Model save
Browse files- README.md +13 -13
- all_results.json +3 -3
- model-00001-of-00003.safetensors +1 -1
- model-00002-of-00003.safetensors +1 -1
- model-00003-of-00003.safetensors +1 -1
- runs/May14_03-30-29_n136-129-074/events.out.tfevents.1715628763.n136-129-074.2348486.0 +2 -2
- train_results.json +3 -3
- trainer_state.json +502 -502
README.md
CHANGED
@@ -15,15 +15,15 @@ should probably proofread and complete it, then remove this comment. -->
|
|
15 |
|
16 |
This model was trained from scratch on the None dataset.
|
17 |
It achieves the following results on the evaluation set:
|
18 |
-
- Loss: 0.
|
19 |
-
- Rewards/chosen: 8.
|
20 |
-
- Rewards/rejected: -16.
|
21 |
-
- Rewards/accuracies: 0.
|
22 |
-
- Rewards/margins: 24.
|
23 |
-
- Logps/rejected: -
|
24 |
-
- Logps/chosen: -121.
|
25 |
-
- Logits/rejected: -1.
|
26 |
-
- Logits/chosen: -1.
|
27 |
|
28 |
## Model description
|
29 |
|
@@ -60,10 +60,10 @@ The following hyperparameters were used during training:
|
|
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 |
-
| 0.1474 | 0.21 | 100 | 0.
|
64 |
-
| 0.
|
65 |
-
| 0.
|
66 |
-
| 0.
|
67 |
|
68 |
|
69 |
### Framework versions
|
|
|
15 |
|
16 |
This model was trained from scratch on the None dataset.
|
17 |
It achieves the following results on the evaluation set:
|
18 |
+
- Loss: 0.1695
|
19 |
+
- Rewards/chosen: 8.1916
|
20 |
+
- Rewards/rejected: -16.5277
|
21 |
+
- Rewards/accuracies: 0.9297
|
22 |
+
- Rewards/margins: 24.7193
|
23 |
+
- Logps/rejected: -126.2961
|
24 |
+
- Logps/chosen: -121.3891
|
25 |
+
- Logits/rejected: -1.8336
|
26 |
+
- Logits/chosen: -1.8336
|
27 |
|
28 |
## Model description
|
29 |
|
|
|
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 |
+
| 0.1474 | 0.21 | 100 | 0.1325 | 6.8739 | -14.1333 | 0.9297 | 21.0072 | -121.5073 | -124.0244 | -1.7724 | -1.7834 |
|
64 |
+
| 0.2092 | 0.42 | 200 | 0.1567 | 8.4112 | -15.0418 | 0.9336 | 23.4530 | -123.3244 | -120.9499 | -1.8442 | -1.8477 |
|
65 |
+
| 0.1925 | 0.63 | 300 | 0.1715 | 7.7458 | -16.7009 | 0.9258 | 24.4467 | -126.6425 | -122.2807 | -1.8047 | -1.8054 |
|
66 |
+
| 0.2762 | 0.84 | 400 | 0.1695 | 8.1916 | -16.5277 | 0.9297 | 24.7193 | -126.2961 | -121.3891 | -1.8336 | -1.8336 |
|
67 |
|
68 |
|
69 |
### Framework versions
|
all_results.json
CHANGED
@@ -1,8 +1,8 @@
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
-
"train_loss": 0.
|
4 |
-
"train_runtime":
|
5 |
"train_samples": 61135,
|
6 |
-
"train_samples_per_second": 8.
|
7 |
"train_steps_per_second": 0.063
|
8 |
}
|
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
+
"train_loss": 0.20260818084413537,
|
4 |
+
"train_runtime": 7642.1158,
|
5 |
"train_samples": 61135,
|
6 |
+
"train_samples_per_second": 8.0,
|
7 |
"train_steps_per_second": 0.063
|
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 4943178720
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:c236809df2b9a9098eb9f2d34c776fda00f7a9d1bf22382bf81f4f3a61e886d2
|
3 |
size 4943178720
|
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 sha256:01733d5f594df90f00ce7815c23e992b809250faa4a0178be321b9b7ac687a96
|
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 4540532728
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:c9bc2f19df9fee5aa320225d07ac06c4aa6645bf7d9d004feb7974633020ea43
|
3 |
size 4540532728
|
runs/May14_03-30-29_n136-129-074/events.out.tfevents.1715628763.n136-129-074.2348486.0
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:c0d4d3d3a226528d79a2ba206d907f12c90d8a75b263159044d331b74b6d8b8e
|
3 |
+
size 41092
|
train_results.json
CHANGED
@@ -1,8 +1,8 @@
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
-
"train_loss": 0.
|
4 |
-
"train_runtime":
|
5 |
"train_samples": 61135,
|
6 |
-
"train_samples_per_second": 8.
|
7 |
"train_steps_per_second": 0.063
|
8 |
}
|
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
+
"train_loss": 0.20260818084413537,
|
4 |
+
"train_runtime": 7642.1158,
|
5 |
"train_samples": 61135,
|
6 |
+
"train_samples_per_second": 8.0,
|
7 |
"train_steps_per_second": 0.063
|
8 |
}
|
trainer_state.json
CHANGED
@@ -10,7 +10,7 @@
|
|
10 |
"log_history": [
|
11 |
{
|
12 |
"epoch": 0.0,
|
13 |
-
"grad_norm": 2193.
|
14 |
"learning_rate": 1.0416666666666666e-08,
|
15 |
"logits/chosen": -1.689455509185791,
|
16 |
"logits/rejected": -1.4794573783874512,
|
@@ -25,780 +25,780 @@
|
|
25 |
},
|
26 |
{
|
27 |
"epoch": 0.02,
|
28 |
-
"grad_norm":
|
29 |
"learning_rate": 1.0416666666666667e-07,
|
30 |
-
"logits/chosen": -1.
|
31 |
-
"logits/rejected": -1.
|
32 |
-
"logps/chosen": -139.
|
33 |
-
"logps/rejected": -91.
|
34 |
-
"loss": 0.
|
35 |
-
"rewards/accuracies": 0.
|
36 |
-
"rewards/chosen": 0.
|
37 |
-
"rewards/margins": 0.
|
38 |
-
"rewards/rejected": -0.
|
39 |
"step": 10
|
40 |
},
|
41 |
{
|
42 |
"epoch": 0.04,
|
43 |
-
"grad_norm":
|
44 |
"learning_rate": 2.0833333333333333e-07,
|
45 |
-
"logits/chosen": -1.
|
46 |
-
"logits/rejected": -1.
|
47 |
-
"logps/chosen": -130.
|
48 |
-
"logps/rejected": -94.
|
49 |
-
"loss": 0.
|
50 |
-
"rewards/accuracies": 0.
|
51 |
-
"rewards/chosen": 1.
|
52 |
-
"rewards/margins": 1.
|
53 |
-
"rewards/rejected": -0.
|
54 |
"step": 20
|
55 |
},
|
56 |
{
|
57 |
"epoch": 0.06,
|
58 |
-
"grad_norm":
|
59 |
"learning_rate": 3.1249999999999997e-07,
|
60 |
-
"logits/chosen": -1.
|
61 |
-
"logits/rejected": -1.
|
62 |
-
"logps/chosen": -131.
|
63 |
-
"logps/rejected": -105.
|
64 |
-
"loss": 0.
|
65 |
-
"rewards/accuracies": 0.
|
66 |
-
"rewards/chosen": 3.
|
67 |
-
"rewards/margins":
|
68 |
-
"rewards/rejected": -1.
|
69 |
"step": 30
|
70 |
},
|
71 |
{
|
72 |
"epoch": 0.08,
|
73 |
-
"grad_norm":
|
74 |
"learning_rate": 4.1666666666666667e-07,
|
75 |
-
"logits/chosen": -1.
|
76 |
-
"logits/rejected": -1.
|
77 |
-
"logps/chosen": -142.
|
78 |
-
"logps/rejected": -108.
|
79 |
-
"loss": 0.
|
80 |
-
"rewards/accuracies": 0.
|
81 |
-
"rewards/chosen": 4.
|
82 |
-
"rewards/margins": 9.
|
83 |
-
"rewards/rejected": -5.
|
84 |
"step": 40
|
85 |
},
|
86 |
{
|
87 |
"epoch": 0.1,
|
88 |
-
"grad_norm":
|
89 |
"learning_rate": 4.999733114418725e-07,
|
90 |
-
"logits/chosen": -1.
|
91 |
-
"logits/rejected": -1.
|
92 |
-
"logps/chosen": -126.
|
93 |
-
"logps/rejected": -116.
|
94 |
-
"loss": 0.
|
95 |
-
"rewards/accuracies": 0.
|
96 |
-
"rewards/chosen": 4.
|
97 |
-
"rewards/margins": 12.
|
98 |
-
"rewards/rejected": -8.
|
99 |
"step": 50
|
100 |
},
|
101 |
{
|
102 |
"epoch": 0.13,
|
103 |
-
"grad_norm":
|
104 |
"learning_rate": 4.990398100856366e-07,
|
105 |
-
"logits/chosen": -1.
|
106 |
-
"logits/rejected": -1.
|
107 |
-
"logps/chosen": -141.
|
108 |
-
"logps/rejected": -125.
|
109 |
-
"loss": 0.
|
110 |
"rewards/accuracies": 0.9437500238418579,
|
111 |
-
"rewards/chosen": 4.
|
112 |
-
"rewards/margins": 16.
|
113 |
-
"rewards/rejected": -11.
|
114 |
"step": 60
|
115 |
},
|
116 |
{
|
117 |
"epoch": 0.15,
|
118 |
-
"grad_norm":
|
119 |
"learning_rate": 4.967775735898179e-07,
|
120 |
-
"logits/chosen": -1.
|
121 |
-
"logits/rejected": -1.
|
122 |
-
"logps/chosen": -134.
|
123 |
-
"logps/rejected": -126.
|
124 |
-
"loss": 0.
|
125 |
"rewards/accuracies": 0.925000011920929,
|
126 |
-
"rewards/chosen": 5.
|
127 |
-
"rewards/margins": 19.
|
128 |
-
"rewards/rejected": -14.
|
129 |
"step": 70
|
130 |
},
|
131 |
{
|
132 |
"epoch": 0.17,
|
133 |
-
"grad_norm":
|
134 |
"learning_rate": 4.931986719649298e-07,
|
135 |
-
"logits/chosen": -1.
|
136 |
-
"logits/rejected": -1.
|
137 |
-
"logps/chosen": -128.
|
138 |
-
"logps/rejected": -119.
|
139 |
-
"loss": 0.
|
140 |
"rewards/accuracies": 0.9125000238418579,
|
141 |
-
"rewards/chosen": 5.
|
142 |
-
"rewards/margins": 18.
|
143 |
-
"rewards/rejected": -12.
|
144 |
"step": 80
|
145 |
},
|
146 |
{
|
147 |
"epoch": 0.19,
|
148 |
-
"grad_norm":
|
149 |
"learning_rate": 4.883222001996351e-07,
|
150 |
-
"logits/chosen": -1.
|
151 |
-
"logits/rejected": -1.
|
152 |
-
"logps/chosen": -136.
|
153 |
-
"logps/rejected": -131.
|
154 |
-
"loss": 0.
|
155 |
-
"rewards/accuracies": 0.
|
156 |
-
"rewards/chosen": 6.
|
157 |
-
"rewards/margins": 21.
|
158 |
-
"rewards/rejected": -14.
|
159 |
"step": 90
|
160 |
},
|
161 |
{
|
162 |
"epoch": 0.21,
|
163 |
-
"grad_norm":
|
164 |
"learning_rate": 4.821741763807186e-07,
|
165 |
-
"logits/chosen": -1.
|
166 |
-
"logits/rejected": -1.
|
167 |
-
"logps/chosen": -116.
|
168 |
-
"logps/rejected": -
|
169 |
"loss": 0.1474,
|
170 |
-
"rewards/accuracies": 0.
|
171 |
-
"rewards/chosen": 5.
|
172 |
-
"rewards/margins": 18.
|
173 |
-
"rewards/rejected": -
|
174 |
"step": 100
|
175 |
},
|
176 |
{
|
177 |
"epoch": 0.21,
|
178 |
-
"eval_logits/chosen": -1.
|
179 |
-
"eval_logits/rejected": -1.
|
180 |
-
"eval_logps/chosen": -
|
181 |
-
"eval_logps/rejected": -
|
182 |
-
"eval_loss": 0.
|
183 |
-
"eval_rewards/accuracies": 0.
|
184 |
-
"eval_rewards/chosen":
|
185 |
-
"eval_rewards/margins":
|
186 |
-
"eval_rewards/rejected": -
|
187 |
-
"eval_runtime": 97.
|
188 |
-
"eval_samples_per_second": 20.
|
189 |
-
"eval_steps_per_second": 0.
|
190 |
"step": 100
|
191 |
},
|
192 |
{
|
193 |
"epoch": 0.23,
|
194 |
-
"grad_norm":
|
195 |
"learning_rate": 4.747874028753375e-07,
|
196 |
-
"logits/chosen": -1.
|
197 |
-
"logits/rejected": -1.
|
198 |
-
"logps/chosen": -120.
|
199 |
-
"logps/rejected": -124.
|
200 |
-
"loss": 0.
|
201 |
-
"rewards/accuracies": 0.
|
202 |
-
"rewards/chosen": 5.
|
203 |
-
"rewards/margins":
|
204 |
-
"rewards/rejected": -11.
|
205 |
"step": 110
|
206 |
},
|
207 |
{
|
208 |
"epoch": 0.25,
|
209 |
-
"grad_norm":
|
210 |
"learning_rate": 4.662012913161997e-07,
|
211 |
-
"logits/chosen": -1.
|
212 |
-
"logits/rejected": -1.
|
213 |
-
"logps/chosen": -118.
|
214 |
-
"logps/rejected": -125.
|
215 |
-
"loss": 0.
|
216 |
-
"rewards/accuracies": 0.
|
217 |
-
"rewards/chosen":
|
218 |
-
"rewards/margins": 19.
|
219 |
-
"rewards/rejected": -
|
220 |
"step": 120
|
221 |
},
|
222 |
{
|
223 |
"epoch": 0.27,
|
224 |
-
"grad_norm":
|
225 |
"learning_rate": 4.5646165232345103e-07,
|
226 |
-
"logits/chosen": -1.
|
227 |
-
"logits/rejected": -1.
|
228 |
-
"logps/chosen": -124.
|
229 |
-
"logps/rejected": -
|
230 |
-
"loss": 0.
|
231 |
"rewards/accuracies": 0.918749988079071,
|
232 |
-
"rewards/chosen": 7.
|
233 |
-
"rewards/margins": 21.
|
234 |
-
"rewards/rejected": -13.
|
235 |
"step": 130
|
236 |
},
|
237 |
{
|
238 |
"epoch": 0.29,
|
239 |
-
"grad_norm":
|
240 |
"learning_rate": 4.456204510851956e-07,
|
241 |
-
"logits/chosen": -1.
|
242 |
-
"logits/rejected": -1.
|
243 |
-
"logps/chosen": -122.
|
244 |
-
"logps/rejected": -117.
|
245 |
-
"loss": 0.
|
246 |
-
"rewards/accuracies": 0.
|
247 |
-
"rewards/chosen": 7.
|
248 |
-
"rewards/margins": 19.
|
249 |
-
"rewards/rejected": -12.
|
250 |
"step": 140
|
251 |
},
|
252 |
{
|
253 |
"epoch": 0.31,
|
254 |
-
"grad_norm":
|
255 |
"learning_rate": 4.337355301007335e-07,
|
256 |
-
"logits/chosen": -1.
|
257 |
-
"logits/rejected": -1.
|
258 |
-
"logps/chosen": -120.
|
259 |
-
"logps/rejected": -
|
260 |
-
"loss": 0.
|
261 |
-
"rewards/accuracies": 0.
|
262 |
-
"rewards/chosen": 7.
|
263 |
-
"rewards/margins": 20.
|
264 |
-
"rewards/rejected": -12.
|
265 |
"step": 150
|
266 |
},
|
267 |
{
|
268 |
"epoch": 0.33,
|
269 |
-
"grad_norm":
|
270 |
"learning_rate": 4.2087030056579986e-07,
|
271 |
-
"logits/chosen": -1.
|
272 |
-
"logits/rejected": -1.
|
273 |
-
"logps/chosen": -
|
274 |
-
"logps/rejected": -122.
|
275 |
-
"loss": 0.
|
276 |
"rewards/accuracies": 0.9125000238418579,
|
277 |
-
"rewards/chosen":
|
278 |
-
"rewards/margins": 18.
|
279 |
-
"rewards/rejected": -11.
|
280 |
"step": 160
|
281 |
},
|
282 |
{
|
283 |
"epoch": 0.36,
|
284 |
-
"grad_norm":
|
285 |
"learning_rate": 4.070934040463998e-07,
|
286 |
-
"logits/chosen": -1.
|
287 |
-
"logits/rejected": -1.
|
288 |
-
"logps/chosen": -
|
289 |
-
"logps/rejected": -
|
290 |
-
"loss": 0.
|
291 |
-
"rewards/accuracies": 0.
|
292 |
-
"rewards/chosen": 8.
|
293 |
-
"rewards/margins": 21.
|
294 |
-
"rewards/rejected": -13.
|
295 |
"step": 170
|
296 |
},
|
297 |
{
|
298 |
"epoch": 0.38,
|
299 |
-
"grad_norm":
|
300 |
"learning_rate": 3.9247834624635404e-07,
|
301 |
-
"logits/chosen": -1.
|
302 |
-
"logits/rejected": -1.
|
303 |
-
"logps/chosen": -
|
304 |
-
"logps/rejected": -118.
|
305 |
-
"loss": 0.
|
306 |
"rewards/accuracies": 0.918749988079071,
|
307 |
-
"rewards/chosen": 8.
|
308 |
-
"rewards/margins": 22.
|
309 |
-
"rewards/rejected": -13.
|
310 |
"step": 180
|
311 |
},
|
312 |
{
|
313 |
"epoch": 0.4,
|
314 |
-
"grad_norm":
|
315 |
"learning_rate": 3.7710310482256523e-07,
|
316 |
-
"logits/chosen": -1.
|
317 |
-
"logits/rejected": -1.
|
318 |
-
"logps/chosen": -
|
319 |
-
"logps/rejected": -
|
320 |
-
"loss": 0.
|
321 |
-
"rewards/accuracies": 0.
|
322 |
-
"rewards/chosen": 8.
|
323 |
-
"rewards/margins": 23.
|
324 |
-
"rewards/rejected": -
|
325 |
"step": 190
|
326 |
},
|
327 |
{
|
328 |
"epoch": 0.42,
|
329 |
-
"grad_norm":
|
330 |
"learning_rate": 3.610497133404795e-07,
|
331 |
-
"logits/chosen": -1.
|
332 |
-
"logits/rejected": -1.
|
333 |
-
"logps/chosen": -
|
334 |
-
"logps/rejected": -
|
335 |
-
"loss": 0.
|
336 |
"rewards/accuracies": 0.9437500238418579,
|
337 |
-
"rewards/chosen": 9.
|
338 |
-
"rewards/margins": 23.
|
339 |
-
"rewards/rejected": -
|
340 |
"step": 200
|
341 |
},
|
342 |
{
|
343 |
"epoch": 0.42,
|
344 |
-
"eval_logits/chosen": -1.
|
345 |
-
"eval_logits/rejected": -1.
|
346 |
-
"eval_logps/chosen": -120.
|
347 |
-
"eval_logps/rejected": -123.
|
348 |
-
"eval_loss": 0.
|
349 |
"eval_rewards/accuracies": 0.93359375,
|
350 |
-
"eval_rewards/chosen": 8.
|
351 |
-
"eval_rewards/margins": 23.
|
352 |
-
"eval_rewards/rejected": -
|
353 |
-
"eval_runtime": 97.
|
354 |
-
"eval_samples_per_second": 20.
|
355 |
"eval_steps_per_second": 0.328,
|
356 |
"step": 200
|
357 |
},
|
358 |
{
|
359 |
"epoch": 0.44,
|
360 |
-
"grad_norm":
|
361 |
"learning_rate": 3.4440382358952115e-07,
|
362 |
-
"logits/chosen": -1.
|
363 |
-
"logits/rejected": -1.
|
364 |
-
"logps/chosen": -119.
|
365 |
-
"logps/rejected": -112.
|
366 |
-
"loss": 0.
|
367 |
-
"rewards/accuracies": 0.
|
368 |
-
"rewards/chosen": 6.
|
369 |
-
"rewards/margins": 18.
|
370 |
-
"rewards/rejected": -11.
|
371 |
"step": 210
|
372 |
},
|
373 |
{
|
374 |
"epoch": 0.46,
|
375 |
-
"grad_norm":
|
376 |
"learning_rate": 3.272542485937368e-07,
|
377 |
-
"logits/chosen": -1.
|
378 |
-
"logits/rejected": -1.
|
379 |
-
"logps/chosen": -120.
|
380 |
-
"logps/rejected": -113.
|
381 |
-
"loss": 0.
|
382 |
-
"rewards/accuracies": 0.
|
383 |
-
"rewards/chosen": 8.
|
384 |
-
"rewards/margins":
|
385 |
-
"rewards/rejected": -
|
386 |
"step": 220
|
387 |
},
|
388 |
{
|
389 |
"epoch": 0.48,
|
390 |
-
"grad_norm":
|
391 |
"learning_rate": 3.096924887558854e-07,
|
392 |
-
"logits/chosen": -1.
|
393 |
-
"logits/rejected": -1.
|
394 |
-
"logps/chosen": -132.
|
395 |
-
"logps/rejected": -132.
|
396 |
-
"loss": 0.
|
397 |
-
"rewards/accuracies": 0.
|
398 |
-
"rewards/chosen": 7.
|
399 |
-
"rewards/margins": 23.
|
400 |
-
"rewards/rejected": -16.
|
401 |
"step": 230
|
402 |
},
|
403 |
{
|
404 |
"epoch": 0.5,
|
405 |
-
"grad_norm":
|
406 |
"learning_rate": 2.9181224366319943e-07,
|
407 |
-
"logits/chosen": -1.
|
408 |
-
"logits/rejected": -1.
|
409 |
-
"logps/chosen": -
|
410 |
-
"logps/rejected": -
|
411 |
-
"loss": 0.
|
412 |
-
"rewards/accuracies": 0.
|
413 |
-
"rewards/chosen": 6.
|
414 |
-
"rewards/margins": 19.
|
415 |
-
"rewards/rejected": -13.
|
416 |
"step": 240
|
417 |
},
|
418 |
{
|
419 |
"epoch": 0.52,
|
420 |
-
"grad_norm": 506.
|
421 |
"learning_rate": 2.7370891215954565e-07,
|
422 |
-
"logits/chosen": -1.
|
423 |
-
"logits/rejected": -1.
|
424 |
-
"logps/chosen": -
|
425 |
-
"logps/rejected": -
|
426 |
-
"loss": 0.
|
427 |
"rewards/accuracies": 0.9312499761581421,
|
428 |
-
"rewards/chosen": 7.
|
429 |
-
"rewards/margins": 21.
|
430 |
-
"rewards/rejected": -14.
|
431 |
"step": 250
|
432 |
},
|
433 |
{
|
434 |
"epoch": 0.54,
|
435 |
-
"grad_norm":
|
436 |
"learning_rate": 2.55479083351317e-07,
|
437 |
-
"logits/chosen": -1.
|
438 |
-
"logits/rejected": -1.
|
439 |
-
"logps/chosen": -
|
440 |
-
"logps/rejected": -
|
441 |
-
"loss": 0.
|
442 |
-
"rewards/accuracies": 0.
|
443 |
-
"rewards/chosen": 8.
|
444 |
-
"rewards/margins": 23.
|
445 |
-
"rewards/rejected": -14.
|
446 |
"step": 260
|
447 |
},
|
448 |
{
|
449 |
"epoch": 0.56,
|
450 |
-
"grad_norm":
|
451 |
"learning_rate": 2.3722002126275822e-07,
|
452 |
-
"logits/chosen": -1.
|
453 |
-
"logits/rejected": -1.
|
454 |
-
"logps/chosen": -
|
455 |
-
"logps/rejected": -
|
456 |
-
"loss": 0.
|
457 |
-
"rewards/accuracies": 0.
|
458 |
-
"rewards/chosen": 7.
|
459 |
-
"rewards/margins": 21.
|
460 |
-
"rewards/rejected": -
|
461 |
"step": 270
|
462 |
},
|
463 |
{
|
464 |
"epoch": 0.59,
|
465 |
-
"grad_norm":
|
466 |
"learning_rate": 2.19029145890313e-07,
|
467 |
-
"logits/chosen": -1.
|
468 |
-
"logits/rejected": -1.
|
469 |
-
"logps/chosen": -
|
470 |
-
"logps/rejected": -
|
471 |
-
"loss": 0.
|
472 |
-
"rewards/accuracies": 0.
|
473 |
-
"rewards/chosen":
|
474 |
-
"rewards/margins": 22.
|
475 |
-
"rewards/rejected": -15.
|
476 |
"step": 280
|
477 |
},
|
478 |
{
|
479 |
"epoch": 0.61,
|
480 |
-
"grad_norm":
|
481 |
"learning_rate": 2.0100351342479216e-07,
|
482 |
-
"logits/chosen": -1.
|
483 |
-
"logits/rejected": -1.
|
484 |
-
"logps/chosen": -114.
|
485 |
-
"logps/rejected": -119.
|
486 |
-
"loss": 0.
|
487 |
-
"rewards/accuracies": 0.
|
488 |
-
"rewards/chosen": 6.
|
489 |
-
"rewards/margins":
|
490 |
-
"rewards/rejected": -
|
491 |
"step": 290
|
492 |
},
|
493 |
{
|
494 |
"epoch": 0.63,
|
495 |
-
"grad_norm":
|
496 |
"learning_rate": 1.8323929841460178e-07,
|
497 |
-
"logits/chosen": -1.
|
498 |
-
"logits/rejected": -1.
|
499 |
-
"logps/chosen": -
|
500 |
-
"logps/rejected": -
|
501 |
-
"loss": 0.
|
502 |
-
"rewards/accuracies": 0.
|
503 |
-
"rewards/chosen": 7.
|
504 |
-
"rewards/margins": 22.
|
505 |
-
"rewards/rejected": -
|
506 |
"step": 300
|
507 |
},
|
508 |
{
|
509 |
"epoch": 0.63,
|
510 |
-
"eval_logits/chosen": -1.
|
511 |
-
"eval_logits/rejected": -1.
|
512 |
-
"eval_logps/chosen": -
|
513 |
-
"eval_logps/rejected": -
|
514 |
-
"eval_loss": 0.
|
515 |
"eval_rewards/accuracies": 0.92578125,
|
516 |
-
"eval_rewards/chosen": 7.
|
517 |
-
"eval_rewards/margins": 24.
|
518 |
-
"eval_rewards/rejected": -16.
|
519 |
-
"eval_runtime": 97.
|
520 |
-
"eval_samples_per_second": 20.
|
521 |
"eval_steps_per_second": 0.328,
|
522 |
"step": 300
|
523 |
},
|
524 |
{
|
525 |
"epoch": 0.65,
|
526 |
-
"grad_norm":
|
527 |
"learning_rate": 1.6583128063291573e-07,
|
528 |
-
"logits/chosen": -1.
|
529 |
-
"logits/rejected": -1.
|
530 |
-
"logps/chosen": -123.
|
531 |
-
"logps/rejected": -
|
532 |
-
"loss": 0.
|
533 |
-
"rewards/accuracies": 0.
|
534 |
-
"rewards/chosen":
|
535 |
-
"rewards/margins": 22.
|
536 |
-
"rewards/rejected": -
|
537 |
"step": 310
|
538 |
},
|
539 |
{
|
540 |
"epoch": 0.67,
|
541 |
-
"grad_norm":
|
542 |
"learning_rate": 1.488723393865766e-07,
|
543 |
-
"logits/chosen": -1.
|
544 |
-
"logits/rejected": -1.
|
545 |
-
"logps/chosen": -
|
546 |
-
"logps/rejected": -
|
547 |
-
"loss": 0.
|
548 |
"rewards/accuracies": 0.887499988079071,
|
549 |
-
"rewards/chosen": 8.
|
550 |
-
"rewards/margins":
|
551 |
-
"rewards/rejected": -
|
552 |
"step": 320
|
553 |
},
|
554 |
{
|
555 |
"epoch": 0.69,
|
556 |
-
"grad_norm":
|
557 |
"learning_rate": 1.3245295796480788e-07,
|
558 |
-
"logits/chosen": -1.
|
559 |
-
"logits/rejected": -1.
|
560 |
-
"logps/chosen": -
|
561 |
-
"logps/rejected": -
|
562 |
-
"loss": 0.
|
563 |
-
"rewards/accuracies": 0.
|
564 |
-
"rewards/chosen": 7.
|
565 |
-
"rewards/margins": 21.
|
566 |
-
"rewards/rejected": -
|
567 |
"step": 330
|
568 |
},
|
569 |
{
|
570 |
"epoch": 0.71,
|
571 |
-
"grad_norm":
|
572 |
"learning_rate": 1.1666074087171627e-07,
|
573 |
-
"logits/chosen": -1.
|
574 |
-
"logits/rejected": -1.
|
575 |
-
"logps/chosen": -
|
576 |
-
"logps/rejected": -
|
577 |
-
"loss": 0.
|
578 |
-
"rewards/accuracies": 0.
|
579 |
-
"rewards/chosen": 6.
|
580 |
-
"rewards/margins":
|
581 |
-
"rewards/rejected": -12.
|
582 |
"step": 340
|
583 |
},
|
584 |
{
|
585 |
"epoch": 0.73,
|
586 |
-
"grad_norm":
|
587 |
"learning_rate": 1.0157994641835734e-07,
|
588 |
-
"logits/chosen": -1.
|
589 |
-
"logits/rejected": -1.
|
590 |
-
"logps/chosen": -117.
|
591 |
-
"logps/rejected": -
|
592 |
-
"loss": 0.
|
593 |
-
"rewards/accuracies": 0.
|
594 |
-
"rewards/chosen": 6.
|
595 |
-
"rewards/margins": 21.
|
596 |
-
"rewards/rejected": -15.
|
597 |
"step": 350
|
598 |
},
|
599 |
{
|
600 |
"epoch": 0.75,
|
601 |
-
"grad_norm":
|
602 |
"learning_rate": 8.729103716819111e-08,
|
603 |
-
"logits/chosen": -1.
|
604 |
-
"logits/rejected": -1.
|
605 |
-
"logps/chosen": -
|
606 |
-
"logps/rejected": -
|
607 |
-
"loss": 0.
|
608 |
-
"rewards/accuracies": 0.
|
609 |
-
"rewards/chosen": 7.
|
610 |
-
"rewards/margins": 21.
|
611 |
-
"rewards/rejected": -
|
612 |
"step": 360
|
613 |
},
|
614 |
{
|
615 |
"epoch": 0.77,
|
616 |
-
"grad_norm":
|
617 |
"learning_rate": 7.387025063449081e-08,
|
618 |
-
"logits/chosen": -1.
|
619 |
-
"logits/rejected": -1.
|
620 |
-
"logps/chosen": -113.
|
621 |
-
"logps/rejected": -117.
|
622 |
-
"loss": 0.
|
623 |
"rewards/accuracies": 0.9312499761581421,
|
624 |
-
"rewards/chosen": 6.
|
625 |
-
"rewards/margins": 18.
|
626 |
-
"rewards/rejected": -12.
|
627 |
"step": 370
|
628 |
},
|
629 |
{
|
630 |
"epoch": 0.79,
|
631 |
-
"grad_norm":
|
632 |
"learning_rate": 6.138919252022435e-08,
|
633 |
-
"logits/chosen": -1.
|
634 |
-
"logits/rejected": -1.
|
635 |
-
"logps/chosen": -
|
636 |
-
"logps/rejected": -
|
637 |
-
"loss": 0.
|
638 |
"rewards/accuracies": 0.9375,
|
639 |
-
"rewards/chosen": 7.
|
640 |
-
"rewards/margins":
|
641 |
-
"rewards/rejected": -
|
642 |
"step": 380
|
643 |
},
|
644 |
{
|
645 |
"epoch": 0.82,
|
646 |
-
"grad_norm":
|
647 |
"learning_rate": 4.991445467064689e-08,
|
648 |
-
"logits/chosen": -1.
|
649 |
-
"logits/rejected": -1.
|
650 |
-
"logps/chosen": -
|
651 |
-
"logps/rejected": -
|
652 |
-
"loss": 0.
|
653 |
"rewards/accuracies": 0.9624999761581421,
|
654 |
-
"rewards/chosen": 7.
|
655 |
-
"rewards/margins": 22.
|
656 |
-
"rewards/rejected": -14.
|
657 |
"step": 390
|
658 |
},
|
659 |
{
|
660 |
"epoch": 0.84,
|
661 |
-
"grad_norm":
|
662 |
"learning_rate": 3.9507259776993954e-08,
|
663 |
-
"logits/chosen": -1.
|
664 |
-
"logits/rejected": -1.
|
665 |
-
"logps/chosen": -119.
|
666 |
-
"logps/rejected": -
|
667 |
-
"loss": 0.
|
668 |
-
"rewards/accuracies": 0.
|
669 |
-
"rewards/chosen": 6.
|
670 |
-
"rewards/margins":
|
671 |
-
"rewards/rejected": -14.
|
672 |
"step": 400
|
673 |
},
|
674 |
{
|
675 |
"epoch": 0.84,
|
676 |
-
"eval_logits/chosen": -1.
|
677 |
-
"eval_logits/rejected": -1.
|
678 |
-
"eval_logps/chosen": -121.
|
679 |
-
"eval_logps/rejected": -
|
680 |
-
"eval_loss": 0.
|
681 |
-
"eval_rewards/accuracies": 0.
|
682 |
-
"eval_rewards/chosen": 8.
|
683 |
-
"eval_rewards/margins": 24.
|
684 |
-
"eval_rewards/rejected": -16.
|
685 |
-
"eval_runtime": 97.
|
686 |
-
"eval_samples_per_second": 20.
|
687 |
-
"eval_steps_per_second": 0.
|
688 |
"step": 400
|
689 |
},
|
690 |
{
|
691 |
"epoch": 0.86,
|
692 |
-
"grad_norm":
|
693 |
"learning_rate": 3.022313472693447e-08,
|
694 |
-
"logits/chosen": -1.
|
695 |
-
"logits/rejected": -1.
|
696 |
-
"logps/chosen": -130.
|
697 |
-
"logps/rejected": -
|
698 |
-
"loss": 0.
|
699 |
-
"rewards/accuracies": 0.
|
700 |
-
"rewards/chosen": 8.
|
701 |
-
"rewards/margins": 24.
|
702 |
-
"rewards/rejected": -
|
703 |
"step": 410
|
704 |
},
|
705 |
{
|
706 |
"epoch": 0.88,
|
707 |
-
"grad_norm":
|
708 |
"learning_rate": 2.2111614344599684e-08,
|
709 |
-
"logits/chosen": -1.
|
710 |
-
"logits/rejected": -1.
|
711 |
-
"logps/chosen": -
|
712 |
-
"logps/rejected": -
|
713 |
-
"loss": 0.
|
714 |
-
"rewards/accuracies": 0.
|
715 |
-
"rewards/chosen": 6.
|
716 |
-
"rewards/margins": 20.
|
717 |
-
"rewards/rejected": -14.
|
718 |
"step": 420
|
719 |
},
|
720 |
{
|
721 |
"epoch": 0.9,
|
722 |
-
"grad_norm":
|
723 |
"learning_rate": 1.521597710086439e-08,
|
724 |
-
"logits/chosen": -1.
|
725 |
-
"logits/rejected": -1.
|
726 |
-
"logps/chosen": -
|
727 |
-
"logps/rejected": -119.
|
728 |
-
"loss": 0.
|
729 |
"rewards/accuracies": 0.9125000238418579,
|
730 |
-
"rewards/chosen": 7.
|
731 |
-
"rewards/margins": 20.
|
732 |
-
"rewards/rejected": -
|
733 |
"step": 430
|
734 |
},
|
735 |
{
|
736 |
"epoch": 0.92,
|
737 |
-
"grad_norm":
|
738 |
"learning_rate": 9.57301420397924e-09,
|
739 |
-
"logits/chosen": -1.
|
740 |
-
"logits/rejected": -1.
|
741 |
-
"logps/chosen": -122.
|
742 |
-
"logps/rejected": -125.
|
743 |
-
"loss": 0.
|
744 |
"rewards/accuracies": 0.9375,
|
745 |
-
"rewards/chosen": 8.
|
746 |
-
"rewards/margins": 24.
|
747 |
-
"rewards/rejected": -
|
748 |
"step": 440
|
749 |
},
|
750 |
{
|
751 |
"epoch": 0.94,
|
752 |
-
"grad_norm":
|
753 |
"learning_rate": 5.212833302556258e-09,
|
754 |
-
"logits/chosen": -1.
|
755 |
-
"logits/rejected": -1.
|
756 |
-
"logps/chosen": -117.
|
757 |
-
"logps/rejected": -121.
|
758 |
-
"loss": 0.
|
759 |
"rewards/accuracies": 0.90625,
|
760 |
-
"rewards/chosen": 8.
|
761 |
-
"rewards/margins": 25.
|
762 |
-
"rewards/rejected": -16.
|
763 |
"step": 450
|
764 |
},
|
765 |
{
|
766 |
"epoch": 0.96,
|
767 |
-
"grad_norm":
|
768 |
"learning_rate": 2.158697848236607e-09,
|
769 |
-
"logits/chosen": -1.
|
770 |
-
"logits/rejected": -1.
|
771 |
-
"logps/chosen": -122.
|
772 |
-
"logps/rejected": -
|
773 |
-
"loss": 0.
|
774 |
-
"rewards/accuracies": 0.
|
775 |
-
"rewards/chosen": 8.
|
776 |
-
"rewards/margins": 23.
|
777 |
-
"rewards/rejected": -14.
|
778 |
"step": 460
|
779 |
},
|
780 |
{
|
781 |
"epoch": 0.98,
|
782 |
-
"grad_norm":
|
783 |
"learning_rate": 4.269029751107489e-10,
|
784 |
-
"logits/chosen": -1.
|
785 |
-
"logits/rejected": -1.
|
786 |
-
"logps/chosen": -
|
787 |
-
"logps/rejected": -
|
788 |
-
"loss": 0.
|
789 |
-
"rewards/accuracies": 0.
|
790 |
-
"rewards/chosen": 6.
|
791 |
-
"rewards/margins": 22.
|
792 |
-
"rewards/rejected": -15.
|
793 |
"step": 470
|
794 |
},
|
795 |
{
|
796 |
"epoch": 1.0,
|
797 |
"step": 478,
|
798 |
"total_flos": 0.0,
|
799 |
-
"train_loss": 0.
|
800 |
-
"train_runtime":
|
801 |
-
"train_samples_per_second": 8.
|
802 |
"train_steps_per_second": 0.063
|
803 |
}
|
804 |
],
|
|
|
10 |
"log_history": [
|
11 |
{
|
12 |
"epoch": 0.0,
|
13 |
+
"grad_norm": 2193.325330109938,
|
14 |
"learning_rate": 1.0416666666666666e-08,
|
15 |
"logits/chosen": -1.689455509185791,
|
16 |
"logits/rejected": -1.4794573783874512,
|
|
|
25 |
},
|
26 |
{
|
27 |
"epoch": 0.02,
|
28 |
+
"grad_norm": 1722.3750513007058,
|
29 |
"learning_rate": 1.0416666666666667e-07,
|
30 |
+
"logits/chosen": -1.7072248458862305,
|
31 |
+
"logits/rejected": -1.6097939014434814,
|
32 |
+
"logps/chosen": -139.610107421875,
|
33 |
+
"logps/rejected": -91.32671356201172,
|
34 |
+
"loss": 0.6992,
|
35 |
+
"rewards/accuracies": 0.4722222089767456,
|
36 |
+
"rewards/chosen": 0.05638541281223297,
|
37 |
+
"rewards/margins": 0.07169164717197418,
|
38 |
+
"rewards/rejected": -0.015306234359741211,
|
39 |
"step": 10
|
40 |
},
|
41 |
{
|
42 |
"epoch": 0.04,
|
43 |
+
"grad_norm": 506.4974128593864,
|
44 |
"learning_rate": 2.0833333333333333e-07,
|
45 |
+
"logits/chosen": -1.6386321783065796,
|
46 |
+
"logits/rejected": -1.6488521099090576,
|
47 |
+
"logps/chosen": -130.5824737548828,
|
48 |
+
"logps/rejected": -94.01884460449219,
|
49 |
+
"loss": 0.3992,
|
50 |
+
"rewards/accuracies": 0.824999988079071,
|
51 |
+
"rewards/chosen": 1.0460855960845947,
|
52 |
+
"rewards/margins": 1.4136831760406494,
|
53 |
+
"rewards/rejected": -0.3675975203514099,
|
54 |
"step": 20
|
55 |
},
|
56 |
{
|
57 |
"epoch": 0.06,
|
58 |
+
"grad_norm": 471.66591122190783,
|
59 |
"learning_rate": 3.1249999999999997e-07,
|
60 |
+
"logits/chosen": -1.6968982219696045,
|
61 |
+
"logits/rejected": -1.6314315795898438,
|
62 |
+
"logps/chosen": -131.55714416503906,
|
63 |
+
"logps/rejected": -105.11051177978516,
|
64 |
+
"loss": 0.2288,
|
65 |
+
"rewards/accuracies": 0.9312499761581421,
|
66 |
+
"rewards/chosen": 3.3202426433563232,
|
67 |
+
"rewards/margins": 5.000655174255371,
|
68 |
+
"rewards/rejected": -1.680413007736206,
|
69 |
"step": 30
|
70 |
},
|
71 |
{
|
72 |
"epoch": 0.08,
|
73 |
+
"grad_norm": 420.8471101103042,
|
74 |
"learning_rate": 4.1666666666666667e-07,
|
75 |
+
"logits/chosen": -1.6393680572509766,
|
76 |
+
"logits/rejected": -1.560689926147461,
|
77 |
+
"logps/chosen": -142.81134033203125,
|
78 |
+
"logps/rejected": -108.94673156738281,
|
79 |
+
"loss": 0.1804,
|
80 |
+
"rewards/accuracies": 0.9375,
|
81 |
+
"rewards/chosen": 4.454810619354248,
|
82 |
+
"rewards/margins": 9.642589569091797,
|
83 |
+
"rewards/rejected": -5.187777996063232,
|
84 |
"step": 40
|
85 |
},
|
86 |
{
|
87 |
"epoch": 0.1,
|
88 |
+
"grad_norm": 759.4656267159737,
|
89 |
"learning_rate": 4.999733114418725e-07,
|
90 |
+
"logits/chosen": -1.6069341897964478,
|
91 |
+
"logits/rejected": -1.631784439086914,
|
92 |
+
"logps/chosen": -126.6369400024414,
|
93 |
+
"logps/rejected": -116.30027770996094,
|
94 |
+
"loss": 0.1844,
|
95 |
+
"rewards/accuracies": 0.856249988079071,
|
96 |
+
"rewards/chosen": 4.2187395095825195,
|
97 |
+
"rewards/margins": 12.810827255249023,
|
98 |
+
"rewards/rejected": -8.59208869934082,
|
99 |
"step": 50
|
100 |
},
|
101 |
{
|
102 |
"epoch": 0.13,
|
103 |
+
"grad_norm": 361.53910661789246,
|
104 |
"learning_rate": 4.990398100856366e-07,
|
105 |
+
"logits/chosen": -1.7161242961883545,
|
106 |
+
"logits/rejected": -1.661266565322876,
|
107 |
+
"logps/chosen": -141.4559326171875,
|
108 |
+
"logps/rejected": -125.5120620727539,
|
109 |
+
"loss": 0.17,
|
110 |
"rewards/accuracies": 0.9437500238418579,
|
111 |
+
"rewards/chosen": 4.772133827209473,
|
112 |
+
"rewards/margins": 16.64890480041504,
|
113 |
+
"rewards/rejected": -11.876771926879883,
|
114 |
"step": 60
|
115 |
},
|
116 |
{
|
117 |
"epoch": 0.15,
|
118 |
+
"grad_norm": 382.2685121568309,
|
119 |
"learning_rate": 4.967775735898179e-07,
|
120 |
+
"logits/chosen": -1.670915961265564,
|
121 |
+
"logits/rejected": -1.6990423202514648,
|
122 |
+
"logps/chosen": -134.56948852539062,
|
123 |
+
"logps/rejected": -126.52008056640625,
|
124 |
+
"loss": 0.1751,
|
125 |
"rewards/accuracies": 0.925000011920929,
|
126 |
+
"rewards/chosen": 5.264647006988525,
|
127 |
+
"rewards/margins": 19.5419979095459,
|
128 |
+
"rewards/rejected": -14.277349472045898,
|
129 |
"step": 70
|
130 |
},
|
131 |
{
|
132 |
"epoch": 0.17,
|
133 |
+
"grad_norm": 337.7204736269515,
|
134 |
"learning_rate": 4.931986719649298e-07,
|
135 |
+
"logits/chosen": -1.658257246017456,
|
136 |
+
"logits/rejected": -1.6444251537322998,
|
137 |
+
"logps/chosen": -128.6245880126953,
|
138 |
+
"logps/rejected": -119.48954010009766,
|
139 |
+
"loss": 0.1656,
|
140 |
"rewards/accuracies": 0.9125000238418579,
|
141 |
+
"rewards/chosen": 5.917626857757568,
|
142 |
+
"rewards/margins": 18.79892349243164,
|
143 |
+
"rewards/rejected": -12.88129711151123,
|
144 |
"step": 80
|
145 |
},
|
146 |
{
|
147 |
"epoch": 0.19,
|
148 |
+
"grad_norm": 658.102190064697,
|
149 |
"learning_rate": 4.883222001996351e-07,
|
150 |
+
"logits/chosen": -1.6909303665161133,
|
151 |
+
"logits/rejected": -1.710211157798767,
|
152 |
+
"logps/chosen": -136.05978393554688,
|
153 |
+
"logps/rejected": -131.3511199951172,
|
154 |
+
"loss": 0.1561,
|
155 |
+
"rewards/accuracies": 0.925000011920929,
|
156 |
+
"rewards/chosen": 6.575238227844238,
|
157 |
+
"rewards/margins": 21.304401397705078,
|
158 |
+
"rewards/rejected": -14.729164123535156,
|
159 |
"step": 90
|
160 |
},
|
161 |
{
|
162 |
"epoch": 0.21,
|
163 |
+
"grad_norm": 303.0257944851465,
|
164 |
"learning_rate": 4.821741763807186e-07,
|
165 |
+
"logits/chosen": -1.739341139793396,
|
166 |
+
"logits/rejected": -1.722266435623169,
|
167 |
+
"logps/chosen": -116.35429382324219,
|
168 |
+
"logps/rejected": -117.0191421508789,
|
169 |
"loss": 0.1474,
|
170 |
+
"rewards/accuracies": 0.9437500238418579,
|
171 |
+
"rewards/chosen": 5.8117475509643555,
|
172 |
+
"rewards/margins": 18.994068145751953,
|
173 |
+
"rewards/rejected": -13.182321548461914,
|
174 |
"step": 100
|
175 |
},
|
176 |
{
|
177 |
"epoch": 0.21,
|
178 |
+
"eval_logits/chosen": -1.783400058746338,
|
179 |
+
"eval_logits/rejected": -1.7724130153656006,
|
180 |
+
"eval_logps/chosen": -124.02442932128906,
|
181 |
+
"eval_logps/rejected": -121.5073013305664,
|
182 |
+
"eval_loss": 0.13251493871212006,
|
183 |
+
"eval_rewards/accuracies": 0.9296875,
|
184 |
+
"eval_rewards/chosen": 6.873911380767822,
|
185 |
+
"eval_rewards/margins": 21.00718879699707,
|
186 |
+
"eval_rewards/rejected": -14.133277893066406,
|
187 |
+
"eval_runtime": 97.5707,
|
188 |
+
"eval_samples_per_second": 20.498,
|
189 |
+
"eval_steps_per_second": 0.328,
|
190 |
"step": 100
|
191 |
},
|
192 |
{
|
193 |
"epoch": 0.23,
|
194 |
+
"grad_norm": 311.0110467993666,
|
195 |
"learning_rate": 4.747874028753375e-07,
|
196 |
+
"logits/chosen": -1.6540296077728271,
|
197 |
+
"logits/rejected": -1.7143176794052124,
|
198 |
+
"logps/chosen": -120.4977035522461,
|
199 |
+
"logps/rejected": -124.83198547363281,
|
200 |
+
"loss": 0.1601,
|
201 |
+
"rewards/accuracies": 0.925000011920929,
|
202 |
+
"rewards/chosen": 5.774500370025635,
|
203 |
+
"rewards/margins": 17.103343963623047,
|
204 |
+
"rewards/rejected": -11.32884407043457,
|
205 |
"step": 110
|
206 |
},
|
207 |
{
|
208 |
"epoch": 0.25,
|
209 |
+
"grad_norm": 317.1259568395023,
|
210 |
"learning_rate": 4.662012913161997e-07,
|
211 |
+
"logits/chosen": -1.6986665725708008,
|
212 |
+
"logits/rejected": -1.6894630193710327,
|
213 |
+
"logps/chosen": -118.8506851196289,
|
214 |
+
"logps/rejected": -125.60331726074219,
|
215 |
+
"loss": 0.1522,
|
216 |
+
"rewards/accuracies": 0.8687499761581421,
|
217 |
+
"rewards/chosen": 6.9937944412231445,
|
218 |
+
"rewards/margins": 19.12407875061035,
|
219 |
+
"rewards/rejected": -12.130284309387207,
|
220 |
"step": 120
|
221 |
},
|
222 |
{
|
223 |
"epoch": 0.27,
|
224 |
+
"grad_norm": 183.0416909345485,
|
225 |
"learning_rate": 4.5646165232345103e-07,
|
226 |
+
"logits/chosen": -1.6666994094848633,
|
227 |
+
"logits/rejected": -1.6851320266723633,
|
228 |
+
"logps/chosen": -124.86369323730469,
|
229 |
+
"logps/rejected": -121.08885192871094,
|
230 |
+
"loss": 0.1572,
|
231 |
"rewards/accuracies": 0.918749988079071,
|
232 |
+
"rewards/chosen": 7.925422668457031,
|
233 |
+
"rewards/margins": 21.82646942138672,
|
234 |
+
"rewards/rejected": -13.901046752929688,
|
235 |
"step": 130
|
236 |
},
|
237 |
{
|
238 |
"epoch": 0.29,
|
239 |
+
"grad_norm": 323.2169573441021,
|
240 |
"learning_rate": 4.456204510851956e-07,
|
241 |
+
"logits/chosen": -1.5770366191864014,
|
242 |
+
"logits/rejected": -1.5510694980621338,
|
243 |
+
"logps/chosen": -122.3641128540039,
|
244 |
+
"logps/rejected": -117.8587646484375,
|
245 |
+
"loss": 0.1593,
|
246 |
+
"rewards/accuracies": 0.9125000238418579,
|
247 |
+
"rewards/chosen": 7.602517604827881,
|
248 |
+
"rewards/margins": 19.99860954284668,
|
249 |
+
"rewards/rejected": -12.396090507507324,
|
250 |
"step": 140
|
251 |
},
|
252 |
{
|
253 |
"epoch": 0.31,
|
254 |
+
"grad_norm": 394.76094620235625,
|
255 |
"learning_rate": 4.337355301007335e-07,
|
256 |
+
"logits/chosen": -1.717551827430725,
|
257 |
+
"logits/rejected": -1.7436870336532593,
|
258 |
+
"logps/chosen": -120.7677001953125,
|
259 |
+
"logps/rejected": -116.0407943725586,
|
260 |
+
"loss": 0.135,
|
261 |
+
"rewards/accuracies": 0.9125000238418579,
|
262 |
+
"rewards/chosen": 7.475555419921875,
|
263 |
+
"rewards/margins": 20.28847885131836,
|
264 |
+
"rewards/rejected": -12.812922477722168,
|
265 |
"step": 150
|
266 |
},
|
267 |
{
|
268 |
"epoch": 0.33,
|
269 |
+
"grad_norm": 696.9618085200943,
|
270 |
"learning_rate": 4.2087030056579986e-07,
|
271 |
+
"logits/chosen": -1.5951331853866577,
|
272 |
+
"logits/rejected": -1.5702898502349854,
|
273 |
+
"logps/chosen": -128.13714599609375,
|
274 |
+
"logps/rejected": -122.9395751953125,
|
275 |
+
"loss": 0.1778,
|
276 |
"rewards/accuracies": 0.9125000238418579,
|
277 |
+
"rewards/chosen": 6.914339542388916,
|
278 |
+
"rewards/margins": 18.91141700744629,
|
279 |
+
"rewards/rejected": -11.997075080871582,
|
280 |
"step": 160
|
281 |
},
|
282 |
{
|
283 |
"epoch": 0.36,
|
284 |
+
"grad_norm": 534.2420908517273,
|
285 |
"learning_rate": 4.070934040463998e-07,
|
286 |
+
"logits/chosen": -1.7780072689056396,
|
287 |
+
"logits/rejected": -1.7594010829925537,
|
288 |
+
"logps/chosen": -123.1135025024414,
|
289 |
+
"logps/rejected": -123.6323013305664,
|
290 |
+
"loss": 0.1817,
|
291 |
+
"rewards/accuracies": 0.918749988079071,
|
292 |
+
"rewards/chosen": 8.014005661010742,
|
293 |
+
"rewards/margins": 21.994348526000977,
|
294 |
+
"rewards/rejected": -13.980344772338867,
|
295 |
"step": 170
|
296 |
},
|
297 |
{
|
298 |
"epoch": 0.38,
|
299 |
+
"grad_norm": 515.866307384091,
|
300 |
"learning_rate": 3.9247834624635404e-07,
|
301 |
+
"logits/chosen": -1.7234785556793213,
|
302 |
+
"logits/rejected": -1.715887427330017,
|
303 |
+
"logps/chosen": -126.28702545166016,
|
304 |
+
"logps/rejected": -118.8428726196289,
|
305 |
+
"loss": 0.1945,
|
306 |
"rewards/accuracies": 0.918749988079071,
|
307 |
+
"rewards/chosen": 8.433076858520508,
|
308 |
+
"rewards/margins": 22.419658660888672,
|
309 |
+
"rewards/rejected": -13.986579895019531,
|
310 |
"step": 180
|
311 |
},
|
312 |
{
|
313 |
"epoch": 0.4,
|
314 |
+
"grad_norm": 442.14688359180957,
|
315 |
"learning_rate": 3.7710310482256523e-07,
|
316 |
+
"logits/chosen": -1.7688430547714233,
|
317 |
+
"logits/rejected": -1.7464864253997803,
|
318 |
+
"logps/chosen": -115.27445983886719,
|
319 |
+
"logps/rejected": -132.3951416015625,
|
320 |
+
"loss": 0.1843,
|
321 |
+
"rewards/accuracies": 0.887499988079071,
|
322 |
+
"rewards/chosen": 8.262548446655273,
|
323 |
+
"rewards/margins": 23.41249656677246,
|
324 |
+
"rewards/rejected": -15.149948120117188,
|
325 |
"step": 190
|
326 |
},
|
327 |
{
|
328 |
"epoch": 0.42,
|
329 |
+
"grad_norm": 436.2379767277873,
|
330 |
"learning_rate": 3.610497133404795e-07,
|
331 |
+
"logits/chosen": -1.7461057901382446,
|
332 |
+
"logits/rejected": -1.7516930103302002,
|
333 |
+
"logps/chosen": -121.12789154052734,
|
334 |
+
"logps/rejected": -116.14985656738281,
|
335 |
+
"loss": 0.2092,
|
336 |
"rewards/accuracies": 0.9437500238418579,
|
337 |
+
"rewards/chosen": 9.004249572753906,
|
338 |
+
"rewards/margins": 23.075119018554688,
|
339 |
+
"rewards/rejected": -14.070871353149414,
|
340 |
"step": 200
|
341 |
},
|
342 |
{
|
343 |
"epoch": 0.42,
|
344 |
+
"eval_logits/chosen": -1.8477407693862915,
|
345 |
+
"eval_logits/rejected": -1.844199538230896,
|
346 |
+
"eval_logps/chosen": -120.94989776611328,
|
347 |
+
"eval_logps/rejected": -123.32437896728516,
|
348 |
+
"eval_loss": 0.15667258203029633,
|
349 |
"eval_rewards/accuracies": 0.93359375,
|
350 |
+
"eval_rewards/chosen": 8.411178588867188,
|
351 |
+
"eval_rewards/margins": 23.453001022338867,
|
352 |
+
"eval_rewards/rejected": -15.04181957244873,
|
353 |
+
"eval_runtime": 97.6977,
|
354 |
+
"eval_samples_per_second": 20.471,
|
355 |
"eval_steps_per_second": 0.328,
|
356 |
"step": 200
|
357 |
},
|
358 |
{
|
359 |
"epoch": 0.44,
|
360 |
+
"grad_norm": 437.460750971005,
|
361 |
"learning_rate": 3.4440382358952115e-07,
|
362 |
+
"logits/chosen": -1.6800906658172607,
|
363 |
+
"logits/rejected": -1.6162738800048828,
|
364 |
+
"logps/chosen": -119.17988586425781,
|
365 |
+
"logps/rejected": -112.8397445678711,
|
366 |
+
"loss": 0.1771,
|
367 |
+
"rewards/accuracies": 0.8812500238418579,
|
368 |
+
"rewards/chosen": 6.705443382263184,
|
369 |
+
"rewards/margins": 18.32439613342285,
|
370 |
+
"rewards/rejected": -11.618951797485352,
|
371 |
"step": 210
|
372 |
},
|
373 |
{
|
374 |
"epoch": 0.46,
|
375 |
+
"grad_norm": 298.75015782601395,
|
376 |
"learning_rate": 3.272542485937368e-07,
|
377 |
+
"logits/chosen": -1.8080447912216187,
|
378 |
+
"logits/rejected": -1.8503248691558838,
|
379 |
+
"logps/chosen": -120.43497467041016,
|
380 |
+
"logps/rejected": -113.6927719116211,
|
381 |
+
"loss": 0.1853,
|
382 |
+
"rewards/accuracies": 0.925000011920929,
|
383 |
+
"rewards/chosen": 8.020153045654297,
|
384 |
+
"rewards/margins": 21.10222053527832,
|
385 |
+
"rewards/rejected": -13.082064628601074,
|
386 |
"step": 220
|
387 |
},
|
388 |
{
|
389 |
"epoch": 0.48,
|
390 |
+
"grad_norm": 290.6447474042267,
|
391 |
"learning_rate": 3.096924887558854e-07,
|
392 |
+
"logits/chosen": -1.7483365535736084,
|
393 |
+
"logits/rejected": -1.6977851390838623,
|
394 |
+
"logps/chosen": -132.64036560058594,
|
395 |
+
"logps/rejected": -132.7303009033203,
|
396 |
+
"loss": 0.2348,
|
397 |
+
"rewards/accuracies": 0.9375,
|
398 |
+
"rewards/chosen": 7.534226894378662,
|
399 |
+
"rewards/margins": 23.992801666259766,
|
400 |
+
"rewards/rejected": -16.458572387695312,
|
401 |
"step": 230
|
402 |
},
|
403 |
{
|
404 |
"epoch": 0.5,
|
405 |
+
"grad_norm": 435.0583755248922,
|
406 |
"learning_rate": 2.9181224366319943e-07,
|
407 |
+
"logits/chosen": -1.7618516683578491,
|
408 |
+
"logits/rejected": -1.7764402627944946,
|
409 |
+
"logps/chosen": -119.7364273071289,
|
410 |
+
"logps/rejected": -124.50996398925781,
|
411 |
+
"loss": 0.2076,
|
412 |
+
"rewards/accuracies": 0.8999999761581421,
|
413 |
+
"rewards/chosen": 6.094353675842285,
|
414 |
+
"rewards/margins": 19.534603118896484,
|
415 |
+
"rewards/rejected": -13.440248489379883,
|
416 |
"step": 240
|
417 |
},
|
418 |
{
|
419 |
"epoch": 0.52,
|
420 |
+
"grad_norm": 506.7749182741835,
|
421 |
"learning_rate": 2.7370891215954565e-07,
|
422 |
+
"logits/chosen": -1.7425343990325928,
|
423 |
+
"logits/rejected": -1.7735121250152588,
|
424 |
+
"logps/chosen": -123.32588195800781,
|
425 |
+
"logps/rejected": -126.53437805175781,
|
426 |
+
"loss": 0.2561,
|
427 |
"rewards/accuracies": 0.9312499761581421,
|
428 |
+
"rewards/chosen": 7.04464054107666,
|
429 |
+
"rewards/margins": 21.820133209228516,
|
430 |
+
"rewards/rejected": -14.775495529174805,
|
431 |
"step": 250
|
432 |
},
|
433 |
{
|
434 |
"epoch": 0.54,
|
435 |
+
"grad_norm": 278.0299175856851,
|
436 |
"learning_rate": 2.55479083351317e-07,
|
437 |
+
"logits/chosen": -1.7782995700836182,
|
438 |
+
"logits/rejected": -1.7973874807357788,
|
439 |
+
"logps/chosen": -127.03205871582031,
|
440 |
+
"logps/rejected": -114.83714294433594,
|
441 |
+
"loss": 0.1749,
|
442 |
+
"rewards/accuracies": 0.887499988079071,
|
443 |
+
"rewards/chosen": 8.397477149963379,
|
444 |
+
"rewards/margins": 23.37790298461914,
|
445 |
+
"rewards/rejected": -14.980427742004395,
|
446 |
"step": 260
|
447 |
},
|
448 |
{
|
449 |
"epoch": 0.56,
|
450 |
+
"grad_norm": 372.30589353402263,
|
451 |
"learning_rate": 2.3722002126275822e-07,
|
452 |
+
"logits/chosen": -1.7808539867401123,
|
453 |
+
"logits/rejected": -1.719679832458496,
|
454 |
+
"logps/chosen": -117.4173355102539,
|
455 |
+
"logps/rejected": -119.30290222167969,
|
456 |
+
"loss": 0.1772,
|
457 |
+
"rewards/accuracies": 0.893750011920929,
|
458 |
+
"rewards/chosen": 7.320569038391113,
|
459 |
+
"rewards/margins": 21.35511589050293,
|
460 |
+
"rewards/rejected": -14.034547805786133,
|
461 |
"step": 270
|
462 |
},
|
463 |
{
|
464 |
"epoch": 0.59,
|
465 |
+
"grad_norm": 470.22249240997917,
|
466 |
"learning_rate": 2.19029145890313e-07,
|
467 |
+
"logits/chosen": -1.6393606662750244,
|
468 |
+
"logits/rejected": -1.7276159524917603,
|
469 |
+
"logps/chosen": -123.42005920410156,
|
470 |
+
"logps/rejected": -132.44947814941406,
|
471 |
+
"loss": 0.2305,
|
472 |
+
"rewards/accuracies": 0.918749988079071,
|
473 |
+
"rewards/chosen": 6.818305969238281,
|
474 |
+
"rewards/margins": 22.368270874023438,
|
475 |
+
"rewards/rejected": -15.549966812133789,
|
476 |
"step": 280
|
477 |
},
|
478 |
{
|
479 |
"epoch": 0.61,
|
480 |
+
"grad_norm": 652.1091329463337,
|
481 |
"learning_rate": 2.0100351342479216e-07,
|
482 |
+
"logits/chosen": -1.7318487167358398,
|
483 |
+
"logits/rejected": -1.706648826599121,
|
484 |
+
"logps/chosen": -114.55335998535156,
|
485 |
+
"logps/rejected": -119.69996643066406,
|
486 |
+
"loss": 0.2179,
|
487 |
+
"rewards/accuracies": 0.887499988079071,
|
488 |
+
"rewards/chosen": 6.804747104644775,
|
489 |
+
"rewards/margins": 22.096506118774414,
|
490 |
+
"rewards/rejected": -15.29175853729248,
|
491 |
"step": 290
|
492 |
},
|
493 |
{
|
494 |
"epoch": 0.63,
|
495 |
+
"grad_norm": 426.10148588025714,
|
496 |
"learning_rate": 1.8323929841460178e-07,
|
497 |
+
"logits/chosen": -1.699927568435669,
|
498 |
+
"logits/rejected": -1.6480505466461182,
|
499 |
+
"logps/chosen": -132.07632446289062,
|
500 |
+
"logps/rejected": -136.4704132080078,
|
501 |
+
"loss": 0.1925,
|
502 |
+
"rewards/accuracies": 0.925000011920929,
|
503 |
+
"rewards/chosen": 7.1819257736206055,
|
504 |
+
"rewards/margins": 22.29709243774414,
|
505 |
+
"rewards/rejected": -15.11516284942627,
|
506 |
"step": 300
|
507 |
},
|
508 |
{
|
509 |
"epoch": 0.63,
|
510 |
+
"eval_logits/chosen": -1.8054125308990479,
|
511 |
+
"eval_logits/rejected": -1.804688572883606,
|
512 |
+
"eval_logps/chosen": -122.28070068359375,
|
513 |
+
"eval_logps/rejected": -126.64250946044922,
|
514 |
+
"eval_loss": 0.17152084410190582,
|
515 |
"eval_rewards/accuracies": 0.92578125,
|
516 |
+
"eval_rewards/chosen": 7.745772361755371,
|
517 |
+
"eval_rewards/margins": 24.44666290283203,
|
518 |
+
"eval_rewards/rejected": -16.700889587402344,
|
519 |
+
"eval_runtime": 97.5667,
|
520 |
+
"eval_samples_per_second": 20.499,
|
521 |
"eval_steps_per_second": 0.328,
|
522 |
"step": 300
|
523 |
},
|
524 |
{
|
525 |
"epoch": 0.65,
|
526 |
+
"grad_norm": 499.48262107192954,
|
527 |
"learning_rate": 1.6583128063291573e-07,
|
528 |
+
"logits/chosen": -1.7035776376724243,
|
529 |
+
"logits/rejected": -1.73648202419281,
|
530 |
+
"logps/chosen": -123.8221664428711,
|
531 |
+
"logps/rejected": -127.0335464477539,
|
532 |
+
"loss": 0.1993,
|
533 |
+
"rewards/accuracies": 0.9312499761581421,
|
534 |
+
"rewards/chosen": 7.757756233215332,
|
535 |
+
"rewards/margins": 22.004987716674805,
|
536 |
+
"rewards/rejected": -14.247230529785156,
|
537 |
"step": 310
|
538 |
},
|
539 |
{
|
540 |
"epoch": 0.67,
|
541 |
+
"grad_norm": 351.5168221076795,
|
542 |
"learning_rate": 1.488723393865766e-07,
|
543 |
+
"logits/chosen": -1.7380039691925049,
|
544 |
+
"logits/rejected": -1.7079343795776367,
|
545 |
+
"logps/chosen": -113.3768081665039,
|
546 |
+
"logps/rejected": -126.8635482788086,
|
547 |
+
"loss": 0.1891,
|
548 |
"rewards/accuracies": 0.887499988079071,
|
549 |
+
"rewards/chosen": 8.71003532409668,
|
550 |
+
"rewards/margins": 24.014118194580078,
|
551 |
+
"rewards/rejected": -15.304081916809082,
|
552 |
"step": 320
|
553 |
},
|
554 |
{
|
555 |
"epoch": 0.69,
|
556 |
+
"grad_norm": 321.010971408508,
|
557 |
"learning_rate": 1.3245295796480788e-07,
|
558 |
+
"logits/chosen": -1.664240837097168,
|
559 |
+
"logits/rejected": -1.7476810216903687,
|
560 |
+
"logps/chosen": -123.50030517578125,
|
561 |
+
"logps/rejected": -122.0934066772461,
|
562 |
+
"loss": 0.1955,
|
563 |
+
"rewards/accuracies": 0.90625,
|
564 |
+
"rewards/chosen": 7.341310024261475,
|
565 |
+
"rewards/margins": 21.513980865478516,
|
566 |
+
"rewards/rejected": -14.172670364379883,
|
567 |
"step": 330
|
568 |
},
|
569 |
{
|
570 |
"epoch": 0.71,
|
571 |
+
"grad_norm": 251.55976817411812,
|
572 |
"learning_rate": 1.1666074087171627e-07,
|
573 |
+
"logits/chosen": -1.6232049465179443,
|
574 |
+
"logits/rejected": -1.6956592798233032,
|
575 |
+
"logps/chosen": -117.71810150146484,
|
576 |
+
"logps/rejected": -113.10723876953125,
|
577 |
+
"loss": 0.2458,
|
578 |
+
"rewards/accuracies": 0.893750011920929,
|
579 |
+
"rewards/chosen": 6.386099815368652,
|
580 |
+
"rewards/margins": 19.121448516845703,
|
581 |
+
"rewards/rejected": -12.73534870147705,
|
582 |
"step": 340
|
583 |
},
|
584 |
{
|
585 |
"epoch": 0.73,
|
586 |
+
"grad_norm": 784.7175940461042,
|
587 |
"learning_rate": 1.0157994641835734e-07,
|
588 |
+
"logits/chosen": -1.6620721817016602,
|
589 |
+
"logits/rejected": -1.6772327423095703,
|
590 |
+
"logps/chosen": -117.80794525146484,
|
591 |
+
"logps/rejected": -132.93991088867188,
|
592 |
+
"loss": 0.1649,
|
593 |
+
"rewards/accuracies": 0.918749988079071,
|
594 |
+
"rewards/chosen": 6.138510704040527,
|
595 |
+
"rewards/margins": 21.829065322875977,
|
596 |
+
"rewards/rejected": -15.690553665161133,
|
597 |
"step": 350
|
598 |
},
|
599 |
{
|
600 |
"epoch": 0.75,
|
601 |
+
"grad_norm": 497.1984887279012,
|
602 |
"learning_rate": 8.729103716819111e-08,
|
603 |
+
"logits/chosen": -1.6556260585784912,
|
604 |
+
"logits/rejected": -1.6805181503295898,
|
605 |
+
"logps/chosen": -124.04801177978516,
|
606 |
+
"logps/rejected": -128.13685607910156,
|
607 |
+
"loss": 0.1743,
|
608 |
+
"rewards/accuracies": 0.856249988079071,
|
609 |
+
"rewards/chosen": 7.051167964935303,
|
610 |
+
"rewards/margins": 21.460834503173828,
|
611 |
+
"rewards/rejected": -14.40966796875,
|
612 |
"step": 360
|
613 |
},
|
614 |
{
|
615 |
"epoch": 0.77,
|
616 |
+
"grad_norm": 701.6355853905161,
|
617 |
"learning_rate": 7.387025063449081e-08,
|
618 |
+
"logits/chosen": -1.7627124786376953,
|
619 |
+
"logits/rejected": -1.7043575048446655,
|
620 |
+
"logps/chosen": -113.8396224975586,
|
621 |
+
"logps/rejected": -117.89778137207031,
|
622 |
+
"loss": 0.1661,
|
623 |
"rewards/accuracies": 0.9312499761581421,
|
624 |
+
"rewards/chosen": 6.275058269500732,
|
625 |
+
"rewards/margins": 18.96691131591797,
|
626 |
+
"rewards/rejected": -12.691854476928711,
|
627 |
"step": 370
|
628 |
},
|
629 |
{
|
630 |
"epoch": 0.79,
|
631 |
+
"grad_norm": 384.3232766408438,
|
632 |
"learning_rate": 6.138919252022435e-08,
|
633 |
+
"logits/chosen": -1.7891247272491455,
|
634 |
+
"logits/rejected": -1.8081142902374268,
|
635 |
+
"logps/chosen": -121.1161880493164,
|
636 |
+
"logps/rejected": -131.83599853515625,
|
637 |
+
"loss": 0.1823,
|
638 |
"rewards/accuracies": 0.9375,
|
639 |
+
"rewards/chosen": 7.768773078918457,
|
640 |
+
"rewards/margins": 24.082279205322266,
|
641 |
+
"rewards/rejected": -16.31351089477539,
|
642 |
"step": 380
|
643 |
},
|
644 |
{
|
645 |
"epoch": 0.82,
|
646 |
+
"grad_norm": 419.5923423572011,
|
647 |
"learning_rate": 4.991445467064689e-08,
|
648 |
+
"logits/chosen": -1.6696172952651978,
|
649 |
+
"logits/rejected": -1.6762924194335938,
|
650 |
+
"logps/chosen": -117.07814025878906,
|
651 |
+
"logps/rejected": -120.15535736083984,
|
652 |
+
"loss": 0.1486,
|
653 |
"rewards/accuracies": 0.9624999761581421,
|
654 |
+
"rewards/chosen": 7.70883846282959,
|
655 |
+
"rewards/margins": 22.634441375732422,
|
656 |
+
"rewards/rejected": -14.9256010055542,
|
657 |
"step": 390
|
658 |
},
|
659 |
{
|
660 |
"epoch": 0.84,
|
661 |
+
"grad_norm": 548.5530090451983,
|
662 |
"learning_rate": 3.9507259776993954e-08,
|
663 |
+
"logits/chosen": -1.6570144891738892,
|
664 |
+
"logits/rejected": -1.7342147827148438,
|
665 |
+
"logps/chosen": -119.59893798828125,
|
666 |
+
"logps/rejected": -120.49696350097656,
|
667 |
+
"loss": 0.2762,
|
668 |
+
"rewards/accuracies": 0.90625,
|
669 |
+
"rewards/chosen": 6.581268310546875,
|
670 |
+
"rewards/margins": 21.039697647094727,
|
671 |
+
"rewards/rejected": -14.4584321975708,
|
672 |
"step": 400
|
673 |
},
|
674 |
{
|
675 |
"epoch": 0.84,
|
676 |
+
"eval_logits/chosen": -1.8335998058319092,
|
677 |
+
"eval_logits/rejected": -1.8335609436035156,
|
678 |
+
"eval_logps/chosen": -121.38908386230469,
|
679 |
+
"eval_logps/rejected": -126.29609680175781,
|
680 |
+
"eval_loss": 0.16946831345558167,
|
681 |
+
"eval_rewards/accuracies": 0.9296875,
|
682 |
+
"eval_rewards/chosen": 8.191582679748535,
|
683 |
+
"eval_rewards/margins": 24.719266891479492,
|
684 |
+
"eval_rewards/rejected": -16.527685165405273,
|
685 |
+
"eval_runtime": 97.6978,
|
686 |
+
"eval_samples_per_second": 20.471,
|
687 |
+
"eval_steps_per_second": 0.328,
|
688 |
"step": 400
|
689 |
},
|
690 |
{
|
691 |
"epoch": 0.86,
|
692 |
+
"grad_norm": 482.27830673225736,
|
693 |
"learning_rate": 3.022313472693447e-08,
|
694 |
+
"logits/chosen": -1.7594387531280518,
|
695 |
+
"logits/rejected": -1.7884505987167358,
|
696 |
+
"logps/chosen": -130.15711975097656,
|
697 |
+
"logps/rejected": -124.42222595214844,
|
698 |
+
"loss": 0.2025,
|
699 |
+
"rewards/accuracies": 0.9125000238418579,
|
700 |
+
"rewards/chosen": 8.578957557678223,
|
701 |
+
"rewards/margins": 24.737140655517578,
|
702 |
+
"rewards/rejected": -16.158184051513672,
|
703 |
"step": 410
|
704 |
},
|
705 |
{
|
706 |
"epoch": 0.88,
|
707 |
+
"grad_norm": 378.46967638932284,
|
708 |
"learning_rate": 2.2111614344599684e-08,
|
709 |
+
"logits/chosen": -1.7343389987945557,
|
710 |
+
"logits/rejected": -1.731032371520996,
|
711 |
+
"logps/chosen": -115.0967025756836,
|
712 |
+
"logps/rejected": -119.22444915771484,
|
713 |
+
"loss": 0.1669,
|
714 |
+
"rewards/accuracies": 0.9312499761581421,
|
715 |
+
"rewards/chosen": 6.606478691101074,
|
716 |
+
"rewards/margins": 20.886428833007812,
|
717 |
+
"rewards/rejected": -14.279950141906738,
|
718 |
"step": 420
|
719 |
},
|
720 |
{
|
721 |
"epoch": 0.9,
|
722 |
+
"grad_norm": 500.1474042235916,
|
723 |
"learning_rate": 1.521597710086439e-08,
|
724 |
+
"logits/chosen": -1.6852420568466187,
|
725 |
+
"logits/rejected": -1.7072862386703491,
|
726 |
+
"logps/chosen": -131.11505126953125,
|
727 |
+
"logps/rejected": -119.86601257324219,
|
728 |
+
"loss": 0.1861,
|
729 |
"rewards/accuracies": 0.9125000238418579,
|
730 |
+
"rewards/chosen": 7.49709939956665,
|
731 |
+
"rewards/margins": 20.68734359741211,
|
732 |
+
"rewards/rejected": -13.190244674682617,
|
733 |
"step": 430
|
734 |
},
|
735 |
{
|
736 |
"epoch": 0.92,
|
737 |
+
"grad_norm": 310.90711668513416,
|
738 |
"learning_rate": 9.57301420397924e-09,
|
739 |
+
"logits/chosen": -1.8146944046020508,
|
740 |
+
"logits/rejected": -1.8261439800262451,
|
741 |
+
"logps/chosen": -122.62892150878906,
|
742 |
+
"logps/rejected": -125.58287048339844,
|
743 |
+
"loss": 0.1762,
|
744 |
"rewards/accuracies": 0.9375,
|
745 |
+
"rewards/chosen": 8.172103881835938,
|
746 |
+
"rewards/margins": 24.26197624206543,
|
747 |
+
"rewards/rejected": -16.089872360229492,
|
748 |
"step": 440
|
749 |
},
|
750 |
{
|
751 |
"epoch": 0.94,
|
752 |
+
"grad_norm": 445.19377861997515,
|
753 |
"learning_rate": 5.212833302556258e-09,
|
754 |
+
"logits/chosen": -1.8277533054351807,
|
755 |
+
"logits/rejected": -1.7913105487823486,
|
756 |
+
"logps/chosen": -117.95567321777344,
|
757 |
+
"logps/rejected": -121.9552001953125,
|
758 |
+
"loss": 0.2023,
|
759 |
"rewards/accuracies": 0.90625,
|
760 |
+
"rewards/chosen": 8.675628662109375,
|
761 |
+
"rewards/margins": 25.467309951782227,
|
762 |
+
"rewards/rejected": -16.79167938232422,
|
763 |
"step": 450
|
764 |
},
|
765 |
{
|
766 |
"epoch": 0.96,
|
767 |
+
"grad_norm": 209.63658036103772,
|
768 |
"learning_rate": 2.158697848236607e-09,
|
769 |
+
"logits/chosen": -1.7152612209320068,
|
770 |
+
"logits/rejected": -1.7372353076934814,
|
771 |
+
"logps/chosen": -122.6450424194336,
|
772 |
+
"logps/rejected": -125.43870544433594,
|
773 |
+
"loss": 0.1507,
|
774 |
+
"rewards/accuracies": 0.9375,
|
775 |
+
"rewards/chosen": 8.354387283325195,
|
776 |
+
"rewards/margins": 23.259435653686523,
|
777 |
+
"rewards/rejected": -14.905046463012695,
|
778 |
"step": 460
|
779 |
},
|
780 |
{
|
781 |
"epoch": 0.98,
|
782 |
+
"grad_norm": 282.1159277085459,
|
783 |
"learning_rate": 4.269029751107489e-10,
|
784 |
+
"logits/chosen": -1.691461205482483,
|
785 |
+
"logits/rejected": -1.706369161605835,
|
786 |
+
"logps/chosen": -116.26470947265625,
|
787 |
+
"logps/rejected": -135.1427459716797,
|
788 |
+
"loss": 0.1833,
|
789 |
+
"rewards/accuracies": 0.918749988079071,
|
790 |
+
"rewards/chosen": 6.723459720611572,
|
791 |
+
"rewards/margins": 22.407346725463867,
|
792 |
+
"rewards/rejected": -15.683886528015137,
|
793 |
"step": 470
|
794 |
},
|
795 |
{
|
796 |
"epoch": 1.0,
|
797 |
"step": 478,
|
798 |
"total_flos": 0.0,
|
799 |
+
"train_loss": 0.20260818084413537,
|
800 |
+
"train_runtime": 7642.1158,
|
801 |
+
"train_samples_per_second": 8.0,
|
802 |
"train_steps_per_second": 0.063
|
803 |
}
|
804 |
],
|