Model save
Browse files- README.md +17 -1
- all_results.json +5 -5
- model-00001-of-00003.safetensors +1 -1
- model-00002-of-00003.safetensors +1 -1
- model-00003-of-00003.safetensors +1 -1
- train_results.json +5 -5
- trainer_state.json +617 -455
README.md
CHANGED
|
@@ -16,6 +16,16 @@ should probably proofread and complete it, then remove this comment. -->
|
|
| 16 |
# zephyr-7b-dpo-full
|
| 17 |
|
| 18 |
This model is a fine-tuned version of [HuggingFaceH4/mistral-7b-sft-beta](https://huggingface.co/HuggingFaceH4/mistral-7b-sft-beta) on the None dataset.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 19 |
|
| 20 |
## Model description
|
| 21 |
|
|
@@ -37,7 +47,7 @@ The following hyperparameters were used during training:
|
|
| 37 |
- learning_rate: 5e-07
|
| 38 |
- train_batch_size: 8
|
| 39 |
- eval_batch_size: 8
|
| 40 |
-
- seed:
|
| 41 |
- distributed_type: multi-GPU
|
| 42 |
- num_devices: 8
|
| 43 |
- gradient_accumulation_steps: 2
|
|
@@ -50,6 +60,12 @@ The following hyperparameters were used during training:
|
|
| 50 |
|
| 51 |
### Training results
|
| 52 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 53 |
|
| 54 |
|
| 55 |
### Framework versions
|
|
|
|
| 16 |
# zephyr-7b-dpo-full
|
| 17 |
|
| 18 |
This model is a fine-tuned version of [HuggingFaceH4/mistral-7b-sft-beta](https://huggingface.co/HuggingFaceH4/mistral-7b-sft-beta) on the None dataset.
|
| 19 |
+
It achieves the following results on the evaluation set:
|
| 20 |
+
- Loss: 0.0069
|
| 21 |
+
- Rewards/chosen: -2.6575
|
| 22 |
+
- Rewards/rejected: -5.6311
|
| 23 |
+
- Rewards/accuracies: 0.6914
|
| 24 |
+
- Rewards/margins: 2.9735
|
| 25 |
+
- Logps/rejected: -820.4592
|
| 26 |
+
- Logps/chosen: -522.7909
|
| 27 |
+
- Logits/rejected: -2.4988
|
| 28 |
+
- Logits/chosen: -2.5238
|
| 29 |
|
| 30 |
## Model description
|
| 31 |
|
|
|
|
| 47 |
- learning_rate: 5e-07
|
| 48 |
- train_batch_size: 8
|
| 49 |
- eval_batch_size: 8
|
| 50 |
+
- seed: 42
|
| 51 |
- distributed_type: multi-GPU
|
| 52 |
- num_devices: 8
|
| 53 |
- gradient_accumulation_steps: 2
|
|
|
|
| 60 |
|
| 61 |
### Training results
|
| 62 |
|
| 63 |
+
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 64 |
+
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 65 |
+
| 0.0071 | 0.21 | 100 | 0.0072 | -0.8922 | -1.6083 | 0.6523 | 0.7161 | -418.1843 | -346.2603 | -2.6342 | -2.6489 |
|
| 66 |
+
| 0.0068 | 0.42 | 200 | 0.0070 | -1.7394 | -3.1970 | 0.6680 | 1.4577 | -577.0542 | -430.9749 | -2.6058 | -2.6219 |
|
| 67 |
+
| 0.0069 | 0.63 | 300 | 0.0069 | -2.2358 | -4.6352 | 0.6992 | 2.3994 | -720.8748 | -480.6185 | -2.5115 | -2.5351 |
|
| 68 |
+
| 0.0067 | 0.84 | 400 | 0.0069 | -2.6575 | -5.6311 | 0.6914 | 2.9735 | -820.4592 | -522.7909 | -2.4988 | -2.5238 |
|
| 69 |
|
| 70 |
|
| 71 |
### Framework versions
|
all_results.json
CHANGED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
-
"train_loss": 0.
|
| 4 |
-
"train_runtime":
|
| 5 |
-
"train_samples":
|
| 6 |
-
"train_samples_per_second":
|
| 7 |
-
"train_steps_per_second": 0.
|
| 8 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
+
"train_loss": 0.007060731609184373,
|
| 4 |
+
"train_runtime": 4232.7551,
|
| 5 |
+
"train_samples": 61134,
|
| 6 |
+
"train_samples_per_second": 14.443,
|
| 7 |
+
"train_steps_per_second": 0.113
|
| 8 |
}
|
model-00001-of-00003.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 4943162336
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8118a565a3d3b99cddea7bd517315c6f04fad9ab28c7adf7a86e47ac8f748a4e
|
| 3 |
size 4943162336
|
model-00002-of-00003.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 4999819336
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:485d1a2bc4d5f0be525df784e2f4b0ace0a55d5990d3f7f85d4aa190ce78c8d1
|
| 3 |
size 4999819336
|
model-00003-of-00003.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 4540516344
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c790488581fcd33e4d3efa897f475860ceddc6631b53ab000d6f2a5fd8380a7c
|
| 3 |
size 4540516344
|
train_results.json
CHANGED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
-
"train_loss": 0.
|
| 4 |
-
"train_runtime":
|
| 5 |
-
"train_samples":
|
| 6 |
-
"train_samples_per_second":
|
| 7 |
-
"train_steps_per_second": 0.
|
| 8 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
+
"train_loss": 0.007060731609184373,
|
| 4 |
+
"train_runtime": 4232.7551,
|
| 5 |
+
"train_samples": 61134,
|
| 6 |
+
"train_samples_per_second": 14.443,
|
| 7 |
+
"train_steps_per_second": 0.113
|
| 8 |
}
|
trainer_state.json
CHANGED
|
@@ -1,21 +1,21 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": null,
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
-
"epoch": 0
|
| 5 |
-
"eval_steps":
|
| 6 |
-
"global_step":
|
| 7 |
"is_hyper_param_search": false,
|
| 8 |
"is_local_process_zero": true,
|
| 9 |
"is_world_process_zero": true,
|
| 10 |
"log_history": [
|
| 11 |
{
|
| 12 |
"epoch": 0.0,
|
| 13 |
-
"learning_rate": 1.
|
| 14 |
-
"logits/chosen": -2.
|
| 15 |
-
"logits/rejected": -2.
|
| 16 |
-
"logps/chosen": -
|
| 17 |
-
"logps/rejected": -
|
| 18 |
-
"loss": 0.
|
| 19 |
"rewards/accuracies": 0.0,
|
| 20 |
"rewards/chosen": 0.0,
|
| 21 |
"rewards/margins": 0.0,
|
|
@@ -24,578 +24,740 @@
|
|
| 24 |
},
|
| 25 |
{
|
| 26 |
"epoch": 0.02,
|
| 27 |
-
"learning_rate": 1.
|
| 28 |
-
"logits/chosen": -2.
|
| 29 |
-
"logits/rejected": -2.
|
| 30 |
-
"logps/chosen": -
|
| 31 |
-
"logps/rejected": -
|
| 32 |
-
"loss": 0.
|
| 33 |
-
"rewards/accuracies": 0.
|
| 34 |
-
"rewards/chosen": 0.
|
| 35 |
-
"rewards/margins":
|
| 36 |
-
"rewards/rejected": -
|
| 37 |
"step": 10
|
| 38 |
},
|
| 39 |
{
|
| 40 |
-
"epoch": 0.
|
| 41 |
-
"learning_rate": 2.
|
| 42 |
-
"logits/chosen": -2.
|
| 43 |
-
"logits/rejected": -2.
|
| 44 |
-
"logps/chosen": -
|
| 45 |
-
"logps/rejected": -
|
| 46 |
-
"loss": 0.
|
| 47 |
-
"rewards/accuracies": 0.
|
| 48 |
-
"rewards/chosen": 0.
|
| 49 |
-
"rewards/margins": 0.
|
| 50 |
-
"rewards/rejected": -0.
|
| 51 |
"step": 20
|
| 52 |
},
|
| 53 |
{
|
| 54 |
-
"epoch": 0.
|
| 55 |
-
"learning_rate": 3.
|
| 56 |
-
"logits/chosen": -2.
|
| 57 |
-
"logits/rejected": -2.
|
| 58 |
-
"logps/chosen": -
|
| 59 |
-
"logps/rejected": -
|
| 60 |
-
"loss": 0.
|
| 61 |
-
"rewards/accuracies": 0.
|
| 62 |
-
"rewards/chosen": 0.
|
| 63 |
-
"rewards/margins": 0.
|
| 64 |
-
"rewards/rejected": -0.
|
| 65 |
"step": 30
|
| 66 |
},
|
| 67 |
{
|
| 68 |
-
"epoch": 0.
|
| 69 |
-
"learning_rate": 4.
|
| 70 |
-
"logits/chosen": -2.
|
| 71 |
-
"logits/rejected": -2.
|
| 72 |
-
"logps/chosen": -
|
| 73 |
-
"logps/rejected": -
|
| 74 |
-
"loss": 0.
|
| 75 |
-
"rewards/accuracies": 0.
|
| 76 |
-
"rewards/chosen": 0.
|
| 77 |
-
"rewards/margins": 0.
|
| 78 |
-
"rewards/rejected": -0.
|
| 79 |
"step": 40
|
| 80 |
},
|
| 81 |
{
|
| 82 |
-
"epoch": 0.
|
| 83 |
-
"learning_rate": 4.
|
| 84 |
-
"logits/chosen": -2.
|
| 85 |
-
"logits/rejected": -2.
|
| 86 |
-
"logps/chosen": -
|
| 87 |
-
"logps/rejected": -
|
| 88 |
-
"loss": 0.
|
| 89 |
-
"rewards/accuracies": 0.
|
| 90 |
-
"rewards/chosen":
|
| 91 |
-
"rewards/margins": 0.
|
| 92 |
-
"rewards/rejected":
|
| 93 |
"step": 50
|
| 94 |
},
|
| 95 |
{
|
| 96 |
-
"epoch": 0.
|
| 97 |
-
"learning_rate": 4.
|
| 98 |
-
"logits/chosen": -2.
|
| 99 |
-
"logits/rejected": -2.
|
| 100 |
-
"logps/chosen": -
|
| 101 |
-
"logps/rejected": -
|
| 102 |
-
"loss": 0.
|
| 103 |
-
"rewards/accuracies": 0.
|
| 104 |
-
"rewards/chosen":
|
| 105 |
-
"rewards/margins":
|
| 106 |
-
"rewards/rejected": -
|
| 107 |
"step": 60
|
| 108 |
},
|
| 109 |
{
|
| 110 |
-
"epoch": 0.
|
| 111 |
-
"learning_rate": 4.
|
| 112 |
-
"logits/chosen": -2.
|
| 113 |
-
"logits/rejected": -2.
|
| 114 |
-
"logps/chosen": -
|
| 115 |
-
"logps/rejected": -
|
| 116 |
-
"loss": 0.
|
| 117 |
-
"rewards/accuracies": 0.
|
| 118 |
-
"rewards/chosen": -0.
|
| 119 |
-
"rewards/margins":
|
| 120 |
-
"rewards/rejected": -
|
| 121 |
"step": 70
|
| 122 |
},
|
| 123 |
{
|
| 124 |
-
"epoch": 0.
|
| 125 |
-
"learning_rate": 4.
|
| 126 |
-
"logits/chosen": -2.
|
| 127 |
-
"logits/rejected": -2.
|
| 128 |
-
"logps/chosen": -
|
| 129 |
-
"logps/rejected": -
|
| 130 |
-
"loss": 0.
|
| 131 |
-
"rewards/accuracies": 0.
|
| 132 |
-
"rewards/chosen": -0.
|
| 133 |
-
"rewards/margins":
|
| 134 |
-
"rewards/rejected": -
|
| 135 |
"step": 80
|
| 136 |
},
|
| 137 |
{
|
| 138 |
-
"epoch": 0.
|
| 139 |
-
"learning_rate": 4.
|
| 140 |
-
"logits/chosen": -2.
|
| 141 |
-
"logits/rejected": -2.
|
| 142 |
-
"logps/chosen": -
|
| 143 |
-
"logps/rejected": -
|
| 144 |
-
"loss": 0.
|
| 145 |
-
"rewards/accuracies": 0.
|
| 146 |
-
"rewards/chosen": -0.
|
| 147 |
-
"rewards/margins":
|
| 148 |
-
"rewards/rejected": -
|
| 149 |
"step": 90
|
| 150 |
},
|
| 151 |
{
|
| 152 |
-
"epoch": 0.
|
| 153 |
-
"learning_rate": 4.
|
| 154 |
-
"logits/chosen": -2.
|
| 155 |
-
"logits/rejected": -2.
|
| 156 |
-
"logps/chosen": -
|
| 157 |
-
"logps/rejected": -
|
| 158 |
-
"loss": 0.
|
| 159 |
-
"rewards/accuracies": 0.
|
| 160 |
-
"rewards/chosen": -
|
| 161 |
-
"rewards/margins":
|
| 162 |
-
"rewards/rejected": -
|
| 163 |
"step": 100
|
| 164 |
},
|
| 165 |
{
|
| 166 |
-
"epoch": 0.
|
| 167 |
-
"
|
| 168 |
-
"
|
| 169 |
-
"
|
| 170 |
-
"
|
| 171 |
-
"
|
| 172 |
-
"
|
| 173 |
-
"
|
| 174 |
-
"
|
| 175 |
-
"
|
| 176 |
-
"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 177 |
"step": 110
|
| 178 |
},
|
| 179 |
{
|
| 180 |
-
"epoch": 0.
|
| 181 |
-
"learning_rate": 4.
|
| 182 |
-
"logits/chosen": -2.
|
| 183 |
-
"logits/rejected": -2.
|
| 184 |
-
"logps/chosen": -
|
| 185 |
-
"logps/rejected": -
|
| 186 |
-
"loss": 0.
|
| 187 |
-
"rewards/accuracies": 0.
|
| 188 |
-
"rewards/chosen": -1.
|
| 189 |
-
"rewards/margins":
|
| 190 |
-
"rewards/rejected": -
|
| 191 |
"step": 120
|
| 192 |
},
|
| 193 |
{
|
| 194 |
-
"epoch": 0.
|
| 195 |
-
"learning_rate": 4.
|
| 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":
|
| 204 |
-
"rewards/rejected": -
|
| 205 |
"step": 130
|
| 206 |
},
|
| 207 |
{
|
| 208 |
-
"epoch": 0.
|
| 209 |
-
"learning_rate": 4.
|
| 210 |
-
"logits/chosen": -2.
|
| 211 |
-
"logits/rejected": -2.
|
| 212 |
-
"logps/chosen": -
|
| 213 |
-
"logps/rejected": -
|
| 214 |
-
"loss": 0.
|
| 215 |
-
"rewards/accuracies": 0.
|
| 216 |
-
"rewards/chosen": -
|
| 217 |
-
"rewards/margins":
|
| 218 |
-
"rewards/rejected": -2.
|
| 219 |
"step": 140
|
| 220 |
},
|
| 221 |
{
|
| 222 |
-
"epoch": 0.
|
| 223 |
-
"learning_rate":
|
| 224 |
-
"logits/chosen": -2.
|
| 225 |
-
"logits/rejected": -2.
|
| 226 |
-
"logps/chosen": -
|
| 227 |
-
"logps/rejected": -
|
| 228 |
-
"loss": 0.
|
| 229 |
-
"rewards/accuracies": 0.
|
| 230 |
-
"rewards/chosen": -0.
|
| 231 |
-
"rewards/margins": 1.
|
| 232 |
-
"rewards/rejected": -2.
|
| 233 |
"step": 150
|
| 234 |
},
|
| 235 |
{
|
| 236 |
-
"epoch": 0.
|
| 237 |
-
"learning_rate":
|
| 238 |
-
"logits/chosen": -2.
|
| 239 |
-
"logits/rejected": -2.
|
| 240 |
-
"logps/chosen": -
|
| 241 |
-
"logps/rejected": -
|
| 242 |
-
"loss": 0.
|
| 243 |
-
"rewards/accuracies": 0.
|
| 244 |
-
"rewards/chosen": -
|
| 245 |
-
"rewards/margins":
|
| 246 |
-
"rewards/rejected": -
|
| 247 |
"step": 160
|
| 248 |
},
|
| 249 |
{
|
| 250 |
-
"epoch": 0.
|
| 251 |
-
"learning_rate":
|
| 252 |
-
"logits/chosen": -2.
|
| 253 |
-
"logits/rejected": -2.
|
| 254 |
-
"logps/chosen": -
|
| 255 |
-
"logps/rejected": -
|
| 256 |
-
"loss": 0.
|
| 257 |
-
"rewards/accuracies": 0.
|
| 258 |
-
"rewards/chosen": -1.
|
| 259 |
-
"rewards/margins": 1.
|
| 260 |
-
"rewards/rejected": -
|
| 261 |
"step": 170
|
| 262 |
},
|
| 263 |
{
|
| 264 |
-
"epoch": 0.
|
| 265 |
-
"learning_rate": 3.
|
| 266 |
-
"logits/chosen": -2.
|
| 267 |
-
"logits/rejected": -2.
|
| 268 |
-
"logps/chosen": -
|
| 269 |
-
"logps/rejected": -
|
| 270 |
-
"loss": 0.
|
| 271 |
-
"rewards/accuracies": 0.
|
| 272 |
-
"rewards/chosen": -1.
|
| 273 |
-
"rewards/margins":
|
| 274 |
-
"rewards/rejected": -
|
| 275 |
"step": 180
|
| 276 |
},
|
| 277 |
{
|
| 278 |
-
"epoch": 0.
|
| 279 |
-
"learning_rate": 3.
|
| 280 |
-
"logits/chosen": -2.
|
| 281 |
-
"logits/rejected": -2.
|
| 282 |
-
"logps/chosen": -
|
| 283 |
-
"logps/rejected": -
|
| 284 |
-
"loss": 0.
|
| 285 |
-
"rewards/accuracies": 0.
|
| 286 |
-
"rewards/chosen": -1.
|
| 287 |
-
"rewards/margins":
|
| 288 |
-
"rewards/rejected": -
|
| 289 |
"step": 190
|
| 290 |
},
|
| 291 |
{
|
| 292 |
-
"epoch": 0.
|
| 293 |
-
"learning_rate":
|
| 294 |
-
"logits/chosen": -2.
|
| 295 |
-
"logits/rejected": -2.
|
| 296 |
-
"logps/chosen": -
|
| 297 |
-
"logps/rejected": -
|
| 298 |
-
"loss": 0.
|
| 299 |
-
"rewards/accuracies": 0.
|
| 300 |
-
"rewards/chosen": -1.
|
| 301 |
-
"rewards/margins":
|
| 302 |
-
"rewards/rejected": -
|
| 303 |
"step": 200
|
| 304 |
},
|
| 305 |
{
|
| 306 |
-
"epoch": 0.
|
| 307 |
-
"
|
| 308 |
-
"
|
| 309 |
-
"
|
| 310 |
-
"
|
| 311 |
-
"
|
| 312 |
-
"
|
| 313 |
-
"
|
| 314 |
-
"
|
| 315 |
-
"
|
| 316 |
-
"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 317 |
"step": 210
|
| 318 |
},
|
| 319 |
{
|
| 320 |
-
"epoch": 0.
|
| 321 |
-
"learning_rate":
|
| 322 |
-
"logits/chosen": -2.
|
| 323 |
-
"logits/rejected": -2.
|
| 324 |
-
"logps/chosen": -
|
| 325 |
-
"logps/rejected": -
|
| 326 |
-
"loss": 0.
|
| 327 |
-
"rewards/accuracies": 0.
|
| 328 |
-
"rewards/chosen": -1.
|
| 329 |
-
"rewards/margins": 1.
|
| 330 |
-
"rewards/rejected": -
|
| 331 |
"step": 220
|
| 332 |
},
|
| 333 |
{
|
| 334 |
-
"epoch": 0.
|
| 335 |
-
"learning_rate":
|
| 336 |
-
"logits/chosen": -2.
|
| 337 |
-
"logits/rejected": -2.
|
| 338 |
-
"logps/chosen": -
|
| 339 |
-
"logps/rejected": -
|
| 340 |
-
"loss": 0.
|
| 341 |
-
"rewards/accuracies": 0.
|
| 342 |
-
"rewards/chosen": -1.
|
| 343 |
-
"rewards/margins":
|
| 344 |
-
"rewards/rejected": -
|
| 345 |
"step": 230
|
| 346 |
},
|
| 347 |
{
|
| 348 |
-
"epoch": 0.
|
| 349 |
-
"learning_rate": 2.
|
| 350 |
-
"logits/chosen": -2.
|
| 351 |
-
"logits/rejected": -2.
|
| 352 |
-
"logps/chosen": -
|
| 353 |
-
"logps/rejected": -
|
| 354 |
-
"loss": 0.
|
| 355 |
-
"rewards/accuracies": 0.
|
| 356 |
-
"rewards/chosen": -
|
| 357 |
-
"rewards/margins":
|
| 358 |
-
"rewards/rejected": -3.
|
| 359 |
"step": 240
|
| 360 |
},
|
| 361 |
{
|
| 362 |
-
"epoch": 0.
|
| 363 |
-
"learning_rate":
|
| 364 |
-
"logits/chosen": -2.
|
| 365 |
-
"logits/rejected": -2.
|
| 366 |
-
"logps/chosen": -
|
| 367 |
-
"logps/rejected": -
|
| 368 |
-
"loss": 0.
|
| 369 |
-
"rewards/accuracies": 0.
|
| 370 |
-
"rewards/chosen": -1.
|
| 371 |
-
"rewards/margins":
|
| 372 |
-
"rewards/rejected": -3.
|
| 373 |
"step": 250
|
| 374 |
},
|
| 375 |
{
|
| 376 |
-
"epoch": 0.
|
| 377 |
-
"learning_rate":
|
| 378 |
-
"logits/chosen": -2.
|
| 379 |
-
"logits/rejected": -2.
|
| 380 |
-
"logps/chosen": -
|
| 381 |
-
"logps/rejected": -
|
| 382 |
-
"loss": 0.
|
| 383 |
-
"rewards/accuracies": 0.
|
| 384 |
-
"rewards/chosen": -1.
|
| 385 |
-
"rewards/margins":
|
| 386 |
-
"rewards/rejected": -3.
|
| 387 |
"step": 260
|
| 388 |
},
|
| 389 |
{
|
| 390 |
-
"epoch": 0.
|
| 391 |
-
"learning_rate":
|
| 392 |
-
"logits/chosen": -2.
|
| 393 |
-
"logits/rejected": -2.
|
| 394 |
-
"logps/chosen": -
|
| 395 |
-
"logps/rejected": -
|
| 396 |
-
"loss": 0.
|
| 397 |
-
"rewards/accuracies": 0.
|
| 398 |
-
"rewards/chosen": -1.
|
| 399 |
-
"rewards/margins": 2.
|
| 400 |
-
"rewards/rejected": -3.
|
| 401 |
"step": 270
|
| 402 |
},
|
| 403 |
{
|
| 404 |
-
"epoch": 0.
|
| 405 |
-
"learning_rate":
|
| 406 |
-
"logits/chosen": -2.
|
| 407 |
-
"logits/rejected": -2.
|
| 408 |
-
"logps/chosen": -
|
| 409 |
-
"logps/rejected": -
|
| 410 |
-
"loss": 0.
|
| 411 |
-
"rewards/accuracies": 0.
|
| 412 |
-
"rewards/chosen": -1.
|
| 413 |
-
"rewards/margins":
|
| 414 |
-
"rewards/rejected": -3.
|
| 415 |
"step": 280
|
| 416 |
},
|
| 417 |
{
|
| 418 |
-
"epoch": 0.
|
| 419 |
-
"learning_rate":
|
| 420 |
-
"logits/chosen": -2.
|
| 421 |
-
"logits/rejected": -2.
|
| 422 |
-
"logps/chosen": -
|
| 423 |
-
"logps/rejected": -
|
| 424 |
-
"loss": 0.
|
| 425 |
-
"rewards/accuracies": 0.
|
| 426 |
-
"rewards/chosen": -
|
| 427 |
-
"rewards/margins":
|
| 428 |
-
"rewards/rejected": -3.
|
| 429 |
"step": 290
|
| 430 |
},
|
| 431 |
{
|
| 432 |
-
"epoch": 0.
|
| 433 |
-
"learning_rate":
|
| 434 |
-
"logits/chosen": -2.
|
| 435 |
-
"logits/rejected": -2.
|
| 436 |
-
"logps/chosen": -523.
|
| 437 |
-
"logps/rejected": -
|
| 438 |
-
"loss": 0.
|
| 439 |
-
"rewards/accuracies": 0.
|
| 440 |
-
"rewards/chosen": -
|
| 441 |
-
"rewards/margins": 2.
|
| 442 |
-
"rewards/rejected": -
|
| 443 |
"step": 300
|
| 444 |
},
|
| 445 |
{
|
| 446 |
-
"epoch": 0.
|
| 447 |
-
"
|
| 448 |
-
"
|
| 449 |
-
"
|
| 450 |
-
"
|
| 451 |
-
"
|
| 452 |
-
"
|
| 453 |
-
"
|
| 454 |
-
"
|
| 455 |
-
"
|
| 456 |
-
"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 457 |
"step": 310
|
| 458 |
},
|
| 459 |
{
|
| 460 |
-
"epoch": 0.
|
| 461 |
-
"learning_rate":
|
| 462 |
-
"logits/chosen": -2.
|
| 463 |
-
"logits/rejected": -2.
|
| 464 |
-
"logps/chosen": -
|
| 465 |
-
"logps/rejected": -
|
| 466 |
-
"loss": 0.
|
| 467 |
-
"rewards/accuracies": 0.
|
| 468 |
-
"rewards/chosen": -
|
| 469 |
-
"rewards/margins": 2.
|
| 470 |
-
"rewards/rejected": -
|
| 471 |
"step": 320
|
| 472 |
},
|
| 473 |
{
|
| 474 |
-
"epoch": 0.
|
| 475 |
-
"learning_rate":
|
| 476 |
-
"logits/chosen": -2.
|
| 477 |
-
"logits/rejected": -2.
|
| 478 |
-
"logps/chosen": -
|
| 479 |
-
"logps/rejected": -
|
| 480 |
-
"loss": 0.
|
| 481 |
-
"rewards/accuracies": 0.
|
| 482 |
-
"rewards/chosen": -
|
| 483 |
-
"rewards/margins":
|
| 484 |
-
"rewards/rejected": -
|
| 485 |
"step": 330
|
| 486 |
},
|
| 487 |
{
|
| 488 |
-
"epoch": 0.
|
| 489 |
-
"learning_rate":
|
| 490 |
-
"logits/chosen": -2.
|
| 491 |
-
"logits/rejected": -2.
|
| 492 |
-
"logps/chosen": -
|
| 493 |
-
"logps/rejected": -
|
| 494 |
-
"loss": 0.
|
| 495 |
-
"rewards/accuracies": 0.
|
| 496 |
-
"rewards/chosen": -
|
| 497 |
-
"rewards/margins": 2.
|
| 498 |
-
"rewards/rejected": -
|
| 499 |
"step": 340
|
| 500 |
},
|
| 501 |
{
|
| 502 |
-
"epoch": 0.
|
| 503 |
-
"learning_rate":
|
| 504 |
-
"logits/chosen": -2.
|
| 505 |
-
"logits/rejected": -2.
|
| 506 |
-
"logps/chosen": -
|
| 507 |
-
"logps/rejected": -
|
| 508 |
-
"loss": 0.
|
| 509 |
-
"rewards/accuracies": 0.
|
| 510 |
-
"rewards/chosen": -
|
| 511 |
-
"rewards/margins": 2.
|
| 512 |
-
"rewards/rejected": -
|
| 513 |
"step": 350
|
| 514 |
},
|
| 515 |
{
|
| 516 |
-
"epoch": 0.
|
| 517 |
-
"learning_rate":
|
| 518 |
-
"logits/chosen": -2.
|
| 519 |
-
"logits/rejected": -2.
|
| 520 |
-
"logps/chosen": -
|
| 521 |
-
"logps/rejected": -
|
| 522 |
-
"loss": 0.
|
| 523 |
-
"rewards/accuracies": 0.
|
| 524 |
-
"rewards/chosen": -
|
| 525 |
-
"rewards/margins":
|
| 526 |
-
"rewards/rejected": -
|
| 527 |
"step": 360
|
| 528 |
},
|
| 529 |
{
|
| 530 |
-
"epoch": 0.
|
| 531 |
-
"learning_rate":
|
| 532 |
-
"logits/chosen": -2.
|
| 533 |
-
"logits/rejected": -2.
|
| 534 |
-
"logps/chosen": -
|
| 535 |
-
"logps/rejected": -
|
| 536 |
-
"loss": 0.
|
| 537 |
-
"rewards/accuracies": 0.
|
| 538 |
-
"rewards/chosen": -
|
| 539 |
-
"rewards/margins":
|
| 540 |
-
"rewards/rejected": -
|
| 541 |
"step": 370
|
| 542 |
},
|
| 543 |
{
|
| 544 |
-
"epoch": 0.
|
| 545 |
-
"learning_rate":
|
| 546 |
-
"logits/chosen": -2.
|
| 547 |
-
"logits/rejected": -2.
|
| 548 |
-
"logps/chosen": -
|
| 549 |
-
"logps/rejected": -
|
| 550 |
-
"loss": 0.
|
| 551 |
-
"rewards/accuracies": 0.
|
| 552 |
-
"rewards/chosen": -
|
| 553 |
-
"rewards/margins": 2.
|
| 554 |
-
"rewards/rejected": -
|
| 555 |
"step": 380
|
| 556 |
},
|
| 557 |
{
|
| 558 |
-
"epoch": 0.
|
| 559 |
-
"learning_rate":
|
| 560 |
-
"logits/chosen": -2.
|
| 561 |
-
"logits/rejected": -2.
|
| 562 |
-
"logps/chosen": -
|
| 563 |
-
"logps/rejected": -
|
| 564 |
-
"loss": 0.
|
| 565 |
-
"rewards/accuracies": 0.
|
| 566 |
-
"rewards/chosen": -
|
| 567 |
-
"rewards/margins":
|
| 568 |
-
"rewards/rejected": -
|
| 569 |
"step": 390
|
| 570 |
},
|
| 571 |
{
|
| 572 |
-
"epoch": 0.
|
| 573 |
-
"learning_rate":
|
| 574 |
-
"logits/chosen": -2.
|
| 575 |
-
"logits/rejected": -2.
|
| 576 |
-
"logps/chosen": -
|
| 577 |
-
"logps/rejected": -
|
| 578 |
-
"loss": 0.
|
| 579 |
-
"rewards/accuracies": 0.
|
| 580 |
-
"rewards/chosen": -
|
| 581 |
-
"rewards/margins":
|
| 582 |
-
"rewards/rejected": -
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 583 |
"step": 400
|
| 584 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 585 |
{
|
| 586 |
"epoch": 1.0,
|
| 587 |
-
"step":
|
| 588 |
"total_flos": 0.0,
|
| 589 |
-
"train_loss": 0.
|
| 590 |
-
"train_runtime":
|
| 591 |
-
"train_samples_per_second":
|
| 592 |
-
"train_steps_per_second": 0.
|
| 593 |
}
|
| 594 |
],
|
| 595 |
"logging_steps": 10,
|
| 596 |
-
"max_steps":
|
| 597 |
"num_train_epochs": 1,
|
| 598 |
-
"save_steps":
|
| 599 |
"total_flos": 0.0,
|
| 600 |
"trial_name": null,
|
| 601 |
"trial_params": null
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": null,
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 1.0,
|
| 5 |
+
"eval_steps": 100,
|
| 6 |
+
"global_step": 478,
|
| 7 |
"is_hyper_param_search": false,
|
| 8 |
"is_local_process_zero": true,
|
| 9 |
"is_world_process_zero": true,
|
| 10 |
"log_history": [
|
| 11 |
{
|
| 12 |
"epoch": 0.0,
|
| 13 |
+
"learning_rate": 1.0416666666666666e-08,
|
| 14 |
+
"logits/chosen": -2.8099329471588135,
|
| 15 |
+
"logits/rejected": -2.7572641372680664,
|
| 16 |
+
"logps/chosen": -241.48843383789062,
|
| 17 |
+
"logps/rejected": -197.4517822265625,
|
| 18 |
+
"loss": 0.0078,
|
| 19 |
"rewards/accuracies": 0.0,
|
| 20 |
"rewards/chosen": 0.0,
|
| 21 |
"rewards/margins": 0.0,
|
|
|
|
| 24 |
},
|
| 25 |
{
|
| 26 |
"epoch": 0.02,
|
| 27 |
+
"learning_rate": 1.0416666666666667e-07,
|
| 28 |
+
"logits/chosen": -2.8321194648742676,
|
| 29 |
+
"logits/rejected": -2.8086512088775635,
|
| 30 |
+
"logps/chosen": -292.6743469238281,
|
| 31 |
+
"logps/rejected": -278.57659912109375,
|
| 32 |
+
"loss": 0.0078,
|
| 33 |
+
"rewards/accuracies": 0.4097222089767456,
|
| 34 |
+
"rewards/chosen": 0.0001647545286687091,
|
| 35 |
+
"rewards/margins": 6.90047672833316e-05,
|
| 36 |
+
"rewards/rejected": 9.574973955750465e-05,
|
| 37 |
"step": 10
|
| 38 |
},
|
| 39 |
{
|
| 40 |
+
"epoch": 0.04,
|
| 41 |
+
"learning_rate": 2.0833333333333333e-07,
|
| 42 |
+
"logits/chosen": -2.812178134918213,
|
| 43 |
+
"logits/rejected": -2.783703565597534,
|
| 44 |
+
"logps/chosen": -290.24627685546875,
|
| 45 |
+
"logps/rejected": -290.8101806640625,
|
| 46 |
+
"loss": 0.0078,
|
| 47 |
+
"rewards/accuracies": 0.5874999761581421,
|
| 48 |
+
"rewards/chosen": 0.0015331681352108717,
|
| 49 |
+
"rewards/margins": 0.002536341082304716,
|
| 50 |
+
"rewards/rejected": -0.0010031728306785226,
|
| 51 |
"step": 20
|
| 52 |
},
|
| 53 |
{
|
| 54 |
+
"epoch": 0.06,
|
| 55 |
+
"learning_rate": 3.1249999999999997e-07,
|
| 56 |
+
"logits/chosen": -2.7607104778289795,
|
| 57 |
+
"logits/rejected": -2.729504108428955,
|
| 58 |
+
"logps/chosen": -245.93002319335938,
|
| 59 |
+
"logps/rejected": -226.58041381835938,
|
| 60 |
+
"loss": 0.0078,
|
| 61 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 62 |
+
"rewards/chosen": 0.004320907406508923,
|
| 63 |
+
"rewards/margins": 0.007326310966163874,
|
| 64 |
+
"rewards/rejected": -0.003005403093993664,
|
| 65 |
"step": 30
|
| 66 |
},
|
| 67 |
{
|
| 68 |
+
"epoch": 0.08,
|
| 69 |
+
"learning_rate": 4.1666666666666667e-07,
|
| 70 |
+
"logits/chosen": -2.802220106124878,
|
| 71 |
+
"logits/rejected": -2.773683547973633,
|
| 72 |
+
"logps/chosen": -296.27679443359375,
|
| 73 |
+
"logps/rejected": -261.5389404296875,
|
| 74 |
+
"loss": 0.0077,
|
| 75 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 76 |
+
"rewards/chosen": 0.03244831785559654,
|
| 77 |
+
"rewards/margins": 0.03520316258072853,
|
| 78 |
+
"rewards/rejected": -0.002754846354946494,
|
| 79 |
"step": 40
|
| 80 |
},
|
| 81 |
{
|
| 82 |
+
"epoch": 0.1,
|
| 83 |
+
"learning_rate": 4.999733114418725e-07,
|
| 84 |
+
"logits/chosen": -2.7273590564727783,
|
| 85 |
+
"logits/rejected": -2.706237316131592,
|
| 86 |
+
"logps/chosen": -270.7037658691406,
|
| 87 |
+
"logps/rejected": -262.9428405761719,
|
| 88 |
+
"loss": 0.0076,
|
| 89 |
+
"rewards/accuracies": 0.675000011920929,
|
| 90 |
+
"rewards/chosen": 0.0672668069601059,
|
| 91 |
+
"rewards/margins": 0.05972485989332199,
|
| 92 |
+
"rewards/rejected": 0.007541947066783905,
|
| 93 |
"step": 50
|
| 94 |
},
|
| 95 |
{
|
| 96 |
+
"epoch": 0.13,
|
| 97 |
+
"learning_rate": 4.990398100856366e-07,
|
| 98 |
+
"logits/chosen": -2.7465662956237793,
|
| 99 |
+
"logits/rejected": -2.7238385677337646,
|
| 100 |
+
"logps/chosen": -255.60317993164062,
|
| 101 |
+
"logps/rejected": -239.0897216796875,
|
| 102 |
+
"loss": 0.0075,
|
| 103 |
+
"rewards/accuracies": 0.6187499761581421,
|
| 104 |
+
"rewards/chosen": 0.03473116084933281,
|
| 105 |
+
"rewards/margins": 0.0994780957698822,
|
| 106 |
+
"rewards/rejected": -0.0647469311952591,
|
| 107 |
"step": 60
|
| 108 |
},
|
| 109 |
{
|
| 110 |
+
"epoch": 0.15,
|
| 111 |
+
"learning_rate": 4.967775735898179e-07,
|
| 112 |
+
"logits/chosen": -2.7161264419555664,
|
| 113 |
+
"logits/rejected": -2.6884379386901855,
|
| 114 |
+
"logps/chosen": -283.9285583496094,
|
| 115 |
+
"logps/rejected": -264.662353515625,
|
| 116 |
+
"loss": 0.0075,
|
| 117 |
+
"rewards/accuracies": 0.6312500238418579,
|
| 118 |
+
"rewards/chosen": -0.15505999326705933,
|
| 119 |
+
"rewards/margins": 0.1427268385887146,
|
| 120 |
+
"rewards/rejected": -0.29778680205345154,
|
| 121 |
"step": 70
|
| 122 |
},
|
| 123 |
{
|
| 124 |
+
"epoch": 0.17,
|
| 125 |
+
"learning_rate": 4.931986719649298e-07,
|
| 126 |
+
"logits/chosen": -2.660313129425049,
|
| 127 |
+
"logits/rejected": -2.6577725410461426,
|
| 128 |
+
"logps/chosen": -321.1432800292969,
|
| 129 |
+
"logps/rejected": -348.98193359375,
|
| 130 |
+
"loss": 0.0073,
|
| 131 |
+
"rewards/accuracies": 0.637499988079071,
|
| 132 |
+
"rewards/chosen": -0.5041145086288452,
|
| 133 |
+
"rewards/margins": 0.3104575276374817,
|
| 134 |
+
"rewards/rejected": -0.8145721554756165,
|
| 135 |
"step": 80
|
| 136 |
},
|
| 137 |
{
|
| 138 |
+
"epoch": 0.19,
|
| 139 |
+
"learning_rate": 4.883222001996351e-07,
|
| 140 |
+
"logits/chosen": -2.6707301139831543,
|
| 141 |
+
"logits/rejected": -2.640599489212036,
|
| 142 |
+
"logps/chosen": -345.442138671875,
|
| 143 |
+
"logps/rejected": -356.21124267578125,
|
| 144 |
+
"loss": 0.0072,
|
| 145 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 146 |
+
"rewards/chosen": -0.9047767519950867,
|
| 147 |
+
"rewards/margins": 0.31039169430732727,
|
| 148 |
+
"rewards/rejected": -1.2151683568954468,
|
| 149 |
"step": 90
|
| 150 |
},
|
| 151 |
{
|
| 152 |
+
"epoch": 0.21,
|
| 153 |
+
"learning_rate": 4.821741763807186e-07,
|
| 154 |
+
"logits/chosen": -2.6787619590759277,
|
| 155 |
+
"logits/rejected": -2.645983934402466,
|
| 156 |
+
"logps/chosen": -396.6234436035156,
|
| 157 |
+
"logps/rejected": -426.6803283691406,
|
| 158 |
+
"loss": 0.0071,
|
| 159 |
+
"rewards/accuracies": 0.606249988079071,
|
| 160 |
+
"rewards/chosen": -1.0718439817428589,
|
| 161 |
+
"rewards/margins": 0.5163585543632507,
|
| 162 |
+
"rewards/rejected": -1.588202714920044,
|
| 163 |
"step": 100
|
| 164 |
},
|
| 165 |
{
|
| 166 |
+
"epoch": 0.21,
|
| 167 |
+
"eval_logits/chosen": -2.648916244506836,
|
| 168 |
+
"eval_logits/rejected": -2.634218215942383,
|
| 169 |
+
"eval_logps/chosen": -346.26025390625,
|
| 170 |
+
"eval_logps/rejected": -418.1842956542969,
|
| 171 |
+
"eval_loss": 0.00723836338147521,
|
| 172 |
+
"eval_rewards/accuracies": 0.65234375,
|
| 173 |
+
"eval_rewards/chosen": -0.8922064900398254,
|
| 174 |
+
"eval_rewards/margins": 0.7161041498184204,
|
| 175 |
+
"eval_rewards/rejected": -1.6083106994628906,
|
| 176 |
+
"eval_runtime": 53.518,
|
| 177 |
+
"eval_samples_per_second": 37.371,
|
| 178 |
+
"eval_steps_per_second": 0.598,
|
| 179 |
+
"step": 100
|
| 180 |
+
},
|
| 181 |
+
{
|
| 182 |
+
"epoch": 0.23,
|
| 183 |
+
"learning_rate": 4.747874028753375e-07,
|
| 184 |
+
"logits/chosen": -2.545974016189575,
|
| 185 |
+
"logits/rejected": -2.5154786109924316,
|
| 186 |
+
"logps/chosen": -365.91827392578125,
|
| 187 |
+
"logps/rejected": -409.01031494140625,
|
| 188 |
+
"loss": 0.0071,
|
| 189 |
+
"rewards/accuracies": 0.6875,
|
| 190 |
+
"rewards/chosen": -0.9582077860832214,
|
| 191 |
+
"rewards/margins": 0.9138003587722778,
|
| 192 |
+
"rewards/rejected": -1.872007966041565,
|
| 193 |
"step": 110
|
| 194 |
},
|
| 195 |
{
|
| 196 |
+
"epoch": 0.25,
|
| 197 |
+
"learning_rate": 4.662012913161997e-07,
|
| 198 |
+
"logits/chosen": -2.684849500656128,
|
| 199 |
+
"logits/rejected": -2.6533303260803223,
|
| 200 |
+
"logps/chosen": -380.91168212890625,
|
| 201 |
+
"logps/rejected": -454.111328125,
|
| 202 |
+
"loss": 0.0071,
|
| 203 |
+
"rewards/accuracies": 0.65625,
|
| 204 |
+
"rewards/chosen": -1.1206060647964478,
|
| 205 |
+
"rewards/margins": 0.7919976711273193,
|
| 206 |
+
"rewards/rejected": -1.912603735923767,
|
| 207 |
"step": 120
|
| 208 |
},
|
| 209 |
{
|
| 210 |
+
"epoch": 0.27,
|
| 211 |
+
"learning_rate": 4.5646165232345103e-07,
|
| 212 |
+
"logits/chosen": -2.69043231010437,
|
| 213 |
+
"logits/rejected": -2.678968906402588,
|
| 214 |
+
"logps/chosen": -380.21856689453125,
|
| 215 |
+
"logps/rejected": -425.89697265625,
|
| 216 |
+
"loss": 0.0071,
|
| 217 |
+
"rewards/accuracies": 0.6312500238418579,
|
| 218 |
+
"rewards/chosen": -1.0572373867034912,
|
| 219 |
+
"rewards/margins": 0.6952055096626282,
|
| 220 |
+
"rewards/rejected": -1.752442717552185,
|
| 221 |
"step": 130
|
| 222 |
},
|
| 223 |
{
|
| 224 |
+
"epoch": 0.29,
|
| 225 |
+
"learning_rate": 4.456204510851956e-07,
|
| 226 |
+
"logits/chosen": -2.6275994777679443,
|
| 227 |
+
"logits/rejected": -2.6046979427337646,
|
| 228 |
+
"logps/chosen": -418.4617614746094,
|
| 229 |
+
"logps/rejected": -492.6680603027344,
|
| 230 |
+
"loss": 0.0071,
|
| 231 |
+
"rewards/accuracies": 0.65625,
|
| 232 |
+
"rewards/chosen": -1.3782728910446167,
|
| 233 |
+
"rewards/margins": 0.8407732248306274,
|
| 234 |
+
"rewards/rejected": -2.219046115875244,
|
| 235 |
"step": 140
|
| 236 |
},
|
| 237 |
{
|
| 238 |
+
"epoch": 0.31,
|
| 239 |
+
"learning_rate": 4.337355301007335e-07,
|
| 240 |
+
"logits/chosen": -2.5452637672424316,
|
| 241 |
+
"logits/rejected": -2.5649313926696777,
|
| 242 |
+
"logps/chosen": -324.2164611816406,
|
| 243 |
+
"logps/rejected": -459.1006774902344,
|
| 244 |
+
"loss": 0.007,
|
| 245 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 246 |
+
"rewards/chosen": -0.9753753542900085,
|
| 247 |
+
"rewards/margins": 1.252770185470581,
|
| 248 |
+
"rewards/rejected": -2.2281455993652344,
|
| 249 |
"step": 150
|
| 250 |
},
|
| 251 |
{
|
| 252 |
+
"epoch": 0.33,
|
| 253 |
+
"learning_rate": 4.2087030056579986e-07,
|
| 254 |
+
"logits/chosen": -2.5417354106903076,
|
| 255 |
+
"logits/rejected": -2.5152485370635986,
|
| 256 |
+
"logps/chosen": -401.59222412109375,
|
| 257 |
+
"logps/rejected": -472.24566650390625,
|
| 258 |
+
"loss": 0.0071,
|
| 259 |
+
"rewards/accuracies": 0.65625,
|
| 260 |
+
"rewards/chosen": -1.4562033414840698,
|
| 261 |
+
"rewards/margins": 0.8074793815612793,
|
| 262 |
+
"rewards/rejected": -2.2636826038360596,
|
| 263 |
"step": 160
|
| 264 |
},
|
| 265 |
{
|
| 266 |
+
"epoch": 0.36,
|
| 267 |
+
"learning_rate": 4.070934040463998e-07,
|
| 268 |
+
"logits/chosen": -2.4884097576141357,
|
| 269 |
+
"logits/rejected": -2.4738173484802246,
|
| 270 |
+
"logps/chosen": -423.27734375,
|
| 271 |
+
"logps/rejected": -517.5540771484375,
|
| 272 |
+
"loss": 0.007,
|
| 273 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 274 |
+
"rewards/chosen": -1.4111053943634033,
|
| 275 |
+
"rewards/margins": 1.162752389907837,
|
| 276 |
+
"rewards/rejected": -2.5738577842712402,
|
| 277 |
"step": 170
|
| 278 |
},
|
| 279 |
{
|
| 280 |
+
"epoch": 0.38,
|
| 281 |
+
"learning_rate": 3.9247834624635404e-07,
|
| 282 |
+
"logits/chosen": -2.5208473205566406,
|
| 283 |
+
"logits/rejected": -2.527306079864502,
|
| 284 |
+
"logps/chosen": -436.7098693847656,
|
| 285 |
+
"logps/rejected": -552.0220947265625,
|
| 286 |
+
"loss": 0.007,
|
| 287 |
+
"rewards/accuracies": 0.706250011920929,
|
| 288 |
+
"rewards/chosen": -1.6753193140029907,
|
| 289 |
+
"rewards/margins": 1.2910652160644531,
|
| 290 |
+
"rewards/rejected": -2.9663846492767334,
|
| 291 |
"step": 180
|
| 292 |
},
|
| 293 |
{
|
| 294 |
+
"epoch": 0.4,
|
| 295 |
+
"learning_rate": 3.7710310482256523e-07,
|
| 296 |
+
"logits/chosen": -2.5808041095733643,
|
| 297 |
+
"logits/rejected": -2.55265474319458,
|
| 298 |
+
"logps/chosen": -418.2471618652344,
|
| 299 |
+
"logps/rejected": -492.46044921875,
|
| 300 |
+
"loss": 0.0071,
|
| 301 |
+
"rewards/accuracies": 0.706250011920929,
|
| 302 |
+
"rewards/chosen": -1.2685142755508423,
|
| 303 |
+
"rewards/margins": 1.015337586402893,
|
| 304 |
+
"rewards/rejected": -2.2838518619537354,
|
| 305 |
"step": 190
|
| 306 |
},
|
| 307 |
{
|
| 308 |
+
"epoch": 0.42,
|
| 309 |
+
"learning_rate": 3.610497133404795e-07,
|
| 310 |
+
"logits/chosen": -2.6063413619995117,
|
| 311 |
+
"logits/rejected": -2.617413282394409,
|
| 312 |
+
"logps/chosen": -435.3294372558594,
|
| 313 |
+
"logps/rejected": -527.3897705078125,
|
| 314 |
+
"loss": 0.0068,
|
| 315 |
+
"rewards/accuracies": 0.6875,
|
| 316 |
+
"rewards/chosen": -1.3916200399398804,
|
| 317 |
+
"rewards/margins": 1.2391204833984375,
|
| 318 |
+
"rewards/rejected": -2.6307404041290283,
|
| 319 |
"step": 200
|
| 320 |
},
|
| 321 |
{
|
| 322 |
+
"epoch": 0.42,
|
| 323 |
+
"eval_logits/chosen": -2.621898651123047,
|
| 324 |
+
"eval_logits/rejected": -2.605832576751709,
|
| 325 |
+
"eval_logps/chosen": -430.9748840332031,
|
| 326 |
+
"eval_logps/rejected": -577.05419921875,
|
| 327 |
+
"eval_loss": 0.0069780959747731686,
|
| 328 |
+
"eval_rewards/accuracies": 0.66796875,
|
| 329 |
+
"eval_rewards/chosen": -1.7393522262573242,
|
| 330 |
+
"eval_rewards/margins": 1.4576572179794312,
|
| 331 |
+
"eval_rewards/rejected": -3.197009325027466,
|
| 332 |
+
"eval_runtime": 53.4259,
|
| 333 |
+
"eval_samples_per_second": 37.435,
|
| 334 |
+
"eval_steps_per_second": 0.599,
|
| 335 |
+
"step": 200
|
| 336 |
+
},
|
| 337 |
+
{
|
| 338 |
+
"epoch": 0.44,
|
| 339 |
+
"learning_rate": 3.4440382358952115e-07,
|
| 340 |
+
"logits/chosen": -2.551345109939575,
|
| 341 |
+
"logits/rejected": -2.5264601707458496,
|
| 342 |
+
"logps/chosen": -426.4783630371094,
|
| 343 |
+
"logps/rejected": -572.7852783203125,
|
| 344 |
+
"loss": 0.0068,
|
| 345 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 346 |
+
"rewards/chosen": -1.5344517230987549,
|
| 347 |
+
"rewards/margins": 1.6740432977676392,
|
| 348 |
+
"rewards/rejected": -3.2084953784942627,
|
| 349 |
"step": 210
|
| 350 |
},
|
| 351 |
{
|
| 352 |
+
"epoch": 0.46,
|
| 353 |
+
"learning_rate": 3.272542485937368e-07,
|
| 354 |
+
"logits/chosen": -2.5313520431518555,
|
| 355 |
+
"logits/rejected": -2.48645281791687,
|
| 356 |
+
"logps/chosen": -433.2318420410156,
|
| 357 |
+
"logps/rejected": -535.4318237304688,
|
| 358 |
+
"loss": 0.0069,
|
| 359 |
+
"rewards/accuracies": 0.637499988079071,
|
| 360 |
+
"rewards/chosen": -1.5826613903045654,
|
| 361 |
+
"rewards/margins": 1.148589849472046,
|
| 362 |
+
"rewards/rejected": -2.7312510013580322,
|
| 363 |
"step": 220
|
| 364 |
},
|
| 365 |
{
|
| 366 |
+
"epoch": 0.48,
|
| 367 |
+
"learning_rate": 3.096924887558854e-07,
|
| 368 |
+
"logits/chosen": -2.553248167037964,
|
| 369 |
+
"logits/rejected": -2.4987640380859375,
|
| 370 |
+
"logps/chosen": -463.6939392089844,
|
| 371 |
+
"logps/rejected": -559.91650390625,
|
| 372 |
+
"loss": 0.0072,
|
| 373 |
+
"rewards/accuracies": 0.6312500238418579,
|
| 374 |
+
"rewards/chosen": -1.8914604187011719,
|
| 375 |
+
"rewards/margins": 1.0075489282608032,
|
| 376 |
+
"rewards/rejected": -2.8990092277526855,
|
| 377 |
"step": 230
|
| 378 |
},
|
| 379 |
{
|
| 380 |
+
"epoch": 0.5,
|
| 381 |
+
"learning_rate": 2.9181224366319943e-07,
|
| 382 |
+
"logits/chosen": -2.509253978729248,
|
| 383 |
+
"logits/rejected": -2.4721405506134033,
|
| 384 |
+
"logps/chosen": -506.30328369140625,
|
| 385 |
+
"logps/rejected": -608.6563720703125,
|
| 386 |
+
"loss": 0.0068,
|
| 387 |
+
"rewards/accuracies": 0.643750011920929,
|
| 388 |
+
"rewards/chosen": -2.2013320922851562,
|
| 389 |
+
"rewards/margins": 1.3522661924362183,
|
| 390 |
+
"rewards/rejected": -3.553598403930664,
|
| 391 |
"step": 240
|
| 392 |
},
|
| 393 |
{
|
| 394 |
+
"epoch": 0.52,
|
| 395 |
+
"learning_rate": 2.7370891215954565e-07,
|
| 396 |
+
"logits/chosen": -2.567660331726074,
|
| 397 |
+
"logits/rejected": -2.519611358642578,
|
| 398 |
+
"logps/chosen": -446.6778259277344,
|
| 399 |
+
"logps/rejected": -590.0146484375,
|
| 400 |
+
"loss": 0.0071,
|
| 401 |
+
"rewards/accuracies": 0.65625,
|
| 402 |
+
"rewards/chosen": -1.7510559558868408,
|
| 403 |
+
"rewards/margins": 1.6048589944839478,
|
| 404 |
+
"rewards/rejected": -3.355915069580078,
|
| 405 |
"step": 250
|
| 406 |
},
|
| 407 |
{
|
| 408 |
+
"epoch": 0.54,
|
| 409 |
+
"learning_rate": 2.55479083351317e-07,
|
| 410 |
+
"logits/chosen": -2.560824155807495,
|
| 411 |
+
"logits/rejected": -2.5370585918426514,
|
| 412 |
+
"logps/chosen": -453.1175842285156,
|
| 413 |
+
"logps/rejected": -639.9363403320312,
|
| 414 |
+
"loss": 0.0069,
|
| 415 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 416 |
+
"rewards/chosen": -1.9397010803222656,
|
| 417 |
+
"rewards/margins": 1.8772062063217163,
|
| 418 |
+
"rewards/rejected": -3.8169074058532715,
|
| 419 |
"step": 260
|
| 420 |
},
|
| 421 |
{
|
| 422 |
+
"epoch": 0.56,
|
| 423 |
+
"learning_rate": 2.3722002126275822e-07,
|
| 424 |
+
"logits/chosen": -2.5935356616973877,
|
| 425 |
+
"logits/rejected": -2.5652170181274414,
|
| 426 |
+
"logps/chosen": -414.54248046875,
|
| 427 |
+
"logps/rejected": -655.6973876953125,
|
| 428 |
+
"loss": 0.007,
|
| 429 |
+
"rewards/accuracies": 0.75,
|
| 430 |
+
"rewards/chosen": -1.2608954906463623,
|
| 431 |
+
"rewards/margins": 2.6454343795776367,
|
| 432 |
+
"rewards/rejected": -3.906330108642578,
|
| 433 |
"step": 270
|
| 434 |
},
|
| 435 |
{
|
| 436 |
+
"epoch": 0.59,
|
| 437 |
+
"learning_rate": 2.19029145890313e-07,
|
| 438 |
+
"logits/chosen": -2.5211143493652344,
|
| 439 |
+
"logits/rejected": -2.514479875564575,
|
| 440 |
+
"logps/chosen": -458.656982421875,
|
| 441 |
+
"logps/rejected": -574.6309814453125,
|
| 442 |
+
"loss": 0.007,
|
| 443 |
+
"rewards/accuracies": 0.762499988079071,
|
| 444 |
+
"rewards/chosen": -1.5091394186019897,
|
| 445 |
+
"rewards/margins": 1.5814958810806274,
|
| 446 |
+
"rewards/rejected": -3.090635299682617,
|
| 447 |
"step": 280
|
| 448 |
},
|
| 449 |
{
|
| 450 |
+
"epoch": 0.61,
|
| 451 |
+
"learning_rate": 2.0100351342479216e-07,
|
| 452 |
+
"logits/chosen": -2.5380799770355225,
|
| 453 |
+
"logits/rejected": -2.5039894580841064,
|
| 454 |
+
"logps/chosen": -482.7261657714844,
|
| 455 |
+
"logps/rejected": -624.1588745117188,
|
| 456 |
+
"loss": 0.0069,
|
| 457 |
+
"rewards/accuracies": 0.699999988079071,
|
| 458 |
+
"rewards/chosen": -2.049624443054199,
|
| 459 |
+
"rewards/margins": 1.658935785293579,
|
| 460 |
+
"rewards/rejected": -3.7085604667663574,
|
| 461 |
"step": 290
|
| 462 |
},
|
| 463 |
{
|
| 464 |
+
"epoch": 0.63,
|
| 465 |
+
"learning_rate": 1.8323929841460178e-07,
|
| 466 |
+
"logits/chosen": -2.5519766807556152,
|
| 467 |
+
"logits/rejected": -2.5217361450195312,
|
| 468 |
+
"logps/chosen": -523.7749633789062,
|
| 469 |
+
"logps/rejected": -727.1653442382812,
|
| 470 |
+
"loss": 0.0069,
|
| 471 |
+
"rewards/accuracies": 0.6875,
|
| 472 |
+
"rewards/chosen": -2.7216830253601074,
|
| 473 |
+
"rewards/margins": 2.1266276836395264,
|
| 474 |
+
"rewards/rejected": -4.848310947418213,
|
| 475 |
"step": 300
|
| 476 |
},
|
| 477 |
{
|
| 478 |
+
"epoch": 0.63,
|
| 479 |
+
"eval_logits/chosen": -2.535085916519165,
|
| 480 |
+
"eval_logits/rejected": -2.5115010738372803,
|
| 481 |
+
"eval_logps/chosen": -480.6185302734375,
|
| 482 |
+
"eval_logps/rejected": -720.874755859375,
|
| 483 |
+
"eval_loss": 0.006893348880112171,
|
| 484 |
+
"eval_rewards/accuracies": 0.69921875,
|
| 485 |
+
"eval_rewards/chosen": -2.2357888221740723,
|
| 486 |
+
"eval_rewards/margins": 2.3994250297546387,
|
| 487 |
+
"eval_rewards/rejected": -4.635214328765869,
|
| 488 |
+
"eval_runtime": 53.4457,
|
| 489 |
+
"eval_samples_per_second": 37.421,
|
| 490 |
+
"eval_steps_per_second": 0.599,
|
| 491 |
+
"step": 300
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"epoch": 0.65,
|
| 495 |
+
"learning_rate": 1.6583128063291573e-07,
|
| 496 |
+
"logits/chosen": -2.52044415473938,
|
| 497 |
+
"logits/rejected": -2.473792552947998,
|
| 498 |
+
"logps/chosen": -516.9593505859375,
|
| 499 |
+
"logps/rejected": -765.6558837890625,
|
| 500 |
+
"loss": 0.0069,
|
| 501 |
+
"rewards/accuracies": 0.699999988079071,
|
| 502 |
+
"rewards/chosen": -2.2925307750701904,
|
| 503 |
+
"rewards/margins": 2.806295871734619,
|
| 504 |
+
"rewards/rejected": -5.0988264083862305,
|
| 505 |
"step": 310
|
| 506 |
},
|
| 507 |
{
|
| 508 |
+
"epoch": 0.67,
|
| 509 |
+
"learning_rate": 1.488723393865766e-07,
|
| 510 |
+
"logits/chosen": -2.486896514892578,
|
| 511 |
+
"logits/rejected": -2.45188570022583,
|
| 512 |
+
"logps/chosen": -513.2415771484375,
|
| 513 |
+
"logps/rejected": -736.2615356445312,
|
| 514 |
+
"loss": 0.0068,
|
| 515 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 516 |
+
"rewards/chosen": -2.2738022804260254,
|
| 517 |
+
"rewards/margins": 2.435603618621826,
|
| 518 |
+
"rewards/rejected": -4.709405899047852,
|
| 519 |
"step": 320
|
| 520 |
},
|
| 521 |
{
|
| 522 |
+
"epoch": 0.69,
|
| 523 |
+
"learning_rate": 1.3245295796480788e-07,
|
| 524 |
+
"logits/chosen": -2.5038528442382812,
|
| 525 |
+
"logits/rejected": -2.45892596244812,
|
| 526 |
+
"logps/chosen": -482.16387939453125,
|
| 527 |
+
"logps/rejected": -711.6837768554688,
|
| 528 |
+
"loss": 0.0069,
|
| 529 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 530 |
+
"rewards/chosen": -2.0813446044921875,
|
| 531 |
+
"rewards/margins": 2.3931524753570557,
|
| 532 |
+
"rewards/rejected": -4.474496841430664,
|
| 533 |
"step": 330
|
| 534 |
},
|
| 535 |
{
|
| 536 |
+
"epoch": 0.71,
|
| 537 |
+
"learning_rate": 1.1666074087171627e-07,
|
| 538 |
+
"logits/chosen": -2.5274105072021484,
|
| 539 |
+
"logits/rejected": -2.496708869934082,
|
| 540 |
+
"logps/chosen": -485.33062744140625,
|
| 541 |
+
"logps/rejected": -762.1112060546875,
|
| 542 |
+
"loss": 0.0069,
|
| 543 |
+
"rewards/accuracies": 0.6875,
|
| 544 |
+
"rewards/chosen": -2.307344675064087,
|
| 545 |
+
"rewards/margins": 2.784510612487793,
|
| 546 |
+
"rewards/rejected": -5.091855525970459,
|
| 547 |
"step": 340
|
| 548 |
},
|
| 549 |
{
|
| 550 |
+
"epoch": 0.73,
|
| 551 |
+
"learning_rate": 1.0157994641835734e-07,
|
| 552 |
+
"logits/chosen": -2.478060483932495,
|
| 553 |
+
"logits/rejected": -2.456634998321533,
|
| 554 |
+
"logps/chosen": -641.193359375,
|
| 555 |
+
"logps/rejected": -830.1126098632812,
|
| 556 |
+
"loss": 0.0071,
|
| 557 |
+
"rewards/accuracies": 0.612500011920929,
|
| 558 |
+
"rewards/chosen": -3.647397994995117,
|
| 559 |
+
"rewards/margins": 2.023047924041748,
|
| 560 |
+
"rewards/rejected": -5.670445442199707,
|
| 561 |
"step": 350
|
| 562 |
},
|
| 563 |
{
|
| 564 |
+
"epoch": 0.75,
|
| 565 |
+
"learning_rate": 8.729103716819111e-08,
|
| 566 |
+
"logits/chosen": -2.452293634414673,
|
| 567 |
+
"logits/rejected": -2.414788007736206,
|
| 568 |
+
"logps/chosen": -563.0314331054688,
|
| 569 |
+
"logps/rejected": -946.1033325195312,
|
| 570 |
+
"loss": 0.0068,
|
| 571 |
+
"rewards/accuracies": 0.71875,
|
| 572 |
+
"rewards/chosen": -3.0464682579040527,
|
| 573 |
+
"rewards/margins": 3.980926036834717,
|
| 574 |
+
"rewards/rejected": -7.0273942947387695,
|
| 575 |
"step": 360
|
| 576 |
},
|
| 577 |
{
|
| 578 |
+
"epoch": 0.77,
|
| 579 |
+
"learning_rate": 7.387025063449081e-08,
|
| 580 |
+
"logits/chosen": -2.489593982696533,
|
| 581 |
+
"logits/rejected": -2.473261594772339,
|
| 582 |
+
"logps/chosen": -690.0401000976562,
|
| 583 |
+
"logps/rejected": -933.4720458984375,
|
| 584 |
+
"loss": 0.0068,
|
| 585 |
+
"rewards/accuracies": 0.643750011920929,
|
| 586 |
+
"rewards/chosen": -4.119799613952637,
|
| 587 |
+
"rewards/margins": 2.5783958435058594,
|
| 588 |
+
"rewards/rejected": -6.698195457458496,
|
| 589 |
"step": 370
|
| 590 |
},
|
| 591 |
{
|
| 592 |
+
"epoch": 0.79,
|
| 593 |
+
"learning_rate": 6.138919252022435e-08,
|
| 594 |
+
"logits/chosen": -2.421910047531128,
|
| 595 |
+
"logits/rejected": -2.4058310985565186,
|
| 596 |
+
"logps/chosen": -612.6002807617188,
|
| 597 |
+
"logps/rejected": -844.525390625,
|
| 598 |
+
"loss": 0.0069,
|
| 599 |
+
"rewards/accuracies": 0.625,
|
| 600 |
+
"rewards/chosen": -3.8975918292999268,
|
| 601 |
+
"rewards/margins": 2.118032693862915,
|
| 602 |
+
"rewards/rejected": -6.015624046325684,
|
| 603 |
"step": 380
|
| 604 |
},
|
| 605 |
{
|
| 606 |
+
"epoch": 0.82,
|
| 607 |
+
"learning_rate": 4.991445467064689e-08,
|
| 608 |
+
"logits/chosen": -2.492781400680542,
|
| 609 |
+
"logits/rejected": -2.4609475135803223,
|
| 610 |
+
"logps/chosen": -508.40472412109375,
|
| 611 |
+
"logps/rejected": -797.5465087890625,
|
| 612 |
+
"loss": 0.007,
|
| 613 |
+
"rewards/accuracies": 0.699999988079071,
|
| 614 |
+
"rewards/chosen": -2.568449020385742,
|
| 615 |
+
"rewards/margins": 3.0519707202911377,
|
| 616 |
+
"rewards/rejected": -5.620419502258301,
|
| 617 |
"step": 390
|
| 618 |
},
|
| 619 |
{
|
| 620 |
+
"epoch": 0.84,
|
| 621 |
+
"learning_rate": 3.9507259776993954e-08,
|
| 622 |
+
"logits/chosen": -2.4738097190856934,
|
| 623 |
+
"logits/rejected": -2.4266116619110107,
|
| 624 |
+
"logps/chosen": -588.7605590820312,
|
| 625 |
+
"logps/rejected": -691.3428344726562,
|
| 626 |
+
"loss": 0.0067,
|
| 627 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 628 |
+
"rewards/chosen": -3.0305678844451904,
|
| 629 |
+
"rewards/margins": 1.4075567722320557,
|
| 630 |
+
"rewards/rejected": -4.438124656677246,
|
| 631 |
+
"step": 400
|
| 632 |
+
},
|
| 633 |
+
{
|
| 634 |
+
"epoch": 0.84,
|
| 635 |
+
"eval_logits/chosen": -2.5238196849823,
|
| 636 |
+
"eval_logits/rejected": -2.498831272125244,
|
| 637 |
+
"eval_logps/chosen": -522.7908935546875,
|
| 638 |
+
"eval_logps/rejected": -820.4591674804688,
|
| 639 |
+
"eval_loss": 0.006906786467880011,
|
| 640 |
+
"eval_rewards/accuracies": 0.69140625,
|
| 641 |
+
"eval_rewards/chosen": -2.6575121879577637,
|
| 642 |
+
"eval_rewards/margins": 2.9735467433929443,
|
| 643 |
+
"eval_rewards/rejected": -5.631058692932129,
|
| 644 |
+
"eval_runtime": 53.4628,
|
| 645 |
+
"eval_samples_per_second": 37.409,
|
| 646 |
+
"eval_steps_per_second": 0.599,
|
| 647 |
"step": 400
|
| 648 |
},
|
| 649 |
+
{
|
| 650 |
+
"epoch": 0.86,
|
| 651 |
+
"learning_rate": 3.022313472693447e-08,
|
| 652 |
+
"logits/chosen": -2.507258653640747,
|
| 653 |
+
"logits/rejected": -2.4417152404785156,
|
| 654 |
+
"logps/chosen": -605.7719116210938,
|
| 655 |
+
"logps/rejected": -805.2724609375,
|
| 656 |
+
"loss": 0.0069,
|
| 657 |
+
"rewards/accuracies": 0.637499988079071,
|
| 658 |
+
"rewards/chosen": -3.4218215942382812,
|
| 659 |
+
"rewards/margins": 1.8162673711776733,
|
| 660 |
+
"rewards/rejected": -5.238089084625244,
|
| 661 |
+
"step": 410
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"epoch": 0.88,
|
| 665 |
+
"learning_rate": 2.2111614344599684e-08,
|
| 666 |
+
"logits/chosen": -2.510485887527466,
|
| 667 |
+
"logits/rejected": -2.476248264312744,
|
| 668 |
+
"logps/chosen": -672.9176025390625,
|
| 669 |
+
"logps/rejected": -961.6502075195312,
|
| 670 |
+
"loss": 0.0068,
|
| 671 |
+
"rewards/accuracies": 0.6312500238418579,
|
| 672 |
+
"rewards/chosen": -3.8610453605651855,
|
| 673 |
+
"rewards/margins": 2.831366539001465,
|
| 674 |
+
"rewards/rejected": -6.692412376403809,
|
| 675 |
+
"step": 420
|
| 676 |
+
},
|
| 677 |
+
{
|
| 678 |
+
"epoch": 0.9,
|
| 679 |
+
"learning_rate": 1.521597710086439e-08,
|
| 680 |
+
"logits/chosen": -2.4765820503234863,
|
| 681 |
+
"logits/rejected": -2.446441173553467,
|
| 682 |
+
"logps/chosen": -610.9924926757812,
|
| 683 |
+
"logps/rejected": -819.8570556640625,
|
| 684 |
+
"loss": 0.0068,
|
| 685 |
+
"rewards/accuracies": 0.637499988079071,
|
| 686 |
+
"rewards/chosen": -3.207393169403076,
|
| 687 |
+
"rewards/margins": 2.247278928756714,
|
| 688 |
+
"rewards/rejected": -5.454671859741211,
|
| 689 |
+
"step": 430
|
| 690 |
+
},
|
| 691 |
+
{
|
| 692 |
+
"epoch": 0.92,
|
| 693 |
+
"learning_rate": 9.57301420397924e-09,
|
| 694 |
+
"logits/chosen": -2.5026493072509766,
|
| 695 |
+
"logits/rejected": -2.4737229347229004,
|
| 696 |
+
"logps/chosen": -692.0069580078125,
|
| 697 |
+
"logps/rejected": -766.7494506835938,
|
| 698 |
+
"loss": 0.0067,
|
| 699 |
+
"rewards/accuracies": 0.606249988079071,
|
| 700 |
+
"rewards/chosen": -3.769559383392334,
|
| 701 |
+
"rewards/margins": 1.2420024871826172,
|
| 702 |
+
"rewards/rejected": -5.011561870574951,
|
| 703 |
+
"step": 440
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 0.94,
|
| 707 |
+
"learning_rate": 5.212833302556258e-09,
|
| 708 |
+
"logits/chosen": -2.4708807468414307,
|
| 709 |
+
"logits/rejected": -2.445148468017578,
|
| 710 |
+
"logps/chosen": -619.1263427734375,
|
| 711 |
+
"logps/rejected": -817.3858032226562,
|
| 712 |
+
"loss": 0.0068,
|
| 713 |
+
"rewards/accuracies": 0.612500011920929,
|
| 714 |
+
"rewards/chosen": -3.3826870918273926,
|
| 715 |
+
"rewards/margins": 2.2892913818359375,
|
| 716 |
+
"rewards/rejected": -5.671978950500488,
|
| 717 |
+
"step": 450
|
| 718 |
+
},
|
| 719 |
+
{
|
| 720 |
+
"epoch": 0.96,
|
| 721 |
+
"learning_rate": 2.158697848236607e-09,
|
| 722 |
+
"logits/chosen": -2.4850497245788574,
|
| 723 |
+
"logits/rejected": -2.4442174434661865,
|
| 724 |
+
"logps/chosen": -721.6383056640625,
|
| 725 |
+
"logps/rejected": -822.6937255859375,
|
| 726 |
+
"loss": 0.0068,
|
| 727 |
+
"rewards/accuracies": 0.625,
|
| 728 |
+
"rewards/chosen": -4.212514400482178,
|
| 729 |
+
"rewards/margins": 1.3080599308013916,
|
| 730 |
+
"rewards/rejected": -5.520575046539307,
|
| 731 |
+
"step": 460
|
| 732 |
+
},
|
| 733 |
+
{
|
| 734 |
+
"epoch": 0.98,
|
| 735 |
+
"learning_rate": 4.269029751107489e-10,
|
| 736 |
+
"logits/chosen": -2.503718852996826,
|
| 737 |
+
"logits/rejected": -2.4400999546051025,
|
| 738 |
+
"logps/chosen": -659.8447875976562,
|
| 739 |
+
"logps/rejected": -753.3057861328125,
|
| 740 |
+
"loss": 0.0067,
|
| 741 |
+
"rewards/accuracies": 0.59375,
|
| 742 |
+
"rewards/chosen": -3.6163833141326904,
|
| 743 |
+
"rewards/margins": 1.5862648487091064,
|
| 744 |
+
"rewards/rejected": -5.202649116516113,
|
| 745 |
+
"step": 470
|
| 746 |
+
},
|
| 747 |
{
|
| 748 |
"epoch": 1.0,
|
| 749 |
+
"step": 478,
|
| 750 |
"total_flos": 0.0,
|
| 751 |
+
"train_loss": 0.007060731609184373,
|
| 752 |
+
"train_runtime": 4232.7551,
|
| 753 |
+
"train_samples_per_second": 14.443,
|
| 754 |
+
"train_steps_per_second": 0.113
|
| 755 |
}
|
| 756 |
],
|
| 757 |
"logging_steps": 10,
|
| 758 |
+
"max_steps": 478,
|
| 759 |
"num_train_epochs": 1,
|
| 760 |
+
"save_steps": 100,
|
| 761 |
"total_flos": 0.0,
|
| 762 |
"trial_name": null,
|
| 763 |
"trial_params": null
|