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 +615 -453
- training_args.bin +1 -1
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: 0.2130
|
| 21 |
+
- Rewards/chosen: -0.8895
|
| 22 |
+
- Rewards/rejected: -1.4961
|
| 23 |
+
- Rewards/accuracies: 0.7070
|
| 24 |
+
- Rewards/margins: 0.6066
|
| 25 |
+
- Logps/rejected: -406.9670
|
| 26 |
+
- Logps/chosen: -345.9897
|
| 27 |
+
- Logits/rejected: -2.3161
|
| 28 |
+
- Logits/chosen: -2.3381
|
| 29 |
|
| 30 |
## Model description
|
| 31 |
|
|
|
|
| 47 |
- learning_rate: 5e-07
|
| 48 |
- train_batch_size: 8
|
| 49 |
- eval_batch_size: 8
|
| 50 |
+
- seed: 4
|
| 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 |
+
| 0.2738 | 0.21 | 100 | 0.2945 | -0.3021 | -0.6399 | 0.7148 | 0.3378 | -321.3399 | -287.2470 | -2.6731 | -2.6882 |
|
| 66 |
+
| 0.1875 | 0.42 | 200 | 0.2041 | -0.9537 | -1.4882 | 0.7227 | 0.5345 | -406.1715 | -352.4083 | -2.3390 | -2.3602 |
|
| 67 |
+
| 0.2103 | 0.63 | 300 | 0.2195 | -0.8401 | -1.4223 | 0.7070 | 0.5822 | -399.5840 | -341.0494 | -2.3533 | -2.3750 |
|
| 68 |
+
| 0.2051 | 0.84 | 400 | 0.2130 | -0.8895 | -1.4961 | 0.7070 | 0.6066 | -406.9670 | -345.9897 | -2.3161 | -2.3381 |
|
| 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": 0.24355335988758997,
|
| 4 |
+
"train_runtime": 3944.1561,
|
| 5 |
+
"train_samples": 61134,
|
| 6 |
+
"train_samples_per_second": 15.5,
|
| 7 |
+
"train_steps_per_second": 0.121
|
| 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:d29920218c4a133ce76bacf8aa2f4ff53ad258527de6d8efd5893da84fc28ec7
|
| 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 sha256:61c2e9bfab01e957057b099136868b877e15e7a35dba4def807e0f8f92bce022
|
| 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 sha256:826a87bf163d8df999f71d47df0d236c315eb251119b93437a07eae548662e31
|
| 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": 0.24355335988758997,
|
| 4 |
+
"train_runtime": 3944.1561,
|
| 5 |
+
"train_samples": 61134,
|
| 6 |
+
"train_samples_per_second": 15.5,
|
| 7 |
+
"train_steps_per_second": 0.121
|
| 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,
|
|
@@ -24,576 +24,738 @@
|
|
| 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":
|
| 35 |
-
"rewards/margins": 0.
|
| 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": -0.
|
| 91 |
-
"rewards/margins": 0.
|
| 92 |
-
"rewards/rejected": -0.
|
| 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": -0.
|
| 105 |
-
"rewards/margins": 0.
|
| 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": -0.
|
| 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": -0.
|
| 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": -0.
|
| 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": -0.
|
| 217 |
-
"rewards/margins":
|
| 218 |
-
"rewards/rejected": -
|
| 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":
|
| 232 |
-
"rewards/rejected": -
|
| 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": -0.
|
| 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": -0.
|
| 259 |
-
"rewards/margins":
|
| 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": -0.
|
| 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": -0.
|
| 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": -
|
| 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": -0.
|
| 329 |
-
"rewards/margins":
|
| 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": -0.
|
| 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": -0.
|
| 357 |
-
"rewards/margins":
|
| 358 |
-
"rewards/rejected": -
|
| 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": -0.
|
| 371 |
-
"rewards/margins":
|
| 372 |
-
"rewards/rejected": -
|
| 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": -0.
|
| 385 |
-
"rewards/margins":
|
| 386 |
-
"rewards/rejected": -
|
| 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": -0.
|
| 399 |
-
"rewards/margins":
|
| 400 |
-
"rewards/rejected": -
|
| 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": -
|
| 413 |
-
"rewards/margins":
|
| 414 |
-
"rewards/rejected": -
|
| 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": -
|
| 429 |
"step": 290
|
| 430 |
},
|
| 431 |
{
|
| 432 |
-
"epoch": 0.
|
| 433 |
-
"learning_rate":
|
| 434 |
-
"logits/chosen": -2.
|
| 435 |
-
"logits/rejected": -2.
|
| 436 |
-
"logps/chosen": -
|
| 437 |
-
"logps/rejected": -
|
| 438 |
-
"loss": 0.
|
| 439 |
-
"rewards/accuracies": 0.
|
| 440 |
-
"rewards/chosen": -0.
|
| 441 |
-
"rewards/margins":
|
| 442 |
-
"rewards/rejected": -
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 443 |
"step": 300
|
| 444 |
},
|
| 445 |
{
|
| 446 |
-
"epoch": 0.
|
| 447 |
-
"learning_rate":
|
| 448 |
-
"logits/chosen": -2.
|
| 449 |
-
"logits/rejected": -2.
|
| 450 |
-
"logps/chosen": -
|
| 451 |
-
"logps/rejected": -
|
| 452 |
-
"loss": 0.
|
| 453 |
-
"rewards/accuracies": 0.
|
| 454 |
-
"rewards/chosen": -0.
|
| 455 |
-
"rewards/margins":
|
| 456 |
-
"rewards/rejected": -
|
| 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":
|
| 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": -0.
|
| 497 |
-
"rewards/margins":
|
| 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": -0.
|
| 511 |
-
"rewards/margins":
|
| 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": -0.
|
| 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": 0.7562500238418579,
|
| 538 |
-
"rewards/chosen": -0.
|
| 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": -0.
|
| 553 |
-
"rewards/margins":
|
| 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": -0.
|
| 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": 1000,
|
| 599 |
"total_flos": 0.0,
|
|
|
|
| 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": -2.7386245727539062,
|
| 15 |
+
"logits/rejected": -2.7273669242858887,
|
| 16 |
+
"logps/chosen": -262.8376159667969,
|
| 17 |
+
"logps/rejected": -255.88758850097656,
|
| 18 |
+
"loss": 0.3809,
|
| 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": -2.7419159412384033,
|
| 29 |
+
"logits/rejected": -2.73586106300354,
|
| 30 |
+
"logps/chosen": -305.9677429199219,
|
| 31 |
+
"logps/rejected": -270.49627685546875,
|
| 32 |
+
"loss": 0.3844,
|
| 33 |
+
"rewards/accuracies": 0.4236111044883728,
|
| 34 |
+
"rewards/chosen": -8.022645488381386e-06,
|
| 35 |
+
"rewards/margins": -0.00040632300078868866,
|
| 36 |
+
"rewards/rejected": 0.0003983003261964768,
|
| 37 |
"step": 10
|
| 38 |
},
|
| 39 |
{
|
| 40 |
+
"epoch": 0.04,
|
| 41 |
+
"learning_rate": 2.0833333333333333e-07,
|
| 42 |
+
"logits/chosen": -2.797684907913208,
|
| 43 |
+
"logits/rejected": -2.7782931327819824,
|
| 44 |
+
"logps/chosen": -296.00689697265625,
|
| 45 |
+
"logps/rejected": -258.25726318359375,
|
| 46 |
+
"loss": 0.3829,
|
| 47 |
+
"rewards/accuracies": 0.606249988079071,
|
| 48 |
+
"rewards/chosen": 0.0003913758264388889,
|
| 49 |
+
"rewards/margins": 0.0022823966573923826,
|
| 50 |
+
"rewards/rejected": -0.0018910206854343414,
|
| 51 |
"step": 20
|
| 52 |
},
|
| 53 |
{
|
| 54 |
+
"epoch": 0.06,
|
| 55 |
+
"learning_rate": 3.1249999999999997e-07,
|
| 56 |
+
"logits/chosen": -2.8223657608032227,
|
| 57 |
+
"logits/rejected": -2.794595718383789,
|
| 58 |
+
"logps/chosen": -300.7593078613281,
|
| 59 |
+
"logps/rejected": -259.7913513183594,
|
| 60 |
+
"loss": 0.3883,
|
| 61 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 62 |
+
"rewards/chosen": -0.0001883889053715393,
|
| 63 |
+
"rewards/margins": 0.009234221652150154,
|
| 64 |
+
"rewards/rejected": -0.009422610513865948,
|
| 65 |
"step": 30
|
| 66 |
},
|
| 67 |
{
|
| 68 |
+
"epoch": 0.08,
|
| 69 |
+
"learning_rate": 4.1666666666666667e-07,
|
| 70 |
+
"logits/chosen": -2.757336139678955,
|
| 71 |
+
"logits/rejected": -2.73185396194458,
|
| 72 |
+
"logps/chosen": -257.5922546386719,
|
| 73 |
+
"logps/rejected": -249.6249237060547,
|
| 74 |
+
"loss": 0.3746,
|
| 75 |
+
"rewards/accuracies": 0.59375,
|
| 76 |
+
"rewards/chosen": -0.006459876894950867,
|
| 77 |
+
"rewards/margins": 0.02206435427069664,
|
| 78 |
+
"rewards/rejected": -0.028524229303002357,
|
| 79 |
"step": 40
|
| 80 |
},
|
| 81 |
{
|
| 82 |
+
"epoch": 0.1,
|
| 83 |
+
"learning_rate": 4.999733114418725e-07,
|
| 84 |
+
"logits/chosen": -2.7224113941192627,
|
| 85 |
+
"logits/rejected": -2.710090160369873,
|
| 86 |
+
"logps/chosen": -256.6429138183594,
|
| 87 |
+
"logps/rejected": -250.55154418945312,
|
| 88 |
+
"loss": 0.3673,
|
| 89 |
+
"rewards/accuracies": 0.6875,
|
| 90 |
+
"rewards/chosen": -0.03372717648744583,
|
| 91 |
+
"rewards/margins": 0.0691945031285286,
|
| 92 |
+
"rewards/rejected": -0.10292167961597443,
|
| 93 |
"step": 50
|
| 94 |
},
|
| 95 |
{
|
| 96 |
+
"epoch": 0.13,
|
| 97 |
+
"learning_rate": 4.990398100856366e-07,
|
| 98 |
+
"logits/chosen": -2.6469366550445557,
|
| 99 |
+
"logits/rejected": -2.6182854175567627,
|
| 100 |
+
"logps/chosen": -267.1353454589844,
|
| 101 |
+
"logps/rejected": -257.6127014160156,
|
| 102 |
+
"loss": 0.3612,
|
| 103 |
+
"rewards/accuracies": 0.612500011920929,
|
| 104 |
+
"rewards/chosen": -0.10208804905414581,
|
| 105 |
+
"rewards/margins": 0.08831791579723358,
|
| 106 |
+
"rewards/rejected": -0.1904059499502182,
|
| 107 |
"step": 60
|
| 108 |
},
|
| 109 |
{
|
| 110 |
+
"epoch": 0.15,
|
| 111 |
+
"learning_rate": 4.967775735898179e-07,
|
| 112 |
+
"logits/chosen": -2.662863254547119,
|
| 113 |
+
"logits/rejected": -2.6449124813079834,
|
| 114 |
+
"logps/chosen": -291.5115051269531,
|
| 115 |
+
"logps/rejected": -291.73223876953125,
|
| 116 |
+
"loss": 0.3288,
|
| 117 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 118 |
+
"rewards/chosen": -0.1798349767923355,
|
| 119 |
+
"rewards/margins": 0.17585726082324982,
|
| 120 |
+
"rewards/rejected": -0.3556922674179077,
|
| 121 |
"step": 70
|
| 122 |
},
|
| 123 |
{
|
| 124 |
+
"epoch": 0.17,
|
| 125 |
+
"learning_rate": 4.931986719649298e-07,
|
| 126 |
+
"logits/chosen": -2.698970079421997,
|
| 127 |
+
"logits/rejected": -2.6765952110290527,
|
| 128 |
+
"logps/chosen": -304.83587646484375,
|
| 129 |
+
"logps/rejected": -304.4337463378906,
|
| 130 |
+
"loss": 0.3139,
|
| 131 |
+
"rewards/accuracies": 0.643750011920929,
|
| 132 |
+
"rewards/chosen": -0.27432456612586975,
|
| 133 |
+
"rewards/margins": 0.17791275680065155,
|
| 134 |
+
"rewards/rejected": -0.4522373080253601,
|
| 135 |
"step": 80
|
| 136 |
},
|
| 137 |
{
|
| 138 |
+
"epoch": 0.19,
|
| 139 |
+
"learning_rate": 4.883222001996351e-07,
|
| 140 |
+
"logits/chosen": -2.6964175701141357,
|
| 141 |
+
"logits/rejected": -2.668422222137451,
|
| 142 |
+
"logps/chosen": -279.7590637207031,
|
| 143 |
+
"logps/rejected": -279.1517639160156,
|
| 144 |
+
"loss": 0.2697,
|
| 145 |
+
"rewards/accuracies": 0.65625,
|
| 146 |
+
"rewards/chosen": -0.34315305948257446,
|
| 147 |
+
"rewards/margins": 0.24895183742046356,
|
| 148 |
+
"rewards/rejected": -0.5921049118041992,
|
| 149 |
"step": 90
|
| 150 |
},
|
| 151 |
{
|
| 152 |
+
"epoch": 0.21,
|
| 153 |
+
"learning_rate": 4.821741763807186e-07,
|
| 154 |
+
"logits/chosen": -2.638505220413208,
|
| 155 |
+
"logits/rejected": -2.639235258102417,
|
| 156 |
+
"logps/chosen": -318.7078857421875,
|
| 157 |
+
"logps/rejected": -320.1378173828125,
|
| 158 |
+
"loss": 0.2738,
|
| 159 |
+
"rewards/accuracies": 0.675000011920929,
|
| 160 |
+
"rewards/chosen": -0.38339221477508545,
|
| 161 |
+
"rewards/margins": 0.3063567578792572,
|
| 162 |
+
"rewards/rejected": -0.689749002456665,
|
| 163 |
"step": 100
|
| 164 |
},
|
| 165 |
{
|
| 166 |
+
"epoch": 0.21,
|
| 167 |
+
"eval_logits/chosen": -2.6881661415100098,
|
| 168 |
+
"eval_logits/rejected": -2.6731417179107666,
|
| 169 |
+
"eval_logps/chosen": -287.24700927734375,
|
| 170 |
+
"eval_logps/rejected": -321.3398742675781,
|
| 171 |
+
"eval_loss": 0.2944861948490143,
|
| 172 |
+
"eval_rewards/accuracies": 0.71484375,
|
| 173 |
+
"eval_rewards/chosen": -0.3020733892917633,
|
| 174 |
+
"eval_rewards/margins": 0.337792307138443,
|
| 175 |
+
"eval_rewards/rejected": -0.6398657560348511,
|
| 176 |
+
"eval_runtime": 53.3678,
|
| 177 |
+
"eval_samples_per_second": 37.476,
|
| 178 |
+
"eval_steps_per_second": 0.6,
|
| 179 |
+
"step": 100
|
| 180 |
+
},
|
| 181 |
+
{
|
| 182 |
+
"epoch": 0.23,
|
| 183 |
+
"learning_rate": 4.747874028753375e-07,
|
| 184 |
+
"logits/chosen": -2.6743226051330566,
|
| 185 |
+
"logits/rejected": -2.6426432132720947,
|
| 186 |
+
"logps/chosen": -332.5823669433594,
|
| 187 |
+
"logps/rejected": -325.58184814453125,
|
| 188 |
+
"loss": 0.2961,
|
| 189 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 190 |
+
"rewards/chosen": -0.36884671449661255,
|
| 191 |
+
"rewards/margins": 0.30729353427886963,
|
| 192 |
+
"rewards/rejected": -0.6761402487754822,
|
| 193 |
"step": 110
|
| 194 |
},
|
| 195 |
{
|
| 196 |
+
"epoch": 0.25,
|
| 197 |
+
"learning_rate": 4.662012913161997e-07,
|
| 198 |
+
"logits/chosen": -2.6002402305603027,
|
| 199 |
+
"logits/rejected": -2.60917329788208,
|
| 200 |
+
"logps/chosen": -267.86614990234375,
|
| 201 |
+
"logps/rejected": -302.95538330078125,
|
| 202 |
+
"loss": 0.2585,
|
| 203 |
+
"rewards/accuracies": 0.699999988079071,
|
| 204 |
+
"rewards/chosen": -0.3666726052761078,
|
| 205 |
+
"rewards/margins": 0.3358650505542755,
|
| 206 |
+
"rewards/rejected": -0.7025377154350281,
|
| 207 |
"step": 120
|
| 208 |
},
|
| 209 |
{
|
| 210 |
+
"epoch": 0.27,
|
| 211 |
+
"learning_rate": 4.5646165232345103e-07,
|
| 212 |
+
"logits/chosen": -2.5904266834259033,
|
| 213 |
+
"logits/rejected": -2.5732221603393555,
|
| 214 |
+
"logps/chosen": -293.5733947753906,
|
| 215 |
+
"logps/rejected": -323.77679443359375,
|
| 216 |
+
"loss": 0.2348,
|
| 217 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 218 |
+
"rewards/chosen": -0.5687397122383118,
|
| 219 |
+
"rewards/margins": 0.2948606014251709,
|
| 220 |
+
"rewards/rejected": -0.8636003732681274,
|
| 221 |
"step": 130
|
| 222 |
},
|
| 223 |
{
|
| 224 |
+
"epoch": 0.29,
|
| 225 |
+
"learning_rate": 4.456204510851956e-07,
|
| 226 |
+
"logits/chosen": -2.5775017738342285,
|
| 227 |
+
"logits/rejected": -2.5546040534973145,
|
| 228 |
+
"logps/chosen": -323.5328674316406,
|
| 229 |
+
"logps/rejected": -327.654052734375,
|
| 230 |
+
"loss": 0.2218,
|
| 231 |
+
"rewards/accuracies": 0.699999988079071,
|
| 232 |
+
"rewards/chosen": -0.5967830419540405,
|
| 233 |
+
"rewards/margins": 0.36974120140075684,
|
| 234 |
+
"rewards/rejected": -0.9665242433547974,
|
| 235 |
"step": 140
|
| 236 |
},
|
| 237 |
{
|
| 238 |
+
"epoch": 0.31,
|
| 239 |
+
"learning_rate": 4.337355301007335e-07,
|
| 240 |
+
"logits/chosen": -2.4853436946868896,
|
| 241 |
+
"logits/rejected": -2.4797251224517822,
|
| 242 |
+
"logps/chosen": -325.6590881347656,
|
| 243 |
+
"logps/rejected": -327.2002868652344,
|
| 244 |
+
"loss": 0.2394,
|
| 245 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 246 |
+
"rewards/chosen": -0.6125308275222778,
|
| 247 |
+
"rewards/margins": 0.353943407535553,
|
| 248 |
+
"rewards/rejected": -0.966474175453186,
|
| 249 |
"step": 150
|
| 250 |
},
|
| 251 |
{
|
| 252 |
+
"epoch": 0.33,
|
| 253 |
+
"learning_rate": 4.2087030056579986e-07,
|
| 254 |
+
"logits/chosen": -2.431196928024292,
|
| 255 |
+
"logits/rejected": -2.418900728225708,
|
| 256 |
+
"logps/chosen": -343.87994384765625,
|
| 257 |
+
"logps/rejected": -368.2380676269531,
|
| 258 |
+
"loss": 0.2398,
|
| 259 |
+
"rewards/accuracies": 0.6875,
|
| 260 |
+
"rewards/chosen": -0.6931950449943542,
|
| 261 |
+
"rewards/margins": 0.4944288730621338,
|
| 262 |
+
"rewards/rejected": -1.1876238584518433,
|
| 263 |
"step": 160
|
| 264 |
},
|
| 265 |
{
|
| 266 |
+
"epoch": 0.36,
|
| 267 |
+
"learning_rate": 4.070934040463998e-07,
|
| 268 |
+
"logits/chosen": -2.5277438163757324,
|
| 269 |
+
"logits/rejected": -2.5002598762512207,
|
| 270 |
+
"logps/chosen": -368.4566650390625,
|
| 271 |
+
"logps/rejected": -341.0594787597656,
|
| 272 |
+
"loss": 0.2543,
|
| 273 |
+
"rewards/accuracies": 0.6875,
|
| 274 |
+
"rewards/chosen": -0.4498482346534729,
|
| 275 |
+
"rewards/margins": 0.39794427156448364,
|
| 276 |
+
"rewards/rejected": -0.8477924466133118,
|
| 277 |
"step": 170
|
| 278 |
},
|
| 279 |
{
|
| 280 |
+
"epoch": 0.38,
|
| 281 |
+
"learning_rate": 3.9247834624635404e-07,
|
| 282 |
+
"logits/chosen": -2.5052051544189453,
|
| 283 |
+
"logits/rejected": -2.4971730709075928,
|
| 284 |
+
"logps/chosen": -310.55523681640625,
|
| 285 |
+
"logps/rejected": -364.71453857421875,
|
| 286 |
+
"loss": 0.2338,
|
| 287 |
+
"rewards/accuracies": 0.71875,
|
| 288 |
+
"rewards/chosen": -0.5203171372413635,
|
| 289 |
+
"rewards/margins": 0.49080243706703186,
|
| 290 |
+
"rewards/rejected": -1.0111196041107178,
|
| 291 |
"step": 180
|
| 292 |
},
|
| 293 |
{
|
| 294 |
+
"epoch": 0.4,
|
| 295 |
+
"learning_rate": 3.7710310482256523e-07,
|
| 296 |
+
"logits/chosen": -2.355747699737549,
|
| 297 |
+
"logits/rejected": -2.3024649620056152,
|
| 298 |
+
"logps/chosen": -370.7752380371094,
|
| 299 |
+
"logps/rejected": -393.657470703125,
|
| 300 |
+
"loss": 0.2009,
|
| 301 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 302 |
+
"rewards/chosen": -0.9750235676765442,
|
| 303 |
+
"rewards/margins": 0.41129952669143677,
|
| 304 |
+
"rewards/rejected": -1.3863232135772705,
|
| 305 |
"step": 190
|
| 306 |
},
|
| 307 |
{
|
| 308 |
+
"epoch": 0.42,
|
| 309 |
+
"learning_rate": 3.610497133404795e-07,
|
| 310 |
+
"logits/chosen": -2.3048462867736816,
|
| 311 |
+
"logits/rejected": -2.3025858402252197,
|
| 312 |
+
"logps/chosen": -336.6009826660156,
|
| 313 |
+
"logps/rejected": -400.9478454589844,
|
| 314 |
+
"loss": 0.1875,
|
| 315 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 316 |
+
"rewards/chosen": -1.0244024991989136,
|
| 317 |
+
"rewards/margins": 0.624415397644043,
|
| 318 |
+
"rewards/rejected": -1.648817777633667,
|
| 319 |
"step": 200
|
| 320 |
},
|
| 321 |
{
|
| 322 |
+
"epoch": 0.42,
|
| 323 |
+
"eval_logits/chosen": -2.3602054119110107,
|
| 324 |
+
"eval_logits/rejected": -2.338970184326172,
|
| 325 |
+
"eval_logps/chosen": -352.4082946777344,
|
| 326 |
+
"eval_logps/rejected": -406.1714782714844,
|
| 327 |
+
"eval_loss": 0.20406465232372284,
|
| 328 |
+
"eval_rewards/accuracies": 0.72265625,
|
| 329 |
+
"eval_rewards/chosen": -0.9536865949630737,
|
| 330 |
+
"eval_rewards/margins": 0.5344952940940857,
|
| 331 |
+
"eval_rewards/rejected": -1.4881819486618042,
|
| 332 |
+
"eval_runtime": 53.3666,
|
| 333 |
+
"eval_samples_per_second": 37.477,
|
| 334 |
+
"eval_steps_per_second": 0.6,
|
| 335 |
+
"step": 200
|
| 336 |
+
},
|
| 337 |
+
{
|
| 338 |
+
"epoch": 0.44,
|
| 339 |
+
"learning_rate": 3.4440382358952115e-07,
|
| 340 |
+
"logits/chosen": -2.33644437789917,
|
| 341 |
+
"logits/rejected": -2.250994920730591,
|
| 342 |
+
"logps/chosen": -398.88409423828125,
|
| 343 |
+
"logps/rejected": -377.0559997558594,
|
| 344 |
+
"loss": 0.2167,
|
| 345 |
+
"rewards/accuracies": 0.668749988079071,
|
| 346 |
+
"rewards/chosen": -1.0204298496246338,
|
| 347 |
+
"rewards/margins": 0.3568393886089325,
|
| 348 |
+
"rewards/rejected": -1.3772691488265991,
|
| 349 |
"step": 210
|
| 350 |
},
|
| 351 |
{
|
| 352 |
+
"epoch": 0.46,
|
| 353 |
+
"learning_rate": 3.272542485937368e-07,
|
| 354 |
+
"logits/chosen": -2.3577969074249268,
|
| 355 |
+
"logits/rejected": -2.2945871353149414,
|
| 356 |
+
"logps/chosen": -364.9277038574219,
|
| 357 |
+
"logps/rejected": -371.2010803222656,
|
| 358 |
+
"loss": 0.2065,
|
| 359 |
+
"rewards/accuracies": 0.668749988079071,
|
| 360 |
+
"rewards/chosen": -0.8520679473876953,
|
| 361 |
+
"rewards/margins": 0.4781356453895569,
|
| 362 |
+
"rewards/rejected": -1.330203652381897,
|
| 363 |
"step": 220
|
| 364 |
},
|
| 365 |
{
|
| 366 |
+
"epoch": 0.48,
|
| 367 |
+
"learning_rate": 3.096924887558854e-07,
|
| 368 |
+
"logits/chosen": -2.3539249897003174,
|
| 369 |
+
"logits/rejected": -2.313048839569092,
|
| 370 |
+
"logps/chosen": -376.7759094238281,
|
| 371 |
+
"logps/rejected": -407.8626708984375,
|
| 372 |
+
"loss": 0.1894,
|
| 373 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 374 |
+
"rewards/chosen": -0.8920862078666687,
|
| 375 |
+
"rewards/margins": 0.5588781237602234,
|
| 376 |
+
"rewards/rejected": -1.4509644508361816,
|
| 377 |
"step": 230
|
| 378 |
},
|
| 379 |
{
|
| 380 |
+
"epoch": 0.5,
|
| 381 |
+
"learning_rate": 2.9181224366319943e-07,
|
| 382 |
+
"logits/chosen": -2.3257548809051514,
|
| 383 |
+
"logits/rejected": -2.310906410217285,
|
| 384 |
+
"logps/chosen": -399.5565490722656,
|
| 385 |
+
"logps/rejected": -400.0947265625,
|
| 386 |
+
"loss": 0.1923,
|
| 387 |
+
"rewards/accuracies": 0.731249988079071,
|
| 388 |
+
"rewards/chosen": -0.9511243104934692,
|
| 389 |
+
"rewards/margins": 0.562946617603302,
|
| 390 |
+
"rewards/rejected": -1.5140708684921265,
|
| 391 |
"step": 240
|
| 392 |
},
|
| 393 |
{
|
| 394 |
+
"epoch": 0.52,
|
| 395 |
+
"learning_rate": 2.7370891215954565e-07,
|
| 396 |
+
"logits/chosen": -2.2992796897888184,
|
| 397 |
+
"logits/rejected": -2.2705976963043213,
|
| 398 |
+
"logps/chosen": -336.1885986328125,
|
| 399 |
+
"logps/rejected": -385.8701477050781,
|
| 400 |
+
"loss": 0.1982,
|
| 401 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 402 |
+
"rewards/chosen": -0.8285441398620605,
|
| 403 |
+
"rewards/margins": 0.6316433548927307,
|
| 404 |
+
"rewards/rejected": -1.4601874351501465,
|
| 405 |
"step": 250
|
| 406 |
},
|
| 407 |
{
|
| 408 |
+
"epoch": 0.54,
|
| 409 |
+
"learning_rate": 2.55479083351317e-07,
|
| 410 |
+
"logits/chosen": -2.327385425567627,
|
| 411 |
+
"logits/rejected": -2.3026115894317627,
|
| 412 |
+
"logps/chosen": -392.62591552734375,
|
| 413 |
+
"logps/rejected": -422.28448486328125,
|
| 414 |
+
"loss": 0.2222,
|
| 415 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 416 |
+
"rewards/chosen": -0.9515058398246765,
|
| 417 |
+
"rewards/margins": 0.49343791604042053,
|
| 418 |
+
"rewards/rejected": -1.4449436664581299,
|
| 419 |
"step": 260
|
| 420 |
},
|
| 421 |
{
|
| 422 |
+
"epoch": 0.56,
|
| 423 |
+
"learning_rate": 2.3722002126275822e-07,
|
| 424 |
+
"logits/chosen": -2.3942108154296875,
|
| 425 |
+
"logits/rejected": -2.3683059215545654,
|
| 426 |
+
"logps/chosen": -345.6385498046875,
|
| 427 |
+
"logps/rejected": -383.3131408691406,
|
| 428 |
+
"loss": 0.2198,
|
| 429 |
+
"rewards/accuracies": 0.625,
|
| 430 |
+
"rewards/chosen": -0.9201357960700989,
|
| 431 |
+
"rewards/margins": 0.3128657937049866,
|
| 432 |
+
"rewards/rejected": -1.233001470565796,
|
| 433 |
"step": 270
|
| 434 |
},
|
| 435 |
{
|
| 436 |
+
"epoch": 0.59,
|
| 437 |
+
"learning_rate": 2.19029145890313e-07,
|
| 438 |
+
"logits/chosen": -2.3505771160125732,
|
| 439 |
+
"logits/rejected": -2.3291995525360107,
|
| 440 |
+
"logps/chosen": -346.64093017578125,
|
| 441 |
+
"logps/rejected": -383.3504943847656,
|
| 442 |
+
"loss": 0.2066,
|
| 443 |
+
"rewards/accuracies": 0.706250011920929,
|
| 444 |
+
"rewards/chosen": -0.8398548364639282,
|
| 445 |
+
"rewards/margins": 0.5200469493865967,
|
| 446 |
+
"rewards/rejected": -1.359901785850525,
|
| 447 |
"step": 280
|
| 448 |
},
|
| 449 |
{
|
| 450 |
+
"epoch": 0.61,
|
| 451 |
+
"learning_rate": 2.0100351342479216e-07,
|
| 452 |
+
"logits/chosen": -2.3983218669891357,
|
| 453 |
+
"logits/rejected": -2.356658458709717,
|
| 454 |
+
"logps/chosen": -379.2950744628906,
|
| 455 |
+
"logps/rejected": -412.7428283691406,
|
| 456 |
+
"loss": 0.2171,
|
| 457 |
+
"rewards/accuracies": 0.699999988079071,
|
| 458 |
+
"rewards/chosen": -0.7873485088348389,
|
| 459 |
+
"rewards/margins": 0.47691774368286133,
|
| 460 |
+
"rewards/rejected": -1.2642663717269897,
|
| 461 |
"step": 290
|
| 462 |
},
|
| 463 |
{
|
| 464 |
+
"epoch": 0.63,
|
| 465 |
+
"learning_rate": 1.8323929841460178e-07,
|
| 466 |
+
"logits/chosen": -2.331897497177124,
|
| 467 |
+
"logits/rejected": -2.3235559463500977,
|
| 468 |
+
"logps/chosen": -356.6403503417969,
|
| 469 |
+
"logps/rejected": -377.82135009765625,
|
| 470 |
+
"loss": 0.2103,
|
| 471 |
+
"rewards/accuracies": 0.6875,
|
| 472 |
+
"rewards/chosen": -0.839586615562439,
|
| 473 |
+
"rewards/margins": 0.46369633078575134,
|
| 474 |
+
"rewards/rejected": -1.3032829761505127,
|
| 475 |
+
"step": 300
|
| 476 |
+
},
|
| 477 |
+
{
|
| 478 |
+
"epoch": 0.63,
|
| 479 |
+
"eval_logits/chosen": -2.3750030994415283,
|
| 480 |
+
"eval_logits/rejected": -2.3532962799072266,
|
| 481 |
+
"eval_logps/chosen": -341.04937744140625,
|
| 482 |
+
"eval_logps/rejected": -399.583984375,
|
| 483 |
+
"eval_loss": 0.21952193975448608,
|
| 484 |
+
"eval_rewards/accuracies": 0.70703125,
|
| 485 |
+
"eval_rewards/chosen": -0.8400974273681641,
|
| 486 |
+
"eval_rewards/margins": 0.5822098255157471,
|
| 487 |
+
"eval_rewards/rejected": -1.4223072528839111,
|
| 488 |
+
"eval_runtime": 53.3824,
|
| 489 |
+
"eval_samples_per_second": 37.466,
|
| 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": -2.345519542694092,
|
| 497 |
+
"logits/rejected": -2.3241662979125977,
|
| 498 |
+
"logps/chosen": -378.9863586425781,
|
| 499 |
+
"logps/rejected": -402.13153076171875,
|
| 500 |
+
"loss": 0.2093,
|
| 501 |
+
"rewards/accuracies": 0.71875,
|
| 502 |
+
"rewards/chosen": -0.8223126530647278,
|
| 503 |
+
"rewards/margins": 0.614983081817627,
|
| 504 |
+
"rewards/rejected": -1.43729567527771,
|
| 505 |
"step": 310
|
| 506 |
},
|
| 507 |
{
|
| 508 |
+
"epoch": 0.67,
|
| 509 |
+
"learning_rate": 1.488723393865766e-07,
|
| 510 |
+
"logits/chosen": -2.3204216957092285,
|
| 511 |
+
"logits/rejected": -2.2891576290130615,
|
| 512 |
+
"logps/chosen": -348.6353454589844,
|
| 513 |
+
"logps/rejected": -385.66278076171875,
|
| 514 |
+
"loss": 0.2022,
|
| 515 |
+
"rewards/accuracies": 0.6187499761581421,
|
| 516 |
+
"rewards/chosen": -1.0812270641326904,
|
| 517 |
+
"rewards/margins": 0.4006832242012024,
|
| 518 |
+
"rewards/rejected": -1.4819103479385376,
|
| 519 |
"step": 320
|
| 520 |
},
|
| 521 |
{
|
| 522 |
+
"epoch": 0.69,
|
| 523 |
+
"learning_rate": 1.3245295796480788e-07,
|
| 524 |
+
"logits/chosen": -2.287698745727539,
|
| 525 |
+
"logits/rejected": -2.279531955718994,
|
| 526 |
+
"logps/chosen": -390.47540283203125,
|
| 527 |
+
"logps/rejected": -438.2626037597656,
|
| 528 |
+
"loss": 0.2029,
|
| 529 |
+
"rewards/accuracies": 0.668749988079071,
|
| 530 |
+
"rewards/chosen": -1.0578622817993164,
|
| 531 |
+
"rewards/margins": 0.43574437499046326,
|
| 532 |
+
"rewards/rejected": -1.4936068058013916,
|
| 533 |
"step": 330
|
| 534 |
},
|
| 535 |
{
|
| 536 |
+
"epoch": 0.71,
|
| 537 |
+
"learning_rate": 1.1666074087171627e-07,
|
| 538 |
+
"logits/chosen": -2.3450725078582764,
|
| 539 |
+
"logits/rejected": -2.3226075172424316,
|
| 540 |
+
"logps/chosen": -381.7784118652344,
|
| 541 |
+
"logps/rejected": -389.5802001953125,
|
| 542 |
+
"loss": 0.2096,
|
| 543 |
+
"rewards/accuracies": 0.6875,
|
| 544 |
+
"rewards/chosen": -0.9645330309867859,
|
| 545 |
+
"rewards/margins": 0.444490522146225,
|
| 546 |
+
"rewards/rejected": -1.409023642539978,
|
| 547 |
"step": 340
|
| 548 |
},
|
| 549 |
{
|
| 550 |
+
"epoch": 0.73,
|
| 551 |
+
"learning_rate": 1.0157994641835734e-07,
|
| 552 |
+
"logits/chosen": -2.356595277786255,
|
| 553 |
+
"logits/rejected": -2.3080544471740723,
|
| 554 |
+
"logps/chosen": -353.982177734375,
|
| 555 |
+
"logps/rejected": -418.7837829589844,
|
| 556 |
+
"loss": 0.2127,
|
| 557 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 558 |
+
"rewards/chosen": -0.979511559009552,
|
| 559 |
+
"rewards/margins": 0.5967382788658142,
|
| 560 |
+
"rewards/rejected": -1.5762498378753662,
|
| 561 |
"step": 350
|
| 562 |
},
|
| 563 |
{
|
| 564 |
+
"epoch": 0.75,
|
| 565 |
+
"learning_rate": 8.729103716819111e-08,
|
| 566 |
+
"logits/chosen": -2.3197410106658936,
|
| 567 |
+
"logits/rejected": -2.27152419090271,
|
| 568 |
+
"logps/chosen": -370.5116271972656,
|
| 569 |
+
"logps/rejected": -402.8948059082031,
|
| 570 |
+
"loss": 0.2121,
|
| 571 |
+
"rewards/accuracies": 0.75,
|
| 572 |
+
"rewards/chosen": -0.8492336273193359,
|
| 573 |
+
"rewards/margins": 0.5515913367271423,
|
| 574 |
+
"rewards/rejected": -1.4008249044418335,
|
| 575 |
"step": 360
|
| 576 |
},
|
| 577 |
{
|
| 578 |
+
"epoch": 0.77,
|
| 579 |
+
"learning_rate": 7.387025063449081e-08,
|
| 580 |
+
"logits/chosen": -2.3385300636291504,
|
| 581 |
+
"logits/rejected": -2.309565544128418,
|
| 582 |
+
"logps/chosen": -328.14276123046875,
|
| 583 |
+
"logps/rejected": -381.6123962402344,
|
| 584 |
+
"loss": 0.2268,
|
| 585 |
"rewards/accuracies": 0.7562500238418579,
|
| 586 |
+
"rewards/chosen": -0.8338375091552734,
|
| 587 |
+
"rewards/margins": 0.5052274465560913,
|
| 588 |
+
"rewards/rejected": -1.3390649557113647,
|
| 589 |
"step": 370
|
| 590 |
},
|
| 591 |
{
|
| 592 |
+
"epoch": 0.79,
|
| 593 |
+
"learning_rate": 6.138919252022435e-08,
|
| 594 |
+
"logits/chosen": -2.331386089324951,
|
| 595 |
+
"logits/rejected": -2.291889190673828,
|
| 596 |
+
"logps/chosen": -350.16119384765625,
|
| 597 |
+
"logps/rejected": -381.3484802246094,
|
| 598 |
+
"loss": 0.2087,
|
| 599 |
+
"rewards/accuracies": 0.699999988079071,
|
| 600 |
+
"rewards/chosen": -0.8641462326049805,
|
| 601 |
+
"rewards/margins": 0.568858802318573,
|
| 602 |
+
"rewards/rejected": -1.4330050945281982,
|
| 603 |
"step": 380
|
| 604 |
},
|
| 605 |
{
|
| 606 |
+
"epoch": 0.82,
|
| 607 |
+
"learning_rate": 4.991445467064689e-08,
|
| 608 |
+
"logits/chosen": -2.2733142375946045,
|
| 609 |
+
"logits/rejected": -2.236274242401123,
|
| 610 |
+
"logps/chosen": -376.81573486328125,
|
| 611 |
+
"logps/rejected": -405.6353759765625,
|
| 612 |
+
"loss": 0.2182,
|
| 613 |
+
"rewards/accuracies": 0.668749988079071,
|
| 614 |
+
"rewards/chosen": -0.889461874961853,
|
| 615 |
+
"rewards/margins": 0.47357290983200073,
|
| 616 |
+
"rewards/rejected": -1.363034725189209,
|
| 617 |
"step": 390
|
| 618 |
},
|
| 619 |
{
|
| 620 |
+
"epoch": 0.84,
|
| 621 |
+
"learning_rate": 3.9507259776993954e-08,
|
| 622 |
+
"logits/chosen": -2.32843017578125,
|
| 623 |
+
"logits/rejected": -2.3155007362365723,
|
| 624 |
+
"logps/chosen": -367.7158508300781,
|
| 625 |
+
"logps/rejected": -422.8672790527344,
|
| 626 |
+
"loss": 0.2051,
|
| 627 |
+
"rewards/accuracies": 0.668749988079071,
|
| 628 |
+
"rewards/chosen": -1.0084705352783203,
|
| 629 |
+
"rewards/margins": 0.5765265226364136,
|
| 630 |
+
"rewards/rejected": -1.5849970579147339,
|
| 631 |
"step": 400
|
| 632 |
},
|
| 633 |
+
{
|
| 634 |
+
"epoch": 0.84,
|
| 635 |
+
"eval_logits/chosen": -2.3381025791168213,
|
| 636 |
+
"eval_logits/rejected": -2.3161401748657227,
|
| 637 |
+
"eval_logps/chosen": -345.9896545410156,
|
| 638 |
+
"eval_logps/rejected": -406.967041015625,
|
| 639 |
+
"eval_loss": 0.21300797164440155,
|
| 640 |
+
"eval_rewards/accuracies": 0.70703125,
|
| 641 |
+
"eval_rewards/chosen": -0.8895003199577332,
|
| 642 |
+
"eval_rewards/margins": 0.6066373586654663,
|
| 643 |
+
"eval_rewards/rejected": -1.4961377382278442,
|
| 644 |
+
"eval_runtime": 53.3639,
|
| 645 |
+
"eval_samples_per_second": 37.479,
|
| 646 |
+
"eval_steps_per_second": 0.6,
|
| 647 |
+
"step": 400
|
| 648 |
+
},
|
| 649 |
+
{
|
| 650 |
+
"epoch": 0.86,
|
| 651 |
+
"learning_rate": 3.022313472693447e-08,
|
| 652 |
+
"logits/chosen": -2.309161901473999,
|
| 653 |
+
"logits/rejected": -2.292067050933838,
|
| 654 |
+
"logps/chosen": -382.9537353515625,
|
| 655 |
+
"logps/rejected": -378.39898681640625,
|
| 656 |
+
"loss": 0.2106,
|
| 657 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 658 |
+
"rewards/chosen": -1.0024274587631226,
|
| 659 |
+
"rewards/margins": 0.4243893623352051,
|
| 660 |
+
"rewards/rejected": -1.426816701889038,
|
| 661 |
+
"step": 410
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"epoch": 0.88,
|
| 665 |
+
"learning_rate": 2.2111614344599684e-08,
|
| 666 |
+
"logits/chosen": -2.372878074645996,
|
| 667 |
+
"logits/rejected": -2.3500359058380127,
|
| 668 |
+
"logps/chosen": -393.245361328125,
|
| 669 |
+
"logps/rejected": -398.61175537109375,
|
| 670 |
+
"loss": 0.2155,
|
| 671 |
+
"rewards/accuracies": 0.637499988079071,
|
| 672 |
+
"rewards/chosen": -0.9859229326248169,
|
| 673 |
+
"rewards/margins": 0.39800676703453064,
|
| 674 |
+
"rewards/rejected": -1.3839296102523804,
|
| 675 |
+
"step": 420
|
| 676 |
+
},
|
| 677 |
+
{
|
| 678 |
+
"epoch": 0.9,
|
| 679 |
+
"learning_rate": 1.521597710086439e-08,
|
| 680 |
+
"logits/chosen": -2.382736921310425,
|
| 681 |
+
"logits/rejected": -2.330305576324463,
|
| 682 |
+
"logps/chosen": -380.0002136230469,
|
| 683 |
+
"logps/rejected": -367.59625244140625,
|
| 684 |
+
"loss": 0.2061,
|
| 685 |
+
"rewards/accuracies": 0.675000011920929,
|
| 686 |
+
"rewards/chosen": -0.9225358963012695,
|
| 687 |
+
"rewards/margins": 0.4716859459877014,
|
| 688 |
+
"rewards/rejected": -1.3942219018936157,
|
| 689 |
+
"step": 430
|
| 690 |
+
},
|
| 691 |
+
{
|
| 692 |
+
"epoch": 0.92,
|
| 693 |
+
"learning_rate": 9.57301420397924e-09,
|
| 694 |
+
"logits/chosen": -2.359827995300293,
|
| 695 |
+
"logits/rejected": -2.3190550804138184,
|
| 696 |
+
"logps/chosen": -396.4924011230469,
|
| 697 |
+
"logps/rejected": -385.8436279296875,
|
| 698 |
+
"loss": 0.2136,
|
| 699 |
+
"rewards/accuracies": 0.706250011920929,
|
| 700 |
+
"rewards/chosen": -0.9304088354110718,
|
| 701 |
+
"rewards/margins": 0.43420737981796265,
|
| 702 |
+
"rewards/rejected": -1.3646161556243896,
|
| 703 |
+
"step": 440
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 0.94,
|
| 707 |
+
"learning_rate": 5.212833302556258e-09,
|
| 708 |
+
"logits/chosen": -2.361069440841675,
|
| 709 |
+
"logits/rejected": -2.330132246017456,
|
| 710 |
+
"logps/chosen": -366.66845703125,
|
| 711 |
+
"logps/rejected": -462.7421875,
|
| 712 |
+
"loss": 0.2099,
|
| 713 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 714 |
+
"rewards/chosen": -0.8973833322525024,
|
| 715 |
+
"rewards/margins": 0.8092821836471558,
|
| 716 |
+
"rewards/rejected": -1.7066656351089478,
|
| 717 |
+
"step": 450
|
| 718 |
+
},
|
| 719 |
+
{
|
| 720 |
+
"epoch": 0.96,
|
| 721 |
+
"learning_rate": 2.158697848236607e-09,
|
| 722 |
+
"logits/chosen": -2.313333034515381,
|
| 723 |
+
"logits/rejected": -2.27656626701355,
|
| 724 |
+
"logps/chosen": -402.6481628417969,
|
| 725 |
+
"logps/rejected": -400.33251953125,
|
| 726 |
+
"loss": 0.2052,
|
| 727 |
+
"rewards/accuracies": 0.65625,
|
| 728 |
+
"rewards/chosen": -0.923121452331543,
|
| 729 |
+
"rewards/margins": 0.4761527478694916,
|
| 730 |
+
"rewards/rejected": -1.399274230003357,
|
| 731 |
+
"step": 460
|
| 732 |
+
},
|
| 733 |
+
{
|
| 734 |
+
"epoch": 0.98,
|
| 735 |
+
"learning_rate": 4.269029751107489e-10,
|
| 736 |
+
"logits/chosen": -2.3397324085235596,
|
| 737 |
+
"logits/rejected": -2.2944626808166504,
|
| 738 |
+
"logps/chosen": -405.10040283203125,
|
| 739 |
+
"logps/rejected": -419.93902587890625,
|
| 740 |
+
"loss": 0.2137,
|
| 741 |
+
"rewards/accuracies": 0.75,
|
| 742 |
+
"rewards/chosen": -1.0206491947174072,
|
| 743 |
+
"rewards/margins": 0.5544232130050659,
|
| 744 |
+
"rewards/rejected": -1.5750725269317627,
|
| 745 |
+
"step": 470
|
| 746 |
+
},
|
| 747 |
{
|
| 748 |
"epoch": 1.0,
|
| 749 |
+
"step": 478,
|
| 750 |
"total_flos": 0.0,
|
| 751 |
+
"train_loss": 0.24355335988758997,
|
| 752 |
+
"train_runtime": 3944.1561,
|
| 753 |
+
"train_samples_per_second": 15.5,
|
| 754 |
+
"train_steps_per_second": 0.121
|
| 755 |
}
|
| 756 |
],
|
| 757 |
"logging_steps": 10,
|
| 758 |
+
"max_steps": 478,
|
| 759 |
"num_train_epochs": 1,
|
| 760 |
"save_steps": 1000,
|
| 761 |
"total_flos": 0.0,
|
training_args.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 5944
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:615833a83dcdf021f9a78aad80acb1d830e478b5789aeee61b4cff876fd4f470
|
| 3 |
size 5944
|