Model save
Browse files- README.md +12 -12
- all_results.json +3 -3
- config.json +1 -1
- model-00001-of-00003.safetensors +1 -1
- model-00002-of-00003.safetensors +1 -1
- model-00003-of-00003.safetensors +1 -1
- runs/May29_00-48-35_n136-100-194/events.out.tfevents.1716915257.n136-100-194.1004174.0 +2 -2
- train_results.json +3 -3
- trainer_state.json +413 -983
- training_args.bin +1 -1
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: 1.
|
19 |
-
- Rewards/chosen: -5.
|
20 |
-
- Rewards/rejected: -5.
|
21 |
-
- Rewards/accuracies: 0.
|
22 |
-
- Rewards/margins: 0.
|
23 |
-
- Logps/rejected: -
|
24 |
-
- Logps/chosen: -
|
25 |
-
- Logits/rejected: -
|
26 |
-
- Logits/chosen: -
|
27 |
|
28 |
## Model description
|
29 |
|
@@ -60,9 +60,9 @@ 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.
|
64 |
-
| 0.
|
65 |
-
| 0.
|
66 |
|
67 |
|
68 |
### 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: 1.1843
|
19 |
+
- Rewards/chosen: -5.6098
|
20 |
+
- Rewards/rejected: -5.9639
|
21 |
+
- Rewards/accuracies: 0.5117
|
22 |
+
- Rewards/margins: 0.3541
|
23 |
+
- Logps/rejected: -1114.7808
|
24 |
+
- Logps/chosen: -951.4574
|
25 |
+
- Logits/rejected: -7.9900
|
26 |
+
- Logits/chosen: -7.4446
|
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 |
+
| 0.4015 | 0.26 | 100 | 0.9856 | -4.3603 | -4.6945 | 0.5273 | 0.3341 | -987.8358 | -826.5081 | -6.7933 | -6.4109 |
|
64 |
+
| 0.3649 | 0.53 | 200 | 1.1239 | -4.8760 | -5.1429 | 0.4883 | 0.2669 | -1032.6809 | -878.0756 | -7.6378 | -7.1525 |
|
65 |
+
| 0.3506 | 0.79 | 300 | 1.1843 | -5.6098 | -5.9639 | 0.5117 | 0.3541 | -1114.7808 | -951.4574 | -7.9900 | -7.4446 |
|
66 |
|
67 |
|
68 |
### 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": 48530,
|
6 |
-
"train_samples_per_second": 8.
|
7 |
"train_steps_per_second": 0.065
|
8 |
}
|
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
+
"train_loss": 0.39733357479830217,
|
4 |
+
"train_runtime": 5845.1464,
|
5 |
"train_samples": 48530,
|
6 |
+
"train_samples_per_second": 8.303,
|
7 |
"train_steps_per_second": 0.065
|
8 |
}
|
config.json
CHANGED
@@ -1,5 +1,5 @@
|
|
1 |
{
|
2 |
-
"_name_or_path": "/mnt/bn/xuruijie-llm/checkpoints/
|
3 |
"architectures": [
|
4 |
"MistralForCausalLM"
|
5 |
],
|
|
|
1 |
{
|
2 |
+
"_name_or_path": "/mnt/bn/xuruijie-llm/checkpoints/new_world/v1-ultral",
|
3 |
"architectures": [
|
4 |
"MistralForCausalLM"
|
5 |
],
|
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:f0448d3619348a37bb87330d8a16f03f2db67a5e69f6d423351f4de49ba701b4
|
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 sha256:0afb427ab7948f6418557c415a2087f26e4a2bcf506844122d917521dfff33e5
|
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:a88ed491b35c8995c182b2bbbff2fb9fb8eab8e0c1d499eee5dfc7d7e795619f
|
3 |
size 4540532728
|
runs/May29_00-48-35_n136-100-194/events.out.tfevents.1716915257.n136-100-194.1004174.0
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 sha256:5ae0e5f7c711b9eb4df161831dead31f4a5f4e24c11dec46a6894b2cd1129f93
|
3 |
+
size 33418
|
train_results.json
CHANGED
@@ -1,8 +1,8 @@
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
-
"train_loss": 0.
|
4 |
-
"train_runtime":
|
5 |
"train_samples": 48530,
|
6 |
-
"train_samples_per_second": 8.
|
7 |
"train_steps_per_second": 0.065
|
8 |
}
|
|
|
1 |
{
|
2 |
"epoch": 1.0,
|
3 |
+
"train_loss": 0.39733357479830217,
|
4 |
+
"train_runtime": 5845.1464,
|
5 |
"train_samples": 48530,
|
6 |
+
"train_samples_per_second": 8.303,
|
7 |
"train_steps_per_second": 0.065
|
8 |
}
|
trainer_state.json
CHANGED
@@ -10,12 +10,12 @@
|
|
10 |
"log_history": [
|
11 |
{
|
12 |
"epoch": 0.0,
|
13 |
-
"grad_norm":
|
14 |
"learning_rate": 1.3157894736842104e-08,
|
15 |
-
"logits/chosen": -
|
16 |
-
"logits/rejected": -
|
17 |
-
"logps/chosen": -
|
18 |
-
"logps/rejected": -
|
19 |
"loss": 0.6931,
|
20 |
"rewards/accuracies": 0.0,
|
21 |
"rewards/chosen": 0.0,
|
@@ -23,1190 +23,620 @@
|
|
23 |
"rewards/rejected": 0.0,
|
24 |
"step": 1
|
25 |
},
|
26 |
-
{
|
27 |
-
"epoch": 0.01,
|
28 |
-
"grad_norm": 35.006735982686784,
|
29 |
-
"learning_rate": 6.578947368421052e-08,
|
30 |
-
"logits/chosen": -2.276287078857422,
|
31 |
-
"logits/rejected": -2.2080204486846924,
|
32 |
-
"logps/chosen": -283.5397644042969,
|
33 |
-
"logps/rejected": -208.59442138671875,
|
34 |
-
"loss": 0.6931,
|
35 |
-
"rewards/accuracies": 0.40625,
|
36 |
-
"rewards/chosen": 0.0013411559630185366,
|
37 |
-
"rewards/margins": 0.00017823810048867017,
|
38 |
-
"rewards/rejected": 0.0011629178188741207,
|
39 |
-
"step": 5
|
40 |
-
},
|
41 |
{
|
42 |
"epoch": 0.03,
|
43 |
-
"grad_norm":
|
44 |
"learning_rate": 1.3157894736842104e-07,
|
45 |
-
"logits/chosen": -
|
46 |
-
"logits/rejected": -
|
47 |
-
"logps/chosen": -
|
48 |
-
"logps/rejected": -
|
49 |
-
"loss": 0.
|
50 |
-
"rewards/accuracies": 0.
|
51 |
-
"rewards/chosen": 0.
|
52 |
-
"rewards/margins":
|
53 |
-
"rewards/rejected": 0.
|
54 |
"step": 10
|
55 |
},
|
56 |
-
{
|
57 |
-
"epoch": 0.04,
|
58 |
-
"grad_norm": 34.873528802219326,
|
59 |
-
"learning_rate": 1.9736842105263157e-07,
|
60 |
-
"logits/chosen": -2.3293230533599854,
|
61 |
-
"logits/rejected": -2.2996506690979004,
|
62 |
-
"logps/chosen": -289.71771240234375,
|
63 |
-
"logps/rejected": -237.6280059814453,
|
64 |
-
"loss": 0.6906,
|
65 |
-
"rewards/accuracies": 0.612500011920929,
|
66 |
-
"rewards/chosen": 0.017419874668121338,
|
67 |
-
"rewards/margins": 0.004653422627598047,
|
68 |
-
"rewards/rejected": 0.012766450643539429,
|
69 |
-
"step": 15
|
70 |
-
},
|
71 |
{
|
72 |
"epoch": 0.05,
|
73 |
-
"grad_norm":
|
74 |
"learning_rate": 2.631578947368421e-07,
|
75 |
-
"logits/chosen": -
|
76 |
-
"logits/rejected": -
|
77 |
-
"logps/chosen": -
|
78 |
-
"logps/rejected": -
|
79 |
-
"loss": 0.
|
80 |
-
"rewards/accuracies": 0.
|
81 |
-
"rewards/chosen": 0.
|
82 |
-
"rewards/margins": 0.
|
83 |
-
"rewards/rejected": 0.
|
84 |
"step": 20
|
85 |
},
|
86 |
-
{
|
87 |
-
"epoch": 0.07,
|
88 |
-
"grad_norm": 28.076450357377052,
|
89 |
-
"learning_rate": 3.2894736842105264e-07,
|
90 |
-
"logits/chosen": -2.3921802043914795,
|
91 |
-
"logits/rejected": -2.377410650253296,
|
92 |
-
"logps/chosen": -260.9800720214844,
|
93 |
-
"logps/rejected": -214.38821411132812,
|
94 |
-
"loss": 0.6763,
|
95 |
-
"rewards/accuracies": 0.762499988079071,
|
96 |
-
"rewards/chosen": 0.13004128634929657,
|
97 |
-
"rewards/margins": 0.03864985704421997,
|
98 |
-
"rewards/rejected": 0.0913914293050766,
|
99 |
-
"step": 25
|
100 |
-
},
|
101 |
{
|
102 |
"epoch": 0.08,
|
103 |
-
"grad_norm":
|
104 |
"learning_rate": 3.9473684210526315e-07,
|
105 |
-
"logits/chosen": -
|
106 |
-
"logits/rejected": -
|
107 |
-
"logps/chosen": -
|
108 |
-
"logps/rejected": -
|
109 |
-
"loss": 0.
|
110 |
-
"rewards/accuracies": 0.
|
111 |
-
"rewards/chosen": 0.
|
112 |
-
"rewards/margins": 0.
|
113 |
-
"rewards/rejected": 0.
|
114 |
"step": 30
|
115 |
},
|
116 |
-
{
|
117 |
-
"epoch": 0.09,
|
118 |
-
"grad_norm": 23.40327238126855,
|
119 |
-
"learning_rate": 4.6052631578947365e-07,
|
120 |
-
"logits/chosen": -2.6838037967681885,
|
121 |
-
"logits/rejected": -2.6064510345458984,
|
122 |
-
"logps/chosen": -272.02825927734375,
|
123 |
-
"logps/rejected": -207.57565307617188,
|
124 |
-
"loss": 0.6576,
|
125 |
-
"rewards/accuracies": 0.800000011920929,
|
126 |
-
"rewards/chosen": 0.3294451832771301,
|
127 |
-
"rewards/margins": 0.12016526609659195,
|
128 |
-
"rewards/rejected": 0.20927992463111877,
|
129 |
-
"step": 35
|
130 |
-
},
|
131 |
{
|
132 |
"epoch": 0.11,
|
133 |
-
"grad_norm":
|
134 |
"learning_rate": 4.999575626062319e-07,
|
135 |
-
"logits/chosen": -
|
136 |
-
"logits/rejected": -
|
137 |
-
"logps/chosen": -
|
138 |
-
"logps/rejected": -
|
139 |
-
"loss": 0.
|
140 |
-
"rewards/accuracies": 0.
|
141 |
-
"rewards/chosen": 0.
|
142 |
-
"rewards/margins": 0.
|
143 |
-
"rewards/rejected": 0.
|
144 |
"step": 40
|
145 |
},
|
146 |
-
{
|
147 |
-
"epoch": 0.12,
|
148 |
-
"grad_norm": 18.3841620762551,
|
149 |
-
"learning_rate": 4.994803073715569e-07,
|
150 |
-
"logits/chosen": -2.788217067718506,
|
151 |
-
"logits/rejected": -2.736720561981201,
|
152 |
-
"logps/chosen": -260.8193359375,
|
153 |
-
"logps/rejected": -202.5912322998047,
|
154 |
-
"loss": 0.6444,
|
155 |
-
"rewards/accuracies": 0.7875000238418579,
|
156 |
-
"rewards/chosen": 0.4205988049507141,
|
157 |
-
"rewards/margins": 0.14394915103912354,
|
158 |
-
"rewards/rejected": 0.2766496539115906,
|
159 |
-
"step": 45
|
160 |
-
},
|
161 |
{
|
162 |
"epoch": 0.13,
|
163 |
-
"grad_norm":
|
164 |
"learning_rate": 4.984737660598186e-07,
|
165 |
-
"logits/chosen": -
|
166 |
-
"logits/rejected": -
|
167 |
-
"logps/chosen": -
|
168 |
-
"logps/rejected": -
|
169 |
-
"loss": 0.
|
170 |
-
"rewards/accuracies": 0.
|
171 |
-
"rewards/chosen": 0.
|
172 |
-
"rewards/margins": 0.
|
173 |
-
"rewards/rejected":
|
174 |
"step": 50
|
175 |
},
|
176 |
-
{
|
177 |
-
"epoch": 0.14,
|
178 |
-
"grad_norm": 20.93264303618464,
|
179 |
-
"learning_rate": 4.969400741032999e-07,
|
180 |
-
"logits/chosen": -2.8368711471557617,
|
181 |
-
"logits/rejected": -2.8192451000213623,
|
182 |
-
"logps/chosen": -234.53515625,
|
183 |
-
"logps/rejected": -199.7665252685547,
|
184 |
-
"loss": 0.6341,
|
185 |
-
"rewards/accuracies": 0.737500011920929,
|
186 |
-
"rewards/chosen": 0.33271390199661255,
|
187 |
-
"rewards/margins": 0.14281830191612244,
|
188 |
-
"rewards/rejected": 0.18989557027816772,
|
189 |
-
"step": 55
|
190 |
-
},
|
191 |
{
|
192 |
"epoch": 0.16,
|
193 |
-
"grad_norm":
|
194 |
"learning_rate": 4.948824853131236e-07,
|
195 |
-
"logits/chosen": -
|
196 |
-
"logits/rejected": -
|
197 |
-
"logps/chosen": -
|
198 |
-
"logps/rejected": -
|
199 |
-
"loss": 0.
|
200 |
-
"rewards/accuracies": 0.
|
201 |
-
"rewards/chosen":
|
202 |
-
"rewards/margins": 0.
|
203 |
-
"rewards/rejected":
|
204 |
"step": 60
|
205 |
},
|
206 |
-
{
|
207 |
-
"epoch": 0.17,
|
208 |
-
"grad_norm": 19.6177452377838,
|
209 |
-
"learning_rate": 4.923053649761152e-07,
|
210 |
-
"logits/chosen": -2.8169569969177246,
|
211 |
-
"logits/rejected": -2.7849762439727783,
|
212 |
-
"logps/chosen": -244.50808715820312,
|
213 |
-
"logps/rejected": -201.7557830810547,
|
214 |
-
"loss": 0.6149,
|
215 |
-
"rewards/accuracies": 0.7875000238418579,
|
216 |
-
"rewards/chosen": 0.35878387093544006,
|
217 |
-
"rewards/margins": 0.19476714730262756,
|
218 |
-
"rewards/rejected": 0.1640167087316513,
|
219 |
-
"step": 65
|
220 |
-
},
|
221 |
{
|
222 |
"epoch": 0.18,
|
223 |
-
"grad_norm":
|
224 |
"learning_rate": 4.892141805936084e-07,
|
225 |
-
"logits/chosen": -
|
226 |
-
"logits/rejected": -
|
227 |
-
"logps/chosen": -
|
228 |
-
"logps/rejected": -
|
229 |
-
"loss": 0.
|
230 |
-
"rewards/accuracies": 0.
|
231 |
-
"rewards/chosen":
|
232 |
-
"rewards/margins": 0.
|
233 |
-
"rewards/rejected":
|
234 |
"step": 70
|
235 |
},
|
236 |
-
{
|
237 |
-
"epoch": 0.2,
|
238 |
-
"grad_norm": 25.663818502097516,
|
239 |
-
"learning_rate": 4.856154902818431e-07,
|
240 |
-
"logits/chosen": -2.8126118183135986,
|
241 |
-
"logits/rejected": -2.7708253860473633,
|
242 |
-
"logps/chosen": -229.9053955078125,
|
243 |
-
"logps/rejected": -206.87258911132812,
|
244 |
-
"loss": 0.6004,
|
245 |
-
"rewards/accuracies": 0.75,
|
246 |
-
"rewards/chosen": 0.2887989580631256,
|
247 |
-
"rewards/margins": 0.22907009720802307,
|
248 |
-
"rewards/rejected": 0.05972885340452194,
|
249 |
-
"step": 75
|
250 |
-
},
|
251 |
{
|
252 |
"epoch": 0.21,
|
253 |
-
"grad_norm":
|
254 |
"learning_rate": 4.81516928858564e-07,
|
255 |
-
"logits/chosen": -
|
256 |
-
"logits/rejected": -
|
257 |
-
"logps/chosen": -
|
258 |
-
"logps/rejected": -
|
259 |
-
"loss": 0.
|
260 |
-
"rewards/accuracies": 0.8125,
|
261 |
-
"rewards/chosen": 0.3578353524208069,
|
262 |
-
"rewards/margins": 0.31964898109436035,
|
263 |
-
"rewards/rejected": 0.03818630054593086,
|
264 |
-
"step": 80
|
265 |
-
},
|
266 |
-
{
|
267 |
-
"epoch": 0.22,
|
268 |
-
"grad_norm": 26.200013285180134,
|
269 |
-
"learning_rate": 4.769271916453385e-07,
|
270 |
-
"logits/chosen": -2.817336320877075,
|
271 |
-
"logits/rejected": -2.7950100898742676,
|
272 |
-
"logps/chosen": -257.9813537597656,
|
273 |
-
"logps/rejected": -234.77108764648438,
|
274 |
-
"loss": 0.5692,
|
275 |
"rewards/accuracies": 0.800000011920929,
|
276 |
-
"rewards/chosen":
|
277 |
-
"rewards/margins":
|
278 |
-
"rewards/rejected": -
|
279 |
-
"step":
|
280 |
},
|
281 |
{
|
282 |
"epoch": 0.24,
|
283 |
-
"grad_norm":
|
284 |
"learning_rate": 4.7185601601995784e-07,
|
285 |
-
"logits/chosen": -
|
286 |
-
"logits/rejected": -
|
287 |
-
"logps/chosen": -
|
288 |
-
"logps/rejected": -
|
289 |
-
"loss": 0.
|
290 |
-
"rewards/accuracies": 0.
|
291 |
-
"rewards/chosen":
|
292 |
-
"rewards/margins":
|
293 |
-
"rewards/rejected": -
|
294 |
"step": 90
|
295 |
},
|
296 |
-
{
|
297 |
-
"epoch": 0.25,
|
298 |
-
"grad_norm": 30.203747907953968,
|
299 |
-
"learning_rate": 4.6631416075805886e-07,
|
300 |
-
"logits/chosen": -2.9183051586151123,
|
301 |
-
"logits/rejected": -2.8942832946777344,
|
302 |
-
"logps/chosen": -292.6710510253906,
|
303 |
-
"logps/rejected": -278.3623962402344,
|
304 |
-
"loss": 0.5293,
|
305 |
-
"rewards/accuracies": 0.762499988079071,
|
306 |
-
"rewards/chosen": 0.0014784678351134062,
|
307 |
-
"rewards/margins": 0.45240503549575806,
|
308 |
-
"rewards/rejected": -0.4509265422821045,
|
309 |
-
"step": 95
|
310 |
-
},
|
311 |
{
|
312 |
"epoch": 0.26,
|
313 |
-
"grad_norm":
|
314 |
"learning_rate": 4.603133832077953e-07,
|
315 |
-
"logits/chosen": -
|
316 |
-
"logits/rejected": -
|
317 |
-
"logps/chosen": -
|
318 |
-
"logps/rejected": -
|
319 |
-
"loss": 0.
|
320 |
-
"rewards/accuracies": 0.
|
321 |
-
"rewards/chosen": -
|
322 |
-
"rewards/margins":
|
323 |
-
"rewards/rejected": -
|
324 |
"step": 100
|
325 |
},
|
326 |
{
|
327 |
"epoch": 0.26,
|
328 |
-
"eval_logits/chosen": -
|
329 |
-
"eval_logits/rejected": -
|
330 |
-
"eval_logps/chosen": -
|
331 |
-
"eval_logps/rejected": -
|
332 |
-
"eval_loss": 0.
|
333 |
-
"eval_rewards/accuracies": 0.
|
334 |
-
"eval_rewards/chosen": -
|
335 |
-
"eval_rewards/margins": 0.
|
336 |
-
"eval_rewards/rejected": -
|
337 |
-
"eval_runtime": 97.
|
338 |
-
"eval_samples_per_second": 20.
|
339 |
"eval_steps_per_second": 0.328,
|
340 |
"step": 100
|
341 |
},
|
342 |
-
{
|
343 |
-
"epoch": 0.28,
|
344 |
-
"grad_norm": 35.37658079116714,
|
345 |
-
"learning_rate": 4.538664143459818e-07,
|
346 |
-
"logits/chosen": -3.1525657176971436,
|
347 |
-
"logits/rejected": -3.170300245285034,
|
348 |
-
"logps/chosen": -301.29736328125,
|
349 |
-
"logps/rejected": -306.07171630859375,
|
350 |
-
"loss": 0.4875,
|
351 |
-
"rewards/accuracies": 0.8374999761581421,
|
352 |
-
"rewards/chosen": -0.40715986490249634,
|
353 |
-
"rewards/margins": 0.6011335849761963,
|
354 |
-
"rewards/rejected": -1.0082933902740479,
|
355 |
-
"step": 105
|
356 |
-
},
|
357 |
{
|
358 |
"epoch": 0.29,
|
359 |
-
"grad_norm":
|
360 |
"learning_rate": 4.4698693176863316e-07,
|
361 |
-
"logits/chosen": -
|
362 |
-
"logits/rejected": -
|
363 |
-
"logps/chosen": -
|
364 |
-
"logps/rejected": -
|
365 |
-
"loss": 0.
|
366 |
-
"rewards/accuracies": 0.800000011920929,
|
367 |
-
"rewards/chosen": -0.6683047413825989,
|
368 |
-
"rewards/margins": 0.806756854057312,
|
369 |
-
"rewards/rejected": -1.4750616550445557,
|
370 |
-
"step": 110
|
371 |
-
},
|
372 |
-
{
|
373 |
-
"epoch": 0.3,
|
374 |
-
"grad_norm": 46.11054779372386,
|
375 |
-
"learning_rate": 4.396895306731977e-07,
|
376 |
-
"logits/chosen": -3.2922375202178955,
|
377 |
-
"logits/rejected": -3.3002562522888184,
|
378 |
-
"logps/chosen": -316.46331787109375,
|
379 |
-
"logps/rejected": -344.7227783203125,
|
380 |
-
"loss": 0.478,
|
381 |
"rewards/accuracies": 0.824999988079071,
|
382 |
-
"rewards/chosen": -
|
383 |
-
"rewards/margins":
|
384 |
-
"rewards/rejected": -
|
385 |
-
"step":
|
386 |
},
|
387 |
{
|
388 |
"epoch": 0.32,
|
389 |
-
"grad_norm":
|
390 |
"learning_rate": 4.319896928940505e-07,
|
391 |
-
"logits/chosen": -
|
392 |
-
"logits/rejected": -
|
393 |
-
"logps/chosen": -
|
394 |
-
"logps/rejected": -
|
395 |
-
"loss": 0.
|
396 |
-
"rewards/accuracies": 0.
|
397 |
-
"rewards/chosen": -
|
398 |
-
"rewards/margins":
|
399 |
-
"rewards/rejected": -
|
400 |
"step": 120
|
401 |
},
|
402 |
-
{
|
403 |
-
"epoch": 0.33,
|
404 |
-
"grad_norm": 36.84462226750656,
|
405 |
-
"learning_rate": 4.2390375405693723e-07,
|
406 |
-
"logits/chosen": -3.474997043609619,
|
407 |
-
"logits/rejected": -3.5272536277770996,
|
408 |
-
"logps/chosen": -415.80438232421875,
|
409 |
-
"logps/rejected": -468.0596618652344,
|
410 |
-
"loss": 0.4422,
|
411 |
-
"rewards/accuracies": 0.762499988079071,
|
412 |
-
"rewards/chosen": -1.5106348991394043,
|
413 |
-
"rewards/margins": 1.0294691324234009,
|
414 |
-
"rewards/rejected": -2.5401039123535156,
|
415 |
-
"step": 125
|
416 |
-
},
|
417 |
{
|
418 |
"epoch": 0.34,
|
419 |
-
"grad_norm":
|
420 |
"learning_rate": 4.1544886892205354e-07,
|
421 |
-
"logits/chosen": -
|
422 |
-
"logits/rejected": -
|
423 |
-
"logps/chosen": -
|
424 |
-
"logps/rejected": -
|
425 |
-
"loss": 0.
|
426 |
-
"rewards/accuracies": 0.
|
427 |
-
"rewards/chosen": -
|
428 |
-
"rewards/margins":
|
429 |
-
"rewards/rejected": -
|
430 |
"step": 130
|
431 |
},
|
432 |
-
{
|
433 |
-
"epoch": 0.36,
|
434 |
-
"grad_norm": 49.77750455648269,
|
435 |
-
"learning_rate": 4.0664297498928534e-07,
|
436 |
-
"logits/chosen": -3.6749653816223145,
|
437 |
-
"logits/rejected": -3.6941752433776855,
|
438 |
-
"logps/chosen": -392.3956604003906,
|
439 |
-
"logps/rejected": -454.9180603027344,
|
440 |
-
"loss": 0.4261,
|
441 |
-
"rewards/accuracies": 0.800000011920929,
|
442 |
-
"rewards/chosen": -1.5022857189178467,
|
443 |
-
"rewards/margins": 1.0955121517181396,
|
444 |
-
"rewards/rejected": -2.5977978706359863,
|
445 |
-
"step": 135
|
446 |
-
},
|
447 |
{
|
448 |
"epoch": 0.37,
|
449 |
-
"grad_norm":
|
450 |
"learning_rate": 3.975047544428254e-07,
|
451 |
-
"logits/chosen": -
|
452 |
-
"logits/rejected": -
|
453 |
-
"logps/chosen": -
|
454 |
-
"logps/rejected": -
|
455 |
-
"loss": 0.
|
456 |
-
"rewards/accuracies": 0.
|
457 |
-
"rewards/chosen": -
|
458 |
-
"rewards/margins": 1.
|
459 |
-
"rewards/rejected": -
|
460 |
"step": 140
|
461 |
},
|
462 |
-
{
|
463 |
-
"epoch": 0.38,
|
464 |
-
"grad_norm": 37.39808184046548,
|
465 |
-
"learning_rate": 3.880535945158997e-07,
|
466 |
-
"logits/chosen": -3.7747676372528076,
|
467 |
-
"logits/rejected": -3.7997519969940186,
|
468 |
-
"logps/chosen": -367.59576416015625,
|
469 |
-
"logps/rejected": -449.857177734375,
|
470 |
-
"loss": 0.4127,
|
471 |
-
"rewards/accuracies": 0.7749999761581421,
|
472 |
-
"rewards/chosen": -1.2651011943817139,
|
473 |
-
"rewards/margins": 1.1386528015136719,
|
474 |
-
"rewards/rejected": -2.403754234313965,
|
475 |
-
"step": 145
|
476 |
-
},
|
477 |
{
|
478 |
"epoch": 0.4,
|
479 |
-
"grad_norm":
|
480 |
"learning_rate": 3.78309546359696e-07,
|
481 |
-
"logits/chosen": -
|
482 |
-
"logits/rejected": -
|
483 |
-
"logps/chosen": -
|
484 |
-
"logps/rejected": -
|
485 |
-
"loss": 0.
|
486 |
-
"rewards/accuracies": 0.
|
487 |
-
"rewards/chosen": -
|
488 |
-
"rewards/margins": 1.
|
489 |
-
"rewards/rejected": -
|
490 |
"step": 150
|
491 |
},
|
492 |
-
{
|
493 |
-
"epoch": 0.41,
|
494 |
-
"grad_norm": 40.77943292899045,
|
495 |
-
"learning_rate": 3.6829328250375227e-07,
|
496 |
-
"logits/chosen": -3.932398557662964,
|
497 |
-
"logits/rejected": -4.020986080169678,
|
498 |
-
"logps/chosen": -427.6490783691406,
|
499 |
-
"logps/rejected": -508.22930908203125,
|
500 |
-
"loss": 0.3809,
|
501 |
-
"rewards/accuracies": 0.824999988079071,
|
502 |
-
"rewards/chosen": -1.581094741821289,
|
503 |
-
"rewards/margins": 1.3298404216766357,
|
504 |
-
"rewards/rejected": -2.910935401916504,
|
505 |
-
"step": 155
|
506 |
-
},
|
507 |
{
|
508 |
"epoch": 0.42,
|
509 |
-
"grad_norm":
|
510 |
"learning_rate": 3.580260529980584e-07,
|
511 |
-
"logits/chosen": -
|
512 |
-
"logits/rejected": -
|
513 |
-
"logps/chosen": -
|
514 |
-
"logps/rejected": -
|
515 |
-
"loss": 0.
|
516 |
-
"rewards/accuracies": 0.
|
517 |
-
"rewards/chosen": -
|
518 |
-
"rewards/margins": 1.
|
519 |
-
"rewards/rejected": -
|
520 |
"step": 160
|
521 |
},
|
522 |
-
{
|
523 |
-
"epoch": 0.43,
|
524 |
-
"grad_norm": 40.85253910181408,
|
525 |
-
"learning_rate": 3.475296403299163e-07,
|
526 |
-
"logits/chosen": -4.092155456542969,
|
527 |
-
"logits/rejected": -4.100491523742676,
|
528 |
-
"logps/chosen": -388.4344482421875,
|
529 |
-
"logps/rejected": -494.84771728515625,
|
530 |
-
"loss": 0.391,
|
531 |
-
"rewards/accuracies": 0.824999988079071,
|
532 |
-
"rewards/chosen": -1.4383682012557983,
|
533 |
-
"rewards/margins": 1.5076179504394531,
|
534 |
-
"rewards/rejected": -2.945986032485962,
|
535 |
-
"step": 165
|
536 |
-
},
|
537 |
{
|
538 |
"epoch": 0.45,
|
539 |
-
"grad_norm":
|
540 |
"learning_rate": 3.36826313211205e-07,
|
541 |
-
"logits/chosen": -
|
542 |
-
"logits/rejected": -
|
543 |
-
"logps/chosen": -
|
544 |
-
"logps/rejected": -
|
545 |
-
"loss": 0.
|
546 |
-
"rewards/accuracies": 0.
|
547 |
-
"rewards/chosen": -
|
548 |
-
"rewards/margins": 1.
|
549 |
-
"rewards/rejected": -
|
550 |
"step": 170
|
551 |
},
|
552 |
-
{
|
553 |
-
"epoch": 0.46,
|
554 |
-
"grad_norm": 40.910837008073884,
|
555 |
-
"learning_rate": 3.259387793340943e-07,
|
556 |
-
"logits/chosen": -4.390842437744141,
|
557 |
-
"logits/rejected": -4.466560363769531,
|
558 |
-
"logps/chosen": -482.41748046875,
|
559 |
-
"logps/rejected": -587.3478393554688,
|
560 |
-
"loss": 0.3845,
|
561 |
-
"rewards/accuracies": 0.7875000238418579,
|
562 |
-
"rewards/chosen": -2.141714096069336,
|
563 |
-
"rewards/margins": 1.4235177040100098,
|
564 |
-
"rewards/rejected": -3.5652313232421875,
|
565 |
-
"step": 175
|
566 |
-
},
|
567 |
{
|
568 |
"epoch": 0.47,
|
569 |
-
"grad_norm":
|
570 |
"learning_rate": 3.14890137195437e-07,
|
571 |
-
"logits/chosen": -
|
572 |
-
"logits/rejected": -
|
573 |
-
"logps/chosen": -
|
574 |
-
"logps/rejected": -
|
575 |
-
"loss": 0.
|
576 |
-
"rewards/accuracies": 0.
|
577 |
-
"rewards/chosen": -2.
|
578 |
-
"rewards/margins": 1.
|
579 |
-
"rewards/rejected": -
|
580 |
"step": 180
|
581 |
},
|
582 |
-
{
|
583 |
-
"epoch": 0.49,
|
584 |
-
"grad_norm": 48.89695189111101,
|
585 |
-
"learning_rate": 3.0370382709204883e-07,
|
586 |
-
"logits/chosen": -4.416136264801025,
|
587 |
-
"logits/rejected": -4.519248962402344,
|
588 |
-
"logps/chosen": -472.10931396484375,
|
589 |
-
"logps/rejected": -609.0546264648438,
|
590 |
-
"loss": 0.3879,
|
591 |
-
"rewards/accuracies": 0.887499988079071,
|
592 |
-
"rewards/chosen": -1.9735286235809326,
|
593 |
-
"rewards/margins": 1.7880016565322876,
|
594 |
-
"rewards/rejected": -3.7615303993225098,
|
595 |
-
"step": 185
|
596 |
-
},
|
597 |
{
|
598 |
"epoch": 0.5,
|
599 |
-
"grad_norm":
|
600 |
"learning_rate": 2.9240358139084013e-07,
|
601 |
-
"logits/chosen": -
|
602 |
-
"logits/rejected": -
|
603 |
-
"logps/chosen": -
|
604 |
-
"logps/rejected": -
|
605 |
-
"loss": 0.
|
606 |
-
"rewards/accuracies": 0.
|
607 |
-
"rewards/chosen": -
|
608 |
-
"rewards/margins":
|
609 |
-
"rewards/rejected": -
|
610 |
"step": 190
|
611 |
},
|
612 |
-
{
|
613 |
-
"epoch": 0.51,
|
614 |
-
"grad_norm": 42.794361716855846,
|
615 |
-
"learning_rate": 2.810133741793052e-07,
|
616 |
-
"logits/chosen": -4.521183490753174,
|
617 |
-
"logits/rejected": -4.65042781829834,
|
618 |
-
"logps/chosen": -472.19500732421875,
|
619 |
-
"logps/rejected": -594.0242309570312,
|
620 |
-
"loss": 0.3875,
|
621 |
-
"rewards/accuracies": 0.862500011920929,
|
622 |
-
"rewards/chosen": -2.0321810245513916,
|
623 |
-
"rewards/margins": 1.7259247303009033,
|
624 |
-
"rewards/rejected": -3.758105516433716,
|
625 |
-
"step": 195
|
626 |
-
},
|
627 |
{
|
628 |
"epoch": 0.53,
|
629 |
-
"grad_norm":
|
630 |
"learning_rate": 2.695573704031885e-07,
|
631 |
-
"logits/chosen": -
|
632 |
-
"logits/rejected": -
|
633 |
-
"logps/chosen": -
|
634 |
-
"logps/rejected": -
|
635 |
-
"loss": 0.
|
636 |
-
"rewards/accuracies": 0.
|
637 |
-
"rewards/chosen": -
|
638 |
-
"rewards/margins": 1.
|
639 |
-
"rewards/rejected": -
|
640 |
"step": 200
|
641 |
},
|
642 |
{
|
643 |
"epoch": 0.53,
|
644 |
-
"eval_logits/chosen": -
|
645 |
-
"eval_logits/rejected": -
|
646 |
-
"eval_logps/chosen": -
|
647 |
-
"eval_logps/rejected": -
|
648 |
-
"eval_loss":
|
649 |
-
"eval_rewards/accuracies": 0.
|
650 |
-
"eval_rewards/chosen": -4.
|
651 |
-
"eval_rewards/margins": 0.
|
652 |
-
"eval_rewards/rejected": -
|
653 |
-
"eval_runtime": 97.
|
654 |
-
"eval_samples_per_second": 20.
|
655 |
-
"eval_steps_per_second": 0.
|
656 |
"step": 200
|
657 |
},
|
658 |
-
{
|
659 |
-
"epoch": 0.54,
|
660 |
-
"grad_norm": 45.90321806356986,
|
661 |
-
"learning_rate": 2.580598745992342e-07,
|
662 |
-
"logits/chosen": -4.592051029205322,
|
663 |
-
"logits/rejected": -4.729592800140381,
|
664 |
-
"logps/chosen": -462.078857421875,
|
665 |
-
"logps/rejected": -578.6028442382812,
|
666 |
-
"loss": 0.3683,
|
667 |
-
"rewards/accuracies": 0.8500000238418579,
|
668 |
-
"rewards/chosen": -1.6925300359725952,
|
669 |
-
"rewards/margins": 1.7680349349975586,
|
670 |
-
"rewards/rejected": -3.4605648517608643,
|
671 |
-
"step": 205
|
672 |
-
},
|
673 |
{
|
674 |
"epoch": 0.55,
|
675 |
-
"grad_norm":
|
676 |
"learning_rate": 2.465452793317865e-07,
|
677 |
-
"logits/chosen": -
|
678 |
-
"logits/rejected": -
|
679 |
-
"logps/chosen": -
|
680 |
-
"logps/rejected": -
|
681 |
-
"loss": 0.
|
682 |
-
"rewards/accuracies": 0.
|
683 |
-
"rewards/chosen": -
|
684 |
-
"rewards/margins": 1.
|
685 |
-
"rewards/rejected": -
|
686 |
"step": 210
|
687 |
},
|
688 |
-
{
|
689 |
-
"epoch": 0.57,
|
690 |
-
"grad_norm": 50.566593421325194,
|
691 |
-
"learning_rate": 2.3503801344263344e-07,
|
692 |
-
"logits/chosen": -4.860222339630127,
|
693 |
-
"logits/rejected": -4.944591045379639,
|
694 |
-
"logps/chosen": -456.22357177734375,
|
695 |
-
"logps/rejected": -573.6387939453125,
|
696 |
-
"loss": 0.3712,
|
697 |
-
"rewards/accuracies": 0.862500011920929,
|
698 |
-
"rewards/chosen": -2.1773338317871094,
|
699 |
-
"rewards/margins": 1.6265901327133179,
|
700 |
-
"rewards/rejected": -3.8039238452911377,
|
701 |
-
"step": 215
|
702 |
-
},
|
703 |
{
|
704 |
"epoch": 0.58,
|
705 |
-
"grad_norm":
|
706 |
"learning_rate": 2.2356249022388789e-07,
|
707 |
-
"logits/chosen": -
|
708 |
-
"logits/rejected": -
|
709 |
-
"logps/chosen": -
|
710 |
-
"logps/rejected": -
|
711 |
-
"loss": 0.
|
712 |
-
"rewards/accuracies": 0.
|
713 |
-
"rewards/chosen": -
|
714 |
-
"rewards/margins":
|
715 |
-
"rewards/rejected": -
|
716 |
"step": 220
|
717 |
},
|
718 |
-
{
|
719 |
-
"epoch": 0.59,
|
720 |
-
"grad_norm": 48.71640334228914,
|
721 |
-
"learning_rate": 2.121430556238559e-07,
|
722 |
-
"logits/chosen": -4.8175554275512695,
|
723 |
-
"logits/rejected": -4.996689319610596,
|
724 |
-
"logps/chosen": -471.17132568359375,
|
725 |
-
"logps/rejected": -604.3876342773438,
|
726 |
-
"loss": 0.3386,
|
727 |
-
"rewards/accuracies": 0.862500011920929,
|
728 |
-
"rewards/chosen": -2.008169412612915,
|
729 |
-
"rewards/margins": 1.8934684991836548,
|
730 |
-
"rewards/rejected": -3.9016380310058594,
|
731 |
-
"step": 225
|
732 |
-
},
|
733 |
{
|
734 |
"epoch": 0.61,
|
735 |
-
"grad_norm":
|
736 |
"learning_rate": 2.0080393659578038e-07,
|
737 |
-
"logits/chosen": -
|
738 |
-
"logits/rejected": -
|
739 |
-
"logps/chosen": -
|
740 |
-
"logps/rejected": -
|
741 |
-
"loss": 0.
|
742 |
-
"rewards/accuracies": 0.
|
743 |
-
"rewards/chosen": -
|
744 |
-
"rewards/margins": 2.
|
745 |
-
"rewards/rejected": -
|
746 |
"step": 230
|
747 |
},
|
748 |
-
{
|
749 |
-
"epoch": 0.62,
|
750 |
-
"grad_norm": 50.20217799621439,
|
751 |
-
"learning_rate": 1.895691896990388e-07,
|
752 |
-
"logits/chosen": -4.822530269622803,
|
753 |
-
"logits/rejected": -4.9560699462890625,
|
754 |
-
"logps/chosen": -474.0023498535156,
|
755 |
-
"logps/rejected": -579.5379638671875,
|
756 |
-
"loss": 0.3354,
|
757 |
-
"rewards/accuracies": 0.800000011920929,
|
758 |
-
"rewards/chosen": -2.1355338096618652,
|
759 |
-
"rewards/margins": 1.5200008153915405,
|
760 |
-
"rewards/rejected": -3.655534267425537,
|
761 |
-
"step": 235
|
762 |
-
},
|
763 |
{
|
764 |
"epoch": 0.63,
|
765 |
-
"grad_norm":
|
766 |
"learning_rate": 1.7846265006183976e-07,
|
767 |
-
"logits/chosen": -
|
768 |
-
"logits/rejected": -
|
769 |
-
"logps/chosen": -
|
770 |
-
"logps/rejected": -
|
771 |
-
"loss": 0.
|
772 |
-
"rewards/accuracies": 0.
|
773 |
-
"rewards/chosen": -
|
774 |
-
"rewards/margins": 1.
|
775 |
-
"rewards/rejected": -
|
776 |
"step": 240
|
777 |
},
|
778 |
-
{
|
779 |
-
"epoch": 0.65,
|
780 |
-
"grad_norm": 50.41005236551354,
|
781 |
-
"learning_rate": 1.6750788081369948e-07,
|
782 |
-
"logits/chosen": -4.760382175445557,
|
783 |
-
"logits/rejected": -4.9731292724609375,
|
784 |
-
"logps/chosen": -475.59765625,
|
785 |
-
"logps/rejected": -638.8780517578125,
|
786 |
-
"loss": 0.3533,
|
787 |
-
"rewards/accuracies": 0.887499988079071,
|
788 |
-
"rewards/chosen": -2.0123469829559326,
|
789 |
-
"rewards/margins": 2.291203737258911,
|
790 |
-
"rewards/rejected": -4.303550720214844,
|
791 |
-
"step": 245
|
792 |
-
},
|
793 |
{
|
794 |
"epoch": 0.66,
|
795 |
-
"grad_norm": 41.
|
796 |
"learning_rate": 1.5672812309497722e-07,
|
797 |
-
"logits/chosen": -
|
798 |
-
"logits/rejected": -
|
799 |
-
"logps/chosen": -
|
800 |
-
"logps/rejected": -
|
801 |
-
"loss": 0.
|
802 |
-
"rewards/accuracies": 0.
|
803 |
-
"rewards/chosen": -
|
804 |
-
"rewards/margins":
|
805 |
-
"rewards/rejected": -
|
806 |
"step": 250
|
807 |
},
|
808 |
-
{
|
809 |
-
"epoch": 0.67,
|
810 |
-
"grad_norm": 44.69671736661672,
|
811 |
-
"learning_rate": 1.461462467495284e-07,
|
812 |
-
"logits/chosen": -4.758819580078125,
|
813 |
-
"logits/rejected": -4.947306156158447,
|
814 |
-
"logps/chosen": -495.4164123535156,
|
815 |
-
"logps/rejected": -593.2689208984375,
|
816 |
-
"loss": 0.3703,
|
817 |
-
"rewards/accuracies": 0.7875000238418579,
|
818 |
-
"rewards/chosen": -2.218562602996826,
|
819 |
-
"rewards/margins": 1.439822793006897,
|
820 |
-
"rewards/rejected": -3.6583850383758545,
|
821 |
-
"step": 255
|
822 |
-
},
|
823 |
{
|
824 |
"epoch": 0.69,
|
825 |
-
"grad_norm":
|
826 |
"learning_rate": 1.357847018050843e-07,
|
827 |
-
"logits/chosen": -
|
828 |
-
"logits/rejected": -
|
829 |
-
"logps/chosen": -
|
830 |
-
"logps/rejected": -
|
831 |
-
"loss": 0.
|
832 |
-
"rewards/accuracies": 0.
|
833 |
-
"rewards/chosen": -2.
|
834 |
-
"rewards/margins":
|
835 |
-
"rewards/rejected": -4.
|
836 |
"step": 260
|
837 |
},
|
838 |
-
{
|
839 |
-
"epoch": 0.7,
|
840 |
-
"grad_norm": 49.07485622430381,
|
841 |
-
"learning_rate": 1.2566547084429324e-07,
|
842 |
-
"logits/chosen": -4.789057731628418,
|
843 |
-
"logits/rejected": -5.023972988128662,
|
844 |
-
"logps/chosen": -472.7372131347656,
|
845 |
-
"logps/rejected": -615.0753784179688,
|
846 |
-
"loss": 0.3591,
|
847 |
-
"rewards/accuracies": 0.800000011920929,
|
848 |
-
"rewards/chosen": -2.2092373371124268,
|
849 |
-
"rewards/margins": 1.8947007656097412,
|
850 |
-
"rewards/rejected": -4.103938579559326,
|
851 |
-
"step": 265
|
852 |
-
},
|
853 |
{
|
854 |
"epoch": 0.71,
|
855 |
-
"grad_norm":
|
856 |
"learning_rate": 1.1581002236747328e-07,
|
857 |
-
"logits/chosen": -
|
858 |
-
"logits/rejected": -
|
859 |
-
"logps/chosen": -470.
|
860 |
-
"logps/rejected": -
|
861 |
-
"loss": 0.
|
862 |
-
"rewards/accuracies": 0.
|
863 |
-
"rewards/chosen": -2.
|
864 |
-
"rewards/margins": 2.
|
865 |
-
"rewards/rejected": -4.
|
866 |
"step": 270
|
867 |
},
|
868 |
-
{
|
869 |
-
"epoch": 0.72,
|
870 |
-
"grad_norm": 53.087469451573526,
|
871 |
-
"learning_rate": 1.062392652460177e-07,
|
872 |
-
"logits/chosen": -4.696314811706543,
|
873 |
-
"logits/rejected": -4.83748197555542,
|
874 |
-
"logps/chosen": -505.9961853027344,
|
875 |
-
"logps/rejected": -626.0050048828125,
|
876 |
-
"loss": 0.3245,
|
877 |
-
"rewards/accuracies": 0.862500011920929,
|
878 |
-
"rewards/chosen": -1.9974615573883057,
|
879 |
-
"rewards/margins": 1.828150987625122,
|
880 |
-
"rewards/rejected": -3.8256123065948486,
|
881 |
-
"step": 275
|
882 |
-
},
|
883 |
{
|
884 |
"epoch": 0.74,
|
885 |
-
"grad_norm":
|
886 |
"learning_rate": 9.697350436308427e-08,
|
887 |
-
"logits/chosen": -
|
888 |
-
"logits/rejected": -
|
889 |
-
"logps/chosen": -
|
890 |
-
"logps/rejected": -
|
891 |
-
"loss": 0.
|
892 |
-
"rewards/accuracies": 0.
|
893 |
-
"rewards/chosen": -
|
894 |
-
"rewards/margins": 1.
|
895 |
-
"rewards/rejected": -
|
896 |
"step": 280
|
897 |
},
|
898 |
-
{
|
899 |
-
"epoch": 0.75,
|
900 |
-
"grad_norm": 44.45010163032659,
|
901 |
-
"learning_rate": 8.803239753567829e-08,
|
902 |
-
"logits/chosen": -4.868664741516113,
|
903 |
-
"logits/rejected": -4.975742340087891,
|
904 |
-
"logps/chosen": -469.70263671875,
|
905 |
-
"logps/rejected": -589.5235595703125,
|
906 |
-
"loss": 0.3498,
|
907 |
-
"rewards/accuracies": 0.875,
|
908 |
-
"rewards/chosen": -2.1795949935913086,
|
909 |
-
"rewards/margins": 1.7358201742172241,
|
910 |
-
"rewards/rejected": -3.9154155254364014,
|
911 |
-
"step": 285
|
912 |
-
},
|
913 |
{
|
914 |
"epoch": 0.76,
|
915 |
-
"grad_norm":
|
916 |
"learning_rate": 7.943491380952188e-08,
|
917 |
-
"logits/chosen": -
|
918 |
-
"logits/rejected": -
|
919 |
-
"logps/chosen": -
|
920 |
-
"logps/rejected": -
|
921 |
-
"loss": 0.
|
922 |
"rewards/accuracies": 0.862500011920929,
|
923 |
-
"rewards/chosen": -
|
924 |
-
"rewards/margins":
|
925 |
-
"rewards/rejected": -
|
926 |
"step": 290
|
927 |
},
|
928 |
-
{
|
929 |
-
"epoch": 0.78,
|
930 |
-
"grad_norm": 47.705363932547336,
|
931 |
-
"learning_rate": 7.119929321518875e-08,
|
932 |
-
"logits/chosen": -4.735751152038574,
|
933 |
-
"logits/rejected": -5.044283866882324,
|
934 |
-
"logps/chosen": -445.70367431640625,
|
935 |
-
"logps/rejected": -583.5177612304688,
|
936 |
-
"loss": 0.3436,
|
937 |
-
"rewards/accuracies": 0.8125,
|
938 |
-
"rewards/chosen": -1.9234256744384766,
|
939 |
-
"rewards/margins": 1.9092767238616943,
|
940 |
-
"rewards/rejected": -3.832702159881592,
|
941 |
-
"step": 295
|
942 |
-
},
|
943 |
{
|
944 |
"epoch": 0.79,
|
945 |
-
"grad_norm":
|
946 |
"learning_rate": 6.334300807088508e-08,
|
947 |
-
"logits/chosen": -
|
948 |
-
"logits/rejected": -
|
949 |
-
"logps/chosen": -
|
950 |
-
"logps/rejected": -
|
951 |
-
"loss": 0.
|
952 |
-
"rewards/accuracies": 0.
|
953 |
-
"rewards/chosen": -
|
954 |
-
"rewards/margins":
|
955 |
-
"rewards/rejected": -
|
956 |
"step": 300
|
957 |
},
|
958 |
{
|
959 |
"epoch": 0.79,
|
960 |
-
"eval_logits/chosen": -
|
961 |
-
"eval_logits/rejected": -
|
962 |
-
"eval_logps/chosen": -
|
963 |
-
"eval_logps/rejected": -
|
964 |
-
"eval_loss": 1.
|
965 |
-
"eval_rewards/accuracies": 0.
|
966 |
-
"eval_rewards/chosen": -5.
|
967 |
-
"eval_rewards/margins": 0.
|
968 |
-
"eval_rewards/rejected": -5.
|
969 |
-
"eval_runtime": 97.
|
970 |
-
"eval_samples_per_second": 20.
|
971 |
-
"eval_steps_per_second": 0.
|
972 |
"step": 300
|
973 |
},
|
974 |
-
{
|
975 |
-
"epoch": 0.8,
|
976 |
-
"grad_norm": 50.197907161154795,
|
977 |
-
"learning_rate": 5.588272591397336e-08,
|
978 |
-
"logits/chosen": -4.786068916320801,
|
979 |
-
"logits/rejected": -4.9945068359375,
|
980 |
-
"logps/chosen": -485.31280517578125,
|
981 |
-
"logps/rejected": -626.9613647460938,
|
982 |
-
"loss": 0.3457,
|
983 |
-
"rewards/accuracies": 0.824999988079071,
|
984 |
-
"rewards/chosen": -2.227161169052124,
|
985 |
-
"rewards/margins": 2.0316004753112793,
|
986 |
-
"rewards/rejected": -4.258761405944824,
|
987 |
-
"step": 305
|
988 |
-
},
|
989 |
{
|
990 |
"epoch": 0.82,
|
991 |
-
"grad_norm":
|
992 |
"learning_rate": 4.8834274139883084e-08,
|
993 |
-
"logits/chosen": -
|
994 |
-
"logits/rejected": -
|
995 |
-
"logps/chosen": -
|
996 |
-
"logps/rejected": -
|
997 |
-
"loss": 0.
|
998 |
-
"rewards/accuracies": 0.
|
999 |
-
"rewards/chosen": -2.
|
1000 |
-
"rewards/margins":
|
1001 |
-
"rewards/rejected": -
|
1002 |
"step": 310
|
1003 |
},
|
1004 |
-
{
|
1005 |
-
"epoch": 0.83,
|
1006 |
-
"grad_norm": 43.92355028448363,
|
1007 |
-
"learning_rate": 4.221260642342786e-08,
|
1008 |
-
"logits/chosen": -4.848982810974121,
|
1009 |
-
"logits/rejected": -4.972110271453857,
|
1010 |
-
"logps/chosen": -502.3720703125,
|
1011 |
-
"logps/rejected": -609.923828125,
|
1012 |
-
"loss": 0.3357,
|
1013 |
-
"rewards/accuracies": 0.8500000238418579,
|
1014 |
-
"rewards/chosen": -2.262599229812622,
|
1015 |
-
"rewards/margins": 1.6350996494293213,
|
1016 |
-
"rewards/rejected": -3.8976986408233643,
|
1017 |
-
"step": 315
|
1018 |
-
},
|
1019 |
{
|
1020 |
"epoch": 0.84,
|
1021 |
-
"grad_norm":
|
1022 |
"learning_rate": 3.60317709937693e-08,
|
1023 |
-
"logits/chosen": -
|
1024 |
-
"logits/rejected": -
|
1025 |
-
"logps/chosen": -
|
1026 |
-
"logps/rejected": -
|
1027 |
-
"loss": 0.
|
1028 |
-
"rewards/accuracies": 0.
|
1029 |
-
"rewards/chosen": -2.
|
1030 |
-
"rewards/margins": 2.
|
1031 |
-
"rewards/rejected": -4.
|
1032 |
"step": 320
|
1033 |
},
|
1034 |
-
{
|
1035 |
-
"epoch": 0.86,
|
1036 |
-
"grad_norm": 38.024316440823085,
|
1037 |
-
"learning_rate": 3.030488083033273e-08,
|
1038 |
-
"logits/chosen": -4.8308892250061035,
|
1039 |
-
"logits/rejected": -5.002086162567139,
|
1040 |
-
"logps/chosen": -500.25579833984375,
|
1041 |
-
"logps/rejected": -703.7189331054688,
|
1042 |
-
"loss": 0.3434,
|
1043 |
-
"rewards/accuracies": 0.8999999761581421,
|
1044 |
-
"rewards/chosen": -2.2706263065338135,
|
1045 |
-
"rewards/margins": 2.5099949836730957,
|
1046 |
-
"rewards/rejected": -4.780620574951172,
|
1047 |
-
"step": 325
|
1048 |
-
},
|
1049 |
{
|
1050 |
"epoch": 0.87,
|
1051 |
-
"grad_norm":
|
1052 |
"learning_rate": 2.5044085842905683e-08,
|
1053 |
-
"logits/chosen": -
|
1054 |
-
"logits/rejected": -
|
1055 |
-
"logps/chosen": -
|
1056 |
-
"logps/rejected": -
|
1057 |
-
"loss": 0.
|
1058 |
-
"rewards/accuracies": 0.
|
1059 |
-
"rewards/chosen": -
|
1060 |
-
"rewards/margins": 2.
|
1061 |
-
"rewards/rejected": -
|
1062 |
"step": 330
|
1063 |
},
|
1064 |
-
{
|
1065 |
-
"epoch": 0.88,
|
1066 |
-
"grad_norm": 44.94685644229088,
|
1067 |
-
"learning_rate": 2.0260547094942348e-08,
|
1068 |
-
"logits/chosen": -4.779486656188965,
|
1069 |
-
"logits/rejected": -4.989696502685547,
|
1070 |
-
"logps/chosen": -506.99920654296875,
|
1071 |
-
"logps/rejected": -654.0205688476562,
|
1072 |
-
"loss": 0.3649,
|
1073 |
-
"rewards/accuracies": 0.8500000238418579,
|
1074 |
-
"rewards/chosen": -2.142956495285034,
|
1075 |
-
"rewards/margins": 2.1117589473724365,
|
1076 |
-
"rewards/rejected": -4.254715919494629,
|
1077 |
-
"step": 335
|
1078 |
-
},
|
1079 |
{
|
1080 |
"epoch": 0.9,
|
1081 |
-
"grad_norm": 44.
|
1082 |
"learning_rate": 1.5964413124758493e-08,
|
1083 |
-
"logits/chosen": -
|
1084 |
-
"logits/rejected": -
|
1085 |
-
"logps/chosen": -
|
1086 |
-
"logps/rejected": -
|
1087 |
-
"loss": 0.
|
1088 |
-
"rewards/accuracies": 0.
|
1089 |
-
"rewards/chosen": -
|
1090 |
-
"rewards/margins":
|
1091 |
-
"rewards/rejected": -
|
1092 |
"step": 340
|
1093 |
},
|
1094 |
-
{
|
1095 |
-
"epoch": 0.91,
|
1096 |
-
"grad_norm": 49.902887693125585,
|
1097 |
-
"learning_rate": 1.2164798414854071e-08,
|
1098 |
-
"logits/chosen": -4.861344814300537,
|
1099 |
-
"logits/rejected": -4.92690896987915,
|
1100 |
-
"logps/chosen": -514.7852783203125,
|
1101 |
-
"logps/rejected": -692.4171142578125,
|
1102 |
-
"loss": 0.3343,
|
1103 |
-
"rewards/accuracies": 0.8125,
|
1104 |
-
"rewards/chosen": -2.289367437362671,
|
1105 |
-
"rewards/margins": 2.1290056705474854,
|
1106 |
-
"rewards/rejected": -4.418373107910156,
|
1107 |
-
"step": 345
|
1108 |
-
},
|
1109 |
{
|
1110 |
"epoch": 0.92,
|
1111 |
-
"grad_norm":
|
1112 |
"learning_rate": 8.869764055041501e-09,
|
1113 |
-
"logits/chosen": -
|
1114 |
-
"logits/rejected": -
|
1115 |
-
"logps/chosen": -
|
1116 |
-
"logps/rejected": -
|
1117 |
-
"loss": 0.
|
1118 |
-
"rewards/accuracies": 0.
|
1119 |
-
"rewards/chosen": -
|
1120 |
-
"rewards/margins":
|
1121 |
-
"rewards/rejected": -
|
1122 |
"step": 350
|
1123 |
},
|
1124 |
-
{
|
1125 |
-
"epoch": 0.94,
|
1126 |
-
"grad_norm": 45.59470033908034,
|
1127 |
-
"learning_rate": 6.086300640404079e-09,
|
1128 |
-
"logits/chosen": -4.7604827880859375,
|
1129 |
-
"logits/rejected": -4.900928020477295,
|
1130 |
-
"logps/chosen": -530.9966430664062,
|
1131 |
-
"logps/rejected": -647.1216430664062,
|
1132 |
-
"loss": 0.3474,
|
1133 |
-
"rewards/accuracies": 0.737500011920929,
|
1134 |
-
"rewards/chosen": -2.3973419666290283,
|
1135 |
-
"rewards/margins": 1.63766610622406,
|
1136 |
-
"rewards/rejected": -4.035007953643799,
|
1137 |
-
"step": 355
|
1138 |
-
},
|
1139 |
{
|
1140 |
"epoch": 0.95,
|
1141 |
-
"grad_norm":
|
1142 |
"learning_rate": 3.82031344036729e-09,
|
1143 |
-
"logits/chosen": -
|
1144 |
-
"logits/rejected": -
|
1145 |
-
"logps/chosen": -
|
1146 |
-
"logps/rejected": -
|
1147 |
-
"loss": 0.
|
1148 |
-
"rewards/accuracies": 0.
|
1149 |
-
"rewards/chosen": -
|
1150 |
-
"rewards/margins": 1.
|
1151 |
-
"rewards/rejected": -
|
1152 |
"step": 360
|
1153 |
},
|
1154 |
-
{
|
1155 |
-
"epoch": 0.96,
|
1156 |
-
"grad_norm": 48.183739221855284,
|
1157 |
-
"learning_rate": 2.0766098703477173e-09,
|
1158 |
-
"logits/chosen": -4.829585552215576,
|
1159 |
-
"logits/rejected": -5.007233619689941,
|
1160 |
-
"logps/chosen": -458.9842834472656,
|
1161 |
-
"logps/rejected": -589.1207275390625,
|
1162 |
-
"loss": 0.3698,
|
1163 |
-
"rewards/accuracies": 0.800000011920929,
|
1164 |
-
"rewards/chosen": -1.9293123483657837,
|
1165 |
-
"rewards/margins": 1.888193130493164,
|
1166 |
-
"rewards/rejected": -3.8175055980682373,
|
1167 |
-
"step": 365
|
1168 |
-
},
|
1169 |
{
|
1170 |
"epoch": 0.97,
|
1171 |
-
"grad_norm":
|
1172 |
"learning_rate": 8.588892925590063e-10,
|
1173 |
-
"logits/chosen": -
|
1174 |
-
"logits/rejected": -
|
1175 |
-
"logps/chosen": -
|
1176 |
-
"logps/rejected": -
|
1177 |
-
"loss": 0.
|
1178 |
-
"rewards/accuracies": 0.
|
1179 |
-
"rewards/chosen": -2.
|
1180 |
-
"rewards/margins": 2.
|
1181 |
-
"rewards/rejected": -
|
1182 |
"step": 370
|
1183 |
},
|
1184 |
-
{
|
1185 |
-
"epoch": 0.99,
|
1186 |
-
"grad_norm": 48.471497735173386,
|
1187 |
-
"learning_rate": 1.6973516761317752e-10,
|
1188 |
-
"logits/chosen": -4.912912845611572,
|
1189 |
-
"logits/rejected": -5.170907020568848,
|
1190 |
-
"logps/chosen": -459.01531982421875,
|
1191 |
-
"logps/rejected": -576.1976318359375,
|
1192 |
-
"loss": 0.3505,
|
1193 |
-
"rewards/accuracies": 0.8374999761581421,
|
1194 |
-
"rewards/chosen": -2.2589094638824463,
|
1195 |
-
"rewards/margins": 1.6588354110717773,
|
1196 |
-
"rewards/rejected": -3.9177448749542236,
|
1197 |
-
"step": 375
|
1198 |
-
},
|
1199 |
{
|
1200 |
"epoch": 1.0,
|
1201 |
"step": 379,
|
1202 |
"total_flos": 0.0,
|
1203 |
-
"train_loss": 0.
|
1204 |
-
"train_runtime":
|
1205 |
-
"train_samples_per_second": 8.
|
1206 |
"train_steps_per_second": 0.065
|
1207 |
}
|
1208 |
],
|
1209 |
-
"logging_steps":
|
1210 |
"max_steps": 379,
|
1211 |
"num_input_tokens_seen": 0,
|
1212 |
"num_train_epochs": 1,
|
|
|
10 |
"log_history": [
|
11 |
{
|
12 |
"epoch": 0.0,
|
13 |
+
"grad_norm": 34.01362102288599,
|
14 |
"learning_rate": 1.3157894736842104e-08,
|
15 |
+
"logits/chosen": -4.685327529907227,
|
16 |
+
"logits/rejected": -4.87608528137207,
|
17 |
+
"logps/chosen": -207.7137451171875,
|
18 |
+
"logps/rejected": -145.5098114013672,
|
19 |
"loss": 0.6931,
|
20 |
"rewards/accuracies": 0.0,
|
21 |
"rewards/chosen": 0.0,
|
|
|
23 |
"rewards/rejected": 0.0,
|
24 |
"step": 1
|
25 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
26 |
{
|
27 |
"epoch": 0.03,
|
28 |
+
"grad_norm": 33.835137410082986,
|
29 |
"learning_rate": 1.3157894736842104e-07,
|
30 |
+
"logits/chosen": -4.499408721923828,
|
31 |
+
"logits/rejected": -4.84108829498291,
|
32 |
+
"logps/chosen": -223.5843048095703,
|
33 |
+
"logps/rejected": -160.73016357421875,
|
34 |
+
"loss": 0.6929,
|
35 |
+
"rewards/accuracies": 0.4375,
|
36 |
+
"rewards/chosen": 0.0008713232818990946,
|
37 |
+
"rewards/margins": 3.9665635995334014e-05,
|
38 |
+
"rewards/rejected": 0.000831657787784934,
|
39 |
"step": 10
|
40 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
41 |
{
|
42 |
"epoch": 0.05,
|
43 |
+
"grad_norm": 32.9122864599857,
|
44 |
"learning_rate": 2.631578947368421e-07,
|
45 |
+
"logits/chosen": -4.521907806396484,
|
46 |
+
"logits/rejected": -4.8204779624938965,
|
47 |
+
"logps/chosen": -220.248779296875,
|
48 |
+
"logps/rejected": -173.30508422851562,
|
49 |
+
"loss": 0.6818,
|
50 |
+
"rewards/accuracies": 0.768750011920929,
|
51 |
+
"rewards/chosen": 0.01710951328277588,
|
52 |
+
"rewards/margins": 0.023774703964591026,
|
53 |
+
"rewards/rejected": -0.006665193475782871,
|
54 |
"step": 20
|
55 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
56 |
{
|
57 |
"epoch": 0.08,
|
58 |
+
"grad_norm": 32.91997500077958,
|
59 |
"learning_rate": 3.9473684210526315e-07,
|
60 |
+
"logits/chosen": -4.713895320892334,
|
61 |
+
"logits/rejected": -5.012445449829102,
|
62 |
+
"logps/chosen": -223.3787841796875,
|
63 |
+
"logps/rejected": -201.9126434326172,
|
64 |
+
"loss": 0.6294,
|
65 |
+
"rewards/accuracies": 0.8125,
|
66 |
+
"rewards/chosen": -0.047253355383872986,
|
67 |
+
"rewards/margins": 0.1270204484462738,
|
68 |
+
"rewards/rejected": -0.1742737889289856,
|
69 |
"step": 30
|
70 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
71 |
{
|
72 |
"epoch": 0.11,
|
73 |
+
"grad_norm": 44.205273980146465,
|
74 |
"learning_rate": 4.999575626062319e-07,
|
75 |
+
"logits/chosen": -4.862967491149902,
|
76 |
+
"logits/rejected": -5.199351787567139,
|
77 |
+
"logps/chosen": -284.15850830078125,
|
78 |
+
"logps/rejected": -260.1568298339844,
|
79 |
+
"loss": 0.5878,
|
80 |
+
"rewards/accuracies": 0.768750011920929,
|
81 |
+
"rewards/chosen": -0.5039564371109009,
|
82 |
+
"rewards/margins": 0.3390721082687378,
|
83 |
+
"rewards/rejected": -0.8430284261703491,
|
84 |
"step": 40
|
85 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
86 |
{
|
87 |
"epoch": 0.13,
|
88 |
+
"grad_norm": 33.79573625552293,
|
89 |
"learning_rate": 4.984737660598186e-07,
|
90 |
+
"logits/chosen": -4.860326766967773,
|
91 |
+
"logits/rejected": -5.1770920753479,
|
92 |
+
"logps/chosen": -287.49212646484375,
|
93 |
+
"logps/rejected": -275.1466979980469,
|
94 |
+
"loss": 0.544,
|
95 |
+
"rewards/accuracies": 0.7749999761581421,
|
96 |
+
"rewards/chosen": -0.6591841578483582,
|
97 |
+
"rewards/margins": 0.3825286030769348,
|
98 |
+
"rewards/rejected": -1.0417125225067139,
|
99 |
"step": 50
|
100 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
101 |
{
|
102 |
"epoch": 0.16,
|
103 |
+
"grad_norm": 37.5548220875479,
|
104 |
"learning_rate": 4.948824853131236e-07,
|
105 |
+
"logits/chosen": -5.250467777252197,
|
106 |
+
"logits/rejected": -5.646960258483887,
|
107 |
+
"logps/chosen": -355.8065185546875,
|
108 |
+
"logps/rejected": -370.16064453125,
|
109 |
+
"loss": 0.4873,
|
110 |
+
"rewards/accuracies": 0.768750011920929,
|
111 |
+
"rewards/chosen": -1.3389923572540283,
|
112 |
+
"rewards/margins": 0.6164921522140503,
|
113 |
+
"rewards/rejected": -1.955484390258789,
|
114 |
"step": 60
|
115 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
116 |
{
|
117 |
"epoch": 0.18,
|
118 |
+
"grad_norm": 40.80845104625175,
|
119 |
"learning_rate": 4.892141805936084e-07,
|
120 |
+
"logits/chosen": -5.506089687347412,
|
121 |
+
"logits/rejected": -5.9388532638549805,
|
122 |
+
"logps/chosen": -373.0067443847656,
|
123 |
+
"logps/rejected": -409.2164001464844,
|
124 |
+
"loss": 0.4619,
|
125 |
+
"rewards/accuracies": 0.831250011920929,
|
126 |
+
"rewards/chosen": -1.4274970293045044,
|
127 |
+
"rewards/margins": 0.8630696535110474,
|
128 |
+
"rewards/rejected": -2.290566921234131,
|
129 |
"step": 70
|
130 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
131 |
{
|
132 |
"epoch": 0.21,
|
133 |
+
"grad_norm": 62.51614299706751,
|
134 |
"learning_rate": 4.81516928858564e-07,
|
135 |
+
"logits/chosen": -5.957489967346191,
|
136 |
+
"logits/rejected": -6.362034797668457,
|
137 |
+
"logps/chosen": -431.8949279785156,
|
138 |
+
"logps/rejected": -484.7469177246094,
|
139 |
+
"loss": 0.4382,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
140 |
"rewards/accuracies": 0.800000011920929,
|
141 |
+
"rewards/chosen": -2.0529427528381348,
|
142 |
+
"rewards/margins": 1.0657070875167847,
|
143 |
+
"rewards/rejected": -3.118650197982788,
|
144 |
+
"step": 80
|
145 |
},
|
146 |
{
|
147 |
"epoch": 0.24,
|
148 |
+
"grad_norm": 42.64563417225672,
|
149 |
"learning_rate": 4.7185601601995784e-07,
|
150 |
+
"logits/chosen": -5.974350929260254,
|
151 |
+
"logits/rejected": -6.67104959487915,
|
152 |
+
"logps/chosen": -387.2274475097656,
|
153 |
+
"logps/rejected": -464.65826416015625,
|
154 |
+
"loss": 0.4236,
|
155 |
+
"rewards/accuracies": 0.856249988079071,
|
156 |
+
"rewards/chosen": -1.6722873449325562,
|
157 |
+
"rewards/margins": 1.3039813041687012,
|
158 |
+
"rewards/rejected": -2.976268768310547,
|
159 |
"step": 90
|
160 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
161 |
{
|
162 |
"epoch": 0.26,
|
163 |
+
"grad_norm": 49.657859383445725,
|
164 |
"learning_rate": 4.603133832077953e-07,
|
165 |
+
"logits/chosen": -6.59436559677124,
|
166 |
+
"logits/rejected": -7.055686950683594,
|
167 |
+
"logps/chosen": -443.9141540527344,
|
168 |
+
"logps/rejected": -517.8334350585938,
|
169 |
+
"loss": 0.4015,
|
170 |
+
"rewards/accuracies": 0.831250011920929,
|
171 |
+
"rewards/chosen": -2.304152011871338,
|
172 |
+
"rewards/margins": 1.163777232170105,
|
173 |
+
"rewards/rejected": -3.4679291248321533,
|
174 |
"step": 100
|
175 |
},
|
176 |
{
|
177 |
"epoch": 0.26,
|
178 |
+
"eval_logits/chosen": -6.410887718200684,
|
179 |
+
"eval_logits/rejected": -6.793334007263184,
|
180 |
+
"eval_logps/chosen": -826.5081176757812,
|
181 |
+
"eval_logps/rejected": -987.8357543945312,
|
182 |
+
"eval_loss": 0.9856035113334656,
|
183 |
+
"eval_rewards/accuracies": 0.52734375,
|
184 |
+
"eval_rewards/chosen": -4.360336780548096,
|
185 |
+
"eval_rewards/margins": 0.3341439962387085,
|
186 |
+
"eval_rewards/rejected": -4.694480895996094,
|
187 |
+
"eval_runtime": 97.5721,
|
188 |
+
"eval_samples_per_second": 20.498,
|
189 |
"eval_steps_per_second": 0.328,
|
190 |
"step": 100
|
191 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
192 |
{
|
193 |
"epoch": 0.29,
|
194 |
+
"grad_norm": 59.660602657641974,
|
195 |
"learning_rate": 4.4698693176863316e-07,
|
196 |
+
"logits/chosen": -6.508014678955078,
|
197 |
+
"logits/rejected": -6.995993137359619,
|
198 |
+
"logps/chosen": -467.217041015625,
|
199 |
+
"logps/rejected": -571.2991943359375,
|
200 |
+
"loss": 0.3652,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
201 |
"rewards/accuracies": 0.824999988079071,
|
202 |
+
"rewards/chosen": -2.5293219089508057,
|
203 |
+
"rewards/margins": 1.5391861200332642,
|
204 |
+
"rewards/rejected": -4.068508148193359,
|
205 |
+
"step": 110
|
206 |
},
|
207 |
{
|
208 |
"epoch": 0.32,
|
209 |
+
"grad_norm": 48.65436374701769,
|
210 |
"learning_rate": 4.319896928940505e-07,
|
211 |
+
"logits/chosen": -6.631407737731934,
|
212 |
+
"logits/rejected": -7.224958896636963,
|
213 |
+
"logps/chosen": -449.9879455566406,
|
214 |
+
"logps/rejected": -551.552734375,
|
215 |
+
"loss": 0.3856,
|
216 |
+
"rewards/accuracies": 0.7875000238418579,
|
217 |
+
"rewards/chosen": -2.370181083679199,
|
218 |
+
"rewards/margins": 1.3927323818206787,
|
219 |
+
"rewards/rejected": -3.762913465499878,
|
220 |
"step": 120
|
221 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
222 |
{
|
223 |
"epoch": 0.34,
|
224 |
+
"grad_norm": 41.65961698696447,
|
225 |
"learning_rate": 4.1544886892205354e-07,
|
226 |
+
"logits/chosen": -6.516014099121094,
|
227 |
+
"logits/rejected": -7.032981872558594,
|
228 |
+
"logps/chosen": -490.3934020996094,
|
229 |
+
"logps/rejected": -610.3671264648438,
|
230 |
+
"loss": 0.3736,
|
231 |
+
"rewards/accuracies": 0.78125,
|
232 |
+
"rewards/chosen": -2.659696578979492,
|
233 |
+
"rewards/margins": 1.6181414127349854,
|
234 |
+
"rewards/rejected": -4.277838230133057,
|
235 |
"step": 130
|
236 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
237 |
{
|
238 |
"epoch": 0.37,
|
239 |
+
"grad_norm": 61.060769075088814,
|
240 |
"learning_rate": 3.975047544428254e-07,
|
241 |
+
"logits/chosen": -6.8892974853515625,
|
242 |
+
"logits/rejected": -7.428150177001953,
|
243 |
+
"logps/chosen": -479.0992736816406,
|
244 |
+
"logps/rejected": -578.5732421875,
|
245 |
+
"loss": 0.3918,
|
246 |
+
"rewards/accuracies": 0.7875000238418579,
|
247 |
+
"rewards/chosen": -2.728147506713867,
|
248 |
+
"rewards/margins": 1.3954027891159058,
|
249 |
+
"rewards/rejected": -4.123549938201904,
|
250 |
"step": 140
|
251 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
252 |
{
|
253 |
"epoch": 0.4,
|
254 |
+
"grad_norm": 40.0921041292044,
|
255 |
"learning_rate": 3.78309546359696e-07,
|
256 |
+
"logits/chosen": -7.085695743560791,
|
257 |
+
"logits/rejected": -7.593710422515869,
|
258 |
+
"logps/chosen": -485.8174743652344,
|
259 |
+
"logps/rejected": -611.1568603515625,
|
260 |
+
"loss": 0.3821,
|
261 |
+
"rewards/accuracies": 0.8125,
|
262 |
+
"rewards/chosen": -2.8462297916412354,
|
263 |
+
"rewards/margins": 1.527190089225769,
|
264 |
+
"rewards/rejected": -4.373419761657715,
|
265 |
"step": 150
|
266 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
267 |
{
|
268 |
"epoch": 0.42,
|
269 |
+
"grad_norm": 45.94809301945095,
|
270 |
"learning_rate": 3.580260529980584e-07,
|
271 |
+
"logits/chosen": -6.79840087890625,
|
272 |
+
"logits/rejected": -7.57172155380249,
|
273 |
+
"logps/chosen": -476.08197021484375,
|
274 |
+
"logps/rejected": -613.65576171875,
|
275 |
+
"loss": 0.3533,
|
276 |
+
"rewards/accuracies": 0.8062499761581421,
|
277 |
+
"rewards/chosen": -2.6100144386291504,
|
278 |
+
"rewards/margins": 1.7850373983383179,
|
279 |
+
"rewards/rejected": -4.3950514793396,
|
280 |
"step": 160
|
281 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
282 |
{
|
283 |
"epoch": 0.45,
|
284 |
+
"grad_norm": 49.925985141071024,
|
285 |
"learning_rate": 3.36826313211205e-07,
|
286 |
+
"logits/chosen": -7.27915096282959,
|
287 |
+
"logits/rejected": -7.999810695648193,
|
288 |
+
"logps/chosen": -464.27606201171875,
|
289 |
+
"logps/rejected": -618.3897705078125,
|
290 |
+
"loss": 0.3586,
|
291 |
+
"rewards/accuracies": 0.8125,
|
292 |
+
"rewards/chosen": -2.638946056365967,
|
293 |
+
"rewards/margins": 1.9340379238128662,
|
294 |
+
"rewards/rejected": -4.572983741760254,
|
295 |
"step": 170
|
296 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
297 |
{
|
298 |
"epoch": 0.47,
|
299 |
+
"grad_norm": 49.77263247276722,
|
300 |
"learning_rate": 3.14890137195437e-07,
|
301 |
+
"logits/chosen": -7.062252998352051,
|
302 |
+
"logits/rejected": -7.757845878601074,
|
303 |
+
"logps/chosen": -499.0604553222656,
|
304 |
+
"logps/rejected": -631.4205322265625,
|
305 |
+
"loss": 0.35,
|
306 |
+
"rewards/accuracies": 0.8187500238418579,
|
307 |
+
"rewards/chosen": -2.841132640838623,
|
308 |
+
"rewards/margins": 1.6729885339736938,
|
309 |
+
"rewards/rejected": -4.514122009277344,
|
310 |
"step": 180
|
311 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
312 |
{
|
313 |
"epoch": 0.5,
|
314 |
+
"grad_norm": 47.32333832777734,
|
315 |
"learning_rate": 2.9240358139084013e-07,
|
316 |
+
"logits/chosen": -7.432755947113037,
|
317 |
+
"logits/rejected": -8.177480697631836,
|
318 |
+
"logps/chosen": -571.7742919921875,
|
319 |
+
"logps/rejected": -753.3114013671875,
|
320 |
+
"loss": 0.3584,
|
321 |
+
"rewards/accuracies": 0.824999988079071,
|
322 |
+
"rewards/chosen": -3.5975048542022705,
|
323 |
+
"rewards/margins": 2.1865601539611816,
|
324 |
+
"rewards/rejected": -5.784065246582031,
|
325 |
"step": 190
|
326 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
327 |
{
|
328 |
"epoch": 0.53,
|
329 |
+
"grad_norm": 47.64189353399637,
|
330 |
"learning_rate": 2.695573704031885e-07,
|
331 |
+
"logits/chosen": -6.939781188964844,
|
332 |
+
"logits/rejected": -7.71111536026001,
|
333 |
+
"logps/chosen": -500.54449462890625,
|
334 |
+
"logps/rejected": -651.9711303710938,
|
335 |
+
"loss": 0.3649,
|
336 |
+
"rewards/accuracies": 0.8500000238418579,
|
337 |
+
"rewards/chosen": -2.752312421798706,
|
338 |
+
"rewards/margins": 1.9267911911010742,
|
339 |
+
"rewards/rejected": -4.679104328155518,
|
340 |
"step": 200
|
341 |
},
|
342 |
{
|
343 |
"epoch": 0.53,
|
344 |
+
"eval_logits/chosen": -7.152472496032715,
|
345 |
+
"eval_logits/rejected": -7.637792110443115,
|
346 |
+
"eval_logps/chosen": -878.0755615234375,
|
347 |
+
"eval_logps/rejected": -1032.680908203125,
|
348 |
+
"eval_loss": 1.1239182949066162,
|
349 |
+
"eval_rewards/accuracies": 0.48828125,
|
350 |
+
"eval_rewards/chosen": -4.876009941101074,
|
351 |
+
"eval_rewards/margins": 0.26692283153533936,
|
352 |
+
"eval_rewards/rejected": -5.142932891845703,
|
353 |
+
"eval_runtime": 97.9349,
|
354 |
+
"eval_samples_per_second": 20.422,
|
355 |
+
"eval_steps_per_second": 0.327,
|
356 |
"step": 200
|
357 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
358 |
{
|
359 |
"epoch": 0.55,
|
360 |
+
"grad_norm": 54.30113996467665,
|
361 |
"learning_rate": 2.465452793317865e-07,
|
362 |
+
"logits/chosen": -6.918679237365723,
|
363 |
+
"logits/rejected": -7.81919002532959,
|
364 |
+
"logps/chosen": -490.4891662597656,
|
365 |
+
"logps/rejected": -645.8287353515625,
|
366 |
+
"loss": 0.351,
|
367 |
+
"rewards/accuracies": 0.8500000238418579,
|
368 |
+
"rewards/chosen": -2.56154727935791,
|
369 |
+
"rewards/margins": 1.9791193008422852,
|
370 |
+
"rewards/rejected": -4.5406670570373535,
|
371 |
"step": 210
|
372 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
373 |
{
|
374 |
"epoch": 0.58,
|
375 |
+
"grad_norm": 48.87773515687126,
|
376 |
"learning_rate": 2.2356249022388789e-07,
|
377 |
+
"logits/chosen": -7.1236982345581055,
|
378 |
+
"logits/rejected": -8.037015914916992,
|
379 |
+
"logps/chosen": -493.3589782714844,
|
380 |
+
"logps/rejected": -644.3180541992188,
|
381 |
+
"loss": 0.3503,
|
382 |
+
"rewards/accuracies": 0.8812500238418579,
|
383 |
+
"rewards/chosen": -2.7893471717834473,
|
384 |
+
"rewards/margins": 2.016085386276245,
|
385 |
+
"rewards/rejected": -4.805432319641113,
|
386 |
"step": 220
|
387 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
388 |
{
|
389 |
"epoch": 0.61,
|
390 |
+
"grad_norm": 42.654084351174774,
|
391 |
"learning_rate": 2.0080393659578038e-07,
|
392 |
+
"logits/chosen": -7.341279029846191,
|
393 |
+
"logits/rejected": -8.349458694458008,
|
394 |
+
"logps/chosen": -546.9940795898438,
|
395 |
+
"logps/rejected": -735.5299072265625,
|
396 |
+
"loss": 0.3336,
|
397 |
+
"rewards/accuracies": 0.875,
|
398 |
+
"rewards/chosen": -3.2596652507781982,
|
399 |
+
"rewards/margins": 2.3632309436798096,
|
400 |
+
"rewards/rejected": -5.622895240783691,
|
401 |
"step": 230
|
402 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
403 |
{
|
404 |
"epoch": 0.63,
|
405 |
+
"grad_norm": 48.836991831948275,
|
406 |
"learning_rate": 1.7846265006183976e-07,
|
407 |
+
"logits/chosen": -7.382364749908447,
|
408 |
+
"logits/rejected": -8.096385955810547,
|
409 |
+
"logps/chosen": -559.1055908203125,
|
410 |
+
"logps/rejected": -690.3636474609375,
|
411 |
+
"loss": 0.3239,
|
412 |
+
"rewards/accuracies": 0.78125,
|
413 |
+
"rewards/chosen": -3.478092670440674,
|
414 |
+
"rewards/margins": 1.7500022649765015,
|
415 |
+
"rewards/rejected": -5.228094577789307,
|
416 |
"step": 240
|
417 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
418 |
{
|
419 |
"epoch": 0.66,
|
420 |
+
"grad_norm": 41.88272757196894,
|
421 |
"learning_rate": 1.5672812309497722e-07,
|
422 |
+
"logits/chosen": -7.3642730712890625,
|
423 |
+
"logits/rejected": -8.137662887573242,
|
424 |
+
"logps/chosen": -481.7484436035156,
|
425 |
+
"logps/rejected": -658.2447509765625,
|
426 |
+
"loss": 0.3435,
|
427 |
+
"rewards/accuracies": 0.831250011920929,
|
428 |
+
"rewards/chosen": -2.742130756378174,
|
429 |
+
"rewards/margins": 2.204786777496338,
|
430 |
+
"rewards/rejected": -4.946917533874512,
|
431 |
"step": 250
|
432 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
433 |
{
|
434 |
"epoch": 0.69,
|
435 |
+
"grad_norm": 42.22161557256122,
|
436 |
"learning_rate": 1.357847018050843e-07,
|
437 |
+
"logits/chosen": -7.285035133361816,
|
438 |
+
"logits/rejected": -8.04423999786377,
|
439 |
+
"logps/chosen": -495.21929931640625,
|
440 |
+
"logps/rejected": -637.4512939453125,
|
441 |
+
"loss": 0.3575,
|
442 |
+
"rewards/accuracies": 0.856249988079071,
|
443 |
+
"rewards/chosen": -2.5677530765533447,
|
444 |
+
"rewards/margins": 1.859423041343689,
|
445 |
+
"rewards/rejected": -4.427175521850586,
|
446 |
"step": 260
|
447 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
448 |
{
|
449 |
"epoch": 0.71,
|
450 |
+
"grad_norm": 36.62148112990462,
|
451 |
"learning_rate": 1.1581002236747328e-07,
|
452 |
+
"logits/chosen": -7.350560188293457,
|
453 |
+
"logits/rejected": -8.334383964538574,
|
454 |
+
"logps/chosen": -470.9756774902344,
|
455 |
+
"logps/rejected": -644.9085083007812,
|
456 |
+
"loss": 0.3419,
|
457 |
+
"rewards/accuracies": 0.831250011920929,
|
458 |
+
"rewards/chosen": -2.7616236209869385,
|
459 |
+
"rewards/margins": 2.1581368446350098,
|
460 |
+
"rewards/rejected": -4.919760704040527,
|
461 |
"step": 270
|
462 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
463 |
{
|
464 |
"epoch": 0.74,
|
465 |
+
"grad_norm": 42.46981787888043,
|
466 |
"learning_rate": 9.697350436308427e-08,
|
467 |
+
"logits/chosen": -7.353733062744141,
|
468 |
+
"logits/rejected": -8.123218536376953,
|
469 |
+
"logps/chosen": -547.2511596679688,
|
470 |
+
"logps/rejected": -669.4581298828125,
|
471 |
+
"loss": 0.3277,
|
472 |
+
"rewards/accuracies": 0.8500000238418579,
|
473 |
+
"rewards/chosen": -3.1215298175811768,
|
474 |
+
"rewards/margins": 1.7408149242401123,
|
475 |
+
"rewards/rejected": -4.862344264984131,
|
476 |
"step": 280
|
477 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
478 |
{
|
479 |
"epoch": 0.76,
|
480 |
+
"grad_norm": 45.06982020026593,
|
481 |
"learning_rate": 7.943491380952188e-08,
|
482 |
+
"logits/chosen": -7.6978254318237305,
|
483 |
+
"logits/rejected": -8.468889236450195,
|
484 |
+
"logps/chosen": -488.8667907714844,
|
485 |
+
"logps/rejected": -652.3041381835938,
|
486 |
+
"loss": 0.3357,
|
487 |
"rewards/accuracies": 0.862500011920929,
|
488 |
+
"rewards/chosen": -2.835703134536743,
|
489 |
+
"rewards/margins": 2.056410789489746,
|
490 |
+
"rewards/rejected": -4.89211368560791,
|
491 |
"step": 290
|
492 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
493 |
{
|
494 |
"epoch": 0.79,
|
495 |
+
"grad_norm": 50.46942762667779,
|
496 |
"learning_rate": 6.334300807088508e-08,
|
497 |
+
"logits/chosen": -7.3240485191345215,
|
498 |
+
"logits/rejected": -8.329621315002441,
|
499 |
+
"logps/chosen": -509.5638122558594,
|
500 |
+
"logps/rejected": -673.7567749023438,
|
501 |
+
"loss": 0.3506,
|
502 |
+
"rewards/accuracies": 0.8500000238418579,
|
503 |
+
"rewards/chosen": -3.0604350566864014,
|
504 |
+
"rewards/margins": 2.1173126697540283,
|
505 |
+
"rewards/rejected": -5.1777472496032715,
|
506 |
"step": 300
|
507 |
},
|
508 |
{
|
509 |
"epoch": 0.79,
|
510 |
+
"eval_logits/chosen": -7.444587707519531,
|
511 |
+
"eval_logits/rejected": -7.9900360107421875,
|
512 |
+
"eval_logps/chosen": -951.4573974609375,
|
513 |
+
"eval_logps/rejected": -1114.78076171875,
|
514 |
+
"eval_loss": 1.1842519044876099,
|
515 |
+
"eval_rewards/accuracies": 0.51171875,
|
516 |
+
"eval_rewards/chosen": -5.609828948974609,
|
517 |
+
"eval_rewards/margins": 0.35410135984420776,
|
518 |
+
"eval_rewards/rejected": -5.963930606842041,
|
519 |
+
"eval_runtime": 97.9073,
|
520 |
+
"eval_samples_per_second": 20.427,
|
521 |
+
"eval_steps_per_second": 0.327,
|
522 |
"step": 300
|
523 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
524 |
{
|
525 |
"epoch": 0.82,
|
526 |
+
"grad_norm": 45.51913834484837,
|
527 |
"learning_rate": 4.8834274139883084e-08,
|
528 |
+
"logits/chosen": -7.380696773529053,
|
529 |
+
"logits/rejected": -8.350247383117676,
|
530 |
+
"logps/chosen": -504.37054443359375,
|
531 |
+
"logps/rejected": -678.6265869140625,
|
532 |
+
"loss": 0.3248,
|
533 |
+
"rewards/accuracies": 0.84375,
|
534 |
+
"rewards/chosen": -2.946779727935791,
|
535 |
+
"rewards/margins": 2.2168285846710205,
|
536 |
+
"rewards/rejected": -5.163609027862549,
|
537 |
"step": 310
|
538 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
539 |
{
|
540 |
"epoch": 0.84,
|
541 |
+
"grad_norm": 53.031113784661194,
|
542 |
"learning_rate": 3.60317709937693e-08,
|
543 |
+
"logits/chosen": -7.483295440673828,
|
544 |
+
"logits/rejected": -8.33633804321289,
|
545 |
+
"logps/chosen": -519.6754150390625,
|
546 |
+
"logps/rejected": -667.0064697265625,
|
547 |
+
"loss": 0.334,
|
548 |
+
"rewards/accuracies": 0.824999988079071,
|
549 |
+
"rewards/chosen": -2.902660369873047,
|
550 |
+
"rewards/margins": 2.0172414779663086,
|
551 |
+
"rewards/rejected": -4.919901371002197,
|
552 |
"step": 320
|
553 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
554 |
{
|
555 |
"epoch": 0.87,
|
556 |
+
"grad_norm": 41.68995945520798,
|
557 |
"learning_rate": 2.5044085842905683e-08,
|
558 |
+
"logits/chosen": -7.596086025238037,
|
559 |
+
"logits/rejected": -8.42108154296875,
|
560 |
+
"logps/chosen": -542.1593627929688,
|
561 |
+
"logps/rejected": -761.1915893554688,
|
562 |
+
"loss": 0.3439,
|
563 |
+
"rewards/accuracies": 0.9312499761581421,
|
564 |
+
"rewards/chosen": -3.237351894378662,
|
565 |
+
"rewards/margins": 2.630371332168579,
|
566 |
+
"rewards/rejected": -5.867722988128662,
|
567 |
"step": 330
|
568 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
569 |
{
|
570 |
"epoch": 0.9,
|
571 |
+
"grad_norm": 44.38600213815557,
|
572 |
"learning_rate": 1.5964413124758493e-08,
|
573 |
+
"logits/chosen": -7.379315376281738,
|
574 |
+
"logits/rejected": -8.234747886657715,
|
575 |
+
"logps/chosen": -516.7742919921875,
|
576 |
+
"logps/rejected": -690.0197143554688,
|
577 |
+
"loss": 0.3394,
|
578 |
+
"rewards/accuracies": 0.875,
|
579 |
+
"rewards/chosen": -2.9615204334259033,
|
580 |
+
"rewards/margins": 2.160275936126709,
|
581 |
+
"rewards/rejected": -5.121796607971191,
|
582 |
"step": 340
|
583 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
584 |
{
|
585 |
"epoch": 0.92,
|
586 |
+
"grad_norm": 49.890826040889124,
|
587 |
"learning_rate": 8.869764055041501e-09,
|
588 |
+
"logits/chosen": -7.525488376617432,
|
589 |
+
"logits/rejected": -8.198974609375,
|
590 |
+
"logps/chosen": -535.8231201171875,
|
591 |
+
"logps/rejected": -737.1505737304688,
|
592 |
+
"loss": 0.3403,
|
593 |
+
"rewards/accuracies": 0.8374999761581421,
|
594 |
+
"rewards/chosen": -3.1437525749206543,
|
595 |
+
"rewards/margins": 2.3052563667297363,
|
596 |
+
"rewards/rejected": -5.449008941650391,
|
597 |
"step": 350
|
598 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
599 |
{
|
600 |
"epoch": 0.95,
|
601 |
+
"grad_norm": 41.28866479778979,
|
602 |
"learning_rate": 3.82031344036729e-09,
|
603 |
+
"logits/chosen": -7.295458793640137,
|
604 |
+
"logits/rejected": -8.147361755371094,
|
605 |
+
"logps/chosen": -541.268798828125,
|
606 |
+
"logps/rejected": -692.4953002929688,
|
607 |
+
"loss": 0.3191,
|
608 |
+
"rewards/accuracies": 0.8187500238418579,
|
609 |
+
"rewards/chosen": -3.197694778442383,
|
610 |
+
"rewards/margins": 1.9286365509033203,
|
611 |
+
"rewards/rejected": -5.126331329345703,
|
612 |
"step": 360
|
613 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
614 |
{
|
615 |
"epoch": 0.97,
|
616 |
+
"grad_norm": 56.54843526627352,
|
617 |
"learning_rate": 8.588892925590063e-10,
|
618 |
+
"logits/chosen": -7.414445400238037,
|
619 |
+
"logits/rejected": -8.447690963745117,
|
620 |
+
"logps/chosen": -499.84197998046875,
|
621 |
+
"logps/rejected": -670.2615966796875,
|
622 |
+
"loss": 0.3342,
|
623 |
+
"rewards/accuracies": 0.8500000238418579,
|
624 |
+
"rewards/chosen": -2.8063206672668457,
|
625 |
+
"rewards/margins": 2.2892918586730957,
|
626 |
+
"rewards/rejected": -5.0956130027771,
|
627 |
"step": 370
|
628 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
629 |
{
|
630 |
"epoch": 1.0,
|
631 |
"step": 379,
|
632 |
"total_flos": 0.0,
|
633 |
+
"train_loss": 0.39733357479830217,
|
634 |
+
"train_runtime": 5845.1464,
|
635 |
+
"train_samples_per_second": 8.303,
|
636 |
"train_steps_per_second": 0.065
|
637 |
}
|
638 |
],
|
639 |
+
"logging_steps": 10,
|
640 |
"max_steps": 379,
|
641 |
"num_input_tokens_seen": 0,
|
642 |
"num_train_epochs": 1,
|
training_args.bin
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 6264
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:9c6ef97f0ae5fdc0c432ac82567593d35a0af306147fdde011f9f421bccdc5eb
|
3 |
size 6264
|