
|
19 |
-
"loss": 0.
|
20 |
"rewards/accuracies": 0.0,
|
21 |
"rewards/chosen": 0.0,
|
22 |
"rewards/margins": 0.0,
|
@@ -24,662 +24,452 @@
|
|
24 |
"step": 1
|
25 |
},
|
26 |
{
|
27 |
-
"epoch": 0.
|
28 |
-
"grad_norm":
|
29 |
-
"learning_rate":
|
30 |
-
"logits/chosen": -1.
|
31 |
-
"logits/rejected": -1.
|
32 |
-
"logps/chosen": -1.
|
33 |
-
"logps/rejected": -1.
|
34 |
"loss": 0.6933,
|
35 |
-
"rewards/accuracies": 0.
|
36 |
-
"rewards/chosen": 0.
|
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": -1.
|
46 |
-
"logits/rejected": -1.
|
47 |
-
"logps/chosen": -1.
|
48 |
-
"logps/rejected": -1.
|
49 |
-
"loss": 0.
|
50 |
-
"rewards/accuracies": 0.
|
51 |
-
"rewards/chosen":
|
52 |
-
"rewards/margins": 0.
|
53 |
-
"rewards/rejected":
|
54 |
"step": 20
|
55 |
},
|
56 |
{
|
57 |
-
"epoch": 0.
|
58 |
-
"grad_norm":
|
59 |
-
"learning_rate":
|
60 |
-
"logits/chosen": -1.
|
61 |
-
"logits/rejected": -1.
|
62 |
-
"logps/chosen": -1.
|
63 |
-
"logps/rejected": -1.
|
64 |
-
"loss": 0.
|
65 |
-
"rewards/accuracies": 0.
|
66 |
-
"rewards/chosen":
|
67 |
-
"rewards/margins": 0.
|
68 |
-
"rewards/rejected": -0.
|
69 |
"step": 30
|
70 |
},
|
71 |
{
|
72 |
-
"epoch": 0.
|
73 |
-
"grad_norm":
|
74 |
-
"learning_rate": 9.
|
75 |
-
"logits/chosen": -1.
|
76 |
-
"logits/rejected": -1.
|
77 |
-
"logps/chosen": -1.
|
78 |
-
"logps/rejected": -1.
|
79 |
-
"loss": 0.
|
80 |
-
"rewards/accuracies": 0.
|
81 |
-
"rewards/chosen":
|
82 |
-
"rewards/margins": 0.
|
83 |
-
"rewards/rejected": -0.
|
84 |
"step": 40
|
85 |
},
|
86 |
{
|
87 |
-
"epoch": 0.
|
88 |
-
"grad_norm":
|
89 |
-
"learning_rate": 9.
|
90 |
-
"logits/chosen": -1.
|
91 |
-
"logits/rejected": -1.
|
92 |
-
"logps/chosen": -1.
|
93 |
-
"logps/rejected": -1.
|
94 |
-
"loss": 0.
|
95 |
-
"rewards/accuracies": 0.
|
96 |
-
"rewards/chosen":
|
97 |
-
"rewards/margins": 0.
|
98 |
-
"rewards/rejected": -0.
|
99 |
"step": 50
|
100 |
},
|
101 |
{
|
102 |
-
"epoch": 0.
|
103 |
-
"grad_norm":
|
104 |
-
"learning_rate": 9.
|
105 |
-
"logits/chosen": -1.
|
106 |
-
"logits/rejected": -1.
|
107 |
-
"logps/chosen": -1.
|
108 |
-
"logps/rejected": -1.
|
109 |
-
"loss": 0.
|
110 |
-
"rewards/accuracies": 0.
|
111 |
-
"rewards/chosen":
|
112 |
-
"rewards/margins": 0.
|
113 |
-
"rewards/rejected": -0.
|
114 |
"step": 60
|
115 |
},
|
116 |
{
|
117 |
-
"epoch": 0.
|
118 |
-
"grad_norm":
|
119 |
-
"learning_rate": 9.
|
120 |
-
"logits/chosen": -1.
|
121 |
-
"logits/rejected": -1.
|
122 |
-
"logps/chosen": -1.
|
123 |
-
"logps/rejected": -1.
|
124 |
-
"loss": 0.
|
125 |
-
"rewards/accuracies": 0.
|
126 |
-
"rewards/chosen":
|
127 |
-
"rewards/margins": 0.
|
128 |
-
"rewards/rejected":
|
129 |
"step": 70
|
130 |
},
|
131 |
{
|
132 |
-
"epoch": 0.
|
133 |
-
"grad_norm":
|
134 |
-
"learning_rate": 9.
|
135 |
-
"logits/chosen": -1.
|
136 |
-
"logits/rejected": -1.
|
137 |
-
"logps/chosen": -1.
|
138 |
-
"logps/rejected": -1.
|
139 |
-
"loss": 0.
|
140 |
-
"rewards/accuracies": 0.
|
141 |
-
"rewards/chosen":
|
142 |
-
"rewards/margins": 0.
|
143 |
-
"rewards/rejected": -
|
144 |
"step": 80
|
145 |
},
|
146 |
{
|
147 |
-
"epoch": 0.
|
148 |
-
"grad_norm":
|
149 |
-
"learning_rate":
|
150 |
-
"logits/chosen": -1.
|
151 |
-
"logits/rejected": -1.
|
152 |
-
"logps/chosen": -1.
|
153 |
-
"logps/rejected": -1.
|
154 |
-
"loss": 0.
|
155 |
-
"rewards/accuracies": 0.
|
156 |
-
"rewards/chosen":
|
157 |
-
"rewards/margins": 0.
|
158 |
-
"rewards/rejected":
|
159 |
"step": 90
|
160 |
},
|
161 |
{
|
162 |
-
"epoch": 0.
|
163 |
-
"grad_norm":
|
164 |
-
"learning_rate":
|
165 |
-
"logits/chosen": -1.
|
166 |
-
"logits/rejected": -1.
|
167 |
-
"logps/chosen": -1.
|
168 |
-
"logps/rejected": -1.
|
169 |
-
"loss": 0.
|
170 |
-
"rewards/accuracies": 0.
|
171 |
-
"rewards/chosen":
|
172 |
-
"rewards/margins": 0.
|
173 |
-
"rewards/rejected": -
|
174 |
"step": 100
|
175 |
},
|
176 |
{
|
177 |
-
"epoch": 0.
|
178 |
-
"grad_norm":
|
179 |
-
"learning_rate":
|
180 |
-
"logits/chosen": -1.
|
181 |
-
"logits/rejected": -1.
|
182 |
-
"logps/chosen": -1.
|
183 |
-
"logps/rejected": -1.
|
184 |
-
"loss": 0.
|
185 |
-
"rewards/accuracies": 0.
|
186 |
-
"rewards/chosen": -
|
187 |
-
"rewards/margins": 0.
|
188 |
-
"rewards/rejected": -
|
189 |
"step": 110
|
190 |
},
|
191 |
{
|
192 |
-
"epoch": 0.
|
193 |
-
"grad_norm":
|
194 |
-
"learning_rate":
|
195 |
-
"logits/chosen": -1.
|
196 |
-
"logits/rejected": -1.
|
197 |
-
"logps/chosen": -1.
|
198 |
-
"logps/rejected": -1.
|
199 |
-
"loss": 0.
|
200 |
-
"rewards/accuracies": 0.
|
201 |
-
"rewards/chosen":
|
202 |
-
"rewards/margins": 0.
|
203 |
-
"rewards/rejected": -
|
204 |
"step": 120
|
205 |
},
|
206 |
{
|
207 |
-
"epoch": 0.
|
208 |
-
"grad_norm":
|
209 |
-
"learning_rate":
|
210 |
-
"logits/chosen": -1.
|
211 |
-
"logits/rejected": -1.
|
212 |
-
"logps/chosen": -1.
|
213 |
-
"logps/rejected": -1.
|
214 |
-
"loss": 0.
|
215 |
-
"rewards/accuracies": 0.
|
216 |
-
"rewards/chosen":
|
217 |
-
"rewards/margins": 0.
|
218 |
-
"rewards/rejected": -
|
219 |
"step": 130
|
220 |
},
|
221 |
{
|
222 |
-
"epoch": 0.
|
223 |
-
"grad_norm":
|
224 |
-
"learning_rate":
|
225 |
-
"logits/chosen": -1.
|
226 |
-
"logits/rejected": -1.
|
227 |
-
"logps/chosen": -1.
|
228 |
-
"logps/rejected": -1.
|
229 |
-
"loss": 0.
|
230 |
-
"rewards/accuracies": 0.
|
231 |
-
"rewards/chosen":
|
232 |
-
"rewards/margins": 0.
|
233 |
-
"rewards/rejected":
|
234 |
"step": 140
|
235 |
},
|
236 |
{
|
237 |
-
"epoch": 0.
|
238 |
-
"grad_norm":
|
239 |
-
"learning_rate":
|
240 |
-
"logits/chosen": -1.
|
241 |
-
"logits/rejected": -1.
|
242 |
-
"logps/chosen": -1.
|
243 |
-
"logps/rejected": -1.
|
244 |
-
"loss": 0.
|
245 |
-
"rewards/accuracies": 0.
|
246 |
-
"rewards/chosen":
|
247 |
-
"rewards/margins": 0.
|
248 |
-
"rewards/rejected": -
|
249 |
"step": 150
|
250 |
},
|
251 |
{
|
252 |
-
"epoch": 0.
|
253 |
-
"grad_norm":
|
254 |
-
"learning_rate":
|
255 |
-
"logits/chosen": -1.
|
256 |
-
"logits/rejected": -1.
|
257 |
-
"logps/chosen": -1.
|
258 |
-
"logps/rejected": -1.
|
259 |
-
"loss": 0.
|
260 |
-
"rewards/accuracies": 0.
|
261 |
-
"rewards/chosen":
|
262 |
-
"rewards/margins": 0.
|
263 |
-
"rewards/rejected": -
|
264 |
"step": 160
|
265 |
},
|
266 |
{
|
267 |
-
"epoch": 0.
|
268 |
-
"grad_norm":
|
269 |
-
"learning_rate":
|
270 |
-
"logits/chosen": -1.
|
271 |
-
"logits/rejected": -1.
|
272 |
-
"logps/chosen": -1.
|
273 |
-
"logps/rejected": -1.
|
274 |
-
"loss": 0.
|
275 |
-
"rewards/accuracies": 0.
|
276 |
-
"rewards/chosen":
|
277 |
-
"rewards/margins": 0.
|
278 |
-
"rewards/rejected":
|
279 |
"step": 170
|
280 |
},
|
281 |
{
|
282 |
-
"epoch": 0.
|
283 |
-
"grad_norm":
|
284 |
-
"learning_rate":
|
285 |
-
"logits/chosen": -1.
|
286 |
-
"logits/rejected": -1.
|
287 |
-
"logps/chosen": -1.
|
288 |
-
"logps/rejected": -1.
|
289 |
-
"loss": 0.
|
290 |
-
"rewards/accuracies": 0.
|
291 |
-
"rewards/chosen":
|
292 |
-
"rewards/margins": 0.
|
293 |
-
"rewards/rejected": -
|
294 |
"step": 180
|
295 |
},
|
296 |
{
|
297 |
-
"epoch": 0.
|
298 |
-
"grad_norm":
|
299 |
-
"learning_rate":
|
300 |
-
"logits/chosen": -1.
|
301 |
-
"logits/rejected": -1.
|
302 |
-
"logps/chosen": -1.
|
303 |
-
"logps/rejected": -1.
|
304 |
-
"loss": 0.
|
305 |
-
"rewards/accuracies": 0.
|
306 |
-
"rewards/chosen":
|
307 |
-
"rewards/margins": 0.
|
308 |
-
"rewards/rejected":
|
309 |
"step": 190
|
310 |
},
|
311 |
{
|
312 |
-
"epoch": 0.
|
313 |
-
"grad_norm": 103.
|
314 |
-
"learning_rate":
|
315 |
-
"logits/chosen": -1.
|
316 |
-
"logits/rejected": -1.
|
317 |
-
"logps/chosen": -1.
|
318 |
-
"logps/rejected": -1.
|
319 |
-
"loss": 0.
|
320 |
-
"rewards/accuracies": 0.
|
321 |
-
"rewards/chosen":
|
322 |
-
"rewards/margins": 0.
|
323 |
-
"rewards/rejected": -
|
324 |
"step": 200
|
325 |
},
|
326 |
{
|
327 |
-
"epoch": 0.
|
328 |
-
"grad_norm":
|
329 |
-
"learning_rate":
|
330 |
-
"logits/chosen": -1.
|
331 |
-
"logits/rejected": -1.
|
332 |
-
"logps/chosen": -1.
|
333 |
-
"logps/rejected": -1.
|
334 |
-
"loss": 0.
|
335 |
-
"rewards/accuracies": 0.
|
336 |
-
"rewards/chosen":
|
337 |
-
"rewards/margins": 0.
|
338 |
-
"rewards/rejected":
|
339 |
"step": 210
|
340 |
},
|
341 |
{
|
342 |
-
"epoch": 0.
|
343 |
-
"grad_norm":
|
344 |
-
"learning_rate":
|
345 |
-
"logits/chosen": -1.
|
346 |
-
"logits/rejected": -1.
|
347 |
-
"logps/chosen": -1.
|
348 |
-
"logps/rejected": -1.
|
349 |
-
"loss": 0.
|
350 |
-
"rewards/accuracies": 0.
|
351 |
-
"rewards/chosen":
|
352 |
-
"rewards/margins": 0.
|
353 |
-
"rewards/rejected": -
|
354 |
"step": 220
|
355 |
},
|
356 |
{
|
357 |
-
"epoch": 0.
|
358 |
-
"grad_norm":
|
359 |
-
"learning_rate":
|
360 |
-
"logits/chosen": -1.
|
361 |
-
"logits/rejected": -1.
|
362 |
-
"logps/chosen": -1.
|
363 |
-
"logps/rejected": -
|
364 |
-
"loss": 0.
|
365 |
-
"rewards/accuracies": 0.
|
366 |
-
"rewards/chosen":
|
367 |
-
"rewards/margins": 0.
|
368 |
-
"rewards/rejected":
|
369 |
"step": 230
|
370 |
},
|
371 |
{
|
372 |
-
"epoch": 0.
|
373 |
-
"grad_norm":
|
374 |
-
"learning_rate":
|
375 |
-
"logits/chosen": -1.
|
376 |
-
"logits/rejected": -1.
|
377 |
-
"logps/chosen": -1.
|
378 |
-
"logps/rejected": -1.
|
379 |
-
"loss": 0.
|
380 |
-
"rewards/accuracies": 0.
|
381 |
-
"rewards/chosen":
|
382 |
-
"rewards/margins": 0.
|
383 |
-
"rewards/rejected": -
|
384 |
"step": 240
|
385 |
},
|
386 |
{
|
387 |
-
"epoch": 0.
|
388 |
-
"grad_norm":
|
389 |
-
"learning_rate":
|
390 |
-
"logits/chosen": -1.
|
391 |
-
"logits/rejected": -1.
|
392 |
-
"logps/chosen": -1.
|
393 |
-
"logps/rejected": -1.
|
394 |
-
"loss": 0.
|
395 |
-
"rewards/accuracies": 0.
|
396 |
-
"rewards/chosen":
|
397 |
-
"rewards/margins": 0.
|
398 |
-
"rewards/rejected":
|
399 |
"step": 250
|
400 |
},
|
401 |
{
|
402 |
-
"epoch": 0.
|
403 |
-
"grad_norm":
|
404 |
-
"learning_rate":
|
405 |
-
"logits/chosen": -1.
|
406 |
-
"logits/rejected": -1.
|
407 |
-
"logps/chosen": -1.
|
408 |
-
"logps/rejected": -1.
|
409 |
-
"loss": 0.
|
410 |
-
"rewards/accuracies": 0.
|
411 |
-
"rewards/chosen":
|
412 |
-
"rewards/margins": 0.
|
413 |
-
"rewards/rejected":
|
414 |
"step": 260
|
415 |
},
|
416 |
{
|
417 |
-
"epoch": 0.
|
418 |
-
"grad_norm":
|
419 |
-
"learning_rate":
|
420 |
-
"logits/chosen": -1.
|
421 |
-
"logits/rejected": -1.
|
422 |
-
"logps/chosen": -
|
423 |
-
"logps/rejected": -1.
|
424 |
-
"loss": 0.
|
425 |
-
"rewards/accuracies": 0.
|
426 |
-
"rewards/chosen":
|
427 |
-
"rewards/margins": 0.
|
428 |
-
"rewards/rejected": -
|
429 |
"step": 270
|
430 |
},
|
431 |
{
|
432 |
-
"epoch": 0.
|
433 |
-
"grad_norm":
|
434 |
-
"learning_rate":
|
435 |
-
"logits/chosen": -1.
|
436 |
-
"logits/rejected": -1.
|
437 |
-
"logps/chosen": -1.
|
438 |
-
"logps/rejected": -1.
|
439 |
-
"loss": 0.
|
440 |
-
"rewards/accuracies": 0.
|
441 |
-
"rewards/chosen":
|
442 |
-
"rewards/margins": 0.
|
443 |
-
"rewards/rejected": -
|
444 |
"step": 280
|
445 |
},
|
446 |
{
|
447 |
-
"epoch": 0.
|
448 |
-
"grad_norm":
|
449 |
-
"learning_rate":
|
450 |
-
"logits/chosen": -1.
|
451 |
-
"logits/rejected": -1.
|
452 |
-
"logps/chosen": -1.
|
453 |
-
"logps/rejected": -1.
|
454 |
-
"loss": 0.
|
455 |
-
"rewards/accuracies": 0.
|
456 |
-
"rewards/chosen":
|
457 |
-
"rewards/margins": 0.
|
458 |
-
"rewards/rejected":
|
459 |
"step": 290
|
460 |
},
|
461 |
-
{
|
462 |
-
"epoch": 0.68,
|
463 |
-
"grad_norm": 87.86932907500237,
|
464 |
-
"learning_rate": 2.7566622175067443e-08,
|
465 |
-
"logits/chosen": -1.7801201343536377,
|
466 |
-
"logits/rejected": -1.7111116647720337,
|
467 |
-
"logps/chosen": -1.3444006443023682,
|
468 |
-
"logps/rejected": -1.494860053062439,
|
469 |
-
"loss": 0.5308,
|
470 |
-
"rewards/accuracies": 0.7124999761581421,
|
471 |
-
"rewards/chosen": -1.3581712245941162,
|
472 |
-
"rewards/margins": 0.7247291803359985,
|
473 |
-
"rewards/rejected": -2.082900285720825,
|
474 |
-
"step": 300
|
475 |
-
},
|
476 |
-
{
|
477 |
-
"epoch": 0.71,
|
478 |
-
"grad_norm": 101.39189225311264,
|
479 |
-
"learning_rate": 2.408731201945432e-08,
|
480 |
-
"logits/chosen": -1.77533757686615,
|
481 |
-
"logits/rejected": -1.7264738082885742,
|
482 |
-
"logps/chosen": -1.313521146774292,
|
483 |
-
"logps/rejected": -1.4012606143951416,
|
484 |
-
"loss": 0.5179,
|
485 |
-
"rewards/accuracies": 0.768750011920929,
|
486 |
-
"rewards/chosen": -1.1962579488754272,
|
487 |
-
"rewards/margins": 0.6758478283882141,
|
488 |
-
"rewards/rejected": -1.8721058368682861,
|
489 |
-
"step": 310
|
490 |
-
},
|
491 |
-
{
|
492 |
-
"epoch": 0.73,
|
493 |
-
"grad_norm": 97.55308734787654,
|
494 |
-
"learning_rate": 2.0771830220378112e-08,
|
495 |
-
"logits/chosen": -1.7309653759002686,
|
496 |
-
"logits/rejected": -1.673135757446289,
|
497 |
-
"logps/chosen": -1.3468959331512451,
|
498 |
-
"logps/rejected": -1.405505895614624,
|
499 |
-
"loss": 0.5257,
|
500 |
-
"rewards/accuracies": 0.7437499761581421,
|
501 |
-
"rewards/chosen": -1.240114450454712,
|
502 |
-
"rewards/margins": 0.6646026968955994,
|
503 |
-
"rewards/rejected": -1.9047170877456665,
|
504 |
-
"step": 320
|
505 |
-
},
|
506 |
-
{
|
507 |
-
"epoch": 0.75,
|
508 |
-
"grad_norm": 79.56679512217426,
|
509 |
-
"learning_rate": 1.7641138321260257e-08,
|
510 |
-
"logits/chosen": -1.7662233114242554,
|
511 |
-
"logits/rejected": -1.6911147832870483,
|
512 |
-
"logps/chosen": -1.3011645078659058,
|
513 |
-
"logps/rejected": -1.418474555015564,
|
514 |
-
"loss": 0.5175,
|
515 |
-
"rewards/accuracies": 0.8125,
|
516 |
-
"rewards/chosen": -1.1730973720550537,
|
517 |
-
"rewards/margins": 0.8977500796318054,
|
518 |
-
"rewards/rejected": -2.070847511291504,
|
519 |
-
"step": 330
|
520 |
-
},
|
521 |
-
{
|
522 |
-
"epoch": 0.77,
|
523 |
-
"grad_norm": 78.95443991883164,
|
524 |
-
"learning_rate": 1.4715029564277793e-08,
|
525 |
-
"logits/chosen": -1.8377494812011719,
|
526 |
-
"logits/rejected": -1.785035490989685,
|
527 |
-
"logps/chosen": -1.3052855730056763,
|
528 |
-
"logps/rejected": -1.4216973781585693,
|
529 |
-
"loss": 0.535,
|
530 |
-
"rewards/accuracies": 0.7562500238418579,
|
531 |
-
"rewards/chosen": -1.0062506198883057,
|
532 |
-
"rewards/margins": 0.7934825420379639,
|
533 |
-
"rewards/rejected": -1.7997331619262695,
|
534 |
-
"step": 340
|
535 |
-
},
|
536 |
-
{
|
537 |
-
"epoch": 0.8,
|
538 |
-
"grad_norm": 85.7404253023672,
|
539 |
-
"learning_rate": 1.2012003751113343e-08,
|
540 |
-
"logits/chosen": -1.8215789794921875,
|
541 |
-
"logits/rejected": -1.7556202411651611,
|
542 |
-
"logps/chosen": -1.2944856882095337,
|
543 |
-
"logps/rejected": -1.427002191543579,
|
544 |
-
"loss": 0.5037,
|
545 |
-
"rewards/accuracies": 0.768750011920929,
|
546 |
-
"rewards/chosen": -1.3231576681137085,
|
547 |
-
"rewards/margins": 0.8232595324516296,
|
548 |
-
"rewards/rejected": -2.1464171409606934,
|
549 |
-
"step": 350
|
550 |
-
},
|
551 |
-
{
|
552 |
-
"epoch": 0.82,
|
553 |
-
"grad_norm": 81.80824877374742,
|
554 |
-
"learning_rate": 9.549150281252633e-09,
|
555 |
-
"logits/chosen": -1.7837785482406616,
|
556 |
-
"logits/rejected": -1.7289537191390991,
|
557 |
-
"logps/chosen": -1.2906091213226318,
|
558 |
-
"logps/rejected": -1.4211328029632568,
|
559 |
-
"loss": 0.5085,
|
560 |
-
"rewards/accuracies": 0.731249988079071,
|
561 |
-
"rewards/chosen": -1.1847398281097412,
|
562 |
-
"rewards/margins": 0.7879296541213989,
|
563 |
-
"rewards/rejected": -1.9726696014404297,
|
564 |
-
"step": 360
|
565 |
-
},
|
566 |
-
{
|
567 |
-
"epoch": 0.84,
|
568 |
-
"grad_norm": 76.45542191386993,
|
569 |
-
"learning_rate": 7.3420401072985306e-09,
|
570 |
-
"logits/chosen": -1.8155953884124756,
|
571 |
-
"logits/rejected": -1.7616589069366455,
|
572 |
-
"logps/chosen": -1.3416122198104858,
|
573 |
-
"logps/rejected": -1.4825923442840576,
|
574 |
-
"loss": 0.512,
|
575 |
-
"rewards/accuracies": 0.731249988079071,
|
576 |
-
"rewards/chosen": -1.2699860334396362,
|
577 |
-
"rewards/margins": 0.8351644277572632,
|
578 |
-
"rewards/rejected": -2.1051506996154785,
|
579 |
-
"step": 370
|
580 |
-
},
|
581 |
-
{
|
582 |
-
"epoch": 0.86,
|
583 |
-
"grad_norm": 100.61500636796463,
|
584 |
-
"learning_rate": 5.404627290395369e-09,
|
585 |
-
"logits/chosen": -1.7809003591537476,
|
586 |
-
"logits/rejected": -1.7117547988891602,
|
587 |
-
"logps/chosen": -1.299863338470459,
|
588 |
-
"logps/rejected": -1.4180821180343628,
|
589 |
-
"loss": 0.5139,
|
590 |
-
"rewards/accuracies": 0.762499988079071,
|
591 |
-
"rewards/chosen": -1.1025749444961548,
|
592 |
-
"rewards/margins": 0.7822272777557373,
|
593 |
-
"rewards/rejected": -1.884802222251892,
|
594 |
-
"step": 380
|
595 |
-
},
|
596 |
-
{
|
597 |
-
"epoch": 0.89,
|
598 |
-
"grad_norm": 143.00077069598655,
|
599 |
-
"learning_rate": 3.74916077816162e-09,
|
600 |
-
"logits/chosen": -1.7901942729949951,
|
601 |
-
"logits/rejected": -1.7254260778427124,
|
602 |
-
"logps/chosen": -1.2824052572250366,
|
603 |
-
"logps/rejected": -1.3659498691558838,
|
604 |
-
"loss": 0.5278,
|
605 |
-
"rewards/accuracies": 0.7124999761581421,
|
606 |
-
"rewards/chosen": -1.326103925704956,
|
607 |
-
"rewards/margins": 0.6741579174995422,
|
608 |
-
"rewards/rejected": -2.0002617835998535,
|
609 |
-
"step": 390
|
610 |
-
},
|
611 |
-
{
|
612 |
-
"epoch": 0.91,
|
613 |
-
"grad_norm": 86.78461219124101,
|
614 |
-
"learning_rate": 2.386106962899165e-09,
|
615 |
-
"logits/chosen": -1.7145483493804932,
|
616 |
-
"logits/rejected": -1.6393749713897705,
|
617 |
-
"logps/chosen": -1.3581172227859497,
|
618 |
-
"logps/rejected": -1.4866178035736084,
|
619 |
-
"loss": 0.5223,
|
620 |
-
"rewards/accuracies": 0.7124999761581421,
|
621 |
-
"rewards/chosen": -1.3169643878936768,
|
622 |
-
"rewards/margins": 0.7216087579727173,
|
623 |
-
"rewards/rejected": -2.0385732650756836,
|
624 |
-
"step": 400
|
625 |
-
},
|
626 |
-
{
|
627 |
-
"epoch": 0.93,
|
628 |
-
"grad_norm": 97.03178172617403,
|
629 |
-
"learning_rate": 1.3240835096913706e-09,
|
630 |
-
"logits/chosen": -1.7489525079727173,
|
631 |
-
"logits/rejected": -1.6416590213775635,
|
632 |
-
"logps/chosen": -1.2802146673202515,
|
633 |
-
"logps/rejected": -1.4399700164794922,
|
634 |
-
"loss": 0.5334,
|
635 |
-
"rewards/accuracies": 0.737500011920929,
|
636 |
-
"rewards/chosen": -1.086639404296875,
|
637 |
-
"rewards/margins": 0.9034107327461243,
|
638 |
-
"rewards/rejected": -1.9900500774383545,
|
639 |
-
"step": 410
|
640 |
-
},
|
641 |
-
{
|
642 |
-
"epoch": 0.96,
|
643 |
-
"grad_norm": 177.7844461853396,
|
644 |
-
"learning_rate": 5.698048727497462e-10,
|
645 |
-
"logits/chosen": -1.7718130350112915,
|
646 |
-
"logits/rejected": -1.6983397006988525,
|
647 |
-
"logps/chosen": -1.3178266286849976,
|
648 |
-
"logps/rejected": -1.4721015691757202,
|
649 |
-
"loss": 0.5219,
|
650 |
-
"rewards/accuracies": 0.7875000238418579,
|
651 |
-
"rewards/chosen": -1.2652876377105713,
|
652 |
-
"rewards/margins": 0.8731800317764282,
|
653 |
-
"rewards/rejected": -2.138467788696289,
|
654 |
-
"step": 420
|
655 |
-
},
|
656 |
-
{
|
657 |
-
"epoch": 0.98,
|
658 |
-
"grad_norm": 191.4766363479098,
|
659 |
-
"learning_rate": 1.2803984447259387e-10,
|
660 |
-
"logits/chosen": -1.7818502187728882,
|
661 |
-
"logits/rejected": -1.721131682395935,
|
662 |
-
"logps/chosen": -1.3105113506317139,
|
663 |
-
"logps/rejected": -1.4450818300247192,
|
664 |
-
"loss": 0.5113,
|
665 |
-
"rewards/accuracies": 0.7250000238418579,
|
666 |
-
"rewards/chosen": -1.2817203998565674,
|
667 |
-
"rewards/margins": 0.869175910949707,
|
668 |
-
"rewards/rejected": -2.1508963108062744,
|
669 |
-
"step": 430
|
670 |
-
},
|
671 |
{
|
672 |
"epoch": 1.0,
|
673 |
-
"step":
|
674 |
"total_flos": 0.0,
|
675 |
-
"train_loss": 0.
|
676 |
-
"train_runtime":
|
677 |
-
"train_samples_per_second": 8.
|
678 |
-
"train_steps_per_second": 0.
|
679 |
}
|
680 |
],
|
681 |
"logging_steps": 10,
|
682 |
-
"max_steps":
|
683 |
"num_input_tokens_seen": 0,
|
684 |
"num_train_epochs": 1,
|
685 |
"save_steps": 100,
|
|
|
1 |
{
|
2 |
"best_metric": null,
|
3 |
"best_model_checkpoint": null,
|
4 |
+
"epoch": 0.9991645781119465,
|
5 |
"eval_steps": 10000000,
|
6 |
+
"global_step": 299,
|
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": 94.51455806402136,
|
14 |
+
"learning_rate": 3.333333333333333e-10,
|
15 |
+
"logits/chosen": -1.693521499633789,
|
16 |
+
"logits/rejected": -1.6753541231155396,
|
17 |
+
"logps/chosen": -1.041430115699768,
|
18 |
+
"logps/rejected": -0.9273841977119446,
|
19 |
+
"loss": 0.6931,
|
20 |
"rewards/accuracies": 0.0,
|
21 |
"rewards/chosen": 0.0,
|
22 |
"rewards/margins": 0.0,
|
|
|
24 |
"step": 1
|
25 |
},
|
26 |
{
|
27 |
+
"epoch": 0.03,
|
28 |
+
"grad_norm": 100.10048800952612,
|
29 |
+
"learning_rate": 3.3333333333333334e-09,
|
30 |
+
"logits/chosen": -1.7426362037658691,
|
31 |
+
"logits/rejected": -1.7463488578796387,
|
32 |
+
"logps/chosen": -1.0522818565368652,
|
33 |
+
"logps/rejected": -1.0174607038497925,
|
34 |
"loss": 0.6933,
|
35 |
+
"rewards/accuracies": 0.4722222089767456,
|
36 |
+
"rewards/chosen": -0.0006895202095620334,
|
37 |
+
"rewards/margins": -0.0014206942869350314,
|
38 |
+
"rewards/rejected": 0.0007311741355806589,
|
39 |
"step": 10
|
40 |
},
|
41 |
{
|
42 |
+
"epoch": 0.07,
|
43 |
+
"grad_norm": 100.2747717498339,
|
44 |
+
"learning_rate": 6.666666666666667e-09,
|
45 |
+
"logits/chosen": -1.876529335975647,
|
46 |
+
"logits/rejected": -1.8286478519439697,
|
47 |
+
"logps/chosen": -1.0717421770095825,
|
48 |
+
"logps/rejected": -1.0434354543685913,
|
49 |
+
"loss": 0.6934,
|
50 |
+
"rewards/accuracies": 0.53125,
|
51 |
+
"rewards/chosen": 0.0013453494757413864,
|
52 |
+
"rewards/margins": 0.001054912805557251,
|
53 |
+
"rewards/rejected": 0.00029043667018413544,
|
54 |
"step": 20
|
55 |
},
|
56 |
{
|
57 |
+
"epoch": 0.1,
|
58 |
+
"grad_norm": 91.78075458726622,
|
59 |
+
"learning_rate": 1e-08,
|
60 |
+
"logits/chosen": -1.7869594097137451,
|
61 |
+
"logits/rejected": -1.7599306106567383,
|
62 |
+
"logps/chosen": -1.066955327987671,
|
63 |
+
"logps/rejected": -1.0448154211044312,
|
64 |
+
"loss": 0.6926,
|
65 |
+
"rewards/accuracies": 0.5562499761581421,
|
66 |
+
"rewards/chosen": 0.0009901206940412521,
|
67 |
+
"rewards/margins": 0.005651786923408508,
|
68 |
+
"rewards/rejected": -0.004661666229367256,
|
69 |
"step": 30
|
70 |
},
|
71 |
{
|
72 |
+
"epoch": 0.13,
|
73 |
+
"grad_norm": 83.1668176080734,
|
74 |
+
"learning_rate": 9.96594024562513e-09,
|
75 |
+
"logits/chosen": -1.8331626653671265,
|
76 |
+
"logits/rejected": -1.8235479593276978,
|
77 |
+
"logps/chosen": -1.069947361946106,
|
78 |
+
"logps/rejected": -1.0194944143295288,
|
79 |
+
"loss": 0.6933,
|
80 |
+
"rewards/accuracies": 0.5625,
|
81 |
+
"rewards/chosen": 0.00406806543469429,
|
82 |
+
"rewards/margins": 0.006294439546763897,
|
83 |
+
"rewards/rejected": -0.0022263741120696068,
|
84 |
"step": 40
|
85 |
},
|
86 |
{
|
87 |
+
"epoch": 0.17,
|
88 |
+
"grad_norm": 97.3484905394127,
|
89 |
+
"learning_rate": 9.86422500924775e-09,
|
90 |
+
"logits/chosen": -1.8598533868789673,
|
91 |
+
"logits/rejected": -1.8249794244766235,
|
92 |
+
"logps/chosen": -1.0844773054122925,
|
93 |
+
"logps/rejected": -1.0546021461486816,
|
94 |
+
"loss": 0.6928,
|
95 |
+
"rewards/accuracies": 0.5,
|
96 |
+
"rewards/chosen": 0.0010452494025230408,
|
97 |
+
"rewards/margins": 0.003718096762895584,
|
98 |
+
"rewards/rejected": -0.0026728473603725433,
|
99 |
"step": 50
|
100 |
},
|
101 |
{
|
102 |
+
"epoch": 0.2,
|
103 |
+
"grad_norm": 89.15270505321199,
|
104 |
+
"learning_rate": 9.696240049254743e-09,
|
105 |
+
"logits/chosen": -1.8162918090820312,
|
106 |
+
"logits/rejected": -1.8095808029174805,
|
107 |
+
"logps/chosen": -1.1147956848144531,
|
108 |
+
"logps/rejected": -1.10584557056427,
|
109 |
+
"loss": 0.692,
|
110 |
+
"rewards/accuracies": 0.574999988079071,
|
111 |
+
"rewards/chosen": 0.003025440499186516,
|
112 |
+
"rewards/margins": 0.006887376308441162,
|
113 |
+
"rewards/rejected": -0.0038619358092546463,
|
114 |
"step": 60
|
115 |
},
|
116 |
{
|
117 |
+
"epoch": 0.23,
|
118 |
+
"grad_norm": 92.66102585547152,
|
119 |
+
"learning_rate": 9.464273976236516e-09,
|
120 |
+
"logits/chosen": -1.7991975545883179,
|
121 |
+
"logits/rejected": -1.7613614797592163,
|
122 |
+
"logps/chosen": -1.0598714351654053,
|
123 |
+
"logps/rejected": -1.0530080795288086,
|
124 |
+
"loss": 0.6927,
|
125 |
+
"rewards/accuracies": 0.5562499761581421,
|
126 |
+
"rewards/chosen": 0.0042045507580041885,
|
127 |
+
"rewards/margins": 0.0019652042537927628,
|
128 |
+
"rewards/rejected": 0.002239346969872713,
|
129 |
"step": 70
|
130 |
},
|
131 |
{
|
132 |
+
"epoch": 0.27,
|
133 |
+
"grad_norm": 194.03821825764928,
|
134 |
+
"learning_rate": 9.171487073181198e-09,
|
135 |
+
"logits/chosen": -1.8440383672714233,
|
136 |
+
"logits/rejected": -1.8225574493408203,
|
137 |
+
"logps/chosen": -1.0219953060150146,
|
138 |
+
"logps/rejected": -1.0044893026351929,
|
139 |
+
"loss": 0.6914,
|
140 |
+
"rewards/accuracies": 0.625,
|
141 |
+
"rewards/chosen": 0.0038851741701364517,
|
142 |
+
"rewards/margins": 0.009898573160171509,
|
143 |
+
"rewards/rejected": -0.006013398990035057,
|
144 |
"step": 80
|
145 |
},
|
146 |
{
|
147 |
+
"epoch": 0.3,
|
148 |
+
"grad_norm": 89.9314461991181,
|
149 |
+
"learning_rate": 8.821868240089676e-09,
|
150 |
+
"logits/chosen": -1.8029550313949585,
|
151 |
+
"logits/rejected": -1.7841434478759766,
|
152 |
+
"logps/chosen": -1.045693039894104,
|
153 |
+
"logps/rejected": -1.0079419612884521,
|
154 |
+
"loss": 0.6912,
|
155 |
+
"rewards/accuracies": 0.48124998807907104,
|
156 |
+
"rewards/chosen": 0.005337424576282501,
|
157 |
+
"rewards/margins": 0.004554492421448231,
|
158 |
+
"rewards/rejected": 0.0007829321548342705,
|
159 |
"step": 90
|
160 |
},
|
161 |
{
|
162 |
+
"epoch": 0.33,
|
163 |
+
"grad_norm": 100.68694545007727,
|
164 |
+
"learning_rate": 8.42018064959393e-09,
|
165 |
+
"logits/chosen": -1.8531291484832764,
|
166 |
+
"logits/rejected": -1.8357194662094116,
|
167 |
+
"logps/chosen": -1.078748345375061,
|
168 |
+
"logps/rejected": -1.0813863277435303,
|
169 |
+
"loss": 0.6903,
|
170 |
+
"rewards/accuracies": 0.606249988079071,
|
171 |
+
"rewards/chosen": 0.007160305976867676,
|
172 |
+
"rewards/margins": 0.007577991578727961,
|
173 |
+
"rewards/rejected": -0.0004176851361989975,
|
174 |
"step": 100
|
175 |
},
|
176 |
{
|
177 |
+
"epoch": 0.37,
|
178 |
+
"grad_norm": 91.99753152902545,
|
179 |
+
"learning_rate": 7.971896853961043e-09,
|
180 |
+
"logits/chosen": -1.8346866369247437,
|
181 |
+
"logits/rejected": -1.8052761554718018,
|
182 |
+
"logps/chosen": -1.0582704544067383,
|
183 |
+
"logps/rejected": -1.0159928798675537,
|
184 |
+
"loss": 0.6901,
|
185 |
+
"rewards/accuracies": 0.550000011920929,
|
186 |
+
"rewards/chosen": -0.0007529235444962978,
|
187 |
+
"rewards/margins": 0.0036198147572577,
|
188 |
+
"rewards/rejected": -0.004372738301753998,
|
189 |
"step": 110
|
190 |
},
|
191 |
{
|
192 |
+
"epoch": 0.4,
|
193 |
+
"grad_norm": 76.03854240063603,
|
194 |
+
"learning_rate": 7.48312422757881e-09,
|
195 |
+
"logits/chosen": -1.8829187154769897,
|
196 |
+
"logits/rejected": -1.8501228094100952,
|
197 |
+
"logps/chosen": -1.0292680263519287,
|
198 |
+
"logps/rejected": -1.019852876663208,
|
199 |
+
"loss": 0.6887,
|
200 |
+
"rewards/accuracies": 0.6875,
|
201 |
+
"rewards/chosen": 0.00865122489631176,
|
202 |
+
"rewards/margins": 0.011260826140642166,
|
203 |
+
"rewards/rejected": -0.0026096017099916935,
|
204 |
"step": 120
|
205 |
},
|
206 |
{
|
207 |
+
"epoch": 0.43,
|
208 |
+
"grad_norm": 91.54881621872774,
|
209 |
+
"learning_rate": 6.96052176068713e-09,
|
210 |
+
"logits/chosen": -1.7626311779022217,
|
211 |
+
"logits/rejected": -1.7299768924713135,
|
212 |
+
"logps/chosen": -1.0229580402374268,
|
213 |
+
"logps/rejected": -1.0213210582733154,
|
214 |
+
"loss": 0.6874,
|
215 |
+
"rewards/accuracies": 0.65625,
|
216 |
+
"rewards/chosen": 0.010313736274838448,
|
217 |
+
"rewards/margins": 0.01158633828163147,
|
218 |
+
"rewards/rejected": -0.0012726020067930222,
|
219 |
"step": 130
|
220 |
},
|
221 |
{
|
222 |
+
"epoch": 0.47,
|
223 |
+
"grad_norm": 88.02230042083127,
|
224 |
+
"learning_rate": 6.4112093379492135e-09,
|
225 |
+
"logits/chosen": -1.796229600906372,
|
226 |
+
"logits/rejected": -1.7824671268463135,
|
227 |
+
"logps/chosen": -1.078906774520874,
|
228 |
+
"logps/rejected": -1.0356519222259521,
|
229 |
+
"loss": 0.6887,
|
230 |
+
"rewards/accuracies": 0.65625,
|
231 |
+
"rewards/chosen": 0.015195205807685852,
|
232 |
+
"rewards/margins": 0.011107077822089195,
|
233 |
+
"rewards/rejected": 0.004088127985596657,
|
234 |
"step": 140
|
235 |
},
|
236 |
{
|
237 |
+
"epoch": 0.5,
|
238 |
+
"grad_norm": 89.70971937959732,
|
239 |
+
"learning_rate": 5.842670737842467e-09,
|
240 |
+
"logits/chosen": -1.8206145763397217,
|
241 |
+
"logits/rejected": -1.788490891456604,
|
242 |
+
"logps/chosen": -1.0580933094024658,
|
243 |
+
"logps/rejected": -1.0225986242294312,
|
244 |
+
"loss": 0.6877,
|
245 |
+
"rewards/accuracies": 0.5625,
|
246 |
+
"rewards/chosen": 0.004929685965180397,
|
247 |
+
"rewards/margins": 0.005628877319395542,
|
248 |
+
"rewards/rejected": -0.0006991913542151451,
|
249 |
"step": 150
|
250 |
},
|
251 |
{
|
252 |
+
"epoch": 0.53,
|
253 |
+
"grad_norm": 81.17857900193553,
|
254 |
+
"learning_rate": 5.262651674395798e-09,
|
255 |
+
"logits/chosen": -1.8310163021087646,
|
256 |
+
"logits/rejected": -1.8322261571884155,
|
257 |
+
"logps/chosen": -1.0157067775726318,
|
258 |
+
"logps/rejected": -1.0004805326461792,
|
259 |
+
"loss": 0.6876,
|
260 |
+
"rewards/accuracies": 0.6187499761581421,
|
261 |
+
"rewards/chosen": 0.010493971407413483,
|
262 |
+
"rewards/margins": 0.013428708538413048,
|
263 |
+
"rewards/rejected": -0.0029347380623221397,
|
264 |
"step": 160
|
265 |
},
|
266 |
{
|
267 |
+
"epoch": 0.57,
|
268 |
+
"grad_norm": 87.11260929204627,
|
269 |
+
"learning_rate": 4.679054270342702e-09,
|
270 |
+
"logits/chosen": -1.8449046611785889,
|
271 |
+
"logits/rejected": -1.7946765422821045,
|
272 |
+
"logps/chosen": -1.0548999309539795,
|
273 |
+
"logps/rejected": -1.051992654800415,
|
274 |
+
"loss": 0.6869,
|
275 |
+
"rewards/accuracies": 0.6937500238418579,
|
276 |
+
"rewards/chosen": 0.010436683893203735,
|
277 |
+
"rewards/margins": 0.010068513453006744,
|
278 |
+
"rewards/rejected": 0.00036817044019699097,
|
279 |
"step": 170
|
280 |
},
|
281 |
{
|
282 |
+
"epoch": 0.6,
|
283 |
+
"grad_norm": 93.41415146032115,
|
284 |
+
"learning_rate": 4.099829399377524e-09,
|
285 |
+
"logits/chosen": -1.8277971744537354,
|
286 |
+
"logits/rejected": -1.7856277227401733,
|
287 |
+
"logps/chosen": -1.0608714818954468,
|
288 |
+
"logps/rejected": -1.0331629514694214,
|
289 |
+
"loss": 0.6855,
|
290 |
+
"rewards/accuracies": 0.6000000238418579,
|
291 |
+
"rewards/chosen": 0.015542459674179554,
|
292 |
+
"rewards/margins": 0.01706361211836338,
|
293 |
+
"rewards/rejected": -0.0015211515128612518,
|
294 |
"step": 180
|
295 |
},
|
296 |
{
|
297 |
+
"epoch": 0.63,
|
298 |
+
"grad_norm": 91.20164752256586,
|
299 |
+
"learning_rate": 3.532868364233416e-09,
|
300 |
+
"logits/chosen": -1.8144668340682983,
|
301 |
+
"logits/rejected": -1.7934455871582031,
|
302 |
+
"logps/chosen": -1.0488895177841187,
|
303 |
+
"logps/rejected": -1.0484153032302856,
|
304 |
+
"loss": 0.6869,
|
305 |
+
"rewards/accuracies": 0.643750011920929,
|
306 |
+
"rewards/chosen": 0.015009616501629353,
|
307 |
+
"rewards/margins": 0.011918185278773308,
|
308 |
+
"rewards/rejected": 0.00309143029153347,
|
309 |
"step": 190
|
310 |
},
|
311 |
{
|
312 |
+
"epoch": 0.67,
|
313 |
+
"grad_norm": 103.72123679470863,
|
314 |
+
"learning_rate": 2.985895386349233e-09,
|
315 |
+
"logits/chosen": -1.783926248550415,
|
316 |
+
"logits/rejected": -1.7509727478027344,
|
317 |
+
"logps/chosen": -1.033827543258667,
|
318 |
+
"logps/rejected": -1.0074162483215332,
|
319 |
+
"loss": 0.6856,
|
320 |
+
"rewards/accuracies": 0.6499999761581421,
|
321 |
+
"rewards/chosen": 0.018476296216249466,
|
322 |
+
"rewards/margins": 0.019362105056643486,
|
323 |
+
"rewards/rejected": -0.0008858110522851348,
|
324 |
"step": 200
|
325 |
},
|
326 |
{
|
327 |
+
"epoch": 0.7,
|
328 |
+
"grad_norm": 86.58629109761597,
|
329 |
+
"learning_rate": 2.4663623718355446e-09,
|
330 |
+
"logits/chosen": -1.842024803161621,
|
331 |
+
"logits/rejected": -1.8078495264053345,
|
332 |
+
"logps/chosen": -1.0763428211212158,
|
333 |
+
"logps/rejected": -1.0434763431549072,
|
334 |
+
"loss": 0.6853,
|
335 |
+
"rewards/accuracies": 0.6625000238418579,
|
336 |
+
"rewards/chosen": 0.018699195235967636,
|
337 |
+
"rewards/margins": 0.017539886757731438,
|
338 |
+
"rewards/rejected": 0.0011593066155910492,
|
339 |
"step": 210
|
340 |
},
|
341 |
{
|
342 |
+
"epoch": 0.74,
|
343 |
+
"grad_norm": 94.3943636617196,
|
344 |
+
"learning_rate": 1.9813473874379397e-09,
|
345 |
+
"logits/chosen": -1.732317328453064,
|
346 |
+
"logits/rejected": -1.7312743663787842,
|
347 |
+
"logps/chosen": -1.073425054550171,
|
348 |
+
"logps/rejected": -1.0629937648773193,
|
349 |
+
"loss": 0.6847,
|
350 |
+
"rewards/accuracies": 0.6312500238418579,
|
351 |
+
"rewards/chosen": 0.013539738953113556,
|
352 |
+
"rewards/margins": 0.013838117942214012,
|
353 |
+
"rewards/rejected": -0.00029837898910045624,
|
354 |
"step": 220
|
355 |
},
|
356 |
{
|
357 |
+
"epoch": 0.77,
|
358 |
+
"grad_norm": 92.64370783949347,
|
359 |
+
"learning_rate": 1.5374582296511053e-09,
|
360 |
+
"logits/chosen": -1.7242523431777954,
|
361 |
+
"logits/rejected": -1.6965806484222412,
|
362 |
+
"logps/chosen": -1.0366003513336182,
|
363 |
+
"logps/rejected": -0.9941840171813965,
|
364 |
+
"loss": 0.6857,
|
365 |
+
"rewards/accuracies": 0.625,
|
366 |
+
"rewards/chosen": 0.021335098892450333,
|
367 |
+
"rewards/margins": 0.015094568021595478,
|
368 |
+
"rewards/rejected": 0.006240529473870993,
|
369 |
"step": 230
|
370 |
},
|
371 |
{
|
372 |
+
"epoch": 0.8,
|
373 |
+
"grad_norm": 91.14853265672438,
|
374 |
+
"learning_rate": 1.1407424007485927e-09,
|
375 |
+
"logits/chosen": -1.881166696548462,
|
376 |
+
"logits/rejected": -1.8595256805419922,
|
377 |
+
"logps/chosen": -1.0768239498138428,
|
378 |
+
"logps/rejected": -1.0254974365234375,
|
379 |
+
"loss": 0.6848,
|
380 |
+
"rewards/accuracies": 0.7250000238418579,
|
381 |
+
"rewards/chosen": 0.021426241844892502,
|
382 |
+
"rewards/margins": 0.021464312449097633,
|
383 |
+
"rewards/rejected": -3.8067344576120377e-05,
|
384 |
"step": 240
|
385 |
},
|
386 |
{
|
387 |
+
"epoch": 0.84,
|
388 |
+
"grad_norm": 88.07457325362287,
|
389 |
+
"learning_rate": 7.966047182060226e-10,
|
390 |
+
"logits/chosen": -1.8788058757781982,
|
391 |
+
"logits/rejected": -1.881800889968872,
|
392 |
+
"logps/chosen": -1.0638504028320312,
|
393 |
+
"logps/rejected": -1.028044581413269,
|
394 |
+
"loss": 0.6828,
|
395 |
+
"rewards/accuracies": 0.637499988079071,
|
396 |
+
"rewards/chosen": 0.018116505816578865,
|
397 |
+
"rewards/margins": 0.015344863757491112,
|
398 |
+
"rewards/rejected": 0.0027716427575796843,
|
399 |
"step": 250
|
400 |
},
|
401 |
{
|
402 |
+
"epoch": 0.87,
|
403 |
+
"grad_norm": 99.87255814500503,
|
404 |
+
"learning_rate": 5.097336799988067e-10,
|
405 |
+
"logits/chosen": -1.8842456340789795,
|
406 |
+
"logits/rejected": -1.8642467260360718,
|
407 |
+
"logps/chosen": -1.0772615671157837,
|
408 |
+
"logps/rejected": -1.0665724277496338,
|
409 |
+
"loss": 0.6848,
|
410 |
+
"rewards/accuracies": 0.675000011920929,
|
411 |
+
"rewards/chosen": 0.023015262559056282,
|
412 |
+
"rewards/margins": 0.02001366578042507,
|
413 |
+
"rewards/rejected": 0.0030015967786312103,
|
414 |
"step": 260
|
415 |
},
|
416 |
{
|
417 |
+
"epoch": 0.9,
|
418 |
+
"grad_norm": 107.74820920122671,
|
419 |
+
"learning_rate": 2.840375889663871e-10,
|
420 |
+
"logits/chosen": -1.8729417324066162,
|
421 |
+
"logits/rejected": -1.8372119665145874,
|
422 |
+
"logps/chosen": -0.9936866760253906,
|
423 |
+
"logps/rejected": -1.0166294574737549,
|
424 |
+
"loss": 0.6847,
|
425 |
+
"rewards/accuracies": 0.7250000238418579,
|
426 |
+
"rewards/chosen": 0.01751965843141079,
|
427 |
+
"rewards/margins": 0.02036522701382637,
|
428 |
+
"rewards/rejected": -0.0028455646242946386,
|
429 |
"step": 270
|
430 |
},
|
431 |
{
|
432 |
+
"epoch": 0.94,
|
433 |
+
"grad_norm": 87.56807066875254,
|
434 |
+
"learning_rate": 1.2259130647833627e-10,
|
435 |
+
"logits/chosen": -1.7625566720962524,
|
436 |
+
"logits/rejected": -1.7224514484405518,
|
437 |
+
"logps/chosen": -1.093827724456787,
|
438 |
+
"logps/rejected": -1.0462143421173096,
|
439 |
+
"loss": 0.6834,
|
440 |
+
"rewards/accuracies": 0.6187499761581421,
|
441 |
+
"rewards/chosen": 0.01200016774237156,
|
442 |
+
"rewards/margins": 0.019106844440102577,
|
443 |
+
"rewards/rejected": -0.0071066757664084435,
|
444 |
"step": 280
|
445 |
},
|
446 |
{
|
447 |
+
"epoch": 0.97,
|
448 |
+
"grad_norm": 81.16688874976322,
|
449 |
+
"learning_rate": 2.7594360825166643e-11,
|
450 |
+
"logits/chosen": -1.780118703842163,
|
451 |
+
"logits/rejected": -1.7672898769378662,
|
452 |
+
"logps/chosen": -1.0616685152053833,
|
453 |
+
"logps/rejected": -1.0320245027542114,
|
454 |
+
"loss": 0.6845,
|
455 |
+
"rewards/accuracies": 0.6499999761581421,
|
456 |
+
"rewards/chosen": 0.028133947402238846,
|
457 |
+
"rewards/margins": 0.020768892019987106,
|
458 |
+
"rewards/rejected": 0.007365054450929165,
|
459 |
"step": 290
|
460 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
461 |
{
|
462 |
"epoch": 1.0,
|
463 |
+
"step": 299,
|
464 |
"total_flos": 0.0,
|
465 |
+
"train_loss": 0.6880149315033469,
|
466 |
+
"train_runtime": 4603.368,
|
467 |
+
"train_samples_per_second": 8.317,
|
468 |
+
"train_steps_per_second": 0.065
|
469 |
}
|
470 |
],
|
471 |
"logging_steps": 10,
|
472 |
+
"max_steps": 299,
|
473 |
"num_input_tokens_seen": 0,
|
474 |
"num_train_epochs": 1,
|
475 |
"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
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:5fc17d04ff8999b784673094bfc94f340bf1b82c104ffe64c18c7d2f2ea8b9f4
|
3 |
+
size 6328
|