hugodk-sch
commited on
Commit
•
450a24e
1
Parent(s):
f42efa5
Model save
Browse files- README.md +14 -16
- adapter_config.json +4 -4
- adapter_model.safetensors +1 -1
- all_results.json +16 -16
- eval_results.json +12 -12
- train_results.json +4 -4
- trainer_state.json +373 -373
- training_args.bin +1 -1
README.md
CHANGED
@@ -1,13 +1,11 @@
|
|
1 |
---
|
2 |
library_name: peft
|
3 |
tags:
|
4 |
-
- alignment-handbook
|
5 |
- trl
|
6 |
- dpo
|
|
|
7 |
- generated_from_trainer
|
8 |
base_model: NbAiLab/nb-gpt-j-6B-v2
|
9 |
-
datasets:
|
10 |
-
- hugodk-sch/aftonposten_title_prefs
|
11 |
model-index:
|
12 |
- name: aftonposten-6b-align-scan
|
13 |
results: []
|
@@ -18,17 +16,17 @@ should probably proofread and complete it, then remove this comment. -->
|
|
18 |
|
19 |
# aftonposten-6b-align-scan
|
20 |
|
21 |
-
This model is a fine-tuned version of [
|
22 |
It achieves the following results on the evaluation set:
|
23 |
-
-
|
24 |
-
-
|
25 |
-
-
|
26 |
-
-
|
27 |
-
-
|
28 |
-
-
|
29 |
-
-
|
30 |
-
-
|
31 |
-
-
|
32 |
|
33 |
## Model description
|
34 |
|
@@ -63,9 +61,9 @@ The following hyperparameters were used during training:
|
|
63 |
|
64 |
| Training Loss | Epoch | Step | Logits/chosen | Logits/rejected | Logps/chosen | Logps/rejected | Validation Loss | Rewards/accuracies | Rewards/chosen | Rewards/margins | Rewards/rejected |
|
65 |
|:-------------:|:-----:|:----:|:-------------:|:---------------:|:------------:|:--------------:|:---------------:|:------------------:|:--------------:|:---------------:|:----------------:|
|
66 |
-
| 0.
|
67 |
-
| 0.
|
68 |
-
| 0.
|
69 |
|
70 |
|
71 |
### Framework versions
|
|
|
1 |
---
|
2 |
library_name: peft
|
3 |
tags:
|
|
|
4 |
- trl
|
5 |
- dpo
|
6 |
+
- alignment-handbook
|
7 |
- generated_from_trainer
|
8 |
base_model: NbAiLab/nb-gpt-j-6B-v2
|
|
|
|
|
9 |
model-index:
|
10 |
- name: aftonposten-6b-align-scan
|
11 |
results: []
|
|
|
16 |
|
17 |
# aftonposten-6b-align-scan
|
18 |
|
19 |
+
This model is a fine-tuned version of [NbAiLab/nb-gpt-j-6B-v2](https://huggingface.co/NbAiLab/nb-gpt-j-6B-v2) on an unknown dataset.
|
20 |
It achieves the following results on the evaluation set:
|
21 |
+
- Logits/chosen: -2.2439
|
22 |
+
- Logits/rejected: -2.2390
|
23 |
+
- Logps/chosen: -34.0176
|
24 |
+
- Logps/rejected: -37.5142
|
25 |
+
- Loss: 0.3276
|
26 |
+
- Rewards/accuracies: 0.5365
|
27 |
+
- Rewards/chosen: 0.0152
|
28 |
+
- Rewards/margins: 0.0131
|
29 |
+
- Rewards/rejected: 0.0022
|
30 |
|
31 |
## Model description
|
32 |
|
|
|
61 |
|
62 |
| Training Loss | Epoch | Step | Logits/chosen | Logits/rejected | Logps/chosen | Logps/rejected | Validation Loss | Rewards/accuracies | Rewards/chosen | Rewards/margins | Rewards/rejected |
|
63 |
|:-------------:|:-----:|:----:|:-------------:|:---------------:|:------------:|:--------------:|:---------------:|:------------------:|:--------------:|:---------------:|:----------------:|
|
64 |
+
| 0.333 | 0.26 | 100 | -2.2438 | -2.2389 | -34.0205 | -37.5035 | 0.3405 | 0.5104 | 0.0126 | 0.0008 | 0.0118 |
|
65 |
+
| 0.3306 | 0.52 | 200 | -2.2435 | -2.2386 | -34.0285 | -37.5113 | 0.3466 | 0.5021 | 0.0054 | 0.0007 | 0.0048 |
|
66 |
+
| 0.3327 | 0.78 | 300 | -2.2439 | -2.2390 | -34.0176 | -37.5142 | 0.3276 | 0.5365 | 0.0152 | 0.0131 | 0.0022 |
|
67 |
|
68 |
|
69 |
### Framework versions
|
adapter_config.json
CHANGED
@@ -19,13 +19,13 @@
|
|
19 |
"rank_pattern": {},
|
20 |
"revision": null,
|
21 |
"target_modules": [
|
22 |
-
"gate_proj",
|
23 |
"o_proj",
|
24 |
-
"
|
25 |
-
"
|
26 |
"down_proj",
|
27 |
"v_proj",
|
28 |
-
"
|
|
|
29 |
],
|
30 |
"task_type": "CAUSAL_LM",
|
31 |
"use_rslora": false
|
|
|
19 |
"rank_pattern": {},
|
20 |
"revision": null,
|
21 |
"target_modules": [
|
|
|
22 |
"o_proj",
|
23 |
+
"gate_proj",
|
24 |
+
"q_proj",
|
25 |
"down_proj",
|
26 |
"v_proj",
|
27 |
+
"k_proj",
|
28 |
+
"up_proj"
|
29 |
],
|
30 |
"task_type": "CAUSAL_LM",
|
31 |
"use_rslora": false
|
adapter_model.safetensors
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 176183216
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:28810f10adaabc08f14ec259ce0b4eff31f65380fdc92f6051a2568fe462f659
|
3 |
size 176183216
|
all_results.json
CHANGED
@@ -1,21 +1,21 @@
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
-
"eval_logits/chosen": -2.
|
4 |
-
"eval_logits/rejected": -2.
|
5 |
-
"eval_logps/chosen": -
|
6 |
-
"eval_logps/rejected": -37.
|
7 |
-
"eval_loss": 0.
|
8 |
-
"eval_rewards/accuracies": 0.
|
9 |
-
"eval_rewards/chosen": 0.
|
10 |
-
"eval_rewards/margins":
|
11 |
-
"eval_rewards/rejected": 0.
|
12 |
-
"eval_runtime":
|
13 |
"eval_samples": 343,
|
14 |
-
"eval_samples_per_second": 2.
|
15 |
-
"eval_steps_per_second": 0.
|
16 |
-
"train_loss": 0.
|
17 |
-
"train_runtime":
|
18 |
"train_samples": 3079,
|
19 |
-
"train_samples_per_second": 4.
|
20 |
-
"train_steps_per_second": 0.
|
21 |
}
|
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
+
"eval_logits/chosen": -2.2436861991882324,
|
4 |
+
"eval_logits/rejected": -2.2388129234313965,
|
5 |
+
"eval_logps/chosen": -34.01886749267578,
|
6 |
+
"eval_logps/rejected": -37.50727081298828,
|
7 |
+
"eval_loss": 0.34474730491638184,
|
8 |
+
"eval_rewards/accuracies": 0.5161960124969482,
|
9 |
+
"eval_rewards/chosen": 0.01411630492657423,
|
10 |
+
"eval_rewards/margins": 0.005696756765246391,
|
11 |
+
"eval_rewards/rejected": 0.008419547230005264,
|
12 |
+
"eval_runtime": 145.8183,
|
13 |
"eval_samples": 343,
|
14 |
+
"eval_samples_per_second": 2.352,
|
15 |
+
"eval_steps_per_second": 0.295,
|
16 |
+
"train_loss": 0.0744757689438857,
|
17 |
+
"train_runtime": 628.0651,
|
18 |
"train_samples": 3079,
|
19 |
+
"train_samples_per_second": 4.902,
|
20 |
+
"train_steps_per_second": 0.613
|
21 |
}
|
eval_results.json
CHANGED
@@ -1,16 +1,16 @@
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
-
"eval_logits/chosen": -2.
|
4 |
-
"eval_logits/rejected": -2.
|
5 |
-
"eval_logps/chosen": -
|
6 |
-
"eval_logps/rejected": -37.
|
7 |
-
"eval_loss": 0.
|
8 |
-
"eval_rewards/accuracies": 0.
|
9 |
-
"eval_rewards/chosen": 0.
|
10 |
-
"eval_rewards/margins":
|
11 |
-
"eval_rewards/rejected": 0.
|
12 |
-
"eval_runtime":
|
13 |
"eval_samples": 343,
|
14 |
-
"eval_samples_per_second": 2.
|
15 |
-
"eval_steps_per_second": 0.
|
16 |
}
|
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
+
"eval_logits/chosen": -2.2436861991882324,
|
4 |
+
"eval_logits/rejected": -2.2388129234313965,
|
5 |
+
"eval_logps/chosen": -34.01886749267578,
|
6 |
+
"eval_logps/rejected": -37.50727081298828,
|
7 |
+
"eval_loss": 0.34474730491638184,
|
8 |
+
"eval_rewards/accuracies": 0.5161960124969482,
|
9 |
+
"eval_rewards/chosen": 0.01411630492657423,
|
10 |
+
"eval_rewards/margins": 0.005696756765246391,
|
11 |
+
"eval_rewards/rejected": 0.008419547230005264,
|
12 |
+
"eval_runtime": 145.8183,
|
13 |
"eval_samples": 343,
|
14 |
+
"eval_samples_per_second": 2.352,
|
15 |
+
"eval_steps_per_second": 0.295
|
16 |
}
|
train_results.json
CHANGED
@@ -1,8 +1,8 @@
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
-
"train_loss": 0.
|
4 |
-
"train_runtime":
|
5 |
"train_samples": 3079,
|
6 |
-
"train_samples_per_second": 4.
|
7 |
-
"train_steps_per_second": 0.
|
8 |
}
|
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
+
"train_loss": 0.0744757689438857,
|
4 |
+
"train_runtime": 628.0651,
|
5 |
"train_samples": 3079,
|
6 |
+
"train_samples_per_second": 4.902,
|
7 |
+
"train_steps_per_second": 0.613
|
8 |
}
|
trainer_state.json
CHANGED
@@ -15,7 +15,7 @@
|
|
15 |
"logits/rejected": -1.7377450466156006,
|
16 |
"logps/chosen": -29.553977966308594,
|
17 |
"logps/rejected": -42.813133239746094,
|
18 |
-
"loss": 0.
|
19 |
"rewards/accuracies": 0.0,
|
20 |
"rewards/chosen": 0.0,
|
21 |
"rewards/margins": 0.0,
|
@@ -25,591 +25,591 @@
|
|
25 |
{
|
26 |
"epoch": 0.03,
|
27 |
"learning_rate": 1.2820512820512818e-07,
|
28 |
-
"logits/chosen": -1.
|
29 |
-
"logits/rejected": -1.
|
30 |
-
"logps/chosen": -36.
|
31 |
-
"logps/rejected": -33.
|
32 |
-
"loss": 0.
|
33 |
-
"rewards/accuracies": 0.
|
34 |
-
"rewards/chosen": 0.
|
35 |
-
"rewards/margins": 0.
|
36 |
-
"rewards/rejected": -0.
|
37 |
"step": 10
|
38 |
},
|
39 |
{
|
40 |
"epoch": 0.05,
|
41 |
"learning_rate": 2.5641025641025636e-07,
|
42 |
-
"logits/chosen": -1.
|
43 |
-
"logits/rejected": -1.
|
44 |
-
"logps/chosen": -29.
|
45 |
-
"logps/rejected": -29.
|
46 |
-
"loss": 0.
|
47 |
-
"rewards/accuracies": 0.
|
48 |
-
"rewards/chosen": -0.
|
49 |
-
"rewards/margins": -0.
|
50 |
-
"rewards/rejected": 0.
|
51 |
"step": 20
|
52 |
},
|
53 |
{
|
54 |
"epoch": 0.08,
|
55 |
"learning_rate": 3.8461538461538463e-07,
|
56 |
-
"logits/chosen": -1.
|
57 |
-
"logits/rejected": -1.
|
58 |
-
"logps/chosen": -31.
|
59 |
-
"logps/rejected": -33.
|
60 |
-
"loss": 0.
|
61 |
-
"rewards/accuracies": 0.
|
62 |
-
"rewards/chosen": 0.
|
63 |
-
"rewards/margins": 0.
|
64 |
-
"rewards/rejected": 0.
|
65 |
"step": 30
|
66 |
},
|
67 |
{
|
68 |
"epoch": 0.1,
|
69 |
"learning_rate": 4.999896948438433e-07,
|
70 |
-
"logits/chosen": -2.
|
71 |
-
"logits/rejected": -2.
|
72 |
-
"logps/chosen": -32.
|
73 |
-
"logps/rejected": -32.
|
74 |
-
"loss": 0.
|
75 |
-
"rewards/accuracies": 0.
|
76 |
-
"rewards/chosen": -0.
|
77 |
-
"rewards/margins": -0.
|
78 |
-
"rewards/rejected": 0.
|
79 |
"step": 40
|
80 |
},
|
81 |
{
|
82 |
"epoch": 0.13,
|
83 |
"learning_rate": 4.987541037542186e-07,
|
84 |
-
"logits/chosen": -1.
|
85 |
-
"logits/rejected": -1.
|
86 |
-
"logps/chosen": -33.
|
87 |
-
"logps/rejected": -35.
|
88 |
-
"loss": 0.
|
89 |
-
"rewards/accuracies": 0.
|
90 |
-
"rewards/chosen": 0.
|
91 |
-
"rewards/margins": -0.
|
92 |
-
"rewards/rejected": 0.
|
93 |
"step": 50
|
94 |
},
|
95 |
{
|
96 |
"epoch": 0.16,
|
97 |
"learning_rate": 4.954691471941118e-07,
|
98 |
-
"logits/chosen": -1.
|
99 |
-
"logits/rejected": -1.
|
100 |
-
"logps/chosen": -32.
|
101 |
-
"logps/rejected": -33.
|
102 |
-
"loss": 0.
|
103 |
-
"rewards/accuracies": 0.
|
104 |
-
"rewards/chosen": 0.
|
105 |
-
"rewards/margins": 0.
|
106 |
-
"rewards/rejected": -0.
|
107 |
"step": 60
|
108 |
},
|
109 |
{
|
110 |
"epoch": 0.18,
|
111 |
"learning_rate": 4.901618883413548e-07,
|
112 |
-
"logits/chosen": -2.
|
113 |
-
"logits/rejected": -2.
|
114 |
-
"logps/chosen": -33.
|
115 |
-
"logps/rejected": -36.
|
116 |
-
"loss": 0.
|
117 |
-
"rewards/accuracies": 0.
|
118 |
-
"rewards/chosen": 0.
|
119 |
-
"rewards/margins": 0.
|
120 |
-
"rewards/rejected": 0.
|
121 |
"step": 70
|
122 |
},
|
123 |
{
|
124 |
"epoch": 0.21,
|
125 |
"learning_rate": 4.828760511501322e-07,
|
126 |
-
"logits/chosen": -1.
|
127 |
-
"logits/rejected": -1.
|
128 |
-
"logps/chosen": -34.
|
129 |
-
"logps/rejected": -34.
|
130 |
-
"loss": 0.
|
131 |
-
"rewards/accuracies": 0.
|
132 |
-
"rewards/chosen": 0.
|
133 |
-
"rewards/margins": 0.
|
134 |
-
"rewards/rejected": 0.
|
135 |
"step": 80
|
136 |
},
|
137 |
{
|
138 |
"epoch": 0.23,
|
139 |
"learning_rate": 4.736716601303429e-07,
|
140 |
-
"logits/chosen": -1.
|
141 |
-
"logits/rejected": -1.
|
142 |
-
"logps/chosen": -32.
|
143 |
-
"logps/rejected": -32.
|
144 |
-
"loss": 0.
|
145 |
-
"rewards/accuracies": 0.
|
146 |
-
"rewards/chosen":
|
147 |
-
"rewards/margins":
|
148 |
-
"rewards/rejected": 0.
|
149 |
"step": 90
|
150 |
},
|
151 |
{
|
152 |
"epoch": 0.26,
|
153 |
"learning_rate": 4.62624545834521e-07,
|
154 |
-
"logits/chosen": -2.
|
155 |
-
"logits/rejected": -2.
|
156 |
-
"logps/chosen": -32.
|
157 |
-
"logps/rejected": -31.
|
158 |
-
"loss": 0.
|
159 |
"rewards/accuracies": 0.512499988079071,
|
160 |
-
"rewards/chosen": 0.
|
161 |
-
"rewards/margins":
|
162 |
-
"rewards/rejected": 0.
|
163 |
"step": 100
|
164 |
},
|
165 |
{
|
166 |
"epoch": 0.26,
|
167 |
-
"eval_logits/chosen": -2.
|
168 |
-
"eval_logits/rejected": -2.
|
169 |
-
"eval_logps/chosen": -34.
|
170 |
-
"eval_logps/rejected": -37.
|
171 |
-
"eval_loss": 0.
|
172 |
-
"eval_rewards/accuracies": 0.
|
173 |
-
"eval_rewards/chosen": 0.
|
174 |
-
"eval_rewards/margins": 0.
|
175 |
-
"eval_rewards/rejected": 0.
|
176 |
-
"eval_runtime": 146.
|
177 |
-
"eval_samples_per_second": 2.
|
178 |
"eval_steps_per_second": 0.294,
|
179 |
"step": 100
|
180 |
},
|
181 |
{
|
182 |
"epoch": 0.29,
|
183 |
"learning_rate": 4.4982572012636904e-07,
|
184 |
-
"logits/chosen": -2.
|
185 |
-
"logits/rejected": -2.
|
186 |
-
"logps/chosen": -33.
|
187 |
-
"logps/rejected": -34.
|
188 |
-
"loss": 0.
|
189 |
-
"rewards/accuracies": 0.
|
190 |
-
"rewards/chosen": 0.
|
191 |
-
"rewards/margins": 0.
|
192 |
-
"rewards/rejected": -0.
|
193 |
"step": 110
|
194 |
},
|
195 |
{
|
196 |
"epoch": 0.31,
|
197 |
"learning_rate": 4.353806263777677e-07,
|
198 |
-
"logits/chosen": -2.
|
199 |
-
"logits/rejected": -2.
|
200 |
-
"logps/chosen": -32.
|
201 |
-
"logps/rejected": -32.
|
202 |
-
"loss": 0.
|
203 |
-
"rewards/accuracies": 0.
|
204 |
-
"rewards/chosen":
|
205 |
-
"rewards/margins": 0.
|
206 |
-
"rewards/rejected": -0.
|
207 |
"step": 120
|
208 |
},
|
209 |
{
|
210 |
"epoch": 0.34,
|
211 |
"learning_rate": 4.194082707715275e-07,
|
212 |
-
"logits/chosen": -2.
|
213 |
-
"logits/rejected": -2.
|
214 |
-
"logps/chosen": -30.
|
215 |
-
"logps/rejected": -32.
|
216 |
-
"loss": 0.
|
217 |
"rewards/accuracies": 0.4625000059604645,
|
218 |
-
"rewards/chosen": -0.
|
219 |
-
"rewards/margins": -0.
|
220 |
-
"rewards/rejected": -0.
|
221 |
"step": 130
|
222 |
},
|
223 |
{
|
224 |
"epoch": 0.36,
|
225 |
"learning_rate": 4.020402418666621e-07,
|
226 |
-
"logits/chosen": -1.
|
227 |
-
"logits/rejected": -1.
|
228 |
-
"logps/chosen": -31.
|
229 |
-
"logps/rejected": -32.
|
230 |
-
"loss": 0.
|
231 |
"rewards/accuracies": 0.550000011920929,
|
232 |
-
"rewards/chosen": 0.
|
233 |
-
"rewards/margins": 0.
|
234 |
-
"rewards/rejected": 0.
|
235 |
"step": 140
|
236 |
},
|
237 |
{
|
238 |
"epoch": 0.39,
|
239 |
"learning_rate": 3.8341962650351185e-07,
|
240 |
-
"logits/chosen": -1.
|
241 |
-
"logits/rejected": -1.
|
242 |
-
"logps/chosen": -34.
|
243 |
-
"logps/rejected": -34.
|
244 |
-
"loss": 0.
|
245 |
-
"rewards/accuracies": 0.
|
246 |
-
"rewards/chosen": -0.
|
247 |
-
"rewards/margins":
|
248 |
-
"rewards/rejected": 0.
|
249 |
"step": 150
|
250 |
},
|
251 |
{
|
252 |
"epoch": 0.42,
|
253 |
"learning_rate": 3.636998309800572e-07,
|
254 |
-
"logits/chosen": -1.
|
255 |
-
"logits/rejected": -1.
|
256 |
-
"logps/chosen": -36.
|
257 |
-
"logps/rejected": -32.
|
258 |
-
"loss": 0.
|
259 |
-
"rewards/accuracies": 0.
|
260 |
-
"rewards/chosen": 0.
|
261 |
-
"rewards/margins": 0.
|
262 |
-
"rewards/rejected": -0.
|
263 |
"step": 160
|
264 |
},
|
265 |
{
|
266 |
"epoch": 0.44,
|
267 |
"learning_rate": 3.430433172111807e-07,
|
268 |
-
"logits/chosen": -2.
|
269 |
-
"logits/rejected": -2.
|
270 |
-
"logps/chosen": -33.
|
271 |
-
"logps/rejected": -31.
|
272 |
-
"loss": 0.
|
273 |
-
"rewards/accuracies": 0.
|
274 |
-
"rewards/chosen": 0.
|
275 |
-
"rewards/margins": 0.
|
276 |
-
"rewards/rejected": -0.
|
277 |
"step": 170
|
278 |
},
|
279 |
{
|
280 |
"epoch": 0.47,
|
281 |
"learning_rate": 3.216202642830543e-07,
|
282 |
-
"logits/chosen": -2.
|
283 |
-
"logits/rejected": -2.
|
284 |
-
"logps/chosen": -32.
|
285 |
-
"logps/rejected": -32.
|
286 |
-
"loss": 0.
|
287 |
-
"rewards/accuracies": 0.
|
288 |
-
"rewards/chosen": 0.
|
289 |
-
"rewards/margins": 0.
|
290 |
-
"rewards/rejected": 0.
|
291 |
"step": 180
|
292 |
},
|
293 |
{
|
294 |
"epoch": 0.49,
|
295 |
"learning_rate": 2.9960716642946403e-07,
|
296 |
-
"logits/chosen": -2.
|
297 |
-
"logits/rejected": -2.
|
298 |
-
"logps/chosen": -31.
|
299 |
-
"logps/rejected": -31.
|
300 |
-
"loss": 0.
|
301 |
-
"rewards/accuracies": 0.
|
302 |
-
"rewards/chosen": 0.
|
303 |
-
"rewards/margins": 0.
|
304 |
-
"rewards/rejected": 0.
|
305 |
"step": 190
|
306 |
},
|
307 |
{
|
308 |
"epoch": 0.52,
|
309 |
"learning_rate": 2.771853789806683e-07,
|
310 |
-
"logits/chosen": -1.
|
311 |
-
"logits/rejected": -1.
|
312 |
-
"logps/chosen": -31.
|
313 |
-
"logps/rejected": -32.
|
314 |
-
"loss": 0.
|
315 |
-
"rewards/accuracies": 0.
|
316 |
-
"rewards/chosen":
|
317 |
-
"rewards/margins": 0.
|
318 |
-
"rewards/rejected": -0.
|
319 |
"step": 200
|
320 |
},
|
321 |
{
|
322 |
"epoch": 0.52,
|
323 |
-
"eval_logits/chosen": -2.
|
324 |
-
"eval_logits/rejected": -2.
|
325 |
-
"eval_logps/chosen": -34.
|
326 |
-
"eval_logps/rejected": -37.
|
327 |
-
"eval_loss": 0.
|
328 |
-
"eval_rewards/accuracies": 0.
|
329 |
-
"eval_rewards/chosen": 0.
|
330 |
-
"eval_rewards/margins": 0.
|
331 |
-
"eval_rewards/rejected": 0.
|
332 |
-
"eval_runtime": 145.
|
333 |
-
"eval_samples_per_second": 2.
|
334 |
"eval_steps_per_second": 0.295,
|
335 |
"step": 200
|
336 |
},
|
337 |
{
|
338 |
"epoch": 0.55,
|
339 |
"learning_rate": 2.5453962426402e-07,
|
340 |
-
"logits/chosen": -2.
|
341 |
-
"logits/rejected": -2.
|
342 |
-
"logps/chosen": -31.
|
343 |
-
"logps/rejected": -33.
|
344 |
-
"loss": 0.
|
345 |
-
"rewards/accuracies": 0.
|
346 |
-
"rewards/chosen": 0.
|
347 |
-
"rewards/margins": 0.
|
348 |
-
"rewards/rejected": 0.
|
349 |
"step": 210
|
350 |
},
|
351 |
{
|
352 |
"epoch": 0.57,
|
353 |
"learning_rate": 2.318564697655179e-07,
|
354 |
-
"logits/chosen": -1.
|
355 |
-
"logits/rejected": -1.
|
356 |
-
"logps/chosen": -30.
|
357 |
-
"logps/rejected": -31.
|
358 |
-
"loss": 0.
|
359 |
-
"rewards/accuracies": 0.
|
360 |
-
"rewards/chosen": 0.
|
361 |
-
"rewards/margins": 0.
|
362 |
-
"rewards/rejected": -0.
|
363 |
"step": 220
|
364 |
},
|
365 |
{
|
366 |
"epoch": 0.6,
|
367 |
"learning_rate": 2.093227910899832e-07,
|
368 |
-
"logits/chosen": -1.
|
369 |
-
"logits/rejected": -1.
|
370 |
-
"logps/chosen": -33.
|
371 |
-
"logps/rejected": -31.
|
372 |
-
"loss": 0.
|
373 |
"rewards/accuracies": 0.550000011920929,
|
374 |
-
"rewards/chosen": 0.
|
375 |
-
"rewards/margins": 0.
|
376 |
-
"rewards/rejected": 0.
|
377 |
"step": 230
|
378 |
},
|
379 |
{
|
380 |
"epoch": 0.62,
|
381 |
"learning_rate": 1.8712423238279356e-07,
|
382 |
-
"logits/chosen": -1.
|
383 |
-
"logits/rejected": -1.
|
384 |
-
"logps/chosen": -34.
|
385 |
-
"logps/rejected": -34.
|
386 |
-
"loss": 0.
|
387 |
"rewards/accuracies": 0.5249999761581421,
|
388 |
-
"rewards/chosen": 0.
|
389 |
-
"rewards/margins":
|
390 |
-
"rewards/rejected": 0.
|
391 |
"step": 240
|
392 |
},
|
393 |
{
|
394 |
"epoch": 0.65,
|
395 |
"learning_rate": 1.654436768970182e-07,
|
396 |
-
"logits/chosen": -2.
|
397 |
-
"logits/rejected": -2.
|
398 |
-
"logps/chosen": -32.
|
399 |
-
"logps/rejected": -36.
|
400 |
-
"loss": 0.
|
401 |
-
"rewards/accuracies": 0.
|
402 |
-
"rewards/chosen": 0.
|
403 |
-
"rewards/margins": 0.
|
404 |
-
"rewards/rejected": 0.
|
405 |
"step": 250
|
406 |
},
|
407 |
{
|
408 |
"epoch": 0.68,
|
409 |
"learning_rate": 1.444597403062196e-07,
|
410 |
-
"logits/chosen": -1.
|
411 |
-
"logits/rejected": -1.
|
412 |
-
"logps/chosen": -34.
|
413 |
-
"logps/rejected": -35.
|
414 |
-
"loss": 0.
|
415 |
-
"rewards/accuracies": 0.
|
416 |
-
"rewards/chosen": 0.
|
417 |
-
"rewards/margins": 0.
|
418 |
-
"rewards/rejected": 0.
|
419 |
"step": 260
|
420 |
},
|
421 |
{
|
422 |
"epoch": 0.7,
|
423 |
"learning_rate": 1.2434529917578887e-07,
|
424 |
-
"logits/chosen": -1.
|
425 |
-
"logits/rejected": -1.
|
426 |
-
"logps/chosen": -34.
|
427 |
-
"logps/rejected": -31.
|
428 |
-
"loss": 0.
|
429 |
"rewards/accuracies": 0.4625000059604645,
|
430 |
-
"rewards/chosen": 0.
|
431 |
-
"rewards/margins": -0.
|
432 |
-
"rewards/rejected": 0.
|
433 |
"step": 270
|
434 |
},
|
435 |
{
|
436 |
"epoch": 0.73,
|
437 |
"learning_rate": 1.0526606671603521e-07,
|
438 |
-
"logits/chosen": -1.
|
439 |
-
"logits/rejected": -1.
|
440 |
-
"logps/chosen": -35.
|
441 |
-
"logps/rejected": -31.
|
442 |
-
"loss": 0.
|
443 |
-
"rewards/accuracies": 0.
|
444 |
-
"rewards/chosen": 0.
|
445 |
-
"rewards/margins": 0.
|
446 |
-
"rewards/rejected": 0.
|
447 |
"step": 280
|
448 |
},
|
449 |
{
|
450 |
"epoch": 0.75,
|
451 |
"learning_rate": 8.737922755071453e-08,
|
452 |
-
"logits/chosen": -2.
|
453 |
-
"logits/rejected": -2.
|
454 |
-
"logps/chosen": -30.
|
455 |
-
"logps/rejected": -32.
|
456 |
-
"loss": 0.
|
457 |
"rewards/accuracies": 0.512499988079071,
|
458 |
-
"rewards/chosen": 0.
|
459 |
-
"rewards/margins": 0.
|
460 |
-
"rewards/rejected": -0.
|
461 |
"step": 290
|
462 |
},
|
463 |
{
|
464 |
"epoch": 0.78,
|
465 |
"learning_rate": 7.08321427484816e-08,
|
466 |
-
"logits/chosen": -1.
|
467 |
-
"logits/rejected": -1.
|
468 |
-
"logps/chosen": -32.
|
469 |
-
"logps/rejected": -30.
|
470 |
-
"loss": 0.
|
471 |
-
"rewards/accuracies": 0.
|
472 |
-
"rewards/chosen": 0.
|
473 |
-
"rewards/margins": 0.
|
474 |
-
"rewards/rejected": 0.
|
475 |
"step": 300
|
476 |
},
|
477 |
{
|
478 |
"epoch": 0.78,
|
479 |
-
"eval_logits/chosen": -2.
|
480 |
-
"eval_logits/rejected": -2.
|
481 |
-
"eval_logps/chosen": -34.
|
482 |
-
"eval_logps/rejected": -37.
|
483 |
-
"eval_loss": 0.
|
484 |
-
"eval_rewards/accuracies": 0.
|
485 |
-
"eval_rewards/chosen": 0.
|
486 |
-
"eval_rewards/margins": 0.
|
487 |
-
"eval_rewards/rejected": 0.
|
488 |
-
"eval_runtime": 145.
|
489 |
-
"eval_samples_per_second": 2.
|
490 |
"eval_steps_per_second": 0.295,
|
491 |
"step": 300
|
492 |
},
|
493 |
{
|
494 |
"epoch": 0.81,
|
495 |
"learning_rate": 5.576113578589034e-08,
|
496 |
-
"logits/chosen": -1.
|
497 |
-
"logits/rejected": -1.
|
498 |
-
"logps/chosen": -31.
|
499 |
-
"logps/rejected": -33.
|
500 |
-
"loss": 0.
|
501 |
-
"rewards/accuracies": 0.
|
502 |
-
"rewards/chosen": 0.
|
503 |
-
"rewards/margins": 0.
|
504 |
-
"rewards/rejected": 0.
|
505 |
"step": 310
|
506 |
},
|
507 |
{
|
508 |
"epoch": 0.83,
|
509 |
"learning_rate": 4.229036944380912e-08,
|
510 |
-
"logits/chosen": -1.
|
511 |
-
"logits/rejected": -1.
|
512 |
-
"logps/chosen": -34.
|
513 |
-
"logps/rejected": -33.
|
514 |
-
"loss": 0.
|
515 |
-
"rewards/accuracies": 0.
|
516 |
-
"rewards/chosen": 0.
|
517 |
-
"rewards/margins": 0.
|
518 |
-
"rewards/rejected": -
|
519 |
"step": 320
|
520 |
},
|
521 |
{
|
522 |
"epoch": 0.86,
|
523 |
"learning_rate": 3.053082288996112e-08,
|
524 |
-
"logits/chosen": -2.
|
525 |
-
"logits/rejected": -2.
|
526 |
-
"logps/chosen": -33.
|
527 |
-
"logps/rejected": -32.
|
528 |
-
"loss": 0.
|
529 |
-
"rewards/accuracies": 0.
|
530 |
-
"rewards/chosen": 0.
|
531 |
-
"rewards/margins": 0.
|
532 |
-
"rewards/rejected": 0.
|
533 |
"step": 330
|
534 |
},
|
535 |
{
|
536 |
"epoch": 0.88,
|
537 |
"learning_rate": 2.05793773749158e-08,
|
538 |
-
"logits/chosen": -2.
|
539 |
-
"logits/rejected": -2.
|
540 |
-
"logps/chosen": -34.
|
541 |
-
"logps/rejected": -33.
|
542 |
-
"loss": 0.
|
543 |
-
"rewards/accuracies": 0.
|
544 |
-
"rewards/chosen": 0.
|
545 |
-
"rewards/margins": -0.
|
546 |
-
"rewards/rejected": 0.
|
547 |
"step": 340
|
548 |
},
|
549 |
{
|
550 |
"epoch": 0.91,
|
551 |
"learning_rate": 1.251801807404168e-08,
|
552 |
-
"logits/chosen": -1.
|
553 |
-
"logits/rejected": -1.
|
554 |
-
"logps/chosen": -33.
|
555 |
-
"logps/rejected": -32.
|
556 |
-
"loss": 0.
|
557 |
-
"rewards/accuracies": 0.
|
558 |
-
"rewards/chosen": 0.
|
559 |
-
"rewards/margins": 0.
|
560 |
-
"rewards/rejected": 0.
|
561 |
"step": 350
|
562 |
},
|
563 |
{
|
564 |
"epoch": 0.94,
|
565 |
"learning_rate": 6.41315865106129e-09,
|
566 |
-
"logits/chosen": -1.
|
567 |
-
"logits/rejected": -1.
|
568 |
-
"logps/chosen": -32.
|
569 |
-
"logps/rejected": -35.
|
570 |
-
"loss": 0.
|
571 |
-
"rewards/accuracies": 0.
|
572 |
-
"rewards/chosen": 0.
|
573 |
-
"rewards/margins": 0.
|
574 |
-
"rewards/rejected": 0.
|
575 |
"step": 360
|
576 |
},
|
577 |
{
|
578 |
"epoch": 0.96,
|
579 |
"learning_rate": 2.3150941078050324e-09,
|
580 |
-
"logits/chosen": -2.
|
581 |
-
"logits/rejected": -2.
|
582 |
-
"logps/chosen": -33.
|
583 |
-
"logps/rejected": -29.
|
584 |
-
"loss": 0.
|
585 |
-
"rewards/accuracies": 0.
|
586 |
-
"rewards/chosen": -0.
|
587 |
-
"rewards/margins": -0.
|
588 |
-
"rewards/rejected": 0.
|
589 |
"step": 370
|
590 |
},
|
591 |
{
|
592 |
"epoch": 0.99,
|
593 |
"learning_rate": 2.575864278703266e-10,
|
594 |
-
"logits/chosen": -1.
|
595 |
-
"logits/rejected": -1.
|
596 |
-
"logps/chosen": -34.
|
597 |
-
"logps/rejected": -30.
|
598 |
-
"loss": 0.
|
599 |
-
"rewards/accuracies": 0.
|
600 |
-
"rewards/chosen": 0.
|
601 |
-
"rewards/margins": 0.
|
602 |
-
"rewards/rejected": -0.
|
603 |
"step": 380
|
604 |
},
|
605 |
{
|
606 |
"epoch": 1.0,
|
607 |
"step": 385,
|
608 |
"total_flos": 0.0,
|
609 |
-
"train_loss": 0.
|
610 |
-
"train_runtime":
|
611 |
-
"train_samples_per_second": 4.
|
612 |
-
"train_steps_per_second": 0.
|
613 |
}
|
614 |
],
|
615 |
"logging_steps": 10,
|
|
|
15 |
"logits/rejected": -1.7377450466156006,
|
16 |
"logps/chosen": -29.553977966308594,
|
17 |
"logps/rejected": -42.813133239746094,
|
18 |
+
"loss": 0.3086,
|
19 |
"rewards/accuracies": 0.0,
|
20 |
"rewards/chosen": 0.0,
|
21 |
"rewards/margins": 0.0,
|
|
|
25 |
{
|
26 |
"epoch": 0.03,
|
27 |
"learning_rate": 1.2820512820512818e-07,
|
28 |
+
"logits/chosen": -1.866453766822815,
|
29 |
+
"logits/rejected": -1.8707627058029175,
|
30 |
+
"logps/chosen": -36.98646545410156,
|
31 |
+
"logps/rejected": -33.6445426940918,
|
32 |
+
"loss": 0.3098,
|
33 |
+
"rewards/accuracies": 0.4444444477558136,
|
34 |
+
"rewards/chosen": 0.018122652545571327,
|
35 |
+
"rewards/margins": 0.031144678592681885,
|
36 |
+
"rewards/rejected": -0.013022023253142834,
|
37 |
"step": 10
|
38 |
},
|
39 |
{
|
40 |
"epoch": 0.05,
|
41 |
"learning_rate": 2.5641025641025636e-07,
|
42 |
+
"logits/chosen": -1.9970738887786865,
|
43 |
+
"logits/rejected": -1.9997297525405884,
|
44 |
+
"logps/chosen": -29.651851654052734,
|
45 |
+
"logps/rejected": -29.048221588134766,
|
46 |
+
"loss": 0.3674,
|
47 |
+
"rewards/accuracies": 0.4375,
|
48 |
+
"rewards/chosen": -0.008690872229635715,
|
49 |
+
"rewards/margins": -0.02239001728594303,
|
50 |
+
"rewards/rejected": 0.01369914598762989,
|
51 |
"step": 20
|
52 |
},
|
53 |
{
|
54 |
"epoch": 0.08,
|
55 |
"learning_rate": 3.8461538461538463e-07,
|
56 |
+
"logits/chosen": -1.919904351234436,
|
57 |
+
"logits/rejected": -1.9172093868255615,
|
58 |
+
"logps/chosen": -31.407363891601562,
|
59 |
+
"logps/rejected": -33.22914505004883,
|
60 |
+
"loss": 0.3289,
|
61 |
+
"rewards/accuracies": 0.5874999761581421,
|
62 |
+
"rewards/chosen": 0.007855755276978016,
|
63 |
+
"rewards/margins": 0.017470674589276314,
|
64 |
+
"rewards/rejected": -0.009614917449653149,
|
65 |
"step": 30
|
66 |
},
|
67 |
{
|
68 |
"epoch": 0.1,
|
69 |
"learning_rate": 4.999896948438433e-07,
|
70 |
+
"logits/chosen": -2.016488552093506,
|
71 |
+
"logits/rejected": -2.0077414512634277,
|
72 |
+
"logps/chosen": -32.588050842285156,
|
73 |
+
"logps/rejected": -32.51618576049805,
|
74 |
+
"loss": 0.3526,
|
75 |
+
"rewards/accuracies": 0.4749999940395355,
|
76 |
+
"rewards/chosen": -0.010120267979800701,
|
77 |
+
"rewards/margins": -0.006958496756851673,
|
78 |
+
"rewards/rejected": -0.0031617730855941772,
|
79 |
"step": 40
|
80 |
},
|
81 |
{
|
82 |
"epoch": 0.13,
|
83 |
"learning_rate": 4.987541037542186e-07,
|
84 |
+
"logits/chosen": -1.8645082712173462,
|
85 |
+
"logits/rejected": -1.8537276983261108,
|
86 |
+
"logps/chosen": -33.5635871887207,
|
87 |
+
"logps/rejected": -35.42169952392578,
|
88 |
+
"loss": 0.3858,
|
89 |
+
"rewards/accuracies": 0.38749998807907104,
|
90 |
+
"rewards/chosen": -0.005497059319168329,
|
91 |
+
"rewards/margins": -0.03307711333036423,
|
92 |
+
"rewards/rejected": 0.027580058202147484,
|
93 |
"step": 50
|
94 |
},
|
95 |
{
|
96 |
"epoch": 0.16,
|
97 |
"learning_rate": 4.954691471941118e-07,
|
98 |
+
"logits/chosen": -1.9458835124969482,
|
99 |
+
"logits/rejected": -1.9478065967559814,
|
100 |
+
"logps/chosen": -32.57460021972656,
|
101 |
+
"logps/rejected": -33.195518493652344,
|
102 |
+
"loss": 0.3015,
|
103 |
+
"rewards/accuracies": 0.574999988079071,
|
104 |
+
"rewards/chosen": 0.02302517369389534,
|
105 |
+
"rewards/margins": 0.04474987834692001,
|
106 |
+
"rewards/rejected": -0.021724706515669823,
|
107 |
"step": 60
|
108 |
},
|
109 |
{
|
110 |
"epoch": 0.18,
|
111 |
"learning_rate": 4.901618883413548e-07,
|
112 |
+
"logits/chosen": -2.0793328285217285,
|
113 |
+
"logits/rejected": -2.084324836730957,
|
114 |
+
"logps/chosen": -33.99799728393555,
|
115 |
+
"logps/rejected": -36.57799530029297,
|
116 |
+
"loss": 0.3658,
|
117 |
+
"rewards/accuracies": 0.48750001192092896,
|
118 |
+
"rewards/chosen": -0.00855917576700449,
|
119 |
+
"rewards/margins": -0.00035296304849907756,
|
120 |
+
"rewards/rejected": -0.008206211030483246,
|
121 |
"step": 70
|
122 |
},
|
123 |
{
|
124 |
"epoch": 0.21,
|
125 |
"learning_rate": 4.828760511501322e-07,
|
126 |
+
"logits/chosen": -1.9421542882919312,
|
127 |
+
"logits/rejected": -1.9453080892562866,
|
128 |
+
"logps/chosen": -34.38970184326172,
|
129 |
+
"logps/rejected": -34.58138656616211,
|
130 |
+
"loss": 0.3181,
|
131 |
+
"rewards/accuracies": 0.5874999761581421,
|
132 |
+
"rewards/chosen": 0.034300822764635086,
|
133 |
+
"rewards/margins": 0.024381538853049278,
|
134 |
+
"rewards/rejected": 0.009919276461005211,
|
135 |
"step": 80
|
136 |
},
|
137 |
{
|
138 |
"epoch": 0.23,
|
139 |
"learning_rate": 4.736716601303429e-07,
|
140 |
+
"logits/chosen": -1.9507806301116943,
|
141 |
+
"logits/rejected": -1.955293893814087,
|
142 |
+
"logps/chosen": -32.461936950683594,
|
143 |
+
"logps/rejected": -32.3816032409668,
|
144 |
+
"loss": 0.3094,
|
145 |
+
"rewards/accuracies": 0.574999988079071,
|
146 |
+
"rewards/chosen": 0.004934861324727535,
|
147 |
+
"rewards/margins": 0.02535998821258545,
|
148 |
+
"rewards/rejected": -0.02042512409389019,
|
149 |
"step": 90
|
150 |
},
|
151 |
{
|
152 |
"epoch": 0.26,
|
153 |
"learning_rate": 4.62624545834521e-07,
|
154 |
+
"logits/chosen": -2.0485286712646484,
|
155 |
+
"logits/rejected": -2.046539783477783,
|
156 |
+
"logps/chosen": -32.23391342163086,
|
157 |
+
"logps/rejected": -31.265544891357422,
|
158 |
+
"loss": 0.333,
|
159 |
"rewards/accuracies": 0.512499988079071,
|
160 |
+
"rewards/chosen": 0.006228716112673283,
|
161 |
+
"rewards/margins": 0.005629505030810833,
|
162 |
+
"rewards/rejected": 0.0005992107326164842,
|
163 |
"step": 100
|
164 |
},
|
165 |
{
|
166 |
"epoch": 0.26,
|
167 |
+
"eval_logits/chosen": -2.243820905685425,
|
168 |
+
"eval_logits/rejected": -2.238938570022583,
|
169 |
+
"eval_logps/chosen": -34.02050018310547,
|
170 |
+
"eval_logps/rejected": -37.50347137451172,
|
171 |
+
"eval_loss": 0.3404548168182373,
|
172 |
+
"eval_rewards/accuracies": 0.5103820562362671,
|
173 |
+
"eval_rewards/chosen": 0.012649022974073887,
|
174 |
+
"eval_rewards/margins": 0.0008147989865392447,
|
175 |
+
"eval_rewards/rejected": 0.011834223754703999,
|
176 |
+
"eval_runtime": 146.2563,
|
177 |
+
"eval_samples_per_second": 2.345,
|
178 |
"eval_steps_per_second": 0.294,
|
179 |
"step": 100
|
180 |
},
|
181 |
{
|
182 |
"epoch": 0.29,
|
183 |
"learning_rate": 4.4982572012636904e-07,
|
184 |
+
"logits/chosen": -2.0055313110351562,
|
185 |
+
"logits/rejected": -2.00311541557312,
|
186 |
+
"logps/chosen": -33.23518753051758,
|
187 |
+
"logps/rejected": -34.03091812133789,
|
188 |
+
"loss": 0.3233,
|
189 |
+
"rewards/accuracies": 0.48750001192092896,
|
190 |
+
"rewards/chosen": 0.007764431647956371,
|
191 |
+
"rewards/margins": 0.009159245528280735,
|
192 |
+
"rewards/rejected": -0.0013948131818324327,
|
193 |
"step": 110
|
194 |
},
|
195 |
{
|
196 |
"epoch": 0.31,
|
197 |
"learning_rate": 4.353806263777677e-07,
|
198 |
+
"logits/chosen": -2.016556978225708,
|
199 |
+
"logits/rejected": -2.0081825256347656,
|
200 |
+
"logps/chosen": -32.435211181640625,
|
201 |
+
"logps/rejected": -32.18501281738281,
|
202 |
+
"loss": 0.3121,
|
203 |
+
"rewards/accuracies": 0.5625,
|
204 |
+
"rewards/chosen": 0.00872363243252039,
|
205 |
+
"rewards/margins": 0.029877260327339172,
|
206 |
+
"rewards/rejected": -0.02115362510085106,
|
207 |
"step": 120
|
208 |
},
|
209 |
{
|
210 |
"epoch": 0.34,
|
211 |
"learning_rate": 4.194082707715275e-07,
|
212 |
+
"logits/chosen": -2.0461266040802,
|
213 |
+
"logits/rejected": -2.03806734085083,
|
214 |
+
"logps/chosen": -30.502710342407227,
|
215 |
+
"logps/rejected": -32.0546760559082,
|
216 |
+
"loss": 0.3674,
|
217 |
"rewards/accuracies": 0.4625000059604645,
|
218 |
+
"rewards/chosen": -0.02429451420903206,
|
219 |
+
"rewards/margins": -0.02279272861778736,
|
220 |
+
"rewards/rejected": -0.0015017850091680884,
|
221 |
"step": 130
|
222 |
},
|
223 |
{
|
224 |
"epoch": 0.36,
|
225 |
"learning_rate": 4.020402418666621e-07,
|
226 |
+
"logits/chosen": -1.9767634868621826,
|
227 |
+
"logits/rejected": -1.9870494604110718,
|
228 |
+
"logps/chosen": -31.402328491210938,
|
229 |
+
"logps/rejected": -32.55817794799805,
|
230 |
+
"loss": 0.3125,
|
231 |
"rewards/accuracies": 0.550000011920929,
|
232 |
+
"rewards/chosen": 0.015601096674799919,
|
233 |
+
"rewards/margins": 0.03657008707523346,
|
234 |
+
"rewards/rejected": -0.02096899226307869,
|
235 |
"step": 140
|
236 |
},
|
237 |
{
|
238 |
"epoch": 0.39,
|
239 |
"learning_rate": 3.8341962650351185e-07,
|
240 |
+
"logits/chosen": -1.8908436298370361,
|
241 |
+
"logits/rejected": -1.8919413089752197,
|
242 |
+
"logps/chosen": -34.207401275634766,
|
243 |
+
"logps/rejected": -34.783695220947266,
|
244 |
+
"loss": 0.3373,
|
245 |
+
"rewards/accuracies": 0.5375000238418579,
|
246 |
+
"rewards/chosen": -0.013235817663371563,
|
247 |
+
"rewards/margins": 0.01055879332125187,
|
248 |
+
"rewards/rejected": -0.023794615641236305,
|
249 |
"step": 150
|
250 |
},
|
251 |
{
|
252 |
"epoch": 0.42,
|
253 |
"learning_rate": 3.636998309800572e-07,
|
254 |
+
"logits/chosen": -1.942582368850708,
|
255 |
+
"logits/rejected": -1.9391053915023804,
|
256 |
+
"logps/chosen": -36.15548324584961,
|
257 |
+
"logps/rejected": -32.729759216308594,
|
258 |
+
"loss": 0.3207,
|
259 |
+
"rewards/accuracies": 0.5,
|
260 |
+
"rewards/chosen": 0.014574882574379444,
|
261 |
+
"rewards/margins": 0.020880455151200294,
|
262 |
+
"rewards/rejected": -0.006305573042482138,
|
263 |
"step": 160
|
264 |
},
|
265 |
{
|
266 |
"epoch": 0.44,
|
267 |
"learning_rate": 3.430433172111807e-07,
|
268 |
+
"logits/chosen": -2.042206287384033,
|
269 |
+
"logits/rejected": -2.034825086593628,
|
270 |
+
"logps/chosen": -33.77631378173828,
|
271 |
+
"logps/rejected": -31.35869789123535,
|
272 |
+
"loss": 0.3201,
|
273 |
+
"rewards/accuracies": 0.5625,
|
274 |
+
"rewards/chosen": 0.019394155591726303,
|
275 |
+
"rewards/margins": 0.025556519627571106,
|
276 |
+
"rewards/rejected": -0.00616236450150609,
|
277 |
"step": 170
|
278 |
},
|
279 |
{
|
280 |
"epoch": 0.47,
|
281 |
"learning_rate": 3.216202642830543e-07,
|
282 |
+
"logits/chosen": -2.0474154949188232,
|
283 |
+
"logits/rejected": -2.0526936054229736,
|
284 |
+
"logps/chosen": -32.521339416503906,
|
285 |
+
"logps/rejected": -32.497581481933594,
|
286 |
+
"loss": 0.2999,
|
287 |
+
"rewards/accuracies": 0.612500011920929,
|
288 |
+
"rewards/chosen": 0.021103620529174805,
|
289 |
+
"rewards/margins": 0.032301973551511765,
|
290 |
+
"rewards/rejected": -0.01119835115969181,
|
291 |
"step": 180
|
292 |
},
|
293 |
{
|
294 |
"epoch": 0.49,
|
295 |
"learning_rate": 2.9960716642946403e-07,
|
296 |
+
"logits/chosen": -2.0485098361968994,
|
297 |
+
"logits/rejected": -2.0457375049591064,
|
298 |
+
"logps/chosen": -31.49017906188965,
|
299 |
+
"logps/rejected": -31.345867156982422,
|
300 |
+
"loss": 0.3315,
|
301 |
+
"rewards/accuracies": 0.5625,
|
302 |
+
"rewards/chosen": -0.001104071387089789,
|
303 |
+
"rewards/margins": 0.023781923577189445,
|
304 |
+
"rewards/rejected": -0.02488599717617035,
|
305 |
"step": 190
|
306 |
},
|
307 |
{
|
308 |
"epoch": 0.52,
|
309 |
"learning_rate": 2.771853789806683e-07,
|
310 |
+
"logits/chosen": -1.9183101654052734,
|
311 |
+
"logits/rejected": -1.9229921102523804,
|
312 |
+
"logps/chosen": -31.60622215270996,
|
313 |
+
"logps/rejected": -32.805049896240234,
|
314 |
+
"loss": 0.3306,
|
315 |
+
"rewards/accuracies": 0.5249999761581421,
|
316 |
+
"rewards/chosen": 5.159676220500842e-05,
|
317 |
+
"rewards/margins": 0.012757277116179466,
|
318 |
+
"rewards/rejected": -0.012705676257610321,
|
319 |
"step": 200
|
320 |
},
|
321 |
{
|
322 |
"epoch": 0.52,
|
323 |
+
"eval_logits/chosen": -2.2434937953948975,
|
324 |
+
"eval_logits/rejected": -2.2386233806610107,
|
325 |
+
"eval_logps/chosen": -34.02851867675781,
|
326 |
+
"eval_logps/rejected": -37.51131057739258,
|
327 |
+
"eval_loss": 0.3466452658176422,
|
328 |
+
"eval_rewards/accuracies": 0.5020764470100403,
|
329 |
+
"eval_rewards/chosen": 0.005428452976047993,
|
330 |
+
"eval_rewards/margins": 0.0006523749325424433,
|
331 |
+
"eval_rewards/rejected": 0.00477607874199748,
|
332 |
+
"eval_runtime": 145.8217,
|
333 |
+
"eval_samples_per_second": 2.352,
|
334 |
"eval_steps_per_second": 0.295,
|
335 |
"step": 200
|
336 |
},
|
337 |
{
|
338 |
"epoch": 0.55,
|
339 |
"learning_rate": 2.5453962426402e-07,
|
340 |
+
"logits/chosen": -2.032118558883667,
|
341 |
+
"logits/rejected": -2.042782783508301,
|
342 |
+
"logps/chosen": -31.939167022705078,
|
343 |
+
"logps/rejected": -33.88642501831055,
|
344 |
+
"loss": 0.2871,
|
345 |
+
"rewards/accuracies": 0.6000000238418579,
|
346 |
+
"rewards/chosen": 0.02892182394862175,
|
347 |
+
"rewards/margins": 0.050276655703783035,
|
348 |
+
"rewards/rejected": -0.021354828029870987,
|
349 |
"step": 210
|
350 |
},
|
351 |
{
|
352 |
"epoch": 0.57,
|
353 |
"learning_rate": 2.318564697655179e-07,
|
354 |
+
"logits/chosen": -1.925293207168579,
|
355 |
+
"logits/rejected": -1.9401395320892334,
|
356 |
+
"logps/chosen": -30.106586456298828,
|
357 |
+
"logps/rejected": -31.56791114807129,
|
358 |
+
"loss": 0.3327,
|
359 |
+
"rewards/accuracies": 0.6000000238418579,
|
360 |
+
"rewards/chosen": 0.0044566914439201355,
|
361 |
+
"rewards/margins": 0.013218318112194538,
|
362 |
+
"rewards/rejected": -0.008761629462242126,
|
363 |
"step": 220
|
364 |
},
|
365 |
{
|
366 |
"epoch": 0.6,
|
367 |
"learning_rate": 2.093227910899832e-07,
|
368 |
+
"logits/chosen": -1.9831531047821045,
|
369 |
+
"logits/rejected": -1.9871160984039307,
|
370 |
+
"logps/chosen": -33.386024475097656,
|
371 |
+
"logps/rejected": -31.5727481842041,
|
372 |
+
"loss": 0.2974,
|
373 |
"rewards/accuracies": 0.550000011920929,
|
374 |
+
"rewards/chosen": 0.028298119083046913,
|
375 |
+
"rewards/margins": 0.04939670115709305,
|
376 |
+
"rewards/rejected": -0.021098580211400986,
|
377 |
"step": 230
|
378 |
},
|
379 |
{
|
380 |
"epoch": 0.62,
|
381 |
"learning_rate": 1.8712423238279356e-07,
|
382 |
+
"logits/chosen": -1.9833450317382812,
|
383 |
+
"logits/rejected": -1.9613698720932007,
|
384 |
+
"logps/chosen": -34.1508903503418,
|
385 |
+
"logps/rejected": -34.95817947387695,
|
386 |
+
"loss": 0.3388,
|
387 |
"rewards/accuracies": 0.5249999761581421,
|
388 |
+
"rewards/chosen": 0.002569653559476137,
|
389 |
+
"rewards/margins": 0.00921549927443266,
|
390 |
+
"rewards/rejected": -0.006645847111940384,
|
391 |
"step": 240
|
392 |
},
|
393 |
{
|
394 |
"epoch": 0.65,
|
395 |
"learning_rate": 1.654436768970182e-07,
|
396 |
+
"logits/chosen": -2.0248680114746094,
|
397 |
+
"logits/rejected": -2.021559238433838,
|
398 |
+
"logps/chosen": -32.907005310058594,
|
399 |
+
"logps/rejected": -36.22737121582031,
|
400 |
+
"loss": 0.3221,
|
401 |
+
"rewards/accuracies": 0.550000011920929,
|
402 |
+
"rewards/chosen": 0.016254058107733727,
|
403 |
+
"rewards/margins": 0.018458303064107895,
|
404 |
+
"rewards/rejected": -0.002204245189204812,
|
405 |
"step": 250
|
406 |
},
|
407 |
{
|
408 |
"epoch": 0.68,
|
409 |
"learning_rate": 1.444597403062196e-07,
|
410 |
+
"logits/chosen": -1.8913663625717163,
|
411 |
+
"logits/rejected": -1.8889217376708984,
|
412 |
+
"logps/chosen": -34.18629455566406,
|
413 |
+
"logps/rejected": -35.490962982177734,
|
414 |
+
"loss": 0.3601,
|
415 |
+
"rewards/accuracies": 0.5,
|
416 |
+
"rewards/chosen": 0.0018952053505927324,
|
417 |
+
"rewards/margins": -0.010397391393780708,
|
418 |
+
"rewards/rejected": 0.012292595580220222,
|
419 |
"step": 260
|
420 |
},
|
421 |
{
|
422 |
"epoch": 0.7,
|
423 |
"learning_rate": 1.2434529917578887e-07,
|
424 |
+
"logits/chosen": -1.8756721019744873,
|
425 |
+
"logits/rejected": -1.8731330633163452,
|
426 |
+
"logps/chosen": -34.39625549316406,
|
427 |
+
"logps/rejected": -31.76236343383789,
|
428 |
+
"loss": 0.3492,
|
429 |
"rewards/accuracies": 0.4625000059604645,
|
430 |
+
"rewards/chosen": 0.001211274997331202,
|
431 |
+
"rewards/margins": -0.006859898567199707,
|
432 |
+
"rewards/rejected": 0.008071173913776875,
|
433 |
"step": 270
|
434 |
},
|
435 |
{
|
436 |
"epoch": 0.73,
|
437 |
"learning_rate": 1.0526606671603521e-07,
|
438 |
+
"logits/chosen": -1.9795773029327393,
|
439 |
+
"logits/rejected": -1.9689537286758423,
|
440 |
+
"logps/chosen": -35.320404052734375,
|
441 |
+
"logps/rejected": -31.86062240600586,
|
442 |
+
"loss": 0.298,
|
443 |
+
"rewards/accuracies": 0.574999988079071,
|
444 |
+
"rewards/chosen": 0.032611697912216187,
|
445 |
+
"rewards/margins": 0.05027870461344719,
|
446 |
+
"rewards/rejected": -0.017666997388005257,
|
447 |
"step": 280
|
448 |
},
|
449 |
{
|
450 |
"epoch": 0.75,
|
451 |
"learning_rate": 8.737922755071453e-08,
|
452 |
+
"logits/chosen": -2.075582981109619,
|
453 |
+
"logits/rejected": -2.0605711936950684,
|
454 |
+
"logps/chosen": -30.907567977905273,
|
455 |
+
"logps/rejected": -32.65256881713867,
|
456 |
+
"loss": 0.3186,
|
457 |
"rewards/accuracies": 0.512499988079071,
|
458 |
+
"rewards/chosen": 0.014904007315635681,
|
459 |
+
"rewards/margins": 0.015564213506877422,
|
460 |
+
"rewards/rejected": -0.0006602049106732011,
|
461 |
"step": 290
|
462 |
},
|
463 |
{
|
464 |
"epoch": 0.78,
|
465 |
"learning_rate": 7.08321427484816e-08,
|
466 |
+
"logits/chosen": -1.9468443393707275,
|
467 |
+
"logits/rejected": -1.944310188293457,
|
468 |
+
"logps/chosen": -32.90850067138672,
|
469 |
+
"logps/rejected": -30.801532745361328,
|
470 |
+
"loss": 0.3327,
|
471 |
+
"rewards/accuracies": 0.4749999940395355,
|
472 |
+
"rewards/chosen": 0.016746358945965767,
|
473 |
+
"rewards/margins": 0.002998964861035347,
|
474 |
+
"rewards/rejected": 0.01374739408493042,
|
475 |
"step": 300
|
476 |
},
|
477 |
{
|
478 |
"epoch": 0.78,
|
479 |
+
"eval_logits/chosen": -2.243875741958618,
|
480 |
+
"eval_logits/rejected": -2.238999366760254,
|
481 |
+
"eval_logps/chosen": -34.01763916015625,
|
482 |
+
"eval_logps/rejected": -37.51422882080078,
|
483 |
+
"eval_loss": 0.32764142751693726,
|
484 |
+
"eval_rewards/accuracies": 0.5365448594093323,
|
485 |
+
"eval_rewards/chosen": 0.015219501219689846,
|
486 |
+
"eval_rewards/margins": 0.01306714303791523,
|
487 |
+
"eval_rewards/rejected": 0.0021523565519601107,
|
488 |
+
"eval_runtime": 145.9076,
|
489 |
+
"eval_samples_per_second": 2.351,
|
490 |
"eval_steps_per_second": 0.295,
|
491 |
"step": 300
|
492 |
},
|
493 |
{
|
494 |
"epoch": 0.81,
|
495 |
"learning_rate": 5.576113578589034e-08,
|
496 |
+
"logits/chosen": -1.9285104274749756,
|
497 |
+
"logits/rejected": -1.925238847732544,
|
498 |
+
"logps/chosen": -31.57692527770996,
|
499 |
+
"logps/rejected": -33.744144439697266,
|
500 |
+
"loss": 0.3143,
|
501 |
+
"rewards/accuracies": 0.550000011920929,
|
502 |
+
"rewards/chosen": 0.023783722892403603,
|
503 |
+
"rewards/margins": 0.025577496737241745,
|
504 |
+
"rewards/rejected": -0.0017937703523784876,
|
505 |
"step": 310
|
506 |
},
|
507 |
{
|
508 |
"epoch": 0.83,
|
509 |
"learning_rate": 4.229036944380912e-08,
|
510 |
+
"logits/chosen": -1.980386734008789,
|
511 |
+
"logits/rejected": -1.9680719375610352,
|
512 |
+
"logps/chosen": -34.57415771484375,
|
513 |
+
"logps/rejected": -33.55961227416992,
|
514 |
+
"loss": 0.31,
|
515 |
+
"rewards/accuracies": 0.5375000238418579,
|
516 |
+
"rewards/chosen": 0.01972355879843235,
|
517 |
+
"rewards/margins": 0.03841744735836983,
|
518 |
+
"rewards/rejected": -0.018693890422582626,
|
519 |
"step": 320
|
520 |
},
|
521 |
{
|
522 |
"epoch": 0.86,
|
523 |
"learning_rate": 3.053082288996112e-08,
|
524 |
+
"logits/chosen": -2.015927314758301,
|
525 |
+
"logits/rejected": -2.014465570449829,
|
526 |
+
"logps/chosen": -33.47549057006836,
|
527 |
+
"logps/rejected": -32.47983169555664,
|
528 |
+
"loss": 0.3454,
|
529 |
+
"rewards/accuracies": 0.512499988079071,
|
530 |
+
"rewards/chosen": 0.009270393289625645,
|
531 |
+
"rewards/margins": 0.009491671808063984,
|
532 |
+
"rewards/rejected": -0.0002212822437286377,
|
533 |
"step": 330
|
534 |
},
|
535 |
{
|
536 |
"epoch": 0.88,
|
537 |
"learning_rate": 2.05793773749158e-08,
|
538 |
+
"logits/chosen": -2.102839946746826,
|
539 |
+
"logits/rejected": -2.0870471000671387,
|
540 |
+
"logps/chosen": -34.18696212768555,
|
541 |
+
"logps/rejected": -33.075828552246094,
|
542 |
+
"loss": 0.4092,
|
543 |
+
"rewards/accuracies": 0.4124999940395355,
|
544 |
+
"rewards/chosen": -0.006814646068960428,
|
545 |
+
"rewards/margins": -0.04479615017771721,
|
546 |
+
"rewards/rejected": 0.03798150271177292,
|
547 |
"step": 340
|
548 |
},
|
549 |
{
|
550 |
"epoch": 0.91,
|
551 |
"learning_rate": 1.251801807404168e-08,
|
552 |
+
"logits/chosen": -1.9747005701065063,
|
553 |
+
"logits/rejected": -1.9737575054168701,
|
554 |
+
"logps/chosen": -33.231842041015625,
|
555 |
+
"logps/rejected": -32.461700439453125,
|
556 |
+
"loss": 0.3207,
|
557 |
+
"rewards/accuracies": 0.5249999761581421,
|
558 |
+
"rewards/chosen": 0.04523162543773651,
|
559 |
+
"rewards/margins": 0.04212886095046997,
|
560 |
+
"rewards/rejected": 0.003102770773693919,
|
561 |
"step": 350
|
562 |
},
|
563 |
{
|
564 |
"epoch": 0.94,
|
565 |
"learning_rate": 6.41315865106129e-09,
|
566 |
+
"logits/chosen": -1.9306046962738037,
|
567 |
+
"logits/rejected": -1.9409637451171875,
|
568 |
+
"logps/chosen": -32.217220306396484,
|
569 |
+
"logps/rejected": -35.306907653808594,
|
570 |
+
"loss": 0.3411,
|
571 |
+
"rewards/accuracies": 0.4625000059604645,
|
572 |
+
"rewards/chosen": -0.011275654658675194,
|
573 |
+
"rewards/margins": 0.005062489304691553,
|
574 |
+
"rewards/rejected": -0.01633814536035061,
|
575 |
"step": 360
|
576 |
},
|
577 |
{
|
578 |
"epoch": 0.96,
|
579 |
"learning_rate": 2.3150941078050324e-09,
|
580 |
+
"logits/chosen": -2.0694775581359863,
|
581 |
+
"logits/rejected": -2.0629212856292725,
|
582 |
+
"logps/chosen": -33.65714645385742,
|
583 |
+
"logps/rejected": -29.221160888671875,
|
584 |
+
"loss": 0.3549,
|
585 |
+
"rewards/accuracies": 0.4749999940395355,
|
586 |
+
"rewards/chosen": -0.004660401493310928,
|
587 |
+
"rewards/margins": -0.009401795454323292,
|
588 |
+
"rewards/rejected": 0.0047413939610123634,
|
589 |
"step": 370
|
590 |
},
|
591 |
{
|
592 |
"epoch": 0.99,
|
593 |
"learning_rate": 2.575864278703266e-10,
|
594 |
+
"logits/chosen": -1.9296157360076904,
|
595 |
+
"logits/rejected": -1.931777000427246,
|
596 |
+
"logps/chosen": -34.22825241088867,
|
597 |
+
"logps/rejected": -30.912120819091797,
|
598 |
+
"loss": 0.2952,
|
599 |
+
"rewards/accuracies": 0.5375000238418579,
|
600 |
+
"rewards/chosen": 0.009714365005493164,
|
601 |
+
"rewards/margins": 0.04525749385356903,
|
602 |
+
"rewards/rejected": -0.035543132573366165,
|
603 |
"step": 380
|
604 |
},
|
605 |
{
|
606 |
"epoch": 1.0,
|
607 |
"step": 385,
|
608 |
"total_flos": 0.0,
|
609 |
+
"train_loss": 0.0744757689438857,
|
610 |
+
"train_runtime": 628.0651,
|
611 |
+
"train_samples_per_second": 4.902,
|
612 |
+
"train_steps_per_second": 0.613
|
613 |
}
|
614 |
],
|
615 |
"logging_steps": 10,
|
training_args.bin
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 4984
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:c79706b878a2ab50513a99855a2b6398ad70408a38470eb167a411505f0bd5e7
|
3 |
size 4984
|