Model save
Browse files- README.md +18 -19
- all_results.json +16 -16
- eval_results.json +12 -12
- train_results.json +5 -5
- trainer_state.json +1896 -506
README.md
CHANGED
|
@@ -1,11 +1,8 @@
|
|
| 1 |
---
|
| 2 |
-
license:
|
| 3 |
-
base_model:
|
| 4 |
tags:
|
| 5 |
-
- alignment-handbook
|
| 6 |
- generated_from_trainer
|
| 7 |
-
datasets:
|
| 8 |
-
- HuggingFaceH4/ultrafeedback_binarized
|
| 9 |
model-index:
|
| 10 |
- name: zephyr-7b-dpo-lora
|
| 11 |
results: []
|
|
@@ -16,17 +13,17 @@ should probably proofread and complete it, then remove this comment. -->
|
|
| 16 |
|
| 17 |
# zephyr-7b-dpo-lora
|
| 18 |
|
| 19 |
-
This model is a fine-tuned version of [
|
| 20 |
It achieves the following results on the evaluation set:
|
| 21 |
-
- Loss: 0.
|
| 22 |
-
- Rewards/chosen: 0.
|
| 23 |
-
- Rewards/rejected: -
|
| 24 |
-
- Rewards/accuracies: 0.
|
| 25 |
-
- Rewards/margins:
|
| 26 |
-
- Logps/rejected: -
|
| 27 |
-
- Logps/chosen: -
|
| 28 |
-
- Logits/rejected: -
|
| 29 |
-
- Logits/chosen: -
|
| 30 |
|
| 31 |
## Model description
|
| 32 |
|
|
@@ -45,7 +42,7 @@ More information needed
|
|
| 45 |
### Training hyperparameters
|
| 46 |
|
| 47 |
The following hyperparameters were used during training:
|
| 48 |
-
- learning_rate:
|
| 49 |
- train_batch_size: 16
|
| 50 |
- eval_batch_size: 16
|
| 51 |
- seed: 42
|
|
@@ -56,18 +53,20 @@ The following hyperparameters were used during training:
|
|
| 56 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 57 |
- lr_scheduler_type: linear
|
| 58 |
- lr_scheduler_warmup_ratio: 0.1
|
| 59 |
-
- num_epochs:
|
| 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.
|
|
|
|
|
|
|
| 66 |
|
| 67 |
|
| 68 |
### Framework versions
|
| 69 |
|
| 70 |
- Transformers 4.35.0
|
| 71 |
-
- Pytorch 2.1.
|
| 72 |
- Datasets 2.14.6
|
| 73 |
- Tokenizers 0.14.1
|
|
|
|
| 1 |
---
|
| 2 |
+
license: mit
|
| 3 |
+
base_model: HuggingFaceH4/mistral-7b-sft-beta
|
| 4 |
tags:
|
|
|
|
| 5 |
- generated_from_trainer
|
|
|
|
|
|
|
| 6 |
model-index:
|
| 7 |
- name: zephyr-7b-dpo-lora
|
| 8 |
results: []
|
|
|
|
| 13 |
|
| 14 |
# zephyr-7b-dpo-lora
|
| 15 |
|
| 16 |
+
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.
|
| 17 |
It achieves the following results on the evaluation set:
|
| 18 |
+
- Loss: 0.4553
|
| 19 |
+
- Rewards/chosen: -0.5876
|
| 20 |
+
- Rewards/rejected: -2.1911
|
| 21 |
+
- Rewards/accuracies: 0.8359
|
| 22 |
+
- Rewards/margins: 1.6035
|
| 23 |
+
- Logps/rejected: -246.6992
|
| 24 |
+
- Logps/chosen: -279.5245
|
| 25 |
+
- Logits/rejected: -2.8331
|
| 26 |
+
- Logits/chosen: -2.8422
|
| 27 |
|
| 28 |
## Model description
|
| 29 |
|
|
|
|
| 42 |
### Training hyperparameters
|
| 43 |
|
| 44 |
The following hyperparameters were used during training:
|
| 45 |
+
- learning_rate: 2e-05
|
| 46 |
- train_batch_size: 16
|
| 47 |
- eval_batch_size: 16
|
| 48 |
- seed: 42
|
|
|
|
| 53 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 54 |
- lr_scheduler_type: linear
|
| 55 |
- lr_scheduler_warmup_ratio: 0.1
|
| 56 |
+
- num_epochs: 3.0
|
| 57 |
|
| 58 |
### Training results
|
| 59 |
|
| 60 |
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 61 |
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 62 |
+
| 0.4771 | 1.0 | 485 | 0.4617 | -0.0843 | -1.3520 | 0.7891 | 1.2677 | -238.3082 | -274.4911 | -2.8501 | -2.8796 |
|
| 63 |
+
| 0.4124 | 2.0 | 970 | 0.4545 | -0.3551 | -1.7590 | 0.8164 | 1.4038 | -242.3781 | -277.1996 | -2.8563 | -2.8659 |
|
| 64 |
+
| 0.3549 | 3.0 | 1455 | 0.4553 | -0.5876 | -2.1911 | 0.8359 | 1.6035 | -246.6992 | -279.5245 | -2.8331 | -2.8422 |
|
| 65 |
|
| 66 |
|
| 67 |
### Framework versions
|
| 68 |
|
| 69 |
- Transformers 4.35.0
|
| 70 |
+
- Pytorch 2.1.0+cu121
|
| 71 |
- Datasets 2.14.6
|
| 72 |
- Tokenizers 0.14.1
|
all_results.json
CHANGED
|
@@ -1,21 +1,21 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch":
|
| 3 |
-
"eval_logits/chosen": -
|
| 4 |
-
"eval_logits/rejected": -
|
| 5 |
-
"eval_logps/chosen": -
|
| 6 |
-
"eval_logps/rejected": -
|
| 7 |
-
"eval_loss": 0.
|
| 8 |
-
"eval_rewards/accuracies": 0.
|
| 9 |
-
"eval_rewards/chosen": 0.
|
| 10 |
-
"eval_rewards/margins":
|
| 11 |
-
"eval_rewards/rejected": -
|
| 12 |
-
"eval_runtime":
|
| 13 |
"eval_samples": 2000,
|
| 14 |
-
"eval_samples_per_second": 7.
|
| 15 |
"eval_steps_per_second": 0.063,
|
| 16 |
-
"train_loss": 0.
|
| 17 |
-
"train_runtime":
|
| 18 |
-
"train_samples":
|
| 19 |
-
"train_samples_per_second":
|
| 20 |
"train_steps_per_second": 0.031
|
| 21 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"eval_logits/chosen": -2.842160701751709,
|
| 4 |
+
"eval_logits/rejected": -2.833141326904297,
|
| 5 |
+
"eval_logps/chosen": -279.5245056152344,
|
| 6 |
+
"eval_logps/rejected": -246.69915771484375,
|
| 7 |
+
"eval_loss": 0.45531293749809265,
|
| 8 |
+
"eval_rewards/accuracies": 0.8359375,
|
| 9 |
+
"eval_rewards/chosen": -0.5876308083534241,
|
| 10 |
+
"eval_rewards/margins": 1.6034575700759888,
|
| 11 |
+
"eval_rewards/rejected": -2.1910881996154785,
|
| 12 |
+
"eval_runtime": 252.3832,
|
| 13 |
"eval_samples": 2000,
|
| 14 |
+
"eval_samples_per_second": 7.924,
|
| 15 |
"eval_steps_per_second": 0.063,
|
| 16 |
+
"train_loss": 0.43281792414557074,
|
| 17 |
+
"train_runtime": 46468.4841,
|
| 18 |
+
"train_samples": 61966,
|
| 19 |
+
"train_samples_per_second": 4.001,
|
| 20 |
"train_steps_per_second": 0.031
|
| 21 |
}
|
eval_results.json
CHANGED
|
@@ -1,16 +1,16 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch":
|
| 3 |
-
"eval_logits/chosen": -
|
| 4 |
-
"eval_logits/rejected": -
|
| 5 |
-
"eval_logps/chosen": -
|
| 6 |
-
"eval_logps/rejected": -
|
| 7 |
-
"eval_loss": 0.
|
| 8 |
-
"eval_rewards/accuracies": 0.
|
| 9 |
-
"eval_rewards/chosen": 0.
|
| 10 |
-
"eval_rewards/margins":
|
| 11 |
-
"eval_rewards/rejected": -
|
| 12 |
-
"eval_runtime":
|
| 13 |
"eval_samples": 2000,
|
| 14 |
-
"eval_samples_per_second": 7.
|
| 15 |
"eval_steps_per_second": 0.063
|
| 16 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"eval_logits/chosen": -2.842160701751709,
|
| 4 |
+
"eval_logits/rejected": -2.833141326904297,
|
| 5 |
+
"eval_logps/chosen": -279.5245056152344,
|
| 6 |
+
"eval_logps/rejected": -246.69915771484375,
|
| 7 |
+
"eval_loss": 0.45531293749809265,
|
| 8 |
+
"eval_rewards/accuracies": 0.8359375,
|
| 9 |
+
"eval_rewards/chosen": -0.5876308083534241,
|
| 10 |
+
"eval_rewards/margins": 1.6034575700759888,
|
| 11 |
+
"eval_rewards/rejected": -2.1910881996154785,
|
| 12 |
+
"eval_runtime": 252.3832,
|
| 13 |
"eval_samples": 2000,
|
| 14 |
+
"eval_samples_per_second": 7.924,
|
| 15 |
"eval_steps_per_second": 0.063
|
| 16 |
}
|
train_results.json
CHANGED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch":
|
| 3 |
-
"train_loss": 0.
|
| 4 |
-
"train_runtime":
|
| 5 |
-
"train_samples":
|
| 6 |
-
"train_samples_per_second":
|
| 7 |
"train_steps_per_second": 0.031
|
| 8 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"train_loss": 0.43281792414557074,
|
| 4 |
+
"train_runtime": 46468.4841,
|
| 5 |
+
"train_samples": 61966,
|
| 6 |
+
"train_samples_per_second": 4.001,
|
| 7 |
"train_steps_per_second": 0.031
|
| 8 |
}
|
trainer_state.json
CHANGED
|
@@ -1,20 +1,20 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": null,
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
-
"epoch":
|
| 5 |
"eval_steps": 100,
|
| 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": -
|
| 15 |
-
"logits/rejected": -
|
| 16 |
-
"logps/chosen": -
|
| 17 |
-
"logps/rejected": -
|
| 18 |
"loss": 0.6931,
|
| 19 |
"rewards/accuracies": 0.0,
|
| 20 |
"rewards/chosen": 0.0,
|
|
@@ -24,705 +24,2095 @@
|
|
| 24 |
},
|
| 25 |
{
|
| 26 |
"epoch": 0.02,
|
| 27 |
-
"learning_rate": 1.
|
| 28 |
-
"logits/chosen": -
|
| 29 |
-
"logits/rejected": -
|
| 30 |
-
"logps/chosen": -
|
| 31 |
-
"logps/rejected": -
|
| 32 |
-
"loss": 0.
|
| 33 |
-
"rewards/accuracies": 0.
|
| 34 |
-
"rewards/chosen": -0.
|
| 35 |
-
"rewards/margins": -0.
|
| 36 |
-
"rewards/rejected": -0.
|
| 37 |
"step": 10
|
| 38 |
},
|
| 39 |
{
|
| 40 |
"epoch": 0.04,
|
| 41 |
-
"learning_rate": 2.
|
| 42 |
-
"logits/chosen": -2.
|
| 43 |
-
"logits/rejected": -2.
|
| 44 |
-
"logps/chosen": -
|
| 45 |
-
"logps/rejected": -
|
| 46 |
-
"loss": 0.
|
| 47 |
-
"rewards/accuracies": 0.
|
| 48 |
-
"rewards/chosen": 0.
|
| 49 |
-
"rewards/margins": 0.
|
| 50 |
-
"rewards/rejected": -0.
|
| 51 |
"step": 20
|
| 52 |
},
|
| 53 |
{
|
| 54 |
"epoch": 0.06,
|
| 55 |
-
"learning_rate":
|
| 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": 0.08,
|
| 69 |
-
"learning_rate":
|
| 70 |
-
"logits/chosen": -
|
| 71 |
-
"logits/rejected": -
|
| 72 |
-
"logps/chosen": -
|
| 73 |
-
"logps/rejected": -
|
| 74 |
-
"loss": 0.
|
| 75 |
-
"rewards/accuracies": 0.
|
| 76 |
-
"rewards/chosen": 0.
|
| 77 |
-
"rewards/margins": 0.
|
| 78 |
-
"rewards/rejected": -0.
|
| 79 |
"step": 40
|
| 80 |
},
|
| 81 |
{
|
| 82 |
"epoch": 0.1,
|
| 83 |
-
"learning_rate":
|
| 84 |
-
"logits/chosen": -
|
| 85 |
-
"logits/rejected": -
|
| 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": 0.12,
|
| 97 |
-
"learning_rate":
|
| 98 |
-
"logits/chosen": -
|
| 99 |
-
"logits/rejected": -
|
| 100 |
-
"logps/chosen": -
|
| 101 |
-
"logps/rejected": -
|
| 102 |
-
"loss": 0.
|
| 103 |
-
"rewards/accuracies": 0.
|
| 104 |
-
"rewards/chosen": 0.
|
| 105 |
-
"rewards/margins": 0.
|
| 106 |
-
"rewards/rejected": -0.
|
| 107 |
"step": 60
|
| 108 |
},
|
| 109 |
{
|
| 110 |
"epoch": 0.14,
|
| 111 |
-
"learning_rate":
|
| 112 |
-
"logits/chosen": -
|
| 113 |
-
"logits/rejected": -
|
| 114 |
-
"logps/chosen": -
|
| 115 |
-
"logps/rejected": -
|
| 116 |
-
"loss": 0.
|
| 117 |
-
"rewards/accuracies": 0.
|
| 118 |
-
"rewards/chosen": 0.
|
| 119 |
-
"rewards/margins": 0.
|
| 120 |
-
"rewards/rejected": -0.
|
| 121 |
"step": 70
|
| 122 |
},
|
| 123 |
{
|
| 124 |
"epoch": 0.16,
|
| 125 |
-
"learning_rate":
|
| 126 |
-
"logits/chosen": -
|
| 127 |
-
"logits/rejected": -2.
|
| 128 |
-
"logps/chosen": -
|
| 129 |
-
"logps/rejected": -
|
| 130 |
-
"loss": 0.
|
| 131 |
-
"rewards/accuracies": 0.
|
| 132 |
-
"rewards/chosen": 0.
|
| 133 |
-
"rewards/margins": 0.
|
| 134 |
-
"rewards/rejected": -0.
|
| 135 |
"step": 80
|
| 136 |
},
|
| 137 |
{
|
| 138 |
"epoch": 0.19,
|
| 139 |
-
"learning_rate":
|
| 140 |
-
"logits/chosen": -
|
| 141 |
-
"logits/rejected": -
|
| 142 |
-
"logps/chosen": -
|
| 143 |
-
"logps/rejected": -
|
| 144 |
-
"loss": 0.
|
| 145 |
-
"rewards/accuracies": 0.
|
| 146 |
-
"rewards/chosen": 0.
|
| 147 |
-
"rewards/margins": 0.
|
| 148 |
-
"rewards/rejected": -0.
|
| 149 |
"step": 90
|
| 150 |
},
|
| 151 |
{
|
| 152 |
"epoch": 0.21,
|
| 153 |
-
"learning_rate":
|
| 154 |
-
"logits/chosen": -
|
| 155 |
-
"logits/rejected": -
|
| 156 |
-
"logps/chosen": -
|
| 157 |
-
"logps/rejected": -
|
| 158 |
-
"loss": 0.
|
| 159 |
-
"rewards/accuracies": 0.
|
| 160 |
-
"rewards/chosen": 0.
|
| 161 |
-
"rewards/margins": 0.
|
| 162 |
-
"rewards/rejected": -0.
|
| 163 |
"step": 100
|
| 164 |
},
|
| 165 |
{
|
| 166 |
"epoch": 0.23,
|
| 167 |
-
"learning_rate":
|
| 168 |
-
"logits/chosen": -
|
| 169 |
-
"logits/rejected": -2.
|
| 170 |
-
"logps/chosen": -
|
| 171 |
-
"logps/rejected": -
|
| 172 |
-
"loss": 0.
|
| 173 |
-
"rewards/accuracies": 0.
|
| 174 |
-
"rewards/chosen": 0.
|
| 175 |
-
"rewards/margins": 0.
|
| 176 |
-
"rewards/rejected": -
|
| 177 |
"step": 110
|
| 178 |
},
|
| 179 |
{
|
| 180 |
"epoch": 0.25,
|
| 181 |
-
"learning_rate":
|
| 182 |
-
"logits/chosen": -
|
| 183 |
-
"logits/rejected": -
|
| 184 |
-
"logps/chosen": -
|
| 185 |
-
"logps/rejected": -
|
| 186 |
-
"loss": 0.
|
| 187 |
-
"rewards/accuracies": 0.
|
| 188 |
-
"rewards/chosen": 0.
|
| 189 |
-
"rewards/margins": 0.
|
| 190 |
-
"rewards/rejected": -
|
| 191 |
"step": 120
|
| 192 |
},
|
| 193 |
{
|
| 194 |
"epoch": 0.27,
|
| 195 |
-
"learning_rate":
|
| 196 |
-
"logits/chosen": -2.
|
| 197 |
-
"logits/rejected": -2.
|
| 198 |
-
"logps/chosen": -
|
| 199 |
-
"logps/rejected": -
|
| 200 |
-
"loss": 0.
|
| 201 |
-
"rewards/accuracies": 0.
|
| 202 |
-
"rewards/chosen": 0.
|
| 203 |
-
"rewards/margins": 0.
|
| 204 |
-
"rewards/rejected": -
|
| 205 |
"step": 130
|
| 206 |
},
|
| 207 |
{
|
| 208 |
"epoch": 0.29,
|
| 209 |
-
"learning_rate":
|
| 210 |
-
"logits/chosen": -
|
| 211 |
-
"logits/rejected": -
|
| 212 |
-
"logps/chosen": -291.
|
| 213 |
-
"logps/rejected": -
|
| 214 |
-
"loss": 0.
|
| 215 |
-
"rewards/accuracies": 0.
|
| 216 |
-
"rewards/chosen": 0.
|
| 217 |
-
"rewards/margins": 0.
|
| 218 |
-
"rewards/rejected": -0.
|
| 219 |
"step": 140
|
| 220 |
},
|
| 221 |
{
|
| 222 |
"epoch": 0.31,
|
| 223 |
-
"learning_rate":
|
| 224 |
-
"logits/chosen": -
|
| 225 |
-
"logits/rejected": -
|
| 226 |
-
"logps/chosen": -
|
| 227 |
-
"logps/rejected": -
|
| 228 |
-
"loss": 0.
|
| 229 |
-
"rewards/accuracies": 0.
|
| 230 |
-
"rewards/chosen": 0.
|
| 231 |
-
"rewards/margins": 0.
|
| 232 |
-
"rewards/rejected": -0.
|
| 233 |
"step": 150
|
| 234 |
},
|
| 235 |
{
|
| 236 |
"epoch": 0.33,
|
| 237 |
-
"learning_rate":
|
| 238 |
-
"logits/chosen": -
|
| 239 |
-
"logits/rejected": -
|
| 240 |
-
"logps/chosen": -
|
| 241 |
-
"logps/rejected": -
|
| 242 |
-
"loss": 0.
|
| 243 |
-
"rewards/accuracies": 0.
|
| 244 |
-
"rewards/chosen": 0.
|
| 245 |
-
"rewards/margins": 0.
|
| 246 |
-
"rewards/rejected": -
|
| 247 |
"step": 160
|
| 248 |
},
|
| 249 |
{
|
| 250 |
"epoch": 0.35,
|
| 251 |
-
"learning_rate":
|
| 252 |
-
"logits/chosen": -
|
| 253 |
-
"logits/rejected": -
|
| 254 |
-
"logps/chosen": -
|
| 255 |
-
"logps/rejected": -
|
| 256 |
-
"loss": 0.
|
| 257 |
-
"rewards/accuracies": 0.
|
| 258 |
-
"rewards/chosen": 0.
|
| 259 |
-
"rewards/margins": 0.
|
| 260 |
-
"rewards/rejected": -0.
|
| 261 |
"step": 170
|
| 262 |
},
|
| 263 |
{
|
| 264 |
"epoch": 0.37,
|
| 265 |
-
"learning_rate":
|
| 266 |
-
"logits/chosen": -
|
| 267 |
-
"logits/rejected": -
|
| 268 |
-
"logps/chosen": -
|
| 269 |
-
"logps/rejected": -
|
| 270 |
-
"loss": 0.
|
| 271 |
-
"rewards/accuracies": 0.
|
| 272 |
-
"rewards/chosen": 0.
|
| 273 |
-
"rewards/margins": 0.
|
| 274 |
-
"rewards/rejected": -0.
|
| 275 |
"step": 180
|
| 276 |
},
|
| 277 |
{
|
| 278 |
"epoch": 0.39,
|
| 279 |
-
"learning_rate":
|
| 280 |
-
"logits/chosen": -
|
| 281 |
-
"logits/rejected": -
|
| 282 |
-
"logps/chosen": -
|
| 283 |
-
"logps/rejected": -
|
| 284 |
-
"loss": 0.
|
| 285 |
-
"rewards/accuracies": 0.
|
| 286 |
-
"rewards/chosen": 0.
|
| 287 |
-
"rewards/margins":
|
| 288 |
-
"rewards/rejected": -
|
| 289 |
"step": 190
|
| 290 |
},
|
| 291 |
{
|
| 292 |
"epoch": 0.41,
|
| 293 |
-
"learning_rate":
|
| 294 |
-
"logits/chosen": -
|
| 295 |
-
"logits/rejected": -
|
| 296 |
-
"logps/chosen": -
|
| 297 |
-
"logps/rejected": -
|
| 298 |
-
"loss": 0.
|
| 299 |
-
"rewards/accuracies": 0.
|
| 300 |
-
"rewards/chosen": 0.
|
| 301 |
-
"rewards/margins": 0.
|
| 302 |
-
"rewards/rejected": -
|
| 303 |
"step": 200
|
| 304 |
},
|
| 305 |
{
|
| 306 |
"epoch": 0.43,
|
| 307 |
-
"learning_rate":
|
| 308 |
-
"logits/chosen": -
|
| 309 |
-
"logits/rejected": -
|
| 310 |
-
"logps/chosen": -
|
| 311 |
-
"logps/rejected": -
|
| 312 |
-
"loss": 0.
|
| 313 |
-
"rewards/accuracies": 0.
|
| 314 |
-
"rewards/chosen": 0.
|
| 315 |
-
"rewards/margins":
|
| 316 |
-
"rewards/rejected": -
|
| 317 |
"step": 210
|
| 318 |
},
|
| 319 |
{
|
| 320 |
"epoch": 0.45,
|
| 321 |
-
"learning_rate":
|
| 322 |
-
"logits/chosen": -
|
| 323 |
-
"logits/rejected": -
|
| 324 |
-
"logps/chosen": -
|
| 325 |
-
"logps/rejected": -
|
| 326 |
-
"loss": 0.
|
| 327 |
-
"rewards/accuracies": 0.
|
| 328 |
-
"rewards/chosen": 0.
|
| 329 |
-
"rewards/margins": 0.
|
| 330 |
-
"rewards/rejected": -
|
| 331 |
"step": 220
|
| 332 |
},
|
| 333 |
{
|
| 334 |
"epoch": 0.47,
|
| 335 |
-
"learning_rate":
|
| 336 |
-
"logits/chosen": -2.
|
| 337 |
-
"logits/rejected": -2.
|
| 338 |
-
"logps/chosen": -
|
| 339 |
-
"logps/rejected": -
|
| 340 |
-
"loss": 0.
|
| 341 |
-
"rewards/accuracies": 0.
|
| 342 |
-
"rewards/chosen": 0.
|
| 343 |
-
"rewards/margins": 0.
|
| 344 |
-
"rewards/rejected": -
|
| 345 |
"step": 230
|
| 346 |
},
|
| 347 |
{
|
| 348 |
"epoch": 0.49,
|
| 349 |
-
"learning_rate":
|
| 350 |
-
"logits/chosen": -
|
| 351 |
-
"logits/rejected": -
|
| 352 |
-
"logps/chosen": -
|
| 353 |
-
"logps/rejected": -
|
| 354 |
-
"loss": 0.
|
| 355 |
-
"rewards/accuracies": 0.
|
| 356 |
-
"rewards/chosen": 0.
|
| 357 |
-
"rewards/margins":
|
| 358 |
-
"rewards/rejected": -
|
| 359 |
"step": 240
|
| 360 |
},
|
| 361 |
{
|
| 362 |
"epoch": 0.52,
|
| 363 |
-
"learning_rate":
|
| 364 |
-
"logits/chosen": -
|
| 365 |
-
"logits/rejected": -
|
| 366 |
-
"logps/chosen": -
|
| 367 |
-
"logps/rejected": -
|
| 368 |
-
"loss": 0.
|
| 369 |
-
"rewards/accuracies": 0.
|
| 370 |
-
"rewards/chosen": 0.
|
| 371 |
-
"rewards/margins": 0.
|
| 372 |
-
"rewards/rejected": -
|
| 373 |
"step": 250
|
| 374 |
},
|
| 375 |
{
|
| 376 |
"epoch": 0.54,
|
| 377 |
-
"learning_rate":
|
| 378 |
-
"logits/chosen": -2.
|
| 379 |
-
"logits/rejected": -2.
|
| 380 |
-
"logps/chosen": -
|
| 381 |
-
"logps/rejected": -
|
| 382 |
-
"loss": 0.
|
| 383 |
-
"rewards/accuracies": 0.
|
| 384 |
-
"rewards/chosen": 0.
|
| 385 |
-
"rewards/margins":
|
| 386 |
-
"rewards/rejected": -
|
| 387 |
"step": 260
|
| 388 |
},
|
| 389 |
{
|
| 390 |
"epoch": 0.56,
|
| 391 |
-
"learning_rate":
|
| 392 |
-
"logits/chosen": -
|
| 393 |
-
"logits/rejected": -
|
| 394 |
-
"logps/chosen": -
|
| 395 |
-
"logps/rejected": -
|
| 396 |
-
"loss": 0.
|
| 397 |
-
"rewards/accuracies": 0.
|
| 398 |
-
"rewards/chosen": 0.
|
| 399 |
-
"rewards/margins":
|
| 400 |
-
"rewards/rejected": -
|
| 401 |
"step": 270
|
| 402 |
},
|
| 403 |
{
|
| 404 |
"epoch": 0.58,
|
| 405 |
-
"learning_rate":
|
| 406 |
-
"logits/chosen": -
|
| 407 |
-
"logits/rejected": -2.
|
| 408 |
-
"logps/chosen": -
|
| 409 |
-
"logps/rejected": -
|
| 410 |
-
"loss": 0.
|
| 411 |
-
"rewards/accuracies": 0.
|
| 412 |
-
"rewards/chosen": 0.
|
| 413 |
-
"rewards/margins": 0.
|
| 414 |
-
"rewards/rejected": -
|
| 415 |
"step": 280
|
| 416 |
},
|
| 417 |
{
|
| 418 |
"epoch": 0.6,
|
| 419 |
-
"learning_rate":
|
| 420 |
-
"logits/chosen": -
|
| 421 |
-
"logits/rejected": -
|
| 422 |
-
"logps/chosen": -
|
| 423 |
-
"logps/rejected": -
|
| 424 |
-
"loss": 0.
|
| 425 |
-
"rewards/accuracies": 0.
|
| 426 |
-
"rewards/chosen": 0.
|
| 427 |
-
"rewards/margins":
|
| 428 |
-
"rewards/rejected": -
|
| 429 |
"step": 290
|
| 430 |
},
|
| 431 |
{
|
| 432 |
"epoch": 0.62,
|
| 433 |
-
"learning_rate":
|
| 434 |
-
"logits/chosen": -2.
|
| 435 |
-
"logits/rejected": -2.
|
| 436 |
-
"logps/chosen": -
|
| 437 |
-
"logps/rejected": -
|
| 438 |
-
"loss": 0.
|
| 439 |
-
"rewards/accuracies": 0.
|
| 440 |
-
"rewards/chosen": 0.
|
| 441 |
-
"rewards/margins": 0.
|
| 442 |
-
"rewards/rejected": -
|
| 443 |
"step": 300
|
| 444 |
},
|
| 445 |
{
|
| 446 |
"epoch": 0.64,
|
| 447 |
-
"learning_rate":
|
| 448 |
-
"logits/chosen": -
|
| 449 |
-
"logits/rejected": -
|
| 450 |
-
"logps/chosen": -
|
| 451 |
-
"logps/rejected": -
|
| 452 |
-
"loss": 0.
|
| 453 |
-
"rewards/accuracies": 0.
|
| 454 |
-
"rewards/chosen": 0.
|
| 455 |
-
"rewards/margins":
|
| 456 |
-
"rewards/rejected": -
|
| 457 |
"step": 310
|
| 458 |
},
|
| 459 |
{
|
| 460 |
"epoch": 0.66,
|
| 461 |
-
"learning_rate": 1.
|
| 462 |
-
"logits/chosen": -
|
| 463 |
-
"logits/rejected": -
|
| 464 |
-
"logps/chosen": -
|
| 465 |
-
"logps/rejected": -
|
| 466 |
-
"loss": 0.
|
| 467 |
-
"rewards/accuracies": 0.
|
| 468 |
-
"rewards/chosen": 0.
|
| 469 |
-
"rewards/margins":
|
| 470 |
-
"rewards/rejected": -
|
| 471 |
"step": 320
|
| 472 |
},
|
| 473 |
{
|
| 474 |
"epoch": 0.68,
|
| 475 |
-
"learning_rate": 1.
|
| 476 |
-
"logits/chosen": -
|
| 477 |
-
"logits/rejected": -
|
| 478 |
-
"logps/chosen": -
|
| 479 |
-
"logps/rejected": -
|
| 480 |
-
"loss": 0.
|
| 481 |
-
"rewards/accuracies": 0.
|
| 482 |
-
"rewards/chosen": 0.
|
| 483 |
-
"rewards/margins":
|
| 484 |
-
"rewards/rejected": -
|
| 485 |
"step": 330
|
| 486 |
},
|
| 487 |
{
|
| 488 |
"epoch": 0.7,
|
| 489 |
-
"learning_rate": 1.
|
| 490 |
-
"logits/chosen": -
|
| 491 |
-
"logits/rejected": -
|
| 492 |
-
"logps/chosen": -
|
| 493 |
-
"logps/rejected": -
|
| 494 |
-
"loss": 0.
|
| 495 |
-
"rewards/accuracies": 0.
|
| 496 |
-
"rewards/chosen": 0.
|
| 497 |
-
"rewards/margins": 0.
|
| 498 |
-
"rewards/rejected": -
|
| 499 |
"step": 340
|
| 500 |
},
|
| 501 |
{
|
| 502 |
"epoch": 0.72,
|
| 503 |
-
"learning_rate": 1.
|
| 504 |
-
"logits/chosen": -
|
| 505 |
-
"logits/rejected": -
|
| 506 |
-
"logps/chosen": -
|
| 507 |
-
"logps/rejected": -
|
| 508 |
-
"loss": 0.
|
| 509 |
-
"rewards/accuracies": 0.
|
| 510 |
-
"rewards/chosen": 0.
|
| 511 |
-
"rewards/margins": 0.
|
| 512 |
-
"rewards/rejected": -
|
| 513 |
"step": 350
|
| 514 |
},
|
| 515 |
{
|
| 516 |
"epoch": 0.74,
|
| 517 |
-
"learning_rate": 1.
|
| 518 |
-
"logits/chosen": -
|
| 519 |
-
"logits/rejected": -
|
| 520 |
-
"logps/chosen": -
|
| 521 |
-
"logps/rejected": -
|
| 522 |
-
"loss": 0.
|
| 523 |
-
"rewards/accuracies": 0.
|
| 524 |
-
"rewards/chosen": 0.
|
| 525 |
-
"rewards/margins": 0.
|
| 526 |
-
"rewards/rejected": -
|
| 527 |
"step": 360
|
| 528 |
},
|
| 529 |
{
|
| 530 |
"epoch": 0.76,
|
| 531 |
-
"learning_rate": 1.
|
| 532 |
-
"logits/chosen": -2.
|
| 533 |
-
"logits/rejected": -2.
|
| 534 |
-
"logps/chosen": -
|
| 535 |
-
"logps/rejected": -
|
| 536 |
-
"loss": 0.
|
| 537 |
-
"rewards/accuracies": 0.
|
| 538 |
-
"rewards/chosen": 0.
|
| 539 |
-
"rewards/margins": 0.
|
| 540 |
-
"rewards/rejected": -
|
| 541 |
"step": 370
|
| 542 |
},
|
| 543 |
{
|
| 544 |
"epoch": 0.78,
|
| 545 |
-
"learning_rate": 1.
|
| 546 |
-
"logits/chosen": -
|
| 547 |
-
"logits/rejected": -
|
| 548 |
-
"logps/chosen": -
|
| 549 |
-
"logps/rejected": -
|
| 550 |
-
"loss": 0.
|
| 551 |
-
"rewards/accuracies": 0.
|
| 552 |
-
"rewards/chosen": 0.
|
| 553 |
-
"rewards/margins":
|
| 554 |
-
"rewards/rejected": -
|
| 555 |
"step": 380
|
| 556 |
},
|
| 557 |
{
|
| 558 |
"epoch": 0.8,
|
| 559 |
-
"learning_rate": 1.
|
| 560 |
-
"logits/chosen": -
|
| 561 |
-
"logits/rejected": -
|
| 562 |
-
"logps/chosen": -
|
| 563 |
-
"logps/rejected": -
|
| 564 |
-
"loss": 0.
|
| 565 |
-
"rewards/accuracies": 0.
|
| 566 |
-
"rewards/chosen": 0.
|
| 567 |
-
"rewards/margins":
|
| 568 |
-
"rewards/rejected": -
|
| 569 |
"step": 390
|
| 570 |
},
|
| 571 |
{
|
| 572 |
"epoch": 0.82,
|
| 573 |
-
"learning_rate":
|
| 574 |
-
"logits/chosen": -
|
| 575 |
-
"logits/rejected": -
|
| 576 |
-
"logps/chosen": -
|
| 577 |
-
"logps/rejected": -
|
| 578 |
-
"loss": 0.
|
| 579 |
-
"rewards/accuracies": 0.
|
| 580 |
-
"rewards/chosen": 0.
|
| 581 |
-
"rewards/margins":
|
| 582 |
-
"rewards/rejected": -
|
| 583 |
"step": 400
|
| 584 |
},
|
| 585 |
{
|
| 586 |
"epoch": 0.85,
|
| 587 |
-
"learning_rate":
|
| 588 |
-
"logits/chosen": -
|
| 589 |
-
"logits/rejected": -
|
| 590 |
-
"logps/chosen": -
|
| 591 |
-
"logps/rejected": -
|
| 592 |
-
"loss": 0.
|
| 593 |
-
"rewards/accuracies": 0.
|
| 594 |
-
"rewards/chosen": 0.
|
| 595 |
-
"rewards/margins":
|
| 596 |
-
"rewards/rejected": -
|
| 597 |
"step": 410
|
| 598 |
},
|
| 599 |
{
|
| 600 |
"epoch": 0.87,
|
| 601 |
-
"learning_rate":
|
| 602 |
-
"logits/chosen": -
|
| 603 |
-
"logits/rejected": -
|
| 604 |
-
"logps/chosen": -
|
| 605 |
-
"logps/rejected": -
|
| 606 |
-
"loss": 0.
|
| 607 |
-
"rewards/accuracies": 0.
|
| 608 |
-
"rewards/chosen": 0.
|
| 609 |
-
"rewards/margins":
|
| 610 |
-
"rewards/rejected": -
|
| 611 |
"step": 420
|
| 612 |
},
|
| 613 |
{
|
| 614 |
"epoch": 0.89,
|
| 615 |
-
"learning_rate":
|
| 616 |
-
"logits/chosen": -
|
| 617 |
-
"logits/rejected": -
|
| 618 |
-
"logps/chosen": -
|
| 619 |
-
"logps/rejected": -
|
| 620 |
-
"loss": 0.
|
| 621 |
-
"rewards/accuracies": 0.
|
| 622 |
-
"rewards/chosen": 0.
|
| 623 |
-
"rewards/margins": 0.
|
| 624 |
-
"rewards/rejected": -
|
| 625 |
"step": 430
|
| 626 |
},
|
| 627 |
{
|
| 628 |
"epoch": 0.91,
|
| 629 |
-
"learning_rate":
|
| 630 |
-
"logits/chosen": -
|
| 631 |
-
"logits/rejected": -
|
| 632 |
-
"logps/chosen": -
|
| 633 |
-
"logps/rejected": -
|
| 634 |
-
"loss": 0.
|
| 635 |
-
"rewards/accuracies": 0.
|
| 636 |
-
"rewards/chosen": 0.
|
| 637 |
-
"rewards/margins":
|
| 638 |
-
"rewards/rejected": -
|
| 639 |
"step": 440
|
| 640 |
},
|
| 641 |
{
|
| 642 |
"epoch": 0.93,
|
| 643 |
-
"learning_rate":
|
| 644 |
-
"logits/chosen": -
|
| 645 |
-
"logits/rejected": -
|
| 646 |
-
"logps/chosen": -
|
| 647 |
-
"logps/rejected": -
|
| 648 |
-
"loss": 0.
|
| 649 |
-
"rewards/accuracies": 0.
|
| 650 |
-
"rewards/chosen": 0.
|
| 651 |
-
"rewards/margins":
|
| 652 |
-
"rewards/rejected": -
|
| 653 |
"step": 450
|
| 654 |
},
|
| 655 |
{
|
| 656 |
"epoch": 0.95,
|
| 657 |
-
"learning_rate":
|
| 658 |
-
"logits/chosen": -
|
| 659 |
-
"logits/rejected": -2.
|
| 660 |
-
"logps/chosen": -
|
| 661 |
-
"logps/rejected": -
|
| 662 |
-
"loss": 0.
|
| 663 |
-
"rewards/accuracies": 0.
|
| 664 |
-
"rewards/chosen": 0.
|
| 665 |
-
"rewards/margins":
|
| 666 |
-
"rewards/rejected": -
|
| 667 |
"step": 460
|
| 668 |
},
|
| 669 |
{
|
| 670 |
"epoch": 0.97,
|
| 671 |
-
"learning_rate": 1.
|
| 672 |
-
"logits/chosen": -
|
| 673 |
-
"logits/rejected": -
|
| 674 |
-
"logps/chosen": -
|
| 675 |
-
"logps/rejected": -
|
| 676 |
-
"loss": 0.
|
| 677 |
-
"rewards/accuracies": 0.
|
| 678 |
-
"rewards/chosen": 0.
|
| 679 |
-
"rewards/margins":
|
| 680 |
-
"rewards/rejected": -
|
| 681 |
"step": 470
|
| 682 |
},
|
| 683 |
{
|
| 684 |
"epoch": 0.99,
|
| 685 |
-
"learning_rate":
|
| 686 |
-
"logits/chosen": -
|
| 687 |
-
"logits/rejected": -2.
|
| 688 |
-
"logps/chosen": -
|
| 689 |
-
"logps/rejected": -
|
| 690 |
-
"loss": 0.
|
| 691 |
-
"rewards/accuracies": 0.
|
| 692 |
-
"rewards/chosen": 0.
|
| 693 |
-
"rewards/margins": 0.
|
| 694 |
-
"rewards/rejected": -
|
| 695 |
"step": 480
|
| 696 |
},
|
| 697 |
{
|
| 698 |
"epoch": 1.0,
|
| 699 |
-
"eval_logits/chosen": -
|
| 700 |
-
"eval_logits/rejected": -
|
| 701 |
-
"eval_logps/chosen": -
|
| 702 |
-
"eval_logps/rejected": -
|
| 703 |
-
"eval_loss": 0.
|
| 704 |
-
"eval_rewards/accuracies": 0.
|
| 705 |
-
"eval_rewards/chosen": 0.
|
| 706 |
-
"eval_rewards/margins":
|
| 707 |
-
"eval_rewards/rejected": -
|
| 708 |
-
"eval_runtime":
|
| 709 |
-
"eval_samples_per_second": 7.
|
| 710 |
-
"eval_steps_per_second": 0.
|
| 711 |
"step": 485
|
| 712 |
},
|
| 713 |
{
|
| 714 |
-
"epoch": 1.
|
| 715 |
-
"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 716 |
"total_flos": 0.0,
|
| 717 |
-
"train_loss": 0.
|
| 718 |
-
"train_runtime":
|
| 719 |
-
"train_samples_per_second":
|
| 720 |
"train_steps_per_second": 0.031
|
| 721 |
}
|
| 722 |
],
|
| 723 |
"logging_steps": 10,
|
| 724 |
-
"max_steps":
|
| 725 |
-
"num_train_epochs":
|
| 726 |
"save_steps": 500,
|
| 727 |
"total_flos": 0.0,
|
| 728 |
"trial_name": null,
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": null,
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 3.0,
|
| 5 |
"eval_steps": 100,
|
| 6 |
+
"global_step": 1455,
|
| 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": 1.36986301369863e-07,
|
| 14 |
+
"logits/chosen": -2.6635093688964844,
|
| 15 |
+
"logits/rejected": -2.7324111461639404,
|
| 16 |
+
"logps/chosen": -135.12002563476562,
|
| 17 |
+
"logps/rejected": -103.28743743896484,
|
| 18 |
"loss": 0.6931,
|
| 19 |
"rewards/accuracies": 0.0,
|
| 20 |
"rewards/chosen": 0.0,
|
|
|
|
| 24 |
},
|
| 25 |
{
|
| 26 |
"epoch": 0.02,
|
| 27 |
+
"learning_rate": 1.3698630136986302e-06,
|
| 28 |
+
"logits/chosen": -2.783435583114624,
|
| 29 |
+
"logits/rejected": -2.754120111465454,
|
| 30 |
+
"logps/chosen": -311.785400390625,
|
| 31 |
+
"logps/rejected": -273.2391357421875,
|
| 32 |
+
"loss": 0.6925,
|
| 33 |
+
"rewards/accuracies": 0.4513888955116272,
|
| 34 |
+
"rewards/chosen": -0.005032465327531099,
|
| 35 |
+
"rewards/margins": -0.004827913362532854,
|
| 36 |
+
"rewards/rejected": -0.00020455113553907722,
|
| 37 |
"step": 10
|
| 38 |
},
|
| 39 |
{
|
| 40 |
"epoch": 0.04,
|
| 41 |
+
"learning_rate": 2.7397260273972604e-06,
|
| 42 |
+
"logits/chosen": -2.8342247009277344,
|
| 43 |
+
"logits/rejected": -2.8470585346221924,
|
| 44 |
+
"logps/chosen": -283.9891662597656,
|
| 45 |
+
"logps/rejected": -250.61019897460938,
|
| 46 |
+
"loss": 0.6861,
|
| 47 |
+
"rewards/accuracies": 0.574999988079071,
|
| 48 |
+
"rewards/chosen": 0.006812130566686392,
|
| 49 |
+
"rewards/margins": 0.014938007108867168,
|
| 50 |
+
"rewards/rejected": -0.008125877007842064,
|
| 51 |
"step": 20
|
| 52 |
},
|
| 53 |
{
|
| 54 |
"epoch": 0.06,
|
| 55 |
+
"learning_rate": 4.109589041095891e-06,
|
| 56 |
+
"logits/chosen": -2.816066026687622,
|
| 57 |
+
"logits/rejected": -2.8595542907714844,
|
| 58 |
+
"logps/chosen": -282.41351318359375,
|
| 59 |
+
"logps/rejected": -200.7602081298828,
|
| 60 |
+
"loss": 0.6719,
|
| 61 |
+
"rewards/accuracies": 0.643750011920929,
|
| 62 |
+
"rewards/chosen": 0.009923343546688557,
|
| 63 |
+
"rewards/margins": 0.052515141665935516,
|
| 64 |
+
"rewards/rejected": -0.04259180277585983,
|
| 65 |
"step": 30
|
| 66 |
},
|
| 67 |
{
|
| 68 |
"epoch": 0.08,
|
| 69 |
+
"learning_rate": 5.479452054794521e-06,
|
| 70 |
+
"logits/chosen": -2.8367600440979004,
|
| 71 |
+
"logits/rejected": -2.777651309967041,
|
| 72 |
+
"logps/chosen": -292.46136474609375,
|
| 73 |
+
"logps/rejected": -239.9720001220703,
|
| 74 |
+
"loss": 0.6352,
|
| 75 |
+
"rewards/accuracies": 0.668749988079071,
|
| 76 |
+
"rewards/chosen": 0.041534725576639175,
|
| 77 |
+
"rewards/margins": 0.1540035903453827,
|
| 78 |
+
"rewards/rejected": -0.11246886104345322,
|
| 79 |
"step": 40
|
| 80 |
},
|
| 81 |
{
|
| 82 |
"epoch": 0.1,
|
| 83 |
+
"learning_rate": 6.849315068493151e-06,
|
| 84 |
+
"logits/chosen": -2.789390802383423,
|
| 85 |
+
"logits/rejected": -2.7937839031219482,
|
| 86 |
+
"logps/chosen": -296.8035583496094,
|
| 87 |
+
"logps/rejected": -220.61669921875,
|
| 88 |
+
"loss": 0.5922,
|
| 89 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 90 |
+
"rewards/chosen": 0.07272736728191376,
|
| 91 |
+
"rewards/margins": 0.3851665258407593,
|
| 92 |
+
"rewards/rejected": -0.3124391436576843,
|
| 93 |
"step": 50
|
| 94 |
},
|
| 95 |
{
|
| 96 |
"epoch": 0.12,
|
| 97 |
+
"learning_rate": 8.219178082191782e-06,
|
| 98 |
+
"logits/chosen": -2.8027291297912598,
|
| 99 |
+
"logits/rejected": -2.8161094188690186,
|
| 100 |
+
"logps/chosen": -298.8342590332031,
|
| 101 |
+
"logps/rejected": -248.74533081054688,
|
| 102 |
+
"loss": 0.5742,
|
| 103 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 104 |
+
"rewards/chosen": -0.023758064955472946,
|
| 105 |
+
"rewards/margins": 0.40244102478027344,
|
| 106 |
+
"rewards/rejected": -0.4261991083621979,
|
| 107 |
"step": 60
|
| 108 |
},
|
| 109 |
{
|
| 110 |
"epoch": 0.14,
|
| 111 |
+
"learning_rate": 9.589041095890411e-06,
|
| 112 |
+
"logits/chosen": -2.8271241188049316,
|
| 113 |
+
"logits/rejected": -2.833664894104004,
|
| 114 |
+
"logps/chosen": -282.1733703613281,
|
| 115 |
+
"logps/rejected": -261.71844482421875,
|
| 116 |
+
"loss": 0.5511,
|
| 117 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 118 |
+
"rewards/chosen": -0.05103034898638725,
|
| 119 |
+
"rewards/margins": 0.4532381594181061,
|
| 120 |
+
"rewards/rejected": -0.5042685270309448,
|
| 121 |
"step": 70
|
| 122 |
},
|
| 123 |
{
|
| 124 |
"epoch": 0.16,
|
| 125 |
+
"learning_rate": 1.0958904109589042e-05,
|
| 126 |
+
"logits/chosen": -2.8094613552093506,
|
| 127 |
+
"logits/rejected": -2.8167314529418945,
|
| 128 |
+
"logps/chosen": -302.01190185546875,
|
| 129 |
+
"logps/rejected": -241.74282836914062,
|
| 130 |
+
"loss": 0.538,
|
| 131 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 132 |
+
"rewards/chosen": -0.05076650530099869,
|
| 133 |
+
"rewards/margins": 0.729707658290863,
|
| 134 |
+
"rewards/rejected": -0.7804741263389587,
|
| 135 |
"step": 80
|
| 136 |
},
|
| 137 |
{
|
| 138 |
"epoch": 0.19,
|
| 139 |
+
"learning_rate": 1.2328767123287673e-05,
|
| 140 |
+
"logits/chosen": -2.7917096614837646,
|
| 141 |
+
"logits/rejected": -2.8236160278320312,
|
| 142 |
+
"logps/chosen": -255.85238647460938,
|
| 143 |
+
"logps/rejected": -210.2568359375,
|
| 144 |
+
"loss": 0.5137,
|
| 145 |
+
"rewards/accuracies": 0.75,
|
| 146 |
+
"rewards/chosen": -0.12663120031356812,
|
| 147 |
+
"rewards/margins": 0.7104201316833496,
|
| 148 |
+
"rewards/rejected": -0.8370513916015625,
|
| 149 |
"step": 90
|
| 150 |
},
|
| 151 |
{
|
| 152 |
"epoch": 0.21,
|
| 153 |
+
"learning_rate": 1.3698630136986302e-05,
|
| 154 |
+
"logits/chosen": -2.8499982357025146,
|
| 155 |
+
"logits/rejected": -2.8203647136688232,
|
| 156 |
+
"logps/chosen": -261.9218444824219,
|
| 157 |
+
"logps/rejected": -243.7086639404297,
|
| 158 |
+
"loss": 0.5222,
|
| 159 |
+
"rewards/accuracies": 0.8125,
|
| 160 |
+
"rewards/chosen": -0.13231949508190155,
|
| 161 |
+
"rewards/margins": 0.815433144569397,
|
| 162 |
+
"rewards/rejected": -0.9477526545524597,
|
| 163 |
"step": 100
|
| 164 |
},
|
| 165 |
{
|
| 166 |
"epoch": 0.23,
|
| 167 |
+
"learning_rate": 1.5068493150684933e-05,
|
| 168 |
+
"logits/chosen": -2.8208096027374268,
|
| 169 |
+
"logits/rejected": -2.8195183277130127,
|
| 170 |
+
"logps/chosen": -278.68377685546875,
|
| 171 |
+
"logps/rejected": -239.5231170654297,
|
| 172 |
+
"loss": 0.4973,
|
| 173 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 174 |
+
"rewards/chosen": -0.1940581500530243,
|
| 175 |
+
"rewards/margins": 0.9298496246337891,
|
| 176 |
+
"rewards/rejected": -1.1239076852798462,
|
| 177 |
"step": 110
|
| 178 |
},
|
| 179 |
{
|
| 180 |
"epoch": 0.25,
|
| 181 |
+
"learning_rate": 1.6438356164383563e-05,
|
| 182 |
+
"logits/chosen": -2.8535244464874268,
|
| 183 |
+
"logits/rejected": -2.8457090854644775,
|
| 184 |
+
"logps/chosen": -299.8865661621094,
|
| 185 |
+
"logps/rejected": -211.28964233398438,
|
| 186 |
+
"loss": 0.503,
|
| 187 |
+
"rewards/accuracies": 0.762499988079071,
|
| 188 |
+
"rewards/chosen": -0.1717432737350464,
|
| 189 |
+
"rewards/margins": 0.9522634744644165,
|
| 190 |
+
"rewards/rejected": -1.124006748199463,
|
| 191 |
"step": 120
|
| 192 |
},
|
| 193 |
{
|
| 194 |
"epoch": 0.27,
|
| 195 |
+
"learning_rate": 1.7808219178082194e-05,
|
| 196 |
+
"logits/chosen": -2.806631565093994,
|
| 197 |
+
"logits/rejected": -2.8321399688720703,
|
| 198 |
+
"logps/chosen": -273.78619384765625,
|
| 199 |
+
"logps/rejected": -232.2666015625,
|
| 200 |
+
"loss": 0.5137,
|
| 201 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 202 |
+
"rewards/chosen": -0.3173758387565613,
|
| 203 |
+
"rewards/margins": 0.8668048977851868,
|
| 204 |
+
"rewards/rejected": -1.1841806173324585,
|
| 205 |
"step": 130
|
| 206 |
},
|
| 207 |
{
|
| 208 |
"epoch": 0.29,
|
| 209 |
+
"learning_rate": 1.9178082191780822e-05,
|
| 210 |
+
"logits/chosen": -2.8575327396392822,
|
| 211 |
+
"logits/rejected": -2.8858590126037598,
|
| 212 |
+
"logps/chosen": -291.01837158203125,
|
| 213 |
+
"logps/rejected": -233.05001831054688,
|
| 214 |
+
"loss": 0.528,
|
| 215 |
+
"rewards/accuracies": 0.706250011920929,
|
| 216 |
+
"rewards/chosen": -0.21833661198616028,
|
| 217 |
+
"rewards/margins": 0.7500525712966919,
|
| 218 |
+
"rewards/rejected": -0.9683893322944641,
|
| 219 |
"step": 140
|
| 220 |
},
|
| 221 |
{
|
| 222 |
"epoch": 0.31,
|
| 223 |
+
"learning_rate": 1.9938884644767e-05,
|
| 224 |
+
"logits/chosen": -2.8324332237243652,
|
| 225 |
+
"logits/rejected": -2.8657970428466797,
|
| 226 |
+
"logps/chosen": -263.90875244140625,
|
| 227 |
+
"logps/rejected": -248.7552490234375,
|
| 228 |
+
"loss": 0.5066,
|
| 229 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 230 |
+
"rewards/chosen": -0.226444810628891,
|
| 231 |
+
"rewards/margins": 0.6770893335342407,
|
| 232 |
+
"rewards/rejected": -0.9035340547561646,
|
| 233 |
"step": 150
|
| 234 |
},
|
| 235 |
{
|
| 236 |
"epoch": 0.33,
|
| 237 |
+
"learning_rate": 1.9786096256684494e-05,
|
| 238 |
+
"logits/chosen": -2.85012149810791,
|
| 239 |
+
"logits/rejected": -2.920163869857788,
|
| 240 |
+
"logps/chosen": -313.2906494140625,
|
| 241 |
+
"logps/rejected": -240.2455596923828,
|
| 242 |
+
"loss": 0.514,
|
| 243 |
+
"rewards/accuracies": 0.75,
|
| 244 |
+
"rewards/chosen": -0.15915456414222717,
|
| 245 |
+
"rewards/margins": 0.9070942997932434,
|
| 246 |
+
"rewards/rejected": -1.066248893737793,
|
| 247 |
"step": 160
|
| 248 |
},
|
| 249 |
{
|
| 250 |
"epoch": 0.35,
|
| 251 |
+
"learning_rate": 1.9633307868601987e-05,
|
| 252 |
+
"logits/chosen": -2.8070321083068848,
|
| 253 |
+
"logits/rejected": -2.8099472522735596,
|
| 254 |
+
"logps/chosen": -302.35736083984375,
|
| 255 |
+
"logps/rejected": -241.8223419189453,
|
| 256 |
+
"loss": 0.5346,
|
| 257 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 258 |
+
"rewards/chosen": -0.07760269939899445,
|
| 259 |
+
"rewards/margins": 0.7661724090576172,
|
| 260 |
+
"rewards/rejected": -0.8437750935554504,
|
| 261 |
"step": 170
|
| 262 |
},
|
| 263 |
{
|
| 264 |
"epoch": 0.37,
|
| 265 |
+
"learning_rate": 1.9480519480519483e-05,
|
| 266 |
+
"logits/chosen": -2.7994544506073,
|
| 267 |
+
"logits/rejected": -2.8084394931793213,
|
| 268 |
+
"logps/chosen": -302.2978820800781,
|
| 269 |
+
"logps/rejected": -248.54586791992188,
|
| 270 |
+
"loss": 0.4909,
|
| 271 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 272 |
+
"rewards/chosen": -0.09264491498470306,
|
| 273 |
+
"rewards/margins": 0.8108696937561035,
|
| 274 |
+
"rewards/rejected": -0.9035146832466125,
|
| 275 |
"step": 180
|
| 276 |
},
|
| 277 |
{
|
| 278 |
"epoch": 0.39,
|
| 279 |
+
"learning_rate": 1.9327731092436976e-05,
|
| 280 |
+
"logits/chosen": -2.867279052734375,
|
| 281 |
+
"logits/rejected": -2.8662352561950684,
|
| 282 |
+
"logps/chosen": -294.74139404296875,
|
| 283 |
+
"logps/rejected": -247.1249542236328,
|
| 284 |
+
"loss": 0.5157,
|
| 285 |
+
"rewards/accuracies": 0.831250011920929,
|
| 286 |
+
"rewards/chosen": -0.023864692077040672,
|
| 287 |
+
"rewards/margins": 1.0177843570709229,
|
| 288 |
+
"rewards/rejected": -1.0416491031646729,
|
| 289 |
"step": 190
|
| 290 |
},
|
| 291 |
{
|
| 292 |
"epoch": 0.41,
|
| 293 |
+
"learning_rate": 1.9174942704354472e-05,
|
| 294 |
+
"logits/chosen": -2.7849340438842773,
|
| 295 |
+
"logits/rejected": -2.8029227256774902,
|
| 296 |
+
"logps/chosen": -274.12408447265625,
|
| 297 |
+
"logps/rejected": -234.2401885986328,
|
| 298 |
+
"loss": 0.4979,
|
| 299 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 300 |
+
"rewards/chosen": -0.18910327553749084,
|
| 301 |
+
"rewards/margins": 0.9559990763664246,
|
| 302 |
+
"rewards/rejected": -1.1451025009155273,
|
| 303 |
"step": 200
|
| 304 |
},
|
| 305 |
{
|
| 306 |
"epoch": 0.43,
|
| 307 |
+
"learning_rate": 1.9022154316271965e-05,
|
| 308 |
+
"logits/chosen": -2.886434555053711,
|
| 309 |
+
"logits/rejected": -2.8876872062683105,
|
| 310 |
+
"logps/chosen": -285.2301940917969,
|
| 311 |
+
"logps/rejected": -270.3388977050781,
|
| 312 |
+
"loss": 0.5246,
|
| 313 |
+
"rewards/accuracies": 0.800000011920929,
|
| 314 |
+
"rewards/chosen": -0.0919012501835823,
|
| 315 |
+
"rewards/margins": 1.0152003765106201,
|
| 316 |
+
"rewards/rejected": -1.1071016788482666,
|
| 317 |
"step": 210
|
| 318 |
},
|
| 319 |
{
|
| 320 |
"epoch": 0.45,
|
| 321 |
+
"learning_rate": 1.8869365928189458e-05,
|
| 322 |
+
"logits/chosen": -2.8153045177459717,
|
| 323 |
+
"logits/rejected": -2.8722925186157227,
|
| 324 |
+
"logps/chosen": -295.7647399902344,
|
| 325 |
+
"logps/rejected": -241.56723022460938,
|
| 326 |
+
"loss": 0.4931,
|
| 327 |
+
"rewards/accuracies": 0.762499988079071,
|
| 328 |
+
"rewards/chosen": -0.16732807457447052,
|
| 329 |
+
"rewards/margins": 0.9812033772468567,
|
| 330 |
+
"rewards/rejected": -1.1485313177108765,
|
| 331 |
"step": 220
|
| 332 |
},
|
| 333 |
{
|
| 334 |
"epoch": 0.47,
|
| 335 |
+
"learning_rate": 1.8716577540106954e-05,
|
| 336 |
+
"logits/chosen": -2.890263080596924,
|
| 337 |
+
"logits/rejected": -2.870927333831787,
|
| 338 |
+
"logps/chosen": -300.0990905761719,
|
| 339 |
+
"logps/rejected": -243.5838165283203,
|
| 340 |
+
"loss": 0.5021,
|
| 341 |
+
"rewards/accuracies": 0.75,
|
| 342 |
+
"rewards/chosen": -0.1739131659269333,
|
| 343 |
+
"rewards/margins": 0.8733876943588257,
|
| 344 |
+
"rewards/rejected": -1.047300934791565,
|
| 345 |
"step": 230
|
| 346 |
},
|
| 347 |
{
|
| 348 |
"epoch": 0.49,
|
| 349 |
+
"learning_rate": 1.8563789152024447e-05,
|
| 350 |
+
"logits/chosen": -2.792865753173828,
|
| 351 |
+
"logits/rejected": -2.8081324100494385,
|
| 352 |
+
"logps/chosen": -249.1541748046875,
|
| 353 |
+
"logps/rejected": -221.9134979248047,
|
| 354 |
+
"loss": 0.4763,
|
| 355 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 356 |
+
"rewards/chosen": -0.2561655640602112,
|
| 357 |
+
"rewards/margins": 1.2132002115249634,
|
| 358 |
+
"rewards/rejected": -1.4693658351898193,
|
| 359 |
"step": 240
|
| 360 |
},
|
| 361 |
{
|
| 362 |
"epoch": 0.52,
|
| 363 |
+
"learning_rate": 1.8411000763941943e-05,
|
| 364 |
+
"logits/chosen": -2.780810594558716,
|
| 365 |
+
"logits/rejected": -2.8007028102874756,
|
| 366 |
+
"logps/chosen": -289.1332702636719,
|
| 367 |
+
"logps/rejected": -259.22491455078125,
|
| 368 |
+
"loss": 0.488,
|
| 369 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 370 |
+
"rewards/chosen": -0.20490197837352753,
|
| 371 |
+
"rewards/margins": 0.9276278614997864,
|
| 372 |
+
"rewards/rejected": -1.132529854774475,
|
| 373 |
"step": 250
|
| 374 |
},
|
| 375 |
{
|
| 376 |
"epoch": 0.54,
|
| 377 |
+
"learning_rate": 1.8258212375859436e-05,
|
| 378 |
+
"logits/chosen": -2.7935876846313477,
|
| 379 |
+
"logits/rejected": -2.795382022857666,
|
| 380 |
+
"logps/chosen": -300.5179748535156,
|
| 381 |
+
"logps/rejected": -260.3360900878906,
|
| 382 |
+
"loss": 0.4823,
|
| 383 |
+
"rewards/accuracies": 0.75,
|
| 384 |
+
"rewards/chosen": -0.22746071219444275,
|
| 385 |
+
"rewards/margins": 1.0280539989471436,
|
| 386 |
+
"rewards/rejected": -1.2555148601531982,
|
| 387 |
"step": 260
|
| 388 |
},
|
| 389 |
{
|
| 390 |
"epoch": 0.56,
|
| 391 |
+
"learning_rate": 1.8105423987776932e-05,
|
| 392 |
+
"logits/chosen": -2.8180222511291504,
|
| 393 |
+
"logits/rejected": -2.7675962448120117,
|
| 394 |
+
"logps/chosen": -264.4036560058594,
|
| 395 |
+
"logps/rejected": -242.5054931640625,
|
| 396 |
+
"loss": 0.4994,
|
| 397 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 398 |
+
"rewards/chosen": -0.12344682216644287,
|
| 399 |
+
"rewards/margins": 1.0876801013946533,
|
| 400 |
+
"rewards/rejected": -1.2111269235610962,
|
| 401 |
"step": 270
|
| 402 |
},
|
| 403 |
{
|
| 404 |
"epoch": 0.58,
|
| 405 |
+
"learning_rate": 1.7952635599694425e-05,
|
| 406 |
+
"logits/chosen": -2.859412908554077,
|
| 407 |
+
"logits/rejected": -2.7997498512268066,
|
| 408 |
+
"logps/chosen": -274.4200744628906,
|
| 409 |
+
"logps/rejected": -253.32235717773438,
|
| 410 |
+
"loss": 0.4921,
|
| 411 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 412 |
+
"rewards/chosen": -0.08681371808052063,
|
| 413 |
+
"rewards/margins": 0.9918006658554077,
|
| 414 |
+
"rewards/rejected": -1.078614354133606,
|
| 415 |
"step": 280
|
| 416 |
},
|
| 417 |
{
|
| 418 |
"epoch": 0.6,
|
| 419 |
+
"learning_rate": 1.7799847211611917e-05,
|
| 420 |
+
"logits/chosen": -2.8540239334106445,
|
| 421 |
+
"logits/rejected": -2.8323752880096436,
|
| 422 |
+
"logps/chosen": -297.4104309082031,
|
| 423 |
+
"logps/rejected": -258.0458984375,
|
| 424 |
+
"loss": 0.4741,
|
| 425 |
+
"rewards/accuracies": 0.71875,
|
| 426 |
+
"rewards/chosen": -0.1035405844449997,
|
| 427 |
+
"rewards/margins": 1.009918451309204,
|
| 428 |
+
"rewards/rejected": -1.1134589910507202,
|
| 429 |
"step": 290
|
| 430 |
},
|
| 431 |
{
|
| 432 |
"epoch": 0.62,
|
| 433 |
+
"learning_rate": 1.7647058823529414e-05,
|
| 434 |
+
"logits/chosen": -2.835106372833252,
|
| 435 |
+
"logits/rejected": -2.8273653984069824,
|
| 436 |
+
"logps/chosen": -251.33984375,
|
| 437 |
+
"logps/rejected": -243.59890747070312,
|
| 438 |
+
"loss": 0.4957,
|
| 439 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 440 |
+
"rewards/chosen": -0.1399732530117035,
|
| 441 |
+
"rewards/margins": 0.9917305111885071,
|
| 442 |
+
"rewards/rejected": -1.1317037343978882,
|
| 443 |
"step": 300
|
| 444 |
},
|
| 445 |
{
|
| 446 |
"epoch": 0.64,
|
| 447 |
+
"learning_rate": 1.7494270435446906e-05,
|
| 448 |
+
"logits/chosen": -2.8415441513061523,
|
| 449 |
+
"logits/rejected": -2.7994866371154785,
|
| 450 |
+
"logps/chosen": -260.0638732910156,
|
| 451 |
+
"logps/rejected": -227.3611602783203,
|
| 452 |
+
"loss": 0.5106,
|
| 453 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 454 |
+
"rewards/chosen": -0.05403571575880051,
|
| 455 |
+
"rewards/margins": 1.2989342212677002,
|
| 456 |
+
"rewards/rejected": -1.3529701232910156,
|
| 457 |
"step": 310
|
| 458 |
},
|
| 459 |
{
|
| 460 |
"epoch": 0.66,
|
| 461 |
+
"learning_rate": 1.7341482047364403e-05,
|
| 462 |
+
"logits/chosen": -2.870739698410034,
|
| 463 |
+
"logits/rejected": -2.808722496032715,
|
| 464 |
+
"logps/chosen": -252.57870483398438,
|
| 465 |
+
"logps/rejected": -238.3760528564453,
|
| 466 |
+
"loss": 0.4714,
|
| 467 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 468 |
+
"rewards/chosen": -0.3325572907924652,
|
| 469 |
+
"rewards/margins": 1.0449466705322266,
|
| 470 |
+
"rewards/rejected": -1.3775039911270142,
|
| 471 |
"step": 320
|
| 472 |
},
|
| 473 |
{
|
| 474 |
"epoch": 0.68,
|
| 475 |
+
"learning_rate": 1.7188693659281895e-05,
|
| 476 |
+
"logits/chosen": -2.870849132537842,
|
| 477 |
+
"logits/rejected": -2.8693222999572754,
|
| 478 |
+
"logps/chosen": -296.0721435546875,
|
| 479 |
+
"logps/rejected": -252.4921875,
|
| 480 |
+
"loss": 0.4776,
|
| 481 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 482 |
+
"rewards/chosen": -0.14332377910614014,
|
| 483 |
+
"rewards/margins": 1.1690990924835205,
|
| 484 |
+
"rewards/rejected": -1.312422752380371,
|
| 485 |
"step": 330
|
| 486 |
},
|
| 487 |
{
|
| 488 |
"epoch": 0.7,
|
| 489 |
+
"learning_rate": 1.703590527119939e-05,
|
| 490 |
+
"logits/chosen": -2.82586669921875,
|
| 491 |
+
"logits/rejected": -2.780189037322998,
|
| 492 |
+
"logps/chosen": -315.26727294921875,
|
| 493 |
+
"logps/rejected": -268.7083740234375,
|
| 494 |
+
"loss": 0.5107,
|
| 495 |
+
"rewards/accuracies": 0.75,
|
| 496 |
+
"rewards/chosen": -0.26427751779556274,
|
| 497 |
+
"rewards/margins": 0.9479537010192871,
|
| 498 |
+
"rewards/rejected": -1.2122312784194946,
|
| 499 |
"step": 340
|
| 500 |
},
|
| 501 |
{
|
| 502 |
"epoch": 0.72,
|
| 503 |
+
"learning_rate": 1.6883116883116884e-05,
|
| 504 |
+
"logits/chosen": -2.8107895851135254,
|
| 505 |
+
"logits/rejected": -2.8060848712921143,
|
| 506 |
+
"logps/chosen": -242.031494140625,
|
| 507 |
+
"logps/rejected": -227.6324920654297,
|
| 508 |
+
"loss": 0.5167,
|
| 509 |
+
"rewards/accuracies": 0.78125,
|
| 510 |
+
"rewards/chosen": -0.1813223510980606,
|
| 511 |
+
"rewards/margins": 0.9393804669380188,
|
| 512 |
+
"rewards/rejected": -1.1207029819488525,
|
| 513 |
"step": 350
|
| 514 |
},
|
| 515 |
{
|
| 516 |
"epoch": 0.74,
|
| 517 |
+
"learning_rate": 1.6730328495034377e-05,
|
| 518 |
+
"logits/chosen": -2.8176794052124023,
|
| 519 |
+
"logits/rejected": -2.76552152633667,
|
| 520 |
+
"logps/chosen": -263.55047607421875,
|
| 521 |
+
"logps/rejected": -240.71484375,
|
| 522 |
+
"loss": 0.5146,
|
| 523 |
+
"rewards/accuracies": 0.75,
|
| 524 |
+
"rewards/chosen": -0.1424541026353836,
|
| 525 |
+
"rewards/margins": 0.9593822360038757,
|
| 526 |
+
"rewards/rejected": -1.1018364429473877,
|
| 527 |
"step": 360
|
| 528 |
},
|
| 529 |
{
|
| 530 |
"epoch": 0.76,
|
| 531 |
+
"learning_rate": 1.6577540106951873e-05,
|
| 532 |
+
"logits/chosen": -2.8674864768981934,
|
| 533 |
+
"logits/rejected": -2.8338735103607178,
|
| 534 |
+
"logps/chosen": -288.48687744140625,
|
| 535 |
+
"logps/rejected": -261.1170349121094,
|
| 536 |
+
"loss": 0.4945,
|
| 537 |
+
"rewards/accuracies": 0.793749988079071,
|
| 538 |
+
"rewards/chosen": -0.24135860800743103,
|
| 539 |
+
"rewards/margins": 0.9816256761550903,
|
| 540 |
+
"rewards/rejected": -1.2229843139648438,
|
| 541 |
"step": 370
|
| 542 |
},
|
| 543 |
{
|
| 544 |
"epoch": 0.78,
|
| 545 |
+
"learning_rate": 1.6424751718869366e-05,
|
| 546 |
+
"logits/chosen": -2.8083739280700684,
|
| 547 |
+
"logits/rejected": -2.7698397636413574,
|
| 548 |
+
"logps/chosen": -243.06356811523438,
|
| 549 |
+
"logps/rejected": -258.90496826171875,
|
| 550 |
+
"loss": 0.4858,
|
| 551 |
+
"rewards/accuracies": 0.731249988079071,
|
| 552 |
+
"rewards/chosen": -0.4036439061164856,
|
| 553 |
+
"rewards/margins": 1.0016567707061768,
|
| 554 |
+
"rewards/rejected": -1.4053006172180176,
|
| 555 |
"step": 380
|
| 556 |
},
|
| 557 |
{
|
| 558 |
"epoch": 0.8,
|
| 559 |
+
"learning_rate": 1.6271963330786862e-05,
|
| 560 |
+
"logits/chosen": -2.8200416564941406,
|
| 561 |
+
"logits/rejected": -2.809105634689331,
|
| 562 |
+
"logps/chosen": -317.0741882324219,
|
| 563 |
+
"logps/rejected": -248.5664825439453,
|
| 564 |
+
"loss": 0.4763,
|
| 565 |
+
"rewards/accuracies": 0.793749988079071,
|
| 566 |
+
"rewards/chosen": -0.23517903685569763,
|
| 567 |
+
"rewards/margins": 1.1138803958892822,
|
| 568 |
+
"rewards/rejected": -1.3490597009658813,
|
| 569 |
"step": 390
|
| 570 |
},
|
| 571 |
{
|
| 572 |
"epoch": 0.82,
|
| 573 |
+
"learning_rate": 1.6119174942704355e-05,
|
| 574 |
+
"logits/chosen": -2.860297679901123,
|
| 575 |
+
"logits/rejected": -2.8406786918640137,
|
| 576 |
+
"logps/chosen": -281.96527099609375,
|
| 577 |
+
"logps/rejected": -261.6462707519531,
|
| 578 |
+
"loss": 0.4637,
|
| 579 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 580 |
+
"rewards/chosen": -0.20879487693309784,
|
| 581 |
+
"rewards/margins": 1.0736795663833618,
|
| 582 |
+
"rewards/rejected": -1.2824745178222656,
|
| 583 |
"step": 400
|
| 584 |
},
|
| 585 |
{
|
| 586 |
"epoch": 0.85,
|
| 587 |
+
"learning_rate": 1.596638655462185e-05,
|
| 588 |
+
"logits/chosen": -2.8639817237854004,
|
| 589 |
+
"logits/rejected": -2.8506665229797363,
|
| 590 |
+
"logps/chosen": -295.0474853515625,
|
| 591 |
+
"logps/rejected": -243.6042938232422,
|
| 592 |
+
"loss": 0.4449,
|
| 593 |
+
"rewards/accuracies": 0.800000011920929,
|
| 594 |
+
"rewards/chosen": -0.20277197659015656,
|
| 595 |
+
"rewards/margins": 1.3167582750320435,
|
| 596 |
+
"rewards/rejected": -1.519530177116394,
|
| 597 |
"step": 410
|
| 598 |
},
|
| 599 |
{
|
| 600 |
"epoch": 0.87,
|
| 601 |
+
"learning_rate": 1.5813598166539344e-05,
|
| 602 |
+
"logits/chosen": -2.83903431892395,
|
| 603 |
+
"logits/rejected": -2.8222603797912598,
|
| 604 |
+
"logps/chosen": -279.1824035644531,
|
| 605 |
+
"logps/rejected": -243.2361297607422,
|
| 606 |
+
"loss": 0.4928,
|
| 607 |
+
"rewards/accuracies": 0.75,
|
| 608 |
+
"rewards/chosen": -0.4378887712955475,
|
| 609 |
+
"rewards/margins": 1.1546775102615356,
|
| 610 |
+
"rewards/rejected": -1.5925662517547607,
|
| 611 |
"step": 420
|
| 612 |
},
|
| 613 |
{
|
| 614 |
"epoch": 0.89,
|
| 615 |
+
"learning_rate": 1.5660809778456837e-05,
|
| 616 |
+
"logits/chosen": -2.8503293991088867,
|
| 617 |
+
"logits/rejected": -2.8215270042419434,
|
| 618 |
+
"logps/chosen": -260.7596130371094,
|
| 619 |
+
"logps/rejected": -272.5841979980469,
|
| 620 |
+
"loss": 0.49,
|
| 621 |
+
"rewards/accuracies": 0.706250011920929,
|
| 622 |
+
"rewards/chosen": -0.2955819368362427,
|
| 623 |
+
"rewards/margins": 0.9733622670173645,
|
| 624 |
+
"rewards/rejected": -1.268944263458252,
|
| 625 |
"step": 430
|
| 626 |
},
|
| 627 |
{
|
| 628 |
"epoch": 0.91,
|
| 629 |
+
"learning_rate": 1.5508021390374333e-05,
|
| 630 |
+
"logits/chosen": -2.820071220397949,
|
| 631 |
+
"logits/rejected": -2.7907521724700928,
|
| 632 |
+
"logps/chosen": -307.09576416015625,
|
| 633 |
+
"logps/rejected": -249.6613006591797,
|
| 634 |
+
"loss": 0.4867,
|
| 635 |
+
"rewards/accuracies": 0.762499988079071,
|
| 636 |
+
"rewards/chosen": -0.3214530348777771,
|
| 637 |
+
"rewards/margins": 1.281368613243103,
|
| 638 |
+
"rewards/rejected": -1.602821707725525,
|
| 639 |
"step": 440
|
| 640 |
},
|
| 641 |
{
|
| 642 |
"epoch": 0.93,
|
| 643 |
+
"learning_rate": 1.5355233002291826e-05,
|
| 644 |
+
"logits/chosen": -2.8464276790618896,
|
| 645 |
+
"logits/rejected": -2.7966537475585938,
|
| 646 |
+
"logps/chosen": -296.7311706542969,
|
| 647 |
+
"logps/rejected": -254.079833984375,
|
| 648 |
+
"loss": 0.4991,
|
| 649 |
+
"rewards/accuracies": 0.793749988079071,
|
| 650 |
+
"rewards/chosen": -0.35059598088264465,
|
| 651 |
+
"rewards/margins": 1.165145993232727,
|
| 652 |
+
"rewards/rejected": -1.5157420635223389,
|
| 653 |
"step": 450
|
| 654 |
},
|
| 655 |
{
|
| 656 |
"epoch": 0.95,
|
| 657 |
+
"learning_rate": 1.5202444614209322e-05,
|
| 658 |
+
"logits/chosen": -2.8699872493743896,
|
| 659 |
+
"logits/rejected": -2.8504984378814697,
|
| 660 |
+
"logps/chosen": -262.0877990722656,
|
| 661 |
+
"logps/rejected": -247.0086212158203,
|
| 662 |
+
"loss": 0.4908,
|
| 663 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 664 |
+
"rewards/chosen": -0.2521810531616211,
|
| 665 |
+
"rewards/margins": 1.0113656520843506,
|
| 666 |
+
"rewards/rejected": -1.2635467052459717,
|
| 667 |
"step": 460
|
| 668 |
},
|
| 669 |
{
|
| 670 |
"epoch": 0.97,
|
| 671 |
+
"learning_rate": 1.5049656226126816e-05,
|
| 672 |
+
"logits/chosen": -2.846945285797119,
|
| 673 |
+
"logits/rejected": -2.849860191345215,
|
| 674 |
+
"logps/chosen": -292.128173828125,
|
| 675 |
+
"logps/rejected": -255.2592315673828,
|
| 676 |
+
"loss": 0.48,
|
| 677 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 678 |
+
"rewards/chosen": -0.06887436658143997,
|
| 679 |
+
"rewards/margins": 1.2138417959213257,
|
| 680 |
+
"rewards/rejected": -1.282716155052185,
|
| 681 |
"step": 470
|
| 682 |
},
|
| 683 |
{
|
| 684 |
"epoch": 0.99,
|
| 685 |
+
"learning_rate": 1.489686783804431e-05,
|
| 686 |
+
"logits/chosen": -2.872967481613159,
|
| 687 |
+
"logits/rejected": -2.8441576957702637,
|
| 688 |
+
"logps/chosen": -263.8342590332031,
|
| 689 |
+
"logps/rejected": -239.5003204345703,
|
| 690 |
+
"loss": 0.4771,
|
| 691 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 692 |
+
"rewards/chosen": -0.3044741153717041,
|
| 693 |
+
"rewards/margins": 0.980197548866272,
|
| 694 |
+
"rewards/rejected": -1.2846715450286865,
|
| 695 |
"step": 480
|
| 696 |
},
|
| 697 |
{
|
| 698 |
"epoch": 1.0,
|
| 699 |
+
"eval_logits/chosen": -2.8796305656433105,
|
| 700 |
+
"eval_logits/rejected": -2.8501133918762207,
|
| 701 |
+
"eval_logps/chosen": -274.4910583496094,
|
| 702 |
+
"eval_logps/rejected": -238.3082275390625,
|
| 703 |
+
"eval_loss": 0.4616946280002594,
|
| 704 |
+
"eval_rewards/accuracies": 0.7890625,
|
| 705 |
+
"eval_rewards/chosen": -0.0842861607670784,
|
| 706 |
+
"eval_rewards/margins": 1.2677102088928223,
|
| 707 |
+
"eval_rewards/rejected": -1.3519961833953857,
|
| 708 |
+
"eval_runtime": 253.7733,
|
| 709 |
+
"eval_samples_per_second": 7.881,
|
| 710 |
+
"eval_steps_per_second": 0.063,
|
| 711 |
"step": 485
|
| 712 |
},
|
| 713 |
{
|
| 714 |
+
"epoch": 1.01,
|
| 715 |
+
"learning_rate": 1.4744079449961804e-05,
|
| 716 |
+
"logits/chosen": -2.8916447162628174,
|
| 717 |
+
"logits/rejected": -2.836836576461792,
|
| 718 |
+
"logps/chosen": -317.30328369140625,
|
| 719 |
+
"logps/rejected": -261.12030029296875,
|
| 720 |
+
"loss": 0.4271,
|
| 721 |
+
"rewards/accuracies": 0.8125,
|
| 722 |
+
"rewards/chosen": -0.1437714546918869,
|
| 723 |
+
"rewards/margins": 1.377268671989441,
|
| 724 |
+
"rewards/rejected": -1.5210400819778442,
|
| 725 |
+
"step": 490
|
| 726 |
+
},
|
| 727 |
+
{
|
| 728 |
+
"epoch": 1.03,
|
| 729 |
+
"learning_rate": 1.4591291061879298e-05,
|
| 730 |
+
"logits/chosen": -2.8427810668945312,
|
| 731 |
+
"logits/rejected": -2.8195979595184326,
|
| 732 |
+
"logps/chosen": -262.42047119140625,
|
| 733 |
+
"logps/rejected": -256.4164733886719,
|
| 734 |
+
"loss": 0.4627,
|
| 735 |
+
"rewards/accuracies": 0.768750011920929,
|
| 736 |
+
"rewards/chosen": -0.3208313584327698,
|
| 737 |
+
"rewards/margins": 1.1283810138702393,
|
| 738 |
+
"rewards/rejected": -1.4492123126983643,
|
| 739 |
+
"step": 500
|
| 740 |
+
},
|
| 741 |
+
{
|
| 742 |
+
"epoch": 1.05,
|
| 743 |
+
"learning_rate": 1.4438502673796793e-05,
|
| 744 |
+
"logits/chosen": -2.8156418800354004,
|
| 745 |
+
"logits/rejected": -2.843306064605713,
|
| 746 |
+
"logps/chosen": -267.56536865234375,
|
| 747 |
+
"logps/rejected": -224.3167266845703,
|
| 748 |
+
"loss": 0.4201,
|
| 749 |
+
"rewards/accuracies": 0.800000011920929,
|
| 750 |
+
"rewards/chosen": -0.24222290515899658,
|
| 751 |
+
"rewards/margins": 1.2655701637268066,
|
| 752 |
+
"rewards/rejected": -1.5077931880950928,
|
| 753 |
+
"step": 510
|
| 754 |
+
},
|
| 755 |
+
{
|
| 756 |
+
"epoch": 1.07,
|
| 757 |
+
"learning_rate": 1.4285714285714287e-05,
|
| 758 |
+
"logits/chosen": -2.7805769443511963,
|
| 759 |
+
"logits/rejected": -2.8112196922302246,
|
| 760 |
+
"logps/chosen": -287.56451416015625,
|
| 761 |
+
"logps/rejected": -262.3890075683594,
|
| 762 |
+
"loss": 0.4203,
|
| 763 |
+
"rewards/accuracies": 0.737500011920929,
|
| 764 |
+
"rewards/chosen": -0.2784760892391205,
|
| 765 |
+
"rewards/margins": 1.08200204372406,
|
| 766 |
+
"rewards/rejected": -1.360478162765503,
|
| 767 |
+
"step": 520
|
| 768 |
+
},
|
| 769 |
+
{
|
| 770 |
+
"epoch": 1.09,
|
| 771 |
+
"learning_rate": 1.4132925897631782e-05,
|
| 772 |
+
"logits/chosen": -2.815563678741455,
|
| 773 |
+
"logits/rejected": -2.839988946914673,
|
| 774 |
+
"logps/chosen": -295.6175842285156,
|
| 775 |
+
"logps/rejected": -235.90164184570312,
|
| 776 |
+
"loss": 0.4009,
|
| 777 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 778 |
+
"rewards/chosen": -0.23429016768932343,
|
| 779 |
+
"rewards/margins": 1.343583106994629,
|
| 780 |
+
"rewards/rejected": -1.5778734683990479,
|
| 781 |
+
"step": 530
|
| 782 |
+
},
|
| 783 |
+
{
|
| 784 |
+
"epoch": 1.11,
|
| 785 |
+
"learning_rate": 1.3980137509549276e-05,
|
| 786 |
+
"logits/chosen": -2.8498940467834473,
|
| 787 |
+
"logits/rejected": -2.830718517303467,
|
| 788 |
+
"logps/chosen": -318.00909423828125,
|
| 789 |
+
"logps/rejected": -257.63348388671875,
|
| 790 |
+
"loss": 0.4186,
|
| 791 |
+
"rewards/accuracies": 0.793749988079071,
|
| 792 |
+
"rewards/chosen": -0.3423912525177002,
|
| 793 |
+
"rewards/margins": 1.4794024229049683,
|
| 794 |
+
"rewards/rejected": -1.821793794631958,
|
| 795 |
+
"step": 540
|
| 796 |
+
},
|
| 797 |
+
{
|
| 798 |
+
"epoch": 1.13,
|
| 799 |
+
"learning_rate": 1.3827349121466769e-05,
|
| 800 |
+
"logits/chosen": -2.7992899417877197,
|
| 801 |
+
"logits/rejected": -2.8059680461883545,
|
| 802 |
+
"logps/chosen": -258.46514892578125,
|
| 803 |
+
"logps/rejected": -210.663818359375,
|
| 804 |
+
"loss": 0.426,
|
| 805 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 806 |
+
"rewards/chosen": -0.40961843729019165,
|
| 807 |
+
"rewards/margins": 1.0720117092132568,
|
| 808 |
+
"rewards/rejected": -1.4816303253173828,
|
| 809 |
+
"step": 550
|
| 810 |
+
},
|
| 811 |
+
{
|
| 812 |
+
"epoch": 1.15,
|
| 813 |
+
"learning_rate": 1.3674560733384263e-05,
|
| 814 |
+
"logits/chosen": -2.862682580947876,
|
| 815 |
+
"logits/rejected": -2.865583658218384,
|
| 816 |
+
"logps/chosen": -281.8017883300781,
|
| 817 |
+
"logps/rejected": -257.26690673828125,
|
| 818 |
+
"loss": 0.4007,
|
| 819 |
+
"rewards/accuracies": 0.8687499761581421,
|
| 820 |
+
"rewards/chosen": -0.2565487325191498,
|
| 821 |
+
"rewards/margins": 1.5748827457427979,
|
| 822 |
+
"rewards/rejected": -1.8314317464828491,
|
| 823 |
+
"step": 560
|
| 824 |
+
},
|
| 825 |
+
{
|
| 826 |
+
"epoch": 1.18,
|
| 827 |
+
"learning_rate": 1.3521772345301758e-05,
|
| 828 |
+
"logits/chosen": -2.8742198944091797,
|
| 829 |
+
"logits/rejected": -2.849208354949951,
|
| 830 |
+
"logps/chosen": -240.13259887695312,
|
| 831 |
+
"logps/rejected": -242.2512664794922,
|
| 832 |
+
"loss": 0.3911,
|
| 833 |
+
"rewards/accuracies": 0.800000011920929,
|
| 834 |
+
"rewards/chosen": -0.2923469543457031,
|
| 835 |
+
"rewards/margins": 1.5814087390899658,
|
| 836 |
+
"rewards/rejected": -1.873755693435669,
|
| 837 |
+
"step": 570
|
| 838 |
+
},
|
| 839 |
+
{
|
| 840 |
+
"epoch": 1.2,
|
| 841 |
+
"learning_rate": 1.3368983957219252e-05,
|
| 842 |
+
"logits/chosen": -2.8210272789001465,
|
| 843 |
+
"logits/rejected": -2.8307366371154785,
|
| 844 |
+
"logps/chosen": -277.41375732421875,
|
| 845 |
+
"logps/rejected": -270.2271728515625,
|
| 846 |
+
"loss": 0.4217,
|
| 847 |
+
"rewards/accuracies": 0.8187500238418579,
|
| 848 |
+
"rewards/chosen": -0.5435327887535095,
|
| 849 |
+
"rewards/margins": 1.276086688041687,
|
| 850 |
+
"rewards/rejected": -1.8196194171905518,
|
| 851 |
+
"step": 580
|
| 852 |
+
},
|
| 853 |
+
{
|
| 854 |
+
"epoch": 1.22,
|
| 855 |
+
"learning_rate": 1.3216195569136747e-05,
|
| 856 |
+
"logits/chosen": -2.8047261238098145,
|
| 857 |
+
"logits/rejected": -2.8207616806030273,
|
| 858 |
+
"logps/chosen": -268.05999755859375,
|
| 859 |
+
"logps/rejected": -250.09188842773438,
|
| 860 |
+
"loss": 0.4192,
|
| 861 |
+
"rewards/accuracies": 0.887499988079071,
|
| 862 |
+
"rewards/chosen": -0.2236345261335373,
|
| 863 |
+
"rewards/margins": 1.6592124700546265,
|
| 864 |
+
"rewards/rejected": -1.8828470706939697,
|
| 865 |
+
"step": 590
|
| 866 |
+
},
|
| 867 |
+
{
|
| 868 |
+
"epoch": 1.24,
|
| 869 |
+
"learning_rate": 1.3063407181054241e-05,
|
| 870 |
+
"logits/chosen": -2.8321800231933594,
|
| 871 |
+
"logits/rejected": -2.813894748687744,
|
| 872 |
+
"logps/chosen": -247.3390655517578,
|
| 873 |
+
"logps/rejected": -245.7059326171875,
|
| 874 |
+
"loss": 0.4081,
|
| 875 |
+
"rewards/accuracies": 0.793749988079071,
|
| 876 |
+
"rewards/chosen": -0.6002678871154785,
|
| 877 |
+
"rewards/margins": 1.3836045265197754,
|
| 878 |
+
"rewards/rejected": -1.983872652053833,
|
| 879 |
+
"step": 600
|
| 880 |
+
},
|
| 881 |
+
{
|
| 882 |
+
"epoch": 1.26,
|
| 883 |
+
"learning_rate": 1.2910618792971734e-05,
|
| 884 |
+
"logits/chosen": -2.8052244186401367,
|
| 885 |
+
"logits/rejected": -2.7987864017486572,
|
| 886 |
+
"logps/chosen": -252.0263214111328,
|
| 887 |
+
"logps/rejected": -265.32666015625,
|
| 888 |
+
"loss": 0.4056,
|
| 889 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 890 |
+
"rewards/chosen": -0.32010939717292786,
|
| 891 |
+
"rewards/margins": 1.5006351470947266,
|
| 892 |
+
"rewards/rejected": -1.820744514465332,
|
| 893 |
+
"step": 610
|
| 894 |
+
},
|
| 895 |
+
{
|
| 896 |
+
"epoch": 1.28,
|
| 897 |
+
"learning_rate": 1.2757830404889229e-05,
|
| 898 |
+
"logits/chosen": -2.852949619293213,
|
| 899 |
+
"logits/rejected": -2.810314655303955,
|
| 900 |
+
"logps/chosen": -305.781982421875,
|
| 901 |
+
"logps/rejected": -255.346435546875,
|
| 902 |
+
"loss": 0.44,
|
| 903 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 904 |
+
"rewards/chosen": -0.2824147343635559,
|
| 905 |
+
"rewards/margins": 1.5166089534759521,
|
| 906 |
+
"rewards/rejected": -1.7990238666534424,
|
| 907 |
+
"step": 620
|
| 908 |
+
},
|
| 909 |
+
{
|
| 910 |
+
"epoch": 1.3,
|
| 911 |
+
"learning_rate": 1.2605042016806723e-05,
|
| 912 |
+
"logits/chosen": -2.8488757610321045,
|
| 913 |
+
"logits/rejected": -2.8255667686462402,
|
| 914 |
+
"logps/chosen": -272.29388427734375,
|
| 915 |
+
"logps/rejected": -245.6341552734375,
|
| 916 |
+
"loss": 0.42,
|
| 917 |
+
"rewards/accuracies": 0.800000011920929,
|
| 918 |
+
"rewards/chosen": -0.3796336054801941,
|
| 919 |
+
"rewards/margins": 1.4470628499984741,
|
| 920 |
+
"rewards/rejected": -1.8266966342926025,
|
| 921 |
+
"step": 630
|
| 922 |
+
},
|
| 923 |
+
{
|
| 924 |
+
"epoch": 1.32,
|
| 925 |
+
"learning_rate": 1.2452253628724218e-05,
|
| 926 |
+
"logits/chosen": -2.8429629802703857,
|
| 927 |
+
"logits/rejected": -2.837463140487671,
|
| 928 |
+
"logps/chosen": -262.43292236328125,
|
| 929 |
+
"logps/rejected": -261.91644287109375,
|
| 930 |
+
"loss": 0.4354,
|
| 931 |
+
"rewards/accuracies": 0.831250011920929,
|
| 932 |
+
"rewards/chosen": -0.4664459228515625,
|
| 933 |
+
"rewards/margins": 1.388948917388916,
|
| 934 |
+
"rewards/rejected": -1.855394721031189,
|
| 935 |
+
"step": 640
|
| 936 |
+
},
|
| 937 |
+
{
|
| 938 |
+
"epoch": 1.34,
|
| 939 |
+
"learning_rate": 1.2299465240641712e-05,
|
| 940 |
+
"logits/chosen": -2.829761028289795,
|
| 941 |
+
"logits/rejected": -2.799344062805176,
|
| 942 |
+
"logps/chosen": -266.4898986816406,
|
| 943 |
+
"logps/rejected": -228.8533935546875,
|
| 944 |
+
"loss": 0.4393,
|
| 945 |
+
"rewards/accuracies": 0.762499988079071,
|
| 946 |
+
"rewards/chosen": -0.6215249300003052,
|
| 947 |
+
"rewards/margins": 1.1069129705429077,
|
| 948 |
+
"rewards/rejected": -1.7284377813339233,
|
| 949 |
+
"step": 650
|
| 950 |
+
},
|
| 951 |
+
{
|
| 952 |
+
"epoch": 1.36,
|
| 953 |
+
"learning_rate": 1.2146676852559206e-05,
|
| 954 |
+
"logits/chosen": -2.8082187175750732,
|
| 955 |
+
"logits/rejected": -2.785370111465454,
|
| 956 |
+
"logps/chosen": -272.84552001953125,
|
| 957 |
+
"logps/rejected": -257.3983154296875,
|
| 958 |
+
"loss": 0.417,
|
| 959 |
+
"rewards/accuracies": 0.762499988079071,
|
| 960 |
+
"rewards/chosen": -0.5420676469802856,
|
| 961 |
+
"rewards/margins": 1.0937607288360596,
|
| 962 |
+
"rewards/rejected": -1.6358283758163452,
|
| 963 |
+
"step": 660
|
| 964 |
+
},
|
| 965 |
+
{
|
| 966 |
+
"epoch": 1.38,
|
| 967 |
+
"learning_rate": 1.1993888464476701e-05,
|
| 968 |
+
"logits/chosen": -2.796501636505127,
|
| 969 |
+
"logits/rejected": -2.753633975982666,
|
| 970 |
+
"logps/chosen": -275.73297119140625,
|
| 971 |
+
"logps/rejected": -261.51495361328125,
|
| 972 |
+
"loss": 0.4048,
|
| 973 |
+
"rewards/accuracies": 0.831250011920929,
|
| 974 |
+
"rewards/chosen": -0.6535481214523315,
|
| 975 |
+
"rewards/margins": 1.5823442935943604,
|
| 976 |
+
"rewards/rejected": -2.2358925342559814,
|
| 977 |
+
"step": 670
|
| 978 |
+
},
|
| 979 |
+
{
|
| 980 |
+
"epoch": 1.4,
|
| 981 |
+
"learning_rate": 1.1841100076394194e-05,
|
| 982 |
+
"logits/chosen": -2.8197669982910156,
|
| 983 |
+
"logits/rejected": -2.8018298149108887,
|
| 984 |
+
"logps/chosen": -279.07794189453125,
|
| 985 |
+
"logps/rejected": -264.37908935546875,
|
| 986 |
+
"loss": 0.4476,
|
| 987 |
+
"rewards/accuracies": 0.78125,
|
| 988 |
+
"rewards/chosen": -0.5116819143295288,
|
| 989 |
+
"rewards/margins": 1.3447668552398682,
|
| 990 |
+
"rewards/rejected": -1.856448769569397,
|
| 991 |
+
"step": 680
|
| 992 |
+
},
|
| 993 |
+
{
|
| 994 |
+
"epoch": 1.42,
|
| 995 |
+
"learning_rate": 1.1688311688311688e-05,
|
| 996 |
+
"logits/chosen": -2.844113826751709,
|
| 997 |
+
"logits/rejected": -2.844442844390869,
|
| 998 |
+
"logps/chosen": -252.5892791748047,
|
| 999 |
+
"logps/rejected": -259.8851623535156,
|
| 1000 |
+
"loss": 0.4359,
|
| 1001 |
+
"rewards/accuracies": 0.762499988079071,
|
| 1002 |
+
"rewards/chosen": -0.6136052012443542,
|
| 1003 |
+
"rewards/margins": 1.2718920707702637,
|
| 1004 |
+
"rewards/rejected": -1.8854974508285522,
|
| 1005 |
+
"step": 690
|
| 1006 |
+
},
|
| 1007 |
+
{
|
| 1008 |
+
"epoch": 1.44,
|
| 1009 |
+
"learning_rate": 1.1535523300229183e-05,
|
| 1010 |
+
"logits/chosen": -2.8703908920288086,
|
| 1011 |
+
"logits/rejected": -2.8482773303985596,
|
| 1012 |
+
"logps/chosen": -300.6127624511719,
|
| 1013 |
+
"logps/rejected": -267.99188232421875,
|
| 1014 |
+
"loss": 0.4299,
|
| 1015 |
+
"rewards/accuracies": 0.800000011920929,
|
| 1016 |
+
"rewards/chosen": -0.30981573462486267,
|
| 1017 |
+
"rewards/margins": 1.310162901878357,
|
| 1018 |
+
"rewards/rejected": -1.619978666305542,
|
| 1019 |
+
"step": 700
|
| 1020 |
+
},
|
| 1021 |
+
{
|
| 1022 |
+
"epoch": 1.46,
|
| 1023 |
+
"learning_rate": 1.1382734912146677e-05,
|
| 1024 |
+
"logits/chosen": -2.807915449142456,
|
| 1025 |
+
"logits/rejected": -2.7632646560668945,
|
| 1026 |
+
"logps/chosen": -244.17822265625,
|
| 1027 |
+
"logps/rejected": -210.9264678955078,
|
| 1028 |
+
"loss": 0.4146,
|
| 1029 |
+
"rewards/accuracies": 0.793749988079071,
|
| 1030 |
+
"rewards/chosen": -0.32631585001945496,
|
| 1031 |
+
"rewards/margins": 1.2793928384780884,
|
| 1032 |
+
"rewards/rejected": -1.6057088375091553,
|
| 1033 |
+
"step": 710
|
| 1034 |
+
},
|
| 1035 |
+
{
|
| 1036 |
+
"epoch": 1.48,
|
| 1037 |
+
"learning_rate": 1.1229946524064172e-05,
|
| 1038 |
+
"logits/chosen": -2.7481131553649902,
|
| 1039 |
+
"logits/rejected": -2.7189228534698486,
|
| 1040 |
+
"logps/chosen": -271.4867858886719,
|
| 1041 |
+
"logps/rejected": -256.009521484375,
|
| 1042 |
+
"loss": 0.4123,
|
| 1043 |
+
"rewards/accuracies": 0.8125,
|
| 1044 |
+
"rewards/chosen": -0.31545475125312805,
|
| 1045 |
+
"rewards/margins": 1.4961215257644653,
|
| 1046 |
+
"rewards/rejected": -1.811576247215271,
|
| 1047 |
+
"step": 720
|
| 1048 |
+
},
|
| 1049 |
+
{
|
| 1050 |
+
"epoch": 1.51,
|
| 1051 |
+
"learning_rate": 1.1077158135981668e-05,
|
| 1052 |
+
"logits/chosen": -2.80031681060791,
|
| 1053 |
+
"logits/rejected": -2.7459282875061035,
|
| 1054 |
+
"logps/chosen": -268.36669921875,
|
| 1055 |
+
"logps/rejected": -244.3325958251953,
|
| 1056 |
+
"loss": 0.4444,
|
| 1057 |
+
"rewards/accuracies": 0.768750011920929,
|
| 1058 |
+
"rewards/chosen": -0.3755728006362915,
|
| 1059 |
+
"rewards/margins": 1.379817008972168,
|
| 1060 |
+
"rewards/rejected": -1.755389928817749,
|
| 1061 |
+
"step": 730
|
| 1062 |
+
},
|
| 1063 |
+
{
|
| 1064 |
+
"epoch": 1.53,
|
| 1065 |
+
"learning_rate": 1.0924369747899159e-05,
|
| 1066 |
+
"logits/chosen": -2.809633255004883,
|
| 1067 |
+
"logits/rejected": -2.85695219039917,
|
| 1068 |
+
"logps/chosen": -293.177734375,
|
| 1069 |
+
"logps/rejected": -231.4840087890625,
|
| 1070 |
+
"loss": 0.4198,
|
| 1071 |
+
"rewards/accuracies": 0.800000011920929,
|
| 1072 |
+
"rewards/chosen": -0.2760697901248932,
|
| 1073 |
+
"rewards/margins": 1.443703293800354,
|
| 1074 |
+
"rewards/rejected": -1.7197730541229248,
|
| 1075 |
+
"step": 740
|
| 1076 |
+
},
|
| 1077 |
+
{
|
| 1078 |
+
"epoch": 1.55,
|
| 1079 |
+
"learning_rate": 1.0771581359816653e-05,
|
| 1080 |
+
"logits/chosen": -2.7558960914611816,
|
| 1081 |
+
"logits/rejected": -2.803246021270752,
|
| 1082 |
+
"logps/chosen": -261.3629455566406,
|
| 1083 |
+
"logps/rejected": -246.388671875,
|
| 1084 |
+
"loss": 0.39,
|
| 1085 |
+
"rewards/accuracies": 0.84375,
|
| 1086 |
+
"rewards/chosen": -0.4588204026222229,
|
| 1087 |
+
"rewards/margins": 1.728859305381775,
|
| 1088 |
+
"rewards/rejected": -2.1876797676086426,
|
| 1089 |
+
"step": 750
|
| 1090 |
+
},
|
| 1091 |
+
{
|
| 1092 |
+
"epoch": 1.57,
|
| 1093 |
+
"learning_rate": 1.0618792971734148e-05,
|
| 1094 |
+
"logits/chosen": -2.78913950920105,
|
| 1095 |
+
"logits/rejected": -2.7678422927856445,
|
| 1096 |
+
"logps/chosen": -256.0678405761719,
|
| 1097 |
+
"logps/rejected": -234.4947509765625,
|
| 1098 |
+
"loss": 0.4066,
|
| 1099 |
+
"rewards/accuracies": 0.8125,
|
| 1100 |
+
"rewards/chosen": -0.471910297870636,
|
| 1101 |
+
"rewards/margins": 1.3505375385284424,
|
| 1102 |
+
"rewards/rejected": -1.8224480152130127,
|
| 1103 |
+
"step": 760
|
| 1104 |
+
},
|
| 1105 |
+
{
|
| 1106 |
+
"epoch": 1.59,
|
| 1107 |
+
"learning_rate": 1.0466004583651644e-05,
|
| 1108 |
+
"logits/chosen": -2.808622121810913,
|
| 1109 |
+
"logits/rejected": -2.8128342628479004,
|
| 1110 |
+
"logps/chosen": -263.07305908203125,
|
| 1111 |
+
"logps/rejected": -242.57357788085938,
|
| 1112 |
+
"loss": 0.4032,
|
| 1113 |
+
"rewards/accuracies": 0.824999988079071,
|
| 1114 |
+
"rewards/chosen": -0.24994757771492004,
|
| 1115 |
+
"rewards/margins": 1.3186602592468262,
|
| 1116 |
+
"rewards/rejected": -1.5686078071594238,
|
| 1117 |
+
"step": 770
|
| 1118 |
+
},
|
| 1119 |
+
{
|
| 1120 |
+
"epoch": 1.61,
|
| 1121 |
+
"learning_rate": 1.0313216195569139e-05,
|
| 1122 |
+
"logits/chosen": -2.8431007862091064,
|
| 1123 |
+
"logits/rejected": -2.8500876426696777,
|
| 1124 |
+
"logps/chosen": -289.2240295410156,
|
| 1125 |
+
"logps/rejected": -240.93447875976562,
|
| 1126 |
+
"loss": 0.3989,
|
| 1127 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1128 |
+
"rewards/chosen": -0.1801755428314209,
|
| 1129 |
+
"rewards/margins": 1.5674879550933838,
|
| 1130 |
+
"rewards/rejected": -1.7476632595062256,
|
| 1131 |
+
"step": 780
|
| 1132 |
+
},
|
| 1133 |
+
{
|
| 1134 |
+
"epoch": 1.63,
|
| 1135 |
+
"learning_rate": 1.0160427807486633e-05,
|
| 1136 |
+
"logits/chosen": -2.896554470062256,
|
| 1137 |
+
"logits/rejected": -2.8891143798828125,
|
| 1138 |
+
"logps/chosen": -304.2243957519531,
|
| 1139 |
+
"logps/rejected": -279.3757019042969,
|
| 1140 |
+
"loss": 0.3938,
|
| 1141 |
+
"rewards/accuracies": 0.78125,
|
| 1142 |
+
"rewards/chosen": -0.3372874855995178,
|
| 1143 |
+
"rewards/margins": 1.5321928262710571,
|
| 1144 |
+
"rewards/rejected": -1.8694803714752197,
|
| 1145 |
+
"step": 790
|
| 1146 |
+
},
|
| 1147 |
+
{
|
| 1148 |
+
"epoch": 1.65,
|
| 1149 |
+
"learning_rate": 1.0007639419404128e-05,
|
| 1150 |
+
"logits/chosen": -2.8317697048187256,
|
| 1151 |
+
"logits/rejected": -2.7945523262023926,
|
| 1152 |
+
"logps/chosen": -292.1405029296875,
|
| 1153 |
+
"logps/rejected": -234.8839569091797,
|
| 1154 |
+
"loss": 0.4294,
|
| 1155 |
+
"rewards/accuracies": 0.8125,
|
| 1156 |
+
"rewards/chosen": -0.5130658149719238,
|
| 1157 |
+
"rewards/margins": 1.4586106538772583,
|
| 1158 |
+
"rewards/rejected": -1.9716764688491821,
|
| 1159 |
+
"step": 800
|
| 1160 |
+
},
|
| 1161 |
+
{
|
| 1162 |
+
"epoch": 1.67,
|
| 1163 |
+
"learning_rate": 9.85485103132162e-06,
|
| 1164 |
+
"logits/chosen": -2.875779151916504,
|
| 1165 |
+
"logits/rejected": -2.8391237258911133,
|
| 1166 |
+
"logps/chosen": -298.0716857910156,
|
| 1167 |
+
"logps/rejected": -242.75662231445312,
|
| 1168 |
+
"loss": 0.4371,
|
| 1169 |
+
"rewards/accuracies": 0.831250011920929,
|
| 1170 |
+
"rewards/chosen": -0.41954541206359863,
|
| 1171 |
+
"rewards/margins": 1.413554072380066,
|
| 1172 |
+
"rewards/rejected": -1.833099603652954,
|
| 1173 |
+
"step": 810
|
| 1174 |
+
},
|
| 1175 |
+
{
|
| 1176 |
+
"epoch": 1.69,
|
| 1177 |
+
"learning_rate": 9.702062643239115e-06,
|
| 1178 |
+
"logits/chosen": -2.842196464538574,
|
| 1179 |
+
"logits/rejected": -2.824432849884033,
|
| 1180 |
+
"logps/chosen": -255.82620239257812,
|
| 1181 |
+
"logps/rejected": -256.314453125,
|
| 1182 |
+
"loss": 0.4196,
|
| 1183 |
+
"rewards/accuracies": 0.8125,
|
| 1184 |
+
"rewards/chosen": -0.4787816107273102,
|
| 1185 |
+
"rewards/margins": 1.262540578842163,
|
| 1186 |
+
"rewards/rejected": -1.7413221597671509,
|
| 1187 |
+
"step": 820
|
| 1188 |
+
},
|
| 1189 |
+
{
|
| 1190 |
+
"epoch": 1.71,
|
| 1191 |
+
"learning_rate": 9.54927425515661e-06,
|
| 1192 |
+
"logits/chosen": -2.8649404048919678,
|
| 1193 |
+
"logits/rejected": -2.807797908782959,
|
| 1194 |
+
"logps/chosen": -296.7332458496094,
|
| 1195 |
+
"logps/rejected": -261.4539489746094,
|
| 1196 |
+
"loss": 0.4274,
|
| 1197 |
+
"rewards/accuracies": 0.768750011920929,
|
| 1198 |
+
"rewards/chosen": -0.34954220056533813,
|
| 1199 |
+
"rewards/margins": 1.1616686582565308,
|
| 1200 |
+
"rewards/rejected": -1.5112109184265137,
|
| 1201 |
+
"step": 830
|
| 1202 |
+
},
|
| 1203 |
+
{
|
| 1204 |
+
"epoch": 1.73,
|
| 1205 |
+
"learning_rate": 9.396485867074104e-06,
|
| 1206 |
+
"logits/chosen": -2.8419387340545654,
|
| 1207 |
+
"logits/rejected": -2.7805838584899902,
|
| 1208 |
+
"logps/chosen": -271.58807373046875,
|
| 1209 |
+
"logps/rejected": -239.6482391357422,
|
| 1210 |
+
"loss": 0.4106,
|
| 1211 |
+
"rewards/accuracies": 0.793749988079071,
|
| 1212 |
+
"rewards/chosen": -0.35656026005744934,
|
| 1213 |
+
"rewards/margins": 1.2432340383529663,
|
| 1214 |
+
"rewards/rejected": -1.5997945070266724,
|
| 1215 |
+
"step": 840
|
| 1216 |
+
},
|
| 1217 |
+
{
|
| 1218 |
+
"epoch": 1.75,
|
| 1219 |
+
"learning_rate": 9.243697478991598e-06,
|
| 1220 |
+
"logits/chosen": -2.851105213165283,
|
| 1221 |
+
"logits/rejected": -2.8537163734436035,
|
| 1222 |
+
"logps/chosen": -285.19293212890625,
|
| 1223 |
+
"logps/rejected": -249.2686004638672,
|
| 1224 |
+
"loss": 0.4067,
|
| 1225 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 1226 |
+
"rewards/chosen": -0.4058365225791931,
|
| 1227 |
+
"rewards/margins": 1.1938354969024658,
|
| 1228 |
+
"rewards/rejected": -1.5996720790863037,
|
| 1229 |
+
"step": 850
|
| 1230 |
+
},
|
| 1231 |
+
{
|
| 1232 |
+
"epoch": 1.77,
|
| 1233 |
+
"learning_rate": 9.090909090909091e-06,
|
| 1234 |
+
"logits/chosen": -2.8415451049804688,
|
| 1235 |
+
"logits/rejected": -2.816315174102783,
|
| 1236 |
+
"logps/chosen": -246.9827117919922,
|
| 1237 |
+
"logps/rejected": -244.11172485351562,
|
| 1238 |
+
"loss": 0.4166,
|
| 1239 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 1240 |
+
"rewards/chosen": -0.5251684784889221,
|
| 1241 |
+
"rewards/margins": 1.321463942527771,
|
| 1242 |
+
"rewards/rejected": -1.8466323614120483,
|
| 1243 |
+
"step": 860
|
| 1244 |
+
},
|
| 1245 |
+
{
|
| 1246 |
+
"epoch": 1.79,
|
| 1247 |
+
"learning_rate": 8.938120702826586e-06,
|
| 1248 |
+
"logits/chosen": -2.8182404041290283,
|
| 1249 |
+
"logits/rejected": -2.7850711345672607,
|
| 1250 |
+
"logps/chosen": -320.24407958984375,
|
| 1251 |
+
"logps/rejected": -251.82949829101562,
|
| 1252 |
+
"loss": 0.409,
|
| 1253 |
+
"rewards/accuracies": 0.78125,
|
| 1254 |
+
"rewards/chosen": -0.2882133424282074,
|
| 1255 |
+
"rewards/margins": 1.439748764038086,
|
| 1256 |
+
"rewards/rejected": -1.7279622554779053,
|
| 1257 |
+
"step": 870
|
| 1258 |
+
},
|
| 1259 |
+
{
|
| 1260 |
+
"epoch": 1.81,
|
| 1261 |
+
"learning_rate": 8.78533231474408e-06,
|
| 1262 |
+
"logits/chosen": -2.859550952911377,
|
| 1263 |
+
"logits/rejected": -2.8387062549591064,
|
| 1264 |
+
"logps/chosen": -251.30764770507812,
|
| 1265 |
+
"logps/rejected": -253.18179321289062,
|
| 1266 |
+
"loss": 0.4071,
|
| 1267 |
+
"rewards/accuracies": 0.793749988079071,
|
| 1268 |
+
"rewards/chosen": -0.4950861930847168,
|
| 1269 |
+
"rewards/margins": 1.2707871198654175,
|
| 1270 |
+
"rewards/rejected": -1.7658733129501343,
|
| 1271 |
+
"step": 880
|
| 1272 |
+
},
|
| 1273 |
+
{
|
| 1274 |
+
"epoch": 1.84,
|
| 1275 |
+
"learning_rate": 8.632543926661574e-06,
|
| 1276 |
+
"logits/chosen": -2.8016982078552246,
|
| 1277 |
+
"logits/rejected": -2.7528939247131348,
|
| 1278 |
+
"logps/chosen": -277.7475891113281,
|
| 1279 |
+
"logps/rejected": -262.52215576171875,
|
| 1280 |
+
"loss": 0.444,
|
| 1281 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 1282 |
+
"rewards/chosen": -0.487938791513443,
|
| 1283 |
+
"rewards/margins": 1.4403693675994873,
|
| 1284 |
+
"rewards/rejected": -1.928308129310608,
|
| 1285 |
+
"step": 890
|
| 1286 |
+
},
|
| 1287 |
+
{
|
| 1288 |
+
"epoch": 1.86,
|
| 1289 |
+
"learning_rate": 8.479755538579069e-06,
|
| 1290 |
+
"logits/chosen": -2.832099199295044,
|
| 1291 |
+
"logits/rejected": -2.786726951599121,
|
| 1292 |
+
"logps/chosen": -232.41787719726562,
|
| 1293 |
+
"logps/rejected": -241.41775512695312,
|
| 1294 |
+
"loss": 0.4149,
|
| 1295 |
+
"rewards/accuracies": 0.793749988079071,
|
| 1296 |
+
"rewards/chosen": -0.5055242776870728,
|
| 1297 |
+
"rewards/margins": 1.4381215572357178,
|
| 1298 |
+
"rewards/rejected": -1.9436458349227905,
|
| 1299 |
+
"step": 900
|
| 1300 |
+
},
|
| 1301 |
+
{
|
| 1302 |
+
"epoch": 1.88,
|
| 1303 |
+
"learning_rate": 8.326967150496563e-06,
|
| 1304 |
+
"logits/chosen": -2.793572425842285,
|
| 1305 |
+
"logits/rejected": -2.8044943809509277,
|
| 1306 |
+
"logps/chosen": -293.1367492675781,
|
| 1307 |
+
"logps/rejected": -241.0701904296875,
|
| 1308 |
+
"loss": 0.4144,
|
| 1309 |
+
"rewards/accuracies": 0.793749988079071,
|
| 1310 |
+
"rewards/chosen": -0.40146392583847046,
|
| 1311 |
+
"rewards/margins": 1.4329584836959839,
|
| 1312 |
+
"rewards/rejected": -1.8344223499298096,
|
| 1313 |
+
"step": 910
|
| 1314 |
+
},
|
| 1315 |
+
{
|
| 1316 |
+
"epoch": 1.9,
|
| 1317 |
+
"learning_rate": 8.174178762414056e-06,
|
| 1318 |
+
"logits/chosen": -2.861027240753174,
|
| 1319 |
+
"logits/rejected": -2.802192449569702,
|
| 1320 |
+
"logps/chosen": -302.29095458984375,
|
| 1321 |
+
"logps/rejected": -264.388916015625,
|
| 1322 |
+
"loss": 0.3921,
|
| 1323 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 1324 |
+
"rewards/chosen": -0.377704381942749,
|
| 1325 |
+
"rewards/margins": 1.487374186515808,
|
| 1326 |
+
"rewards/rejected": -1.8650786876678467,
|
| 1327 |
+
"step": 920
|
| 1328 |
+
},
|
| 1329 |
+
{
|
| 1330 |
+
"epoch": 1.92,
|
| 1331 |
+
"learning_rate": 8.02139037433155e-06,
|
| 1332 |
+
"logits/chosen": -2.856374502182007,
|
| 1333 |
+
"logits/rejected": -2.875044584274292,
|
| 1334 |
+
"logps/chosen": -282.31585693359375,
|
| 1335 |
+
"logps/rejected": -235.76748657226562,
|
| 1336 |
+
"loss": 0.4205,
|
| 1337 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 1338 |
+
"rewards/chosen": -0.39734259247779846,
|
| 1339 |
+
"rewards/margins": 1.3766673803329468,
|
| 1340 |
+
"rewards/rejected": -1.7740100622177124,
|
| 1341 |
+
"step": 930
|
| 1342 |
+
},
|
| 1343 |
+
{
|
| 1344 |
+
"epoch": 1.94,
|
| 1345 |
+
"learning_rate": 7.868601986249045e-06,
|
| 1346 |
+
"logits/chosen": -2.8075997829437256,
|
| 1347 |
+
"logits/rejected": -2.77669620513916,
|
| 1348 |
+
"logps/chosen": -256.1700134277344,
|
| 1349 |
+
"logps/rejected": -247.21896362304688,
|
| 1350 |
+
"loss": 0.4157,
|
| 1351 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 1352 |
+
"rewards/chosen": -0.3420465886592865,
|
| 1353 |
+
"rewards/margins": 1.5288991928100586,
|
| 1354 |
+
"rewards/rejected": -1.8709455728530884,
|
| 1355 |
+
"step": 940
|
| 1356 |
+
},
|
| 1357 |
+
{
|
| 1358 |
+
"epoch": 1.96,
|
| 1359 |
+
"learning_rate": 7.71581359816654e-06,
|
| 1360 |
+
"logits/chosen": -2.8214058876037598,
|
| 1361 |
+
"logits/rejected": -2.812399387359619,
|
| 1362 |
+
"logps/chosen": -271.9144592285156,
|
| 1363 |
+
"logps/rejected": -248.6433563232422,
|
| 1364 |
+
"loss": 0.4086,
|
| 1365 |
+
"rewards/accuracies": 0.8187500238418579,
|
| 1366 |
+
"rewards/chosen": -0.5668919682502747,
|
| 1367 |
+
"rewards/margins": 1.462934970855713,
|
| 1368 |
+
"rewards/rejected": -2.0298266410827637,
|
| 1369 |
+
"step": 950
|
| 1370 |
+
},
|
| 1371 |
+
{
|
| 1372 |
+
"epoch": 1.98,
|
| 1373 |
+
"learning_rate": 7.563025210084034e-06,
|
| 1374 |
+
"logits/chosen": -2.855318069458008,
|
| 1375 |
+
"logits/rejected": -2.8766427040100098,
|
| 1376 |
+
"logps/chosen": -305.3939514160156,
|
| 1377 |
+
"logps/rejected": -259.01739501953125,
|
| 1378 |
+
"loss": 0.4274,
|
| 1379 |
+
"rewards/accuracies": 0.8125,
|
| 1380 |
+
"rewards/chosen": -0.4556516706943512,
|
| 1381 |
+
"rewards/margins": 1.3332924842834473,
|
| 1382 |
+
"rewards/rejected": -1.7889440059661865,
|
| 1383 |
+
"step": 960
|
| 1384 |
+
},
|
| 1385 |
+
{
|
| 1386 |
+
"epoch": 2.0,
|
| 1387 |
+
"learning_rate": 7.410236822001529e-06,
|
| 1388 |
+
"logits/chosen": -2.804309368133545,
|
| 1389 |
+
"logits/rejected": -2.7984023094177246,
|
| 1390 |
+
"logps/chosen": -269.3817443847656,
|
| 1391 |
+
"logps/rejected": -236.9586639404297,
|
| 1392 |
+
"loss": 0.4124,
|
| 1393 |
+
"rewards/accuracies": 0.824999988079071,
|
| 1394 |
+
"rewards/chosen": -0.4466172754764557,
|
| 1395 |
+
"rewards/margins": 1.4143751859664917,
|
| 1396 |
+
"rewards/rejected": -1.860992431640625,
|
| 1397 |
+
"step": 970
|
| 1398 |
+
},
|
| 1399 |
+
{
|
| 1400 |
+
"epoch": 2.0,
|
| 1401 |
+
"eval_logits/chosen": -2.865886688232422,
|
| 1402 |
+
"eval_logits/rejected": -2.8562960624694824,
|
| 1403 |
+
"eval_logps/chosen": -277.1995849609375,
|
| 1404 |
+
"eval_logps/rejected": -242.37806701660156,
|
| 1405 |
+
"eval_loss": 0.4544542133808136,
|
| 1406 |
+
"eval_rewards/accuracies": 0.81640625,
|
| 1407 |
+
"eval_rewards/chosen": -0.3551396429538727,
|
| 1408 |
+
"eval_rewards/margins": 1.4038398265838623,
|
| 1409 |
+
"eval_rewards/rejected": -1.7589795589447021,
|
| 1410 |
+
"eval_runtime": 253.9486,
|
| 1411 |
+
"eval_samples_per_second": 7.876,
|
| 1412 |
+
"eval_steps_per_second": 0.063,
|
| 1413 |
+
"step": 970
|
| 1414 |
+
},
|
| 1415 |
+
{
|
| 1416 |
+
"epoch": 2.02,
|
| 1417 |
+
"learning_rate": 7.257448433919023e-06,
|
| 1418 |
+
"logits/chosen": -2.8543519973754883,
|
| 1419 |
+
"logits/rejected": -2.8081583976745605,
|
| 1420 |
+
"logps/chosen": -279.414306640625,
|
| 1421 |
+
"logps/rejected": -276.77752685546875,
|
| 1422 |
+
"loss": 0.3522,
|
| 1423 |
+
"rewards/accuracies": 0.8187500238418579,
|
| 1424 |
+
"rewards/chosen": -0.36832618713378906,
|
| 1425 |
+
"rewards/margins": 1.5619885921478271,
|
| 1426 |
+
"rewards/rejected": -1.9303147792816162,
|
| 1427 |
+
"step": 980
|
| 1428 |
+
},
|
| 1429 |
+
{
|
| 1430 |
+
"epoch": 2.04,
|
| 1431 |
+
"learning_rate": 7.104660045836517e-06,
|
| 1432 |
+
"logits/chosen": -2.8730249404907227,
|
| 1433 |
+
"logits/rejected": -2.8948845863342285,
|
| 1434 |
+
"logps/chosen": -318.85736083984375,
|
| 1435 |
+
"logps/rejected": -257.4644470214844,
|
| 1436 |
+
"loss": 0.3859,
|
| 1437 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 1438 |
+
"rewards/chosen": -0.29034391045570374,
|
| 1439 |
+
"rewards/margins": 1.8795220851898193,
|
| 1440 |
+
"rewards/rejected": -2.1698660850524902,
|
| 1441 |
+
"step": 990
|
| 1442 |
+
},
|
| 1443 |
+
{
|
| 1444 |
+
"epoch": 2.06,
|
| 1445 |
+
"learning_rate": 6.951871657754011e-06,
|
| 1446 |
+
"logits/chosen": -2.8562302589416504,
|
| 1447 |
+
"logits/rejected": -2.834265947341919,
|
| 1448 |
+
"logps/chosen": -259.0556640625,
|
| 1449 |
+
"logps/rejected": -247.9607696533203,
|
| 1450 |
+
"loss": 0.3666,
|
| 1451 |
+
"rewards/accuracies": 0.831250011920929,
|
| 1452 |
+
"rewards/chosen": -0.5741233229637146,
|
| 1453 |
+
"rewards/margins": 1.5469454526901245,
|
| 1454 |
+
"rewards/rejected": -2.1210689544677734,
|
| 1455 |
+
"step": 1000
|
| 1456 |
+
},
|
| 1457 |
+
{
|
| 1458 |
+
"epoch": 2.08,
|
| 1459 |
+
"learning_rate": 6.799083269671506e-06,
|
| 1460 |
+
"logits/chosen": -2.905282974243164,
|
| 1461 |
+
"logits/rejected": -2.836683988571167,
|
| 1462 |
+
"logps/chosen": -293.37139892578125,
|
| 1463 |
+
"logps/rejected": -253.64810180664062,
|
| 1464 |
+
"loss": 0.3664,
|
| 1465 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 1466 |
+
"rewards/chosen": -0.5017197728157043,
|
| 1467 |
+
"rewards/margins": 1.8150691986083984,
|
| 1468 |
+
"rewards/rejected": -2.316789150238037,
|
| 1469 |
+
"step": 1010
|
| 1470 |
+
},
|
| 1471 |
+
{
|
| 1472 |
+
"epoch": 2.1,
|
| 1473 |
+
"learning_rate": 6.646294881588999e-06,
|
| 1474 |
+
"logits/chosen": -2.8647093772888184,
|
| 1475 |
+
"logits/rejected": -2.8602373600006104,
|
| 1476 |
+
"logps/chosen": -253.71450805664062,
|
| 1477 |
+
"logps/rejected": -248.3654022216797,
|
| 1478 |
+
"loss": 0.369,
|
| 1479 |
+
"rewards/accuracies": 0.831250011920929,
|
| 1480 |
+
"rewards/chosen": -0.581963300704956,
|
| 1481 |
+
"rewards/margins": 1.589311957359314,
|
| 1482 |
+
"rewards/rejected": -2.1712751388549805,
|
| 1483 |
+
"step": 1020
|
| 1484 |
+
},
|
| 1485 |
+
{
|
| 1486 |
+
"epoch": 2.12,
|
| 1487 |
+
"learning_rate": 6.493506493506494e-06,
|
| 1488 |
+
"logits/chosen": -2.837463617324829,
|
| 1489 |
+
"logits/rejected": -2.817621946334839,
|
| 1490 |
+
"logps/chosen": -285.40142822265625,
|
| 1491 |
+
"logps/rejected": -252.75448608398438,
|
| 1492 |
+
"loss": 0.354,
|
| 1493 |
+
"rewards/accuracies": 0.84375,
|
| 1494 |
+
"rewards/chosen": -0.5188383460044861,
|
| 1495 |
+
"rewards/margins": 1.7174046039581299,
|
| 1496 |
+
"rewards/rejected": -2.2362427711486816,
|
| 1497 |
+
"step": 1030
|
| 1498 |
+
},
|
| 1499 |
+
{
|
| 1500 |
+
"epoch": 2.14,
|
| 1501 |
+
"learning_rate": 6.340718105423988e-06,
|
| 1502 |
+
"logits/chosen": -2.901369571685791,
|
| 1503 |
+
"logits/rejected": -2.8450589179992676,
|
| 1504 |
+
"logps/chosen": -270.55145263671875,
|
| 1505 |
+
"logps/rejected": -257.68060302734375,
|
| 1506 |
+
"loss": 0.3873,
|
| 1507 |
+
"rewards/accuracies": 0.8687499761581421,
|
| 1508 |
+
"rewards/chosen": -0.28947222232818604,
|
| 1509 |
+
"rewards/margins": 1.6885324716567993,
|
| 1510 |
+
"rewards/rejected": -1.9780044555664062,
|
| 1511 |
+
"step": 1040
|
| 1512 |
+
},
|
| 1513 |
+
{
|
| 1514 |
+
"epoch": 2.16,
|
| 1515 |
+
"learning_rate": 6.187929717341482e-06,
|
| 1516 |
+
"logits/chosen": -2.8084139823913574,
|
| 1517 |
+
"logits/rejected": -2.83860445022583,
|
| 1518 |
+
"logps/chosen": -272.42822265625,
|
| 1519 |
+
"logps/rejected": -243.85299682617188,
|
| 1520 |
+
"loss": 0.3468,
|
| 1521 |
+
"rewards/accuracies": 0.8125,
|
| 1522 |
+
"rewards/chosen": -0.4739387035369873,
|
| 1523 |
+
"rewards/margins": 1.55910325050354,
|
| 1524 |
+
"rewards/rejected": -2.0330419540405273,
|
| 1525 |
+
"step": 1050
|
| 1526 |
+
},
|
| 1527 |
+
{
|
| 1528 |
+
"epoch": 2.19,
|
| 1529 |
+
"learning_rate": 6.0351413292589764e-06,
|
| 1530 |
+
"logits/chosen": -2.8273186683654785,
|
| 1531 |
+
"logits/rejected": -2.8402295112609863,
|
| 1532 |
+
"logps/chosen": -282.76422119140625,
|
| 1533 |
+
"logps/rejected": -263.18548583984375,
|
| 1534 |
+
"loss": 0.3782,
|
| 1535 |
+
"rewards/accuracies": 0.875,
|
| 1536 |
+
"rewards/chosen": -0.4047788083553314,
|
| 1537 |
+
"rewards/margins": 1.8494300842285156,
|
| 1538 |
+
"rewards/rejected": -2.254209041595459,
|
| 1539 |
+
"step": 1060
|
| 1540 |
+
},
|
| 1541 |
+
{
|
| 1542 |
+
"epoch": 2.21,
|
| 1543 |
+
"learning_rate": 5.882352941176471e-06,
|
| 1544 |
+
"logits/chosen": -2.8585734367370605,
|
| 1545 |
+
"logits/rejected": -2.8870089054107666,
|
| 1546 |
+
"logps/chosen": -242.56851196289062,
|
| 1547 |
+
"logps/rejected": -221.9137725830078,
|
| 1548 |
+
"loss": 0.3691,
|
| 1549 |
+
"rewards/accuracies": 0.762499988079071,
|
| 1550 |
+
"rewards/chosen": -0.5514971017837524,
|
| 1551 |
+
"rewards/margins": 1.47260582447052,
|
| 1552 |
+
"rewards/rejected": -2.0241026878356934,
|
| 1553 |
+
"step": 1070
|
| 1554 |
+
},
|
| 1555 |
+
{
|
| 1556 |
+
"epoch": 2.23,
|
| 1557 |
+
"learning_rate": 5.729564553093966e-06,
|
| 1558 |
+
"logits/chosen": -2.8390653133392334,
|
| 1559 |
+
"logits/rejected": -2.827547788619995,
|
| 1560 |
+
"logps/chosen": -336.2854309082031,
|
| 1561 |
+
"logps/rejected": -258.2242126464844,
|
| 1562 |
+
"loss": 0.373,
|
| 1563 |
+
"rewards/accuracies": 0.831250011920929,
|
| 1564 |
+
"rewards/chosen": -0.47643327713012695,
|
| 1565 |
+
"rewards/margins": 1.622065544128418,
|
| 1566 |
+
"rewards/rejected": -2.098498821258545,
|
| 1567 |
+
"step": 1080
|
| 1568 |
+
},
|
| 1569 |
+
{
|
| 1570 |
+
"epoch": 2.25,
|
| 1571 |
+
"learning_rate": 5.576776165011459e-06,
|
| 1572 |
+
"logits/chosen": -2.864114284515381,
|
| 1573 |
+
"logits/rejected": -2.8293545246124268,
|
| 1574 |
+
"logps/chosen": -273.42315673828125,
|
| 1575 |
+
"logps/rejected": -267.7647399902344,
|
| 1576 |
+
"loss": 0.3652,
|
| 1577 |
+
"rewards/accuracies": 0.8187500238418579,
|
| 1578 |
+
"rewards/chosen": -0.6307514905929565,
|
| 1579 |
+
"rewards/margins": 1.7940418720245361,
|
| 1580 |
+
"rewards/rejected": -2.424793243408203,
|
| 1581 |
+
"step": 1090
|
| 1582 |
+
},
|
| 1583 |
+
{
|
| 1584 |
+
"epoch": 2.27,
|
| 1585 |
+
"learning_rate": 5.423987776928954e-06,
|
| 1586 |
+
"logits/chosen": -2.847562313079834,
|
| 1587 |
+
"logits/rejected": -2.8787343502044678,
|
| 1588 |
+
"logps/chosen": -271.5540771484375,
|
| 1589 |
+
"logps/rejected": -237.20877075195312,
|
| 1590 |
+
"loss": 0.3705,
|
| 1591 |
+
"rewards/accuracies": 0.8187500238418579,
|
| 1592 |
+
"rewards/chosen": -0.7378746271133423,
|
| 1593 |
+
"rewards/margins": 1.611467719078064,
|
| 1594 |
+
"rewards/rejected": -2.3493425846099854,
|
| 1595 |
+
"step": 1100
|
| 1596 |
+
},
|
| 1597 |
+
{
|
| 1598 |
+
"epoch": 2.29,
|
| 1599 |
+
"learning_rate": 5.271199388846449e-06,
|
| 1600 |
+
"logits/chosen": -2.844463348388672,
|
| 1601 |
+
"logits/rejected": -2.7723591327667236,
|
| 1602 |
+
"logps/chosen": -289.79632568359375,
|
| 1603 |
+
"logps/rejected": -266.92529296875,
|
| 1604 |
+
"loss": 0.3637,
|
| 1605 |
+
"rewards/accuracies": 0.856249988079071,
|
| 1606 |
+
"rewards/chosen": -0.6152908205986023,
|
| 1607 |
+
"rewards/margins": 1.741233229637146,
|
| 1608 |
+
"rewards/rejected": -2.3565244674682617,
|
| 1609 |
+
"step": 1110
|
| 1610 |
+
},
|
| 1611 |
+
{
|
| 1612 |
+
"epoch": 2.31,
|
| 1613 |
+
"learning_rate": 5.118411000763942e-06,
|
| 1614 |
+
"logits/chosen": -2.8476827144622803,
|
| 1615 |
+
"logits/rejected": -2.8521735668182373,
|
| 1616 |
+
"logps/chosen": -273.8808288574219,
|
| 1617 |
+
"logps/rejected": -241.986572265625,
|
| 1618 |
+
"loss": 0.3575,
|
| 1619 |
+
"rewards/accuracies": 0.831250011920929,
|
| 1620 |
+
"rewards/chosen": -0.7439392805099487,
|
| 1621 |
+
"rewards/margins": 1.6854461431503296,
|
| 1622 |
+
"rewards/rejected": -2.4293856620788574,
|
| 1623 |
+
"step": 1120
|
| 1624 |
+
},
|
| 1625 |
+
{
|
| 1626 |
+
"epoch": 2.33,
|
| 1627 |
+
"learning_rate": 4.965622612681437e-06,
|
| 1628 |
+
"logits/chosen": -2.8075308799743652,
|
| 1629 |
+
"logits/rejected": -2.7614858150482178,
|
| 1630 |
+
"logps/chosen": -285.82550048828125,
|
| 1631 |
+
"logps/rejected": -258.968994140625,
|
| 1632 |
+
"loss": 0.3651,
|
| 1633 |
+
"rewards/accuracies": 0.856249988079071,
|
| 1634 |
+
"rewards/chosen": -0.7141876220703125,
|
| 1635 |
+
"rewards/margins": 1.7358572483062744,
|
| 1636 |
+
"rewards/rejected": -2.450045108795166,
|
| 1637 |
+
"step": 1130
|
| 1638 |
+
},
|
| 1639 |
+
{
|
| 1640 |
+
"epoch": 2.35,
|
| 1641 |
+
"learning_rate": 4.812834224598931e-06,
|
| 1642 |
+
"logits/chosen": -2.8139889240264893,
|
| 1643 |
+
"logits/rejected": -2.8023717403411865,
|
| 1644 |
+
"logps/chosen": -304.9195251464844,
|
| 1645 |
+
"logps/rejected": -261.23919677734375,
|
| 1646 |
+
"loss": 0.3568,
|
| 1647 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 1648 |
+
"rewards/chosen": -0.511367678642273,
|
| 1649 |
+
"rewards/margins": 1.4749786853790283,
|
| 1650 |
+
"rewards/rejected": -1.9863464832305908,
|
| 1651 |
+
"step": 1140
|
| 1652 |
+
},
|
| 1653 |
+
{
|
| 1654 |
+
"epoch": 2.37,
|
| 1655 |
+
"learning_rate": 4.660045836516425e-06,
|
| 1656 |
+
"logits/chosen": -2.7972934246063232,
|
| 1657 |
+
"logits/rejected": -2.7768425941467285,
|
| 1658 |
+
"logps/chosen": -277.6763916015625,
|
| 1659 |
+
"logps/rejected": -251.0581817626953,
|
| 1660 |
+
"loss": 0.3389,
|
| 1661 |
+
"rewards/accuracies": 0.800000011920929,
|
| 1662 |
+
"rewards/chosen": -0.7810333967208862,
|
| 1663 |
+
"rewards/margins": 1.623008370399475,
|
| 1664 |
+
"rewards/rejected": -2.4040417671203613,
|
| 1665 |
+
"step": 1150
|
| 1666 |
+
},
|
| 1667 |
+
{
|
| 1668 |
+
"epoch": 2.39,
|
| 1669 |
+
"learning_rate": 4.5072574484339196e-06,
|
| 1670 |
+
"logits/chosen": -2.842956781387329,
|
| 1671 |
+
"logits/rejected": -2.8352267742156982,
|
| 1672 |
+
"logps/chosen": -246.30056762695312,
|
| 1673 |
+
"logps/rejected": -273.0400390625,
|
| 1674 |
+
"loss": 0.3685,
|
| 1675 |
+
"rewards/accuracies": 0.800000011920929,
|
| 1676 |
+
"rewards/chosen": -0.6348632574081421,
|
| 1677 |
+
"rewards/margins": 1.7184776067733765,
|
| 1678 |
+
"rewards/rejected": -2.3533406257629395,
|
| 1679 |
+
"step": 1160
|
| 1680 |
+
},
|
| 1681 |
+
{
|
| 1682 |
+
"epoch": 2.41,
|
| 1683 |
+
"learning_rate": 4.354469060351414e-06,
|
| 1684 |
+
"logits/chosen": -2.801945447921753,
|
| 1685 |
+
"logits/rejected": -2.791245937347412,
|
| 1686 |
+
"logps/chosen": -279.83709716796875,
|
| 1687 |
+
"logps/rejected": -284.6191711425781,
|
| 1688 |
+
"loss": 0.3482,
|
| 1689 |
+
"rewards/accuracies": 0.831250011920929,
|
| 1690 |
+
"rewards/chosen": -0.7146093249320984,
|
| 1691 |
+
"rewards/margins": 1.7736566066741943,
|
| 1692 |
+
"rewards/rejected": -2.4882664680480957,
|
| 1693 |
+
"step": 1170
|
| 1694 |
+
},
|
| 1695 |
+
{
|
| 1696 |
+
"epoch": 2.43,
|
| 1697 |
+
"learning_rate": 4.201680672268908e-06,
|
| 1698 |
+
"logits/chosen": -2.752314329147339,
|
| 1699 |
+
"logits/rejected": -2.751218318939209,
|
| 1700 |
+
"logps/chosen": -285.19122314453125,
|
| 1701 |
+
"logps/rejected": -262.41162109375,
|
| 1702 |
+
"loss": 0.3726,
|
| 1703 |
+
"rewards/accuracies": 0.824999988079071,
|
| 1704 |
+
"rewards/chosen": -0.7081433534622192,
|
| 1705 |
+
"rewards/margins": 1.631823182106018,
|
| 1706 |
+
"rewards/rejected": -2.339966297149658,
|
| 1707 |
+
"step": 1180
|
| 1708 |
+
},
|
| 1709 |
+
{
|
| 1710 |
+
"epoch": 2.45,
|
| 1711 |
+
"learning_rate": 4.048892284186402e-06,
|
| 1712 |
+
"logits/chosen": -2.7520546913146973,
|
| 1713 |
+
"logits/rejected": -2.706789493560791,
|
| 1714 |
+
"logps/chosen": -274.19903564453125,
|
| 1715 |
+
"logps/rejected": -271.12530517578125,
|
| 1716 |
+
"loss": 0.37,
|
| 1717 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 1718 |
+
"rewards/chosen": -0.5612797737121582,
|
| 1719 |
+
"rewards/margins": 1.6494731903076172,
|
| 1720 |
+
"rewards/rejected": -2.2107529640197754,
|
| 1721 |
+
"step": 1190
|
| 1722 |
+
},
|
| 1723 |
+
{
|
| 1724 |
+
"epoch": 2.47,
|
| 1725 |
+
"learning_rate": 3.896103896103897e-06,
|
| 1726 |
+
"logits/chosen": -2.8326988220214844,
|
| 1727 |
+
"logits/rejected": -2.7695236206054688,
|
| 1728 |
+
"logps/chosen": -272.9900817871094,
|
| 1729 |
+
"logps/rejected": -257.9608154296875,
|
| 1730 |
+
"loss": 0.3775,
|
| 1731 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 1732 |
+
"rewards/chosen": -0.7353371381759644,
|
| 1733 |
+
"rewards/margins": 1.5964215993881226,
|
| 1734 |
+
"rewards/rejected": -2.331758737564087,
|
| 1735 |
+
"step": 1200
|
| 1736 |
+
},
|
| 1737 |
+
{
|
| 1738 |
+
"epoch": 2.49,
|
| 1739 |
+
"learning_rate": 3.7433155080213907e-06,
|
| 1740 |
+
"logits/chosen": -2.7996091842651367,
|
| 1741 |
+
"logits/rejected": -2.8057217597961426,
|
| 1742 |
+
"logps/chosen": -290.14324951171875,
|
| 1743 |
+
"logps/rejected": -257.6935119628906,
|
| 1744 |
+
"loss": 0.3824,
|
| 1745 |
+
"rewards/accuracies": 0.862500011920929,
|
| 1746 |
+
"rewards/chosen": -0.5703123807907104,
|
| 1747 |
+
"rewards/margins": 1.7026172876358032,
|
| 1748 |
+
"rewards/rejected": -2.2729296684265137,
|
| 1749 |
+
"step": 1210
|
| 1750 |
+
},
|
| 1751 |
+
{
|
| 1752 |
+
"epoch": 2.52,
|
| 1753 |
+
"learning_rate": 3.5905271199388848e-06,
|
| 1754 |
+
"logits/chosen": -2.834031343460083,
|
| 1755 |
+
"logits/rejected": -2.8236374855041504,
|
| 1756 |
+
"logps/chosen": -279.32598876953125,
|
| 1757 |
+
"logps/rejected": -250.78292846679688,
|
| 1758 |
+
"loss": 0.376,
|
| 1759 |
+
"rewards/accuracies": 0.824999988079071,
|
| 1760 |
+
"rewards/chosen": -0.5887583494186401,
|
| 1761 |
+
"rewards/margins": 1.6018693447113037,
|
| 1762 |
+
"rewards/rejected": -2.1906275749206543,
|
| 1763 |
+
"step": 1220
|
| 1764 |
+
},
|
| 1765 |
+
{
|
| 1766 |
+
"epoch": 2.54,
|
| 1767 |
+
"learning_rate": 3.4377387318563792e-06,
|
| 1768 |
+
"logits/chosen": -2.8245434761047363,
|
| 1769 |
+
"logits/rejected": -2.852121353149414,
|
| 1770 |
+
"logps/chosen": -323.82940673828125,
|
| 1771 |
+
"logps/rejected": -270.3510437011719,
|
| 1772 |
+
"loss": 0.3703,
|
| 1773 |
+
"rewards/accuracies": 0.8062499761581421,
|
| 1774 |
+
"rewards/chosen": -0.7247987985610962,
|
| 1775 |
+
"rewards/margins": 1.512142539024353,
|
| 1776 |
+
"rewards/rejected": -2.236941337585449,
|
| 1777 |
+
"step": 1230
|
| 1778 |
+
},
|
| 1779 |
+
{
|
| 1780 |
+
"epoch": 2.56,
|
| 1781 |
+
"learning_rate": 3.2849503437738733e-06,
|
| 1782 |
+
"logits/chosen": -2.842057466506958,
|
| 1783 |
+
"logits/rejected": -2.855686902999878,
|
| 1784 |
+
"logps/chosen": -275.85418701171875,
|
| 1785 |
+
"logps/rejected": -251.3174591064453,
|
| 1786 |
+
"loss": 0.3878,
|
| 1787 |
+
"rewards/accuracies": 0.84375,
|
| 1788 |
+
"rewards/chosen": -0.4698792099952698,
|
| 1789 |
+
"rewards/margins": 1.7577711343765259,
|
| 1790 |
+
"rewards/rejected": -2.2276504039764404,
|
| 1791 |
+
"step": 1240
|
| 1792 |
+
},
|
| 1793 |
+
{
|
| 1794 |
+
"epoch": 2.58,
|
| 1795 |
+
"learning_rate": 3.1321619556913678e-06,
|
| 1796 |
+
"logits/chosen": -2.842163562774658,
|
| 1797 |
+
"logits/rejected": -2.8198578357696533,
|
| 1798 |
+
"logps/chosen": -293.8838195800781,
|
| 1799 |
+
"logps/rejected": -260.05194091796875,
|
| 1800 |
+
"loss": 0.3441,
|
| 1801 |
+
"rewards/accuracies": 0.84375,
|
| 1802 |
+
"rewards/chosen": -0.48028483986854553,
|
| 1803 |
+
"rewards/margins": 1.8665335178375244,
|
| 1804 |
+
"rewards/rejected": -2.346818447113037,
|
| 1805 |
+
"step": 1250
|
| 1806 |
+
},
|
| 1807 |
+
{
|
| 1808 |
+
"epoch": 2.6,
|
| 1809 |
+
"learning_rate": 2.979373567608862e-06,
|
| 1810 |
+
"logits/chosen": -2.835662364959717,
|
| 1811 |
+
"logits/rejected": -2.8839526176452637,
|
| 1812 |
+
"logps/chosen": -257.24334716796875,
|
| 1813 |
+
"logps/rejected": -238.25588989257812,
|
| 1814 |
+
"loss": 0.368,
|
| 1815 |
+
"rewards/accuracies": 0.824999988079071,
|
| 1816 |
+
"rewards/chosen": -0.6586586236953735,
|
| 1817 |
+
"rewards/margins": 1.7249507904052734,
|
| 1818 |
+
"rewards/rejected": -2.3836092948913574,
|
| 1819 |
+
"step": 1260
|
| 1820 |
+
},
|
| 1821 |
+
{
|
| 1822 |
+
"epoch": 2.62,
|
| 1823 |
+
"learning_rate": 2.826585179526356e-06,
|
| 1824 |
+
"logits/chosen": -2.7938365936279297,
|
| 1825 |
+
"logits/rejected": -2.767805576324463,
|
| 1826 |
+
"logps/chosen": -249.7235107421875,
|
| 1827 |
+
"logps/rejected": -234.31283569335938,
|
| 1828 |
+
"loss": 0.369,
|
| 1829 |
+
"rewards/accuracies": 0.8125,
|
| 1830 |
+
"rewards/chosen": -0.7289739847183228,
|
| 1831 |
+
"rewards/margins": 1.7214456796646118,
|
| 1832 |
+
"rewards/rejected": -2.4504194259643555,
|
| 1833 |
+
"step": 1270
|
| 1834 |
+
},
|
| 1835 |
+
{
|
| 1836 |
+
"epoch": 2.64,
|
| 1837 |
+
"learning_rate": 2.673796791443851e-06,
|
| 1838 |
+
"logits/chosen": -2.7239506244659424,
|
| 1839 |
+
"logits/rejected": -2.751171588897705,
|
| 1840 |
+
"logps/chosen": -329.4594421386719,
|
| 1841 |
+
"logps/rejected": -260.408935546875,
|
| 1842 |
+
"loss": 0.3679,
|
| 1843 |
+
"rewards/accuracies": 0.90625,
|
| 1844 |
+
"rewards/chosen": -0.6046456694602966,
|
| 1845 |
+
"rewards/margins": 2.0207152366638184,
|
| 1846 |
+
"rewards/rejected": -2.6253609657287598,
|
| 1847 |
+
"step": 1280
|
| 1848 |
+
},
|
| 1849 |
+
{
|
| 1850 |
+
"epoch": 2.66,
|
| 1851 |
+
"learning_rate": 2.521008403361345e-06,
|
| 1852 |
+
"logits/chosen": -2.797870397567749,
|
| 1853 |
+
"logits/rejected": -2.8259756565093994,
|
| 1854 |
+
"logps/chosen": -305.6458435058594,
|
| 1855 |
+
"logps/rejected": -244.69601440429688,
|
| 1856 |
+
"loss": 0.3668,
|
| 1857 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 1858 |
+
"rewards/chosen": -0.6542243957519531,
|
| 1859 |
+
"rewards/margins": 1.900058388710022,
|
| 1860 |
+
"rewards/rejected": -2.5542826652526855,
|
| 1861 |
+
"step": 1290
|
| 1862 |
+
},
|
| 1863 |
+
{
|
| 1864 |
+
"epoch": 2.68,
|
| 1865 |
+
"learning_rate": 2.368220015278839e-06,
|
| 1866 |
+
"logits/chosen": -2.7966253757476807,
|
| 1867 |
+
"logits/rejected": -2.8371598720550537,
|
| 1868 |
+
"logps/chosen": -281.93939208984375,
|
| 1869 |
+
"logps/rejected": -261.60150146484375,
|
| 1870 |
+
"loss": 0.3348,
|
| 1871 |
+
"rewards/accuracies": 0.793749988079071,
|
| 1872 |
+
"rewards/chosen": -0.5757613182067871,
|
| 1873 |
+
"rewards/margins": 1.6483405828475952,
|
| 1874 |
+
"rewards/rejected": -2.224102020263672,
|
| 1875 |
+
"step": 1300
|
| 1876 |
+
},
|
| 1877 |
+
{
|
| 1878 |
+
"epoch": 2.7,
|
| 1879 |
+
"learning_rate": 2.2154316271963334e-06,
|
| 1880 |
+
"logits/chosen": -2.8347060680389404,
|
| 1881 |
+
"logits/rejected": -2.818382501602173,
|
| 1882 |
+
"logps/chosen": -266.34979248046875,
|
| 1883 |
+
"logps/rejected": -248.8013153076172,
|
| 1884 |
+
"loss": 0.3448,
|
| 1885 |
+
"rewards/accuracies": 0.8687499761581421,
|
| 1886 |
+
"rewards/chosen": -0.5766314268112183,
|
| 1887 |
+
"rewards/margins": 1.5551669597625732,
|
| 1888 |
+
"rewards/rejected": -2.131798505783081,
|
| 1889 |
+
"step": 1310
|
| 1890 |
+
},
|
| 1891 |
+
{
|
| 1892 |
+
"epoch": 2.72,
|
| 1893 |
+
"learning_rate": 2.0626432391138275e-06,
|
| 1894 |
+
"logits/chosen": -2.780651569366455,
|
| 1895 |
+
"logits/rejected": -2.7655837535858154,
|
| 1896 |
+
"logps/chosen": -317.9383544921875,
|
| 1897 |
+
"logps/rejected": -282.8804931640625,
|
| 1898 |
+
"loss": 0.3784,
|
| 1899 |
+
"rewards/accuracies": 0.875,
|
| 1900 |
+
"rewards/chosen": -0.5049678683280945,
|
| 1901 |
+
"rewards/margins": 1.9375810623168945,
|
| 1902 |
+
"rewards/rejected": -2.442549228668213,
|
| 1903 |
+
"step": 1320
|
| 1904 |
+
},
|
| 1905 |
+
{
|
| 1906 |
+
"epoch": 2.74,
|
| 1907 |
+
"learning_rate": 1.9098548510313215e-06,
|
| 1908 |
+
"logits/chosen": -2.8032174110412598,
|
| 1909 |
+
"logits/rejected": -2.7949814796447754,
|
| 1910 |
+
"logps/chosen": -260.9715576171875,
|
| 1911 |
+
"logps/rejected": -260.29913330078125,
|
| 1912 |
+
"loss": 0.3628,
|
| 1913 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 1914 |
+
"rewards/chosen": -0.6280083060264587,
|
| 1915 |
+
"rewards/margins": 1.837847113609314,
|
| 1916 |
+
"rewards/rejected": -2.465855360031128,
|
| 1917 |
+
"step": 1330
|
| 1918 |
+
},
|
| 1919 |
+
{
|
| 1920 |
+
"epoch": 2.76,
|
| 1921 |
+
"learning_rate": 1.757066462948816e-06,
|
| 1922 |
+
"logits/chosen": -2.8185904026031494,
|
| 1923 |
+
"logits/rejected": -2.7385153770446777,
|
| 1924 |
+
"logps/chosen": -278.1485290527344,
|
| 1925 |
+
"logps/rejected": -278.1205139160156,
|
| 1926 |
+
"loss": 0.3636,
|
| 1927 |
+
"rewards/accuracies": 0.862500011920929,
|
| 1928 |
+
"rewards/chosen": -0.6104816198348999,
|
| 1929 |
+
"rewards/margins": 2.0218350887298584,
|
| 1930 |
+
"rewards/rejected": -2.6323161125183105,
|
| 1931 |
+
"step": 1340
|
| 1932 |
+
},
|
| 1933 |
+
{
|
| 1934 |
+
"epoch": 2.78,
|
| 1935 |
+
"learning_rate": 1.6042780748663103e-06,
|
| 1936 |
+
"logits/chosen": -2.8218092918395996,
|
| 1937 |
+
"logits/rejected": -2.7967371940612793,
|
| 1938 |
+
"logps/chosen": -297.7966613769531,
|
| 1939 |
+
"logps/rejected": -240.7419891357422,
|
| 1940 |
+
"loss": 0.3763,
|
| 1941 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 1942 |
+
"rewards/chosen": -0.5741168260574341,
|
| 1943 |
+
"rewards/margins": 1.761366605758667,
|
| 1944 |
+
"rewards/rejected": -2.3354835510253906,
|
| 1945 |
+
"step": 1350
|
| 1946 |
+
},
|
| 1947 |
+
{
|
| 1948 |
+
"epoch": 2.8,
|
| 1949 |
+
"learning_rate": 1.4514896867838045e-06,
|
| 1950 |
+
"logits/chosen": -2.8747620582580566,
|
| 1951 |
+
"logits/rejected": -2.818427562713623,
|
| 1952 |
+
"logps/chosen": -280.43756103515625,
|
| 1953 |
+
"logps/rejected": -234.6200714111328,
|
| 1954 |
+
"loss": 0.3859,
|
| 1955 |
+
"rewards/accuracies": 0.8687499761581421,
|
| 1956 |
+
"rewards/chosen": -0.5531826019287109,
|
| 1957 |
+
"rewards/margins": 1.6504430770874023,
|
| 1958 |
+
"rewards/rejected": -2.2036256790161133,
|
| 1959 |
+
"step": 1360
|
| 1960 |
+
},
|
| 1961 |
+
{
|
| 1962 |
+
"epoch": 2.82,
|
| 1963 |
+
"learning_rate": 1.2987012987012986e-06,
|
| 1964 |
+
"logits/chosen": -2.8544864654541016,
|
| 1965 |
+
"logits/rejected": -2.8276607990264893,
|
| 1966 |
+
"logps/chosen": -298.3465881347656,
|
| 1967 |
+
"logps/rejected": -288.8724365234375,
|
| 1968 |
+
"loss": 0.3595,
|
| 1969 |
+
"rewards/accuracies": 0.8187500238418579,
|
| 1970 |
+
"rewards/chosen": -0.5492029786109924,
|
| 1971 |
+
"rewards/margins": 1.664731740951538,
|
| 1972 |
+
"rewards/rejected": -2.2139344215393066,
|
| 1973 |
+
"step": 1370
|
| 1974 |
+
},
|
| 1975 |
+
{
|
| 1976 |
+
"epoch": 2.85,
|
| 1977 |
+
"learning_rate": 1.145912910618793e-06,
|
| 1978 |
+
"logits/chosen": -2.7500030994415283,
|
| 1979 |
+
"logits/rejected": -2.7478601932525635,
|
| 1980 |
+
"logps/chosen": -245.6803741455078,
|
| 1981 |
+
"logps/rejected": -244.2843475341797,
|
| 1982 |
+
"loss": 0.3457,
|
| 1983 |
+
"rewards/accuracies": 0.831250011920929,
|
| 1984 |
+
"rewards/chosen": -0.5498741269111633,
|
| 1985 |
+
"rewards/margins": 1.8164262771606445,
|
| 1986 |
+
"rewards/rejected": -2.366300344467163,
|
| 1987 |
+
"step": 1380
|
| 1988 |
+
},
|
| 1989 |
+
{
|
| 1990 |
+
"epoch": 2.87,
|
| 1991 |
+
"learning_rate": 9.931245225362874e-07,
|
| 1992 |
+
"logits/chosen": -2.78765869140625,
|
| 1993 |
+
"logits/rejected": -2.7899107933044434,
|
| 1994 |
+
"logps/chosen": -269.9845886230469,
|
| 1995 |
+
"logps/rejected": -260.02557373046875,
|
| 1996 |
+
"loss": 0.366,
|
| 1997 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 1998 |
+
"rewards/chosen": -0.5584810376167297,
|
| 1999 |
+
"rewards/margins": 1.836942434310913,
|
| 2000 |
+
"rewards/rejected": -2.395423412322998,
|
| 2001 |
+
"step": 1390
|
| 2002 |
+
},
|
| 2003 |
+
{
|
| 2004 |
+
"epoch": 2.89,
|
| 2005 |
+
"learning_rate": 8.403361344537816e-07,
|
| 2006 |
+
"logits/chosen": -2.822734832763672,
|
| 2007 |
+
"logits/rejected": -2.7889413833618164,
|
| 2008 |
+
"logps/chosen": -295.0702819824219,
|
| 2009 |
+
"logps/rejected": -263.552978515625,
|
| 2010 |
+
"loss": 0.3664,
|
| 2011 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 2012 |
+
"rewards/chosen": -0.6814947128295898,
|
| 2013 |
+
"rewards/margins": 1.600311040878296,
|
| 2014 |
+
"rewards/rejected": -2.2818057537078857,
|
| 2015 |
+
"step": 1400
|
| 2016 |
+
},
|
| 2017 |
+
{
|
| 2018 |
+
"epoch": 2.91,
|
| 2019 |
+
"learning_rate": 6.875477463712758e-07,
|
| 2020 |
+
"logits/chosen": -2.7872085571289062,
|
| 2021 |
+
"logits/rejected": -2.813737392425537,
|
| 2022 |
+
"logps/chosen": -271.7468566894531,
|
| 2023 |
+
"logps/rejected": -254.4673309326172,
|
| 2024 |
+
"loss": 0.3564,
|
| 2025 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 2026 |
+
"rewards/chosen": -0.44644269347190857,
|
| 2027 |
+
"rewards/margins": 1.852097511291504,
|
| 2028 |
+
"rewards/rejected": -2.298539876937866,
|
| 2029 |
+
"step": 1410
|
| 2030 |
+
},
|
| 2031 |
+
{
|
| 2032 |
+
"epoch": 2.93,
|
| 2033 |
+
"learning_rate": 5.347593582887701e-07,
|
| 2034 |
+
"logits/chosen": -2.8677477836608887,
|
| 2035 |
+
"logits/rejected": -2.8729989528656006,
|
| 2036 |
+
"logps/chosen": -321.89447021484375,
|
| 2037 |
+
"logps/rejected": -266.3446350097656,
|
| 2038 |
+
"loss": 0.3638,
|
| 2039 |
+
"rewards/accuracies": 0.8125,
|
| 2040 |
+
"rewards/chosen": -0.6643561124801636,
|
| 2041 |
+
"rewards/margins": 1.5856568813323975,
|
| 2042 |
+
"rewards/rejected": -2.2500128746032715,
|
| 2043 |
+
"step": 1420
|
| 2044 |
+
},
|
| 2045 |
+
{
|
| 2046 |
+
"epoch": 2.95,
|
| 2047 |
+
"learning_rate": 3.819709702062643e-07,
|
| 2048 |
+
"logits/chosen": -2.8617660999298096,
|
| 2049 |
+
"logits/rejected": -2.8304145336151123,
|
| 2050 |
+
"logps/chosen": -289.00482177734375,
|
| 2051 |
+
"logps/rejected": -255.83004760742188,
|
| 2052 |
+
"loss": 0.3575,
|
| 2053 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 2054 |
+
"rewards/chosen": -0.6355488300323486,
|
| 2055 |
+
"rewards/margins": 1.5412448644638062,
|
| 2056 |
+
"rewards/rejected": -2.1767935752868652,
|
| 2057 |
+
"step": 1430
|
| 2058 |
+
},
|
| 2059 |
+
{
|
| 2060 |
+
"epoch": 2.97,
|
| 2061 |
+
"learning_rate": 2.2918258212375862e-07,
|
| 2062 |
+
"logits/chosen": -2.813324451446533,
|
| 2063 |
+
"logits/rejected": -2.822589635848999,
|
| 2064 |
+
"logps/chosen": -270.77374267578125,
|
| 2065 |
+
"logps/rejected": -270.3390808105469,
|
| 2066 |
+
"loss": 0.3543,
|
| 2067 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 2068 |
+
"rewards/chosen": -0.6036959886550903,
|
| 2069 |
+
"rewards/margins": 1.9311864376068115,
|
| 2070 |
+
"rewards/rejected": -2.5348825454711914,
|
| 2071 |
+
"step": 1440
|
| 2072 |
+
},
|
| 2073 |
+
{
|
| 2074 |
+
"epoch": 2.99,
|
| 2075 |
+
"learning_rate": 7.639419404125288e-08,
|
| 2076 |
+
"logits/chosen": -2.8171768188476562,
|
| 2077 |
+
"logits/rejected": -2.779967784881592,
|
| 2078 |
+
"logps/chosen": -303.8172912597656,
|
| 2079 |
+
"logps/rejected": -260.3236389160156,
|
| 2080 |
+
"loss": 0.3549,
|
| 2081 |
+
"rewards/accuracies": 0.831250011920929,
|
| 2082 |
+
"rewards/chosen": -0.6331970691680908,
|
| 2083 |
+
"rewards/margins": 1.8079423904418945,
|
| 2084 |
+
"rewards/rejected": -2.4411392211914062,
|
| 2085 |
+
"step": 1450
|
| 2086 |
+
},
|
| 2087 |
+
{
|
| 2088 |
+
"epoch": 3.0,
|
| 2089 |
+
"eval_logits/chosen": -2.842160701751709,
|
| 2090 |
+
"eval_logits/rejected": -2.833141326904297,
|
| 2091 |
+
"eval_logps/chosen": -279.5245056152344,
|
| 2092 |
+
"eval_logps/rejected": -246.69915771484375,
|
| 2093 |
+
"eval_loss": 0.45531293749809265,
|
| 2094 |
+
"eval_rewards/accuracies": 0.8359375,
|
| 2095 |
+
"eval_rewards/chosen": -0.5876308083534241,
|
| 2096 |
+
"eval_rewards/margins": 1.6034575700759888,
|
| 2097 |
+
"eval_rewards/rejected": -2.1910881996154785,
|
| 2098 |
+
"eval_runtime": 254.2995,
|
| 2099 |
+
"eval_samples_per_second": 7.865,
|
| 2100 |
+
"eval_steps_per_second": 0.063,
|
| 2101 |
+
"step": 1455
|
| 2102 |
+
},
|
| 2103 |
+
{
|
| 2104 |
+
"epoch": 3.0,
|
| 2105 |
+
"step": 1455,
|
| 2106 |
"total_flos": 0.0,
|
| 2107 |
+
"train_loss": 0.43281792414557074,
|
| 2108 |
+
"train_runtime": 46468.4841,
|
| 2109 |
+
"train_samples_per_second": 4.001,
|
| 2110 |
"train_steps_per_second": 0.031
|
| 2111 |
}
|
| 2112 |
],
|
| 2113 |
"logging_steps": 10,
|
| 2114 |
+
"max_steps": 1455,
|
| 2115 |
+
"num_train_epochs": 3,
|
| 2116 |
"save_steps": 500,
|
| 2117 |
"total_flos": 0.0,
|
| 2118 |
"trial_name": null,
|