Model save
Browse files- README.md +23 -25
- all_results.json +4 -4
- train_results.json +4 -4
- trainer_state.json +963 -963
README.md
CHANGED
|
@@ -3,33 +3,31 @@ library_name: transformers
|
|
| 3 |
license: apache-2.0
|
| 4 |
base_model: alignment-handbook/zephyr-7b-sft-full
|
| 5 |
tags:
|
| 6 |
-
- alignment-handbook
|
| 7 |
- trl
|
| 8 |
- dpo
|
|
|
|
| 9 |
- generated_from_trainer
|
| 10 |
-
datasets:
|
| 11 |
-
- HuggingFaceH4/ultrafeedback_binarized
|
| 12 |
model-index:
|
| 13 |
-
- name:
|
| 14 |
results: []
|
| 15 |
---
|
| 16 |
|
| 17 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 18 |
should probably proofread and complete it, then remove this comment. -->
|
| 19 |
|
| 20 |
-
#
|
| 21 |
|
| 22 |
-
This model is a fine-tuned version of [alignment-handbook/zephyr-7b-sft-full](https://huggingface.co/alignment-handbook/zephyr-7b-sft-full) on the
|
| 23 |
It achieves the following results on the evaluation set:
|
| 24 |
-
-
|
| 25 |
-
-
|
| 26 |
-
-
|
| 27 |
-
-
|
| 28 |
-
-
|
| 29 |
-
-
|
| 30 |
-
-
|
| 31 |
-
-
|
| 32 |
-
-
|
| 33 |
|
| 34 |
## Model description
|
| 35 |
|
|
@@ -64,16 +62,16 @@ The following hyperparameters were used during training:
|
|
| 64 |
|
| 65 |
### Training results
|
| 66 |
|
| 67 |
-
| Training Loss | Epoch | Step |
|
| 68 |
-
|
| 69 |
-
| 0.
|
| 70 |
-
| 0.
|
| 71 |
-
| 0.
|
| 72 |
-
| 0.
|
| 73 |
-
| 0.
|
| 74 |
-
| 0.
|
| 75 |
-
| 0.
|
| 76 |
-
| 0.
|
| 77 |
|
| 78 |
|
| 79 |
### Framework versions
|
|
|
|
| 3 |
license: apache-2.0
|
| 4 |
base_model: alignment-handbook/zephyr-7b-sft-full
|
| 5 |
tags:
|
|
|
|
| 6 |
- trl
|
| 7 |
- dpo
|
| 8 |
+
- alignment-handbook
|
| 9 |
- generated_from_trainer
|
|
|
|
|
|
|
| 10 |
model-index:
|
| 11 |
+
- name: zephyr-7b-dpo-full
|
| 12 |
results: []
|
| 13 |
---
|
| 14 |
|
| 15 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 16 |
should probably proofread and complete it, then remove this comment. -->
|
| 17 |
|
| 18 |
+
# zephyr-7b-dpo-full
|
| 19 |
|
| 20 |
+
This model is a fine-tuned version of [alignment-handbook/zephyr-7b-sft-full](https://huggingface.co/alignment-handbook/zephyr-7b-sft-full) on the None dataset.
|
| 21 |
It achieves the following results on the evaluation set:
|
| 22 |
+
- Loss: 0.3161
|
| 23 |
+
- Rewards/chosen: -4.1174
|
| 24 |
+
- Rewards/rejected: -8.3387
|
| 25 |
+
- Rewards/accuracies: 0.8212
|
| 26 |
+
- Rewards/margins: 4.2213
|
| 27 |
+
- Logps/rejected: -1123.9625
|
| 28 |
+
- Logps/chosen: -702.7068
|
| 29 |
+
- Logits/rejected: 0.5558
|
| 30 |
+
- Logits/chosen: -0.4246
|
| 31 |
|
| 32 |
## Model description
|
| 33 |
|
|
|
|
| 62 |
|
| 63 |
### Training results
|
| 64 |
|
| 65 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 66 |
+
|:-------------:|:------:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 67 |
+
| 0.5381 | 0.1152 | 100 | 0.4758 | -1.9882 | -2.9171 | 0.7270 | 0.9288 | -581.7981 | -489.7893 | -2.8822 | -2.9045 |
|
| 68 |
+
| 0.4268 | 0.2303 | 200 | 0.3577 | -3.9068 | -6.8487 | 0.7976 | 2.9419 | -974.9606 | -681.6494 | -0.6781 | -0.9791 |
|
| 69 |
+
| 0.4067 | 0.3455 | 300 | 0.3411 | -3.9757 | -7.6481 | 0.8094 | 3.6724 | -1054.9027 | -688.5351 | -0.6642 | -1.2474 |
|
| 70 |
+
| 0.4011 | 0.4607 | 400 | 0.3295 | -4.4449 | -8.4011 | 0.8156 | 3.9562 | -1130.1991 | -735.4550 | 0.1183 | -0.7429 |
|
| 71 |
+
| 0.3727 | 0.5759 | 500 | 0.3260 | -3.7203 | -7.6540 | 0.8161 | 3.9337 | -1055.4913 | -662.9987 | -0.4066 | -1.3009 |
|
| 72 |
+
| 0.3933 | 0.6910 | 600 | 0.3190 | -3.7331 | -7.5182 | 0.8257 | 3.7851 | -1041.9088 | -664.2776 | 0.3247 | -0.5819 |
|
| 73 |
+
| 0.3858 | 0.8062 | 700 | 0.3166 | -3.9569 | -8.0356 | 0.8246 | 4.0787 | -1093.6547 | -686.6614 | 0.3586 | -0.6058 |
|
| 74 |
+
| 0.3785 | 0.9214 | 800 | 0.3161 | -4.1174 | -8.3387 | 0.8212 | 4.2213 | -1123.9625 | -702.7068 | 0.5558 | -0.4246 |
|
| 75 |
|
| 76 |
|
| 77 |
### Framework versions
|
all_results.json
CHANGED
|
@@ -14,9 +14,9 @@
|
|
| 14 |
"eval_samples_per_second": 11.048,
|
| 15 |
"eval_steps_per_second": 0.346,
|
| 16 |
"total_flos": 0.0,
|
| 17 |
-
"train_loss": 0.
|
| 18 |
-
"train_runtime":
|
| 19 |
"train_samples": 111134,
|
| 20 |
-
"train_samples_per_second":
|
| 21 |
-
"train_steps_per_second":
|
| 22 |
}
|
|
|
|
| 14 |
"eval_samples_per_second": 11.048,
|
| 15 |
"eval_steps_per_second": 0.346,
|
| 16 |
"total_flos": 0.0,
|
| 17 |
+
"train_loss": 0.42924998652550483,
|
| 18 |
+
"train_runtime": 32201.1967,
|
| 19 |
"train_samples": 111134,
|
| 20 |
+
"train_samples_per_second": 3.451,
|
| 21 |
+
"train_steps_per_second": 0.027
|
| 22 |
}
|
train_results.json
CHANGED
|
@@ -1,9 +1,9 @@
|
|
| 1 |
{
|
| 2 |
"epoch": 0.9997120644975526,
|
| 3 |
"total_flos": 0.0,
|
| 4 |
-
"train_loss": 0.
|
| 5 |
-
"train_runtime":
|
| 6 |
"train_samples": 111134,
|
| 7 |
-
"train_samples_per_second":
|
| 8 |
-
"train_steps_per_second":
|
| 9 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"epoch": 0.9997120644975526,
|
| 3 |
"total_flos": 0.0,
|
| 4 |
+
"train_loss": 0.42924998652550483,
|
| 5 |
+
"train_runtime": 32201.1967,
|
| 6 |
"train_samples": 111134,
|
| 7 |
+
"train_samples_per_second": 3.451,
|
| 8 |
+
"train_steps_per_second": 0.027
|
| 9 |
}
|
trainer_state.json
CHANGED
|
@@ -9,13 +9,13 @@
|
|
| 9 |
"is_world_process_zero": true,
|
| 10 |
"log_history": [
|
| 11 |
{
|
| 12 |
-
"epoch":
|
| 13 |
-
"grad_norm":
|
| 14 |
-
"learning_rate":
|
| 15 |
-
"logits/chosen": -2.
|
| 16 |
-
"logits/rejected": -2.
|
| 17 |
-
"logps/chosen": -
|
| 18 |
-
"logps/rejected": -
|
| 19 |
"loss": 0.6931,
|
| 20 |
"rewards/accuracies": 0.0,
|
| 21 |
"rewards/chosen": 0.0,
|
|
@@ -25,1430 +25,1430 @@
|
|
| 25 |
},
|
| 26 |
{
|
| 27 |
"epoch": 0.01151742009789807,
|
| 28 |
-
"grad_norm": 8.
|
| 29 |
"learning_rate": 5.747126436781609e-08,
|
| 30 |
-
"logits/chosen": -2.
|
| 31 |
-
"logits/rejected": -2.
|
| 32 |
-
"logps/chosen": -
|
| 33 |
-
"logps/rejected": -
|
| 34 |
-
"loss": 0.
|
| 35 |
-
"rewards/accuracies": 0.
|
| 36 |
-
"rewards/chosen":
|
| 37 |
-
"rewards/margins": -0.
|
| 38 |
-
"rewards/rejected":
|
| 39 |
"step": 10
|
| 40 |
},
|
| 41 |
{
|
| 42 |
"epoch": 0.02303484019579614,
|
| 43 |
-
"grad_norm": 8.
|
| 44 |
"learning_rate": 1.1494252873563217e-07,
|
| 45 |
-
"logits/chosen": -2.
|
| 46 |
-
"logits/rejected": -2.
|
| 47 |
-
"logps/chosen": -
|
| 48 |
-
"logps/rejected": -
|
| 49 |
-
"loss": 0.
|
| 50 |
"rewards/accuracies": 0.5406249761581421,
|
| 51 |
-
"rewards/chosen": 0.
|
| 52 |
-
"rewards/margins": 0.
|
| 53 |
-
"rewards/rejected": -0.
|
| 54 |
"step": 20
|
| 55 |
},
|
| 56 |
{
|
| 57 |
"epoch": 0.03455226029369421,
|
| 58 |
-
"grad_norm": 8.
|
| 59 |
"learning_rate": 1.7241379310344828e-07,
|
| 60 |
-
"logits/chosen": -2.
|
| 61 |
-
"logits/rejected": -2.
|
| 62 |
-
"logps/chosen": -
|
| 63 |
-
"logps/rejected": -
|
| 64 |
-
"loss": 0.
|
| 65 |
-
"rewards/accuracies": 0.
|
| 66 |
-
"rewards/chosen": 0.
|
| 67 |
-
"rewards/margins": 0.
|
| 68 |
-
"rewards/rejected": -0.
|
| 69 |
"step": 30
|
| 70 |
},
|
| 71 |
{
|
| 72 |
"epoch": 0.04606968039159228,
|
| 73 |
-
"grad_norm": 8.
|
| 74 |
"learning_rate": 2.2988505747126435e-07,
|
| 75 |
-
"logits/chosen": -2.
|
| 76 |
-
"logits/rejected": -2.
|
| 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": 40
|
| 85 |
},
|
| 86 |
{
|
| 87 |
"epoch": 0.05758710048949035,
|
| 88 |
-
"grad_norm":
|
| 89 |
"learning_rate": 2.873563218390804e-07,
|
| 90 |
-
"logits/chosen": -2.
|
| 91 |
-
"logits/rejected": -2.
|
| 92 |
-
"logps/chosen": -
|
| 93 |
-
"logps/rejected": -
|
| 94 |
-
"loss": 0.
|
| 95 |
-
"rewards/accuracies": 0.
|
| 96 |
-
"rewards/chosen": 0.
|
| 97 |
-
"rewards/margins": 0.
|
| 98 |
-
"rewards/rejected": -0.
|
| 99 |
"step": 50
|
| 100 |
},
|
| 101 |
{
|
| 102 |
"epoch": 0.06910452058738842,
|
| 103 |
-
"grad_norm": 8.
|
| 104 |
"learning_rate": 3.4482758620689656e-07,
|
| 105 |
-
"logits/chosen": -2.
|
| 106 |
-
"logits/rejected": -2.
|
| 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": 60
|
| 115 |
},
|
| 116 |
{
|
| 117 |
"epoch": 0.0806219406852865,
|
| 118 |
-
"grad_norm":
|
| 119 |
"learning_rate": 4.0229885057471266e-07,
|
| 120 |
-
"logits/chosen": -2.
|
| 121 |
-
"logits/rejected": -2.
|
| 122 |
-
"logps/chosen": -
|
| 123 |
-
"logps/rejected": -
|
| 124 |
-
"loss": 0.
|
| 125 |
-
"rewards/accuracies": 0.
|
| 126 |
-
"rewards/chosen": -0.
|
| 127 |
-
"rewards/margins": 0.
|
| 128 |
-
"rewards/rejected": -0.
|
| 129 |
"step": 70
|
| 130 |
},
|
| 131 |
{
|
| 132 |
"epoch": 0.09213936078318456,
|
| 133 |
-
"grad_norm": 11.
|
| 134 |
"learning_rate": 4.597701149425287e-07,
|
| 135 |
-
"logits/chosen": -2.
|
| 136 |
-
"logits/rejected": -2.
|
| 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": 80
|
| 145 |
},
|
| 146 |
{
|
| 147 |
"epoch": 0.10365678088108264,
|
| 148 |
-
"grad_norm":
|
| 149 |
"learning_rate": 4.999817969178237e-07,
|
| 150 |
-
"logits/chosen": -2.
|
| 151 |
-
"logits/rejected": -2.
|
| 152 |
-
"logps/chosen": -
|
| 153 |
-
"logps/rejected": -
|
| 154 |
-
"loss": 0.
|
| 155 |
-
"rewards/accuracies": 0.
|
| 156 |
-
"rewards/chosen": -0.
|
| 157 |
-
"rewards/margins": 0.
|
| 158 |
-
"rewards/rejected": -1.
|
| 159 |
"step": 90
|
| 160 |
},
|
| 161 |
{
|
| 162 |
"epoch": 0.1151742009789807,
|
| 163 |
-
"grad_norm":
|
| 164 |
"learning_rate": 4.996582603056428e-07,
|
| 165 |
-
"logits/chosen": -2.
|
| 166 |
-
"logits/rejected": -2.
|
| 167 |
-
"logps/chosen": -
|
| 168 |
-
"logps/rejected": -
|
| 169 |
-
"loss": 0.
|
| 170 |
-
"rewards/accuracies": 0.
|
| 171 |
-
"rewards/chosen": -1.
|
| 172 |
-
"rewards/margins": 0.
|
| 173 |
-
"rewards/rejected": -
|
| 174 |
"step": 100
|
| 175 |
},
|
| 176 |
{
|
| 177 |
"epoch": 0.1151742009789807,
|
| 178 |
-
"eval_logits/chosen": -2.
|
| 179 |
-
"eval_logits/rejected": -2.
|
| 180 |
-
"eval_logps/chosen": -
|
| 181 |
-
"eval_logps/rejected": -
|
| 182 |
-
"eval_loss": 0.
|
| 183 |
-
"eval_rewards/accuracies": 0.
|
| 184 |
-
"eval_rewards/chosen": -1.
|
| 185 |
-
"eval_rewards/margins": 0.
|
| 186 |
-
"eval_rewards/rejected": -2.
|
| 187 |
-
"eval_runtime":
|
| 188 |
-
"eval_samples_per_second":
|
| 189 |
-
"eval_steps_per_second": 0.
|
| 190 |
"step": 100
|
| 191 |
},
|
| 192 |
{
|
| 193 |
"epoch": 0.12669162107687879,
|
| 194 |
-
"grad_norm":
|
| 195 |
"learning_rate": 4.989308132738126e-07,
|
| 196 |
-
"logits/chosen": -2.
|
| 197 |
-
"logits/rejected": -2.
|
| 198 |
-
"logps/chosen": -
|
| 199 |
-
"logps/rejected": -
|
| 200 |
-
"loss": 0.
|
| 201 |
-
"rewards/accuracies": 0.
|
| 202 |
-
"rewards/chosen": -1.
|
| 203 |
-
"rewards/margins": 0.
|
| 204 |
-
"rewards/rejected": -
|
| 205 |
"step": 110
|
| 206 |
},
|
| 207 |
{
|
| 208 |
"epoch": 0.13820904117477684,
|
| 209 |
-
"grad_norm":
|
| 210 |
"learning_rate": 4.978006327248536e-07,
|
| 211 |
-
"logits/chosen": -2.
|
| 212 |
-
"logits/rejected": -2.
|
| 213 |
-
"logps/chosen": -
|
| 214 |
-
"logps/rejected": -
|
| 215 |
-
"loss": 0.
|
| 216 |
-
"rewards/accuracies": 0.
|
| 217 |
-
"rewards/chosen": -1.
|
| 218 |
-
"rewards/margins":
|
| 219 |
-
"rewards/rejected": -2.
|
| 220 |
"step": 120
|
| 221 |
},
|
| 222 |
{
|
| 223 |
"epoch": 0.14972646127267492,
|
| 224 |
-
"grad_norm":
|
| 225 |
"learning_rate": 4.962695471250032e-07,
|
| 226 |
-
"logits/chosen": -2.
|
| 227 |
-
"logits/rejected": -2.
|
| 228 |
-
"logps/chosen": -
|
| 229 |
-
"logps/rejected": -
|
| 230 |
-
"loss": 0.
|
| 231 |
-
"rewards/accuracies": 0.
|
| 232 |
-
"rewards/chosen": -
|
| 233 |
-
"rewards/margins": 1.
|
| 234 |
-
"rewards/rejected": -
|
| 235 |
"step": 130
|
| 236 |
},
|
| 237 |
{
|
| 238 |
"epoch": 0.161243881370573,
|
| 239 |
-
"grad_norm":
|
| 240 |
"learning_rate": 4.94340033546025e-07,
|
| 241 |
-
"logits/chosen": -1.
|
| 242 |
-
"logits/rejected": -1.
|
| 243 |
-
"logps/chosen": -
|
| 244 |
-
"logps/rejected": -
|
| 245 |
-
"loss": 0.
|
| 246 |
-
"rewards/accuracies": 0.
|
| 247 |
-
"rewards/chosen": -2.
|
| 248 |
-
"rewards/margins": 1.
|
| 249 |
-
"rewards/rejected": -3.
|
| 250 |
"step": 140
|
| 251 |
},
|
| 252 |
{
|
| 253 |
"epoch": 0.17276130146847107,
|
| 254 |
-
"grad_norm":
|
| 255 |
"learning_rate": 4.920152136576705e-07,
|
| 256 |
-
"logits/chosen": -1.
|
| 257 |
-
"logits/rejected": -1.
|
| 258 |
-
"logps/chosen": -
|
| 259 |
-
"logps/rejected": -
|
| 260 |
-
"loss": 0.
|
| 261 |
-
"rewards/accuracies": 0.
|
| 262 |
-
"rewards/chosen": -
|
| 263 |
-
"rewards/margins": 1.
|
| 264 |
-
"rewards/rejected": -
|
| 265 |
"step": 150
|
| 266 |
},
|
| 267 |
{
|
| 268 |
"epoch": 0.18427872156636912,
|
| 269 |
-
"grad_norm":
|
| 270 |
"learning_rate": 4.892988486772756e-07,
|
| 271 |
-
"logits/chosen": -1.
|
| 272 |
-
"logits/rejected": -1.
|
| 273 |
-
"logps/chosen": -
|
| 274 |
-
"logps/rejected": -
|
| 275 |
-
"loss": 0.
|
| 276 |
-
"rewards/accuracies": 0.
|
| 277 |
-
"rewards/chosen": -
|
| 278 |
-
"rewards/margins": 1.
|
| 279 |
-
"rewards/rejected": -3.
|
| 280 |
"step": 160
|
| 281 |
},
|
| 282 |
{
|
| 283 |
"epoch": 0.1957961416642672,
|
| 284 |
-
"grad_norm":
|
| 285 |
"learning_rate": 4.861953332846629e-07,
|
| 286 |
-
"logits/chosen": -1.
|
| 287 |
-
"logits/rejected": -1.
|
| 288 |
-
"logps/chosen": -
|
| 289 |
-
"logps/rejected": -
|
| 290 |
-
"loss": 0.
|
| 291 |
-
"rewards/accuracies": 0.
|
| 292 |
-
"rewards/chosen": -2.
|
| 293 |
-
"rewards/margins": 1.
|
| 294 |
-
"rewards/rejected": -3.
|
| 295 |
"step": 170
|
| 296 |
},
|
| 297 |
{
|
| 298 |
"epoch": 0.20731356176216528,
|
| 299 |
-
"grad_norm":
|
| 300 |
"learning_rate": 4.827096885121953e-07,
|
| 301 |
-
"logits/chosen": -
|
| 302 |
-
"logits/rejected": -0.
|
| 303 |
-
"logps/chosen": -
|
| 304 |
-
"logps/rejected": -
|
| 305 |
-
"loss": 0.
|
| 306 |
-
"rewards/accuracies": 0.
|
| 307 |
-
"rewards/chosen": -
|
| 308 |
-
"rewards/margins": 1.
|
| 309 |
-
"rewards/rejected": -
|
| 310 |
"step": 180
|
| 311 |
},
|
| 312 |
{
|
| 313 |
"epoch": 0.21883098186006333,
|
| 314 |
-
"grad_norm":
|
| 315 |
"learning_rate": 4.788475536214821e-07,
|
| 316 |
-
"logits/chosen": -
|
| 317 |
-
"logits/rejected": -0.
|
| 318 |
-
"logps/chosen": -
|
| 319 |
-
"logps/rejected": -
|
| 320 |
-
"loss": 0.
|
| 321 |
-
"rewards/accuracies": 0.
|
| 322 |
-
"rewards/chosen": -2.
|
| 323 |
-
"rewards/margins": 1.
|
| 324 |
-
"rewards/rejected": -
|
| 325 |
"step": 190
|
| 326 |
},
|
| 327 |
{
|
| 328 |
"epoch": 0.2303484019579614,
|
| 329 |
-
"grad_norm":
|
| 330 |
"learning_rate": 4.746151769798818e-07,
|
| 331 |
-
"logits/chosen": -0.
|
| 332 |
-
"logits/rejected":
|
| 333 |
-
"logps/chosen": -
|
| 334 |
-
"logps/rejected": -
|
| 335 |
-
"loss": 0.
|
| 336 |
-
"rewards/accuracies": 0.
|
| 337 |
-
"rewards/chosen": -
|
| 338 |
-
"rewards/margins":
|
| 339 |
-
"rewards/rejected": -
|
| 340 |
"step": 200
|
| 341 |
},
|
| 342 |
{
|
| 343 |
"epoch": 0.2303484019579614,
|
| 344 |
-
"eval_logits/chosen": -
|
| 345 |
-
"eval_logits/rejected": -
|
| 346 |
-
"eval_logps/chosen": -
|
| 347 |
-
"eval_logps/rejected": -
|
| 348 |
-
"eval_loss": 0.
|
| 349 |
-
"eval_rewards/accuracies": 0.
|
| 350 |
-
"eval_rewards/chosen": -3.
|
| 351 |
-
"eval_rewards/margins": 2.
|
| 352 |
-
"eval_rewards/rejected": -
|
| 353 |
-
"eval_runtime":
|
| 354 |
-
"eval_samples_per_second":
|
| 355 |
-
"eval_steps_per_second": 0.
|
| 356 |
"step": 200
|
| 357 |
},
|
| 358 |
{
|
| 359 |
"epoch": 0.2418658220558595,
|
| 360 |
-
"grad_norm":
|
| 361 |
"learning_rate": 4.7001940595156055e-07,
|
| 362 |
-
"logits/chosen": -0.
|
| 363 |
-
"logits/rejected":
|
| 364 |
-
"logps/chosen": -
|
| 365 |
-
"logps/rejected": -
|
| 366 |
-
"loss": 0.
|
| 367 |
-
"rewards/accuracies": 0.
|
| 368 |
-
"rewards/chosen": -2.
|
| 369 |
-
"rewards/margins": 1.
|
| 370 |
-
"rewards/rejected": -4.
|
| 371 |
"step": 210
|
| 372 |
},
|
| 373 |
{
|
| 374 |
"epoch": 0.25338324215375757,
|
| 375 |
-
"grad_norm":
|
| 376 |
"learning_rate": 4.650676758194623e-07,
|
| 377 |
-
"logits/chosen": -0.
|
| 378 |
-
"logits/rejected": -0.
|
| 379 |
-
"logps/chosen": -
|
| 380 |
-
"logps/rejected": -
|
| 381 |
-
"loss": 0.
|
| 382 |
-
"rewards/accuracies": 0.
|
| 383 |
-
"rewards/chosen": -
|
| 384 |
-
"rewards/margins": 2.
|
| 385 |
-
"rewards/rejected": -
|
| 386 |
"step": 220
|
| 387 |
},
|
| 388 |
{
|
| 389 |
"epoch": 0.26490066225165565,
|
| 390 |
-
"grad_norm":
|
| 391 |
"learning_rate": 4.5976799775611215e-07,
|
| 392 |
-
"logits/chosen": -0.
|
| 393 |
-
"logits/rejected": 0.
|
| 394 |
-
"logps/chosen": -
|
| 395 |
-
"logps/rejected": -
|
| 396 |
-
"loss": 0.
|
| 397 |
-
"rewards/accuracies": 0.
|
| 398 |
-
"rewards/chosen": -2.
|
| 399 |
-
"rewards/margins":
|
| 400 |
-
"rewards/rejected": -
|
| 401 |
"step": 230
|
| 402 |
},
|
| 403 |
{
|
| 404 |
"epoch": 0.2764180823495537,
|
| 405 |
-
"grad_norm": 25.
|
| 406 |
"learning_rate": 4.5412894586271543e-07,
|
| 407 |
-
"logits/chosen": -0.
|
| 408 |
-
"logits/rejected": 0.
|
| 409 |
-
"logps/chosen": -
|
| 410 |
-
"logps/rejected": -
|
| 411 |
-
"loss": 0.
|
| 412 |
-
"rewards/accuracies": 0.
|
| 413 |
-
"rewards/chosen": -2.
|
| 414 |
-
"rewards/margins": 1.
|
| 415 |
-
"rewards/rejected": -4.
|
| 416 |
"step": 240
|
| 417 |
},
|
| 418 |
{
|
| 419 |
"epoch": 0.28793550244745175,
|
| 420 |
-
"grad_norm":
|
| 421 |
"learning_rate": 4.481596432975201e-07,
|
| 422 |
-
"logits/chosen": -
|
| 423 |
-
"logits/rejected": -0.
|
| 424 |
-
"logps/chosen": -
|
| 425 |
-
"logps/rejected": -
|
| 426 |
-
"loss": 0.
|
| 427 |
-
"rewards/accuracies": 0.
|
| 428 |
-
"rewards/chosen": -
|
| 429 |
-
"rewards/margins": 2.
|
| 430 |
-
"rewards/rejected": -
|
| 431 |
"step": 250
|
| 432 |
},
|
| 433 |
{
|
| 434 |
"epoch": 0.29945292254534983,
|
| 435 |
-
"grad_norm":
|
| 436 |
"learning_rate": 4.41869747515886e-07,
|
| 437 |
-
"logits/chosen": -
|
| 438 |
-
"logits/rejected": 0.
|
| 439 |
-
"logps/chosen": -
|
| 440 |
-
"logps/rejected": -
|
| 441 |
-
"loss": 0.
|
| 442 |
-
"rewards/accuracies": 0.
|
| 443 |
-
"rewards/chosen": -
|
| 444 |
-
"rewards/margins": 2.
|
| 445 |
-
"rewards/rejected": -5.
|
| 446 |
"step": 260
|
| 447 |
},
|
| 448 |
{
|
| 449 |
"epoch": 0.3109703426432479,
|
| 450 |
-
"grad_norm":
|
| 451 |
"learning_rate": 4.352694346459396e-07,
|
| 452 |
-
"logits/chosen": -
|
| 453 |
-
"logits/rejected": 0.
|
| 454 |
-
"logps/chosen": -
|
| 455 |
-
"logps/rejected": -
|
| 456 |
-
"loss": 0.
|
| 457 |
"rewards/accuracies": 0.784375011920929,
|
| 458 |
-
"rewards/chosen": -
|
| 459 |
-
"rewards/margins": 2.
|
| 460 |
-
"rewards/rejected": -
|
| 461 |
"step": 270
|
| 462 |
},
|
| 463 |
{
|
| 464 |
"epoch": 0.322487762741146,
|
| 465 |
-
"grad_norm": 26.
|
| 466 |
"learning_rate": 4.2836938302509256e-07,
|
| 467 |
-
"logits/chosen": -
|
| 468 |
-
"logits/rejected": 0.
|
| 469 |
-
"logps/chosen": -
|
| 470 |
-
"logps/rejected": -
|
| 471 |
-
"loss": 0.
|
| 472 |
-
"rewards/accuracies": 0.
|
| 473 |
-
"rewards/chosen": -2.
|
| 474 |
-
"rewards/margins": 2.
|
| 475 |
-
"rewards/rejected": -5.
|
| 476 |
"step": 280
|
| 477 |
},
|
| 478 |
{
|
| 479 |
"epoch": 0.33400518283904407,
|
| 480 |
-
"grad_norm":
|
| 481 |
"learning_rate": 4.2118075592405874e-07,
|
| 482 |
-
"logits/chosen": -0.
|
| 483 |
-
"logits/rejected": 0.
|
| 484 |
-
"logps/chosen": -
|
| 485 |
-
"logps/rejected": -
|
| 486 |
-
"loss": 0.
|
| 487 |
-
"rewards/accuracies": 0.
|
| 488 |
-
"rewards/chosen": -2.
|
| 489 |
-
"rewards/margins": 2.
|
| 490 |
-
"rewards/rejected": -5.
|
| 491 |
"step": 290
|
| 492 |
},
|
| 493 |
{
|
| 494 |
"epoch": 0.34552260293694215,
|
| 495 |
-
"grad_norm":
|
| 496 |
"learning_rate": 4.137151834863213e-07,
|
| 497 |
-
"logits/chosen": -0.
|
| 498 |
-
"logits/rejected": 0.
|
| 499 |
-
"logps/chosen": -
|
| 500 |
-
"logps/rejected": -
|
| 501 |
-
"loss": 0.
|
| 502 |
-
"rewards/accuracies": 0.
|
| 503 |
-
"rewards/chosen": -3.
|
| 504 |
-
"rewards/margins": 2.
|
| 505 |
-
"rewards/rejected": -5.
|
| 506 |
"step": 300
|
| 507 |
},
|
| 508 |
{
|
| 509 |
"epoch": 0.34552260293694215,
|
| 510 |
-
"eval_logits/chosen": -
|
| 511 |
-
"eval_logits/rejected": -0.
|
| 512 |
-
"eval_logps/chosen": -
|
| 513 |
-
"eval_logps/rejected": -
|
| 514 |
-
"eval_loss": 0.
|
| 515 |
-
"eval_rewards/accuracies": 0.
|
| 516 |
-
"eval_rewards/chosen": -3.
|
| 517 |
-
"eval_rewards/margins": 3.
|
| 518 |
-
"eval_rewards/rejected": -7.
|
| 519 |
-
"eval_runtime":
|
| 520 |
-
"eval_samples_per_second":
|
| 521 |
-
"eval_steps_per_second": 0.
|
| 522 |
"step": 300
|
| 523 |
},
|
| 524 |
{
|
| 525 |
"epoch": 0.35704002303484017,
|
| 526 |
-
"grad_norm":
|
| 527 |
"learning_rate": 4.059847439122671e-07,
|
| 528 |
-
"logits/chosen": -0.
|
| 529 |
-
"logits/rejected": 0.
|
| 530 |
-
"logps/chosen": -
|
| 531 |
-
"logps/rejected": -
|
| 532 |
-
"loss": 0.
|
| 533 |
-
"rewards/accuracies": 0.
|
| 534 |
-
"rewards/chosen": -2.
|
| 535 |
-
"rewards/margins": 2.
|
| 536 |
-
"rewards/rejected": -
|
| 537 |
"step": 310
|
| 538 |
},
|
| 539 |
{
|
| 540 |
"epoch": 0.36855744313273825,
|
| 541 |
-
"grad_norm":
|
| 542 |
"learning_rate": 3.98001943918432e-07,
|
| 543 |
-
"logits/chosen": -
|
| 544 |
-
"logits/rejected": -0.
|
| 545 |
-
"logps/chosen": -
|
| 546 |
-
"logps/rejected": -
|
| 547 |
-
"loss": 0.
|
| 548 |
-
"rewards/accuracies": 0.
|
| 549 |
-
"rewards/chosen": -2.
|
| 550 |
-
"rewards/margins": 2.
|
| 551 |
-
"rewards/rejected": -4.
|
| 552 |
"step": 320
|
| 553 |
},
|
| 554 |
{
|
| 555 |
"epoch": 0.38007486323063633,
|
| 556 |
-
"grad_norm":
|
| 557 |
"learning_rate": 3.8977969850346866e-07,
|
| 558 |
-
"logits/chosen": 0.
|
| 559 |
-
"logits/rejected": 0.
|
| 560 |
-
"logps/chosen": -
|
| 561 |
-
"logps/rejected": -
|
| 562 |
-
"loss": 0.
|
| 563 |
-
"rewards/accuracies": 0.
|
| 564 |
-
"rewards/chosen": -
|
| 565 |
-
"rewards/margins":
|
| 566 |
-
"rewards/rejected": -
|
| 567 |
"step": 330
|
| 568 |
},
|
| 569 |
{
|
| 570 |
"epoch": 0.3915922833285344,
|
| 571 |
-
"grad_norm":
|
| 572 |
"learning_rate": 3.8133131005357465e-07,
|
| 573 |
-
"logits/chosen": -0.
|
| 574 |
-
"logits/rejected": 0.
|
| 575 |
-
"logps/chosen": -
|
| 576 |
-
"logps/rejected": -
|
| 577 |
-
"loss": 0.
|
| 578 |
-
"rewards/accuracies": 0.
|
| 579 |
-
"rewards/chosen": -3.
|
| 580 |
-
"rewards/margins":
|
| 581 |
-
"rewards/rejected": -
|
| 582 |
"step": 340
|
| 583 |
},
|
| 584 |
{
|
| 585 |
"epoch": 0.4031097034264325,
|
| 586 |
-
"grad_norm":
|
| 587 |
"learning_rate": 3.7267044682118435e-07,
|
| 588 |
-
"logits/chosen": -0.
|
| 589 |
-
"logits/rejected": 0.
|
| 590 |
-
"logps/chosen": -
|
| 591 |
-
"logps/rejected": -
|
| 592 |
-
"loss": 0.
|
| 593 |
-
"rewards/accuracies": 0.
|
| 594 |
-
"rewards/chosen": -
|
| 595 |
-
"rewards/margins": 2.
|
| 596 |
-
"rewards/rejected": -
|
| 597 |
"step": 350
|
| 598 |
},
|
| 599 |
{
|
| 600 |
"epoch": 0.41462712352433057,
|
| 601 |
-
"grad_norm":
|
| 602 |
"learning_rate": 3.638111208117425e-07,
|
| 603 |
-
"logits/chosen": -0.
|
| 604 |
-
"logits/rejected": 0.
|
| 605 |
-
"logps/chosen": -
|
| 606 |
-
"logps/rejected": -
|
| 607 |
-
"loss": 0.
|
| 608 |
-
"rewards/accuracies": 0.
|
| 609 |
-
"rewards/chosen": -3.
|
| 610 |
-
"rewards/margins":
|
| 611 |
-
"rewards/rejected": -
|
| 612 |
"step": 360
|
| 613 |
},
|
| 614 |
{
|
| 615 |
"epoch": 0.42614454362222864,
|
| 616 |
-
"grad_norm":
|
| 617 |
"learning_rate": 3.5476766511433605e-07,
|
| 618 |
-
"logits/chosen": -0.
|
| 619 |
-
"logits/rejected": 0.
|
| 620 |
-
"logps/chosen": -
|
| 621 |
-
"logps/rejected": -
|
| 622 |
-
"loss": 0.
|
| 623 |
-
"rewards/accuracies": 0.
|
| 624 |
-
"rewards/chosen": -3.
|
| 625 |
-
"rewards/margins": 2.
|
| 626 |
-
"rewards/rejected": -5.
|
| 627 |
"step": 370
|
| 628 |
},
|
| 629 |
{
|
| 630 |
"epoch": 0.43766196372012667,
|
| 631 |
-
"grad_norm":
|
| 632 |
"learning_rate": 3.455547107128602e-07,
|
| 633 |
-
"logits/chosen": -0.
|
| 634 |
-
"logits/rejected": 0.
|
| 635 |
-
"logps/chosen": -
|
| 636 |
-
"logps/rejected": -
|
| 637 |
-
"loss": 0.
|
| 638 |
-
"rewards/accuracies": 0.
|
| 639 |
-
"rewards/chosen": -
|
| 640 |
-
"rewards/margins": 2.
|
| 641 |
-
"rewards/rejected": -5.
|
| 642 |
"step": 380
|
| 643 |
},
|
| 644 |
{
|
| 645 |
"epoch": 0.44917938381802475,
|
| 646 |
-
"grad_norm":
|
| 647 |
"learning_rate": 3.361871628152338e-07,
|
| 648 |
-
"logits/chosen":
|
| 649 |
-
"logits/rejected": 0.
|
| 650 |
-
"logps/chosen": -
|
| 651 |
-
"logps/rejected": -
|
| 652 |
-
"loss": 0.
|
| 653 |
-
"rewards/accuracies": 0.
|
| 654 |
-
"rewards/chosen": -
|
| 655 |
-
"rewards/margins":
|
| 656 |
-
"rewards/rejected": -
|
| 657 |
"step": 390
|
| 658 |
},
|
| 659 |
{
|
| 660 |
"epoch": 0.4606968039159228,
|
| 661 |
-
"grad_norm":
|
| 662 |
"learning_rate": 3.2668017673896077e-07,
|
| 663 |
-
"logits/chosen": -0.
|
| 664 |
-
"logits/rejected": 0.
|
| 665 |
-
"logps/chosen": -
|
| 666 |
-
"logps/rejected": -
|
| 667 |
-
"loss": 0.
|
| 668 |
-
"rewards/accuracies": 0.
|
| 669 |
-
"rewards/chosen": -3.
|
| 670 |
-
"rewards/margins":
|
| 671 |
-
"rewards/rejected": -
|
| 672 |
"step": 400
|
| 673 |
},
|
| 674 |
{
|
| 675 |
"epoch": 0.4606968039159228,
|
| 676 |
-
"eval_logits/chosen": -0.
|
| 677 |
-
"eval_logits/rejected": 0.
|
| 678 |
-
"eval_logps/chosen": -
|
| 679 |
-
"eval_logps/rejected": -
|
| 680 |
-
"eval_loss": 0.
|
| 681 |
-
"eval_rewards/accuracies": 0.
|
| 682 |
-
"eval_rewards/chosen": -4.
|
| 683 |
-
"eval_rewards/margins": 3.
|
| 684 |
-
"eval_rewards/rejected": -8.
|
| 685 |
-
"eval_runtime":
|
| 686 |
-
"eval_samples_per_second":
|
| 687 |
-
"eval_steps_per_second": 0.
|
| 688 |
"step": 400
|
| 689 |
},
|
| 690 |
{
|
| 691 |
"epoch": 0.4722142240138209,
|
| 692 |
-
"grad_norm":
|
| 693 |
"learning_rate": 3.1704913339205103e-07,
|
| 694 |
-
"logits/chosen": 0.
|
| 695 |
-
"logits/rejected":
|
| 696 |
-
"logps/chosen": -
|
| 697 |
-
"logps/rejected": -
|
| 698 |
-
"loss": 0.
|
| 699 |
-
"rewards/accuracies": 0.
|
| 700 |
-
"rewards/chosen": -
|
| 701 |
-
"rewards/margins":
|
| 702 |
-
"rewards/rejected": -5.
|
| 703 |
"step": 410
|
| 704 |
},
|
| 705 |
{
|
| 706 |
"epoch": 0.483731644111719,
|
| 707 |
-
"grad_norm":
|
| 708 |
"learning_rate": 3.0730961438896885e-07,
|
| 709 |
-
"logits/chosen": -
|
| 710 |
-
"logits/rejected": 0.
|
| 711 |
-
"logps/chosen": -
|
| 712 |
-
"logps/rejected": -
|
| 713 |
-
"loss": 0.
|
| 714 |
-
"rewards/accuracies": 0.
|
| 715 |
-
"rewards/chosen": -
|
| 716 |
-
"rewards/margins": 2.
|
| 717 |
-
"rewards/rejected": -
|
| 718 |
"step": 420
|
| 719 |
},
|
| 720 |
{
|
| 721 |
"epoch": 0.49524906420961706,
|
| 722 |
-
"grad_norm":
|
| 723 |
"learning_rate": 2.9747737684186795e-07,
|
| 724 |
-
"logits/chosen": -
|
| 725 |
-
"logits/rejected": 0.
|
| 726 |
-
"logps/chosen": -
|
| 727 |
-
"logps/rejected": -
|
| 728 |
-
"loss": 0.
|
| 729 |
-
"rewards/accuracies": 0.
|
| 730 |
-
"rewards/chosen": -
|
| 731 |
-
"rewards/margins": 2.
|
| 732 |
-
"rewards/rejected": -
|
| 733 |
"step": 430
|
| 734 |
},
|
| 735 |
{
|
| 736 |
"epoch": 0.5067664843075151,
|
| 737 |
-
"grad_norm":
|
| 738 |
"learning_rate": 2.8756832786789663e-07,
|
| 739 |
-
"logits/chosen": -
|
| 740 |
-
"logits/rejected": 0.
|
| 741 |
-
"logps/chosen": -
|
| 742 |
-
"logps/rejected": -
|
| 743 |
-
"loss": 0.
|
| 744 |
-
"rewards/accuracies": 0.
|
| 745 |
-
"rewards/chosen": -2.
|
| 746 |
-
"rewards/margins":
|
| 747 |
-
"rewards/rejected": -5.
|
| 748 |
"step": 440
|
| 749 |
},
|
| 750 |
{
|
| 751 |
"epoch": 0.5182839044054132,
|
| 752 |
-
"grad_norm":
|
| 753 |
"learning_rate": 2.7759849885381747e-07,
|
| 754 |
-
"logits/chosen": -
|
| 755 |
-
"logits/rejected": 0.
|
| 756 |
-
"logps/chosen": -
|
| 757 |
-
"logps/rejected": -
|
| 758 |
-
"loss": 0.
|
| 759 |
-
"rewards/accuracies": 0.
|
| 760 |
-
"rewards/chosen": -2.
|
| 761 |
-
"rewards/margins": 2.
|
| 762 |
-
"rewards/rejected": -
|
| 763 |
"step": 450
|
| 764 |
},
|
| 765 |
{
|
| 766 |
"epoch": 0.5298013245033113,
|
| 767 |
-
"grad_norm":
|
| 768 |
"learning_rate": 2.675840195195762e-07,
|
| 769 |
-
"logits/chosen": -
|
| 770 |
-
"logits/rejected": 0.
|
| 771 |
-
"logps/chosen": -
|
| 772 |
-
"logps/rejected": -
|
| 773 |
-
"loss": 0.
|
| 774 |
-
"rewards/accuracies": 0.
|
| 775 |
-
"rewards/chosen": -2.
|
| 776 |
-
"rewards/margins": 2.
|
| 777 |
-
"rewards/rejected": -
|
| 778 |
"step": 460
|
| 779 |
},
|
| 780 |
{
|
| 781 |
"epoch": 0.5413187446012093,
|
| 782 |
-
"grad_norm":
|
| 783 |
"learning_rate": 2.575410918227829e-07,
|
| 784 |
-
"logits/chosen": -0.
|
| 785 |
-
"logits/rejected": 0.
|
| 786 |
-
"logps/chosen": -
|
| 787 |
-
"logps/rejected": -
|
| 788 |
-
"loss": 0.
|
| 789 |
-
"rewards/accuracies": 0.
|
| 790 |
-
"rewards/chosen": -2.
|
| 791 |
-
"rewards/margins": 2.
|
| 792 |
-
"rewards/rejected": -5.
|
| 793 |
"step": 470
|
| 794 |
},
|
| 795 |
{
|
| 796 |
"epoch": 0.5528361646991073,
|
| 797 |
-
"grad_norm":
|
| 798 |
"learning_rate": 2.474859637463226e-07,
|
| 799 |
-
"logits/chosen": 0.
|
| 800 |
-
"logits/rejected": 0.
|
| 801 |
-
"logps/chosen": -
|
| 802 |
-
"logps/rejected": -
|
| 803 |
-
"loss": 0.
|
| 804 |
-
"rewards/accuracies": 0.
|
| 805 |
-
"rewards/chosen": -
|
| 806 |
-
"rewards/margins": 2.
|
| 807 |
-
"rewards/rejected": -5.
|
| 808 |
"step": 480
|
| 809 |
},
|
| 810 |
{
|
| 811 |
"epoch": 0.5643535847970055,
|
| 812 |
-
"grad_norm":
|
| 813 |
"learning_rate": 2.3743490301150355e-07,
|
| 814 |
-
"logits/chosen": 0.
|
| 815 |
-
"logits/rejected": 0.
|
| 816 |
-
"logps/chosen": -
|
| 817 |
-
"logps/rejected": -
|
| 818 |
-
"loss": 0.
|
| 819 |
-
"rewards/accuracies": 0.
|
| 820 |
-
"rewards/chosen": -
|
| 821 |
-
"rewards/margins": 2.
|
| 822 |
-
"rewards/rejected": -5.
|
| 823 |
"step": 490
|
| 824 |
},
|
| 825 |
{
|
| 826 |
"epoch": 0.5758710048949035,
|
| 827 |
-
"grad_norm":
|
| 828 |
"learning_rate": 2.274041707592724e-07,
|
| 829 |
-
"logits/chosen": -
|
| 830 |
-
"logits/rejected": 0.
|
| 831 |
-
"logps/chosen": -
|
| 832 |
-
"logps/rejected": -
|
| 833 |
-
"loss": 0.
|
| 834 |
-
"rewards/accuracies": 0.
|
| 835 |
-
"rewards/chosen": -
|
| 836 |
-
"rewards/margins": 2.
|
| 837 |
-
"rewards/rejected": -
|
| 838 |
"step": 500
|
| 839 |
},
|
| 840 |
{
|
| 841 |
"epoch": 0.5758710048949035,
|
| 842 |
-
"eval_logits/chosen": -
|
| 843 |
-
"eval_logits/rejected": 0.
|
| 844 |
-
"eval_logps/chosen": -
|
| 845 |
-
"eval_logps/rejected": -
|
| 846 |
-
"eval_loss": 0.
|
| 847 |
-
"eval_rewards/accuracies": 0.
|
| 848 |
-
"eval_rewards/chosen": -
|
| 849 |
-
"eval_rewards/margins": 3.
|
| 850 |
-
"eval_rewards/rejected": -7.
|
| 851 |
-
"eval_runtime":
|
| 852 |
-
"eval_samples_per_second":
|
| 853 |
-
"eval_steps_per_second": 0.
|
| 854 |
"step": 500
|
| 855 |
},
|
| 856 |
{
|
| 857 |
"epoch": 0.5873884249928016,
|
| 858 |
-
"grad_norm":
|
| 859 |
"learning_rate": 2.17409995242075e-07,
|
| 860 |
-
"logits/chosen": -0.
|
| 861 |
-
"logits/rejected": 0.
|
| 862 |
-
"logps/chosen": -
|
| 863 |
-
"logps/rejected": -
|
| 864 |
-
"loss": 0.
|
| 865 |
-
"rewards/accuracies": 0.
|
| 866 |
-
"rewards/chosen": -
|
| 867 |
-
"rewards/margins":
|
| 868 |
-
"rewards/rejected": -5.
|
| 869 |
"step": 510
|
| 870 |
},
|
| 871 |
{
|
| 872 |
"epoch": 0.5989058450906997,
|
| 873 |
-
"grad_norm":
|
| 874 |
"learning_rate": 2.0746854556892544e-07,
|
| 875 |
-
"logits/chosen": -0.
|
| 876 |
-
"logits/rejected": 0.
|
| 877 |
-
"logps/chosen": -
|
| 878 |
-
"logps/rejected": -
|
| 879 |
-
"loss": 0.
|
| 880 |
"rewards/accuracies": 0.8187500238418579,
|
| 881 |
-
"rewards/chosen": -3.
|
| 882 |
-
"rewards/margins": 2.
|
| 883 |
-
"rewards/rejected": -
|
| 884 |
"step": 520
|
| 885 |
},
|
| 886 |
{
|
| 887 |
"epoch": 0.6104232651885978,
|
| 888 |
-
"grad_norm": 31.
|
| 889 |
"learning_rate": 1.9759590554616173e-07,
|
| 890 |
-
"logits/chosen": -0.
|
| 891 |
-
"logits/rejected": 0.
|
| 892 |
-
"logps/chosen": -
|
| 893 |
-
"logps/rejected": -
|
| 894 |
-
"loss": 0.
|
| 895 |
-
"rewards/accuracies": 0.
|
| 896 |
-
"rewards/chosen": -3.
|
| 897 |
-
"rewards/margins": 2.
|
| 898 |
-
"rewards/rejected": -5.
|
| 899 |
"step": 530
|
| 900 |
},
|
| 901 |
{
|
| 902 |
"epoch": 0.6219406852864958,
|
| 903 |
-
"grad_norm":
|
| 904 |
"learning_rate": 1.8780804765620746e-07,
|
| 905 |
-
"logits/chosen": -0.
|
| 906 |
-
"logits/rejected":
|
| 907 |
-
"logps/chosen": -
|
| 908 |
-
"logps/rejected": -
|
| 909 |
-
"loss": 0.
|
| 910 |
-
"rewards/accuracies": 0.
|
| 911 |
-
"rewards/chosen": -2.
|
| 912 |
-
"rewards/margins": 2.
|
| 913 |
-
"rewards/rejected": -5.
|
| 914 |
"step": 540
|
| 915 |
},
|
| 916 |
{
|
| 917 |
"epoch": 0.6334581053843938,
|
| 918 |
-
"grad_norm":
|
| 919 |
"learning_rate": 1.7812080721643973e-07,
|
| 920 |
-
"logits/chosen": -0.
|
| 921 |
-
"logits/rejected": 0.
|
| 922 |
-
"logps/chosen": -
|
| 923 |
-
"logps/rejected": -
|
| 924 |
-
"loss": 0.
|
| 925 |
-
"rewards/accuracies": 0.
|
| 926 |
-
"rewards/chosen": -2.
|
| 927 |
-
"rewards/margins":
|
| 928 |
-
"rewards/rejected": -
|
| 929 |
"step": 550
|
| 930 |
},
|
| 931 |
{
|
| 932 |
"epoch": 0.644975525482292,
|
| 933 |
-
"grad_norm":
|
| 934 |
"learning_rate": 1.6854985675997063e-07,
|
| 935 |
-
"logits/chosen": -0.
|
| 936 |
-
"logits/rejected": 0.
|
| 937 |
-
"logps/chosen": -
|
| 938 |
-
"logps/rejected": -
|
| 939 |
-
"loss": 0.
|
| 940 |
-
"rewards/accuracies": 0.
|
| 941 |
-
"rewards/chosen": -2.
|
| 942 |
-
"rewards/margins": 2.
|
| 943 |
-
"rewards/rejected": -5.
|
| 944 |
"step": 560
|
| 945 |
},
|
| 946 |
{
|
| 947 |
"epoch": 0.65649294558019,
|
| 948 |
-
"grad_norm":
|
| 949 |
"learning_rate": 1.5911068067978818e-07,
|
| 950 |
-
"logits/chosen": -0.
|
| 951 |
-
"logits/rejected":
|
| 952 |
-
"logps/chosen": -
|
| 953 |
-
"logps/rejected": -
|
| 954 |
-
"loss": 0.
|
| 955 |
-
"rewards/accuracies": 0.
|
| 956 |
-
"rewards/chosen": -2.
|
| 957 |
-
"rewards/margins": 2.
|
| 958 |
-
"rewards/rejected": -5.
|
| 959 |
"step": 570
|
| 960 |
},
|
| 961 |
{
|
| 962 |
"epoch": 0.6680103656780881,
|
| 963 |
-
"grad_norm":
|
| 964 |
"learning_rate": 1.4981855017728197e-07,
|
| 965 |
-
"logits/chosen": 0.
|
| 966 |
-
"logits/rejected": 0.
|
| 967 |
-
"logps/chosen": -
|
| 968 |
-
"logps/rejected": -
|
| 969 |
-
"loss": 0.
|
| 970 |
-
"rewards/accuracies": 0.
|
| 971 |
-
"rewards/chosen": -
|
| 972 |
-
"rewards/margins": 2.
|
| 973 |
-
"rewards/rejected": -5.
|
| 974 |
"step": 580
|
| 975 |
},
|
| 976 |
{
|
| 977 |
"epoch": 0.6795277857759862,
|
| 978 |
-
"grad_norm":
|
| 979 |
"learning_rate": 1.406884985556804e-07,
|
| 980 |
-
"logits/chosen": -0.
|
| 981 |
-
"logits/rejected":
|
| 982 |
-
"logps/chosen": -
|
| 983 |
-
"logps/rejected": -
|
| 984 |
-
"loss": 0.
|
| 985 |
-
"rewards/accuracies": 0.
|
| 986 |
-
"rewards/chosen": -
|
| 987 |
-
"rewards/margins": 2.
|
| 988 |
-
"rewards/rejected": -
|
| 989 |
"step": 590
|
| 990 |
},
|
| 991 |
{
|
| 992 |
"epoch": 0.6910452058738843,
|
| 993 |
-
"grad_norm":
|
| 994 |
"learning_rate": 1.3173529689837354e-07,
|
| 995 |
-
"logits/chosen": -0.
|
| 996 |
-
"logits/rejected": 0.
|
| 997 |
-
"logps/chosen": -
|
| 998 |
-
"logps/rejected": -
|
| 999 |
-
"loss": 0.
|
| 1000 |
-
"rewards/accuracies": 0.
|
| 1001 |
-
"rewards/chosen": -
|
| 1002 |
-
"rewards/margins": 3.
|
| 1003 |
-
"rewards/rejected": -6.
|
| 1004 |
"step": 600
|
| 1005 |
},
|
| 1006 |
{
|
| 1007 |
"epoch": 0.6910452058738843,
|
| 1008 |
-
"eval_logits/chosen": -0.
|
| 1009 |
-
"eval_logits/rejected": 0.
|
| 1010 |
-
"eval_logps/chosen": -
|
| 1011 |
-
"eval_logps/rejected": -
|
| 1012 |
-
"eval_loss": 0.
|
| 1013 |
-
"eval_rewards/accuracies": 0.
|
| 1014 |
-
"eval_rewards/chosen": -
|
| 1015 |
-
"eval_rewards/margins": 3.
|
| 1016 |
-
"eval_rewards/rejected": -
|
| 1017 |
-
"eval_runtime":
|
| 1018 |
-
"eval_samples_per_second":
|
| 1019 |
-
"eval_steps_per_second": 0.
|
| 1020 |
"step": 600
|
| 1021 |
},
|
| 1022 |
{
|
| 1023 |
"epoch": 0.7025626259717823,
|
| 1024 |
-
"grad_norm":
|
| 1025 |
"learning_rate": 1.2297343017146726e-07,
|
| 1026 |
-
"logits/chosen": 0.
|
| 1027 |
-
"logits/rejected":
|
| 1028 |
-
"logps/chosen": -
|
| 1029 |
-
"logps/rejected": -
|
| 1030 |
-
"loss": 0.
|
| 1031 |
-
"rewards/accuracies": 0.
|
| 1032 |
-
"rewards/chosen": -
|
| 1033 |
-
"rewards/margins": 2.
|
| 1034 |
-
"rewards/rejected": -
|
| 1035 |
"step": 610
|
| 1036 |
},
|
| 1037 |
{
|
| 1038 |
"epoch": 0.7140800460696803,
|
| 1039 |
-
"grad_norm":
|
| 1040 |
"learning_rate": 1.1441707378923474e-07,
|
| 1041 |
-
"logits/chosen": 0.
|
| 1042 |
-
"logits/rejected":
|
| 1043 |
-
"logps/chosen": -
|
| 1044 |
-
"logps/rejected": -
|
| 1045 |
-
"loss": 0.
|
| 1046 |
-
"rewards/accuracies": 0.
|
| 1047 |
-
"rewards/chosen": -
|
| 1048 |
-
"rewards/margins":
|
| 1049 |
-
"rewards/rejected": -
|
| 1050 |
"step": 620
|
| 1051 |
},
|
| 1052 |
{
|
| 1053 |
"epoch": 0.7255974661675785,
|
| 1054 |
-
"grad_norm": 22.
|
| 1055 |
"learning_rate": 1.06080070680377e-07,
|
| 1056 |
-
"logits/chosen": 0.
|
| 1057 |
-
"logits/rejected":
|
| 1058 |
-
"logps/chosen": -
|
| 1059 |
-
"logps/rejected": -
|
| 1060 |
-
"loss": 0.
|
| 1061 |
-
"rewards/accuracies": 0.
|
| 1062 |
-
"rewards/chosen": -
|
| 1063 |
-
"rewards/margins": 2.
|
| 1064 |
-
"rewards/rejected": -5.
|
| 1065 |
"step": 630
|
| 1066 |
},
|
| 1067 |
{
|
| 1068 |
"epoch": 0.7371148862654765,
|
| 1069 |
-
"grad_norm":
|
| 1070 |
"learning_rate": 9.797590889219587e-08,
|
| 1071 |
-
"logits/chosen":
|
| 1072 |
-
"logits/rejected": 0.
|
| 1073 |
-
"logps/chosen": -
|
| 1074 |
-
"logps/rejected": -
|
| 1075 |
-
"loss": 0.
|
| 1076 |
-
"rewards/accuracies": 0.
|
| 1077 |
-
"rewards/chosen": -3.
|
| 1078 |
-
"rewards/margins":
|
| 1079 |
-
"rewards/rejected": -
|
| 1080 |
"step": 640
|
| 1081 |
},
|
| 1082 |
{
|
| 1083 |
"epoch": 0.7486323063633746,
|
| 1084 |
-
"grad_norm":
|
| 1085 |
"learning_rate": 9.011769976891367e-08,
|
| 1086 |
-
"logits/chosen": 0.
|
| 1087 |
-
"logits/rejected": 1.
|
| 1088 |
-
"logps/chosen": -
|
| 1089 |
-
"logps/rejected": -
|
| 1090 |
-
"loss": 0.
|
| 1091 |
-
"rewards/accuracies": 0.
|
| 1092 |
-
"rewards/chosen": -2.
|
| 1093 |
-
"rewards/margins": 2.
|
| 1094 |
-
"rewards/rejected": -5.
|
| 1095 |
"step": 650
|
| 1096 |
},
|
| 1097 |
{
|
| 1098 |
"epoch": 0.7601497264612727,
|
| 1099 |
-
"grad_norm":
|
| 1100 |
"learning_rate": 8.251815673944218e-08,
|
| 1101 |
-
"logits/chosen": -0.
|
| 1102 |
-
"logits/rejected": 0.
|
| 1103 |
-
"logps/chosen": -
|
| 1104 |
-
"logps/rejected": -
|
| 1105 |
-
"loss": 0.
|
| 1106 |
-
"rewards/accuracies": 0.
|
| 1107 |
-
"rewards/chosen": -3.
|
| 1108 |
-
"rewards/margins": 3.
|
| 1109 |
-
"rewards/rejected": -6.
|
| 1110 |
"step": 660
|
| 1111 |
},
|
| 1112 |
{
|
| 1113 |
"epoch": 0.7716671465591708,
|
| 1114 |
-
"grad_norm":
|
| 1115 |
"learning_rate": 7.518957474892148e-08,
|
| 1116 |
-
"logits/chosen": 0.
|
| 1117 |
-
"logits/rejected": 0.
|
| 1118 |
-
"logps/chosen": -
|
| 1119 |
-
"logps/rejected": -
|
| 1120 |
-
"loss": 0.
|
| 1121 |
-
"rewards/accuracies": 0.
|
| 1122 |
-
"rewards/chosen": -
|
| 1123 |
-
"rewards/margins":
|
| 1124 |
-
"rewards/rejected": -
|
| 1125 |
"step": 670
|
| 1126 |
},
|
| 1127 |
{
|
| 1128 |
"epoch": 0.7831845666570688,
|
| 1129 |
-
"grad_norm":
|
| 1130 |
"learning_rate": 6.814381036730274e-08,
|
| 1131 |
-
"logits/chosen": -0.
|
| 1132 |
-
"logits/rejected": 0.
|
| 1133 |
-
"logps/chosen": -
|
| 1134 |
-
"logps/rejected": -
|
| 1135 |
-
"loss": 0.
|
| 1136 |
-
"rewards/accuracies": 0.
|
| 1137 |
-
"rewards/chosen": -3.
|
| 1138 |
-
"rewards/margins":
|
| 1139 |
-
"rewards/rejected": -
|
| 1140 |
"step": 680
|
| 1141 |
},
|
| 1142 |
{
|
| 1143 |
"epoch": 0.7947019867549668,
|
| 1144 |
-
"grad_norm":
|
| 1145 |
"learning_rate": 6.139226260715872e-08,
|
| 1146 |
-
"logits/chosen": -0.
|
| 1147 |
-
"logits/rejected": 0.
|
| 1148 |
-
"logps/chosen": -
|
| 1149 |
-
"logps/rejected": -
|
| 1150 |
-
"loss": 0.
|
| 1151 |
-
"rewards/accuracies": 0.
|
| 1152 |
-
"rewards/chosen": -3.
|
| 1153 |
-
"rewards/margins":
|
| 1154 |
-
"rewards/rejected": -6.
|
| 1155 |
"step": 690
|
| 1156 |
},
|
| 1157 |
{
|
| 1158 |
"epoch": 0.806219406852865,
|
| 1159 |
-
"grad_norm":
|
| 1160 |
"learning_rate": 5.4945854481754734e-08,
|
| 1161 |
-
"logits/chosen": 0.
|
| 1162 |
-
"logits/rejected": 0.
|
| 1163 |
-
"logps/chosen": -
|
| 1164 |
-
"logps/rejected": -
|
| 1165 |
-
"loss": 0.
|
| 1166 |
-
"rewards/accuracies": 0.
|
| 1167 |
-
"rewards/chosen": -3.
|
| 1168 |
-
"rewards/margins": 3.
|
| 1169 |
-
"rewards/rejected": -
|
| 1170 |
"step": 700
|
| 1171 |
},
|
| 1172 |
{
|
| 1173 |
"epoch": 0.806219406852865,
|
| 1174 |
-
"eval_logits/chosen": -0.
|
| 1175 |
-
"eval_logits/rejected": 0.
|
| 1176 |
-
"eval_logps/chosen": -
|
| 1177 |
-
"eval_logps/rejected": -
|
| 1178 |
-
"eval_loss": 0.
|
| 1179 |
-
"eval_rewards/accuracies": 0.
|
| 1180 |
-
"eval_rewards/chosen": -
|
| 1181 |
-
"eval_rewards/margins": 4.
|
| 1182 |
-
"eval_rewards/rejected": -8.
|
| 1183 |
-
"eval_runtime":
|
| 1184 |
-
"eval_samples_per_second":
|
| 1185 |
-
"eval_steps_per_second": 0.
|
| 1186 |
"step": 700
|
| 1187 |
},
|
| 1188 |
{
|
| 1189 |
"epoch": 0.817736826950763,
|
| 1190 |
-
"grad_norm": 26.
|
| 1191 |
"learning_rate": 4.881501533321605e-08,
|
| 1192 |
-
"logits/chosen": -0.
|
| 1193 |
-
"logits/rejected": 0.
|
| 1194 |
-
"logps/chosen": -
|
| 1195 |
-
"logps/rejected": -
|
| 1196 |
-
"loss": 0.
|
| 1197 |
-
"rewards/accuracies": 0.
|
| 1198 |
-
"rewards/chosen": -
|
| 1199 |
-
"rewards/margins":
|
| 1200 |
-
"rewards/rejected": -6.
|
| 1201 |
"step": 710
|
| 1202 |
},
|
| 1203 |
{
|
| 1204 |
"epoch": 0.8292542470486611,
|
| 1205 |
-
"grad_norm":
|
| 1206 |
"learning_rate": 4.300966395938377e-08,
|
| 1207 |
-
"logits/chosen": -0.
|
| 1208 |
-
"logits/rejected": 0.
|
| 1209 |
-
"logps/chosen": -
|
| 1210 |
-
"logps/rejected": -
|
| 1211 |
-
"loss": 0.
|
| 1212 |
-
"rewards/accuracies": 0.
|
| 1213 |
-
"rewards/chosen": -3.
|
| 1214 |
-
"rewards/margins": 3.
|
| 1215 |
-
"rewards/rejected": -6.
|
| 1216 |
"step": 720
|
| 1217 |
},
|
| 1218 |
{
|
| 1219 |
"epoch": 0.8407716671465592,
|
| 1220 |
-
"grad_norm": 27.
|
| 1221 |
"learning_rate": 3.7539192566655246e-08,
|
| 1222 |
-
"logits/chosen": -0.
|
| 1223 |
-
"logits/rejected": 0.
|
| 1224 |
-
"logps/chosen": -
|
| 1225 |
-
"logps/rejected": -
|
| 1226 |
-
"loss": 0.
|
| 1227 |
-
"rewards/accuracies": 0.
|
| 1228 |
-
"rewards/chosen": -3.
|
| 1229 |
-
"rewards/margins":
|
| 1230 |
-
"rewards/rejected": -6.
|
| 1231 |
"step": 730
|
| 1232 |
},
|
| 1233 |
{
|
| 1234 |
"epoch": 0.8522890872444573,
|
| 1235 |
-
"grad_norm":
|
| 1236 |
"learning_rate": 3.24124515747731e-08,
|
| 1237 |
-
"logits/chosen":
|
| 1238 |
-
"logits/rejected": 1.
|
| 1239 |
-
"logps/chosen": -
|
| 1240 |
-
"logps/rejected": -
|
| 1241 |
-
"loss": 0.
|
| 1242 |
-
"rewards/accuracies": 0.
|
| 1243 |
-
"rewards/chosen": -3.
|
| 1244 |
-
"rewards/margins": 3.
|
| 1245 |
-
"rewards/rejected": -
|
| 1246 |
"step": 740
|
| 1247 |
},
|
| 1248 |
{
|
| 1249 |
"epoch": 0.8638065073423553,
|
| 1250 |
-
"grad_norm":
|
| 1251 |
"learning_rate": 2.763773529814506e-08,
|
| 1252 |
-
"logits/chosen": 0.
|
| 1253 |
-
"logits/rejected":
|
| 1254 |
-
"logps/chosen": -
|
| 1255 |
-
"logps/rejected": -
|
| 1256 |
-
"loss": 0.
|
| 1257 |
-
"rewards/accuracies": 0.
|
| 1258 |
-
"rewards/chosen": -3.
|
| 1259 |
-
"rewards/margins": 3.
|
| 1260 |
-
"rewards/rejected": -6.
|
| 1261 |
"step": 750
|
| 1262 |
},
|
| 1263 |
{
|
| 1264 |
"epoch": 0.8753239274402533,
|
| 1265 |
-
"grad_norm":
|
| 1266 |
"learning_rate": 2.3222768526860698e-08,
|
| 1267 |
-
"logits/chosen": 0.
|
| 1268 |
-
"logits/rejected":
|
| 1269 |
-
"logps/chosen": -613.
|
| 1270 |
-
"logps/rejected": -
|
| 1271 |
-
"loss": 0.
|
| 1272 |
-
"rewards/accuracies": 0.
|
| 1273 |
-
"rewards/chosen": -3.
|
| 1274 |
-
"rewards/margins": 2.
|
| 1275 |
-
"rewards/rejected": -6.
|
| 1276 |
"step": 760
|
| 1277 |
},
|
| 1278 |
{
|
| 1279 |
"epoch": 0.8868413475381515,
|
| 1280 |
-
"grad_norm":
|
| 1281 |
"learning_rate": 1.9174694029115146e-08,
|
| 1282 |
-
"logits/chosen": 0.
|
| 1283 |
-
"logits/rejected": 1.
|
| 1284 |
-
"logps/chosen": -
|
| 1285 |
-
"logps/rejected": -
|
| 1286 |
-
"loss": 0.
|
| 1287 |
-
"rewards/accuracies": 0.
|
| 1288 |
-
"rewards/chosen": -3.
|
| 1289 |
-
"rewards/margins": 3.
|
| 1290 |
-
"rewards/rejected": -6.
|
| 1291 |
"step": 770
|
| 1292 |
},
|
| 1293 |
{
|
| 1294 |
"epoch": 0.8983587676360495,
|
| 1295 |
-
"grad_norm": 29.
|
| 1296 |
"learning_rate": 1.5500060995258134e-08,
|
| 1297 |
-
"logits/chosen": 0.
|
| 1298 |
-
"logits/rejected": 1.
|
| 1299 |
-
"logps/chosen": -
|
| 1300 |
-
"logps/rejected": -
|
| 1301 |
-
"loss": 0.
|
| 1302 |
-
"rewards/accuracies": 0.
|
| 1303 |
-
"rewards/chosen": -3.
|
| 1304 |
-
"rewards/margins": 2.
|
| 1305 |
-
"rewards/rejected": -
|
| 1306 |
"step": 780
|
| 1307 |
},
|
| 1308 |
{
|
| 1309 |
"epoch": 0.9098761877339476,
|
| 1310 |
-
"grad_norm": 31.
|
| 1311 |
"learning_rate": 1.2204814442165812e-08,
|
| 1312 |
-
"logits/chosen": 0.
|
| 1313 |
-
"logits/rejected":
|
| 1314 |
-
"logps/chosen": -
|
| 1315 |
-
"logps/rejected": -
|
| 1316 |
-
"loss": 0.
|
| 1317 |
-
"rewards/accuracies": 0.
|
| 1318 |
-
"rewards/chosen": -3.
|
| 1319 |
-
"rewards/margins":
|
| 1320 |
-
"rewards/rejected": -6.
|
| 1321 |
"step": 790
|
| 1322 |
},
|
| 1323 |
{
|
| 1324 |
"epoch": 0.9213936078318457,
|
| 1325 |
-
"grad_norm":
|
| 1326 |
"learning_rate": 9.294285595075669e-09,
|
| 1327 |
-
"logits/chosen": 0.
|
| 1328 |
-
"logits/rejected": 1.
|
| 1329 |
-
"logps/chosen": -
|
| 1330 |
-
"logps/rejected": -
|
| 1331 |
-
"loss": 0.
|
| 1332 |
-
"rewards/accuracies": 0.
|
| 1333 |
-
"rewards/chosen": -3.
|
| 1334 |
-
"rewards/margins": 3.
|
| 1335 |
-
"rewards/rejected": -6.
|
| 1336 |
"step": 800
|
| 1337 |
},
|
| 1338 |
{
|
| 1339 |
"epoch": 0.9213936078318457,
|
| 1340 |
-
"eval_logits/chosen": -0.
|
| 1341 |
-
"eval_logits/rejected": 0.
|
| 1342 |
-
"eval_logps/chosen": -
|
| 1343 |
-
"eval_logps/rejected": -
|
| 1344 |
-
"eval_loss": 0.
|
| 1345 |
-
"eval_rewards/accuracies": 0.
|
| 1346 |
-
"eval_rewards/chosen": -4.
|
| 1347 |
-
"eval_rewards/margins": 4.
|
| 1348 |
-
"eval_rewards/rejected": -
|
| 1349 |
-
"eval_runtime":
|
| 1350 |
-
"eval_samples_per_second":
|
| 1351 |
-
"eval_steps_per_second": 0.
|
| 1352 |
"step": 800
|
| 1353 |
},
|
| 1354 |
{
|
| 1355 |
"epoch": 0.9329110279297438,
|
| 1356 |
-
"grad_norm":
|
| 1357 |
"learning_rate": 6.773183262446914e-09,
|
| 1358 |
-
"logits/chosen": 0.
|
| 1359 |
-
"logits/rejected": 1.
|
| 1360 |
-
"logps/chosen": -
|
| 1361 |
-
"logps/rejected": -
|
| 1362 |
-
"loss": 0.
|
| 1363 |
-
"rewards/accuracies": 0.
|
| 1364 |
-
"rewards/chosen": -3.
|
| 1365 |
-
"rewards/margins": 3.
|
| 1366 |
-
"rewards/rejected": -6.
|
| 1367 |
"step": 810
|
| 1368 |
},
|
| 1369 |
{
|
| 1370 |
"epoch": 0.9444284480276418,
|
| 1371 |
-
"grad_norm":
|
| 1372 |
"learning_rate": 4.645586217799452e-09,
|
| 1373 |
-
"logits/chosen": -0.
|
| 1374 |
-
"logits/rejected":
|
| 1375 |
-
"logps/chosen": -
|
| 1376 |
-
"logps/rejected": -
|
| 1377 |
-
"loss": 0.
|
| 1378 |
-
"rewards/accuracies": 0.
|
| 1379 |
-
"rewards/chosen": -
|
| 1380 |
-
"rewards/margins": 3.
|
| 1381 |
-
"rewards/rejected": -6.
|
| 1382 |
"step": 820
|
| 1383 |
},
|
| 1384 |
{
|
| 1385 |
"epoch": 0.9559458681255398,
|
| 1386 |
-
"grad_norm":
|
| 1387 |
"learning_rate": 2.9149366008568987e-09,
|
| 1388 |
-
"logits/chosen": 0.
|
| 1389 |
-
"logits/rejected":
|
| 1390 |
-
"logps/chosen": -
|
| 1391 |
-
"logps/rejected": -
|
| 1392 |
-
"loss": 0.
|
| 1393 |
"rewards/accuracies": 0.778124988079071,
|
| 1394 |
-
"rewards/chosen": -
|
| 1395 |
-
"rewards/margins":
|
| 1396 |
-
"rewards/rejected": -
|
| 1397 |
"step": 830
|
| 1398 |
},
|
| 1399 |
{
|
| 1400 |
"epoch": 0.967463288223438,
|
| 1401 |
-
"grad_norm":
|
| 1402 |
"learning_rate": 1.5840343486700215e-09,
|
| 1403 |
-
"logits/chosen": 0.
|
| 1404 |
-
"logits/rejected": 1.
|
| 1405 |
-
"logps/chosen": -
|
| 1406 |
-
"logps/rejected": -
|
| 1407 |
-
"loss": 0.
|
| 1408 |
-
"rewards/accuracies": 0.
|
| 1409 |
-
"rewards/chosen": -3.
|
| 1410 |
-
"rewards/margins": 3.
|
| 1411 |
-
"rewards/rejected": -
|
| 1412 |
"step": 840
|
| 1413 |
},
|
| 1414 |
{
|
| 1415 |
"epoch": 0.978980708321336,
|
| 1416 |
-
"grad_norm":
|
| 1417 |
"learning_rate": 6.550326657293881e-10,
|
| 1418 |
-
"logits/chosen": 0.
|
| 1419 |
-
"logits/rejected": 1.
|
| 1420 |
-
"logps/chosen": -
|
| 1421 |
-
"logps/rejected": -
|
| 1422 |
-
"loss": 0.
|
| 1423 |
-
"rewards/accuracies": 0.
|
| 1424 |
-
"rewards/chosen": -3.
|
| 1425 |
-
"rewards/margins": 2.
|
| 1426 |
-
"rewards/rejected": -
|
| 1427 |
"step": 850
|
| 1428 |
},
|
| 1429 |
{
|
| 1430 |
"epoch": 0.9904981284192341,
|
| 1431 |
-
"grad_norm":
|
| 1432 |
"learning_rate": 1.2943454039654467e-10,
|
| 1433 |
-
"logits/chosen": 0.
|
| 1434 |
-
"logits/rejected": 1.
|
| 1435 |
-
"logps/chosen": -
|
| 1436 |
-
"logps/rejected": -
|
| 1437 |
-
"loss": 0.
|
| 1438 |
-
"rewards/accuracies": 0.
|
| 1439 |
-
"rewards/chosen": -
|
| 1440 |
-
"rewards/margins": 2.
|
| 1441 |
-
"rewards/rejected": -5.
|
| 1442 |
"step": 860
|
| 1443 |
},
|
| 1444 |
{
|
| 1445 |
"epoch": 0.9997120644975526,
|
| 1446 |
"step": 868,
|
| 1447 |
"total_flos": 0.0,
|
| 1448 |
-
"train_loss": 0.
|
| 1449 |
-
"train_runtime":
|
| 1450 |
-
"train_samples_per_second":
|
| 1451 |
-
"train_steps_per_second":
|
| 1452 |
}
|
| 1453 |
],
|
| 1454 |
"logging_steps": 10,
|
|
|
|
| 9 |
"is_world_process_zero": true,
|
| 10 |
"log_history": [
|
| 11 |
{
|
| 12 |
+
"epoch": 0.001151742009789807,
|
| 13 |
+
"grad_norm": 8.954344956678737,
|
| 14 |
+
"learning_rate": 5.747126436781609e-09,
|
| 15 |
+
"logits/chosen": -2.759351968765259,
|
| 16 |
+
"logits/rejected": -2.762708902359009,
|
| 17 |
+
"logps/chosen": -256.92901611328125,
|
| 18 |
+
"logps/rejected": -268.4883728027344,
|
| 19 |
"loss": 0.6931,
|
| 20 |
"rewards/accuracies": 0.0,
|
| 21 |
"rewards/chosen": 0.0,
|
|
|
|
| 25 |
},
|
| 26 |
{
|
| 27 |
"epoch": 0.01151742009789807,
|
| 28 |
+
"grad_norm": 8.455865944439507,
|
| 29 |
"learning_rate": 5.747126436781609e-08,
|
| 30 |
+
"logits/chosen": -2.8186073303222656,
|
| 31 |
+
"logits/rejected": -2.7818901538848877,
|
| 32 |
+
"logps/chosen": -280.477783203125,
|
| 33 |
+
"logps/rejected": -273.6023864746094,
|
| 34 |
+
"loss": 0.6932,
|
| 35 |
+
"rewards/accuracies": 0.3854166567325592,
|
| 36 |
+
"rewards/chosen": 0.0001983554830076173,
|
| 37 |
+
"rewards/margins": -0.00018551234097685665,
|
| 38 |
+
"rewards/rejected": 0.0003838678530883044,
|
| 39 |
"step": 10
|
| 40 |
},
|
| 41 |
{
|
| 42 |
"epoch": 0.02303484019579614,
|
| 43 |
+
"grad_norm": 8.281991111156,
|
| 44 |
"learning_rate": 1.1494252873563217e-07,
|
| 45 |
+
"logits/chosen": -2.780491828918457,
|
| 46 |
+
"logits/rejected": -2.7574048042297363,
|
| 47 |
+
"logps/chosen": -288.68408203125,
|
| 48 |
+
"logps/rejected": -274.33233642578125,
|
| 49 |
+
"loss": 0.6929,
|
| 50 |
"rewards/accuracies": 0.5406249761581421,
|
| 51 |
+
"rewards/chosen": 0.000404498161515221,
|
| 52 |
+
"rewards/margins": 0.0008455432835035026,
|
| 53 |
+
"rewards/rejected": -0.00044104509288445115,
|
| 54 |
"step": 20
|
| 55 |
},
|
| 56 |
{
|
| 57 |
"epoch": 0.03455226029369421,
|
| 58 |
+
"grad_norm": 8.294840577575956,
|
| 59 |
"learning_rate": 1.7241379310344828e-07,
|
| 60 |
+
"logits/chosen": -2.7976105213165283,
|
| 61 |
+
"logits/rejected": -2.771127462387085,
|
| 62 |
+
"logps/chosen": -282.4824523925781,
|
| 63 |
+
"logps/rejected": -267.7333984375,
|
| 64 |
+
"loss": 0.6915,
|
| 65 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 66 |
+
"rewards/chosen": 0.002527938922867179,
|
| 67 |
+
"rewards/margins": 0.0036745104007422924,
|
| 68 |
+
"rewards/rejected": -0.0011465717107057571,
|
| 69 |
"step": 30
|
| 70 |
},
|
| 71 |
{
|
| 72 |
"epoch": 0.04606968039159228,
|
| 73 |
+
"grad_norm": 8.712130617823888,
|
| 74 |
"learning_rate": 2.2988505747126435e-07,
|
| 75 |
+
"logits/chosen": -2.7874083518981934,
|
| 76 |
+
"logits/rejected": -2.7691152095794678,
|
| 77 |
+
"logps/chosen": -276.18780517578125,
|
| 78 |
+
"logps/rejected": -266.81781005859375,
|
| 79 |
+
"loss": 0.6874,
|
| 80 |
+
"rewards/accuracies": 0.706250011920929,
|
| 81 |
+
"rewards/chosen": 0.0077004628255963326,
|
| 82 |
+
"rewards/margins": 0.010546171106398106,
|
| 83 |
+
"rewards/rejected": -0.0028457094449549913,
|
| 84 |
"step": 40
|
| 85 |
},
|
| 86 |
{
|
| 87 |
"epoch": 0.05758710048949035,
|
| 88 |
+
"grad_norm": 10.31218146006529,
|
| 89 |
"learning_rate": 2.873563218390804e-07,
|
| 90 |
+
"logits/chosen": -2.8169891834259033,
|
| 91 |
+
"logits/rejected": -2.780351161956787,
|
| 92 |
+
"logps/chosen": -288.07293701171875,
|
| 93 |
+
"logps/rejected": -277.56207275390625,
|
| 94 |
+
"loss": 0.6779,
|
| 95 |
+
"rewards/accuracies": 0.746874988079071,
|
| 96 |
+
"rewards/chosen": 0.016184702515602112,
|
| 97 |
+
"rewards/margins": 0.033592261373996735,
|
| 98 |
+
"rewards/rejected": -0.01740756258368492,
|
| 99 |
"step": 50
|
| 100 |
},
|
| 101 |
{
|
| 102 |
"epoch": 0.06910452058738842,
|
| 103 |
+
"grad_norm": 8.630622631020962,
|
| 104 |
"learning_rate": 3.4482758620689656e-07,
|
| 105 |
+
"logits/chosen": -2.8177428245544434,
|
| 106 |
+
"logits/rejected": -2.7997875213623047,
|
| 107 |
+
"logps/chosen": -284.9295959472656,
|
| 108 |
+
"logps/rejected": -300.76239013671875,
|
| 109 |
+
"loss": 0.6642,
|
| 110 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 111 |
+
"rewards/chosen": 0.014625328592956066,
|
| 112 |
+
"rewards/margins": 0.06306789815425873,
|
| 113 |
+
"rewards/rejected": -0.04844257980585098,
|
| 114 |
"step": 60
|
| 115 |
},
|
| 116 |
{
|
| 117 |
"epoch": 0.0806219406852865,
|
| 118 |
+
"grad_norm": 10.734027101004664,
|
| 119 |
"learning_rate": 4.0229885057471266e-07,
|
| 120 |
+
"logits/chosen": -2.7529358863830566,
|
| 121 |
+
"logits/rejected": -2.7322020530700684,
|
| 122 |
+
"logps/chosen": -290.8133850097656,
|
| 123 |
+
"logps/rejected": -287.9371643066406,
|
| 124 |
+
"loss": 0.6367,
|
| 125 |
+
"rewards/accuracies": 0.715624988079071,
|
| 126 |
+
"rewards/chosen": -0.06497061997652054,
|
| 127 |
+
"rewards/margins": 0.13046525418758392,
|
| 128 |
+
"rewards/rejected": -0.19543585181236267,
|
| 129 |
"step": 70
|
| 130 |
},
|
| 131 |
{
|
| 132 |
"epoch": 0.09213936078318456,
|
| 133 |
+
"grad_norm": 11.33124626301948,
|
| 134 |
"learning_rate": 4.597701149425287e-07,
|
| 135 |
+
"logits/chosen": -2.861381769180298,
|
| 136 |
+
"logits/rejected": -2.830373525619507,
|
| 137 |
+
"logps/chosen": -326.0346374511719,
|
| 138 |
+
"logps/rejected": -342.4071960449219,
|
| 139 |
+
"loss": 0.5943,
|
| 140 |
+
"rewards/accuracies": 0.7718750238418579,
|
| 141 |
+
"rewards/chosen": -0.21955294907093048,
|
| 142 |
+
"rewards/margins": 0.28150954842567444,
|
| 143 |
+
"rewards/rejected": -0.5010625123977661,
|
| 144 |
"step": 80
|
| 145 |
},
|
| 146 |
{
|
| 147 |
"epoch": 0.10365678088108264,
|
| 148 |
+
"grad_norm": 20.262969117511794,
|
| 149 |
"learning_rate": 4.999817969178237e-07,
|
| 150 |
+
"logits/chosen": -2.802300214767456,
|
| 151 |
+
"logits/rejected": -2.766491174697876,
|
| 152 |
+
"logps/chosen": -352.80755615234375,
|
| 153 |
+
"logps/rejected": -383.49725341796875,
|
| 154 |
+
"loss": 0.5561,
|
| 155 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 156 |
+
"rewards/chosen": -0.6216552257537842,
|
| 157 |
+
"rewards/margins": 0.41057389974594116,
|
| 158 |
+
"rewards/rejected": -1.0322291851043701,
|
| 159 |
"step": 90
|
| 160 |
},
|
| 161 |
{
|
| 162 |
"epoch": 0.1151742009789807,
|
| 163 |
+
"grad_norm": 21.31118240013579,
|
| 164 |
"learning_rate": 4.996582603056428e-07,
|
| 165 |
+
"logits/chosen": -2.789545774459839,
|
| 166 |
+
"logits/rejected": -2.761136293411255,
|
| 167 |
+
"logps/chosen": -439.2823791503906,
|
| 168 |
+
"logps/rejected": -501.868408203125,
|
| 169 |
+
"loss": 0.5381,
|
| 170 |
+
"rewards/accuracies": 0.699999988079071,
|
| 171 |
+
"rewards/chosen": -1.3656787872314453,
|
| 172 |
+
"rewards/margins": 0.7182655334472656,
|
| 173 |
+
"rewards/rejected": -2.08394455909729,
|
| 174 |
"step": 100
|
| 175 |
},
|
| 176 |
{
|
| 177 |
"epoch": 0.1151742009789807,
|
| 178 |
+
"eval_logits/chosen": -2.9044759273529053,
|
| 179 |
+
"eval_logits/rejected": -2.882232189178467,
|
| 180 |
+
"eval_logps/chosen": -489.789306640625,
|
| 181 |
+
"eval_logps/rejected": -581.798095703125,
|
| 182 |
+
"eval_loss": 0.4757886528968811,
|
| 183 |
+
"eval_rewards/accuracies": 0.7270179390907288,
|
| 184 |
+
"eval_rewards/chosen": -1.9882127046585083,
|
| 185 |
+
"eval_rewards/margins": 0.9288455843925476,
|
| 186 |
+
"eval_rewards/rejected": -2.9170584678649902,
|
| 187 |
+
"eval_runtime": 641.6076,
|
| 188 |
+
"eval_samples_per_second": 11.106,
|
| 189 |
+
"eval_steps_per_second": 0.348,
|
| 190 |
"step": 100
|
| 191 |
},
|
| 192 |
{
|
| 193 |
"epoch": 0.12669162107687879,
|
| 194 |
+
"grad_norm": 19.990999467237586,
|
| 195 |
"learning_rate": 4.989308132738126e-07,
|
| 196 |
+
"logits/chosen": -2.7357916831970215,
|
| 197 |
+
"logits/rejected": -2.711280345916748,
|
| 198 |
+
"logps/chosen": -394.81048583984375,
|
| 199 |
+
"logps/rejected": -465.7745056152344,
|
| 200 |
+
"loss": 0.5215,
|
| 201 |
+
"rewards/accuracies": 0.690625011920929,
|
| 202 |
+
"rewards/chosen": -1.2416951656341553,
|
| 203 |
+
"rewards/margins": 0.7623735666275024,
|
| 204 |
+
"rewards/rejected": -2.0040688514709473,
|
| 205 |
"step": 110
|
| 206 |
},
|
| 207 |
{
|
| 208 |
"epoch": 0.13820904117477684,
|
| 209 |
+
"grad_norm": 22.41253635277838,
|
| 210 |
"learning_rate": 4.978006327248536e-07,
|
| 211 |
+
"logits/chosen": -2.7071425914764404,
|
| 212 |
+
"logits/rejected": -2.6982598304748535,
|
| 213 |
+
"logps/chosen": -415.25634765625,
|
| 214 |
+
"logps/rejected": -507.17376708984375,
|
| 215 |
+
"loss": 0.4829,
|
| 216 |
+
"rewards/accuracies": 0.768750011920929,
|
| 217 |
+
"rewards/chosen": -1.099837303161621,
|
| 218 |
+
"rewards/margins": 0.9606343507766724,
|
| 219 |
+
"rewards/rejected": -2.060471773147583,
|
| 220 |
"step": 120
|
| 221 |
},
|
| 222 |
{
|
| 223 |
"epoch": 0.14972646127267492,
|
| 224 |
+
"grad_norm": 28.3351388035627,
|
| 225 |
"learning_rate": 4.962695471250032e-07,
|
| 226 |
+
"logits/chosen": -2.509568929672241,
|
| 227 |
+
"logits/rejected": -2.4706530570983887,
|
| 228 |
+
"logps/chosen": -475.9378356933594,
|
| 229 |
+
"logps/rejected": -565.5125732421875,
|
| 230 |
+
"loss": 0.4913,
|
| 231 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 232 |
+
"rewards/chosen": -1.896477460861206,
|
| 233 |
+
"rewards/margins": 1.053924798965454,
|
| 234 |
+
"rewards/rejected": -2.950402021408081,
|
| 235 |
"step": 130
|
| 236 |
},
|
| 237 |
{
|
| 238 |
"epoch": 0.161243881370573,
|
| 239 |
+
"grad_norm": 28.57986478281954,
|
| 240 |
"learning_rate": 4.94340033546025e-07,
|
| 241 |
+
"logits/chosen": -1.8837181329727173,
|
| 242 |
+
"logits/rejected": -1.7757914066314697,
|
| 243 |
+
"logps/chosen": -481.201416015625,
|
| 244 |
+
"logps/rejected": -605.1053466796875,
|
| 245 |
+
"loss": 0.4783,
|
| 246 |
+
"rewards/accuracies": 0.753125011920929,
|
| 247 |
+
"rewards/chosen": -2.0652859210968018,
|
| 248 |
+
"rewards/margins": 1.3595540523529053,
|
| 249 |
+
"rewards/rejected": -3.424839735031128,
|
| 250 |
"step": 140
|
| 251 |
},
|
| 252 |
{
|
| 253 |
"epoch": 0.17276130146847107,
|
| 254 |
+
"grad_norm": 19.68624301597443,
|
| 255 |
"learning_rate": 4.920152136576705e-07,
|
| 256 |
+
"logits/chosen": -1.7746684551239014,
|
| 257 |
+
"logits/rejected": -1.6012051105499268,
|
| 258 |
+
"logps/chosen": -478.12646484375,
|
| 259 |
+
"logps/rejected": -569.0491333007812,
|
| 260 |
+
"loss": 0.456,
|
| 261 |
+
"rewards/accuracies": 0.721875011920929,
|
| 262 |
+
"rewards/chosen": -1.7061989307403564,
|
| 263 |
+
"rewards/margins": 1.1741154193878174,
|
| 264 |
+
"rewards/rejected": -2.880314350128174,
|
| 265 |
"step": 150
|
| 266 |
},
|
| 267 |
{
|
| 268 |
"epoch": 0.18427872156636912,
|
| 269 |
+
"grad_norm": 22.179872671918822,
|
| 270 |
"learning_rate": 4.892988486772756e-07,
|
| 271 |
+
"logits/chosen": -1.5278130769729614,
|
| 272 |
+
"logits/rejected": -1.3506165742874146,
|
| 273 |
+
"logps/chosen": -477.80853271484375,
|
| 274 |
+
"logps/rejected": -651.3321533203125,
|
| 275 |
+
"loss": 0.4295,
|
| 276 |
+
"rewards/accuracies": 0.778124988079071,
|
| 277 |
+
"rewards/chosen": -2.0188517570495605,
|
| 278 |
+
"rewards/margins": 1.7111384868621826,
|
| 279 |
+
"rewards/rejected": -3.7299904823303223,
|
| 280 |
"step": 160
|
| 281 |
},
|
| 282 |
{
|
| 283 |
"epoch": 0.1957961416642672,
|
| 284 |
+
"grad_norm": 22.622363623676517,
|
| 285 |
"learning_rate": 4.861953332846629e-07,
|
| 286 |
+
"logits/chosen": -1.2652065753936768,
|
| 287 |
+
"logits/rejected": -1.1065785884857178,
|
| 288 |
+
"logps/chosen": -485.0794372558594,
|
| 289 |
+
"logps/rejected": -644.7564697265625,
|
| 290 |
+
"loss": 0.4356,
|
| 291 |
+
"rewards/accuracies": 0.7406250238418579,
|
| 292 |
+
"rewards/chosen": -2.1487345695495605,
|
| 293 |
+
"rewards/margins": 1.5242230892181396,
|
| 294 |
+
"rewards/rejected": -3.6729576587677,
|
| 295 |
"step": 170
|
| 296 |
},
|
| 297 |
{
|
| 298 |
"epoch": 0.20731356176216528,
|
| 299 |
+
"grad_norm": 26.087556223120203,
|
| 300 |
"learning_rate": 4.827096885121953e-07,
|
| 301 |
+
"logits/chosen": -1.0748474597930908,
|
| 302 |
+
"logits/rejected": -0.7963994145393372,
|
| 303 |
+
"logps/chosen": -554.6792602539062,
|
| 304 |
+
"logps/rejected": -734.0240478515625,
|
| 305 |
+
"loss": 0.4457,
|
| 306 |
+
"rewards/accuracies": 0.75,
|
| 307 |
+
"rewards/chosen": -2.8119685649871826,
|
| 308 |
+
"rewards/margins": 1.7995468378067017,
|
| 309 |
+
"rewards/rejected": -4.611515045166016,
|
| 310 |
"step": 180
|
| 311 |
},
|
| 312 |
{
|
| 313 |
"epoch": 0.21883098186006333,
|
| 314 |
+
"grad_norm": 23.82706622737464,
|
| 315 |
"learning_rate": 4.788475536214821e-07,
|
| 316 |
+
"logits/chosen": -0.7968783974647522,
|
| 317 |
+
"logits/rejected": -0.5444242358207703,
|
| 318 |
+
"logps/chosen": -570.3772583007812,
|
| 319 |
+
"logps/rejected": -743.9293212890625,
|
| 320 |
+
"loss": 0.404,
|
| 321 |
+
"rewards/accuracies": 0.800000011920929,
|
| 322 |
+
"rewards/chosen": -2.7484724521636963,
|
| 323 |
+
"rewards/margins": 1.847896933555603,
|
| 324 |
+
"rewards/rejected": -4.59636926651001,
|
| 325 |
"step": 190
|
| 326 |
},
|
| 327 |
{
|
| 328 |
"epoch": 0.2303484019579614,
|
| 329 |
+
"grad_norm": 22.12633066649394,
|
| 330 |
"learning_rate": 4.746151769798818e-07,
|
| 331 |
+
"logits/chosen": -0.33320680260658264,
|
| 332 |
+
"logits/rejected": 0.0012704581022262573,
|
| 333 |
+
"logps/chosen": -587.2708740234375,
|
| 334 |
+
"logps/rejected": -802.9293212890625,
|
| 335 |
+
"loss": 0.4268,
|
| 336 |
+
"rewards/accuracies": 0.7593749761581421,
|
| 337 |
+
"rewards/chosen": -3.0321972370147705,
|
| 338 |
+
"rewards/margins": 2.202017068862915,
|
| 339 |
+
"rewards/rejected": -5.2342143058776855,
|
| 340 |
"step": 200
|
| 341 |
},
|
| 342 |
{
|
| 343 |
"epoch": 0.2303484019579614,
|
| 344 |
+
"eval_logits/chosen": -0.9790740013122559,
|
| 345 |
+
"eval_logits/rejected": -0.6780607104301453,
|
| 346 |
+
"eval_logps/chosen": -681.6493530273438,
|
| 347 |
+
"eval_logps/rejected": -974.9606323242188,
|
| 348 |
+
"eval_loss": 0.3576536774635315,
|
| 349 |
+
"eval_rewards/accuracies": 0.7976457476615906,
|
| 350 |
+
"eval_rewards/chosen": -3.906813621520996,
|
| 351 |
+
"eval_rewards/margins": 2.9418699741363525,
|
| 352 |
+
"eval_rewards/rejected": -6.848682403564453,
|
| 353 |
+
"eval_runtime": 642.8141,
|
| 354 |
+
"eval_samples_per_second": 11.086,
|
| 355 |
+
"eval_steps_per_second": 0.347,
|
| 356 |
"step": 200
|
| 357 |
},
|
| 358 |
{
|
| 359 |
"epoch": 0.2418658220558595,
|
| 360 |
+
"grad_norm": 28.55770787025379,
|
| 361 |
"learning_rate": 4.7001940595156055e-07,
|
| 362 |
+
"logits/chosen": -0.15793052315711975,
|
| 363 |
+
"logits/rejected": 0.05452694371342659,
|
| 364 |
+
"logps/chosen": -531.8611450195312,
|
| 365 |
+
"logps/rejected": -715.461181640625,
|
| 366 |
+
"loss": 0.4386,
|
| 367 |
+
"rewards/accuracies": 0.731249988079071,
|
| 368 |
+
"rewards/chosen": -2.547398328781128,
|
| 369 |
+
"rewards/margins": 1.7863715887069702,
|
| 370 |
+
"rewards/rejected": -4.333769798278809,
|
| 371 |
"step": 210
|
| 372 |
},
|
| 373 |
{
|
| 374 |
"epoch": 0.25338324215375757,
|
| 375 |
+
"grad_norm": 34.18952768278615,
|
| 376 |
"learning_rate": 4.650676758194623e-07,
|
| 377 |
+
"logits/chosen": -0.5843445062637329,
|
| 378 |
+
"logits/rejected": -0.08485187590122223,
|
| 379 |
+
"logps/chosen": -509.3975524902344,
|
| 380 |
+
"logps/rejected": -702.2587280273438,
|
| 381 |
+
"loss": 0.4169,
|
| 382 |
+
"rewards/accuracies": 0.78125,
|
| 383 |
+
"rewards/chosen": -2.2745628356933594,
|
| 384 |
+
"rewards/margins": 2.145096778869629,
|
| 385 |
+
"rewards/rejected": -4.419659614562988,
|
| 386 |
"step": 220
|
| 387 |
},
|
| 388 |
{
|
| 389 |
"epoch": 0.26490066225165565,
|
| 390 |
+
"grad_norm": 33.36988457876833,
|
| 391 |
"learning_rate": 4.5976799775611215e-07,
|
| 392 |
+
"logits/chosen": -0.3160017728805542,
|
| 393 |
+
"logits/rejected": 0.20093998312950134,
|
| 394 |
+
"logps/chosen": -516.4368896484375,
|
| 395 |
+
"logps/rejected": -678.5955200195312,
|
| 396 |
+
"loss": 0.4413,
|
| 397 |
+
"rewards/accuracies": 0.75,
|
| 398 |
+
"rewards/chosen": -2.2482521533966064,
|
| 399 |
+
"rewards/margins": 1.7269527912139893,
|
| 400 |
+
"rewards/rejected": -3.9752049446105957,
|
| 401 |
"step": 230
|
| 402 |
},
|
| 403 |
{
|
| 404 |
"epoch": 0.2764180823495537,
|
| 405 |
+
"grad_norm": 25.800936315454518,
|
| 406 |
"learning_rate": 4.5412894586271543e-07,
|
| 407 |
+
"logits/chosen": -0.25688791275024414,
|
| 408 |
+
"logits/rejected": 0.23107607662677765,
|
| 409 |
+
"logps/chosen": -507.2642517089844,
|
| 410 |
+
"logps/rejected": -678.391357421875,
|
| 411 |
+
"loss": 0.42,
|
| 412 |
+
"rewards/accuracies": 0.765625,
|
| 413 |
+
"rewards/chosen": -2.0940096378326416,
|
| 414 |
+
"rewards/margins": 1.9648818969726562,
|
| 415 |
+
"rewards/rejected": -4.058891773223877,
|
| 416 |
"step": 240
|
| 417 |
},
|
| 418 |
{
|
| 419 |
"epoch": 0.28793550244745175,
|
| 420 |
+
"grad_norm": 25.786461976756396,
|
| 421 |
"learning_rate": 4.481596432975201e-07,
|
| 422 |
+
"logits/chosen": -1.0171256065368652,
|
| 423 |
+
"logits/rejected": -0.5706368684768677,
|
| 424 |
+
"logps/chosen": -517.12451171875,
|
| 425 |
+
"logps/rejected": -737.7791137695312,
|
| 426 |
+
"loss": 0.4331,
|
| 427 |
+
"rewards/accuracies": 0.768750011920929,
|
| 428 |
+
"rewards/chosen": -2.2817940711975098,
|
| 429 |
+
"rewards/margins": 2.181729793548584,
|
| 430 |
+
"rewards/rejected": -4.463524341583252,
|
| 431 |
"step": 250
|
| 432 |
},
|
| 433 |
{
|
| 434 |
"epoch": 0.29945292254534983,
|
| 435 |
+
"grad_norm": 47.2235188357849,
|
| 436 |
"learning_rate": 4.41869747515886e-07,
|
| 437 |
+
"logits/chosen": -1.1118611097335815,
|
| 438 |
+
"logits/rejected": -0.6688307523727417,
|
| 439 |
+
"logps/chosen": -620.9531860351562,
|
| 440 |
+
"logps/rejected": -872.623046875,
|
| 441 |
+
"loss": 0.4144,
|
| 442 |
+
"rewards/accuracies": 0.753125011920929,
|
| 443 |
+
"rewards/chosen": -3.3720595836639404,
|
| 444 |
+
"rewards/margins": 2.488572359085083,
|
| 445 |
+
"rewards/rejected": -5.860631465911865,
|
| 446 |
"step": 260
|
| 447 |
},
|
| 448 |
{
|
| 449 |
"epoch": 0.3109703426432479,
|
| 450 |
+
"grad_norm": 26.838626915471743,
|
| 451 |
"learning_rate": 4.352694346459396e-07,
|
| 452 |
+
"logits/chosen": -1.2210652828216553,
|
| 453 |
+
"logits/rejected": -0.7195647358894348,
|
| 454 |
+
"logps/chosen": -534.5902099609375,
|
| 455 |
+
"logps/rejected": -761.0294189453125,
|
| 456 |
+
"loss": 0.4083,
|
| 457 |
"rewards/accuracies": 0.784375011920929,
|
| 458 |
+
"rewards/chosen": -2.6057441234588623,
|
| 459 |
+
"rewards/margins": 2.184957504272461,
|
| 460 |
+
"rewards/rejected": -4.790701866149902,
|
| 461 |
"step": 270
|
| 462 |
},
|
| 463 |
{
|
| 464 |
"epoch": 0.322487762741146,
|
| 465 |
+
"grad_norm": 26.363036179443302,
|
| 466 |
"learning_rate": 4.2836938302509256e-07,
|
| 467 |
+
"logits/chosen": -1.2423580884933472,
|
| 468 |
+
"logits/rejected": -0.7629604935646057,
|
| 469 |
+
"logps/chosen": -559.0585327148438,
|
| 470 |
+
"logps/rejected": -827.2657470703125,
|
| 471 |
+
"loss": 0.3904,
|
| 472 |
+
"rewards/accuracies": 0.7718750238418579,
|
| 473 |
+
"rewards/chosen": -2.7791800498962402,
|
| 474 |
+
"rewards/margins": 2.684356689453125,
|
| 475 |
+
"rewards/rejected": -5.463536739349365,
|
| 476 |
"step": 280
|
| 477 |
},
|
| 478 |
{
|
| 479 |
"epoch": 0.33400518283904407,
|
| 480 |
+
"grad_norm": 25.387456130173874,
|
| 481 |
"learning_rate": 4.2118075592405874e-07,
|
| 482 |
+
"logits/chosen": -0.6111725568771362,
|
| 483 |
+
"logits/rejected": -0.15174560248851776,
|
| 484 |
+
"logps/chosen": -587.7760009765625,
|
| 485 |
+
"logps/rejected": -839.3088989257812,
|
| 486 |
+
"loss": 0.4086,
|
| 487 |
+
"rewards/accuracies": 0.8125,
|
| 488 |
+
"rewards/chosen": -2.9200491905212402,
|
| 489 |
+
"rewards/margins": 2.618988037109375,
|
| 490 |
+
"rewards/rejected": -5.539036750793457,
|
| 491 |
"step": 290
|
| 492 |
},
|
| 493 |
{
|
| 494 |
"epoch": 0.34552260293694215,
|
| 495 |
+
"grad_norm": 24.386747936025703,
|
| 496 |
"learning_rate": 4.137151834863213e-07,
|
| 497 |
+
"logits/chosen": -0.8588908910751343,
|
| 498 |
+
"logits/rejected": -0.17885461449623108,
|
| 499 |
+
"logps/chosen": -585.371337890625,
|
| 500 |
+
"logps/rejected": -798.4925537109375,
|
| 501 |
+
"loss": 0.4067,
|
| 502 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 503 |
+
"rewards/chosen": -3.0237374305725098,
|
| 504 |
+
"rewards/margins": 2.431835651397705,
|
| 505 |
+
"rewards/rejected": -5.455573558807373,
|
| 506 |
"step": 300
|
| 507 |
},
|
| 508 |
{
|
| 509 |
"epoch": 0.34552260293694215,
|
| 510 |
+
"eval_logits/chosen": -1.2473794221878052,
|
| 511 |
+
"eval_logits/rejected": -0.6641976833343506,
|
| 512 |
+
"eval_logps/chosen": -688.5350952148438,
|
| 513 |
+
"eval_logps/rejected": -1054.9027099609375,
|
| 514 |
+
"eval_loss": 0.34113186597824097,
|
| 515 |
+
"eval_rewards/accuracies": 0.8094170689582825,
|
| 516 |
+
"eval_rewards/chosen": -3.9756710529327393,
|
| 517 |
+
"eval_rewards/margins": 3.6724324226379395,
|
| 518 |
+
"eval_rewards/rejected": -7.648103713989258,
|
| 519 |
+
"eval_runtime": 641.4555,
|
| 520 |
+
"eval_samples_per_second": 11.109,
|
| 521 |
+
"eval_steps_per_second": 0.348,
|
| 522 |
"step": 300
|
| 523 |
},
|
| 524 |
{
|
| 525 |
"epoch": 0.35704002303484017,
|
| 526 |
+
"grad_norm": 31.743815715159577,
|
| 527 |
"learning_rate": 4.059847439122671e-07,
|
| 528 |
+
"logits/chosen": -0.9840304255485535,
|
| 529 |
+
"logits/rejected": -0.4123767018318176,
|
| 530 |
+
"logps/chosen": -538.8982543945312,
|
| 531 |
+
"logps/rejected": -785.3285522460938,
|
| 532 |
+
"loss": 0.4144,
|
| 533 |
+
"rewards/accuracies": 0.7718750238418579,
|
| 534 |
+
"rewards/chosen": -2.550431489944458,
|
| 535 |
+
"rewards/margins": 2.4609429836273193,
|
| 536 |
+
"rewards/rejected": -5.011374473571777,
|
| 537 |
"step": 310
|
| 538 |
},
|
| 539 |
{
|
| 540 |
"epoch": 0.36855744313273825,
|
| 541 |
+
"grad_norm": 24.806762291664178,
|
| 542 |
"learning_rate": 3.98001943918432e-07,
|
| 543 |
+
"logits/chosen": -1.2074978351593018,
|
| 544 |
+
"logits/rejected": -0.2775883674621582,
|
| 545 |
+
"logps/chosen": -509.53887939453125,
|
| 546 |
+
"logps/rejected": -702.5384521484375,
|
| 547 |
+
"loss": 0.4077,
|
| 548 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 549 |
+
"rewards/chosen": -2.1766436100006104,
|
| 550 |
+
"rewards/margins": 2.217984914779663,
|
| 551 |
+
"rewards/rejected": -4.394628524780273,
|
| 552 |
"step": 320
|
| 553 |
},
|
| 554 |
{
|
| 555 |
"epoch": 0.38007486323063633,
|
| 556 |
+
"grad_norm": 31.303160160987616,
|
| 557 |
"learning_rate": 3.8977969850346866e-07,
|
| 558 |
+
"logits/chosen": -0.7918741106987,
|
| 559 |
+
"logits/rejected": 0.004956415388733149,
|
| 560 |
+
"logps/chosen": -578.205078125,
|
| 561 |
+
"logps/rejected": -870.2096557617188,
|
| 562 |
+
"loss": 0.3752,
|
| 563 |
+
"rewards/accuracies": 0.765625,
|
| 564 |
+
"rewards/chosen": -2.821563720703125,
|
| 565 |
+
"rewards/margins": 3.012451648712158,
|
| 566 |
+
"rewards/rejected": -5.834015369415283,
|
| 567 |
"step": 330
|
| 568 |
},
|
| 569 |
{
|
| 570 |
"epoch": 0.3915922833285344,
|
| 571 |
+
"grad_norm": 27.692313832407383,
|
| 572 |
"learning_rate": 3.8133131005357465e-07,
|
| 573 |
+
"logits/chosen": -0.9087894558906555,
|
| 574 |
+
"logits/rejected": -0.1920526772737503,
|
| 575 |
+
"logps/chosen": -602.9824829101562,
|
| 576 |
+
"logps/rejected": -872.3701171875,
|
| 577 |
+
"loss": 0.4233,
|
| 578 |
+
"rewards/accuracies": 0.7906249761581421,
|
| 579 |
+
"rewards/chosen": -3.1107020378112793,
|
| 580 |
+
"rewards/margins": 2.777942180633545,
|
| 581 |
+
"rewards/rejected": -5.888644218444824,
|
| 582 |
"step": 340
|
| 583 |
},
|
| 584 |
{
|
| 585 |
"epoch": 0.4031097034264325,
|
| 586 |
+
"grad_norm": 22.459786517410773,
|
| 587 |
"learning_rate": 3.7267044682118435e-07,
|
| 588 |
+
"logits/chosen": -0.6368467211723328,
|
| 589 |
+
"logits/rejected": 0.05047903582453728,
|
| 590 |
+
"logps/chosen": -546.3031616210938,
|
| 591 |
+
"logps/rejected": -768.5010986328125,
|
| 592 |
+
"loss": 0.4038,
|
| 593 |
+
"rewards/accuracies": 0.753125011920929,
|
| 594 |
+
"rewards/chosen": -2.607321262359619,
|
| 595 |
+
"rewards/margins": 2.1534266471862793,
|
| 596 |
+
"rewards/rejected": -4.76074743270874,
|
| 597 |
"step": 350
|
| 598 |
},
|
| 599 |
{
|
| 600 |
"epoch": 0.41462712352433057,
|
| 601 |
+
"grad_norm": 54.30836966222414,
|
| 602 |
"learning_rate": 3.638111208117425e-07,
|
| 603 |
+
"logits/chosen": -0.4448986053466797,
|
| 604 |
+
"logits/rejected": 0.26401039958000183,
|
| 605 |
+
"logps/chosen": -660.1737060546875,
|
| 606 |
+
"logps/rejected": -936.7528076171875,
|
| 607 |
+
"loss": 0.3945,
|
| 608 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 609 |
+
"rewards/chosen": -3.779724597930908,
|
| 610 |
+
"rewards/margins": 2.7942070960998535,
|
| 611 |
+
"rewards/rejected": -6.573931694030762,
|
| 612 |
"step": 360
|
| 613 |
},
|
| 614 |
{
|
| 615 |
"epoch": 0.42614454362222864,
|
| 616 |
+
"grad_norm": 30.943136192122527,
|
| 617 |
"learning_rate": 3.5476766511433605e-07,
|
| 618 |
+
"logits/chosen": -0.6570574641227722,
|
| 619 |
+
"logits/rejected": 0.12815245985984802,
|
| 620 |
+
"logps/chosen": -589.9466552734375,
|
| 621 |
+
"logps/rejected": -808.5029296875,
|
| 622 |
+
"loss": 0.4008,
|
| 623 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 624 |
+
"rewards/chosen": -3.0944387912750244,
|
| 625 |
+
"rewards/margins": 2.3009159564971924,
|
| 626 |
+
"rewards/rejected": -5.395354270935059,
|
| 627 |
"step": 370
|
| 628 |
},
|
| 629 |
{
|
| 630 |
"epoch": 0.43766196372012667,
|
| 631 |
+
"grad_norm": 19.09835141181145,
|
| 632 |
"learning_rate": 3.455547107128602e-07,
|
| 633 |
+
"logits/chosen": -0.45787113904953003,
|
| 634 |
+
"logits/rejected": 0.38415655493736267,
|
| 635 |
+
"logps/chosen": -527.8428955078125,
|
| 636 |
+
"logps/rejected": -821.2969970703125,
|
| 637 |
+
"loss": 0.3953,
|
| 638 |
+
"rewards/accuracies": 0.8125,
|
| 639 |
+
"rewards/chosen": -2.6933910846710205,
|
| 640 |
+
"rewards/margins": 2.8653125762939453,
|
| 641 |
+
"rewards/rejected": -5.558703422546387,
|
| 642 |
"step": 380
|
| 643 |
},
|
| 644 |
{
|
| 645 |
"epoch": 0.44917938381802475,
|
| 646 |
+
"grad_norm": 25.986133062638014,
|
| 647 |
"learning_rate": 3.361871628152338e-07,
|
| 648 |
+
"logits/chosen": 0.1055336743593216,
|
| 649 |
+
"logits/rejected": 0.8421304821968079,
|
| 650 |
+
"logps/chosen": -574.3500366210938,
|
| 651 |
+
"logps/rejected": -818.9120483398438,
|
| 652 |
+
"loss": 0.4116,
|
| 653 |
+
"rewards/accuracies": 0.800000011920929,
|
| 654 |
+
"rewards/chosen": -2.960313558578491,
|
| 655 |
+
"rewards/margins": 2.6313090324401855,
|
| 656 |
+
"rewards/rejected": -5.591622829437256,
|
| 657 |
"step": 390
|
| 658 |
},
|
| 659 |
{
|
| 660 |
"epoch": 0.4606968039159228,
|
| 661 |
+
"grad_norm": 33.121972667621534,
|
| 662 |
"learning_rate": 3.2668017673896077e-07,
|
| 663 |
+
"logits/chosen": -0.19368359446525574,
|
| 664 |
+
"logits/rejected": 0.5146237015724182,
|
| 665 |
+
"logps/chosen": -617.7755737304688,
|
| 666 |
+
"logps/rejected": -860.82861328125,
|
| 667 |
+
"loss": 0.4011,
|
| 668 |
+
"rewards/accuracies": 0.762499988079071,
|
| 669 |
+
"rewards/chosen": -3.3203601837158203,
|
| 670 |
+
"rewards/margins": 2.387831211090088,
|
| 671 |
+
"rewards/rejected": -5.708191871643066,
|
| 672 |
"step": 400
|
| 673 |
},
|
| 674 |
{
|
| 675 |
"epoch": 0.4606968039159228,
|
| 676 |
+
"eval_logits/chosen": -0.7429019808769226,
|
| 677 |
+
"eval_logits/rejected": 0.1183277890086174,
|
| 678 |
+
"eval_logps/chosen": -735.4549560546875,
|
| 679 |
+
"eval_logps/rejected": -1130.1990966796875,
|
| 680 |
+
"eval_loss": 0.32945430278778076,
|
| 681 |
+
"eval_rewards/accuracies": 0.8155829310417175,
|
| 682 |
+
"eval_rewards/chosen": -4.444869518280029,
|
| 683 |
+
"eval_rewards/margins": 3.9561986923217773,
|
| 684 |
+
"eval_rewards/rejected": -8.401067733764648,
|
| 685 |
+
"eval_runtime": 645.3491,
|
| 686 |
+
"eval_samples_per_second": 11.042,
|
| 687 |
+
"eval_steps_per_second": 0.346,
|
| 688 |
"step": 400
|
| 689 |
},
|
| 690 |
{
|
| 691 |
"epoch": 0.4722142240138209,
|
| 692 |
+
"grad_norm": 28.677116827222893,
|
| 693 |
"learning_rate": 3.1704913339205103e-07,
|
| 694 |
+
"logits/chosen": -0.49749231338500977,
|
| 695 |
+
"logits/rejected": 0.5208367109298706,
|
| 696 |
+
"logps/chosen": -578.421875,
|
| 697 |
+
"logps/rejected": -866.33544921875,
|
| 698 |
+
"loss": 0.4082,
|
| 699 |
+
"rewards/accuracies": 0.8031250238418579,
|
| 700 |
+
"rewards/chosen": -2.9447290897369385,
|
| 701 |
+
"rewards/margins": 3.001622438430786,
|
| 702 |
+
"rewards/rejected": -5.946351528167725,
|
| 703 |
"step": 410
|
| 704 |
},
|
| 705 |
{
|
| 706 |
"epoch": 0.483731644111719,
|
| 707 |
+
"grad_norm": 25.93977899429838,
|
| 708 |
"learning_rate": 3.0730961438896885e-07,
|
| 709 |
+
"logits/chosen": -1.0381872653961182,
|
| 710 |
+
"logits/rejected": -0.06657940149307251,
|
| 711 |
+
"logps/chosen": -550.3687744140625,
|
| 712 |
+
"logps/rejected": -820.9139404296875,
|
| 713 |
+
"loss": 0.4,
|
| 714 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 715 |
+
"rewards/chosen": -2.6255903244018555,
|
| 716 |
+
"rewards/margins": 2.71159029006958,
|
| 717 |
+
"rewards/rejected": -5.3371806144714355,
|
| 718 |
"step": 420
|
| 719 |
},
|
| 720 |
{
|
| 721 |
"epoch": 0.49524906420961706,
|
| 722 |
+
"grad_norm": 24.865636277888157,
|
| 723 |
"learning_rate": 2.9747737684186795e-07,
|
| 724 |
+
"logits/chosen": -1.131136417388916,
|
| 725 |
+
"logits/rejected": -0.22055275738239288,
|
| 726 |
+
"logps/chosen": -534.1951293945312,
|
| 727 |
+
"logps/rejected": -773.7057495117188,
|
| 728 |
+
"loss": 0.3973,
|
| 729 |
+
"rewards/accuracies": 0.793749988079071,
|
| 730 |
+
"rewards/chosen": -2.4572598934173584,
|
| 731 |
+
"rewards/margins": 2.522524356842041,
|
| 732 |
+
"rewards/rejected": -4.9797844886779785,
|
| 733 |
"step": 430
|
| 734 |
},
|
| 735 |
{
|
| 736 |
"epoch": 0.5067664843075151,
|
| 737 |
+
"grad_norm": 36.423518981463275,
|
| 738 |
"learning_rate": 2.8756832786789663e-07,
|
| 739 |
+
"logits/chosen": -1.2586928606033325,
|
| 740 |
+
"logits/rejected": -0.14189645648002625,
|
| 741 |
+
"logps/chosen": -513.0911254882812,
|
| 742 |
+
"logps/rejected": -821.0222778320312,
|
| 743 |
+
"loss": 0.4053,
|
| 744 |
+
"rewards/accuracies": 0.815625011920929,
|
| 745 |
+
"rewards/chosen": -2.5397918224334717,
|
| 746 |
+
"rewards/margins": 3.0112507343292236,
|
| 747 |
+
"rewards/rejected": -5.551042556762695,
|
| 748 |
"step": 440
|
| 749 |
},
|
| 750 |
{
|
| 751 |
"epoch": 0.5182839044054132,
|
| 752 |
+
"grad_norm": 27.76293529869272,
|
| 753 |
"learning_rate": 2.7759849885381747e-07,
|
| 754 |
+
"logits/chosen": -1.2538650035858154,
|
| 755 |
+
"logits/rejected": -0.25470593571662903,
|
| 756 |
+
"logps/chosen": -517.45751953125,
|
| 757 |
+
"logps/rejected": -752.2701416015625,
|
| 758 |
+
"loss": 0.3845,
|
| 759 |
+
"rewards/accuracies": 0.793749988079071,
|
| 760 |
+
"rewards/chosen": -2.279754877090454,
|
| 761 |
+
"rewards/margins": 2.4887535572052,
|
| 762 |
+
"rewards/rejected": -4.768507957458496,
|
| 763 |
"step": 450
|
| 764 |
},
|
| 765 |
{
|
| 766 |
"epoch": 0.5298013245033113,
|
| 767 |
+
"grad_norm": 24.085341189256496,
|
| 768 |
"learning_rate": 2.675840195195762e-07,
|
| 769 |
+
"logits/chosen": -1.382927417755127,
|
| 770 |
+
"logits/rejected": -0.48569250106811523,
|
| 771 |
+
"logps/chosen": -513.8163452148438,
|
| 772 |
+
"logps/rejected": -766.2885131835938,
|
| 773 |
+
"loss": 0.3799,
|
| 774 |
+
"rewards/accuracies": 0.793749988079071,
|
| 775 |
+
"rewards/chosen": -2.302953004837036,
|
| 776 |
+
"rewards/margins": 2.5440595149993896,
|
| 777 |
+
"rewards/rejected": -4.847012519836426,
|
| 778 |
"step": 460
|
| 779 |
},
|
| 780 |
{
|
| 781 |
"epoch": 0.5413187446012093,
|
| 782 |
+
"grad_norm": 29.101645338578166,
|
| 783 |
"learning_rate": 2.575410918227829e-07,
|
| 784 |
+
"logits/chosen": -0.9476648569107056,
|
| 785 |
+
"logits/rejected": -0.29435592889785767,
|
| 786 |
+
"logps/chosen": -573.229736328125,
|
| 787 |
+
"logps/rejected": -794.494140625,
|
| 788 |
+
"loss": 0.3896,
|
| 789 |
+
"rewards/accuracies": 0.71875,
|
| 790 |
+
"rewards/chosen": -2.8009214401245117,
|
| 791 |
+
"rewards/margins": 2.2908072471618652,
|
| 792 |
+
"rewards/rejected": -5.091729164123535,
|
| 793 |
"step": 470
|
| 794 |
},
|
| 795 |
{
|
| 796 |
"epoch": 0.5528361646991073,
|
| 797 |
+
"grad_norm": 30.198868045523568,
|
| 798 |
"learning_rate": 2.474859637463226e-07,
|
| 799 |
+
"logits/chosen": -0.4918448328971863,
|
| 800 |
+
"logits/rejected": 0.3972582221031189,
|
| 801 |
+
"logps/chosen": -556.0670776367188,
|
| 802 |
+
"logps/rejected": -804.6024169921875,
|
| 803 |
+
"loss": 0.3886,
|
| 804 |
+
"rewards/accuracies": 0.8218749761581421,
|
| 805 |
+
"rewards/chosen": -2.8543224334716797,
|
| 806 |
+
"rewards/margins": 2.5813965797424316,
|
| 807 |
+
"rewards/rejected": -5.435718536376953,
|
| 808 |
"step": 480
|
| 809 |
},
|
| 810 |
{
|
| 811 |
"epoch": 0.5643535847970055,
|
| 812 |
+
"grad_norm": 28.52494731969291,
|
| 813 |
"learning_rate": 2.3743490301150355e-07,
|
| 814 |
+
"logits/chosen": -0.5242654085159302,
|
| 815 |
+
"logits/rejected": 0.405693382024765,
|
| 816 |
+
"logps/chosen": -570.1974487304688,
|
| 817 |
+
"logps/rejected": -811.2410278320312,
|
| 818 |
+
"loss": 0.3959,
|
| 819 |
+
"rewards/accuracies": 0.778124988079071,
|
| 820 |
+
"rewards/chosen": -2.862198829650879,
|
| 821 |
+
"rewards/margins": 2.4608964920043945,
|
| 822 |
+
"rewards/rejected": -5.323095321655273,
|
| 823 |
"step": 490
|
| 824 |
},
|
| 825 |
{
|
| 826 |
"epoch": 0.5758710048949035,
|
| 827 |
+
"grad_norm": 27.925871853424944,
|
| 828 |
"learning_rate": 2.274041707592724e-07,
|
| 829 |
+
"logits/chosen": -1.0720345973968506,
|
| 830 |
+
"logits/rejected": -0.10801704227924347,
|
| 831 |
+
"logps/chosen": -542.1129150390625,
|
| 832 |
+
"logps/rejected": -843.2058715820312,
|
| 833 |
+
"loss": 0.3727,
|
| 834 |
+
"rewards/accuracies": 0.800000011920929,
|
| 835 |
+
"rewards/chosen": -2.6700572967529297,
|
| 836 |
+
"rewards/margins": 2.9847068786621094,
|
| 837 |
+
"rewards/rejected": -5.654764175415039,
|
| 838 |
"step": 500
|
| 839 |
},
|
| 840 |
{
|
| 841 |
"epoch": 0.5758710048949035,
|
| 842 |
+
"eval_logits/chosen": -1.3008555173873901,
|
| 843 |
+
"eval_logits/rejected": -0.40656155347824097,
|
| 844 |
+
"eval_logps/chosen": -662.9987182617188,
|
| 845 |
+
"eval_logps/rejected": -1055.4913330078125,
|
| 846 |
+
"eval_loss": 0.3260224759578705,
|
| 847 |
+
"eval_rewards/accuracies": 0.8161435127258301,
|
| 848 |
+
"eval_rewards/chosen": -3.720306634902954,
|
| 849 |
+
"eval_rewards/margins": 3.9336841106414795,
|
| 850 |
+
"eval_rewards/rejected": -7.653989791870117,
|
| 851 |
+
"eval_runtime": 644.1686,
|
| 852 |
+
"eval_samples_per_second": 11.062,
|
| 853 |
+
"eval_steps_per_second": 0.346,
|
| 854 |
"step": 500
|
| 855 |
},
|
| 856 |
{
|
| 857 |
"epoch": 0.5873884249928016,
|
| 858 |
+
"grad_norm": 23.857944610065903,
|
| 859 |
"learning_rate": 2.17409995242075e-07,
|
| 860 |
+
"logits/chosen": -0.8062444925308228,
|
| 861 |
+
"logits/rejected": 0.046298883855342865,
|
| 862 |
+
"logps/chosen": -568.9137573242188,
|
| 863 |
+
"logps/rejected": -859.3053588867188,
|
| 864 |
+
"loss": 0.3716,
|
| 865 |
+
"rewards/accuracies": 0.7906249761581421,
|
| 866 |
+
"rewards/chosen": -2.9011380672454834,
|
| 867 |
+
"rewards/margins": 3.026390552520752,
|
| 868 |
+
"rewards/rejected": -5.927529335021973,
|
| 869 |
"step": 510
|
| 870 |
},
|
| 871 |
{
|
| 872 |
"epoch": 0.5989058450906997,
|
| 873 |
+
"grad_norm": 34.73513393097343,
|
| 874 |
"learning_rate": 2.0746854556892544e-07,
|
| 875 |
+
"logits/chosen": -0.23699383437633514,
|
| 876 |
+
"logits/rejected": 0.7075563073158264,
|
| 877 |
+
"logps/chosen": -630.9346313476562,
|
| 878 |
+
"logps/rejected": -914.3643798828125,
|
| 879 |
+
"loss": 0.3669,
|
| 880 |
"rewards/accuracies": 0.8187500238418579,
|
| 881 |
+
"rewards/chosen": -3.393051862716675,
|
| 882 |
+
"rewards/margins": 2.9863784313201904,
|
| 883 |
+
"rewards/rejected": -6.379430294036865,
|
| 884 |
"step": 520
|
| 885 |
},
|
| 886 |
{
|
| 887 |
"epoch": 0.6104232651885978,
|
| 888 |
+
"grad_norm": 31.950736125786133,
|
| 889 |
"learning_rate": 1.9759590554616173e-07,
|
| 890 |
+
"logits/chosen": -0.5588125586509705,
|
| 891 |
+
"logits/rejected": 0.502726674079895,
|
| 892 |
+
"logps/chosen": -579.3170776367188,
|
| 893 |
+
"logps/rejected": -851.6565551757812,
|
| 894 |
+
"loss": 0.3844,
|
| 895 |
+
"rewards/accuracies": 0.78125,
|
| 896 |
+
"rewards/chosen": -3.0607731342315674,
|
| 897 |
+
"rewards/margins": 2.8013463020324707,
|
| 898 |
+
"rewards/rejected": -5.862119674682617,
|
| 899 |
"step": 530
|
| 900 |
},
|
| 901 |
{
|
| 902 |
"epoch": 0.6219406852864958,
|
| 903 |
+
"grad_norm": 31.535492973313232,
|
| 904 |
"learning_rate": 1.8780804765620746e-07,
|
| 905 |
+
"logits/chosen": -0.7882386445999146,
|
| 906 |
+
"logits/rejected": 0.33689114451408386,
|
| 907 |
+
"logps/chosen": -505.6971740722656,
|
| 908 |
+
"logps/rejected": -770.9715576171875,
|
| 909 |
+
"loss": 0.3714,
|
| 910 |
+
"rewards/accuracies": 0.8125,
|
| 911 |
+
"rewards/chosen": -2.2458343505859375,
|
| 912 |
+
"rewards/margins": 2.7756423950195312,
|
| 913 |
+
"rewards/rejected": -5.021476745605469,
|
| 914 |
"step": 540
|
| 915 |
},
|
| 916 |
{
|
| 917 |
"epoch": 0.6334581053843938,
|
| 918 |
+
"grad_norm": 24.518899300596807,
|
| 919 |
"learning_rate": 1.7812080721643973e-07,
|
| 920 |
+
"logits/chosen": -0.6544414758682251,
|
| 921 |
+
"logits/rejected": 0.6189062595367432,
|
| 922 |
+
"logps/chosen": -587.9449462890625,
|
| 923 |
+
"logps/rejected": -899.134765625,
|
| 924 |
+
"loss": 0.367,
|
| 925 |
+
"rewards/accuracies": 0.828125,
|
| 926 |
+
"rewards/chosen": -2.8350448608398438,
|
| 927 |
+
"rewards/margins": 3.2712531089782715,
|
| 928 |
+
"rewards/rejected": -6.106298446655273,
|
| 929 |
"step": 550
|
| 930 |
},
|
| 931 |
{
|
| 932 |
"epoch": 0.644975525482292,
|
| 933 |
+
"grad_norm": 30.6763264648463,
|
| 934 |
"learning_rate": 1.6854985675997063e-07,
|
| 935 |
+
"logits/chosen": -0.7213582396507263,
|
| 936 |
+
"logits/rejected": 0.4497779905796051,
|
| 937 |
+
"logps/chosen": -577.9434814453125,
|
| 938 |
+
"logps/rejected": -847.3308715820312,
|
| 939 |
+
"loss": 0.3895,
|
| 940 |
+
"rewards/accuracies": 0.800000011920929,
|
| 941 |
+
"rewards/chosen": -2.8931756019592285,
|
| 942 |
+
"rewards/margins": 2.903951644897461,
|
| 943 |
+
"rewards/rejected": -5.797126293182373,
|
| 944 |
"step": 560
|
| 945 |
},
|
| 946 |
{
|
| 947 |
"epoch": 0.65649294558019,
|
| 948 |
+
"grad_norm": 25.945256821159443,
|
| 949 |
"learning_rate": 1.5911068067978818e-07,
|
| 950 |
+
"logits/chosen": -0.48560142517089844,
|
| 951 |
+
"logits/rejected": 0.6249833106994629,
|
| 952 |
+
"logps/chosen": -559.338623046875,
|
| 953 |
+
"logps/rejected": -765.5469970703125,
|
| 954 |
+
"loss": 0.4061,
|
| 955 |
+
"rewards/accuracies": 0.737500011920929,
|
| 956 |
+
"rewards/chosen": -2.6765408515930176,
|
| 957 |
+
"rewards/margins": 2.3505935668945312,
|
| 958 |
+
"rewards/rejected": -5.027134895324707,
|
| 959 |
"step": 570
|
| 960 |
},
|
| 961 |
{
|
| 962 |
"epoch": 0.6680103656780881,
|
| 963 |
+
"grad_norm": 27.330253535761116,
|
| 964 |
"learning_rate": 1.4981855017728197e-07,
|
| 965 |
+
"logits/chosen": -0.3748374879360199,
|
| 966 |
+
"logits/rejected": 0.31157660484313965,
|
| 967 |
+
"logps/chosen": -530.5482177734375,
|
| 968 |
+
"logps/rejected": -811.3858032226562,
|
| 969 |
+
"loss": 0.3952,
|
| 970 |
+
"rewards/accuracies": 0.778124988079071,
|
| 971 |
+
"rewards/chosen": -2.6320903301239014,
|
| 972 |
+
"rewards/margins": 2.578404188156128,
|
| 973 |
+
"rewards/rejected": -5.2104949951171875,
|
| 974 |
"step": 580
|
| 975 |
},
|
| 976 |
{
|
| 977 |
"epoch": 0.6795277857759862,
|
| 978 |
+
"grad_norm": 31.016503489930162,
|
| 979 |
"learning_rate": 1.406884985556804e-07,
|
| 980 |
+
"logits/chosen": -0.366117388010025,
|
| 981 |
+
"logits/rejected": 0.6988335251808167,
|
| 982 |
+
"logps/chosen": -560.417236328125,
|
| 983 |
+
"logps/rejected": -811.5693359375,
|
| 984 |
+
"loss": 0.3842,
|
| 985 |
+
"rewards/accuracies": 0.7593749761581421,
|
| 986 |
+
"rewards/chosen": -2.7148842811584473,
|
| 987 |
+
"rewards/margins": 2.6223959922790527,
|
| 988 |
+
"rewards/rejected": -5.337280750274658,
|
| 989 |
"step": 590
|
| 990 |
},
|
| 991 |
{
|
| 992 |
"epoch": 0.6910452058738843,
|
| 993 |
+
"grad_norm": 33.49108857296992,
|
| 994 |
"learning_rate": 1.3173529689837354e-07,
|
| 995 |
+
"logits/chosen": -0.42413753271102905,
|
| 996 |
+
"logits/rejected": 0.8382568359375,
|
| 997 |
+
"logps/chosen": -592.3516845703125,
|
| 998 |
+
"logps/rejected": -904.6168212890625,
|
| 999 |
+
"loss": 0.3933,
|
| 1000 |
+
"rewards/accuracies": 0.800000011920929,
|
| 1001 |
+
"rewards/chosen": -2.979618549346924,
|
| 1002 |
+
"rewards/margins": 3.4002768993377686,
|
| 1003 |
+
"rewards/rejected": -6.3798956871032715,
|
| 1004 |
"step": 600
|
| 1005 |
},
|
| 1006 |
{
|
| 1007 |
"epoch": 0.6910452058738843,
|
| 1008 |
+
"eval_logits/chosen": -0.5819162130355835,
|
| 1009 |
+
"eval_logits/rejected": 0.32466375827789307,
|
| 1010 |
+
"eval_logps/chosen": -664.277587890625,
|
| 1011 |
+
"eval_logps/rejected": -1041.9088134765625,
|
| 1012 |
+
"eval_loss": 0.3190486431121826,
|
| 1013 |
+
"eval_rewards/accuracies": 0.8256726264953613,
|
| 1014 |
+
"eval_rewards/chosen": -3.733095169067383,
|
| 1015 |
+
"eval_rewards/margins": 3.7850706577301025,
|
| 1016 |
+
"eval_rewards/rejected": -7.5181660652160645,
|
| 1017 |
+
"eval_runtime": 642.0739,
|
| 1018 |
+
"eval_samples_per_second": 11.098,
|
| 1019 |
+
"eval_steps_per_second": 0.347,
|
| 1020 |
"step": 600
|
| 1021 |
},
|
| 1022 |
{
|
| 1023 |
"epoch": 0.7025626259717823,
|
| 1024 |
+
"grad_norm": 33.29931962582806,
|
| 1025 |
"learning_rate": 1.2297343017146726e-07,
|
| 1026 |
+
"logits/chosen": -0.12166979163885117,
|
| 1027 |
+
"logits/rejected": 0.9423264265060425,
|
| 1028 |
+
"logps/chosen": -573.4078979492188,
|
| 1029 |
+
"logps/rejected": -842.9945068359375,
|
| 1030 |
+
"loss": 0.38,
|
| 1031 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 1032 |
+
"rewards/chosen": -2.8864645957946777,
|
| 1033 |
+
"rewards/margins": 2.795393705368042,
|
| 1034 |
+
"rewards/rejected": -5.681858539581299,
|
| 1035 |
"step": 610
|
| 1036 |
},
|
| 1037 |
{
|
| 1038 |
"epoch": 0.7140800460696803,
|
| 1039 |
+
"grad_norm": 27.906297465558936,
|
| 1040 |
"learning_rate": 1.1441707378923474e-07,
|
| 1041 |
+
"logits/chosen": 0.07007602602243423,
|
| 1042 |
+
"logits/rejected": 0.9558390378952026,
|
| 1043 |
+
"logps/chosen": -573.7945556640625,
|
| 1044 |
+
"logps/rejected": -877.9972534179688,
|
| 1045 |
+
"loss": 0.4068,
|
| 1046 |
+
"rewards/accuracies": 0.8031250238418579,
|
| 1047 |
+
"rewards/chosen": -2.903010368347168,
|
| 1048 |
+
"rewards/margins": 3.0875983238220215,
|
| 1049 |
+
"rewards/rejected": -5.990609169006348,
|
| 1050 |
"step": 620
|
| 1051 |
},
|
| 1052 |
{
|
| 1053 |
"epoch": 0.7255974661675785,
|
| 1054 |
+
"grad_norm": 22.67579385883027,
|
| 1055 |
"learning_rate": 1.06080070680377e-07,
|
| 1056 |
+
"logits/chosen": 0.022813748568296432,
|
| 1057 |
+
"logits/rejected": 0.8800600171089172,
|
| 1058 |
+
"logps/chosen": -573.94482421875,
|
| 1059 |
+
"logps/rejected": -831.4425048828125,
|
| 1060 |
+
"loss": 0.3693,
|
| 1061 |
+
"rewards/accuracies": 0.765625,
|
| 1062 |
+
"rewards/chosen": -2.894199848175049,
|
| 1063 |
+
"rewards/margins": 2.732788324356079,
|
| 1064 |
+
"rewards/rejected": -5.626988410949707,
|
| 1065 |
"step": 630
|
| 1066 |
},
|
| 1067 |
{
|
| 1068 |
"epoch": 0.7371148862654765,
|
| 1069 |
+
"grad_norm": 24.736748332046908,
|
| 1070 |
"learning_rate": 9.797590889219587e-08,
|
| 1071 |
+
"logits/chosen": 0.08141092956066132,
|
| 1072 |
+
"logits/rejected": 0.9846137166023254,
|
| 1073 |
+
"logps/chosen": -573.6690673828125,
|
| 1074 |
+
"logps/rejected": -854.0896606445312,
|
| 1075 |
+
"loss": 0.3694,
|
| 1076 |
+
"rewards/accuracies": 0.7718750238418579,
|
| 1077 |
+
"rewards/chosen": -3.009357213973999,
|
| 1078 |
+
"rewards/margins": 2.850132465362549,
|
| 1079 |
+
"rewards/rejected": -5.859489440917969,
|
| 1080 |
"step": 640
|
| 1081 |
},
|
| 1082 |
{
|
| 1083 |
"epoch": 0.7486323063633746,
|
| 1084 |
+
"grad_norm": 19.955009545836916,
|
| 1085 |
"learning_rate": 9.011769976891367e-08,
|
| 1086 |
+
"logits/chosen": 0.18348722159862518,
|
| 1087 |
+
"logits/rejected": 1.4492700099945068,
|
| 1088 |
+
"logps/chosen": -584.7655029296875,
|
| 1089 |
+
"logps/rejected": -836.3447265625,
|
| 1090 |
+
"loss": 0.3854,
|
| 1091 |
+
"rewards/accuracies": 0.809374988079071,
|
| 1092 |
+
"rewards/chosen": -2.9547104835510254,
|
| 1093 |
+
"rewards/margins": 2.7604758739471436,
|
| 1094 |
+
"rewards/rejected": -5.71518611907959,
|
| 1095 |
"step": 650
|
| 1096 |
},
|
| 1097 |
{
|
| 1098 |
"epoch": 0.7601497264612727,
|
| 1099 |
+
"grad_norm": 30.274437174678905,
|
| 1100 |
"learning_rate": 8.251815673944218e-08,
|
| 1101 |
+
"logits/chosen": -0.350322425365448,
|
| 1102 |
+
"logits/rejected": 0.7839977741241455,
|
| 1103 |
+
"logps/chosen": -641.0626220703125,
|
| 1104 |
+
"logps/rejected": -957.6082763671875,
|
| 1105 |
+
"loss": 0.3594,
|
| 1106 |
+
"rewards/accuracies": 0.8218749761581421,
|
| 1107 |
+
"rewards/chosen": -3.33996844291687,
|
| 1108 |
+
"rewards/margins": 3.434263229370117,
|
| 1109 |
+
"rewards/rejected": -6.774231910705566,
|
| 1110 |
"step": 660
|
| 1111 |
},
|
| 1112 |
{
|
| 1113 |
"epoch": 0.7716671465591708,
|
| 1114 |
+
"grad_norm": 27.553152868967715,
|
| 1115 |
"learning_rate": 7.518957474892148e-08,
|
| 1116 |
+
"logits/chosen": -0.08678195625543594,
|
| 1117 |
+
"logits/rejected": 0.7337055206298828,
|
| 1118 |
+
"logps/chosen": -591.7120971679688,
|
| 1119 |
+
"logps/rejected": -890.9392700195312,
|
| 1120 |
+
"loss": 0.3827,
|
| 1121 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 1122 |
+
"rewards/chosen": -2.974543333053589,
|
| 1123 |
+
"rewards/margins": 3.0520572662353516,
|
| 1124 |
+
"rewards/rejected": -6.0266008377075195,
|
| 1125 |
"step": 670
|
| 1126 |
},
|
| 1127 |
{
|
| 1128 |
"epoch": 0.7831845666570688,
|
| 1129 |
+
"grad_norm": 30.232454471420272,
|
| 1130 |
"learning_rate": 6.814381036730274e-08,
|
| 1131 |
+
"logits/chosen": -0.2323370724916458,
|
| 1132 |
+
"logits/rejected": 0.9590624570846558,
|
| 1133 |
+
"logps/chosen": -585.7045288085938,
|
| 1134 |
+
"logps/rejected": -862.7799072265625,
|
| 1135 |
+
"loss": 0.4032,
|
| 1136 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 1137 |
+
"rewards/chosen": -3.0005805492401123,
|
| 1138 |
+
"rewards/margins": 2.9160284996032715,
|
| 1139 |
+
"rewards/rejected": -5.916609287261963,
|
| 1140 |
"step": 680
|
| 1141 |
},
|
| 1142 |
{
|
| 1143 |
"epoch": 0.7947019867549668,
|
| 1144 |
+
"grad_norm": 28.297784072359498,
|
| 1145 |
"learning_rate": 6.139226260715872e-08,
|
| 1146 |
+
"logits/chosen": -0.23143115639686584,
|
| 1147 |
+
"logits/rejected": 0.6265262365341187,
|
| 1148 |
+
"logps/chosen": -598.2958374023438,
|
| 1149 |
+
"logps/rejected": -911.5656127929688,
|
| 1150 |
+
"loss": 0.3747,
|
| 1151 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 1152 |
+
"rewards/chosen": -3.1650335788726807,
|
| 1153 |
+
"rewards/margins": 3.1202616691589355,
|
| 1154 |
+
"rewards/rejected": -6.285294532775879,
|
| 1155 |
"step": 690
|
| 1156 |
},
|
| 1157 |
{
|
| 1158 |
"epoch": 0.806219406852865,
|
| 1159 |
+
"grad_norm": 27.17313115678408,
|
| 1160 |
"learning_rate": 5.4945854481754734e-08,
|
| 1161 |
+
"logits/chosen": -0.07786539942026138,
|
| 1162 |
+
"logits/rejected": 0.7690663933753967,
|
| 1163 |
+
"logps/chosen": -588.6981201171875,
|
| 1164 |
+
"logps/rejected": -892.0836181640625,
|
| 1165 |
+
"loss": 0.3858,
|
| 1166 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 1167 |
+
"rewards/chosen": -3.089141845703125,
|
| 1168 |
+
"rewards/margins": 3.093146800994873,
|
| 1169 |
+
"rewards/rejected": -6.182288646697998,
|
| 1170 |
"step": 700
|
| 1171 |
},
|
| 1172 |
{
|
| 1173 |
"epoch": 0.806219406852865,
|
| 1174 |
+
"eval_logits/chosen": -0.6058293581008911,
|
| 1175 |
+
"eval_logits/rejected": 0.35863998532295227,
|
| 1176 |
+
"eval_logps/chosen": -686.6614379882812,
|
| 1177 |
+
"eval_logps/rejected": -1093.6546630859375,
|
| 1178 |
+
"eval_loss": 0.31664812564849854,
|
| 1179 |
+
"eval_rewards/accuracies": 0.8245515823364258,
|
| 1180 |
+
"eval_rewards/chosen": -3.9569337368011475,
|
| 1181 |
+
"eval_rewards/margins": 4.078691005706787,
|
| 1182 |
+
"eval_rewards/rejected": -8.035624504089355,
|
| 1183 |
+
"eval_runtime": 644.0322,
|
| 1184 |
+
"eval_samples_per_second": 11.065,
|
| 1185 |
+
"eval_steps_per_second": 0.346,
|
| 1186 |
"step": 700
|
| 1187 |
},
|
| 1188 |
{
|
| 1189 |
"epoch": 0.817736826950763,
|
| 1190 |
+
"grad_norm": 26.61646646409602,
|
| 1191 |
"learning_rate": 4.881501533321605e-08,
|
| 1192 |
+
"logits/chosen": -0.4410906732082367,
|
| 1193 |
+
"logits/rejected": 0.5729657411575317,
|
| 1194 |
+
"logps/chosen": -575.347900390625,
|
| 1195 |
+
"logps/rejected": -934.3646240234375,
|
| 1196 |
+
"loss": 0.3728,
|
| 1197 |
+
"rewards/accuracies": 0.8187500238418579,
|
| 1198 |
+
"rewards/chosen": -2.907287359237671,
|
| 1199 |
+
"rewards/margins": 3.6123032569885254,
|
| 1200 |
+
"rewards/rejected": -6.519589900970459,
|
| 1201 |
"step": 710
|
| 1202 |
},
|
| 1203 |
{
|
| 1204 |
"epoch": 0.8292542470486611,
|
| 1205 |
+
"grad_norm": 29.443243927116985,
|
| 1206 |
"learning_rate": 4.300966395938377e-08,
|
| 1207 |
+
"logits/chosen": -0.38687664270401,
|
| 1208 |
+
"logits/rejected": 0.5803302526473999,
|
| 1209 |
+
"logps/chosen": -589.1840209960938,
|
| 1210 |
+
"logps/rejected": -924.7605590820312,
|
| 1211 |
+
"loss": 0.3718,
|
| 1212 |
+
"rewards/accuracies": 0.8125,
|
| 1213 |
+
"rewards/chosen": -3.089477062225342,
|
| 1214 |
+
"rewards/margins": 3.404120683670044,
|
| 1215 |
+
"rewards/rejected": -6.493597507476807,
|
| 1216 |
"step": 720
|
| 1217 |
},
|
| 1218 |
{
|
| 1219 |
"epoch": 0.8407716671465592,
|
| 1220 |
+
"grad_norm": 27.075379532779515,
|
| 1221 |
"learning_rate": 3.7539192566655246e-08,
|
| 1222 |
+
"logits/chosen": -0.2331455647945404,
|
| 1223 |
+
"logits/rejected": 0.7184351086616516,
|
| 1224 |
+
"logps/chosen": -597.6744384765625,
|
| 1225 |
+
"logps/rejected": -883.7164916992188,
|
| 1226 |
+
"loss": 0.3743,
|
| 1227 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 1228 |
+
"rewards/chosen": -3.0997419357299805,
|
| 1229 |
+
"rewards/margins": 2.9383721351623535,
|
| 1230 |
+
"rewards/rejected": -6.038114547729492,
|
| 1231 |
"step": 730
|
| 1232 |
},
|
| 1233 |
{
|
| 1234 |
"epoch": 0.8522890872444573,
|
| 1235 |
+
"grad_norm": 34.009147703744425,
|
| 1236 |
"learning_rate": 3.24124515747731e-08,
|
| 1237 |
+
"logits/chosen": 0.15024222433567047,
|
| 1238 |
+
"logits/rejected": 1.2283384799957275,
|
| 1239 |
+
"logps/chosen": -628.3982543945312,
|
| 1240 |
+
"logps/rejected": -985.0546875,
|
| 1241 |
+
"loss": 0.3708,
|
| 1242 |
+
"rewards/accuracies": 0.7906249761581421,
|
| 1243 |
+
"rewards/chosen": -3.4668242931365967,
|
| 1244 |
+
"rewards/margins": 3.758965253829956,
|
| 1245 |
+
"rewards/rejected": -7.2257890701293945,
|
| 1246 |
"step": 740
|
| 1247 |
},
|
| 1248 |
{
|
| 1249 |
"epoch": 0.8638065073423553,
|
| 1250 |
+
"grad_norm": 32.52714315668832,
|
| 1251 |
"learning_rate": 2.763773529814506e-08,
|
| 1252 |
+
"logits/chosen": 0.2603093981742859,
|
| 1253 |
+
"logits/rejected": 1.1775026321411133,
|
| 1254 |
+
"logps/chosen": -583.00439453125,
|
| 1255 |
+
"logps/rejected": -904.5720825195312,
|
| 1256 |
+
"loss": 0.3746,
|
| 1257 |
+
"rewards/accuracies": 0.778124988079071,
|
| 1258 |
+
"rewards/chosen": -3.006887912750244,
|
| 1259 |
+
"rewards/margins": 3.236215114593506,
|
| 1260 |
+
"rewards/rejected": -6.24310302734375,
|
| 1261 |
"step": 750
|
| 1262 |
},
|
| 1263 |
{
|
| 1264 |
"epoch": 0.8753239274402533,
|
| 1265 |
+
"grad_norm": 27.98558075342816,
|
| 1266 |
"learning_rate": 2.3222768526860698e-08,
|
| 1267 |
+
"logits/chosen": 0.1701681762933731,
|
| 1268 |
+
"logits/rejected": 1.1987775564193726,
|
| 1269 |
+
"logps/chosen": -613.1824951171875,
|
| 1270 |
+
"logps/rejected": -900.89111328125,
|
| 1271 |
+
"loss": 0.3922,
|
| 1272 |
+
"rewards/accuracies": 0.753125011920929,
|
| 1273 |
+
"rewards/chosen": -3.226149320602417,
|
| 1274 |
+
"rewards/margins": 2.7997944355010986,
|
| 1275 |
+
"rewards/rejected": -6.025943756103516,
|
| 1276 |
"step": 760
|
| 1277 |
},
|
| 1278 |
{
|
| 1279 |
"epoch": 0.8868413475381515,
|
| 1280 |
+
"grad_norm": 29.982562987973527,
|
| 1281 |
"learning_rate": 1.9174694029115146e-08,
|
| 1282 |
+
"logits/chosen": 0.12872493267059326,
|
| 1283 |
+
"logits/rejected": 1.1761186122894287,
|
| 1284 |
+
"logps/chosen": -600.1962890625,
|
| 1285 |
+
"logps/rejected": -947.6414794921875,
|
| 1286 |
+
"loss": 0.3739,
|
| 1287 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 1288 |
+
"rewards/chosen": -3.291241407394409,
|
| 1289 |
+
"rewards/margins": 3.434460401535034,
|
| 1290 |
+
"rewards/rejected": -6.725701808929443,
|
| 1291 |
"step": 770
|
| 1292 |
},
|
| 1293 |
{
|
| 1294 |
"epoch": 0.8983587676360495,
|
| 1295 |
+
"grad_norm": 29.902322430965125,
|
| 1296 |
"learning_rate": 1.5500060995258134e-08,
|
| 1297 |
+
"logits/chosen": 0.07875040918588638,
|
| 1298 |
+
"logits/rejected": 1.1803662776947021,
|
| 1299 |
+
"logps/chosen": -587.24951171875,
|
| 1300 |
+
"logps/rejected": -848.0086059570312,
|
| 1301 |
+
"loss": 0.3833,
|
| 1302 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 1303 |
+
"rewards/chosen": -3.145979642868042,
|
| 1304 |
+
"rewards/margins": 2.7399418354034424,
|
| 1305 |
+
"rewards/rejected": -5.885921955108643,
|
| 1306 |
"step": 780
|
| 1307 |
},
|
| 1308 |
{
|
| 1309 |
"epoch": 0.9098761877339476,
|
| 1310 |
+
"grad_norm": 31.87265554134813,
|
| 1311 |
"learning_rate": 1.2204814442165812e-08,
|
| 1312 |
+
"logits/chosen": 0.0372554175555706,
|
| 1313 |
+
"logits/rejected": 0.9403896331787109,
|
| 1314 |
+
"logps/chosen": -605.0985107421875,
|
| 1315 |
+
"logps/rejected": -893.2574462890625,
|
| 1316 |
+
"loss": 0.3836,
|
| 1317 |
+
"rewards/accuracies": 0.8031250238418579,
|
| 1318 |
+
"rewards/chosen": -3.219792127609253,
|
| 1319 |
+
"rewards/margins": 2.947915554046631,
|
| 1320 |
+
"rewards/rejected": -6.1677069664001465,
|
| 1321 |
"step": 790
|
| 1322 |
},
|
| 1323 |
{
|
| 1324 |
"epoch": 0.9213936078318457,
|
| 1325 |
+
"grad_norm": 28.29913492171739,
|
| 1326 |
"learning_rate": 9.294285595075669e-09,
|
| 1327 |
+
"logits/chosen": 0.1543000489473343,
|
| 1328 |
+
"logits/rejected": 1.00014328956604,
|
| 1329 |
+
"logps/chosen": -594.6954956054688,
|
| 1330 |
+
"logps/rejected": -909.7340087890625,
|
| 1331 |
+
"loss": 0.3785,
|
| 1332 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 1333 |
+
"rewards/chosen": -3.133711338043213,
|
| 1334 |
+
"rewards/margins": 3.2508773803710938,
|
| 1335 |
+
"rewards/rejected": -6.384588241577148,
|
| 1336 |
"step": 800
|
| 1337 |
},
|
| 1338 |
{
|
| 1339 |
"epoch": 0.9213936078318457,
|
| 1340 |
+
"eval_logits/chosen": -0.424582302570343,
|
| 1341 |
+
"eval_logits/rejected": 0.5558030605316162,
|
| 1342 |
+
"eval_logps/chosen": -702.706787109375,
|
| 1343 |
+
"eval_logps/rejected": -1123.9625244140625,
|
| 1344 |
+
"eval_loss": 0.31607937812805176,
|
| 1345 |
+
"eval_rewards/accuracies": 0.8211883306503296,
|
| 1346 |
+
"eval_rewards/chosen": -4.117387771606445,
|
| 1347 |
+
"eval_rewards/margins": 4.221314430236816,
|
| 1348 |
+
"eval_rewards/rejected": -8.338702201843262,
|
| 1349 |
+
"eval_runtime": 644.1715,
|
| 1350 |
+
"eval_samples_per_second": 11.062,
|
| 1351 |
+
"eval_steps_per_second": 0.346,
|
| 1352 |
"step": 800
|
| 1353 |
},
|
| 1354 |
{
|
| 1355 |
"epoch": 0.9329110279297438,
|
| 1356 |
+
"grad_norm": 28.271393767746897,
|
| 1357 |
"learning_rate": 6.773183262446914e-09,
|
| 1358 |
+
"logits/chosen": 0.015259919688105583,
|
| 1359 |
+
"logits/rejected": 1.0111353397369385,
|
| 1360 |
+
"logps/chosen": -610.5919189453125,
|
| 1361 |
+
"logps/rejected": -938.6796875,
|
| 1362 |
+
"loss": 0.3958,
|
| 1363 |
+
"rewards/accuracies": 0.778124988079071,
|
| 1364 |
+
"rewards/chosen": -3.225220203399658,
|
| 1365 |
+
"rewards/margins": 3.3511810302734375,
|
| 1366 |
+
"rewards/rejected": -6.5764007568359375,
|
| 1367 |
"step": 810
|
| 1368 |
},
|
| 1369 |
{
|
| 1370 |
"epoch": 0.9444284480276418,
|
| 1371 |
+
"grad_norm": 36.747504831392355,
|
| 1372 |
"learning_rate": 4.645586217799452e-09,
|
| 1373 |
+
"logits/chosen": -0.05728811025619507,
|
| 1374 |
+
"logits/rejected": 1.0568214654922485,
|
| 1375 |
+
"logps/chosen": -602.8958129882812,
|
| 1376 |
+
"logps/rejected": -914.1162109375,
|
| 1377 |
+
"loss": 0.392,
|
| 1378 |
+
"rewards/accuracies": 0.784375011920929,
|
| 1379 |
+
"rewards/chosen": -2.9873690605163574,
|
| 1380 |
+
"rewards/margins": 3.2894864082336426,
|
| 1381 |
+
"rewards/rejected": -6.27685546875,
|
| 1382 |
"step": 820
|
| 1383 |
},
|
| 1384 |
{
|
| 1385 |
"epoch": 0.9559458681255398,
|
| 1386 |
+
"grad_norm": 30.708105872731245,
|
| 1387 |
"learning_rate": 2.9149366008568987e-09,
|
| 1388 |
+
"logits/chosen": 0.1167169064283371,
|
| 1389 |
+
"logits/rejected": 1.0058209896087646,
|
| 1390 |
+
"logps/chosen": -580.9910278320312,
|
| 1391 |
+
"logps/rejected": -903.1005859375,
|
| 1392 |
+
"loss": 0.405,
|
| 1393 |
"rewards/accuracies": 0.778124988079071,
|
| 1394 |
+
"rewards/chosen": -2.9704513549804688,
|
| 1395 |
+
"rewards/margins": 3.3174147605895996,
|
| 1396 |
+
"rewards/rejected": -6.287866115570068,
|
| 1397 |
"step": 830
|
| 1398 |
},
|
| 1399 |
{
|
| 1400 |
"epoch": 0.967463288223438,
|
| 1401 |
+
"grad_norm": 26.852523446236688,
|
| 1402 |
"learning_rate": 1.5840343486700215e-09,
|
| 1403 |
+
"logits/chosen": 0.10119374096393585,
|
| 1404 |
+
"logits/rejected": 1.3030340671539307,
|
| 1405 |
+
"logps/chosen": -624.4727783203125,
|
| 1406 |
+
"logps/rejected": -907.2607421875,
|
| 1407 |
+
"loss": 0.4136,
|
| 1408 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 1409 |
+
"rewards/chosen": -3.329115629196167,
|
| 1410 |
+
"rewards/margins": 3.0215249061584473,
|
| 1411 |
+
"rewards/rejected": -6.350640296936035,
|
| 1412 |
"step": 840
|
| 1413 |
},
|
| 1414 |
{
|
| 1415 |
"epoch": 0.978980708321336,
|
| 1416 |
+
"grad_norm": 29.02811217591628,
|
| 1417 |
"learning_rate": 6.550326657293881e-10,
|
| 1418 |
+
"logits/chosen": 0.16462787985801697,
|
| 1419 |
+
"logits/rejected": 1.183337926864624,
|
| 1420 |
+
"logps/chosen": -576.1724853515625,
|
| 1421 |
+
"logps/rejected": -849.2581176757812,
|
| 1422 |
+
"loss": 0.4037,
|
| 1423 |
+
"rewards/accuracies": 0.793749988079071,
|
| 1424 |
+
"rewards/chosen": -3.109165668487549,
|
| 1425 |
+
"rewards/margins": 2.8826241493225098,
|
| 1426 |
+
"rewards/rejected": -5.991789817810059,
|
| 1427 |
"step": 850
|
| 1428 |
},
|
| 1429 |
{
|
| 1430 |
"epoch": 0.9904981284192341,
|
| 1431 |
+
"grad_norm": 38.83754661046954,
|
| 1432 |
"learning_rate": 1.2943454039654467e-10,
|
| 1433 |
+
"logits/chosen": 0.04645358771085739,
|
| 1434 |
+
"logits/rejected": 1.1813570261001587,
|
| 1435 |
+
"logps/chosen": -563.9542236328125,
|
| 1436 |
+
"logps/rejected": -806.0048828125,
|
| 1437 |
+
"loss": 0.385,
|
| 1438 |
+
"rewards/accuracies": 0.784375011920929,
|
| 1439 |
+
"rewards/chosen": -2.8706278800964355,
|
| 1440 |
+
"rewards/margins": 2.5064380168914795,
|
| 1441 |
+
"rewards/rejected": -5.377066135406494,
|
| 1442 |
"step": 860
|
| 1443 |
},
|
| 1444 |
{
|
| 1445 |
"epoch": 0.9997120644975526,
|
| 1446 |
"step": 868,
|
| 1447 |
"total_flos": 0.0,
|
| 1448 |
+
"train_loss": 0.42924998652550483,
|
| 1449 |
+
"train_runtime": 32201.1967,
|
| 1450 |
+
"train_samples_per_second": 3.451,
|
| 1451 |
+
"train_steps_per_second": 0.027
|
| 1452 |
}
|
| 1453 |
],
|
| 1454 |
"logging_steps": 10,
|