Model save
Browse files- README.md +1 -17
- 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 +454 -616
- training_args.bin +1 -1
README.md
CHANGED
|
@@ -16,16 +16,6 @@ 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 |
-
It achieves the following results on the evaluation set:
|
| 20 |
-
- Loss: 0.2143
|
| 21 |
-
- Rewards/chosen: -0.8956
|
| 22 |
-
- Rewards/rejected: -1.5167
|
| 23 |
-
- Rewards/accuracies: 0.7031
|
| 24 |
-
- Rewards/margins: 0.6212
|
| 25 |
-
- Logps/rejected: -409.0278
|
| 26 |
-
- Logps/chosen: -346.5983
|
| 27 |
-
- Logits/rejected: -2.4275
|
| 28 |
-
- Logits/chosen: -2.4425
|
| 29 |
|
| 30 |
## Model description
|
| 31 |
|
|
@@ -47,7 +37,7 @@ The following hyperparameters were used during training:
|
|
| 47 |
- learning_rate: 5e-07
|
| 48 |
- train_batch_size: 8
|
| 49 |
- eval_batch_size: 8
|
| 50 |
-
- seed:
|
| 51 |
- distributed_type: multi-GPU
|
| 52 |
- num_devices: 8
|
| 53 |
- gradient_accumulation_steps: 2
|
|
@@ -60,12 +50,6 @@ The following hyperparameters were used during training:
|
|
| 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 |
-
| 0.2667 | 0.21 | 100 | 0.2670 | -0.4530 | -0.7921 | 0.6797 | 0.3391 | -336.5636 | -302.3352 | -2.7593 | -2.7741 |
|
| 66 |
-
| 0.2068 | 0.42 | 200 | 0.2087 | -0.8343 | -1.3671 | 0.6836 | 0.5328 | -394.0588 | -340.4660 | -2.5512 | -2.5673 |
|
| 67 |
-
| 0.2095 | 0.63 | 300 | 0.2233 | -0.8384 | -1.4377 | 0.7109 | 0.5993 | -401.1194 | -340.8771 | -2.4645 | -2.4791 |
|
| 68 |
-
| 0.204 | 0.84 | 400 | 0.2143 | -0.8956 | -1.5167 | 0.7031 | 0.6212 | -409.0278 | -346.5983 | -2.4275 | -2.4425 |
|
| 69 |
|
| 70 |
|
| 71 |
### 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 |
|
| 20 |
## Model description
|
| 21 |
|
|
|
|
| 37 |
- learning_rate: 5e-07
|
| 38 |
- train_batch_size: 8
|
| 39 |
- eval_batch_size: 8
|
| 40 |
+
- seed: 4
|
| 41 |
- distributed_type: multi-GPU
|
| 42 |
- num_devices: 8
|
| 43 |
- gradient_accumulation_steps: 2
|
|
|
|
| 50 |
|
| 51 |
### Training results
|
| 52 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 53 |
|
| 54 |
|
| 55 |
### 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": 0.14271833528707056,
|
| 4 |
+
"train_runtime": 3205.3137,
|
| 5 |
+
"train_samples": 51894,
|
| 6 |
+
"train_samples_per_second": 16.19,
|
| 7 |
+
"train_steps_per_second": 0.126
|
| 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 sha256:1adf15ad14a70a15f2da17273d04658d31bcb80a7cf71fd8739ea6ece32a9dc0
|
| 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:d6840860a9c0b139fea3960dfc4ec0c766e717e85de1ad3968cc0068318f6714
|
| 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 sha256:741f22e485d0728e3f4f8f4f6ed06bb1f8dbf7720e43f83f9cde68c7f1dd445c
|
| 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": 0.14271833528707056,
|
| 4 |
+
"train_runtime": 3205.3137,
|
| 5 |
+
"train_samples": 51894,
|
| 6 |
+
"train_samples_per_second": 16.19,
|
| 7 |
+
"train_steps_per_second": 0.126
|
| 8 |
}
|
trainer_state.json
CHANGED
|
@@ -1,21 +1,21 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": null,
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
-
"epoch":
|
| 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,
|
|
@@ -24,738 +24,576 @@
|
|
| 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":
|
| 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":
|
| 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":
|
| 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":
|
| 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": 0.21,
|
| 167 |
-
"eval_logits/chosen": -2.7741379737854004,
|
| 168 |
-
"eval_logits/rejected": -2.7592568397521973,
|
| 169 |
-
"eval_logps/chosen": -302.3351745605469,
|
| 170 |
-
"eval_logps/rejected": -336.5635986328125,
|
| 171 |
-
"eval_loss": 0.26699498295783997,
|
| 172 |
-
"eval_rewards/accuracies": 0.6796875,
|
| 173 |
-
"eval_rewards/chosen": -0.4529554545879364,
|
| 174 |
-
"eval_rewards/margins": 0.3391478657722473,
|
| 175 |
-
"eval_rewards/rejected": -0.7921034097671509,
|
| 176 |
-
"eval_runtime": 53.381,
|
| 177 |
-
"eval_samples_per_second": 37.466,
|
| 178 |
-
"eval_steps_per_second": 0.599,
|
| 179 |
"step": 100
|
| 180 |
},
|
| 181 |
{
|
| 182 |
-
"epoch": 0.
|
| 183 |
-
"learning_rate": 4.
|
| 184 |
-
"logits/chosen": -2.
|
| 185 |
-
"logits/rejected": -2.
|
| 186 |
-
"logps/chosen": -
|
| 187 |
-
"logps/rejected": -
|
| 188 |
-
"loss": 0.
|
| 189 |
-
"rewards/accuracies": 0.
|
| 190 |
-
"rewards/chosen": -
|
| 191 |
-
"rewards/margins":
|
| 192 |
-
"rewards/rejected": -
|
| 193 |
"step": 110
|
| 194 |
},
|
| 195 |
{
|
| 196 |
-
"epoch": 0.
|
| 197 |
-
"learning_rate": 4.
|
| 198 |
-
"logits/chosen": -2.
|
| 199 |
-
"logits/rejected": -2.
|
| 200 |
-
"logps/chosen": -
|
| 201 |
-
"logps/rejected": -
|
| 202 |
-
"loss": 0.
|
| 203 |
-
"rewards/accuracies": 0.
|
| 204 |
-
"rewards/chosen": -0.
|
| 205 |
-
"rewards/margins":
|
| 206 |
-
"rewards/rejected": -
|
| 207 |
"step": 120
|
| 208 |
},
|
| 209 |
{
|
| 210 |
-
"epoch": 0.
|
| 211 |
-
"learning_rate": 4.
|
| 212 |
-
"logits/chosen": -2.
|
| 213 |
-
"logits/rejected": -2.
|
| 214 |
-
"logps/chosen": -
|
| 215 |
-
"logps/rejected": -
|
| 216 |
-
"loss": 0.
|
| 217 |
-
"rewards/accuracies": 0.
|
| 218 |
-
"rewards/chosen": -
|
| 219 |
-
"rewards/margins":
|
| 220 |
-
"rewards/rejected": -
|
| 221 |
"step": 130
|
| 222 |
},
|
| 223 |
{
|
| 224 |
-
"epoch": 0.
|
| 225 |
-
"learning_rate": 4.
|
| 226 |
-
"logits/chosen": -2.
|
| 227 |
-
"logits/rejected": -2.
|
| 228 |
-
"logps/chosen": -
|
| 229 |
-
"logps/rejected": -
|
| 230 |
-
"loss": 0.
|
| 231 |
-
"rewards/accuracies": 0.
|
| 232 |
-
"rewards/chosen": -0.
|
| 233 |
-
"rewards/margins":
|
| 234 |
-
"rewards/rejected": -
|
| 235 |
"step": 140
|
| 236 |
},
|
| 237 |
{
|
| 238 |
-
"epoch": 0.
|
| 239 |
-
"learning_rate":
|
| 240 |
-
"logits/chosen": -2.
|
| 241 |
-
"logits/rejected": -2.
|
| 242 |
-
"logps/chosen": -
|
| 243 |
-
"logps/rejected": -
|
| 244 |
-
"loss": 0.
|
| 245 |
-
"rewards/accuracies": 0.
|
| 246 |
-
"rewards/chosen": -
|
| 247 |
-
"rewards/margins":
|
| 248 |
-
"rewards/rejected": -
|
| 249 |
"step": 150
|
| 250 |
},
|
| 251 |
{
|
| 252 |
-
"epoch": 0.
|
| 253 |
-
"learning_rate":
|
| 254 |
-
"logits/chosen": -2.
|
| 255 |
-
"logits/rejected": -2.
|
| 256 |
-
"logps/chosen": -
|
| 257 |
-
"logps/rejected": -
|
| 258 |
-
"loss": 0.
|
| 259 |
-
"rewards/accuracies": 0.
|
| 260 |
-
"rewards/chosen": -
|
| 261 |
-
"rewards/margins":
|
| 262 |
-
"rewards/rejected": -
|
| 263 |
"step": 160
|
| 264 |
},
|
| 265 |
{
|
| 266 |
-
"epoch": 0.
|
| 267 |
-
"learning_rate":
|
| 268 |
-
"logits/chosen": -2.
|
| 269 |
-
"logits/rejected": -2.
|
| 270 |
-
"logps/chosen": -
|
| 271 |
-
"logps/rejected": -
|
| 272 |
-
"loss": 0.
|
| 273 |
-
"rewards/accuracies": 0.
|
| 274 |
-
"rewards/chosen": -
|
| 275 |
-
"rewards/margins":
|
| 276 |
-
"rewards/rejected": -
|
| 277 |
"step": 170
|
| 278 |
},
|
| 279 |
{
|
| 280 |
-
"epoch": 0.
|
| 281 |
-
"learning_rate": 3.
|
| 282 |
-
"logits/chosen": -2.
|
| 283 |
-
"logits/rejected": -2.
|
| 284 |
-
"logps/chosen": -
|
| 285 |
-
"logps/rejected": -
|
| 286 |
-
"loss": 0.
|
| 287 |
-
"rewards/accuracies": 0.
|
| 288 |
-
"rewards/chosen": -0.
|
| 289 |
-
"rewards/margins":
|
| 290 |
-
"rewards/rejected": -
|
| 291 |
"step": 180
|
| 292 |
},
|
| 293 |
{
|
| 294 |
-
"epoch": 0.
|
| 295 |
-
"learning_rate": 3.
|
| 296 |
-
"logits/chosen": -2.
|
| 297 |
-
"logits/rejected": -2.
|
| 298 |
-
"logps/chosen": -
|
| 299 |
-
"logps/rejected": -
|
| 300 |
-
"loss": 0.
|
| 301 |
-
"rewards/accuracies": 0.
|
| 302 |
-
"rewards/chosen": -
|
| 303 |
-
"rewards/margins":
|
| 304 |
-
"rewards/rejected": -
|
| 305 |
"step": 190
|
| 306 |
},
|
| 307 |
{
|
| 308 |
-
"epoch": 0.
|
| 309 |
-
"learning_rate":
|
| 310 |
-
"logits/chosen": -2.
|
| 311 |
-
"logits/rejected": -2.
|
| 312 |
-
"logps/chosen": -
|
| 313 |
-
"logps/rejected": -
|
| 314 |
-
"loss": 0.
|
| 315 |
-
"rewards/accuracies": 0.
|
| 316 |
-
"rewards/chosen": -
|
| 317 |
-
"rewards/margins":
|
| 318 |
-
"rewards/rejected": -
|
| 319 |
-
"step": 200
|
| 320 |
-
},
|
| 321 |
-
{
|
| 322 |
-
"epoch": 0.42,
|
| 323 |
-
"eval_logits/chosen": -2.5673255920410156,
|
| 324 |
-
"eval_logits/rejected": -2.5512256622314453,
|
| 325 |
-
"eval_logps/chosen": -340.4659729003906,
|
| 326 |
-
"eval_logps/rejected": -394.05877685546875,
|
| 327 |
-
"eval_loss": 0.2086561620235443,
|
| 328 |
-
"eval_rewards/accuracies": 0.68359375,
|
| 329 |
-
"eval_rewards/chosen": -0.8342636823654175,
|
| 330 |
-
"eval_rewards/margins": 0.532791256904602,
|
| 331 |
-
"eval_rewards/rejected": -1.3670549392700195,
|
| 332 |
-
"eval_runtime": 53.3257,
|
| 333 |
-
"eval_samples_per_second": 37.505,
|
| 334 |
-
"eval_steps_per_second": 0.6,
|
| 335 |
"step": 200
|
| 336 |
},
|
| 337 |
{
|
| 338 |
-
"epoch": 0.
|
| 339 |
-
"learning_rate":
|
| 340 |
-
"logits/chosen": -2.
|
| 341 |
-
"logits/rejected": -2.
|
| 342 |
-
"logps/chosen": -
|
| 343 |
-
"logps/rejected": -
|
| 344 |
-
"loss": 0.
|
| 345 |
-
"rewards/accuracies": 0.
|
| 346 |
-
"rewards/chosen": -
|
| 347 |
-
"rewards/margins":
|
| 348 |
-
"rewards/rejected": -
|
| 349 |
"step": 210
|
| 350 |
},
|
| 351 |
{
|
| 352 |
-
"epoch": 0.
|
| 353 |
-
"learning_rate":
|
| 354 |
-
"logits/chosen": -2.
|
| 355 |
-
"logits/rejected": -2.
|
| 356 |
-
"logps/chosen": -
|
| 357 |
-
"logps/rejected": -
|
| 358 |
-
"loss": 0.
|
| 359 |
-
"rewards/accuracies": 0.
|
| 360 |
-
"rewards/chosen": -
|
| 361 |
-
"rewards/margins":
|
| 362 |
-
"rewards/rejected": -
|
| 363 |
"step": 220
|
| 364 |
},
|
| 365 |
{
|
| 366 |
-
"epoch": 0.
|
| 367 |
-
"learning_rate":
|
| 368 |
-
"logits/chosen": -2.
|
| 369 |
-
"logits/rejected": -2.
|
| 370 |
-
"logps/chosen": -
|
| 371 |
-
"logps/rejected": -
|
| 372 |
-
"loss": 0.
|
| 373 |
-
"rewards/accuracies": 0.
|
| 374 |
-
"rewards/chosen": -0.
|
| 375 |
-
"rewards/margins":
|
| 376 |
-
"rewards/rejected": -
|
| 377 |
"step": 230
|
| 378 |
},
|
| 379 |
{
|
| 380 |
-
"epoch": 0.
|
| 381 |
-
"learning_rate": 2.
|
| 382 |
-
"logits/chosen": -2.
|
| 383 |
-
"logits/rejected": -2.
|
| 384 |
-
"logps/chosen": -
|
| 385 |
-
"logps/rejected": -
|
| 386 |
-
"loss": 0.
|
| 387 |
-
"rewards/accuracies": 0.
|
| 388 |
-
"rewards/chosen": -
|
| 389 |
-
"rewards/margins":
|
| 390 |
-
"rewards/rejected": -
|
| 391 |
"step": 240
|
| 392 |
},
|
| 393 |
{
|
| 394 |
-
"epoch": 0.
|
| 395 |
-
"learning_rate":
|
| 396 |
-
"logits/chosen": -2.
|
| 397 |
-
"logits/rejected": -2.
|
| 398 |
-
"logps/chosen": -
|
| 399 |
-
"logps/rejected": -
|
| 400 |
-
"loss": 0.
|
| 401 |
-
"rewards/accuracies": 0.
|
| 402 |
-
"rewards/chosen": -0.
|
| 403 |
-
"rewards/margins":
|
| 404 |
-
"rewards/rejected": -
|
| 405 |
"step": 250
|
| 406 |
},
|
| 407 |
{
|
| 408 |
-
"epoch": 0.
|
| 409 |
-
"learning_rate":
|
| 410 |
-
"logits/chosen": -2.
|
| 411 |
-
"logits/rejected": -2.
|
| 412 |
-
"logps/chosen": -
|
| 413 |
-
"logps/rejected": -
|
| 414 |
-
"loss": 0.
|
| 415 |
-
"rewards/accuracies": 0.
|
| 416 |
-
"rewards/chosen": -0.
|
| 417 |
-
"rewards/margins":
|
| 418 |
-
"rewards/rejected": -
|
| 419 |
"step": 260
|
| 420 |
},
|
| 421 |
{
|
| 422 |
-
"epoch": 0.
|
| 423 |
-
"learning_rate":
|
| 424 |
-
"logits/chosen": -2.
|
| 425 |
-
"logits/rejected": -2.
|
| 426 |
-
"logps/chosen": -
|
| 427 |
-
"logps/rejected": -
|
| 428 |
-
"loss": 0.
|
| 429 |
-
"rewards/accuracies": 0.
|
| 430 |
-
"rewards/chosen": -0.
|
| 431 |
-
"rewards/margins":
|
| 432 |
-
"rewards/rejected": -
|
| 433 |
"step": 270
|
| 434 |
},
|
| 435 |
{
|
| 436 |
-
"epoch": 0.
|
| 437 |
-
"learning_rate":
|
| 438 |
-
"logits/chosen": -2.
|
| 439 |
-
"logits/rejected": -2.
|
| 440 |
-
"logps/chosen": -
|
| 441 |
-
"logps/rejected": -
|
| 442 |
-
"loss": 0.
|
| 443 |
-
"rewards/accuracies": 0.
|
| 444 |
-
"rewards/chosen": -1.
|
| 445 |
-
"rewards/margins":
|
| 446 |
-
"rewards/rejected": -
|
| 447 |
"step": 280
|
| 448 |
},
|
| 449 |
{
|
| 450 |
-
"epoch": 0.
|
| 451 |
-
"learning_rate":
|
| 452 |
-
"logits/chosen": -2.
|
| 453 |
-
"logits/rejected": -2.
|
| 454 |
-
"logps/chosen": -
|
| 455 |
-
"logps/rejected": -
|
| 456 |
-
"loss": 0.
|
| 457 |
-
"rewards/accuracies": 0.
|
| 458 |
-
"rewards/chosen": -
|
| 459 |
-
"rewards/margins":
|
| 460 |
-
"rewards/rejected": -
|
| 461 |
"step": 290
|
| 462 |
},
|
| 463 |
{
|
| 464 |
-
"epoch": 0.
|
| 465 |
-
"learning_rate":
|
| 466 |
-
"logits/chosen": -2.
|
| 467 |
-
"logits/rejected": -2.
|
| 468 |
-
"logps/chosen": -
|
| 469 |
-
"logps/rejected": -
|
| 470 |
-
"loss": 0.
|
| 471 |
-
"rewards/accuracies": 0.
|
| 472 |
-
"rewards/chosen": -1.
|
| 473 |
-
"rewards/margins":
|
| 474 |
-
"rewards/rejected": -
|
| 475 |
"step": 300
|
| 476 |
},
|
| 477 |
{
|
| 478 |
-
"epoch": 0.
|
| 479 |
-
"
|
| 480 |
-
"
|
| 481 |
-
"
|
| 482 |
-
"
|
| 483 |
-
"
|
| 484 |
-
"
|
| 485 |
-
"
|
| 486 |
-
"
|
| 487 |
-
"
|
| 488 |
-
"
|
| 489 |
-
"eval_samples_per_second": 37.515,
|
| 490 |
-
"eval_steps_per_second": 0.6,
|
| 491 |
-
"step": 300
|
| 492 |
-
},
|
| 493 |
-
{
|
| 494 |
-
"epoch": 0.65,
|
| 495 |
-
"learning_rate": 1.6583128063291573e-07,
|
| 496 |
-
"logits/chosen": -2.459578037261963,
|
| 497 |
-
"logits/rejected": -2.4456562995910645,
|
| 498 |
-
"logps/chosen": -363.7841796875,
|
| 499 |
-
"logps/rejected": -435.59478759765625,
|
| 500 |
-
"loss": 0.2183,
|
| 501 |
-
"rewards/accuracies": 0.737500011920929,
|
| 502 |
-
"rewards/chosen": -0.8949993848800659,
|
| 503 |
-
"rewards/margins": 0.6970807313919067,
|
| 504 |
-
"rewards/rejected": -1.592079997062683,
|
| 505 |
"step": 310
|
| 506 |
},
|
| 507 |
{
|
| 508 |
-
"epoch": 0.
|
| 509 |
-
"learning_rate":
|
| 510 |
-
"logits/chosen": -2.
|
| 511 |
-
"logits/rejected": -2.
|
| 512 |
-
"logps/chosen": -
|
| 513 |
-
"logps/rejected": -
|
| 514 |
-
"loss": 0.
|
| 515 |
-
"rewards/accuracies": 0.
|
| 516 |
-
"rewards/chosen": -
|
| 517 |
-
"rewards/margins":
|
| 518 |
-
"rewards/rejected": -
|
| 519 |
"step": 320
|
| 520 |
},
|
| 521 |
{
|
| 522 |
-
"epoch": 0.
|
| 523 |
-
"learning_rate":
|
| 524 |
-
"logits/chosen": -2.
|
| 525 |
-
"logits/rejected": -2.
|
| 526 |
-
"logps/chosen": -
|
| 527 |
-
"logps/rejected": -
|
| 528 |
-
"loss": 0.
|
| 529 |
-
"rewards/accuracies": 0.
|
| 530 |
-
"rewards/chosen": -
|
| 531 |
-
"rewards/margins":
|
| 532 |
-
"rewards/rejected": -
|
| 533 |
"step": 330
|
| 534 |
},
|
| 535 |
{
|
| 536 |
-
"epoch": 0.
|
| 537 |
-
"learning_rate":
|
| 538 |
-
"logits/chosen": -2.
|
| 539 |
-
"logits/rejected": -2.
|
| 540 |
-
"logps/chosen": -
|
| 541 |
-
"logps/rejected": -
|
| 542 |
-
"loss": 0.
|
| 543 |
-
"rewards/accuracies": 0.
|
| 544 |
-
"rewards/chosen": -1.
|
| 545 |
-
"rewards/margins":
|
| 546 |
-
"rewards/rejected": -
|
| 547 |
"step": 340
|
| 548 |
},
|
| 549 |
{
|
| 550 |
-
"epoch": 0.
|
| 551 |
-
"learning_rate":
|
| 552 |
-
"logits/chosen": -2.
|
| 553 |
-
"logits/rejected": -2.
|
| 554 |
-
"logps/chosen": -
|
| 555 |
-
"logps/rejected": -
|
| 556 |
-
"loss": 0.
|
| 557 |
-
"rewards/accuracies": 0.
|
| 558 |
-
"rewards/chosen": -
|
| 559 |
-
"rewards/margins":
|
| 560 |
-
"rewards/rejected": -
|
| 561 |
"step": 350
|
| 562 |
},
|
| 563 |
{
|
| 564 |
-
"epoch": 0.
|
| 565 |
-
"learning_rate":
|
| 566 |
-
"logits/chosen": -2.
|
| 567 |
-
"logits/rejected": -2.
|
| 568 |
-
"logps/chosen": -
|
| 569 |
-
"logps/rejected": -
|
| 570 |
-
"loss": 0.
|
| 571 |
-
"rewards/accuracies": 0.
|
| 572 |
-
"rewards/chosen": -
|
| 573 |
-
"rewards/margins":
|
| 574 |
-
"rewards/rejected": -
|
| 575 |
"step": 360
|
| 576 |
},
|
| 577 |
{
|
| 578 |
-
"epoch": 0.
|
| 579 |
-
"learning_rate":
|
| 580 |
-
"logits/chosen": -2.
|
| 581 |
-
"logits/rejected": -2.
|
| 582 |
-
"logps/chosen": -
|
| 583 |
-
"logps/rejected": -
|
| 584 |
-
"loss": 0.
|
| 585 |
-
"rewards/accuracies": 0.
|
| 586 |
-
"rewards/chosen": -
|
| 587 |
-
"rewards/margins":
|
| 588 |
-
"rewards/rejected": -
|
| 589 |
"step": 370
|
| 590 |
},
|
| 591 |
{
|
| 592 |
-
"epoch": 0.
|
| 593 |
-
"learning_rate":
|
| 594 |
-
"logits/chosen": -2.
|
| 595 |
-
"logits/rejected": -2.
|
| 596 |
-
"logps/chosen": -
|
| 597 |
-
"logps/rejected": -
|
| 598 |
-
"loss": 0.
|
| 599 |
-
"rewards/accuracies": 0.
|
| 600 |
-
"rewards/chosen": -1.
|
| 601 |
-
"rewards/margins":
|
| 602 |
-
"rewards/rejected": -
|
| 603 |
"step": 380
|
| 604 |
},
|
| 605 |
{
|
| 606 |
-
"epoch": 0.
|
| 607 |
-
"learning_rate":
|
| 608 |
-
"logits/chosen": -2.
|
| 609 |
-
"logits/rejected": -2.
|
| 610 |
-
"logps/chosen": -
|
| 611 |
-
"logps/rejected": -
|
| 612 |
-
"loss": 0.
|
| 613 |
-
"rewards/accuracies": 0.
|
| 614 |
-
"rewards/chosen": -
|
| 615 |
-
"rewards/margins":
|
| 616 |
-
"rewards/rejected": -
|
| 617 |
"step": 390
|
| 618 |
},
|
| 619 |
{
|
| 620 |
-
"epoch": 0.
|
| 621 |
-
"learning_rate":
|
| 622 |
-
"logits/chosen": -2.
|
| 623 |
-
"logits/rejected": -2.
|
| 624 |
-
"logps/chosen": -
|
| 625 |
-
"logps/rejected": -
|
| 626 |
-
"loss": 0.
|
| 627 |
-
"rewards/accuracies": 0.
|
| 628 |
-
"rewards/chosen": -
|
| 629 |
-
"rewards/margins":
|
| 630 |
-
"rewards/rejected": -
|
| 631 |
-
"step": 400
|
| 632 |
-
},
|
| 633 |
-
{
|
| 634 |
-
"epoch": 0.84,
|
| 635 |
-
"eval_logits/chosen": -2.4425089359283447,
|
| 636 |
-
"eval_logits/rejected": -2.4274942874908447,
|
| 637 |
-
"eval_logps/chosen": -346.5982666015625,
|
| 638 |
-
"eval_logps/rejected": -409.02783203125,
|
| 639 |
-
"eval_loss": 0.21434982120990753,
|
| 640 |
-
"eval_rewards/accuracies": 0.703125,
|
| 641 |
-
"eval_rewards/chosen": -0.8955863118171692,
|
| 642 |
-
"eval_rewards/margins": 0.6211590766906738,
|
| 643 |
-
"eval_rewards/rejected": -1.5167453289031982,
|
| 644 |
-
"eval_runtime": 53.3158,
|
| 645 |
-
"eval_samples_per_second": 37.512,
|
| 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": -2.346083164215088,
|
| 653 |
-
"logits/rejected": -2.3530008792877197,
|
| 654 |
-
"logps/chosen": -359.29718017578125,
|
| 655 |
-
"logps/rejected": -380.91094970703125,
|
| 656 |
-
"loss": 0.2113,
|
| 657 |
-
"rewards/accuracies": 0.7250000238418579,
|
| 658 |
-
"rewards/chosen": -0.9364844560623169,
|
| 659 |
-
"rewards/margins": 0.4689885973930359,
|
| 660 |
-
"rewards/rejected": -1.4054731130599976,
|
| 661 |
-
"step": 410
|
| 662 |
-
},
|
| 663 |
-
{
|
| 664 |
-
"epoch": 0.88,
|
| 665 |
-
"learning_rate": 2.2111614344599684e-08,
|
| 666 |
-
"logits/chosen": -2.3877267837524414,
|
| 667 |
-
"logits/rejected": -2.3703396320343018,
|
| 668 |
-
"logps/chosen": -347.065673828125,
|
| 669 |
-
"logps/rejected": -389.3710021972656,
|
| 670 |
-
"loss": 0.2072,
|
| 671 |
-
"rewards/accuracies": 0.7124999761581421,
|
| 672 |
-
"rewards/chosen": -0.8826552629470825,
|
| 673 |
-
"rewards/margins": 0.5836518406867981,
|
| 674 |
-
"rewards/rejected": -1.4663069248199463,
|
| 675 |
-
"step": 420
|
| 676 |
-
},
|
| 677 |
-
{
|
| 678 |
-
"epoch": 0.9,
|
| 679 |
-
"learning_rate": 1.521597710086439e-08,
|
| 680 |
-
"logits/chosen": -2.427062511444092,
|
| 681 |
-
"logits/rejected": -2.3968257904052734,
|
| 682 |
-
"logps/chosen": -384.0503845214844,
|
| 683 |
-
"logps/rejected": -394.9156799316406,
|
| 684 |
-
"loss": 0.2142,
|
| 685 |
-
"rewards/accuracies": 0.699999988079071,
|
| 686 |
-
"rewards/chosen": -1.0325697660446167,
|
| 687 |
-
"rewards/margins": 0.47820836305618286,
|
| 688 |
-
"rewards/rejected": -1.5107781887054443,
|
| 689 |
-
"step": 430
|
| 690 |
-
},
|
| 691 |
-
{
|
| 692 |
-
"epoch": 0.92,
|
| 693 |
-
"learning_rate": 9.57301420397924e-09,
|
| 694 |
-
"logits/chosen": -2.4799857139587402,
|
| 695 |
-
"logits/rejected": -2.432332992553711,
|
| 696 |
-
"logps/chosen": -368.3861083984375,
|
| 697 |
-
"logps/rejected": -357.58380126953125,
|
| 698 |
-
"loss": 0.2142,
|
| 699 |
-
"rewards/accuracies": 0.7124999761581421,
|
| 700 |
-
"rewards/chosen": -0.9250670671463013,
|
| 701 |
-
"rewards/margins": 0.450716108083725,
|
| 702 |
-
"rewards/rejected": -1.375783085823059,
|
| 703 |
-
"step": 440
|
| 704 |
-
},
|
| 705 |
-
{
|
| 706 |
-
"epoch": 0.94,
|
| 707 |
-
"learning_rate": 5.212833302556258e-09,
|
| 708 |
-
"logits/chosen": -2.4059414863586426,
|
| 709 |
-
"logits/rejected": -2.3920960426330566,
|
| 710 |
-
"logps/chosen": -371.8204650878906,
|
| 711 |
-
"logps/rejected": -386.82269287109375,
|
| 712 |
-
"loss": 0.215,
|
| 713 |
-
"rewards/accuracies": 0.71875,
|
| 714 |
-
"rewards/chosen": -0.8487032651901245,
|
| 715 |
-
"rewards/margins": 0.4881046414375305,
|
| 716 |
-
"rewards/rejected": -1.3368079662322998,
|
| 717 |
-
"step": 450
|
| 718 |
-
},
|
| 719 |
-
{
|
| 720 |
-
"epoch": 0.96,
|
| 721 |
-
"learning_rate": 2.158697848236607e-09,
|
| 722 |
-
"logits/chosen": -2.456326961517334,
|
| 723 |
-
"logits/rejected": -2.4320716857910156,
|
| 724 |
-
"logps/chosen": -348.0980529785156,
|
| 725 |
-
"logps/rejected": -381.2003173828125,
|
| 726 |
-
"loss": 0.2155,
|
| 727 |
-
"rewards/accuracies": 0.706250011920929,
|
| 728 |
-
"rewards/chosen": -0.8624986410140991,
|
| 729 |
-
"rewards/margins": 0.5545080900192261,
|
| 730 |
-
"rewards/rejected": -1.4170067310333252,
|
| 731 |
-
"step": 460
|
| 732 |
-
},
|
| 733 |
-
{
|
| 734 |
-
"epoch": 0.98,
|
| 735 |
-
"learning_rate": 4.269029751107489e-10,
|
| 736 |
-
"logits/chosen": -2.4110119342803955,
|
| 737 |
-
"logits/rejected": -2.383718967437744,
|
| 738 |
-
"logps/chosen": -323.4446716308594,
|
| 739 |
-
"logps/rejected": -392.60064697265625,
|
| 740 |
-
"loss": 0.2178,
|
| 741 |
-
"rewards/accuracies": 0.768750011920929,
|
| 742 |
-
"rewards/chosen": -0.8438596725463867,
|
| 743 |
-
"rewards/margins": 0.6573068499565125,
|
| 744 |
-
"rewards/rejected": -1.501166582107544,
|
| 745 |
-
"step": 470
|
| 746 |
-
},
|
| 747 |
{
|
| 748 |
"epoch": 1.0,
|
| 749 |
-
"step":
|
| 750 |
"total_flos": 0.0,
|
| 751 |
-
"train_loss": 0.
|
| 752 |
-
"train_runtime":
|
| 753 |
-
"train_samples_per_second":
|
| 754 |
-
"train_steps_per_second": 0.
|
| 755 |
}
|
| 756 |
],
|
| 757 |
"logging_steps": 10,
|
| 758 |
-
"max_steps":
|
| 759 |
"num_train_epochs": 1,
|
| 760 |
"save_steps": 1000,
|
| 761 |
"total_flos": 0.0,
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": null,
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.998766954377312,
|
| 5 |
+
"eval_steps": 1000,
|
| 6 |
+
"global_step": 405,
|
| 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": 1.2195121951219512e-08,
|
| 14 |
+
"logits/chosen": -2.8681135177612305,
|
| 15 |
+
"logits/rejected": -2.8858838081359863,
|
| 16 |
+
"logps/chosen": -518.1907958984375,
|
| 17 |
+
"logps/rejected": -109.31971740722656,
|
| 18 |
+
"loss": 0.5582,
|
| 19 |
"rewards/accuracies": 0.0,
|
| 20 |
"rewards/chosen": 0.0,
|
| 21 |
"rewards/margins": 0.0,
|
|
|
|
| 24 |
},
|
| 25 |
{
|
| 26 |
"epoch": 0.02,
|
| 27 |
+
"learning_rate": 1.219512195121951e-07,
|
| 28 |
+
"logits/chosen": -2.7987241744995117,
|
| 29 |
+
"logits/rejected": -2.7522923946380615,
|
| 30 |
+
"logps/chosen": -434.1634521484375,
|
| 31 |
+
"logps/rejected": -114.18580627441406,
|
| 32 |
+
"loss": 0.5551,
|
| 33 |
+
"rewards/accuracies": 0.5069444179534912,
|
| 34 |
+
"rewards/chosen": 0.0007126106647774577,
|
| 35 |
+
"rewards/margins": 0.0012787013547495008,
|
| 36 |
+
"rewards/rejected": -0.000566090689972043,
|
| 37 |
"step": 10
|
| 38 |
},
|
| 39 |
{
|
| 40 |
+
"epoch": 0.05,
|
| 41 |
+
"learning_rate": 2.439024390243902e-07,
|
| 42 |
+
"logits/chosen": -2.820725917816162,
|
| 43 |
+
"logits/rejected": -2.7975618839263916,
|
| 44 |
+
"logps/chosen": -417.19793701171875,
|
| 45 |
+
"logps/rejected": -118.11334228515625,
|
| 46 |
+
"loss": 0.5389,
|
| 47 |
+
"rewards/accuracies": 0.762499988079071,
|
| 48 |
+
"rewards/chosen": 0.02059975638985634,
|
| 49 |
+
"rewards/margins": 0.03748037666082382,
|
| 50 |
+
"rewards/rejected": -0.016880620270967484,
|
| 51 |
"step": 20
|
| 52 |
},
|
| 53 |
{
|
| 54 |
+
"epoch": 0.07,
|
| 55 |
+
"learning_rate": 3.6585365853658536e-07,
|
| 56 |
+
"logits/chosen": -2.654127597808838,
|
| 57 |
+
"logits/rejected": -2.641984224319458,
|
| 58 |
+
"logps/chosen": -398.6944580078125,
|
| 59 |
+
"logps/rejected": -126.52714538574219,
|
| 60 |
+
"loss": 0.4667,
|
| 61 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 62 |
+
"rewards/chosen": 0.07748116552829742,
|
| 63 |
+
"rewards/margins": 0.2089122235774994,
|
| 64 |
+
"rewards/rejected": -0.13143105804920197,
|
| 65 |
"step": 30
|
| 66 |
},
|
| 67 |
{
|
| 68 |
+
"epoch": 0.1,
|
| 69 |
+
"learning_rate": 4.878048780487804e-07,
|
| 70 |
+
"logits/chosen": -2.538156509399414,
|
| 71 |
+
"logits/rejected": -2.523983955383301,
|
| 72 |
+
"logps/chosen": -384.8261413574219,
|
| 73 |
+
"logps/rejected": -170.25672912597656,
|
| 74 |
+
"loss": 0.3743,
|
| 75 |
+
"rewards/accuracies": 0.737500011920929,
|
| 76 |
+
"rewards/chosen": 0.021691232919692993,
|
| 77 |
+
"rewards/margins": 0.43257051706314087,
|
| 78 |
+
"rewards/rejected": -0.4108792245388031,
|
| 79 |
"step": 40
|
| 80 |
},
|
| 81 |
{
|
| 82 |
+
"epoch": 0.12,
|
| 83 |
+
"learning_rate": 4.992461696250783e-07,
|
| 84 |
+
"logits/chosen": -2.4283156394958496,
|
| 85 |
+
"logits/rejected": -2.394186496734619,
|
| 86 |
+
"logps/chosen": -435.05364990234375,
|
| 87 |
+
"logps/rejected": -218.2919464111328,
|
| 88 |
+
"loss": 0.2788,
|
| 89 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 90 |
+
"rewards/chosen": -0.15316179394721985,
|
| 91 |
+
"rewards/margins": 0.794213056564331,
|
| 92 |
+
"rewards/rejected": -0.947374701499939,
|
| 93 |
"step": 50
|
| 94 |
},
|
| 95 |
{
|
| 96 |
+
"epoch": 0.15,
|
| 97 |
+
"learning_rate": 4.966461721767899e-07,
|
| 98 |
+
"logits/chosen": -2.3560433387756348,
|
| 99 |
+
"logits/rejected": -2.3106768131256104,
|
| 100 |
+
"logps/chosen": -437.6812438964844,
|
| 101 |
+
"logps/rejected": -239.7769012451172,
|
| 102 |
+
"loss": 0.2266,
|
| 103 |
+
"rewards/accuracies": 0.762499988079071,
|
| 104 |
+
"rewards/chosen": -0.3268889784812927,
|
| 105 |
+
"rewards/margins": 0.9203512072563171,
|
| 106 |
+
"rewards/rejected": -1.2472401857376099,
|
| 107 |
"step": 60
|
| 108 |
},
|
| 109 |
{
|
| 110 |
+
"epoch": 0.17,
|
| 111 |
+
"learning_rate": 4.922100518015975e-07,
|
| 112 |
+
"logits/chosen": -2.331510543823242,
|
| 113 |
+
"logits/rejected": -2.284951686859131,
|
| 114 |
+
"logps/chosen": -424.73333740234375,
|
| 115 |
+
"logps/rejected": -268.073486328125,
|
| 116 |
+
"loss": 0.1727,
|
| 117 |
+
"rewards/accuracies": 0.762499988079071,
|
| 118 |
+
"rewards/chosen": -0.44431695342063904,
|
| 119 |
+
"rewards/margins": 1.1743532419204712,
|
| 120 |
+
"rewards/rejected": -1.618670105934143,
|
| 121 |
"step": 70
|
| 122 |
},
|
| 123 |
{
|
| 124 |
+
"epoch": 0.2,
|
| 125 |
+
"learning_rate": 4.859708325770919e-07,
|
| 126 |
+
"logits/chosen": -2.393051862716675,
|
| 127 |
+
"logits/rejected": -2.33725643157959,
|
| 128 |
+
"logps/chosen": -468.8316345214844,
|
| 129 |
+
"logps/rejected": -316.95086669921875,
|
| 130 |
+
"loss": 0.1441,
|
| 131 |
+
"rewards/accuracies": 0.8125,
|
| 132 |
+
"rewards/chosen": -0.5438361167907715,
|
| 133 |
+
"rewards/margins": 1.4411370754241943,
|
| 134 |
+
"rewards/rejected": -1.9849731922149658,
|
| 135 |
"step": 80
|
| 136 |
},
|
| 137 |
{
|
| 138 |
+
"epoch": 0.22,
|
| 139 |
+
"learning_rate": 4.779749614980225e-07,
|
| 140 |
+
"logits/chosen": -2.372483491897583,
|
| 141 |
+
"logits/rejected": -2.3344340324401855,
|
| 142 |
+
"logps/chosen": -466.607421875,
|
| 143 |
+
"logps/rejected": -327.7405700683594,
|
| 144 |
+
"loss": 0.1325,
|
| 145 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 146 |
+
"rewards/chosen": -0.45430439710617065,
|
| 147 |
+
"rewards/margins": 1.7159345149993896,
|
| 148 |
+
"rewards/rejected": -2.170238733291626,
|
| 149 |
"step": 90
|
| 150 |
},
|
| 151 |
{
|
| 152 |
+
"epoch": 0.25,
|
| 153 |
+
"learning_rate": 4.682819627081427e-07,
|
| 154 |
+
"logits/chosen": -2.343848466873169,
|
| 155 |
+
"logits/rejected": -2.300412654876709,
|
| 156 |
+
"logps/chosen": -498.5199279785156,
|
| 157 |
+
"logps/rejected": -356.87860107421875,
|
| 158 |
+
"loss": 0.1141,
|
| 159 |
+
"rewards/accuracies": 0.800000011920929,
|
| 160 |
+
"rewards/chosen": -0.500357449054718,
|
| 161 |
+
"rewards/margins": 1.946504831314087,
|
| 162 |
+
"rewards/rejected": -2.446861743927002,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 163 |
"step": 100
|
| 164 |
},
|
| 165 |
{
|
| 166 |
+
"epoch": 0.27,
|
| 167 |
+
"learning_rate": 4.569639943810477e-07,
|
| 168 |
+
"logits/chosen": -2.2943625450134277,
|
| 169 |
+
"logits/rejected": -2.2243731021881104,
|
| 170 |
+
"logps/chosen": -475.5547790527344,
|
| 171 |
+
"logps/rejected": -396.2657165527344,
|
| 172 |
+
"loss": 0.0898,
|
| 173 |
+
"rewards/accuracies": 0.75,
|
| 174 |
+
"rewards/chosen": -1.1803233623504639,
|
| 175 |
+
"rewards/margins": 1.67868971824646,
|
| 176 |
+
"rewards/rejected": -2.859013080596924,
|
| 177 |
"step": 110
|
| 178 |
},
|
| 179 |
{
|
| 180 |
+
"epoch": 0.3,
|
| 181 |
+
"learning_rate": 4.4410531154874543e-07,
|
| 182 |
+
"logits/chosen": -2.313946485519409,
|
| 183 |
+
"logits/rejected": -2.2655868530273438,
|
| 184 |
+
"logps/chosen": -490.6239318847656,
|
| 185 |
+
"logps/rejected": -387.259765625,
|
| 186 |
+
"loss": 0.0963,
|
| 187 |
+
"rewards/accuracies": 0.824999988079071,
|
| 188 |
+
"rewards/chosen": -0.9641086459159851,
|
| 189 |
+
"rewards/margins": 1.7083423137664795,
|
| 190 |
+
"rewards/rejected": -2.6724514961242676,
|
| 191 |
"step": 120
|
| 192 |
},
|
| 193 |
{
|
| 194 |
+
"epoch": 0.32,
|
| 195 |
+
"learning_rate": 4.298016388768561e-07,
|
| 196 |
+
"logits/chosen": -2.270263910293579,
|
| 197 |
+
"logits/rejected": -2.234313488006592,
|
| 198 |
+
"logps/chosen": -485.26605224609375,
|
| 199 |
+
"logps/rejected": -381.05609130859375,
|
| 200 |
+
"loss": 0.0985,
|
| 201 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 202 |
+
"rewards/chosen": -1.1396474838256836,
|
| 203 |
+
"rewards/margins": 1.5876901149749756,
|
| 204 |
+
"rewards/rejected": -2.727337598800659,
|
| 205 |
"step": 130
|
| 206 |
},
|
| 207 |
{
|
| 208 |
+
"epoch": 0.35,
|
| 209 |
+
"learning_rate": 4.1415945805573005e-07,
|
| 210 |
+
"logits/chosen": -2.252371311187744,
|
| 211 |
+
"logits/rejected": -2.2070510387420654,
|
| 212 |
+
"logps/chosen": -512.5037841796875,
|
| 213 |
+
"logps/rejected": -396.6216735839844,
|
| 214 |
+
"loss": 0.0999,
|
| 215 |
+
"rewards/accuracies": 0.887499988079071,
|
| 216 |
+
"rewards/chosen": -0.8260290026664734,
|
| 217 |
+
"rewards/margins": 1.9724273681640625,
|
| 218 |
+
"rewards/rejected": -2.7984564304351807,
|
| 219 |
"step": 140
|
| 220 |
},
|
| 221 |
{
|
| 222 |
+
"epoch": 0.37,
|
| 223 |
+
"learning_rate": 3.972952151123984e-07,
|
| 224 |
+
"logits/chosen": -2.2577097415924072,
|
| 225 |
+
"logits/rejected": -2.193302869796753,
|
| 226 |
+
"logps/chosen": -537.8040771484375,
|
| 227 |
+
"logps/rejected": -436.078369140625,
|
| 228 |
+
"loss": 0.0769,
|
| 229 |
+
"rewards/accuracies": 0.793749988079071,
|
| 230 |
+
"rewards/chosen": -1.2730529308319092,
|
| 231 |
+
"rewards/margins": 1.9622818231582642,
|
| 232 |
+
"rewards/rejected": -3.235334873199463,
|
| 233 |
"step": 150
|
| 234 |
},
|
| 235 |
{
|
| 236 |
+
"epoch": 0.39,
|
| 237 |
+
"learning_rate": 3.793344535444142e-07,
|
| 238 |
+
"logits/chosen": -2.242403984069824,
|
| 239 |
+
"logits/rejected": -2.1921000480651855,
|
| 240 |
+
"logps/chosen": -575.3533325195312,
|
| 241 |
+
"logps/rejected": -464.304443359375,
|
| 242 |
+
"loss": 0.0831,
|
| 243 |
+
"rewards/accuracies": 0.762499988079071,
|
| 244 |
+
"rewards/chosen": -1.6633422374725342,
|
| 245 |
+
"rewards/margins": 1.7485841512680054,
|
| 246 |
+
"rewards/rejected": -3.41192626953125,
|
| 247 |
"step": 160
|
| 248 |
},
|
| 249 |
{
|
| 250 |
+
"epoch": 0.42,
|
| 251 |
+
"learning_rate": 3.604108797288461e-07,
|
| 252 |
+
"logits/chosen": -2.2311742305755615,
|
| 253 |
+
"logits/rejected": -2.215085983276367,
|
| 254 |
+
"logps/chosen": -447.465576171875,
|
| 255 |
+
"logps/rejected": -357.7833557128906,
|
| 256 |
+
"loss": 0.1106,
|
| 257 |
+
"rewards/accuracies": 0.75,
|
| 258 |
+
"rewards/chosen": -1.085645318031311,
|
| 259 |
+
"rewards/margins": 1.4540026187896729,
|
| 260 |
+
"rewards/rejected": -2.5396475791931152,
|
| 261 |
"step": 170
|
| 262 |
},
|
| 263 |
{
|
| 264 |
+
"epoch": 0.44,
|
| 265 |
+
"learning_rate": 3.40665367563858e-07,
|
| 266 |
+
"logits/chosen": -2.283315658569336,
|
| 267 |
+
"logits/rejected": -2.2039597034454346,
|
| 268 |
+
"logps/chosen": -531.3043212890625,
|
| 269 |
+
"logps/rejected": -403.9816589355469,
|
| 270 |
+
"loss": 0.1024,
|
| 271 |
+
"rewards/accuracies": 0.8125,
|
| 272 |
+
"rewards/chosen": -0.712272047996521,
|
| 273 |
+
"rewards/margins": 2.0688250064849854,
|
| 274 |
+
"rewards/rejected": -2.781096935272217,
|
| 275 |
"step": 180
|
| 276 |
},
|
| 277 |
{
|
| 278 |
+
"epoch": 0.47,
|
| 279 |
+
"learning_rate": 3.202449097526798e-07,
|
| 280 |
+
"logits/chosen": -2.248110294342041,
|
| 281 |
+
"logits/rejected": -2.2098121643066406,
|
| 282 |
+
"logps/chosen": -515.7914428710938,
|
| 283 |
+
"logps/rejected": -415.6000061035156,
|
| 284 |
+
"loss": 0.087,
|
| 285 |
+
"rewards/accuracies": 0.762499988079071,
|
| 286 |
+
"rewards/chosen": -1.106316328048706,
|
| 287 |
+
"rewards/margins": 1.8899608850479126,
|
| 288 |
+
"rewards/rejected": -2.9962775707244873,
|
| 289 |
"step": 190
|
| 290 |
},
|
| 291 |
{
|
| 292 |
+
"epoch": 0.49,
|
| 293 |
+
"learning_rate": 2.993015235369905e-07,
|
| 294 |
+
"logits/chosen": -2.2014596462249756,
|
| 295 |
+
"logits/rejected": -2.1617417335510254,
|
| 296 |
+
"logps/chosen": -526.7307739257812,
|
| 297 |
+
"logps/rejected": -452.57415771484375,
|
| 298 |
+
"loss": 0.0786,
|
| 299 |
+
"rewards/accuracies": 0.78125,
|
| 300 |
+
"rewards/chosen": -1.4753978252410889,
|
| 301 |
+
"rewards/margins": 1.9113044738769531,
|
| 302 |
+
"rewards/rejected": -3.386702060699463,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 303 |
"step": 200
|
| 304 |
},
|
| 305 |
{
|
| 306 |
+
"epoch": 0.52,
|
| 307 |
+
"learning_rate": 2.7799111902582693e-07,
|
| 308 |
+
"logits/chosen": -2.2283272743225098,
|
| 309 |
+
"logits/rejected": -2.1539306640625,
|
| 310 |
+
"logps/chosen": -547.9268798828125,
|
| 311 |
+
"logps/rejected": -442.84716796875,
|
| 312 |
+
"loss": 0.0712,
|
| 313 |
+
"rewards/accuracies": 0.793749988079071,
|
| 314 |
+
"rewards/chosen": -1.0882245302200317,
|
| 315 |
+
"rewards/margins": 2.2409071922302246,
|
| 316 |
+
"rewards/rejected": -3.329131603240967,
|
| 317 |
"step": 210
|
| 318 |
},
|
| 319 |
{
|
| 320 |
+
"epoch": 0.54,
|
| 321 |
+
"learning_rate": 2.564723385445869e-07,
|
| 322 |
+
"logits/chosen": -2.300952911376953,
|
| 323 |
+
"logits/rejected": -2.2398998737335205,
|
| 324 |
+
"logps/chosen": -535.0980224609375,
|
| 325 |
+
"logps/rejected": -421.75799560546875,
|
| 326 |
+
"loss": 0.1106,
|
| 327 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 328 |
+
"rewards/chosen": -1.0268903970718384,
|
| 329 |
+
"rewards/margins": 1.9092063903808594,
|
| 330 |
+
"rewards/rejected": -2.9360969066619873,
|
| 331 |
"step": 220
|
| 332 |
},
|
| 333 |
{
|
| 334 |
+
"epoch": 0.57,
|
| 335 |
+
"learning_rate": 2.3490537564442845e-07,
|
| 336 |
+
"logits/chosen": -2.2512927055358887,
|
| 337 |
+
"logits/rejected": -2.1920220851898193,
|
| 338 |
+
"logps/chosen": -481.20947265625,
|
| 339 |
+
"logps/rejected": -388.5928649902344,
|
| 340 |
+
"loss": 0.1072,
|
| 341 |
+
"rewards/accuracies": 0.731249988079071,
|
| 342 |
+
"rewards/chosen": -0.936708927154541,
|
| 343 |
+
"rewards/margins": 1.7015047073364258,
|
| 344 |
+
"rewards/rejected": -2.638213634490967,
|
| 345 |
"step": 230
|
| 346 |
},
|
| 347 |
{
|
| 348 |
+
"epoch": 0.59,
|
| 349 |
+
"learning_rate": 2.1345078256378801e-07,
|
| 350 |
+
"logits/chosen": -2.2840850353240967,
|
| 351 |
+
"logits/rejected": -2.223998546600342,
|
| 352 |
+
"logps/chosen": -525.611083984375,
|
| 353 |
+
"logps/rejected": -408.7984924316406,
|
| 354 |
+
"loss": 0.0977,
|
| 355 |
+
"rewards/accuracies": 0.768750011920929,
|
| 356 |
+
"rewards/chosen": -1.13532555103302,
|
| 357 |
+
"rewards/margins": 1.788002371788025,
|
| 358 |
+
"rewards/rejected": -2.923327922821045,
|
| 359 |
"step": 240
|
| 360 |
},
|
| 361 |
{
|
| 362 |
+
"epoch": 0.62,
|
| 363 |
+
"learning_rate": 1.9226827501969865e-07,
|
| 364 |
+
"logits/chosen": -2.2815299034118652,
|
| 365 |
+
"logits/rejected": -2.2289319038391113,
|
| 366 |
+
"logps/chosen": -510.8829040527344,
|
| 367 |
+
"logps/rejected": -408.6348571777344,
|
| 368 |
+
"loss": 0.1022,
|
| 369 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 370 |
+
"rewards/chosen": -0.9617199897766113,
|
| 371 |
+
"rewards/margins": 2.0366153717041016,
|
| 372 |
+
"rewards/rejected": -2.998335361480713,
|
| 373 |
"step": 250
|
| 374 |
},
|
| 375 |
{
|
| 376 |
+
"epoch": 0.64,
|
| 377 |
+
"learning_rate": 1.715155432264775e-07,
|
| 378 |
+
"logits/chosen": -2.257248640060425,
|
| 379 |
+
"logits/rejected": -2.162848949432373,
|
| 380 |
+
"logps/chosen": -476.12603759765625,
|
| 381 |
+
"logps/rejected": -383.61993408203125,
|
| 382 |
+
"loss": 0.0978,
|
| 383 |
+
"rewards/accuracies": 0.78125,
|
| 384 |
+
"rewards/chosen": -0.8627266883850098,
|
| 385 |
+
"rewards/margins": 1.9056837558746338,
|
| 386 |
+
"rewards/rejected": -2.7684102058410645,
|
| 387 |
"step": 260
|
| 388 |
},
|
| 389 |
{
|
| 390 |
+
"epoch": 0.67,
|
| 391 |
+
"learning_rate": 1.51347077992983e-07,
|
| 392 |
+
"logits/chosen": -2.2847957611083984,
|
| 393 |
+
"logits/rejected": -2.197834014892578,
|
| 394 |
+
"logps/chosen": -536.037109375,
|
| 395 |
+
"logps/rejected": -406.7389221191406,
|
| 396 |
+
"loss": 0.0889,
|
| 397 |
+
"rewards/accuracies": 0.8125,
|
| 398 |
+
"rewards/chosen": -0.9227035641670227,
|
| 399 |
+
"rewards/margins": 1.997865915298462,
|
| 400 |
+
"rewards/rejected": -2.9205691814422607,
|
| 401 |
"step": 270
|
| 402 |
},
|
| 403 |
{
|
| 404 |
+
"epoch": 0.69,
|
| 405 |
+
"learning_rate": 1.3191302063739906e-07,
|
| 406 |
+
"logits/chosen": -2.288609743118286,
|
| 407 |
+
"logits/rejected": -2.211268424987793,
|
| 408 |
+
"logps/chosen": -509.29156494140625,
|
| 409 |
+
"logps/rejected": -429.5159606933594,
|
| 410 |
+
"loss": 0.1021,
|
| 411 |
+
"rewards/accuracies": 0.8125,
|
| 412 |
+
"rewards/chosen": -1.05027174949646,
|
| 413 |
+
"rewards/margins": 1.9743726253509521,
|
| 414 |
+
"rewards/rejected": -3.024644136428833,
|
| 415 |
"step": 280
|
| 416 |
},
|
| 417 |
{
|
| 418 |
+
"epoch": 0.72,
|
| 419 |
+
"learning_rate": 1.1335804528119475e-07,
|
| 420 |
+
"logits/chosen": -2.2901360988616943,
|
| 421 |
+
"logits/rejected": -2.2138819694519043,
|
| 422 |
+
"logps/chosen": -530.1524658203125,
|
| 423 |
+
"logps/rejected": -411.786376953125,
|
| 424 |
+
"loss": 0.1097,
|
| 425 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 426 |
+
"rewards/chosen": -0.8590289950370789,
|
| 427 |
+
"rewards/margins": 2.134665012359619,
|
| 428 |
+
"rewards/rejected": -2.993694305419922,
|
| 429 |
"step": 290
|
| 430 |
},
|
| 431 |
{
|
| 432 |
+
"epoch": 0.74,
|
| 433 |
+
"learning_rate": 9.582028184286423e-08,
|
| 434 |
+
"logits/chosen": -2.3320322036743164,
|
| 435 |
+
"logits/rejected": -2.3061583042144775,
|
| 436 |
+
"logps/chosen": -540.7392578125,
|
| 437 |
+
"logps/rejected": -452.91729736328125,
|
| 438 |
+
"loss": 0.0991,
|
| 439 |
+
"rewards/accuracies": 0.78125,
|
| 440 |
+
"rewards/chosen": -1.0294259786605835,
|
| 441 |
+
"rewards/margins": 2.121091365814209,
|
| 442 |
+
"rewards/rejected": -3.150517225265503,
|
| 443 |
"step": 300
|
| 444 |
},
|
| 445 |
{
|
| 446 |
+
"epoch": 0.76,
|
| 447 |
+
"learning_rate": 7.943028774907065e-08,
|
| 448 |
+
"logits/chosen": -2.298715353012085,
|
| 449 |
+
"logits/rejected": -2.212947130203247,
|
| 450 |
+
"logps/chosen": -518.6718139648438,
|
| 451 |
+
"logps/rejected": -410.0641174316406,
|
| 452 |
+
"loss": 0.0799,
|
| 453 |
+
"rewards/accuracies": 0.8125,
|
| 454 |
+
"rewards/chosen": -0.9973870515823364,
|
| 455 |
+
"rewards/margins": 2.0150723457336426,
|
| 456 |
+
"rewards/rejected": -3.0124592781066895,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 457 |
"step": 310
|
| 458 |
},
|
| 459 |
{
|
| 460 |
+
"epoch": 0.79,
|
| 461 |
+
"learning_rate": 6.431007601814637e-08,
|
| 462 |
+
"logits/chosen": -2.262784242630005,
|
| 463 |
+
"logits/rejected": -2.179920196533203,
|
| 464 |
+
"logps/chosen": -515.6483154296875,
|
| 465 |
+
"logps/rejected": -430.5453186035156,
|
| 466 |
+
"loss": 0.0831,
|
| 467 |
+
"rewards/accuracies": 0.706250011920929,
|
| 468 |
+
"rewards/chosen": -1.3527257442474365,
|
| 469 |
+
"rewards/margins": 1.8855050802230835,
|
| 470 |
+
"rewards/rejected": -3.2382309436798096,
|
| 471 |
"step": 320
|
| 472 |
},
|
| 473 |
{
|
| 474 |
+
"epoch": 0.81,
|
| 475 |
+
"learning_rate": 5.0572206951246e-08,
|
| 476 |
+
"logits/chosen": -2.2783610820770264,
|
| 477 |
+
"logits/rejected": -2.207493305206299,
|
| 478 |
+
"logps/chosen": -545.3638916015625,
|
| 479 |
+
"logps/rejected": -452.2981872558594,
|
| 480 |
+
"loss": 0.0755,
|
| 481 |
+
"rewards/accuracies": 0.8125,
|
| 482 |
+
"rewards/chosen": -1.209948182106018,
|
| 483 |
+
"rewards/margins": 2.0318679809570312,
|
| 484 |
+
"rewards/rejected": -3.241816759109497,
|
| 485 |
"step": 330
|
| 486 |
},
|
| 487 |
{
|
| 488 |
+
"epoch": 0.84,
|
| 489 |
+
"learning_rate": 3.831895019292897e-08,
|
| 490 |
+
"logits/chosen": -2.311371088027954,
|
| 491 |
+
"logits/rejected": -2.2250900268554688,
|
| 492 |
+
"logps/chosen": -602.3062133789062,
|
| 493 |
+
"logps/rejected": -488.07012939453125,
|
| 494 |
+
"loss": 0.0825,
|
| 495 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 496 |
+
"rewards/chosen": -1.3970798254013062,
|
| 497 |
+
"rewards/margins": 2.1450533866882324,
|
| 498 |
+
"rewards/rejected": -3.542132616043091,
|
| 499 |
"step": 340
|
| 500 |
},
|
| 501 |
{
|
| 502 |
+
"epoch": 0.86,
|
| 503 |
+
"learning_rate": 2.764152339909756e-08,
|
| 504 |
+
"logits/chosen": -2.3025100231170654,
|
| 505 |
+
"logits/rejected": -2.220672845840454,
|
| 506 |
+
"logps/chosen": -548.1705322265625,
|
| 507 |
+
"logps/rejected": -444.21856689453125,
|
| 508 |
+
"loss": 0.0776,
|
| 509 |
+
"rewards/accuracies": 0.824999988079071,
|
| 510 |
+
"rewards/chosen": -1.1419382095336914,
|
| 511 |
+
"rewards/margins": 2.2051339149475098,
|
| 512 |
+
"rewards/rejected": -3.347072124481201,
|
| 513 |
"step": 350
|
| 514 |
},
|
| 515 |
{
|
| 516 |
+
"epoch": 0.89,
|
| 517 |
+
"learning_rate": 1.861941317991664e-08,
|
| 518 |
+
"logits/chosen": -2.3118605613708496,
|
| 519 |
+
"logits/rejected": -2.235429286956787,
|
| 520 |
+
"logps/chosen": -549.4410400390625,
|
| 521 |
+
"logps/rejected": -460.76947021484375,
|
| 522 |
+
"loss": 0.0783,
|
| 523 |
+
"rewards/accuracies": 0.800000011920929,
|
| 524 |
+
"rewards/chosen": -1.140019178390503,
|
| 525 |
+
"rewards/margins": 2.2883784770965576,
|
| 526 |
+
"rewards/rejected": -3.4283976554870605,
|
| 527 |
"step": 360
|
| 528 |
},
|
| 529 |
{
|
| 530 |
+
"epoch": 0.91,
|
| 531 |
+
"learning_rate": 1.13197833728636e-08,
|
| 532 |
+
"logits/chosen": -2.2796883583068848,
|
| 533 |
+
"logits/rejected": -2.2151246070861816,
|
| 534 |
+
"logps/chosen": -561.2109375,
|
| 535 |
+
"logps/rejected": -477.9205627441406,
|
| 536 |
+
"loss": 0.087,
|
| 537 |
+
"rewards/accuracies": 0.8187500238418579,
|
| 538 |
+
"rewards/chosen": -1.0488560199737549,
|
| 539 |
+
"rewards/margins": 2.396134853363037,
|
| 540 |
+
"rewards/rejected": -3.444990873336792,
|
| 541 |
"step": 370
|
| 542 |
},
|
| 543 |
{
|
| 544 |
+
"epoch": 0.94,
|
| 545 |
+
"learning_rate": 5.79697505093521e-09,
|
| 546 |
+
"logits/chosen": -2.2897768020629883,
|
| 547 |
+
"logits/rejected": -2.192831039428711,
|
| 548 |
+
"logps/chosen": -544.9513549804688,
|
| 549 |
+
"logps/rejected": -457.36456298828125,
|
| 550 |
+
"loss": 0.0966,
|
| 551 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 552 |
+
"rewards/chosen": -1.2722972631454468,
|
| 553 |
+
"rewards/margins": 2.0332696437835693,
|
| 554 |
+
"rewards/rejected": -3.3055667877197266,
|
| 555 |
"step": 380
|
| 556 |
},
|
| 557 |
{
|
| 558 |
+
"epoch": 0.96,
|
| 559 |
+
"learning_rate": 2.092101988131256e-09,
|
| 560 |
+
"logits/chosen": -2.308785915374756,
|
| 561 |
+
"logits/rejected": -2.224362850189209,
|
| 562 |
+
"logps/chosen": -571.474609375,
|
| 563 |
+
"logps/rejected": -455.10498046875,
|
| 564 |
+
"loss": 0.0847,
|
| 565 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 566 |
+
"rewards/chosen": -1.0817571878433228,
|
| 567 |
+
"rewards/margins": 2.259777545928955,
|
| 568 |
+
"rewards/rejected": -3.3415348529815674,
|
| 569 |
"step": 390
|
| 570 |
},
|
| 571 |
{
|
| 572 |
+
"epoch": 0.99,
|
| 573 |
+
"learning_rate": 2.327445937151673e-10,
|
| 574 |
+
"logits/chosen": -2.303267002105713,
|
| 575 |
+
"logits/rejected": -2.222599983215332,
|
| 576 |
+
"logps/chosen": -541.7862548828125,
|
| 577 |
+
"logps/rejected": -435.1748962402344,
|
| 578 |
+
"loss": 0.0925,
|
| 579 |
+
"rewards/accuracies": 0.75,
|
| 580 |
+
"rewards/chosen": -1.486748456954956,
|
| 581 |
+
"rewards/margins": 1.7116472721099854,
|
| 582 |
+
"rewards/rejected": -3.1983959674835205,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 583 |
"step": 400
|
| 584 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 585 |
{
|
| 586 |
"epoch": 1.0,
|
| 587 |
+
"step": 405,
|
| 588 |
"total_flos": 0.0,
|
| 589 |
+
"train_loss": 0.14271833528707056,
|
| 590 |
+
"train_runtime": 3205.3137,
|
| 591 |
+
"train_samples_per_second": 16.19,
|
| 592 |
+
"train_steps_per_second": 0.126
|
| 593 |
}
|
| 594 |
],
|
| 595 |
"logging_steps": 10,
|
| 596 |
+
"max_steps": 405,
|
| 597 |
"num_train_epochs": 1,
|
| 598 |
"save_steps": 1000,
|
| 599 |
"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 sha256:6820e20a1ed318ae7453ea4028656b79247178ba1388ae104c90cce83fcf4bd3
|
| 3 |
size 5944
|