Model save
Browse files- README.md +21 -24
- adapter_model.safetensors +1 -1
- all_results.json +3 -18
- train_results.json +3 -3
- trainer_state.json +670 -670
README.md
CHANGED
|
@@ -2,13 +2,10 @@
|
|
| 2 |
license: apache-2.0
|
| 3 |
library_name: peft
|
| 4 |
tags:
|
| 5 |
-
- alignment-handbook
|
| 6 |
- trl
|
| 7 |
- dpo
|
| 8 |
- generated_from_trainer
|
| 9 |
base_model: mistralai/Mistral-7B-v0.1
|
| 10 |
-
datasets:
|
| 11 |
-
- HuggingFaceH4/ultrafeedback_binarized
|
| 12 |
model-index:
|
| 13 |
- name: zephyr-7b
|
| 14 |
results: []
|
|
@@ -19,19 +16,19 @@ should probably proofread and complete it, then remove this comment. -->
|
|
| 19 |
|
| 20 |
# zephyr-7b
|
| 21 |
|
| 22 |
-
This model is a fine-tuned version of [
|
| 23 |
It achieves the following results on the evaluation set:
|
| 24 |
-
- Loss: 0.
|
| 25 |
-
- Rewards/chosen: -0.
|
| 26 |
-
- Rewards/rejected: -0.
|
| 27 |
-
- Rewards/accuracies: 0.
|
| 28 |
-
- Rewards/margins: 0.
|
| 29 |
-
- Logps/rejected: -
|
| 30 |
-
- Logps/chosen: -
|
| 31 |
-
- Logits/rejected: -
|
| 32 |
-
- Logits/chosen: -
|
| 33 |
-
- Use Label:
|
| 34 |
-
- Pred Label:
|
| 35 |
|
| 36 |
## Model description
|
| 37 |
|
|
@@ -68,15 +65,15 @@ The following hyperparameters were used during training:
|
|
| 68 |
|
| 69 |
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen | Use Label | Pred Label |
|
| 70 |
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|:----------:|:----------:|
|
| 71 |
-
| 0.
|
| 72 |
-
| 0.
|
| 73 |
-
| 0.
|
| 74 |
-
| 0.
|
| 75 |
-
| 0.
|
| 76 |
-
| 0.
|
| 77 |
-
| 0.
|
| 78 |
-
| 0.
|
| 79 |
-
| 0.
|
| 80 |
|
| 81 |
|
| 82 |
### Framework versions
|
|
|
|
| 2 |
license: apache-2.0
|
| 3 |
library_name: peft
|
| 4 |
tags:
|
|
|
|
| 5 |
- trl
|
| 6 |
- dpo
|
| 7 |
- generated_from_trainer
|
| 8 |
base_model: mistralai/Mistral-7B-v0.1
|
|
|
|
|
|
|
| 9 |
model-index:
|
| 10 |
- name: zephyr-7b
|
| 11 |
results: []
|
|
|
|
| 16 |
|
| 17 |
# zephyr-7b
|
| 18 |
|
| 19 |
+
This model is a fine-tuned version of [mistralai/Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) on the None dataset.
|
| 20 |
It achieves the following results on the evaluation set:
|
| 21 |
+
- Loss: 0.6928
|
| 22 |
+
- Rewards/chosen: -0.0288
|
| 23 |
+
- Rewards/rejected: -0.1012
|
| 24 |
+
- Rewards/accuracies: 0.3492
|
| 25 |
+
- Rewards/margins: 0.0723
|
| 26 |
+
- Logps/rejected: -85.5160
|
| 27 |
+
- Logps/chosen: -71.7842
|
| 28 |
+
- Logits/rejected: -2.1139
|
| 29 |
+
- Logits/chosen: -2.1428
|
| 30 |
+
- Use Label: 13461.3809
|
| 31 |
+
- Pred Label: 5226.6191
|
| 32 |
|
| 33 |
## Model description
|
| 34 |
|
|
|
|
| 65 |
|
| 66 |
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen | Use Label | Pred Label |
|
| 67 |
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|:----------:|:----------:|
|
| 68 |
+
| 0.6911 | 0.1 | 100 | 0.6919 | -0.0053 | -0.0356 | 0.3393 | 0.0303 | -78.9541 | -69.4262 | -2.0935 | -2.1210 | 1705.8572 | 150.1429 |
|
| 69 |
+
| 0.692 | 0.21 | 200 | 0.6927 | -0.0264 | -0.0695 | 0.3433 | 0.0431 | -82.3504 | -71.5409 | -2.1057 | -2.1268 | 3337.0476 | 622.9524 |
|
| 70 |
+
| 0.6924 | 0.31 | 300 | 0.6929 | -0.0369 | -0.0896 | 0.3393 | 0.0527 | -84.3537 | -72.5877 | -2.1933 | -2.2169 | 4863.7300 | 1200.2699 |
|
| 71 |
+
| 0.6927 | 0.42 | 400 | 0.6925 | -0.0211 | -0.0804 | 0.3413 | 0.0593 | -83.4364 | -71.0104 | -2.0934 | -2.1190 | 6324.0796 | 1843.9207 |
|
| 72 |
+
| 0.6924 | 0.52 | 500 | 0.6929 | -0.0206 | -0.0831 | 0.3433 | 0.0625 | -83.7112 | -70.9618 | -2.1518 | -2.1762 | 7772.7778 | 2499.2222 |
|
| 73 |
+
| 0.6929 | 0.63 | 600 | 0.6927 | -0.0452 | -0.1160 | 0.3512 | 0.0708 | -86.9945 | -73.4171 | -2.1125 | -2.1408 | 9198.8574 | 3177.1428 |
|
| 74 |
+
| 0.6928 | 0.73 | 700 | 0.6930 | -0.0507 | -0.1231 | 0.3512 | 0.0724 | -87.7077 | -73.9657 | -2.1086 | -2.1372 | 10627.2695 | 3852.7302 |
|
| 75 |
+
| 0.6927 | 0.84 | 800 | 0.6928 | -0.0272 | -0.0999 | 0.3552 | 0.0726 | -85.3832 | -71.6247 | -2.1141 | -2.1431 | 12045.5234 | 4538.4761 |
|
| 76 |
+
| 0.6929 | 0.94 | 900 | 0.6928 | -0.0288 | -0.1012 | 0.3492 | 0.0723 | -85.5160 | -71.7842 | -2.1139 | -2.1428 | 13461.3809 | 5226.6191 |
|
| 77 |
|
| 78 |
|
| 79 |
### Framework versions
|
adapter_model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 671150064
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7756f5d0cc022294ccf38c261284a9de8e425fd482035784aa767bab75061bc0
|
| 3 |
size 671150064
|
all_results.json
CHANGED
|
@@ -1,23 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
-
"
|
| 4 |
-
"
|
| 5 |
-
"eval_logps/chosen": -77.5232162475586,
|
| 6 |
-
"eval_logps/rejected": -95.19373321533203,
|
| 7 |
-
"eval_loss": 0.6917868852615356,
|
| 8 |
-
"eval_pred_label": 4738.58740234375,
|
| 9 |
-
"eval_rewards/accuracies": 0.3591269850730896,
|
| 10 |
-
"eval_rewards/chosen": -0.0862266793847084,
|
| 11 |
-
"eval_rewards/margins": 0.11172995716333389,
|
| 12 |
-
"eval_rewards/rejected": -0.19795666635036469,
|
| 13 |
-
"eval_runtime": 247.3331,
|
| 14 |
-
"eval_samples": 2000,
|
| 15 |
-
"eval_samples_per_second": 8.086,
|
| 16 |
-
"eval_steps_per_second": 0.255,
|
| 17 |
-
"eval_use_label": 15333.4130859375,
|
| 18 |
-
"train_loss": 0.6906769273168754,
|
| 19 |
-
"train_runtime": 20027.4031,
|
| 20 |
"train_samples": 61135,
|
| 21 |
-
"train_samples_per_second": 3.
|
| 22 |
"train_steps_per_second": 0.048
|
| 23 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
+
"train_loss": 0.692275420283772,
|
| 4 |
+
"train_runtime": 20019.5915,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 5 |
"train_samples": 61135,
|
| 6 |
+
"train_samples_per_second": 3.054,
|
| 7 |
"train_steps_per_second": 0.048
|
| 8 |
}
|
train_results.json
CHANGED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
-
"train_loss": 0.
|
| 4 |
-
"train_runtime":
|
| 5 |
"train_samples": 61135,
|
| 6 |
-
"train_samples_per_second": 3.
|
| 7 |
"train_steps_per_second": 0.048
|
| 8 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
+
"train_loss": 0.692275420283772,
|
| 4 |
+
"train_runtime": 20019.5915,
|
| 5 |
"train_samples": 61135,
|
| 6 |
+
"train_samples_per_second": 3.054,
|
| 7 |
"train_steps_per_second": 0.048
|
| 8 |
}
|
trainer_state.json
CHANGED
|
@@ -29,970 +29,970 @@
|
|
| 29 |
"epoch": 0.02,
|
| 30 |
"grad_norm": 0.6796875,
|
| 31 |
"learning_rate": 1.0416666666666667e-06,
|
| 32 |
-
"logits/chosen": -2.
|
| 33 |
-
"logits/rejected": -2.
|
| 34 |
-
"logps/chosen": -57.
|
| 35 |
-
"logps/rejected": -66.
|
| 36 |
"loss": 0.6927,
|
| 37 |
"pred_label": 0.0,
|
| 38 |
-
"rewards/accuracies": 0.
|
| 39 |
-
"rewards/chosen": 0.
|
| 40 |
-
"rewards/margins": 0.
|
| 41 |
-
"rewards/rejected": 0.
|
| 42 |
"step": 20,
|
| 43 |
"use_label": 170.0
|
| 44 |
},
|
| 45 |
{
|
| 46 |
"epoch": 0.04,
|
| 47 |
-
"grad_norm": 0.
|
| 48 |
"learning_rate": 2.0833333333333334e-06,
|
| 49 |
-
"logits/chosen": -2.
|
| 50 |
-
"logits/rejected": -2.
|
| 51 |
-
"logps/chosen": -54.
|
| 52 |
-
"logps/rejected": -67.
|
| 53 |
-
"loss": 0.
|
| 54 |
"pred_label": 0.0,
|
| 55 |
"rewards/accuracies": 0.24687500298023224,
|
| 56 |
-
"rewards/chosen": 0.
|
| 57 |
-
"rewards/margins": 0.
|
| 58 |
-
"rewards/rejected": 0.
|
| 59 |
"step": 40,
|
| 60 |
"use_label": 482.0
|
| 61 |
},
|
| 62 |
{
|
| 63 |
"epoch": 0.06,
|
| 64 |
-
"grad_norm": 0.
|
| 65 |
"learning_rate": 3.125e-06,
|
| 66 |
-
"logits/chosen": -2.
|
| 67 |
-
"logits/rejected": -2.
|
| 68 |
-
"logps/chosen": -75.
|
| 69 |
-
"logps/rejected": -87.
|
| 70 |
-
"loss": 0.
|
| 71 |
-
"pred_label":
|
| 72 |
-
"rewards/accuracies": 0.
|
| 73 |
-
"rewards/chosen": 0.
|
| 74 |
-
"rewards/margins": 0.
|
| 75 |
-
"rewards/rejected": 0.
|
| 76 |
"step": 60,
|
| 77 |
-
"use_label":
|
| 78 |
},
|
| 79 |
{
|
| 80 |
"epoch": 0.08,
|
| 81 |
-
"grad_norm": 0.
|
| 82 |
"learning_rate": 4.166666666666667e-06,
|
| 83 |
-
"logits/chosen": -2.
|
| 84 |
-
"logits/rejected": -2.
|
| 85 |
-
"logps/chosen": -
|
| 86 |
-
"logps/rejected": -
|
| 87 |
-
"loss": 0.
|
| 88 |
-
"pred_label":
|
| 89 |
"rewards/accuracies": 0.3343749940395355,
|
| 90 |
-
"rewards/chosen": 0.
|
| 91 |
-
"rewards/margins": 0.
|
| 92 |
-
"rewards/rejected":
|
| 93 |
"step": 80,
|
| 94 |
-
"use_label":
|
| 95 |
},
|
| 96 |
{
|
| 97 |
"epoch": 0.1,
|
| 98 |
-
"grad_norm": 0.
|
| 99 |
"learning_rate": 4.9997324926814375e-06,
|
| 100 |
-
"logits/chosen": -2.
|
| 101 |
-
"logits/rejected": -2.
|
| 102 |
-
"logps/chosen": -
|
| 103 |
-
"logps/rejected": -
|
| 104 |
-
"loss": 0.
|
| 105 |
-
"pred_label":
|
| 106 |
-
"rewards/accuracies": 0.
|
| 107 |
-
"rewards/chosen":
|
| 108 |
-
"rewards/margins": 0.
|
| 109 |
-
"rewards/rejected": -0.
|
| 110 |
"step": 100,
|
| 111 |
-
"use_label":
|
| 112 |
},
|
| 113 |
{
|
| 114 |
"epoch": 0.1,
|
| 115 |
-
"eval_logits/chosen": -2.
|
| 116 |
-
"eval_logits/rejected": -2.
|
| 117 |
-
"eval_logps/chosen": -
|
| 118 |
-
"eval_logps/rejected": -
|
| 119 |
-
"eval_loss": 0.
|
| 120 |
-
"eval_pred_label":
|
| 121 |
-
"eval_rewards/accuracies": 0.
|
| 122 |
-
"eval_rewards/chosen": -0.
|
| 123 |
-
"eval_rewards/margins": 0.
|
| 124 |
-
"eval_rewards/rejected": -0.
|
| 125 |
-
"eval_runtime": 247.
|
| 126 |
-
"eval_samples_per_second": 8.
|
| 127 |
"eval_steps_per_second": 0.254,
|
| 128 |
-
"eval_use_label":
|
| 129 |
"step": 100
|
| 130 |
},
|
| 131 |
{
|
| 132 |
"epoch": 0.13,
|
| 133 |
-
"grad_norm": 0.
|
| 134 |
"learning_rate": 4.9903757462135984e-06,
|
| 135 |
-
"logits/chosen": -2.
|
| 136 |
-
"logits/rejected": -2.
|
| 137 |
-
"logps/chosen": -
|
| 138 |
-
"logps/rejected": -
|
| 139 |
-
"loss": 0.
|
| 140 |
-
"pred_label":
|
| 141 |
-
"rewards/accuracies": 0.
|
| 142 |
-
"rewards/chosen":
|
| 143 |
-
"rewards/margins": 0.
|
| 144 |
-
"rewards/rejected": -0.
|
| 145 |
"step": 120,
|
| 146 |
-
"use_label":
|
| 147 |
},
|
| 148 |
{
|
| 149 |
"epoch": 0.15,
|
| 150 |
-
"grad_norm": 0.
|
| 151 |
"learning_rate": 4.967700826904229e-06,
|
| 152 |
-
"logits/chosen": -2.
|
| 153 |
-
"logits/rejected": -2.
|
| 154 |
-
"logps/chosen": -
|
| 155 |
-
"logps/rejected": -
|
| 156 |
-
"loss": 0.
|
| 157 |
-
"pred_label":
|
| 158 |
-
"rewards/accuracies": 0.
|
| 159 |
-
"rewards/chosen":
|
| 160 |
-
"rewards/margins": 0.
|
| 161 |
-
"rewards/rejected": -0.
|
| 162 |
"step": 140,
|
| 163 |
-
"use_label":
|
| 164 |
},
|
| 165 |
{
|
| 166 |
"epoch": 0.17,
|
| 167 |
-
"grad_norm": 0.
|
| 168 |
"learning_rate": 4.931828996974498e-06,
|
| 169 |
-
"logits/chosen": -2.
|
| 170 |
-
"logits/rejected": -2.
|
| 171 |
-
"logps/chosen": -66.
|
| 172 |
-
"logps/rejected": -
|
| 173 |
-
"loss": 0.
|
| 174 |
-
"pred_label":
|
| 175 |
-
"rewards/accuracies": 0.
|
| 176 |
-
"rewards/chosen": -0.
|
| 177 |
-
"rewards/margins": 0.
|
| 178 |
-
"rewards/rejected": -0.
|
| 179 |
"step": 160,
|
| 180 |
-
"use_label":
|
| 181 |
},
|
| 182 |
{
|
| 183 |
"epoch": 0.19,
|
| 184 |
-
"grad_norm": 0.
|
| 185 |
"learning_rate": 4.882952093833628e-06,
|
| 186 |
-
"logits/chosen": -2.
|
| 187 |
-
"logits/rejected": -2.
|
| 188 |
-
"logps/chosen": -
|
| 189 |
-
"logps/rejected": -
|
| 190 |
-
"loss": 0.
|
| 191 |
-
"pred_label":
|
| 192 |
-
"rewards/accuracies": 0.
|
| 193 |
-
"rewards/chosen": -0.
|
| 194 |
-
"rewards/margins": 0.
|
| 195 |
-
"rewards/rejected": -0.
|
| 196 |
"step": 180,
|
| 197 |
-
"use_label":
|
| 198 |
},
|
| 199 |
{
|
| 200 |
"epoch": 0.21,
|
| 201 |
-
"grad_norm": 0.
|
| 202 |
"learning_rate": 4.821331504159906e-06,
|
| 203 |
-
"logits/chosen": -2.
|
| 204 |
-
"logits/rejected": -2.
|
| 205 |
-
"logps/chosen": -
|
| 206 |
-
"logps/rejected": -
|
| 207 |
-
"loss": 0.
|
| 208 |
-
"pred_label":
|
| 209 |
-
"rewards/accuracies": 0.
|
| 210 |
-
"rewards/chosen": -0.
|
| 211 |
-
"rewards/margins": 0.
|
| 212 |
-
"rewards/rejected": -0.
|
| 213 |
"step": 200,
|
| 214 |
-
"use_label":
|
| 215 |
},
|
| 216 |
{
|
| 217 |
"epoch": 0.21,
|
| 218 |
-
"eval_logits/chosen": -2.
|
| 219 |
-
"eval_logits/rejected": -2.
|
| 220 |
-
"eval_logps/chosen": -
|
| 221 |
-
"eval_logps/rejected": -
|
| 222 |
-
"eval_loss": 0.
|
| 223 |
-
"eval_pred_label":
|
| 224 |
-
"eval_rewards/accuracies": 0.
|
| 225 |
-
"eval_rewards/chosen": -0.
|
| 226 |
-
"eval_rewards/margins": 0.
|
| 227 |
-
"eval_rewards/rejected": -0.
|
| 228 |
-
"eval_runtime":
|
| 229 |
-
"eval_samples_per_second": 8.
|
| 230 |
"eval_steps_per_second": 0.254,
|
| 231 |
-
"eval_use_label":
|
| 232 |
"step": 200
|
| 233 |
},
|
| 234 |
{
|
| 235 |
"epoch": 0.23,
|
| 236 |
-
"grad_norm": 0.
|
| 237 |
"learning_rate": 4.747296766042161e-06,
|
| 238 |
-
"logits/chosen": -2.
|
| 239 |
-
"logits/rejected": -2.
|
| 240 |
-
"logps/chosen": -
|
| 241 |
-
"logps/rejected": -
|
| 242 |
-
"loss": 0.
|
| 243 |
-
"pred_label":
|
| 244 |
-
"rewards/accuracies": 0.
|
| 245 |
-
"rewards/chosen": -0.
|
| 246 |
-
"rewards/margins": 0.
|
| 247 |
-
"rewards/rejected": -0.
|
| 248 |
"step": 220,
|
| 249 |
-
"use_label":
|
| 250 |
},
|
| 251 |
{
|
| 252 |
"epoch": 0.25,
|
| 253 |
-
"grad_norm": 0.
|
| 254 |
"learning_rate": 4.661243806657256e-06,
|
| 255 |
-
"logits/chosen": -2.
|
| 256 |
-
"logits/rejected": -2.
|
| 257 |
-
"logps/chosen": -
|
| 258 |
-
"logps/rejected": -
|
| 259 |
-
"loss": 0.
|
| 260 |
-
"pred_label":
|
| 261 |
-
"rewards/accuracies": 0.
|
| 262 |
-
"rewards/chosen": -0.
|
| 263 |
-
"rewards/margins": 0.
|
| 264 |
-
"rewards/rejected": -0.
|
| 265 |
"step": 240,
|
| 266 |
-
"use_label":
|
| 267 |
},
|
| 268 |
{
|
| 269 |
"epoch": 0.27,
|
| 270 |
-
"grad_norm": 0.
|
| 271 |
"learning_rate": 4.563632824908252e-06,
|
| 272 |
-
"logits/chosen": -2.
|
| 273 |
-
"logits/rejected": -2.
|
| 274 |
-
"logps/chosen": -
|
| 275 |
-
"logps/rejected": -
|
| 276 |
-
"loss": 0.
|
| 277 |
-
"pred_label":
|
| 278 |
-
"rewards/accuracies": 0.
|
| 279 |
-
"rewards/chosen": -0.
|
| 280 |
-
"rewards/margins": 0.
|
| 281 |
-
"rewards/rejected": -0.
|
| 282 |
"step": 260,
|
| 283 |
-
"use_label":
|
| 284 |
},
|
| 285 |
{
|
| 286 |
"epoch": 0.29,
|
| 287 |
-
"grad_norm": 0.
|
| 288 |
"learning_rate": 4.454985830346574e-06,
|
| 289 |
-
"logits/chosen": -2.
|
| 290 |
-
"logits/rejected": -2.
|
| 291 |
-
"logps/chosen": -
|
| 292 |
-
"logps/rejected": -
|
| 293 |
-
"loss": 0.
|
| 294 |
-
"pred_label":
|
| 295 |
-
"rewards/accuracies": 0.
|
| 296 |
-
"rewards/chosen": -0.
|
| 297 |
-
"rewards/margins": 0.
|
| 298 |
-
"rewards/rejected": -0.
|
| 299 |
"step": 280,
|
| 300 |
-
"use_label":
|
| 301 |
},
|
| 302 |
{
|
| 303 |
"epoch": 0.31,
|
| 304 |
-
"grad_norm": 0.
|
| 305 |
"learning_rate": 4.335883851539693e-06,
|
| 306 |
-
"logits/chosen": -2.
|
| 307 |
-
"logits/rejected": -2.
|
| 308 |
-
"logps/chosen": -
|
| 309 |
-
"logps/rejected": -
|
| 310 |
-
"loss": 0.
|
| 311 |
-
"pred_label":
|
| 312 |
-
"rewards/accuracies": 0.
|
| 313 |
-
"rewards/chosen": -0.
|
| 314 |
-
"rewards/margins": 0.
|
| 315 |
-
"rewards/rejected": -0.
|
| 316 |
"step": 300,
|
| 317 |
-
"use_label":
|
| 318 |
},
|
| 319 |
{
|
| 320 |
"epoch": 0.31,
|
| 321 |
-
"eval_logits/chosen": -2.
|
| 322 |
-
"eval_logits/rejected": -
|
| 323 |
-
"eval_logps/chosen": -
|
| 324 |
-
"eval_logps/rejected": -
|
| 325 |
-
"eval_loss": 0.
|
| 326 |
-
"eval_pred_label":
|
| 327 |
-
"eval_rewards/accuracies": 0.
|
| 328 |
-
"eval_rewards/chosen": -0.
|
| 329 |
-
"eval_rewards/margins": 0.
|
| 330 |
-
"eval_rewards/rejected": -0.
|
| 331 |
-
"eval_runtime": 247.
|
| 332 |
-
"eval_samples_per_second": 8.
|
| 333 |
"eval_steps_per_second": 0.254,
|
| 334 |
-
"eval_use_label":
|
| 335 |
"step": 300
|
| 336 |
},
|
| 337 |
{
|
| 338 |
"epoch": 0.33,
|
| 339 |
-
"grad_norm": 0.
|
| 340 |
"learning_rate": 4.206963828813555e-06,
|
| 341 |
-
"logits/chosen": -2.
|
| 342 |
-
"logits/rejected": -2.
|
| 343 |
-
"logps/chosen": -
|
| 344 |
-
"logps/rejected": -
|
| 345 |
-
"loss": 0.
|
| 346 |
-
"pred_label":
|
| 347 |
-
"rewards/accuracies": 0.
|
| 348 |
-
"rewards/chosen": -0.
|
| 349 |
-
"rewards/margins": 0.
|
| 350 |
-
"rewards/rejected": -0.
|
| 351 |
"step": 320,
|
| 352 |
-
"use_label":
|
| 353 |
},
|
| 354 |
{
|
| 355 |
"epoch": 0.36,
|
| 356 |
-
"grad_norm": 0.
|
| 357 |
"learning_rate": 4.068915207986931e-06,
|
| 358 |
-
"logits/chosen": -2.
|
| 359 |
-
"logits/rejected": -
|
| 360 |
-
"logps/chosen": -
|
| 361 |
-
"logps/rejected": -
|
| 362 |
-
"loss": 0.
|
| 363 |
-
"pred_label":
|
| 364 |
-
"rewards/accuracies": 0.
|
| 365 |
-
"rewards/chosen": -0.
|
| 366 |
-
"rewards/margins": 0.
|
| 367 |
-
"rewards/rejected": -0.
|
| 368 |
"step": 340,
|
| 369 |
-
"use_label":
|
| 370 |
},
|
| 371 |
{
|
| 372 |
"epoch": 0.38,
|
| 373 |
-
"grad_norm": 0.
|
| 374 |
"learning_rate": 3.922476253313921e-06,
|
| 375 |
-
"logits/chosen": -2.
|
| 376 |
-
"logits/rejected": -2.
|
| 377 |
-
"logps/chosen": -
|
| 378 |
-
"logps/rejected": -
|
| 379 |
-
"loss": 0.
|
| 380 |
-
"pred_label":
|
| 381 |
-
"rewards/accuracies": 0.
|
| 382 |
-
"rewards/chosen": -0.
|
| 383 |
-
"rewards/margins": 0.
|
| 384 |
-
"rewards/rejected": -0.
|
| 385 |
"step": 360,
|
| 386 |
-
"use_label":
|
| 387 |
},
|
| 388 |
{
|
| 389 |
"epoch": 0.4,
|
| 390 |
-
"grad_norm": 0.
|
| 391 |
"learning_rate": 3.768430099352445e-06,
|
| 392 |
-
"logits/chosen": -2.
|
| 393 |
-
"logits/rejected": -2.
|
| 394 |
-
"logps/chosen": -
|
| 395 |
-
"logps/rejected": -
|
| 396 |
-
"loss": 0.
|
| 397 |
-
"pred_label":
|
| 398 |
-
"rewards/accuracies": 0.
|
| 399 |
-
"rewards/chosen": -0.
|
| 400 |
-
"rewards/margins": 0.
|
| 401 |
-
"rewards/rejected": -0.
|
| 402 |
"step": 380,
|
| 403 |
-
"use_label":
|
| 404 |
},
|
| 405 |
{
|
| 406 |
"epoch": 0.42,
|
| 407 |
-
"grad_norm": 0.
|
| 408 |
"learning_rate": 3.607600562872785e-06,
|
| 409 |
-
"logits/chosen": -2.
|
| 410 |
-
"logits/rejected": -2.
|
| 411 |
-
"logps/chosen": -
|
| 412 |
-
"logps/rejected": -
|
| 413 |
-
"loss": 0.
|
| 414 |
-
"pred_label":
|
| 415 |
-
"rewards/accuracies": 0.
|
| 416 |
-
"rewards/chosen": -0.
|
| 417 |
-
"rewards/margins": 0.
|
| 418 |
-
"rewards/rejected": -0.
|
| 419 |
"step": 400,
|
| 420 |
-
"use_label":
|
| 421 |
},
|
| 422 |
{
|
| 423 |
"epoch": 0.42,
|
| 424 |
-
"eval_logits/chosen": -2.
|
| 425 |
-
"eval_logits/rejected": -2.
|
| 426 |
-
"eval_logps/chosen": -
|
| 427 |
-
"eval_logps/rejected": -
|
| 428 |
-
"eval_loss": 0.
|
| 429 |
-
"eval_pred_label":
|
| 430 |
-
"eval_rewards/accuracies": 0.
|
| 431 |
-
"eval_rewards/chosen": -0.
|
| 432 |
-
"eval_rewards/margins": 0.
|
| 433 |
-
"eval_rewards/rejected": -0.
|
| 434 |
-
"eval_runtime": 248.
|
| 435 |
"eval_samples_per_second": 8.064,
|
| 436 |
"eval_steps_per_second": 0.254,
|
| 437 |
-
"eval_use_label":
|
| 438 |
"step": 400
|
| 439 |
},
|
| 440 |
{
|
| 441 |
"epoch": 0.44,
|
| 442 |
-
"grad_norm": 0.
|
| 443 |
"learning_rate": 3.4408477372034743e-06,
|
| 444 |
-
"logits/chosen": -2.
|
| 445 |
-
"logits/rejected": -2.
|
| 446 |
-
"logps/chosen": -
|
| 447 |
-
"logps/rejected": -
|
| 448 |
-
"loss": 0.
|
| 449 |
-
"pred_label":
|
| 450 |
-
"rewards/accuracies": 0.
|
| 451 |
-
"rewards/chosen": -0.
|
| 452 |
-
"rewards/margins": 0.
|
| 453 |
-
"rewards/rejected": -0.
|
| 454 |
"step": 420,
|
| 455 |
-
"use_label":
|
| 456 |
},
|
| 457 |
{
|
| 458 |
"epoch": 0.46,
|
| 459 |
-
"grad_norm": 0.
|
| 460 |
"learning_rate": 3.269063392575352e-06,
|
| 461 |
-
"logits/chosen": -2.
|
| 462 |
-
"logits/rejected": -2.
|
| 463 |
-
"logps/chosen": -
|
| 464 |
-
"logps/rejected": -
|
| 465 |
-
"loss": 0.
|
| 466 |
-
"pred_label":
|
| 467 |
-
"rewards/accuracies": 0.
|
| 468 |
-
"rewards/chosen": -0.
|
| 469 |
-
"rewards/margins": 0.
|
| 470 |
-
"rewards/rejected": -0.
|
| 471 |
"step": 440,
|
| 472 |
-
"use_label":
|
| 473 |
},
|
| 474 |
{
|
| 475 |
"epoch": 0.48,
|
| 476 |
-
"grad_norm": 0.
|
| 477 |
"learning_rate": 3.09316620706208e-06,
|
| 478 |
-
"logits/chosen": -2.
|
| 479 |
-
"logits/rejected": -2.
|
| 480 |
-
"logps/chosen": -
|
| 481 |
-
"logps/rejected": -
|
| 482 |
-
"loss": 0.
|
| 483 |
-
"pred_label":
|
| 484 |
-
"rewards/accuracies": 0.
|
| 485 |
-
"rewards/chosen": -0.
|
| 486 |
-
"rewards/margins": 0.
|
| 487 |
-
"rewards/rejected": -0.
|
| 488 |
"step": 460,
|
| 489 |
-
"use_label":
|
| 490 |
},
|
| 491 |
{
|
| 492 |
"epoch": 0.5,
|
| 493 |
-
"grad_norm": 0.
|
| 494 |
"learning_rate": 2.91409685362137e-06,
|
| 495 |
-
"logits/chosen": -2.
|
| 496 |
-
"logits/rejected": -2.
|
| 497 |
-
"logps/chosen": -
|
| 498 |
-
"logps/rejected": -
|
| 499 |
-
"loss": 0.
|
| 500 |
-
"pred_label":
|
| 501 |
-
"rewards/accuracies": 0.
|
| 502 |
-
"rewards/chosen": -0.
|
| 503 |
-
"rewards/margins": 0.
|
| 504 |
-
"rewards/rejected": -0.
|
| 505 |
"step": 480,
|
| 506 |
-
"use_label":
|
| 507 |
},
|
| 508 |
{
|
| 509 |
"epoch": 0.52,
|
| 510 |
-
"grad_norm": 0.
|
| 511 |
"learning_rate": 2.7328129695107205e-06,
|
| 512 |
-
"logits/chosen": -2.
|
| 513 |
-
"logits/rejected": -2.
|
| 514 |
-
"logps/chosen": -
|
| 515 |
-
"logps/rejected": -
|
| 516 |
-
"loss": 0.
|
| 517 |
-
"pred_label":
|
| 518 |
-
"rewards/accuracies": 0.
|
| 519 |
-
"rewards/chosen": -0.
|
| 520 |
-
"rewards/margins": 0.
|
| 521 |
-
"rewards/rejected": -0.
|
| 522 |
"step": 500,
|
| 523 |
-
"use_label":
|
| 524 |
},
|
| 525 |
{
|
| 526 |
"epoch": 0.52,
|
| 527 |
-
"eval_logits/chosen": -2.
|
| 528 |
-
"eval_logits/rejected": -
|
| 529 |
-
"eval_logps/chosen": -
|
| 530 |
-
"eval_logps/rejected": -
|
| 531 |
-
"eval_loss": 0.
|
| 532 |
-
"eval_pred_label":
|
| 533 |
-
"eval_rewards/accuracies": 0.
|
| 534 |
-
"eval_rewards/chosen": -0.
|
| 535 |
-
"eval_rewards/margins": 0.
|
| 536 |
-
"eval_rewards/rejected": -0.
|
| 537 |
-
"eval_runtime":
|
| 538 |
-
"eval_samples_per_second": 8.
|
| 539 |
"eval_steps_per_second": 0.254,
|
| 540 |
-
"eval_use_label":
|
| 541 |
"step": 500
|
| 542 |
},
|
| 543 |
{
|
| 544 |
"epoch": 0.54,
|
| 545 |
-
"grad_norm": 0.
|
| 546 |
"learning_rate": 2.5502840349805074e-06,
|
| 547 |
-
"logits/chosen": -2.
|
| 548 |
-
"logits/rejected": -2.
|
| 549 |
-
"logps/chosen": -
|
| 550 |
-
"logps/rejected": -
|
| 551 |
-
"loss": 0.
|
| 552 |
-
"pred_label":
|
| 553 |
-
"rewards/accuracies": 0.
|
| 554 |
-
"rewards/chosen": -0.
|
| 555 |
-
"rewards/margins": 0.
|
| 556 |
-
"rewards/rejected": -0.
|
| 557 |
"step": 520,
|
| 558 |
-
"use_label":
|
| 559 |
},
|
| 560 |
{
|
| 561 |
"epoch": 0.57,
|
| 562 |
-
"grad_norm":
|
| 563 |
"learning_rate": 2.367486188632446e-06,
|
| 564 |
-
"logits/chosen": -2.
|
| 565 |
-
"logits/rejected": -2.
|
| 566 |
-
"logps/chosen": -
|
| 567 |
-
"logps/rejected": -
|
| 568 |
-
"loss": 0.
|
| 569 |
-
"pred_label":
|
| 570 |
-
"rewards/accuracies": 0.
|
| 571 |
-
"rewards/chosen": -0.
|
| 572 |
-
"rewards/margins": 0.
|
| 573 |
-
"rewards/rejected": -0.
|
| 574 |
"step": 540,
|
| 575 |
-
"use_label":
|
| 576 |
},
|
| 577 |
{
|
| 578 |
"epoch": 0.59,
|
| 579 |
-
"grad_norm": 0.
|
| 580 |
"learning_rate": 2.1853970071701415e-06,
|
| 581 |
-
"logits/chosen": -2.
|
| 582 |
-
"logits/rejected": -2.
|
| 583 |
-
"logps/chosen": -
|
| 584 |
-
"logps/rejected": -
|
| 585 |
-
"loss": 0.
|
| 586 |
-
"pred_label":
|
| 587 |
"rewards/accuracies": 0.2874999940395355,
|
| 588 |
-
"rewards/chosen": -0.
|
| 589 |
-
"rewards/margins": 0.
|
| 590 |
-
"rewards/rejected": -0.
|
| 591 |
"step": 560,
|
| 592 |
-
"use_label":
|
| 593 |
},
|
| 594 |
{
|
| 595 |
"epoch": 0.61,
|
| 596 |
-
"grad_norm": 0.
|
| 597 |
"learning_rate": 2.00499027745888e-06,
|
| 598 |
-
"logits/chosen": -2.
|
| 599 |
-
"logits/rejected": -2.
|
| 600 |
-
"logps/chosen": -
|
| 601 |
-
"logps/rejected": -
|
| 602 |
-
"loss": 0.
|
| 603 |
-
"pred_label":
|
| 604 |
-
"rewards/accuracies": 0.
|
| 605 |
-
"rewards/chosen": -0.
|
| 606 |
-
"rewards/margins": 0.
|
| 607 |
-
"rewards/rejected": -0.
|
| 608 |
"step": 580,
|
| 609 |
-
"use_label":
|
| 610 |
},
|
| 611 |
{
|
| 612 |
"epoch": 0.63,
|
| 613 |
-
"grad_norm": 0.
|
| 614 |
"learning_rate": 1.8272307888529276e-06,
|
| 615 |
-
"logits/chosen": -2.
|
| 616 |
-
"logits/rejected": -2.
|
| 617 |
-
"logps/chosen": -
|
| 618 |
-
"logps/rejected": -
|
| 619 |
-
"loss": 0.
|
| 620 |
-
"pred_label":
|
| 621 |
-
"rewards/accuracies": 0.
|
| 622 |
-
"rewards/chosen": -0.
|
| 623 |
-
"rewards/margins": 0.
|
| 624 |
-
"rewards/rejected": -0.
|
| 625 |
"step": 600,
|
| 626 |
-
"use_label":
|
| 627 |
},
|
| 628 |
{
|
| 629 |
"epoch": 0.63,
|
| 630 |
-
"eval_logits/chosen": -
|
| 631 |
-
"eval_logits/rejected": -
|
| 632 |
-
"eval_logps/chosen": -
|
| 633 |
-
"eval_logps/rejected": -
|
| 634 |
-
"eval_loss": 0.
|
| 635 |
-
"eval_pred_label":
|
| 636 |
-
"eval_rewards/accuracies": 0.
|
| 637 |
-
"eval_rewards/chosen": -0.
|
| 638 |
-
"eval_rewards/margins": 0.
|
| 639 |
-
"eval_rewards/rejected": -0.
|
| 640 |
-
"eval_runtime":
|
| 641 |
-
"eval_samples_per_second": 8.
|
| 642 |
"eval_steps_per_second": 0.254,
|
| 643 |
-
"eval_use_label":
|
| 644 |
"step": 600
|
| 645 |
},
|
| 646 |
{
|
| 647 |
"epoch": 0.65,
|
| 648 |
-
"grad_norm": 0.
|
| 649 |
"learning_rate": 1.6530691736402317e-06,
|
| 650 |
-
"logits/chosen": -
|
| 651 |
-
"logits/rejected": -2.
|
| 652 |
-
"logps/chosen": -
|
| 653 |
-
"logps/rejected": -
|
| 654 |
-
"loss": 0.
|
| 655 |
-
"pred_label":
|
| 656 |
-
"rewards/accuracies": 0.
|
| 657 |
-
"rewards/chosen": -0.
|
| 658 |
-
"rewards/margins": 0.
|
| 659 |
-
"rewards/rejected": -0.
|
| 660 |
"step": 620,
|
| 661 |
-
"use_label":
|
| 662 |
},
|
| 663 |
{
|
| 664 |
"epoch": 0.67,
|
| 665 |
-
"grad_norm": 0.
|
| 666 |
"learning_rate": 1.4834368231970922e-06,
|
| 667 |
-
"logits/chosen": -2.
|
| 668 |
-
"logits/rejected": -2.
|
| 669 |
-
"logps/chosen": -
|
| 670 |
-
"logps/rejected": -
|
| 671 |
-
"loss": 0.
|
| 672 |
-
"pred_label":
|
| 673 |
-
"rewards/accuracies": 0.
|
| 674 |
-
"rewards/chosen": -0.
|
| 675 |
-
"rewards/margins": 0.
|
| 676 |
-
"rewards/rejected": -0.
|
| 677 |
"step": 640,
|
| 678 |
-
"use_label":
|
| 679 |
},
|
| 680 |
{
|
| 681 |
"epoch": 0.69,
|
| 682 |
-
"grad_norm": 0.
|
| 683 |
"learning_rate": 1.3192409070404582e-06,
|
| 684 |
-
"logits/chosen": -2.
|
| 685 |
-
"logits/rejected": -2.
|
| 686 |
-
"logps/chosen": -
|
| 687 |
-
"logps/rejected": -
|
| 688 |
-
"loss": 0.
|
| 689 |
-
"pred_label":
|
| 690 |
-
"rewards/accuracies": 0.
|
| 691 |
-
"rewards/chosen": -0.
|
| 692 |
-
"rewards/margins": 0.
|
| 693 |
-
"rewards/rejected": -0.
|
| 694 |
"step": 660,
|
| 695 |
-
"use_label":
|
| 696 |
},
|
| 697 |
{
|
| 698 |
"epoch": 0.71,
|
| 699 |
-
"grad_norm": 0.
|
| 700 |
"learning_rate": 1.1613595214152713e-06,
|
| 701 |
-
"logits/chosen": -2.
|
| 702 |
-
"logits/rejected": -2.
|
| 703 |
-
"logps/chosen": -
|
| 704 |
-
"logps/rejected": -
|
| 705 |
-
"loss": 0.
|
| 706 |
-
"pred_label":
|
| 707 |
-
"rewards/accuracies": 0.
|
| 708 |
-
"rewards/chosen": -0.
|
| 709 |
-
"rewards/margins": 0.
|
| 710 |
-
"rewards/rejected": -0.
|
| 711 |
"step": 680,
|
| 712 |
-
"use_label":
|
| 713 |
},
|
| 714 |
{
|
| 715 |
"epoch": 0.73,
|
| 716 |
-
"grad_norm": 0.
|
| 717 |
"learning_rate": 1.0106369933615043e-06,
|
| 718 |
-
"logits/chosen": -2.
|
| 719 |
-
"logits/rejected": -2.
|
| 720 |
-
"logps/chosen": -
|
| 721 |
-
"logps/rejected": -
|
| 722 |
-
"loss": 0.
|
| 723 |
-
"pred_label":
|
| 724 |
-
"rewards/accuracies": 0.
|
| 725 |
-
"rewards/chosen": -0.
|
| 726 |
-
"rewards/margins": 0.
|
| 727 |
-
"rewards/rejected": -0.
|
| 728 |
"step": 700,
|
| 729 |
-
"use_label":
|
| 730 |
},
|
| 731 |
{
|
| 732 |
"epoch": 0.73,
|
| 733 |
-
"eval_logits/chosen": -
|
| 734 |
-
"eval_logits/rejected": -
|
| 735 |
-
"eval_logps/chosen": -
|
| 736 |
-
"eval_logps/rejected": -
|
| 737 |
-
"eval_loss": 0.
|
| 738 |
-
"eval_pred_label":
|
| 739 |
"eval_rewards/accuracies": 0.3511904776096344,
|
| 740 |
-
"eval_rewards/chosen": -0.
|
| 741 |
-
"eval_rewards/margins": 0.
|
| 742 |
-
"eval_rewards/rejected": -0.
|
| 743 |
-
"eval_runtime":
|
| 744 |
-
"eval_samples_per_second": 8.
|
| 745 |
"eval_steps_per_second": 0.254,
|
| 746 |
-
"eval_use_label":
|
| 747 |
"step": 700
|
| 748 |
},
|
| 749 |
{
|
| 750 |
"epoch": 0.75,
|
| 751 |
-
"grad_norm": 0.
|
| 752 |
"learning_rate": 8.678793653740633e-07,
|
| 753 |
-
"logits/chosen": -2.
|
| 754 |
-
"logits/rejected": -2.
|
| 755 |
-
"logps/chosen": -
|
| 756 |
-
"logps/rejected": -
|
| 757 |
-
"loss": 0.
|
| 758 |
-
"pred_label":
|
| 759 |
-
"rewards/accuracies": 0.
|
| 760 |
-
"rewards/chosen": -0.
|
| 761 |
-
"rewards/margins": 0.
|
| 762 |
-
"rewards/rejected": -0.
|
| 763 |
"step": 720,
|
| 764 |
-
"use_label":
|
| 765 |
},
|
| 766 |
{
|
| 767 |
"epoch": 0.77,
|
| 768 |
-
"grad_norm": 0.
|
| 769 |
"learning_rate": 7.338500848029603e-07,
|
| 770 |
-
"logits/chosen": -2.
|
| 771 |
-
"logits/rejected": -2.
|
| 772 |
-
"logps/chosen": -
|
| 773 |
-
"logps/rejected": -
|
| 774 |
-
"loss": 0.
|
| 775 |
-
"pred_label":
|
| 776 |
-
"rewards/accuracies": 0.
|
| 777 |
-
"rewards/chosen": -0.
|
| 778 |
-
"rewards/margins": 0.
|
| 779 |
-
"rewards/rejected": -0.
|
| 780 |
"step": 740,
|
| 781 |
-
"use_label":
|
| 782 |
},
|
| 783 |
{
|
| 784 |
"epoch": 0.8,
|
| 785 |
-
"grad_norm": 0.
|
| 786 |
"learning_rate": 6.092659210462232e-07,
|
| 787 |
-
"logits/chosen": -2.
|
| 788 |
-
"logits/rejected": -2.
|
| 789 |
-
"logps/chosen": -
|
| 790 |
-
"logps/rejected": -
|
| 791 |
-
"loss": 0.
|
| 792 |
-
"pred_label":
|
| 793 |
"rewards/accuracies": 0.33125001192092896,
|
| 794 |
-
"rewards/chosen": -0.
|
| 795 |
-
"rewards/margins": 0.
|
| 796 |
-
"rewards/rejected": -0.
|
| 797 |
"step": 760,
|
| 798 |
-
"use_label":
|
| 799 |
},
|
| 800 |
{
|
| 801 |
"epoch": 0.82,
|
| 802 |
-
"grad_norm": 0.
|
| 803 |
"learning_rate": 4.947931323697983e-07,
|
| 804 |
-
"logits/chosen": -2.
|
| 805 |
-
"logits/rejected": -2.
|
| 806 |
-
"logps/chosen": -
|
| 807 |
-
"logps/rejected": -
|
| 808 |
-
"loss": 0.
|
| 809 |
-
"pred_label":
|
| 810 |
"rewards/accuracies": 0.375,
|
| 811 |
-
"rewards/chosen": -0.
|
| 812 |
-
"rewards/margins": 0.
|
| 813 |
-
"rewards/rejected": -0.
|
| 814 |
"step": 780,
|
| 815 |
-
"use_label":
|
| 816 |
},
|
| 817 |
{
|
| 818 |
"epoch": 0.84,
|
| 819 |
-
"grad_norm": 0.
|
| 820 |
"learning_rate": 3.910439028537638e-07,
|
| 821 |
-
"logits/chosen": -2.
|
| 822 |
-
"logits/rejected": -
|
| 823 |
-
"logps/chosen": -
|
| 824 |
-
"logps/rejected": -
|
| 825 |
-
"loss": 0.
|
| 826 |
-
"pred_label":
|
| 827 |
-
"rewards/accuracies": 0.
|
| 828 |
-
"rewards/chosen": -0.
|
| 829 |
-
"rewards/margins": 0.
|
| 830 |
-
"rewards/rejected": -0.
|
| 831 |
"step": 800,
|
| 832 |
-
"use_label":
|
| 833 |
},
|
| 834 |
{
|
| 835 |
"epoch": 0.84,
|
| 836 |
-
"eval_logits/chosen": -
|
| 837 |
-
"eval_logits/rejected": -
|
| 838 |
-
"eval_logps/chosen": -
|
| 839 |
-
"eval_logps/rejected": -
|
| 840 |
-
"eval_loss": 0.
|
| 841 |
-
"eval_pred_label":
|
| 842 |
-
"eval_rewards/accuracies": 0.
|
| 843 |
-
"eval_rewards/chosen": -0.
|
| 844 |
-
"eval_rewards/margins": 0.
|
| 845 |
-
"eval_rewards/rejected": -0.
|
| 846 |
-
"eval_runtime": 247.
|
| 847 |
-
"eval_samples_per_second": 8.
|
| 848 |
"eval_steps_per_second": 0.254,
|
| 849 |
-
"eval_use_label":
|
| 850 |
"step": 800
|
| 851 |
},
|
| 852 |
{
|
| 853 |
"epoch": 0.86,
|
| 854 |
-
"grad_norm": 0.
|
| 855 |
"learning_rate": 2.98573068519539e-07,
|
| 856 |
-
"logits/chosen": -2.
|
| 857 |
-
"logits/rejected": -2.
|
| 858 |
-
"logps/chosen": -
|
| 859 |
-
"logps/rejected": -
|
| 860 |
-
"loss": 0.
|
| 861 |
-
"pred_label":
|
| 862 |
-
"rewards/accuracies": 0.
|
| 863 |
-
"rewards/chosen": -0.
|
| 864 |
-
"rewards/margins": 0.
|
| 865 |
-
"rewards/rejected": -0.
|
| 866 |
"step": 820,
|
| 867 |
-
"use_label":
|
| 868 |
},
|
| 869 |
{
|
| 870 |
"epoch": 0.88,
|
| 871 |
-
"grad_norm": 0.
|
| 872 |
"learning_rate": 2.178751501463036e-07,
|
| 873 |
-
"logits/chosen": -2.
|
| 874 |
-
"logits/rejected": -2.
|
| 875 |
-
"logps/chosen": -
|
| 876 |
-
"logps/rejected": -
|
| 877 |
-
"loss": 0.
|
| 878 |
-
"pred_label":
|
| 879 |
-
"rewards/accuracies": 0.
|
| 880 |
-
"rewards/chosen": -0.
|
| 881 |
-
"rewards/margins": 0.
|
| 882 |
-
"rewards/rejected": -0.
|
| 883 |
"step": 840,
|
| 884 |
-
"use_label":
|
| 885 |
},
|
| 886 |
{
|
| 887 |
"epoch": 0.9,
|
| 888 |
-
"grad_norm": 0.
|
| 889 |
"learning_rate": 1.4938170864468636e-07,
|
| 890 |
-
"logits/chosen": -2.
|
| 891 |
-
"logits/rejected": -2.
|
| 892 |
-
"logps/chosen": -
|
| 893 |
-
"logps/rejected": -
|
| 894 |
-
"loss": 0.
|
| 895 |
-
"pred_label":
|
| 896 |
-
"rewards/accuracies": 0.
|
| 897 |
-
"rewards/chosen": -0.
|
| 898 |
-
"rewards/margins": 0.
|
| 899 |
-
"rewards/rejected": -0.
|
| 900 |
"step": 860,
|
| 901 |
-
"use_label":
|
| 902 |
},
|
| 903 |
{
|
| 904 |
"epoch": 0.92,
|
| 905 |
-
"grad_norm": 0.
|
| 906 |
"learning_rate": 9.345903713082305e-08,
|
| 907 |
-
"logits/chosen": -2.
|
| 908 |
-
"logits/rejected": -2.
|
| 909 |
-
"logps/chosen": -
|
| 910 |
-
"logps/rejected": -
|
| 911 |
-
"loss": 0.
|
| 912 |
-
"pred_label":
|
| 913 |
-
"rewards/accuracies": 0.
|
| 914 |
-
"rewards/chosen": -0.
|
| 915 |
-
"rewards/margins": 0.
|
| 916 |
-
"rewards/rejected": -0.
|
| 917 |
"step": 880,
|
| 918 |
-
"use_label":
|
| 919 |
},
|
| 920 |
{
|
| 921 |
"epoch": 0.94,
|
| 922 |
-
"grad_norm": 0.
|
| 923 |
"learning_rate": 5.0406202043228604e-08,
|
| 924 |
-
"logits/chosen": -
|
| 925 |
-
"logits/rejected": -
|
| 926 |
-
"logps/chosen": -
|
| 927 |
-
"logps/rejected": -
|
| 928 |
-
"loss": 0.
|
| 929 |
-
"pred_label":
|
| 930 |
"rewards/accuracies": 0.3375000059604645,
|
| 931 |
-
"rewards/chosen": -0.
|
| 932 |
-
"rewards/margins": 0.
|
| 933 |
-
"rewards/rejected": -0.
|
| 934 |
"step": 900,
|
| 935 |
-
"use_label":
|
| 936 |
},
|
| 937 |
{
|
| 938 |
"epoch": 0.94,
|
| 939 |
-
"eval_logits/chosen": -
|
| 940 |
-
"eval_logits/rejected": -
|
| 941 |
-
"eval_logps/chosen": -
|
| 942 |
-
"eval_logps/rejected": -
|
| 943 |
-
"eval_loss": 0.
|
| 944 |
-
"eval_pred_label":
|
| 945 |
-
"eval_rewards/accuracies": 0.
|
| 946 |
-
"eval_rewards/chosen": -0.
|
| 947 |
-
"eval_rewards/margins": 0.
|
| 948 |
-
"eval_rewards/rejected": -0.
|
| 949 |
-
"eval_runtime":
|
| 950 |
-
"eval_samples_per_second": 8.
|
| 951 |
-
"eval_steps_per_second": 0.
|
| 952 |
-
"eval_use_label":
|
| 953 |
"step": 900
|
| 954 |
},
|
| 955 |
{
|
| 956 |
"epoch": 0.96,
|
| 957 |
-
"grad_norm": 0.
|
| 958 |
"learning_rate": 2.0453443778310766e-08,
|
| 959 |
-
"logits/chosen": -
|
| 960 |
-
"logits/rejected": -
|
| 961 |
-
"logps/chosen": -
|
| 962 |
-
"logps/rejected": -
|
| 963 |
-
"loss": 0.
|
| 964 |
-
"pred_label":
|
| 965 |
-
"rewards/accuracies": 0.
|
| 966 |
-
"rewards/chosen": -0.
|
| 967 |
-
"rewards/margins": 0.
|
| 968 |
-
"rewards/rejected": -0.
|
| 969 |
"step": 920,
|
| 970 |
-
"use_label":
|
| 971 |
},
|
| 972 |
{
|
| 973 |
"epoch": 0.98,
|
| 974 |
-
"grad_norm": 0.
|
| 975 |
"learning_rate": 3.760945397705828e-09,
|
| 976 |
-
"logits/chosen": -
|
| 977 |
-
"logits/rejected": -
|
| 978 |
-
"logps/chosen": -
|
| 979 |
-
"logps/rejected": -
|
| 980 |
-
"loss": 0.
|
| 981 |
-
"pred_label":
|
| 982 |
"rewards/accuracies": 0.32499998807907104,
|
| 983 |
-
"rewards/chosen": -0.
|
| 984 |
-
"rewards/margins": 0.
|
| 985 |
-
"rewards/rejected": -0.
|
| 986 |
"step": 940,
|
| 987 |
-
"use_label":
|
| 988 |
},
|
| 989 |
{
|
| 990 |
"epoch": 1.0,
|
| 991 |
"step": 955,
|
| 992 |
"total_flos": 0.0,
|
| 993 |
-
"train_loss": 0.
|
| 994 |
-
"train_runtime":
|
| 995 |
-
"train_samples_per_second": 3.
|
| 996 |
"train_steps_per_second": 0.048
|
| 997 |
}
|
| 998 |
],
|
|
|
|
| 29 |
"epoch": 0.02,
|
| 30 |
"grad_norm": 0.6796875,
|
| 31 |
"learning_rate": 1.0416666666666667e-06,
|
| 32 |
+
"logits/chosen": -2.227864980697632,
|
| 33 |
+
"logits/rejected": -2.276106834411621,
|
| 34 |
+
"logps/chosen": -57.02927780151367,
|
| 35 |
+
"logps/rejected": -66.8729019165039,
|
| 36 |
"loss": 0.6927,
|
| 37 |
"pred_label": 0.0,
|
| 38 |
+
"rewards/accuracies": 0.24671052396297455,
|
| 39 |
+
"rewards/chosen": 0.003993770573288202,
|
| 40 |
+
"rewards/margins": 0.0009077258291654289,
|
| 41 |
+
"rewards/rejected": 0.003086044918745756,
|
| 42 |
"step": 20,
|
| 43 |
"use_label": 170.0
|
| 44 |
},
|
| 45 |
{
|
| 46 |
"epoch": 0.04,
|
| 47 |
+
"grad_norm": 0.546875,
|
| 48 |
"learning_rate": 2.0833333333333334e-06,
|
| 49 |
+
"logits/chosen": -2.2728817462921143,
|
| 50 |
+
"logits/rejected": -2.261592388153076,
|
| 51 |
+
"logps/chosen": -54.7827033996582,
|
| 52 |
+
"logps/rejected": -67.2376708984375,
|
| 53 |
+
"loss": 0.6915,
|
| 54 |
"pred_label": 0.0,
|
| 55 |
"rewards/accuracies": 0.24687500298023224,
|
| 56 |
+
"rewards/chosen": 0.017464743927121162,
|
| 57 |
+
"rewards/margins": 0.0016005486249923706,
|
| 58 |
+
"rewards/rejected": 0.015864195302128792,
|
| 59 |
"step": 40,
|
| 60 |
"use_label": 482.0
|
| 61 |
},
|
| 62 |
{
|
| 63 |
"epoch": 0.06,
|
| 64 |
+
"grad_norm": 0.625,
|
| 65 |
"learning_rate": 3.125e-06,
|
| 66 |
+
"logits/chosen": -2.321017026901245,
|
| 67 |
+
"logits/rejected": -2.318946123123169,
|
| 68 |
+
"logps/chosen": -75.58020782470703,
|
| 69 |
+
"logps/rejected": -87.66261291503906,
|
| 70 |
+
"loss": 0.6905,
|
| 71 |
+
"pred_label": 4.987500190734863,
|
| 72 |
+
"rewards/accuracies": 0.3187499940395355,
|
| 73 |
+
"rewards/chosen": 0.031644098460674286,
|
| 74 |
+
"rewards/margins": 0.009459299966692924,
|
| 75 |
+
"rewards/rejected": 0.02218480221927166,
|
| 76 |
"step": 60,
|
| 77 |
+
"use_label": 797.0125122070312
|
| 78 |
},
|
| 79 |
{
|
| 80 |
"epoch": 0.08,
|
| 81 |
+
"grad_norm": 0.57421875,
|
| 82 |
"learning_rate": 4.166666666666667e-06,
|
| 83 |
+
"logits/chosen": -2.2973294258117676,
|
| 84 |
+
"logits/rejected": -2.2655692100524902,
|
| 85 |
+
"logps/chosen": -77.97566223144531,
|
| 86 |
+
"logps/rejected": -81.31121826171875,
|
| 87 |
+
"loss": 0.6909,
|
| 88 |
+
"pred_label": 29.850000381469727,
|
| 89 |
"rewards/accuracies": 0.3343749940395355,
|
| 90 |
+
"rewards/chosen": 0.02917659282684326,
|
| 91 |
+
"rewards/margins": 0.014682939276099205,
|
| 92 |
+
"rewards/rejected": 0.014493651688098907,
|
| 93 |
"step": 80,
|
| 94 |
+
"use_label": 1092.1500244140625
|
| 95 |
},
|
| 96 |
{
|
| 97 |
"epoch": 0.1,
|
| 98 |
+
"grad_norm": 0.625,
|
| 99 |
"learning_rate": 4.9997324926814375e-06,
|
| 100 |
+
"logits/chosen": -2.210732936859131,
|
| 101 |
+
"logits/rejected": -2.2176434993743896,
|
| 102 |
+
"logps/chosen": -66.4733657836914,
|
| 103 |
+
"logps/rejected": -74.55338287353516,
|
| 104 |
+
"loss": 0.6911,
|
| 105 |
+
"pred_label": 68.07499694824219,
|
| 106 |
+
"rewards/accuracies": 0.33125001192092896,
|
| 107 |
+
"rewards/chosen": 0.013588580302894115,
|
| 108 |
+
"rewards/margins": 0.026675995439291,
|
| 109 |
+
"rewards/rejected": -0.01308741606771946,
|
| 110 |
"step": 100,
|
| 111 |
+
"use_label": 1373.925048828125
|
| 112 |
},
|
| 113 |
{
|
| 114 |
"epoch": 0.1,
|
| 115 |
+
"eval_logits/chosen": -2.120985984802246,
|
| 116 |
+
"eval_logits/rejected": -2.093513250350952,
|
| 117 |
+
"eval_logps/chosen": -69.42622375488281,
|
| 118 |
+
"eval_logps/rejected": -78.9540786743164,
|
| 119 |
+
"eval_loss": 0.691917359828949,
|
| 120 |
+
"eval_pred_label": 150.14285278320312,
|
| 121 |
+
"eval_rewards/accuracies": 0.3392857015132904,
|
| 122 |
+
"eval_rewards/chosen": -0.005256766453385353,
|
| 123 |
+
"eval_rewards/margins": 0.030303288251161575,
|
| 124 |
+
"eval_rewards/rejected": -0.03556005656719208,
|
| 125 |
+
"eval_runtime": 247.9513,
|
| 126 |
+
"eval_samples_per_second": 8.066,
|
| 127 |
"eval_steps_per_second": 0.254,
|
| 128 |
+
"eval_use_label": 1705.857177734375,
|
| 129 |
"step": 100
|
| 130 |
},
|
| 131 |
{
|
| 132 |
"epoch": 0.13,
|
| 133 |
+
"grad_norm": 0.53125,
|
| 134 |
"learning_rate": 4.9903757462135984e-06,
|
| 135 |
+
"logits/chosen": -2.24790620803833,
|
| 136 |
+
"logits/rejected": -2.1782658100128174,
|
| 137 |
+
"logps/chosen": -67.23531341552734,
|
| 138 |
+
"logps/rejected": -80.04717254638672,
|
| 139 |
+
"loss": 0.6914,
|
| 140 |
+
"pred_label": 243.0,
|
| 141 |
+
"rewards/accuracies": 0.3125,
|
| 142 |
+
"rewards/chosen": 0.00682856747880578,
|
| 143 |
+
"rewards/margins": 0.03467796370387077,
|
| 144 |
+
"rewards/rejected": -0.02784939482808113,
|
| 145 |
"step": 120,
|
| 146 |
+
"use_label": 2023.0
|
| 147 |
},
|
| 148 |
{
|
| 149 |
"epoch": 0.15,
|
| 150 |
+
"grad_norm": 0.341796875,
|
| 151 |
"learning_rate": 4.967700826904229e-06,
|
| 152 |
+
"logits/chosen": -2.1205825805664062,
|
| 153 |
+
"logits/rejected": -2.150360584259033,
|
| 154 |
+
"logps/chosen": -58.376564025878906,
|
| 155 |
+
"logps/rejected": -71.84730529785156,
|
| 156 |
+
"loss": 0.6917,
|
| 157 |
+
"pred_label": 304.9125061035156,
|
| 158 |
+
"rewards/accuracies": 0.2874999940395355,
|
| 159 |
+
"rewards/chosen": 0.0049698068760335445,
|
| 160 |
+
"rewards/margins": 0.025117725133895874,
|
| 161 |
+
"rewards/rejected": -0.02014791965484619,
|
| 162 |
"step": 140,
|
| 163 |
+
"use_label": 2281.08740234375
|
| 164 |
},
|
| 165 |
{
|
| 166 |
"epoch": 0.17,
|
| 167 |
+
"grad_norm": 0.515625,
|
| 168 |
"learning_rate": 4.931828996974498e-06,
|
| 169 |
+
"logits/chosen": -2.1872148513793945,
|
| 170 |
+
"logits/rejected": -2.1555256843566895,
|
| 171 |
+
"logps/chosen": -66.3367919921875,
|
| 172 |
+
"logps/rejected": -69.24983215332031,
|
| 173 |
+
"loss": 0.6918,
|
| 174 |
+
"pred_label": 371.5375061035156,
|
| 175 |
+
"rewards/accuracies": 0.3218750059604645,
|
| 176 |
+
"rewards/chosen": -0.017877796664834023,
|
| 177 |
+
"rewards/margins": 0.03909943252801895,
|
| 178 |
+
"rewards/rejected": -0.05697723478078842,
|
| 179 |
"step": 160,
|
| 180 |
+
"use_label": 2534.46240234375
|
| 181 |
},
|
| 182 |
{
|
| 183 |
"epoch": 0.19,
|
| 184 |
+
"grad_norm": 0.4921875,
|
| 185 |
"learning_rate": 4.882952093833628e-06,
|
| 186 |
+
"logits/chosen": -2.1010584831237793,
|
| 187 |
+
"logits/rejected": -2.112929582595825,
|
| 188 |
+
"logps/chosen": -67.18075561523438,
|
| 189 |
+
"logps/rejected": -77.23786163330078,
|
| 190 |
+
"loss": 0.6925,
|
| 191 |
+
"pred_label": 444.2124938964844,
|
| 192 |
+
"rewards/accuracies": 0.3031249940395355,
|
| 193 |
+
"rewards/chosen": -0.03951570764183998,
|
| 194 |
+
"rewards/margins": 0.03356783464550972,
|
| 195 |
+
"rewards/rejected": -0.0730835422873497,
|
| 196 |
"step": 180,
|
| 197 |
+
"use_label": 2781.78759765625
|
| 198 |
},
|
| 199 |
{
|
| 200 |
"epoch": 0.21,
|
| 201 |
+
"grad_norm": 0.546875,
|
| 202 |
"learning_rate": 4.821331504159906e-06,
|
| 203 |
+
"logits/chosen": -2.181281805038452,
|
| 204 |
+
"logits/rejected": -2.155298948287964,
|
| 205 |
+
"logps/chosen": -78.88096618652344,
|
| 206 |
+
"logps/rejected": -77.27136993408203,
|
| 207 |
+
"loss": 0.692,
|
| 208 |
+
"pred_label": 513.2125244140625,
|
| 209 |
+
"rewards/accuracies": 0.3812499940395355,
|
| 210 |
+
"rewards/chosen": -0.019123973324894905,
|
| 211 |
+
"rewards/margins": 0.040298379957675934,
|
| 212 |
+
"rewards/rejected": -0.05942235141992569,
|
| 213 |
"step": 200,
|
| 214 |
+
"use_label": 3032.78759765625
|
| 215 |
},
|
| 216 |
{
|
| 217 |
"epoch": 0.21,
|
| 218 |
+
"eval_logits/chosen": -2.1267549991607666,
|
| 219 |
+
"eval_logits/rejected": -2.1057066917419434,
|
| 220 |
+
"eval_logps/chosen": -71.54093170166016,
|
| 221 |
+
"eval_logps/rejected": -82.35039520263672,
|
| 222 |
+
"eval_loss": 0.6926834583282471,
|
| 223 |
+
"eval_pred_label": 622.952392578125,
|
| 224 |
+
"eval_rewards/accuracies": 0.3432539701461792,
|
| 225 |
+
"eval_rewards/chosen": -0.026403911411762238,
|
| 226 |
+
"eval_rewards/margins": 0.043119337409734726,
|
| 227 |
+
"eval_rewards/rejected": -0.06952324509620667,
|
| 228 |
+
"eval_runtime": 248.2687,
|
| 229 |
+
"eval_samples_per_second": 8.056,
|
| 230 |
"eval_steps_per_second": 0.254,
|
| 231 |
+
"eval_use_label": 3337.047607421875,
|
| 232 |
"step": 200
|
| 233 |
},
|
| 234 |
{
|
| 235 |
"epoch": 0.23,
|
| 236 |
+
"grad_norm": 0.609375,
|
| 237 |
"learning_rate": 4.747296766042161e-06,
|
| 238 |
+
"logits/chosen": -2.2548727989196777,
|
| 239 |
+
"logits/rejected": -2.2427258491516113,
|
| 240 |
+
"logps/chosen": -74.4991683959961,
|
| 241 |
+
"logps/rejected": -75.8321762084961,
|
| 242 |
+
"loss": 0.6924,
|
| 243 |
+
"pred_label": 738.5,
|
| 244 |
+
"rewards/accuracies": 0.3531250059604645,
|
| 245 |
+
"rewards/chosen": -0.024670986458659172,
|
| 246 |
+
"rewards/margins": 0.04779377579689026,
|
| 247 |
+
"rewards/rejected": -0.07246476411819458,
|
| 248 |
"step": 220,
|
| 249 |
+
"use_label": 3631.5
|
| 250 |
},
|
| 251 |
{
|
| 252 |
"epoch": 0.25,
|
| 253 |
+
"grad_norm": 0.46875,
|
| 254 |
"learning_rate": 4.661243806657256e-06,
|
| 255 |
+
"logits/chosen": -2.2358717918395996,
|
| 256 |
+
"logits/rejected": -2.216477870941162,
|
| 257 |
+
"logps/chosen": -72.57451629638672,
|
| 258 |
+
"logps/rejected": -79.20014953613281,
|
| 259 |
+
"loss": 0.6921,
|
| 260 |
+
"pred_label": 830.7750244140625,
|
| 261 |
+
"rewards/accuracies": 0.3499999940395355,
|
| 262 |
+
"rewards/chosen": -0.013481785543262959,
|
| 263 |
+
"rewards/margins": 0.0440317802131176,
|
| 264 |
+
"rewards/rejected": -0.05751357227563858,
|
| 265 |
"step": 240,
|
| 266 |
+
"use_label": 3859.22509765625
|
| 267 |
},
|
| 268 |
{
|
| 269 |
"epoch": 0.27,
|
| 270 |
+
"grad_norm": 0.5390625,
|
| 271 |
"learning_rate": 4.563632824908252e-06,
|
| 272 |
+
"logits/chosen": -2.204738140106201,
|
| 273 |
+
"logits/rejected": -2.2045350074768066,
|
| 274 |
+
"logps/chosen": -64.52825164794922,
|
| 275 |
+
"logps/rejected": -74.71345520019531,
|
| 276 |
+
"loss": 0.6919,
|
| 277 |
+
"pred_label": 912.1624755859375,
|
| 278 |
+
"rewards/accuracies": 0.3125,
|
| 279 |
+
"rewards/chosen": -0.01493888907134533,
|
| 280 |
+
"rewards/margins": 0.038629818707704544,
|
| 281 |
+
"rewards/rejected": -0.05356870964169502,
|
| 282 |
"step": 260,
|
| 283 |
+
"use_label": 4097.83740234375
|
| 284 |
},
|
| 285 |
{
|
| 286 |
"epoch": 0.29,
|
| 287 |
+
"grad_norm": 0.431640625,
|
| 288 |
"learning_rate": 4.454985830346574e-06,
|
| 289 |
+
"logits/chosen": -2.224844455718994,
|
| 290 |
+
"logits/rejected": -2.247999668121338,
|
| 291 |
+
"logps/chosen": -72.3452377319336,
|
| 292 |
+
"logps/rejected": -75.01800537109375,
|
| 293 |
+
"loss": 0.6916,
|
| 294 |
+
"pred_label": 993.7874755859375,
|
| 295 |
+
"rewards/accuracies": 0.3125,
|
| 296 |
+
"rewards/chosen": -0.04014473780989647,
|
| 297 |
+
"rewards/margins": 0.030534306541085243,
|
| 298 |
+
"rewards/rejected": -0.07067903876304626,
|
| 299 |
"step": 280,
|
| 300 |
+
"use_label": 4336.21240234375
|
| 301 |
},
|
| 302 |
{
|
| 303 |
"epoch": 0.31,
|
| 304 |
+
"grad_norm": 0.423828125,
|
| 305 |
"learning_rate": 4.335883851539693e-06,
|
| 306 |
+
"logits/chosen": -2.2155380249023438,
|
| 307 |
+
"logits/rejected": -2.2151846885681152,
|
| 308 |
+
"logps/chosen": -67.15587615966797,
|
| 309 |
+
"logps/rejected": -74.2086181640625,
|
| 310 |
+
"loss": 0.6924,
|
| 311 |
+
"pred_label": 1083.4625244140625,
|
| 312 |
+
"rewards/accuracies": 0.34687501192092896,
|
| 313 |
+
"rewards/chosen": -0.020541679114103317,
|
| 314 |
+
"rewards/margins": 0.06299655884504318,
|
| 315 |
+
"rewards/rejected": -0.0835382491350174,
|
| 316 |
"step": 300,
|
| 317 |
+
"use_label": 4566.53759765625
|
| 318 |
},
|
| 319 |
{
|
| 320 |
"epoch": 0.31,
|
| 321 |
+
"eval_logits/chosen": -2.2169294357299805,
|
| 322 |
+
"eval_logits/rejected": -2.1932876110076904,
|
| 323 |
+
"eval_logps/chosen": -72.5876693725586,
|
| 324 |
+
"eval_logps/rejected": -84.35366821289062,
|
| 325 |
+
"eval_loss": 0.6928625702857971,
|
| 326 |
+
"eval_pred_label": 1200.2698974609375,
|
| 327 |
+
"eval_rewards/accuracies": 0.3392857015132904,
|
| 328 |
+
"eval_rewards/chosen": -0.03687124699354172,
|
| 329 |
+
"eval_rewards/margins": 0.0526847243309021,
|
| 330 |
+
"eval_rewards/rejected": -0.08955597132444382,
|
| 331 |
+
"eval_runtime": 247.9119,
|
| 332 |
+
"eval_samples_per_second": 8.067,
|
| 333 |
"eval_steps_per_second": 0.254,
|
| 334 |
+
"eval_use_label": 4863.72998046875,
|
| 335 |
"step": 300
|
| 336 |
},
|
| 337 |
{
|
| 338 |
"epoch": 0.33,
|
| 339 |
+
"grad_norm": 0.61328125,
|
| 340 |
"learning_rate": 4.206963828813555e-06,
|
| 341 |
+
"logits/chosen": -2.291391134262085,
|
| 342 |
+
"logits/rejected": -2.3002986907958984,
|
| 343 |
+
"logps/chosen": -68.5405502319336,
|
| 344 |
+
"logps/rejected": -83.0180435180664,
|
| 345 |
+
"loss": 0.6927,
|
| 346 |
+
"pred_label": 1323.074951171875,
|
| 347 |
+
"rewards/accuracies": 0.30000001192092896,
|
| 348 |
+
"rewards/chosen": -0.07074997574090958,
|
| 349 |
+
"rewards/margins": 0.04067195579409599,
|
| 350 |
+
"rewards/rejected": -0.11142192780971527,
|
| 351 |
"step": 320,
|
| 352 |
+
"use_label": 5150.9248046875
|
| 353 |
},
|
| 354 |
{
|
| 355 |
"epoch": 0.36,
|
| 356 |
+
"grad_norm": 0.455078125,
|
| 357 |
"learning_rate": 4.068915207986931e-06,
|
| 358 |
+
"logits/chosen": -2.2867865562438965,
|
| 359 |
+
"logits/rejected": -2.2617173194885254,
|
| 360 |
+
"logps/chosen": -64.90373229980469,
|
| 361 |
+
"logps/rejected": -74.42888641357422,
|
| 362 |
+
"loss": 0.692,
|
| 363 |
+
"pred_label": 1427.7750244140625,
|
| 364 |
+
"rewards/accuracies": 0.3531250059604645,
|
| 365 |
+
"rewards/chosen": -0.016644436866044998,
|
| 366 |
+
"rewards/margins": 0.052551619708538055,
|
| 367 |
+
"rewards/rejected": -0.06919606029987335,
|
| 368 |
"step": 340,
|
| 369 |
+
"use_label": 5366.22509765625
|
| 370 |
},
|
| 371 |
{
|
| 372 |
"epoch": 0.38,
|
| 373 |
+
"grad_norm": 0.458984375,
|
| 374 |
"learning_rate": 3.922476253313921e-06,
|
| 375 |
+
"logits/chosen": -2.249298572540283,
|
| 376 |
+
"logits/rejected": -2.253566265106201,
|
| 377 |
+
"logps/chosen": -68.57295989990234,
|
| 378 |
+
"logps/rejected": -73.1113510131836,
|
| 379 |
+
"loss": 0.693,
|
| 380 |
+
"pred_label": 1522.0999755859375,
|
| 381 |
+
"rewards/accuracies": 0.328125,
|
| 382 |
+
"rewards/chosen": -0.037180084735155106,
|
| 383 |
+
"rewards/margins": 0.045733559876680374,
|
| 384 |
+
"rewards/rejected": -0.08291363716125488,
|
| 385 |
"step": 360,
|
| 386 |
+
"use_label": 5591.89990234375
|
| 387 |
},
|
| 388 |
{
|
| 389 |
"epoch": 0.4,
|
| 390 |
+
"grad_norm": 0.4453125,
|
| 391 |
"learning_rate": 3.768430099352445e-06,
|
| 392 |
+
"logits/chosen": -2.2458603382110596,
|
| 393 |
+
"logits/rejected": -2.2051453590393066,
|
| 394 |
+
"logps/chosen": -70.38607788085938,
|
| 395 |
+
"logps/rejected": -78.15666198730469,
|
| 396 |
+
"loss": 0.6923,
|
| 397 |
+
"pred_label": 1625.5374755859375,
|
| 398 |
+
"rewards/accuracies": 0.3531250059604645,
|
| 399 |
+
"rewards/chosen": -0.03562153875827789,
|
| 400 |
+
"rewards/margins": 0.054723359644412994,
|
| 401 |
+
"rewards/rejected": -0.09034489840269089,
|
| 402 |
"step": 380,
|
| 403 |
+
"use_label": 5808.46240234375
|
| 404 |
},
|
| 405 |
{
|
| 406 |
"epoch": 0.42,
|
| 407 |
+
"grad_norm": 0.59765625,
|
| 408 |
"learning_rate": 3.607600562872785e-06,
|
| 409 |
+
"logits/chosen": -2.196977138519287,
|
| 410 |
+
"logits/rejected": -2.197218656539917,
|
| 411 |
+
"logps/chosen": -81.0395736694336,
|
| 412 |
+
"logps/rejected": -81.44091033935547,
|
| 413 |
+
"loss": 0.6927,
|
| 414 |
+
"pred_label": 1725.362548828125,
|
| 415 |
+
"rewards/accuracies": 0.32499998807907104,
|
| 416 |
+
"rewards/chosen": -0.03092697635293007,
|
| 417 |
+
"rewards/margins": 0.049932099878787994,
|
| 418 |
+
"rewards/rejected": -0.08085907250642776,
|
| 419 |
"step": 400,
|
| 420 |
+
"use_label": 6028.6376953125
|
| 421 |
},
|
| 422 |
{
|
| 423 |
"epoch": 0.42,
|
| 424 |
+
"eval_logits/chosen": -2.118962526321411,
|
| 425 |
+
"eval_logits/rejected": -2.093430995941162,
|
| 426 |
+
"eval_logps/chosen": -71.01036071777344,
|
| 427 |
+
"eval_logps/rejected": -83.43638610839844,
|
| 428 |
+
"eval_loss": 0.6925376653671265,
|
| 429 |
+
"eval_pred_label": 1843.920654296875,
|
| 430 |
+
"eval_rewards/accuracies": 0.341269850730896,
|
| 431 |
+
"eval_rewards/chosen": -0.021098149940371513,
|
| 432 |
+
"eval_rewards/margins": 0.05928494408726692,
|
| 433 |
+
"eval_rewards/rejected": -0.08038310706615448,
|
| 434 |
+
"eval_runtime": 248.0095,
|
| 435 |
"eval_samples_per_second": 8.064,
|
| 436 |
"eval_steps_per_second": 0.254,
|
| 437 |
+
"eval_use_label": 6324.07958984375,
|
| 438 |
"step": 400
|
| 439 |
},
|
| 440 |
{
|
| 441 |
"epoch": 0.44,
|
| 442 |
+
"grad_norm": 0.30078125,
|
| 443 |
"learning_rate": 3.4408477372034743e-06,
|
| 444 |
+
"logits/chosen": -2.146075487136841,
|
| 445 |
+
"logits/rejected": -2.152238607406616,
|
| 446 |
+
"logps/chosen": -65.8438720703125,
|
| 447 |
+
"logps/rejected": -70.74162292480469,
|
| 448 |
+
"loss": 0.692,
|
| 449 |
+
"pred_label": 1975.637451171875,
|
| 450 |
+
"rewards/accuracies": 0.3531250059604645,
|
| 451 |
+
"rewards/chosen": -0.017682421952486038,
|
| 452 |
+
"rewards/margins": 0.05984373763203621,
|
| 453 |
+
"rewards/rejected": -0.07752615213394165,
|
| 454 |
"step": 420,
|
| 455 |
+
"use_label": 6602.3623046875
|
| 456 |
},
|
| 457 |
{
|
| 458 |
"epoch": 0.46,
|
| 459 |
+
"grad_norm": 0.9296875,
|
| 460 |
"learning_rate": 3.269063392575352e-06,
|
| 461 |
+
"logits/chosen": -2.2523856163024902,
|
| 462 |
+
"logits/rejected": -2.2490224838256836,
|
| 463 |
+
"logps/chosen": -74.74308013916016,
|
| 464 |
+
"logps/rejected": -74.57176208496094,
|
| 465 |
+
"loss": 0.6927,
|
| 466 |
+
"pred_label": 2072.27490234375,
|
| 467 |
+
"rewards/accuracies": 0.3218750059604645,
|
| 468 |
+
"rewards/chosen": -0.027858540415763855,
|
| 469 |
+
"rewards/margins": 0.05976608395576477,
|
| 470 |
+
"rewards/rejected": -0.08762462437152863,
|
| 471 |
"step": 440,
|
| 472 |
+
"use_label": 6825.72509765625
|
| 473 |
},
|
| 474 |
{
|
| 475 |
"epoch": 0.48,
|
| 476 |
+
"grad_norm": 0.34375,
|
| 477 |
"learning_rate": 3.09316620706208e-06,
|
| 478 |
+
"logits/chosen": -2.2484962940216064,
|
| 479 |
+
"logits/rejected": -2.253873109817505,
|
| 480 |
+
"logps/chosen": -68.02134704589844,
|
| 481 |
+
"logps/rejected": -73.40286254882812,
|
| 482 |
+
"loss": 0.6929,
|
| 483 |
+
"pred_label": 2175.53759765625,
|
| 484 |
+
"rewards/accuracies": 0.328125,
|
| 485 |
+
"rewards/chosen": -0.03111925721168518,
|
| 486 |
+
"rewards/margins": 0.06376632302999496,
|
| 487 |
+
"rewards/rejected": -0.09488557279109955,
|
| 488 |
"step": 460,
|
| 489 |
+
"use_label": 7042.46240234375
|
| 490 |
},
|
| 491 |
{
|
| 492 |
"epoch": 0.5,
|
| 493 |
+
"grad_norm": 0.3984375,
|
| 494 |
"learning_rate": 2.91409685362137e-06,
|
| 495 |
+
"logits/chosen": -2.2359812259674072,
|
| 496 |
+
"logits/rejected": -2.2330563068389893,
|
| 497 |
+
"logps/chosen": -75.03883361816406,
|
| 498 |
+
"logps/rejected": -84.55928039550781,
|
| 499 |
+
"loss": 0.6922,
|
| 500 |
+
"pred_label": 2276.949951171875,
|
| 501 |
+
"rewards/accuracies": 0.3687500059604645,
|
| 502 |
+
"rewards/chosen": -0.040116917341947556,
|
| 503 |
+
"rewards/margins": 0.0741645023226738,
|
| 504 |
+
"rewards/rejected": -0.11428143084049225,
|
| 505 |
"step": 480,
|
| 506 |
+
"use_label": 7261.0498046875
|
| 507 |
},
|
| 508 |
{
|
| 509 |
"epoch": 0.52,
|
| 510 |
+
"grad_norm": 0.5703125,
|
| 511 |
"learning_rate": 2.7328129695107205e-06,
|
| 512 |
+
"logits/chosen": -2.2053210735321045,
|
| 513 |
+
"logits/rejected": -2.2094616889953613,
|
| 514 |
+
"logps/chosen": -75.30181121826172,
|
| 515 |
+
"logps/rejected": -77.61902618408203,
|
| 516 |
+
"loss": 0.6924,
|
| 517 |
+
"pred_label": 2379.137451171875,
|
| 518 |
+
"rewards/accuracies": 0.3656249940395355,
|
| 519 |
+
"rewards/chosen": -0.039206866174936295,
|
| 520 |
+
"rewards/margins": 0.05418051406741142,
|
| 521 |
+
"rewards/rejected": -0.09338738024234772,
|
| 522 |
"step": 500,
|
| 523 |
+
"use_label": 7478.8623046875
|
| 524 |
},
|
| 525 |
{
|
| 526 |
"epoch": 0.52,
|
| 527 |
+
"eval_logits/chosen": -2.176236152648926,
|
| 528 |
+
"eval_logits/rejected": -2.151799201965332,
|
| 529 |
+
"eval_logps/chosen": -70.96183776855469,
|
| 530 |
+
"eval_logps/rejected": -83.7112045288086,
|
| 531 |
+
"eval_loss": 0.6929337382316589,
|
| 532 |
+
"eval_pred_label": 2499.22216796875,
|
| 533 |
+
"eval_rewards/accuracies": 0.3432539701461792,
|
| 534 |
+
"eval_rewards/chosen": -0.02061287872493267,
|
| 535 |
+
"eval_rewards/margins": 0.06251849234104156,
|
| 536 |
+
"eval_rewards/rejected": -0.08313137292861938,
|
| 537 |
+
"eval_runtime": 248.0888,
|
| 538 |
+
"eval_samples_per_second": 8.062,
|
| 539 |
"eval_steps_per_second": 0.254,
|
| 540 |
+
"eval_use_label": 7772.77783203125,
|
| 541 |
"step": 500
|
| 542 |
},
|
| 543 |
{
|
| 544 |
"epoch": 0.54,
|
| 545 |
+
"grad_norm": 0.61328125,
|
| 546 |
"learning_rate": 2.5502840349805074e-06,
|
| 547 |
+
"logits/chosen": -2.195094347000122,
|
| 548 |
+
"logits/rejected": -2.237112045288086,
|
| 549 |
+
"logps/chosen": -70.13484954833984,
|
| 550 |
+
"logps/rejected": -79.53434753417969,
|
| 551 |
+
"loss": 0.692,
|
| 552 |
+
"pred_label": 2632.125,
|
| 553 |
+
"rewards/accuracies": 0.3656249940395355,
|
| 554 |
+
"rewards/chosen": -0.018158430233597755,
|
| 555 |
+
"rewards/margins": 0.061979226768016815,
|
| 556 |
+
"rewards/rejected": -0.08013766258955002,
|
| 557 |
"step": 520,
|
| 558 |
+
"use_label": 8049.875
|
| 559 |
},
|
| 560 |
{
|
| 561 |
"epoch": 0.57,
|
| 562 |
+
"grad_norm": 0.55078125,
|
| 563 |
"learning_rate": 2.367486188632446e-06,
|
| 564 |
+
"logits/chosen": -2.1844329833984375,
|
| 565 |
+
"logits/rejected": -2.1980721950531006,
|
| 566 |
+
"logps/chosen": -78.40437316894531,
|
| 567 |
+
"logps/rejected": -80.49110412597656,
|
| 568 |
+
"loss": 0.6925,
|
| 569 |
+
"pred_label": 2729.66259765625,
|
| 570 |
+
"rewards/accuracies": 0.3656249940395355,
|
| 571 |
+
"rewards/chosen": -0.028946753591299057,
|
| 572 |
+
"rewards/margins": 0.0717843621969223,
|
| 573 |
+
"rewards/rejected": -0.10073111951351166,
|
| 574 |
"step": 540,
|
| 575 |
+
"use_label": 8272.337890625
|
| 576 |
},
|
| 577 |
{
|
| 578 |
"epoch": 0.59,
|
| 579 |
+
"grad_norm": 0.48828125,
|
| 580 |
"learning_rate": 2.1853970071701415e-06,
|
| 581 |
+
"logits/chosen": -2.19417667388916,
|
| 582 |
+
"logits/rejected": -2.1900599002838135,
|
| 583 |
+
"logps/chosen": -73.69783020019531,
|
| 584 |
+
"logps/rejected": -72.62937927246094,
|
| 585 |
+
"loss": 0.6926,
|
| 586 |
+
"pred_label": 2827.875,
|
| 587 |
"rewards/accuracies": 0.2874999940395355,
|
| 588 |
+
"rewards/chosen": -0.04889845848083496,
|
| 589 |
+
"rewards/margins": 0.04425561800599098,
|
| 590 |
+
"rewards/rejected": -0.09315408021211624,
|
| 591 |
"step": 560,
|
| 592 |
+
"use_label": 8494.125
|
| 593 |
},
|
| 594 |
{
|
| 595 |
"epoch": 0.61,
|
| 596 |
+
"grad_norm": 0.328125,
|
| 597 |
"learning_rate": 2.00499027745888e-06,
|
| 598 |
+
"logits/chosen": -2.224670171737671,
|
| 599 |
+
"logits/rejected": -2.230435371398926,
|
| 600 |
+
"logps/chosen": -76.27436065673828,
|
| 601 |
+
"logps/rejected": -87.6956787109375,
|
| 602 |
+
"loss": 0.6922,
|
| 603 |
+
"pred_label": 2926.862548828125,
|
| 604 |
+
"rewards/accuracies": 0.3687500059604645,
|
| 605 |
+
"rewards/chosen": -0.055185507982969284,
|
| 606 |
+
"rewards/margins": 0.05776001885533333,
|
| 607 |
+
"rewards/rejected": -0.11294553428888321,
|
| 608 |
"step": 580,
|
| 609 |
+
"use_label": 8715.1376953125
|
| 610 |
},
|
| 611 |
{
|
| 612 |
"epoch": 0.63,
|
| 613 |
+
"grad_norm": 0.5546875,
|
| 614 |
"learning_rate": 1.8272307888529276e-06,
|
| 615 |
+
"logits/chosen": -2.231316089630127,
|
| 616 |
+
"logits/rejected": -2.258852481842041,
|
| 617 |
+
"logps/chosen": -84.25640106201172,
|
| 618 |
+
"logps/rejected": -99.73040771484375,
|
| 619 |
+
"loss": 0.6929,
|
| 620 |
+
"pred_label": 3042.83740234375,
|
| 621 |
+
"rewards/accuracies": 0.40312498807907104,
|
| 622 |
+
"rewards/chosen": -0.07248945534229279,
|
| 623 |
+
"rewards/margins": 0.06686891615390778,
|
| 624 |
+
"rewards/rejected": -0.13935837149620056,
|
| 625 |
"step": 600,
|
| 626 |
+
"use_label": 8919.162109375
|
| 627 |
},
|
| 628 |
{
|
| 629 |
"epoch": 0.63,
|
| 630 |
+
"eval_logits/chosen": -2.1407980918884277,
|
| 631 |
+
"eval_logits/rejected": -2.1125032901763916,
|
| 632 |
+
"eval_logps/chosen": -73.41705322265625,
|
| 633 |
+
"eval_logps/rejected": -86.9944839477539,
|
| 634 |
+
"eval_loss": 0.6927017569541931,
|
| 635 |
+
"eval_pred_label": 3177.142822265625,
|
| 636 |
+
"eval_rewards/accuracies": 0.3511904776096344,
|
| 637 |
+
"eval_rewards/chosen": -0.04516514018177986,
|
| 638 |
+
"eval_rewards/margins": 0.07079902291297913,
|
| 639 |
+
"eval_rewards/rejected": -0.11596415936946869,
|
| 640 |
+
"eval_runtime": 248.1359,
|
| 641 |
+
"eval_samples_per_second": 8.06,
|
| 642 |
"eval_steps_per_second": 0.254,
|
| 643 |
+
"eval_use_label": 9198.857421875,
|
| 644 |
"step": 600
|
| 645 |
},
|
| 646 |
{
|
| 647 |
"epoch": 0.65,
|
| 648 |
+
"grad_norm": 0.404296875,
|
| 649 |
"learning_rate": 1.6530691736402317e-06,
|
| 650 |
+
"logits/chosen": -2.1386609077453613,
|
| 651 |
+
"logits/rejected": -2.1743404865264893,
|
| 652 |
+
"logps/chosen": -65.55394744873047,
|
| 653 |
+
"logps/rejected": -88.32081604003906,
|
| 654 |
+
"loss": 0.6924,
|
| 655 |
+
"pred_label": 3318.58740234375,
|
| 656 |
+
"rewards/accuracies": 0.34375,
|
| 657 |
+
"rewards/chosen": -0.05246468633413315,
|
| 658 |
+
"rewards/margins": 0.059620797634124756,
|
| 659 |
+
"rewards/rejected": -0.1120854839682579,
|
| 660 |
"step": 620,
|
| 661 |
+
"use_label": 9467.412109375
|
| 662 |
},
|
| 663 |
{
|
| 664 |
"epoch": 0.67,
|
| 665 |
+
"grad_norm": 0.51171875,
|
| 666 |
"learning_rate": 1.4834368231970922e-06,
|
| 667 |
+
"logits/chosen": -2.1956310272216797,
|
| 668 |
+
"logits/rejected": -2.2024998664855957,
|
| 669 |
+
"logps/chosen": -77.41986846923828,
|
| 670 |
+
"logps/rejected": -82.58815002441406,
|
| 671 |
+
"loss": 0.692,
|
| 672 |
+
"pred_label": 3414.199951171875,
|
| 673 |
+
"rewards/accuracies": 0.3656249940395355,
|
| 674 |
+
"rewards/chosen": -0.05061299726366997,
|
| 675 |
+
"rewards/margins": 0.07674984633922577,
|
| 676 |
+
"rewards/rejected": -0.12736284732818604,
|
| 677 |
"step": 640,
|
| 678 |
+
"use_label": 9691.7998046875
|
| 679 |
},
|
| 680 |
{
|
| 681 |
"epoch": 0.69,
|
| 682 |
+
"grad_norm": 0.2890625,
|
| 683 |
"learning_rate": 1.3192409070404582e-06,
|
| 684 |
+
"logits/chosen": -2.1827545166015625,
|
| 685 |
+
"logits/rejected": -2.1392319202423096,
|
| 686 |
+
"logps/chosen": -71.07948303222656,
|
| 687 |
+
"logps/rejected": -78.78751373291016,
|
| 688 |
+
"loss": 0.6924,
|
| 689 |
+
"pred_label": 3519.35009765625,
|
| 690 |
+
"rewards/accuracies": 0.3375000059604645,
|
| 691 |
+
"rewards/chosen": -0.0542152114212513,
|
| 692 |
+
"rewards/margins": 0.06142013147473335,
|
| 693 |
+
"rewards/rejected": -0.11563535034656525,
|
| 694 |
"step": 660,
|
| 695 |
+
"use_label": 9906.650390625
|
| 696 |
},
|
| 697 |
{
|
| 698 |
"epoch": 0.71,
|
| 699 |
+
"grad_norm": 0.435546875,
|
| 700 |
"learning_rate": 1.1613595214152713e-06,
|
| 701 |
+
"logits/chosen": -2.2185826301574707,
|
| 702 |
+
"logits/rejected": -2.2344555854797363,
|
| 703 |
+
"logps/chosen": -81.96281433105469,
|
| 704 |
+
"logps/rejected": -87.13890838623047,
|
| 705 |
+
"loss": 0.6923,
|
| 706 |
+
"pred_label": 3610.012451171875,
|
| 707 |
+
"rewards/accuracies": 0.34375,
|
| 708 |
+
"rewards/chosen": -0.06083091348409653,
|
| 709 |
+
"rewards/margins": 0.0633452981710434,
|
| 710 |
+
"rewards/rejected": -0.12417621910572052,
|
| 711 |
"step": 680,
|
| 712 |
+
"use_label": 10135.9873046875
|
| 713 |
},
|
| 714 |
{
|
| 715 |
"epoch": 0.73,
|
| 716 |
+
"grad_norm": 0.5078125,
|
| 717 |
"learning_rate": 1.0106369933615043e-06,
|
| 718 |
+
"logits/chosen": -2.2393274307250977,
|
| 719 |
+
"logits/rejected": -2.2085208892822266,
|
| 720 |
+
"logps/chosen": -90.31179809570312,
|
| 721 |
+
"logps/rejected": -96.00973510742188,
|
| 722 |
+
"loss": 0.6928,
|
| 723 |
+
"pred_label": 3716.97509765625,
|
| 724 |
+
"rewards/accuracies": 0.3843750059604645,
|
| 725 |
+
"rewards/chosen": -0.06285654008388519,
|
| 726 |
+
"rewards/margins": 0.07485760748386383,
|
| 727 |
+
"rewards/rejected": -0.13771414756774902,
|
| 728 |
"step": 700,
|
| 729 |
+
"use_label": 10349.025390625
|
| 730 |
},
|
| 731 |
{
|
| 732 |
"epoch": 0.73,
|
| 733 |
+
"eval_logits/chosen": -2.1372170448303223,
|
| 734 |
+
"eval_logits/rejected": -2.1086459159851074,
|
| 735 |
+
"eval_logps/chosen": -73.96572875976562,
|
| 736 |
+
"eval_logps/rejected": -87.70773315429688,
|
| 737 |
+
"eval_loss": 0.6929500102996826,
|
| 738 |
+
"eval_pred_label": 3852.730224609375,
|
| 739 |
"eval_rewards/accuracies": 0.3511904776096344,
|
| 740 |
+
"eval_rewards/chosen": -0.05065184459090233,
|
| 741 |
+
"eval_rewards/margins": 0.07244490087032318,
|
| 742 |
+
"eval_rewards/rejected": -0.12309674173593521,
|
| 743 |
+
"eval_runtime": 248.0038,
|
| 744 |
+
"eval_samples_per_second": 8.064,
|
| 745 |
"eval_steps_per_second": 0.254,
|
| 746 |
+
"eval_use_label": 10627.26953125,
|
| 747 |
"step": 700
|
| 748 |
},
|
| 749 |
{
|
| 750 |
"epoch": 0.75,
|
| 751 |
+
"grad_norm": 0.55078125,
|
| 752 |
"learning_rate": 8.678793653740633e-07,
|
| 753 |
+
"logits/chosen": -2.1876041889190674,
|
| 754 |
+
"logits/rejected": -2.1966712474823,
|
| 755 |
+
"logps/chosen": -64.94602966308594,
|
| 756 |
+
"logps/rejected": -77.46949005126953,
|
| 757 |
+
"loss": 0.6927,
|
| 758 |
+
"pred_label": 3992.16259765625,
|
| 759 |
+
"rewards/accuracies": 0.31562501192092896,
|
| 760 |
+
"rewards/chosen": -0.04975567013025284,
|
| 761 |
+
"rewards/margins": 0.06240048259496689,
|
| 762 |
+
"rewards/rejected": -0.11215615272521973,
|
| 763 |
"step": 720,
|
| 764 |
+
"use_label": 10897.837890625
|
| 765 |
},
|
| 766 |
{
|
| 767 |
"epoch": 0.77,
|
| 768 |
+
"grad_norm": 0.416015625,
|
| 769 |
"learning_rate": 7.338500848029603e-07,
|
| 770 |
+
"logits/chosen": -2.194794178009033,
|
| 771 |
+
"logits/rejected": -2.2083091735839844,
|
| 772 |
+
"logps/chosen": -69.16300201416016,
|
| 773 |
+
"logps/rejected": -74.87442779541016,
|
| 774 |
+
"loss": 0.6927,
|
| 775 |
+
"pred_label": 4088.0625,
|
| 776 |
+
"rewards/accuracies": 0.3187499940395355,
|
| 777 |
+
"rewards/chosen": -0.03673207014799118,
|
| 778 |
+
"rewards/margins": 0.07390830665826797,
|
| 779 |
+
"rewards/rejected": -0.11064038425683975,
|
| 780 |
"step": 740,
|
| 781 |
+
"use_label": 11121.9375
|
| 782 |
},
|
| 783 |
{
|
| 784 |
"epoch": 0.8,
|
| 785 |
+
"grad_norm": 0.47265625,
|
| 786 |
"learning_rate": 6.092659210462232e-07,
|
| 787 |
+
"logits/chosen": -2.2297511100769043,
|
| 788 |
+
"logits/rejected": -2.232818841934204,
|
| 789 |
+
"logps/chosen": -70.27059173583984,
|
| 790 |
+
"logps/rejected": -88.61542510986328,
|
| 791 |
+
"loss": 0.6927,
|
| 792 |
+
"pred_label": 4190.375,
|
| 793 |
"rewards/accuracies": 0.33125001192092896,
|
| 794 |
+
"rewards/chosen": -0.04521505907177925,
|
| 795 |
+
"rewards/margins": 0.05956338718533516,
|
| 796 |
+
"rewards/rejected": -0.10477845370769501,
|
| 797 |
"step": 760,
|
| 798 |
+
"use_label": 11339.625
|
| 799 |
},
|
| 800 |
{
|
| 801 |
"epoch": 0.82,
|
| 802 |
+
"grad_norm": 0.515625,
|
| 803 |
"learning_rate": 4.947931323697983e-07,
|
| 804 |
+
"logits/chosen": -2.224112033843994,
|
| 805 |
+
"logits/rejected": -2.241053581237793,
|
| 806 |
+
"logps/chosen": -82.8070068359375,
|
| 807 |
+
"logps/rejected": -85.62196350097656,
|
| 808 |
+
"loss": 0.6927,
|
| 809 |
+
"pred_label": 4299.97509765625,
|
| 810 |
"rewards/accuracies": 0.375,
|
| 811 |
+
"rewards/chosen": -0.04633576422929764,
|
| 812 |
+
"rewards/margins": 0.0844966396689415,
|
| 813 |
+
"rewards/rejected": -0.13083240389823914,
|
| 814 |
"step": 780,
|
| 815 |
+
"use_label": 11550.025390625
|
| 816 |
},
|
| 817 |
{
|
| 818 |
"epoch": 0.84,
|
| 819 |
+
"grad_norm": 0.498046875,
|
| 820 |
"learning_rate": 3.910439028537638e-07,
|
| 821 |
+
"logits/chosen": -2.201280117034912,
|
| 822 |
+
"logits/rejected": -2.177452325820923,
|
| 823 |
+
"logps/chosen": -65.0578842163086,
|
| 824 |
+
"logps/rejected": -66.19197082519531,
|
| 825 |
+
"loss": 0.6927,
|
| 826 |
+
"pred_label": 4407.78759765625,
|
| 827 |
+
"rewards/accuracies": 0.3499999940395355,
|
| 828 |
+
"rewards/chosen": -0.026169428601861,
|
| 829 |
+
"rewards/margins": 0.06455135345458984,
|
| 830 |
+
"rewards/rejected": -0.0907207801938057,
|
| 831 |
"step": 800,
|
| 832 |
+
"use_label": 11762.212890625
|
| 833 |
},
|
| 834 |
{
|
| 835 |
"epoch": 0.84,
|
| 836 |
+
"eval_logits/chosen": -2.1430623531341553,
|
| 837 |
+
"eval_logits/rejected": -2.1141114234924316,
|
| 838 |
+
"eval_logps/chosen": -71.62469482421875,
|
| 839 |
+
"eval_logps/rejected": -85.3831787109375,
|
| 840 |
+
"eval_loss": 0.6928467750549316,
|
| 841 |
+
"eval_pred_label": 4538.47607421875,
|
| 842 |
+
"eval_rewards/accuracies": 0.3551587164402008,
|
| 843 |
+
"eval_rewards/chosen": -0.027241550385951996,
|
| 844 |
+
"eval_rewards/margins": 0.072609543800354,
|
| 845 |
+
"eval_rewards/rejected": -0.099851094186306,
|
| 846 |
+
"eval_runtime": 247.951,
|
| 847 |
+
"eval_samples_per_second": 8.066,
|
| 848 |
"eval_steps_per_second": 0.254,
|
| 849 |
+
"eval_use_label": 12045.5234375,
|
| 850 |
"step": 800
|
| 851 |
},
|
| 852 |
{
|
| 853 |
"epoch": 0.86,
|
| 854 |
+
"grad_norm": 0.48046875,
|
| 855 |
"learning_rate": 2.98573068519539e-07,
|
| 856 |
+
"logits/chosen": -2.228102684020996,
|
| 857 |
+
"logits/rejected": -2.2112691402435303,
|
| 858 |
+
"logps/chosen": -68.63658142089844,
|
| 859 |
+
"logps/rejected": -75.33064270019531,
|
| 860 |
+
"loss": 0.6923,
|
| 861 |
+
"pred_label": 4678.53759765625,
|
| 862 |
+
"rewards/accuracies": 0.3218750059604645,
|
| 863 |
+
"rewards/chosen": -0.03714119642972946,
|
| 864 |
+
"rewards/margins": 0.05530167371034622,
|
| 865 |
+
"rewards/rejected": -0.09244287014007568,
|
| 866 |
"step": 820,
|
| 867 |
+
"use_label": 12315.462890625
|
| 868 |
},
|
| 869 |
{
|
| 870 |
"epoch": 0.88,
|
| 871 |
+
"grad_norm": 0.56640625,
|
| 872 |
"learning_rate": 2.178751501463036e-07,
|
| 873 |
+
"logits/chosen": -2.204557418823242,
|
| 874 |
+
"logits/rejected": -2.2018847465515137,
|
| 875 |
+
"logps/chosen": -61.4800910949707,
|
| 876 |
+
"logps/rejected": -63.1760139465332,
|
| 877 |
+
"loss": 0.6929,
|
| 878 |
+
"pred_label": 4777.375,
|
| 879 |
+
"rewards/accuracies": 0.29374998807907104,
|
| 880 |
+
"rewards/chosen": -0.02809613011777401,
|
| 881 |
+
"rewards/margins": 0.05226613208651543,
|
| 882 |
+
"rewards/rejected": -0.08036227524280548,
|
| 883 |
"step": 840,
|
| 884 |
+
"use_label": 12536.625
|
| 885 |
},
|
| 886 |
{
|
| 887 |
"epoch": 0.9,
|
| 888 |
+
"grad_norm": 0.8515625,
|
| 889 |
"learning_rate": 1.4938170864468636e-07,
|
| 890 |
+
"logits/chosen": -2.252244234085083,
|
| 891 |
+
"logits/rejected": -2.242299795150757,
|
| 892 |
+
"logps/chosen": -84.9459228515625,
|
| 893 |
+
"logps/rejected": -90.69441223144531,
|
| 894 |
+
"loss": 0.6922,
|
| 895 |
+
"pred_label": 4874.3251953125,
|
| 896 |
+
"rewards/accuracies": 0.4124999940395355,
|
| 897 |
+
"rewards/chosen": -0.022363774478435516,
|
| 898 |
+
"rewards/margins": 0.09057153016328812,
|
| 899 |
+
"rewards/rejected": -0.11293530464172363,
|
| 900 |
"step": 860,
|
| 901 |
+
"use_label": 12759.6748046875
|
| 902 |
},
|
| 903 |
{
|
| 904 |
"epoch": 0.92,
|
| 905 |
+
"grad_norm": 0.4296875,
|
| 906 |
"learning_rate": 9.345903713082305e-08,
|
| 907 |
+
"logits/chosen": -2.2364704608917236,
|
| 908 |
+
"logits/rejected": -2.224773406982422,
|
| 909 |
+
"logps/chosen": -75.7426528930664,
|
| 910 |
+
"logps/rejected": -91.20499420166016,
|
| 911 |
+
"loss": 0.6925,
|
| 912 |
+
"pred_label": 4988.4873046875,
|
| 913 |
+
"rewards/accuracies": 0.3843750059604645,
|
| 914 |
+
"rewards/chosen": -0.03711060434579849,
|
| 915 |
+
"rewards/margins": 0.08993253856897354,
|
| 916 |
+
"rewards/rejected": -0.12704312801361084,
|
| 917 |
"step": 880,
|
| 918 |
+
"use_label": 12965.5126953125
|
| 919 |
},
|
| 920 |
{
|
| 921 |
"epoch": 0.94,
|
| 922 |
+
"grad_norm": 0.6015625,
|
| 923 |
"learning_rate": 5.0406202043228604e-08,
|
| 924 |
+
"logits/chosen": -2.121796131134033,
|
| 925 |
+
"logits/rejected": -2.15610671043396,
|
| 926 |
+
"logps/chosen": -69.87088775634766,
|
| 927 |
+
"logps/rejected": -90.85367584228516,
|
| 928 |
+
"loss": 0.6929,
|
| 929 |
+
"pred_label": 5089.85009765625,
|
| 930 |
"rewards/accuracies": 0.3375000059604645,
|
| 931 |
+
"rewards/chosen": -0.032511431723833084,
|
| 932 |
+
"rewards/margins": 0.06634987145662308,
|
| 933 |
+
"rewards/rejected": -0.09886129945516586,
|
| 934 |
"step": 900,
|
| 935 |
+
"use_label": 13184.150390625
|
| 936 |
},
|
| 937 |
{
|
| 938 |
"epoch": 0.94,
|
| 939 |
+
"eval_logits/chosen": -2.1427581310272217,
|
| 940 |
+
"eval_logits/rejected": -2.113929510116577,
|
| 941 |
+
"eval_logps/chosen": -71.7841567993164,
|
| 942 |
+
"eval_logps/rejected": -85.5160140991211,
|
| 943 |
+
"eval_loss": 0.6928035020828247,
|
| 944 |
+
"eval_pred_label": 5226.619140625,
|
| 945 |
+
"eval_rewards/accuracies": 0.3492063581943512,
|
| 946 |
+
"eval_rewards/chosen": -0.02883605659008026,
|
| 947 |
+
"eval_rewards/margins": 0.0723433569073677,
|
| 948 |
+
"eval_rewards/rejected": -0.10117942094802856,
|
| 949 |
+
"eval_runtime": 246.4796,
|
| 950 |
+
"eval_samples_per_second": 8.114,
|
| 951 |
+
"eval_steps_per_second": 0.256,
|
| 952 |
+
"eval_use_label": 13461.380859375,
|
| 953 |
"step": 900
|
| 954 |
},
|
| 955 |
{
|
| 956 |
"epoch": 0.96,
|
| 957 |
+
"grad_norm": 0.52734375,
|
| 958 |
"learning_rate": 2.0453443778310766e-08,
|
| 959 |
+
"logits/chosen": -2.1679275035858154,
|
| 960 |
+
"logits/rejected": -2.1737468242645264,
|
| 961 |
+
"logps/chosen": -59.419395446777344,
|
| 962 |
+
"logps/rejected": -76.71382141113281,
|
| 963 |
+
"loss": 0.6925,
|
| 964 |
+
"pred_label": 5365.3876953125,
|
| 965 |
+
"rewards/accuracies": 0.3187499940395355,
|
| 966 |
+
"rewards/chosen": -0.021122563630342484,
|
| 967 |
+
"rewards/margins": 0.0707126036286354,
|
| 968 |
+
"rewards/rejected": -0.09183517098426819,
|
| 969 |
"step": 920,
|
| 970 |
+
"use_label": 13732.6123046875
|
| 971 |
},
|
| 972 |
{
|
| 973 |
"epoch": 0.98,
|
| 974 |
+
"grad_norm": 0.64453125,
|
| 975 |
"learning_rate": 3.760945397705828e-09,
|
| 976 |
+
"logits/chosen": -2.1522116661071777,
|
| 977 |
+
"logits/rejected": -2.1893556118011475,
|
| 978 |
+
"logps/chosen": -68.75323486328125,
|
| 979 |
+
"logps/rejected": -82.70423889160156,
|
| 980 |
+
"loss": 0.6926,
|
| 981 |
+
"pred_label": 5459.0751953125,
|
| 982 |
"rewards/accuracies": 0.32499998807907104,
|
| 983 |
+
"rewards/chosen": -0.02701050415635109,
|
| 984 |
+
"rewards/margins": 0.06467042118310928,
|
| 985 |
+
"rewards/rejected": -0.09168092906475067,
|
| 986 |
"step": 940,
|
| 987 |
+
"use_label": 13958.9248046875
|
| 988 |
},
|
| 989 |
{
|
| 990 |
"epoch": 1.0,
|
| 991 |
"step": 955,
|
| 992 |
"total_flos": 0.0,
|
| 993 |
+
"train_loss": 0.692275420283772,
|
| 994 |
+
"train_runtime": 20019.5915,
|
| 995 |
+
"train_samples_per_second": 3.054,
|
| 996 |
"train_steps_per_second": 0.048
|
| 997 |
}
|
| 998 |
],
|