Training in progress, step 100
Browse files- README.md +28 -19
- all_results.json +5 -5
- model-00001-of-00003.safetensors +1 -1
- model-00002-of-00003.safetensors +1 -1
- model-00003-of-00003.safetensors +1 -1
- train_results.json +5 -5
- trainer_state.json +1798 -578
- training_args.bin +1 -1
README.md
CHANGED
|
@@ -2,30 +2,31 @@
|
|
| 2 |
license: mit
|
| 3 |
base_model: HuggingFaceH4/mistral-7b-sft-beta
|
| 4 |
tags:
|
| 5 |
-
-
|
| 6 |
-
- dpo
|
| 7 |
- generated_from_trainer
|
|
|
|
|
|
|
| 8 |
model-index:
|
| 9 |
-
- name:
|
| 10 |
results: []
|
| 11 |
---
|
| 12 |
|
| 13 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 14 |
should probably proofread and complete it, then remove this comment. -->
|
| 15 |
|
| 16 |
-
#
|
| 17 |
|
| 18 |
-
This model is a fine-tuned version of [HuggingFaceH4/mistral-7b-sft-beta](https://huggingface.co/HuggingFaceH4/mistral-7b-sft-beta) on the
|
| 19 |
It achieves the following results on the evaluation set:
|
| 20 |
-
- Loss: 0.
|
| 21 |
-
- Rewards/chosen: -
|
| 22 |
-
- Rewards/rejected: -
|
| 23 |
-
- Rewards/accuracies: 0.
|
| 24 |
-
- Rewards/margins: 0.
|
| 25 |
-
- Logps/rejected: -
|
| 26 |
-
- Logps/chosen: -
|
| 27 |
-
- Logits/rejected:
|
| 28 |
-
- Logits/chosen:
|
| 29 |
|
| 30 |
## Model description
|
| 31 |
|
|
@@ -47,7 +48,7 @@ The following hyperparameters were used during training:
|
|
| 47 |
- learning_rate: 5e-07
|
| 48 |
- train_batch_size: 8
|
| 49 |
- eval_batch_size: 8
|
| 50 |
-
- seed:
|
| 51 |
- distributed_type: multi-GPU
|
| 52 |
- num_devices: 8
|
| 53 |
- gradient_accumulation_steps: 2
|
|
@@ -62,10 +63,18 @@ The following hyperparameters were used during training:
|
|
| 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 |
-
| 0.
|
| 67 |
-
| 0.
|
| 68 |
-
| 0.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 69 |
|
| 70 |
|
| 71 |
### Framework versions
|
|
|
|
| 2 |
license: mit
|
| 3 |
base_model: HuggingFaceH4/mistral-7b-sft-beta
|
| 4 |
tags:
|
| 5 |
+
- alignment-handbook
|
|
|
|
| 6 |
- generated_from_trainer
|
| 7 |
+
datasets:
|
| 8 |
+
- HuggingFaceH4/hh-rlhf-h4
|
| 9 |
model-index:
|
| 10 |
+
- name: dpo_hh
|
| 11 |
results: []
|
| 12 |
---
|
| 13 |
|
| 14 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 15 |
should probably proofread and complete it, then remove this comment. -->
|
| 16 |
|
| 17 |
+
# dpo_hh
|
| 18 |
|
| 19 |
+
This model is a fine-tuned version of [HuggingFaceH4/mistral-7b-sft-beta](https://huggingface.co/HuggingFaceH4/mistral-7b-sft-beta) on the HuggingFaceH4/hh-rlhf-h4 dataset.
|
| 20 |
It achieves the following results on the evaluation set:
|
| 21 |
+
- Loss: 0.5440
|
| 22 |
+
- Rewards/chosen: -2.2940
|
| 23 |
+
- Rewards/rejected: -3.0054
|
| 24 |
+
- Rewards/accuracies: 0.7090
|
| 25 |
+
- Rewards/margins: 0.7114
|
| 26 |
+
- Logps/rejected: -451.6765
|
| 27 |
+
- Logps/chosen: -373.9785
|
| 28 |
+
- Logits/rejected: 0.3244
|
| 29 |
+
- Logits/chosen: 0.0742
|
| 30 |
|
| 31 |
## Model description
|
| 32 |
|
|
|
|
| 48 |
- learning_rate: 5e-07
|
| 49 |
- train_batch_size: 8
|
| 50 |
- eval_batch_size: 8
|
| 51 |
+
- seed: 42
|
| 52 |
- distributed_type: multi-GPU
|
| 53 |
- num_devices: 8
|
| 54 |
- gradient_accumulation_steps: 2
|
|
|
|
| 63 |
|
| 64 |
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 65 |
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 66 |
+
| 0.6789 | 0.08 | 100 | 0.6770 | -0.1062 | -0.1422 | 0.5914 | 0.0360 | -165.3552 | -155.1927 | -2.7255 | -2.7337 |
|
| 67 |
+
| 0.6062 | 0.16 | 200 | 0.6079 | -1.0212 | -1.3873 | 0.6670 | 0.3660 | -289.8622 | -246.6971 | -2.3696 | -2.3856 |
|
| 68 |
+
| 0.5965 | 0.24 | 300 | 0.5907 | -1.3779 | -1.8008 | 0.6623 | 0.4229 | -331.2100 | -282.3621 | -2.2450 | -2.2656 |
|
| 69 |
+
| 0.5729 | 0.32 | 400 | 0.5711 | -1.6763 | -2.2404 | 0.6828 | 0.5640 | -375.1720 | -312.2064 | -1.2920 | -1.3760 |
|
| 70 |
+
| 0.5645 | 0.4 | 500 | 0.5639 | -2.0721 | -2.6869 | 0.6987 | 0.6147 | -419.8194 | -351.7883 | -0.6091 | -0.7860 |
|
| 71 |
+
| 0.5513 | 0.48 | 600 | 0.5582 | -2.9237 | -3.5389 | 0.7108 | 0.6152 | -505.0223 | -436.9386 | 0.1224 | -0.1054 |
|
| 72 |
+
| 0.5571 | 0.56 | 700 | 0.5559 | -2.7971 | -3.5456 | 0.7043 | 0.7485 | -505.6961 | -424.2823 | 0.2980 | 0.0356 |
|
| 73 |
+
| 0.5609 | 0.64 | 800 | 0.5469 | -2.4314 | -3.0831 | 0.7108 | 0.6517 | -459.4439 | -387.7092 | 0.1922 | -0.0312 |
|
| 74 |
+
| 0.5514 | 0.72 | 900 | 0.5474 | -2.4774 | -3.2082 | 0.6996 | 0.7308 | -471.9533 | -392.3096 | 0.5382 | 0.2860 |
|
| 75 |
+
| 0.527 | 0.8 | 1000 | 0.5454 | -2.5040 | -3.2071 | 0.7080 | 0.7031 | -471.8454 | -394.9711 | 0.6372 | 0.3871 |
|
| 76 |
+
| 0.5487 | 0.88 | 1100 | 0.5444 | -2.2851 | -2.9963 | 0.7090 | 0.7112 | -450.7599 | -373.0831 | 0.4336 | 0.1858 |
|
| 77 |
+
| 0.5483 | 0.96 | 1200 | 0.5440 | -2.2940 | -3.0054 | 0.7090 | 0.7114 | -451.6765 | -373.9785 | 0.3244 | 0.0742 |
|
| 78 |
|
| 79 |
|
| 80 |
### Framework versions
|
all_results.json
CHANGED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
-
"train_loss": 0.
|
| 4 |
-
"train_runtime":
|
| 5 |
-
"train_samples":
|
| 6 |
-
"train_samples_per_second":
|
| 7 |
-
"train_steps_per_second": 0.
|
| 8 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
+
"train_loss": 0.5712926928784438,
|
| 4 |
+
"train_runtime": 11573.5451,
|
| 5 |
+
"train_samples": 160800,
|
| 6 |
+
"train_samples_per_second": 13.894,
|
| 7 |
+
"train_steps_per_second": 0.109
|
| 8 |
}
|
model-00001-of-00003.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 4943162336
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:18e1cec63bd40f863dc594533ae9ac02d7bcdd4f57a17c1ef5d63193122a0814
|
| 3 |
size 4943162336
|
model-00002-of-00003.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 4999819336
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:565d4244afeda54e7f62be9e162a16c6892085c081422f02c7a001ecce587eb6
|
| 3 |
size 4999819336
|
model-00003-of-00003.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 4540516344
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0debf1533b3a9f2ffea91ddec7f947ba3d1c43476aedcef3273235a227bb4ce5
|
| 3 |
size 4540516344
|
train_results.json
CHANGED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
-
"train_loss": 0.
|
| 4 |
-
"train_runtime":
|
| 5 |
-
"train_samples":
|
| 6 |
-
"train_samples_per_second":
|
| 7 |
-
"train_steps_per_second": 0.
|
| 8 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
+
"train_loss": 0.5712926928784438,
|
| 4 |
+
"train_runtime": 11573.5451,
|
| 5 |
+
"train_samples": 160800,
|
| 6 |
+
"train_samples_per_second": 13.894,
|
| 7 |
+
"train_steps_per_second": 0.109
|
| 8 |
}
|
trainer_state.json
CHANGED
|
@@ -1,21 +1,21 @@
|
|
| 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":
|
| 14 |
-
"logits/chosen": -2.
|
| 15 |
-
"logits/rejected": -2.
|
| 16 |
-
"logps/chosen": -
|
| 17 |
-
"logps/rejected": -
|
| 18 |
-
"loss": 0.
|
| 19 |
"rewards/accuracies": 0.0,
|
| 20 |
"rewards/chosen": 0.0,
|
| 21 |
"rewards/margins": 0.0,
|
|
@@ -23,739 +23,1959 @@
|
|
| 23 |
"step": 1
|
| 24 |
},
|
| 25 |
{
|
| 26 |
-
"epoch": 0.
|
| 27 |
-
"learning_rate":
|
| 28 |
-
"logits/chosen": -2.
|
| 29 |
-
"logits/rejected": -2.
|
| 30 |
-
"logps/chosen": -
|
| 31 |
-
"logps/rejected": -
|
| 32 |
-
"loss": 0.
|
| 33 |
-
"rewards/accuracies": 0.
|
| 34 |
-
"rewards/chosen": 0.
|
| 35 |
-
"rewards/margins": 0.
|
| 36 |
-
"rewards/rejected": 0.
|
| 37 |
"step": 10
|
| 38 |
},
|
| 39 |
{
|
| 40 |
-
"epoch": 0.
|
| 41 |
-
"learning_rate":
|
| 42 |
-
"logits/chosen": -2.
|
| 43 |
-
"logits/rejected": -2.
|
| 44 |
-
"logps/chosen": -
|
| 45 |
-
"logps/rejected": -
|
| 46 |
-
"loss": 0.
|
| 47 |
"rewards/accuracies": 0.550000011920929,
|
| 48 |
-
"rewards/chosen": -
|
| 49 |
-
"rewards/margins": 0.
|
| 50 |
-
"rewards/rejected": -
|
| 51 |
"step": 20
|
| 52 |
},
|
| 53 |
{
|
| 54 |
-
"epoch": 0.
|
| 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":
|
| 65 |
"step": 30
|
| 66 |
},
|
| 67 |
{
|
| 68 |
-
"epoch": 0.
|
| 69 |
-
"learning_rate":
|
| 70 |
-
"logits/chosen": -2.
|
| 71 |
-
"logits/rejected": -2.
|
| 72 |
-
"logps/chosen": -
|
| 73 |
-
"logps/rejected": -
|
| 74 |
-
"loss": 0.
|
| 75 |
-
"rewards/accuracies": 0.
|
| 76 |
-
"rewards/chosen":
|
| 77 |
-
"rewards/margins": 0.
|
| 78 |
-
"rewards/rejected":
|
| 79 |
"step": 40
|
| 80 |
},
|
| 81 |
{
|
| 82 |
-
"epoch": 0.
|
| 83 |
-
"learning_rate":
|
| 84 |
-
"logits/chosen": -2.
|
| 85 |
-
"logits/rejected": -2.
|
| 86 |
-
"logps/chosen": -
|
| 87 |
-
"logps/rejected": -
|
| 88 |
-
"loss": 0.
|
| 89 |
-
"rewards/accuracies": 0.
|
| 90 |
-
"rewards/chosen":
|
| 91 |
-
"rewards/margins": 0.
|
| 92 |
-
"rewards/rejected":
|
| 93 |
"step": 50
|
| 94 |
},
|
| 95 |
{
|
| 96 |
-
"epoch": 0.
|
| 97 |
-
"learning_rate":
|
| 98 |
-
"logits/chosen": -2.
|
| 99 |
-
"logits/rejected": -2.
|
| 100 |
-
"logps/chosen": -
|
| 101 |
-
"logps/rejected": -
|
| 102 |
-
"loss": 0.
|
| 103 |
-
"rewards/accuracies": 0.
|
| 104 |
-
"rewards/chosen":
|
| 105 |
-
"rewards/margins": 0.
|
| 106 |
-
"rewards/rejected":
|
| 107 |
"step": 60
|
| 108 |
},
|
| 109 |
{
|
| 110 |
-
"epoch": 0.
|
| 111 |
-
"learning_rate":
|
| 112 |
-
"logits/chosen": -2.
|
| 113 |
-
"logits/rejected": -2.
|
| 114 |
-
"logps/chosen": -
|
| 115 |
-
"logps/rejected": -
|
| 116 |
-
"loss": 0.
|
| 117 |
-
"rewards/accuracies": 0.
|
| 118 |
-
"rewards/chosen":
|
| 119 |
-
"rewards/margins": 0.
|
| 120 |
-
"rewards/rejected":
|
| 121 |
"step": 70
|
| 122 |
},
|
| 123 |
{
|
| 124 |
-
"epoch": 0.
|
| 125 |
-
"learning_rate":
|
| 126 |
-
"logits/chosen": -2.
|
| 127 |
-
"logits/rejected": -2.
|
| 128 |
-
"logps/chosen": -
|
| 129 |
-
"logps/rejected": -
|
| 130 |
-
"loss": 0.
|
| 131 |
-
"rewards/accuracies": 0.
|
| 132 |
-
"rewards/chosen":
|
| 133 |
-
"rewards/margins": 0.
|
| 134 |
-
"rewards/rejected": -0.
|
| 135 |
"step": 80
|
| 136 |
},
|
| 137 |
{
|
| 138 |
-
"epoch": 0.
|
| 139 |
-
"learning_rate":
|
| 140 |
-
"logits/chosen": -2.
|
| 141 |
-
"logits/rejected": -2.
|
| 142 |
-
"logps/chosen": -
|
| 143 |
-
"logps/rejected": -
|
| 144 |
-
"loss": 0.
|
| 145 |
-
"rewards/accuracies": 0.
|
| 146 |
-
"rewards/chosen": -0.
|
| 147 |
-
"rewards/margins": 0.
|
| 148 |
-
"rewards/rejected": -0.
|
| 149 |
"step": 90
|
| 150 |
},
|
| 151 |
{
|
| 152 |
-
"epoch": 0.
|
| 153 |
-
"learning_rate":
|
| 154 |
-
"logits/chosen": -2.
|
| 155 |
-
"logits/rejected": -2.
|
| 156 |
-
"logps/chosen": -
|
| 157 |
-
"logps/rejected": -
|
| 158 |
-
"loss": 0.
|
| 159 |
-
"rewards/accuracies": 0.
|
| 160 |
-
"rewards/chosen": -0.
|
| 161 |
-
"rewards/margins": 0.
|
| 162 |
-
"rewards/rejected": -0.
|
| 163 |
"step": 100
|
| 164 |
},
|
| 165 |
{
|
| 166 |
-
"epoch": 0.
|
| 167 |
-
"eval_logits/chosen": -2.
|
| 168 |
-
"eval_logits/rejected": -2.
|
| 169 |
-
"eval_logps/chosen": -
|
| 170 |
-
"eval_logps/rejected": -
|
| 171 |
-
"eval_loss": 0.
|
| 172 |
-
"eval_rewards/accuracies": 0.
|
| 173 |
-
"eval_rewards/chosen": -0.
|
| 174 |
-
"eval_rewards/margins": 0.
|
| 175 |
-
"eval_rewards/rejected": -
|
| 176 |
-
"eval_runtime":
|
| 177 |
-
"eval_samples_per_second":
|
| 178 |
-
"eval_steps_per_second": 0.
|
| 179 |
"step": 100
|
| 180 |
},
|
| 181 |
{
|
| 182 |
-
"epoch": 0.
|
| 183 |
-
"learning_rate": 4.
|
| 184 |
-
"logits/chosen": -2.
|
| 185 |
-
"logits/rejected": -2.
|
| 186 |
-
"logps/chosen": -
|
| 187 |
-
"logps/rejected": -
|
| 188 |
-
"loss": 0.
|
| 189 |
-
"rewards/accuracies": 0.
|
| 190 |
-
"rewards/chosen": -0.
|
| 191 |
-
"rewards/margins": 0.
|
| 192 |
-
"rewards/rejected": -
|
| 193 |
"step": 110
|
| 194 |
},
|
| 195 |
{
|
| 196 |
-
"epoch": 0.
|
| 197 |
-
"learning_rate": 4.
|
| 198 |
-
"logits/chosen": -2.
|
| 199 |
-
"logits/rejected": -2.
|
| 200 |
-
"logps/chosen": -
|
| 201 |
-
"logps/rejected": -
|
| 202 |
-
"loss": 0.
|
| 203 |
-
"rewards/accuracies": 0.
|
| 204 |
-
"rewards/chosen": -0.
|
| 205 |
-
"rewards/margins": 0.
|
| 206 |
-
"rewards/rejected": -
|
| 207 |
"step": 120
|
| 208 |
},
|
| 209 |
{
|
| 210 |
-
"epoch": 0.
|
| 211 |
-
"learning_rate": 4.
|
| 212 |
-
"logits/chosen": -2.
|
| 213 |
-
"logits/rejected": -2.
|
| 214 |
-
"logps/chosen": -
|
| 215 |
-
"logps/rejected": -
|
| 216 |
-
"loss": 0.
|
| 217 |
-
"rewards/accuracies": 0.
|
| 218 |
-
"rewards/chosen": -0.
|
| 219 |
-
"rewards/margins": 0.
|
| 220 |
-
"rewards/rejected": -
|
| 221 |
"step": 130
|
| 222 |
},
|
| 223 |
{
|
| 224 |
-
"epoch": 0.
|
| 225 |
-
"learning_rate": 4.
|
| 226 |
-
"logits/chosen": -2.
|
| 227 |
-
"logits/rejected": -2.
|
| 228 |
-
"logps/chosen": -
|
| 229 |
-
"logps/rejected": -
|
| 230 |
-
"loss": 0.
|
| 231 |
-
"rewards/accuracies": 0.
|
| 232 |
-
"rewards/chosen": -0.
|
| 233 |
-
"rewards/margins": 0.
|
| 234 |
-
"rewards/rejected": -
|
| 235 |
"step": 140
|
| 236 |
},
|
| 237 |
{
|
| 238 |
-
"epoch": 0.
|
| 239 |
-
"learning_rate": 4.
|
| 240 |
-
"logits/chosen": -2.
|
| 241 |
-
"logits/rejected": -2.
|
| 242 |
-
"logps/chosen": -
|
| 243 |
-
"logps/rejected": -
|
| 244 |
-
"loss": 0.
|
| 245 |
-
"rewards/accuracies": 0.
|
| 246 |
-
"rewards/chosen": -0.
|
| 247 |
-
"rewards/margins": 0.
|
| 248 |
-
"rewards/rejected": -
|
| 249 |
"step": 150
|
| 250 |
},
|
| 251 |
{
|
| 252 |
-
"epoch": 0.
|
| 253 |
-
"learning_rate": 4.
|
| 254 |
-
"logits/chosen": -2.
|
| 255 |
-
"logits/rejected": -2.
|
| 256 |
-
"logps/chosen": -
|
| 257 |
-
"logps/rejected": -
|
| 258 |
-
"loss": 0.
|
| 259 |
-
"rewards/accuracies": 0.
|
| 260 |
-
"rewards/chosen": -0.
|
| 261 |
-
"rewards/margins": 0.
|
| 262 |
-
"rewards/rejected": -
|
| 263 |
"step": 160
|
| 264 |
},
|
| 265 |
{
|
| 266 |
-
"epoch": 0.
|
| 267 |
-
"learning_rate": 4.
|
| 268 |
-
"logits/chosen": -2.
|
| 269 |
-
"logits/rejected": -2.
|
| 270 |
-
"logps/chosen": -
|
| 271 |
-
"logps/rejected": -
|
| 272 |
-
"loss": 0.
|
| 273 |
-
"rewards/accuracies": 0.
|
| 274 |
-
"rewards/chosen": -0.
|
| 275 |
-
"rewards/margins": 0.
|
| 276 |
-
"rewards/rejected": -
|
| 277 |
"step": 170
|
| 278 |
},
|
| 279 |
{
|
| 280 |
-
"epoch": 0.
|
| 281 |
-
"learning_rate":
|
| 282 |
-
"logits/chosen": -2.
|
| 283 |
-
"logits/rejected": -2.
|
| 284 |
-
"logps/chosen": -
|
| 285 |
-
"logps/rejected": -
|
| 286 |
-
"loss": 0.
|
| 287 |
-
"rewards/accuracies": 0.
|
| 288 |
-
"rewards/chosen": -0.
|
| 289 |
-
"rewards/margins": 0.
|
| 290 |
-
"rewards/rejected": -
|
| 291 |
"step": 180
|
| 292 |
},
|
| 293 |
{
|
| 294 |
-
"epoch": 0.
|
| 295 |
-
"learning_rate":
|
| 296 |
-
"logits/chosen": -2.
|
| 297 |
-
"logits/rejected": -2.
|
| 298 |
-
"logps/chosen": -
|
| 299 |
-
"logps/rejected": -
|
| 300 |
-
"loss": 0.
|
| 301 |
"rewards/accuracies": 0.699999988079071,
|
| 302 |
-
"rewards/chosen": -
|
| 303 |
-
"rewards/margins": 0.
|
| 304 |
-
"rewards/rejected": -1.
|
| 305 |
"step": 190
|
| 306 |
},
|
| 307 |
{
|
| 308 |
-
"epoch": 0.
|
| 309 |
-
"learning_rate":
|
| 310 |
-
"logits/chosen": -2.
|
| 311 |
-
"logits/rejected": -2.
|
| 312 |
-
"logps/chosen": -
|
| 313 |
-
"logps/rejected": -
|
| 314 |
-
"loss": 0.
|
| 315 |
-
"rewards/accuracies": 0.
|
| 316 |
-
"rewards/chosen": -
|
| 317 |
-
"rewards/margins": 0.
|
| 318 |
-
"rewards/rejected": -1.
|
| 319 |
"step": 200
|
| 320 |
},
|
| 321 |
{
|
| 322 |
-
"epoch": 0.
|
| 323 |
-
"eval_logits/chosen": -2.
|
| 324 |
-
"eval_logits/rejected": -2.
|
| 325 |
-
"eval_logps/chosen": -
|
| 326 |
-
"eval_logps/rejected": -
|
| 327 |
-
"eval_loss": 0.
|
| 328 |
-
"eval_rewards/accuracies": 0.
|
| 329 |
-
"eval_rewards/chosen": -1.
|
| 330 |
-
"eval_rewards/margins": 0.
|
| 331 |
-
"eval_rewards/rejected": -1.
|
| 332 |
-
"eval_runtime":
|
| 333 |
-
"eval_samples_per_second":
|
| 334 |
-
"eval_steps_per_second": 0.
|
| 335 |
"step": 200
|
| 336 |
},
|
| 337 |
{
|
| 338 |
-
"epoch": 0.
|
| 339 |
-
"learning_rate":
|
| 340 |
-
"logits/chosen": -2.
|
| 341 |
-
"logits/rejected": -2.
|
| 342 |
-
"logps/chosen": -
|
| 343 |
-
"logps/rejected": -
|
| 344 |
-
"loss": 0.
|
| 345 |
-
"rewards/accuracies": 0.
|
| 346 |
-
"rewards/chosen": -1.
|
| 347 |
-
"rewards/margins": 0.
|
| 348 |
-
"rewards/rejected": -1.
|
| 349 |
"step": 210
|
| 350 |
},
|
| 351 |
{
|
| 352 |
-
"epoch": 0.
|
| 353 |
-
"learning_rate":
|
| 354 |
-
"logits/chosen": -2.
|
| 355 |
-
"logits/rejected": -2.
|
| 356 |
-
"logps/chosen": -
|
| 357 |
-
"logps/rejected": -
|
| 358 |
-
"loss": 0.
|
| 359 |
-
"rewards/accuracies": 0.
|
| 360 |
-
"rewards/chosen": -1.
|
| 361 |
-
"rewards/margins": 0.
|
| 362 |
-
"rewards/rejected": -
|
| 363 |
"step": 220
|
| 364 |
},
|
| 365 |
{
|
| 366 |
-
"epoch": 0.
|
| 367 |
-
"learning_rate":
|
| 368 |
-
"logits/chosen": -2.
|
| 369 |
-
"logits/rejected": -2.
|
| 370 |
-
"logps/chosen": -
|
| 371 |
-
"logps/rejected": -
|
| 372 |
-
"loss": 0.
|
| 373 |
-
"rewards/accuracies": 0.
|
| 374 |
-
"rewards/chosen": -1.
|
| 375 |
-
"rewards/margins": 0.
|
| 376 |
-
"rewards/rejected": -1.
|
| 377 |
"step": 230
|
| 378 |
},
|
| 379 |
{
|
| 380 |
-
"epoch": 0.
|
| 381 |
-
"learning_rate":
|
| 382 |
-
"logits/chosen": -2.
|
| 383 |
-
"logits/rejected": -2.
|
| 384 |
-
"logps/chosen": -
|
| 385 |
-
"logps/rejected": -
|
| 386 |
-
"loss": 0.
|
| 387 |
-
"rewards/accuracies": 0.
|
| 388 |
-
"rewards/chosen": -1.
|
| 389 |
-
"rewards/margins": 0.
|
| 390 |
-
"rewards/rejected": -
|
| 391 |
"step": 240
|
| 392 |
},
|
| 393 |
{
|
| 394 |
-
"epoch": 0.
|
| 395 |
-
"learning_rate":
|
| 396 |
-
"logits/chosen": -2.
|
| 397 |
-
"logits/rejected": -2.
|
| 398 |
-
"logps/chosen": -
|
| 399 |
-
"logps/rejected": -
|
| 400 |
-
"loss": 0.
|
| 401 |
-
"rewards/accuracies": 0.
|
| 402 |
-
"rewards/chosen": -
|
| 403 |
-
"rewards/margins": 0.
|
| 404 |
-
"rewards/rejected": -
|
| 405 |
"step": 250
|
| 406 |
},
|
| 407 |
{
|
| 408 |
-
"epoch": 0.
|
| 409 |
-
"learning_rate":
|
| 410 |
-
"logits/chosen": -2.
|
| 411 |
-
"logits/rejected": -2.
|
| 412 |
-
"logps/chosen": -
|
| 413 |
-
"logps/rejected": -
|
| 414 |
-
"loss": 0.
|
| 415 |
-
"rewards/accuracies": 0.
|
| 416 |
-
"rewards/chosen": -1.
|
| 417 |
-
"rewards/margins": 0.
|
| 418 |
-
"rewards/rejected": -
|
| 419 |
"step": 260
|
| 420 |
},
|
| 421 |
{
|
| 422 |
-
"epoch": 0.
|
| 423 |
-
"learning_rate":
|
| 424 |
-
"logits/chosen": -2.
|
| 425 |
-
"logits/rejected": -2.
|
| 426 |
-
"logps/chosen": -
|
| 427 |
-
"logps/rejected": -
|
| 428 |
-
"loss": 0.
|
| 429 |
-
"rewards/accuracies": 0.
|
| 430 |
-
"rewards/chosen": -1.
|
| 431 |
-
"rewards/margins": 0.
|
| 432 |
-
"rewards/rejected": -
|
| 433 |
"step": 270
|
| 434 |
},
|
| 435 |
{
|
| 436 |
-
"epoch": 0.
|
| 437 |
-
"learning_rate":
|
| 438 |
-
"logits/chosen": -2.
|
| 439 |
-
"logits/rejected": -2.
|
| 440 |
-
"logps/chosen": -
|
| 441 |
-
"logps/rejected": -
|
| 442 |
-
"loss": 0.
|
| 443 |
-
"rewards/accuracies": 0.
|
| 444 |
-
"rewards/chosen": -1.
|
| 445 |
-
"rewards/margins": 0.
|
| 446 |
-
"rewards/rejected": -
|
| 447 |
"step": 280
|
| 448 |
},
|
| 449 |
{
|
| 450 |
-
"epoch": 0.
|
| 451 |
-
"learning_rate":
|
| 452 |
-
"logits/chosen": -2.
|
| 453 |
-
"logits/rejected": -2.
|
| 454 |
-
"logps/chosen": -
|
| 455 |
-
"logps/rejected": -
|
| 456 |
-
"loss": 0.
|
| 457 |
-
"rewards/accuracies": 0.
|
| 458 |
-
"rewards/chosen": -1.
|
| 459 |
-
"rewards/margins": 0.
|
| 460 |
-
"rewards/rejected": -
|
| 461 |
"step": 290
|
| 462 |
},
|
| 463 |
{
|
| 464 |
-
"epoch": 0.
|
| 465 |
-
"learning_rate":
|
| 466 |
-
"logits/chosen": -2.
|
| 467 |
-
"logits/rejected": -2.
|
| 468 |
-
"logps/chosen": -
|
| 469 |
-
"logps/rejected": -
|
| 470 |
-
"loss": 0.
|
| 471 |
"rewards/accuracies": 0.675000011920929,
|
| 472 |
-
"rewards/chosen": -1.
|
| 473 |
-
"rewards/margins": 0.
|
| 474 |
-
"rewards/rejected": -
|
| 475 |
"step": 300
|
| 476 |
},
|
| 477 |
{
|
| 478 |
-
"epoch": 0.
|
| 479 |
-
"eval_logits/chosen": -2.
|
| 480 |
-
"eval_logits/rejected": -2.
|
| 481 |
-
"eval_logps/chosen": -
|
| 482 |
-
"eval_logps/rejected": -
|
| 483 |
-
"eval_loss": 0.
|
| 484 |
-
"eval_rewards/accuracies": 0.
|
| 485 |
-
"eval_rewards/chosen": -1.
|
| 486 |
-
"eval_rewards/margins": 0.
|
| 487 |
-
"eval_rewards/rejected": -
|
| 488 |
-
"eval_runtime":
|
| 489 |
-
"eval_samples_per_second":
|
| 490 |
-
"eval_steps_per_second": 0.
|
| 491 |
"step": 300
|
| 492 |
},
|
| 493 |
{
|
| 494 |
-
"epoch": 0.
|
| 495 |
-
"learning_rate":
|
| 496 |
-
"logits/chosen": -2.
|
| 497 |
-
"logits/rejected": -2.
|
| 498 |
-
"logps/chosen": -
|
| 499 |
-
"logps/rejected": -
|
| 500 |
-
"loss": 0.
|
| 501 |
-
"rewards/accuracies": 0.
|
| 502 |
-
"rewards/chosen": -1.
|
| 503 |
-
"rewards/margins": 0.
|
| 504 |
-
"rewards/rejected": -
|
| 505 |
"step": 310
|
| 506 |
},
|
| 507 |
{
|
| 508 |
-
"epoch": 0.
|
| 509 |
-
"learning_rate":
|
| 510 |
-
"logits/chosen": -2.
|
| 511 |
-
"logits/rejected": -2.
|
| 512 |
-
"logps/chosen": -
|
| 513 |
-
"logps/rejected": -
|
| 514 |
-
"loss": 0.
|
| 515 |
-
"rewards/accuracies": 0.
|
| 516 |
-
"rewards/chosen": -1.
|
| 517 |
-
"rewards/margins": 0.
|
| 518 |
-
"rewards/rejected": -
|
| 519 |
"step": 320
|
| 520 |
},
|
| 521 |
{
|
| 522 |
-
"epoch": 0.
|
| 523 |
-
"learning_rate":
|
| 524 |
-
"logits/chosen": -2.
|
| 525 |
-
"logits/rejected": -2.
|
| 526 |
-
"logps/chosen": -
|
| 527 |
-
"logps/rejected": -
|
| 528 |
-
"loss": 0.
|
| 529 |
-
"rewards/accuracies": 0.
|
| 530 |
-
"rewards/chosen": -1.
|
| 531 |
-
"rewards/margins": 0.
|
| 532 |
-
"rewards/rejected": -
|
| 533 |
"step": 330
|
| 534 |
},
|
| 535 |
{
|
| 536 |
-
"epoch": 0.
|
| 537 |
-
"learning_rate":
|
| 538 |
-
"logits/chosen": -2.
|
| 539 |
-
"logits/rejected": -2.
|
| 540 |
-
"logps/chosen": -
|
| 541 |
-
"logps/rejected": -
|
| 542 |
-
"loss": 0.
|
| 543 |
-
"rewards/accuracies": 0.
|
| 544 |
-
"rewards/chosen": -1.
|
| 545 |
-
"rewards/margins": 0.
|
| 546 |
-
"rewards/rejected": -2.
|
| 547 |
"step": 340
|
| 548 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 549 |
{
|
| 550 |
"epoch": 0.73,
|
| 551 |
-
"learning_rate": 1.
|
| 552 |
-
"logits/chosen":
|
| 553 |
-
"logits/rejected":
|
| 554 |
-
"logps/chosen": -
|
| 555 |
-
"logps/rejected": -
|
| 556 |
-
"loss": 0.
|
| 557 |
"rewards/accuracies": 0.706250011920929,
|
| 558 |
-
"rewards/chosen": -
|
| 559 |
-
"rewards/margins": 0.
|
| 560 |
-
"rewards/rejected": -2.
|
| 561 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 562 |
},
|
| 563 |
{
|
| 564 |
"epoch": 0.75,
|
| 565 |
-
"learning_rate":
|
| 566 |
-
"logits/chosen": -
|
| 567 |
-
"logits/rejected":
|
| 568 |
-
"logps/chosen": -
|
| 569 |
-
"logps/rejected": -
|
| 570 |
-
"loss": 0.
|
| 571 |
-
"rewards/accuracies": 0.
|
| 572 |
-
"rewards/chosen": -
|
| 573 |
-
"rewards/margins": 0.
|
| 574 |
-
"rewards/rejected": -2.
|
| 575 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 576 |
},
|
| 577 |
{
|
| 578 |
"epoch": 0.77,
|
| 579 |
-
"learning_rate": 7.
|
| 580 |
-
"logits/chosen":
|
| 581 |
-
"logits/rejected":
|
| 582 |
-
"logps/chosen": -
|
| 583 |
-
"logps/rejected": -
|
| 584 |
-
"loss": 0.
|
| 585 |
-
"rewards/accuracies": 0.
|
| 586 |
-
"rewards/chosen": -
|
| 587 |
-
"rewards/margins": 0.
|
| 588 |
-
"rewards/rejected": -
|
| 589 |
-
"step":
|
| 590 |
},
|
| 591 |
{
|
| 592 |
-
"epoch": 0.
|
| 593 |
-
"learning_rate":
|
| 594 |
-
"logits/chosen":
|
| 595 |
-
"logits/rejected":
|
| 596 |
-
"logps/chosen": -
|
| 597 |
-
"logps/rejected": -
|
| 598 |
-
"loss": 0.
|
| 599 |
"rewards/accuracies": 0.737500011920929,
|
| 600 |
-
"rewards/chosen": -
|
| 601 |
-
"rewards/margins": 0.
|
| 602 |
-
"rewards/rejected": -
|
| 603 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 604 |
},
|
| 605 |
{
|
| 606 |
"epoch": 0.82,
|
| 607 |
-
"learning_rate": 4.
|
| 608 |
-
"logits/chosen":
|
| 609 |
-
"logits/rejected":
|
| 610 |
-
"logps/chosen": -
|
| 611 |
-
"logps/rejected": -
|
| 612 |
-
"loss": 0.
|
| 613 |
-
"rewards/accuracies": 0.
|
| 614 |
-
"rewards/chosen": -
|
| 615 |
-
"rewards/margins": 0.
|
| 616 |
-
"rewards/rejected": -2.
|
| 617 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 618 |
},
|
| 619 |
{
|
| 620 |
"epoch": 0.84,
|
| 621 |
-
"learning_rate": 3.
|
| 622 |
-
"logits/chosen":
|
| 623 |
-
"logits/rejected":
|
| 624 |
-
"logps/chosen": -
|
| 625 |
-
"logps/rejected": -
|
| 626 |
-
"loss": 0.
|
| 627 |
-
"rewards/accuracies": 0.
|
| 628 |
-
"rewards/chosen": -
|
| 629 |
-
"rewards/margins": 0.
|
| 630 |
-
"rewards/rejected": -2.
|
| 631 |
-
"step":
|
| 632 |
},
|
| 633 |
{
|
| 634 |
"epoch": 0.84,
|
| 635 |
-
"
|
| 636 |
-
"
|
| 637 |
-
"
|
| 638 |
-
"
|
| 639 |
-
"
|
| 640 |
-
"
|
| 641 |
-
"
|
| 642 |
-
"
|
| 643 |
-
"
|
| 644 |
-
"
|
| 645 |
-
"
|
| 646 |
-
|
| 647 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 648 |
},
|
| 649 |
{
|
| 650 |
"epoch": 0.86,
|
| 651 |
-
"learning_rate":
|
| 652 |
-
"logits/chosen":
|
| 653 |
-
"logits/rejected":
|
| 654 |
-
"logps/chosen": -
|
| 655 |
-
"logps/rejected": -
|
| 656 |
-
"loss": 0.
|
| 657 |
-
"rewards/accuracies": 0.
|
| 658 |
-
"rewards/chosen": -
|
| 659 |
-
"rewards/margins": 0.
|
| 660 |
-
"rewards/rejected": -2.
|
| 661 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 662 |
},
|
| 663 |
{
|
| 664 |
"epoch": 0.88,
|
| 665 |
-
"learning_rate": 2.
|
| 666 |
-
"logits/chosen":
|
| 667 |
-
"logits/rejected":
|
| 668 |
-
"logps/chosen": -
|
| 669 |
-
"logps/rejected": -
|
| 670 |
-
"loss": 0.
|
| 671 |
-
"rewards/accuracies": 0.
|
| 672 |
-
"rewards/chosen": -
|
| 673 |
-
"rewards/margins": 0.
|
| 674 |
-
"rewards/rejected": -2.
|
| 675 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 676 |
},
|
| 677 |
{
|
| 678 |
"epoch": 0.9,
|
| 679 |
-
"learning_rate": 1.
|
| 680 |
-
"logits/chosen":
|
| 681 |
-
"logits/rejected":
|
| 682 |
-
"logps/chosen": -
|
| 683 |
-
"logps/rejected": -
|
| 684 |
-
"loss": 0.
|
| 685 |
-
"rewards/accuracies": 0.
|
| 686 |
-
"rewards/chosen": -
|
| 687 |
-
"rewards/margins": 0.
|
| 688 |
-
"rewards/rejected": -2.
|
| 689 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 690 |
},
|
| 691 |
{
|
| 692 |
"epoch": 0.92,
|
| 693 |
-
"learning_rate":
|
| 694 |
-
"logits/chosen":
|
| 695 |
-
"logits/rejected":
|
| 696 |
-
"logps/chosen": -
|
| 697 |
-
"logps/rejected": -
|
| 698 |
-
"loss": 0.
|
| 699 |
-
"rewards/accuracies": 0.
|
| 700 |
-
"rewards/chosen": -
|
| 701 |
-
"rewards/margins": 0.
|
| 702 |
-
"rewards/rejected": -
|
| 703 |
-
"step":
|
| 704 |
},
|
| 705 |
{
|
| 706 |
-
"epoch": 0.
|
| 707 |
-
"learning_rate":
|
| 708 |
-
"logits/chosen":
|
| 709 |
-
"logits/rejected":
|
| 710 |
-
"logps/chosen": -
|
| 711 |
-
"logps/rejected": -
|
| 712 |
-
"loss": 0.
|
| 713 |
"rewards/accuracies": 0.6937500238418579,
|
| 714 |
-
"rewards/chosen": -
|
| 715 |
-
"rewards/margins": 0.
|
| 716 |
-
"rewards/rejected": -2.
|
| 717 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 718 |
},
|
| 719 |
{
|
| 720 |
"epoch": 0.96,
|
| 721 |
-
"learning_rate":
|
| 722 |
-
"logits/chosen":
|
| 723 |
-
"logits/rejected":
|
| 724 |
-
"logps/chosen": -
|
| 725 |
-
"logps/rejected": -
|
| 726 |
-
"loss": 0.
|
| 727 |
-
"rewards/accuracies": 0.
|
| 728 |
-
"rewards/chosen": -
|
| 729 |
-
"rewards/margins": 0.
|
| 730 |
-
"rewards/rejected": -2.
|
| 731 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 732 |
},
|
| 733 |
{
|
| 734 |
"epoch": 0.98,
|
| 735 |
-
"learning_rate":
|
| 736 |
-
"logits/chosen":
|
| 737 |
-
"logits/rejected":
|
| 738 |
-
"logps/chosen": -
|
| 739 |
-
"logps/rejected": -
|
| 740 |
-
"loss": 0.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 741 |
"rewards/accuracies": 0.737500011920929,
|
| 742 |
-
"rewards/chosen": -
|
| 743 |
-
"rewards/margins": 0.
|
| 744 |
-
"rewards/rejected": -
|
| 745 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 746 |
},
|
| 747 |
{
|
| 748 |
"epoch": 1.0,
|
| 749 |
-
"step":
|
| 750 |
"total_flos": 0.0,
|
| 751 |
-
"train_loss": 0.
|
| 752 |
-
"train_runtime":
|
| 753 |
-
"train_samples_per_second":
|
| 754 |
-
"train_steps_per_second": 0.
|
| 755 |
}
|
| 756 |
],
|
| 757 |
"logging_steps": 10,
|
| 758 |
-
"max_steps":
|
| 759 |
"num_train_epochs": 1,
|
| 760 |
"save_steps": 100,
|
| 761 |
"total_flos": 0.0,
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": null,
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.9996020692399522,
|
| 5 |
"eval_steps": 100,
|
| 6 |
+
"global_step": 1256,
|
| 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": 3.968253968253968e-09,
|
| 14 |
+
"logits/chosen": -2.7193620204925537,
|
| 15 |
+
"logits/rejected": -2.698728084564209,
|
| 16 |
+
"logps/chosen": -182.0961456298828,
|
| 17 |
+
"logps/rejected": -172.47128295898438,
|
| 18 |
+
"loss": 0.6931,
|
| 19 |
"rewards/accuracies": 0.0,
|
| 20 |
"rewards/chosen": 0.0,
|
| 21 |
"rewards/margins": 0.0,
|
|
|
|
| 23 |
"step": 1
|
| 24 |
},
|
| 25 |
{
|
| 26 |
+
"epoch": 0.01,
|
| 27 |
+
"learning_rate": 3.968253968253968e-08,
|
| 28 |
+
"logits/chosen": -2.7041964530944824,
|
| 29 |
+
"logits/rejected": -2.6794540882110596,
|
| 30 |
+
"logps/chosen": -162.45831298828125,
|
| 31 |
+
"logps/rejected": -140.5693359375,
|
| 32 |
+
"loss": 0.6931,
|
| 33 |
+
"rewards/accuracies": 0.5486111044883728,
|
| 34 |
+
"rewards/chosen": 0.00032037965138442814,
|
| 35 |
+
"rewards/margins": 0.0004935775068588555,
|
| 36 |
+
"rewards/rejected": -0.00017319784092251211,
|
| 37 |
"step": 10
|
| 38 |
},
|
| 39 |
{
|
| 40 |
+
"epoch": 0.02,
|
| 41 |
+
"learning_rate": 7.936507936507936e-08,
|
| 42 |
+
"logits/chosen": -2.7177577018737793,
|
| 43 |
+
"logits/rejected": -2.7136425971984863,
|
| 44 |
+
"logps/chosen": -134.47242736816406,
|
| 45 |
+
"logps/rejected": -143.55604553222656,
|
| 46 |
+
"loss": 0.6931,
|
| 47 |
"rewards/accuracies": 0.550000011920929,
|
| 48 |
+
"rewards/chosen": 8.780837379163131e-05,
|
| 49 |
+
"rewards/margins": 0.00010721785656642169,
|
| 50 |
+
"rewards/rejected": -1.940951551659964e-05,
|
| 51 |
"step": 20
|
| 52 |
},
|
| 53 |
{
|
| 54 |
+
"epoch": 0.02,
|
| 55 |
+
"learning_rate": 1.1904761904761903e-07,
|
| 56 |
+
"logits/chosen": -2.6898293495178223,
|
| 57 |
+
"logits/rejected": -2.676154613494873,
|
| 58 |
+
"logps/chosen": -140.94692993164062,
|
| 59 |
+
"logps/rejected": -136.50369262695312,
|
| 60 |
+
"loss": 0.6931,
|
| 61 |
+
"rewards/accuracies": 0.4375,
|
| 62 |
+
"rewards/chosen": 0.0005466601578518748,
|
| 63 |
+
"rewards/margins": -0.00021456097601912916,
|
| 64 |
+
"rewards/rejected": 0.0007612211629748344,
|
| 65 |
"step": 30
|
| 66 |
},
|
| 67 |
{
|
| 68 |
+
"epoch": 0.03,
|
| 69 |
+
"learning_rate": 1.5873015873015872e-07,
|
| 70 |
+
"logits/chosen": -2.6958394050598145,
|
| 71 |
+
"logits/rejected": -2.686532974243164,
|
| 72 |
+
"logps/chosen": -134.98963928222656,
|
| 73 |
+
"logps/rejected": -144.46652221679688,
|
| 74 |
+
"loss": 0.6928,
|
| 75 |
+
"rewards/accuracies": 0.581250011920929,
|
| 76 |
+
"rewards/chosen": 0.0015748919686302543,
|
| 77 |
+
"rewards/margins": 0.0009769219905138016,
|
| 78 |
+
"rewards/rejected": 0.0005979698617011309,
|
| 79 |
"step": 40
|
| 80 |
},
|
| 81 |
{
|
| 82 |
+
"epoch": 0.04,
|
| 83 |
+
"learning_rate": 1.984126984126984e-07,
|
| 84 |
+
"logits/chosen": -2.7042899131774902,
|
| 85 |
+
"logits/rejected": -2.6861345767974854,
|
| 86 |
+
"logps/chosen": -149.71768188476562,
|
| 87 |
+
"logps/rejected": -145.0757293701172,
|
| 88 |
+
"loss": 0.6921,
|
| 89 |
+
"rewards/accuracies": 0.6000000238418579,
|
| 90 |
+
"rewards/chosen": 0.005199921317398548,
|
| 91 |
+
"rewards/margins": 0.0022330707870423794,
|
| 92 |
+
"rewards/rejected": 0.0029668500646948814,
|
| 93 |
"step": 50
|
| 94 |
},
|
| 95 |
{
|
| 96 |
+
"epoch": 0.05,
|
| 97 |
+
"learning_rate": 2.3809523809523806e-07,
|
| 98 |
+
"logits/chosen": -2.705153703689575,
|
| 99 |
+
"logits/rejected": -2.685439348220825,
|
| 100 |
+
"logps/chosen": -154.3783416748047,
|
| 101 |
+
"logps/rejected": -151.54519653320312,
|
| 102 |
+
"loss": 0.6912,
|
| 103 |
+
"rewards/accuracies": 0.543749988079071,
|
| 104 |
+
"rewards/chosen": 0.00569504126906395,
|
| 105 |
+
"rewards/margins": 0.0022000311873853207,
|
| 106 |
+
"rewards/rejected": 0.003495010081678629,
|
| 107 |
"step": 60
|
| 108 |
},
|
| 109 |
{
|
| 110 |
+
"epoch": 0.06,
|
| 111 |
+
"learning_rate": 2.7777777777777776e-07,
|
| 112 |
+
"logits/chosen": -2.7017154693603516,
|
| 113 |
+
"logits/rejected": -2.6924962997436523,
|
| 114 |
+
"logps/chosen": -146.3284149169922,
|
| 115 |
+
"logps/rejected": -138.79405212402344,
|
| 116 |
+
"loss": 0.6885,
|
| 117 |
+
"rewards/accuracies": 0.550000011920929,
|
| 118 |
+
"rewards/chosen": 0.010588793084025383,
|
| 119 |
+
"rewards/margins": 0.010192448273301125,
|
| 120 |
+
"rewards/rejected": 0.00039634370477870107,
|
| 121 |
"step": 70
|
| 122 |
},
|
| 123 |
{
|
| 124 |
+
"epoch": 0.06,
|
| 125 |
+
"learning_rate": 3.1746031746031743e-07,
|
| 126 |
+
"logits/chosen": -2.7155232429504395,
|
| 127 |
+
"logits/rejected": -2.696071147918701,
|
| 128 |
+
"logps/chosen": -141.80067443847656,
|
| 129 |
+
"logps/rejected": -147.0068817138672,
|
| 130 |
+
"loss": 0.6867,
|
| 131 |
+
"rewards/accuracies": 0.5562499761581421,
|
| 132 |
+
"rewards/chosen": 0.0049073463305830956,
|
| 133 |
+
"rewards/margins": 0.013599385507404804,
|
| 134 |
+
"rewards/rejected": -0.008692039176821709,
|
| 135 |
"step": 80
|
| 136 |
},
|
| 137 |
{
|
| 138 |
+
"epoch": 0.07,
|
| 139 |
+
"learning_rate": 3.5714285714285716e-07,
|
| 140 |
+
"logits/chosen": -2.7175304889678955,
|
| 141 |
+
"logits/rejected": -2.7080624103546143,
|
| 142 |
+
"logps/chosen": -153.12509155273438,
|
| 143 |
+
"logps/rejected": -146.53590393066406,
|
| 144 |
+
"loss": 0.6847,
|
| 145 |
+
"rewards/accuracies": 0.5874999761581421,
|
| 146 |
+
"rewards/chosen": -0.028871387243270874,
|
| 147 |
+
"rewards/margins": 0.017175236716866493,
|
| 148 |
+
"rewards/rejected": -0.046046625822782516,
|
| 149 |
"step": 90
|
| 150 |
},
|
| 151 |
{
|
| 152 |
+
"epoch": 0.08,
|
| 153 |
+
"learning_rate": 3.968253968253968e-07,
|
| 154 |
+
"logits/chosen": -2.7524733543395996,
|
| 155 |
+
"logits/rejected": -2.7452526092529297,
|
| 156 |
+
"logps/chosen": -163.88070678710938,
|
| 157 |
+
"logps/rejected": -163.61032104492188,
|
| 158 |
+
"loss": 0.6789,
|
| 159 |
+
"rewards/accuracies": 0.612500011920929,
|
| 160 |
+
"rewards/chosen": -0.0662173330783844,
|
| 161 |
+
"rewards/margins": 0.02977912127971649,
|
| 162 |
+
"rewards/rejected": -0.09599645435810089,
|
| 163 |
"step": 100
|
| 164 |
},
|
| 165 |
{
|
| 166 |
+
"epoch": 0.08,
|
| 167 |
+
"eval_logits/chosen": -2.7336502075195312,
|
| 168 |
+
"eval_logits/rejected": -2.7255024909973145,
|
| 169 |
+
"eval_logps/chosen": -155.19271850585938,
|
| 170 |
+
"eval_logps/rejected": -165.35523986816406,
|
| 171 |
+
"eval_loss": 0.6769910454750061,
|
| 172 |
+
"eval_rewards/accuracies": 0.5914179086685181,
|
| 173 |
+
"eval_rewards/chosen": -0.10619194805622101,
|
| 174 |
+
"eval_rewards/margins": 0.03601696714758873,
|
| 175 |
+
"eval_rewards/rejected": -0.14220890402793884,
|
| 176 |
+
"eval_runtime": 184.3291,
|
| 177 |
+
"eval_samples_per_second": 46.395,
|
| 178 |
+
"eval_steps_per_second": 0.727,
|
| 179 |
"step": 100
|
| 180 |
},
|
| 181 |
{
|
| 182 |
+
"epoch": 0.09,
|
| 183 |
+
"learning_rate": 4.365079365079365e-07,
|
| 184 |
+
"logits/chosen": -2.738532543182373,
|
| 185 |
+
"logits/rejected": -2.7273170948028564,
|
| 186 |
+
"logps/chosen": -164.2928009033203,
|
| 187 |
+
"logps/rejected": -160.19398498535156,
|
| 188 |
+
"loss": 0.6738,
|
| 189 |
+
"rewards/accuracies": 0.5625,
|
| 190 |
+
"rewards/chosen": -0.16211798787117004,
|
| 191 |
+
"rewards/margins": 0.03163355216383934,
|
| 192 |
+
"rewards/rejected": -0.19375154376029968,
|
| 193 |
"step": 110
|
| 194 |
},
|
| 195 |
{
|
| 196 |
+
"epoch": 0.1,
|
| 197 |
+
"learning_rate": 4.761904761904761e-07,
|
| 198 |
+
"logits/chosen": -2.7289297580718994,
|
| 199 |
+
"logits/rejected": -2.705962657928467,
|
| 200 |
+
"logps/chosen": -196.69662475585938,
|
| 201 |
+
"logps/rejected": -197.2833251953125,
|
| 202 |
+
"loss": 0.661,
|
| 203 |
+
"rewards/accuracies": 0.59375,
|
| 204 |
+
"rewards/chosen": -0.2917623221874237,
|
| 205 |
+
"rewards/margins": 0.08966299891471863,
|
| 206 |
+
"rewards/rejected": -0.38142532110214233,
|
| 207 |
"step": 120
|
| 208 |
},
|
| 209 |
{
|
| 210 |
+
"epoch": 0.1,
|
| 211 |
+
"learning_rate": 4.999845414634076e-07,
|
| 212 |
+
"logits/chosen": -2.658005475997925,
|
| 213 |
+
"logits/rejected": -2.6317684650421143,
|
| 214 |
+
"logps/chosen": -187.4532928466797,
|
| 215 |
+
"logps/rejected": -188.37689208984375,
|
| 216 |
+
"loss": 0.6542,
|
| 217 |
+
"rewards/accuracies": 0.6312500238418579,
|
| 218 |
+
"rewards/chosen": -0.3956056833267212,
|
| 219 |
+
"rewards/margins": 0.12092368304729462,
|
| 220 |
+
"rewards/rejected": -0.5165294408798218,
|
| 221 |
"step": 130
|
| 222 |
},
|
| 223 |
{
|
| 224 |
+
"epoch": 0.11,
|
| 225 |
+
"learning_rate": 4.998106548810311e-07,
|
| 226 |
+
"logits/chosen": -2.6906683444976807,
|
| 227 |
+
"logits/rejected": -2.6913747787475586,
|
| 228 |
+
"logps/chosen": -199.67568969726562,
|
| 229 |
+
"logps/rejected": -253.02487182617188,
|
| 230 |
+
"loss": 0.6171,
|
| 231 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 232 |
+
"rewards/chosen": -0.4227059781551361,
|
| 233 |
+
"rewards/margins": 0.27536457777023315,
|
| 234 |
+
"rewards/rejected": -0.6980706453323364,
|
| 235 |
"step": 140
|
| 236 |
},
|
| 237 |
{
|
| 238 |
+
"epoch": 0.12,
|
| 239 |
+
"learning_rate": 4.994436933879359e-07,
|
| 240 |
+
"logits/chosen": -2.6662166118621826,
|
| 241 |
+
"logits/rejected": -2.644784927368164,
|
| 242 |
+
"logps/chosen": -197.07180786132812,
|
| 243 |
+
"logps/rejected": -198.4012908935547,
|
| 244 |
+
"loss": 0.6395,
|
| 245 |
+
"rewards/accuracies": 0.6187499761581421,
|
| 246 |
+
"rewards/chosen": -0.3597154915332794,
|
| 247 |
+
"rewards/margins": 0.13716872036457062,
|
| 248 |
+
"rewards/rejected": -0.49688419699668884,
|
| 249 |
"step": 150
|
| 250 |
},
|
| 251 |
{
|
| 252 |
+
"epoch": 0.13,
|
| 253 |
+
"learning_rate": 4.988839406031596e-07,
|
| 254 |
+
"logits/chosen": -2.647681474685669,
|
| 255 |
+
"logits/rejected": -2.6395888328552246,
|
| 256 |
+
"logps/chosen": -182.04420471191406,
|
| 257 |
+
"logps/rejected": -206.59780883789062,
|
| 258 |
+
"loss": 0.629,
|
| 259 |
+
"rewards/accuracies": 0.6499999761581421,
|
| 260 |
+
"rewards/chosen": -0.3573477864265442,
|
| 261 |
+
"rewards/margins": 0.2222837507724762,
|
| 262 |
+
"rewards/rejected": -0.579631507396698,
|
| 263 |
"step": 160
|
| 264 |
},
|
| 265 |
{
|
| 266 |
+
"epoch": 0.14,
|
| 267 |
+
"learning_rate": 4.981318291512395e-07,
|
| 268 |
+
"logits/chosen": -2.619232654571533,
|
| 269 |
+
"logits/rejected": -2.598362684249878,
|
| 270 |
+
"logps/chosen": -227.0933380126953,
|
| 271 |
+
"logps/rejected": -230.9747772216797,
|
| 272 |
+
"loss": 0.6242,
|
| 273 |
+
"rewards/accuracies": 0.637499988079071,
|
| 274 |
+
"rewards/chosen": -0.7434185743331909,
|
| 275 |
+
"rewards/margins": 0.21749505400657654,
|
| 276 |
+
"rewards/rejected": -0.9609137773513794,
|
| 277 |
"step": 170
|
| 278 |
},
|
| 279 |
{
|
| 280 |
+
"epoch": 0.14,
|
| 281 |
+
"learning_rate": 4.971879403278432e-07,
|
| 282 |
+
"logits/chosen": -2.5654754638671875,
|
| 283 |
+
"logits/rejected": -2.5364232063293457,
|
| 284 |
+
"logps/chosen": -241.6617431640625,
|
| 285 |
+
"logps/rejected": -245.66268920898438,
|
| 286 |
+
"loss": 0.6151,
|
| 287 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 288 |
+
"rewards/chosen": -0.7342535257339478,
|
| 289 |
+
"rewards/margins": 0.23685339093208313,
|
| 290 |
+
"rewards/rejected": -0.9711068868637085,
|
| 291 |
"step": 180
|
| 292 |
},
|
| 293 |
{
|
| 294 |
+
"epoch": 0.15,
|
| 295 |
+
"learning_rate": 4.960530036504941e-07,
|
| 296 |
+
"logits/chosen": -2.5271048545837402,
|
| 297 |
+
"logits/rejected": -2.486818790435791,
|
| 298 |
+
"logps/chosen": -235.6089630126953,
|
| 299 |
+
"logps/rejected": -251.17758178710938,
|
| 300 |
+
"loss": 0.6215,
|
| 301 |
"rewards/accuracies": 0.699999988079071,
|
| 302 |
+
"rewards/chosen": -0.807177722454071,
|
| 303 |
+
"rewards/margins": 0.28561535477638245,
|
| 304 |
+
"rewards/rejected": -1.0927931070327759,
|
| 305 |
"step": 190
|
| 306 |
},
|
| 307 |
{
|
| 308 |
+
"epoch": 0.16,
|
| 309 |
+
"learning_rate": 4.947278962947386e-07,
|
| 310 |
+
"logits/chosen": -2.4217896461486816,
|
| 311 |
+
"logits/rejected": -2.413295269012451,
|
| 312 |
+
"logps/chosen": -251.0736083984375,
|
| 313 |
+
"logps/rejected": -268.6098937988281,
|
| 314 |
+
"loss": 0.6062,
|
| 315 |
+
"rewards/accuracies": 0.6625000238418579,
|
| 316 |
+
"rewards/chosen": -1.086307406425476,
|
| 317 |
+
"rewards/margins": 0.24874301254749298,
|
| 318 |
+
"rewards/rejected": -1.3350504636764526,
|
| 319 |
"step": 200
|
| 320 |
},
|
| 321 |
{
|
| 322 |
+
"epoch": 0.16,
|
| 323 |
+
"eval_logits/chosen": -2.3855514526367188,
|
| 324 |
+
"eval_logits/rejected": -2.369593858718872,
|
| 325 |
+
"eval_logps/chosen": -246.6970672607422,
|
| 326 |
+
"eval_logps/rejected": -289.8621826171875,
|
| 327 |
+
"eval_loss": 0.6079375743865967,
|
| 328 |
+
"eval_rewards/accuracies": 0.66697758436203,
|
| 329 |
+
"eval_rewards/chosen": -1.021235704421997,
|
| 330 |
+
"eval_rewards/margins": 0.3660426437854767,
|
| 331 |
+
"eval_rewards/rejected": -1.3872781991958618,
|
| 332 |
+
"eval_runtime": 184.172,
|
| 333 |
+
"eval_samples_per_second": 46.435,
|
| 334 |
+
"eval_steps_per_second": 0.728,
|
| 335 |
"step": 200
|
| 336 |
},
|
| 337 |
{
|
| 338 |
+
"epoch": 0.17,
|
| 339 |
+
"learning_rate": 4.932136424161899e-07,
|
| 340 |
+
"logits/chosen": -2.3366785049438477,
|
| 341 |
+
"logits/rejected": -2.3228511810302734,
|
| 342 |
+
"logps/chosen": -266.292236328125,
|
| 343 |
+
"logps/rejected": -300.22894287109375,
|
| 344 |
+
"loss": 0.5893,
|
| 345 |
+
"rewards/accuracies": 0.5874999761581421,
|
| 346 |
+
"rewards/chosen": -1.2130026817321777,
|
| 347 |
+
"rewards/margins": 0.3487839996814728,
|
| 348 |
+
"rewards/rejected": -1.5617868900299072,
|
| 349 |
"step": 210
|
| 350 |
},
|
| 351 |
{
|
| 352 |
+
"epoch": 0.18,
|
| 353 |
+
"learning_rate": 4.915114123589732e-07,
|
| 354 |
+
"logits/chosen": -2.321228504180908,
|
| 355 |
+
"logits/rejected": -2.3033699989318848,
|
| 356 |
+
"logps/chosen": -336.34161376953125,
|
| 357 |
+
"logps/rejected": -373.39935302734375,
|
| 358 |
+
"loss": 0.612,
|
| 359 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 360 |
+
"rewards/chosen": -1.9529145956039429,
|
| 361 |
+
"rewards/margins": 0.2863468527793884,
|
| 362 |
+
"rewards/rejected": -2.2392613887786865,
|
| 363 |
"step": 220
|
| 364 |
},
|
| 365 |
{
|
| 366 |
+
"epoch": 0.18,
|
| 367 |
+
"learning_rate": 4.896225217511849e-07,
|
| 368 |
+
"logits/chosen": -2.4310107231140137,
|
| 369 |
+
"logits/rejected": -2.422048568725586,
|
| 370 |
+
"logps/chosen": -291.1025695800781,
|
| 371 |
+
"logps/rejected": -328.18963623046875,
|
| 372 |
+
"loss": 0.6079,
|
| 373 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 374 |
+
"rewards/chosen": -1.4314143657684326,
|
| 375 |
+
"rewards/margins": 0.3364001214504242,
|
| 376 |
+
"rewards/rejected": -1.7678143978118896,
|
| 377 |
"step": 230
|
| 378 |
},
|
| 379 |
{
|
| 380 |
+
"epoch": 0.19,
|
| 381 |
+
"learning_rate": 4.875484304880629e-07,
|
| 382 |
+
"logits/chosen": -2.3412394523620605,
|
| 383 |
+
"logits/rejected": -2.309183120727539,
|
| 384 |
+
"logps/chosen": -280.8785705566406,
|
| 385 |
+
"logps/rejected": -308.54132080078125,
|
| 386 |
+
"loss": 0.613,
|
| 387 |
+
"rewards/accuracies": 0.612500011920929,
|
| 388 |
+
"rewards/chosen": -1.309309482574463,
|
| 389 |
+
"rewards/margins": 0.3731766939163208,
|
| 390 |
+
"rewards/rejected": -1.6824861764907837,
|
| 391 |
"step": 240
|
| 392 |
},
|
| 393 |
{
|
| 394 |
+
"epoch": 0.2,
|
| 395 |
+
"learning_rate": 4.852907416036558e-07,
|
| 396 |
+
"logits/chosen": -2.415271282196045,
|
| 397 |
+
"logits/rejected": -2.4072234630584717,
|
| 398 |
+
"logps/chosen": -243.56332397460938,
|
| 399 |
+
"logps/rejected": -298.7532043457031,
|
| 400 |
+
"loss": 0.591,
|
| 401 |
+
"rewards/accuracies": 0.737500011920929,
|
| 402 |
+
"rewards/chosen": -0.967076301574707,
|
| 403 |
+
"rewards/margins": 0.4581146240234375,
|
| 404 |
+
"rewards/rejected": -1.4251911640167236,
|
| 405 |
"step": 250
|
| 406 |
},
|
| 407 |
{
|
| 408 |
+
"epoch": 0.21,
|
| 409 |
+
"learning_rate": 4.828512000318616e-07,
|
| 410 |
+
"logits/chosen": -2.3924427032470703,
|
| 411 |
+
"logits/rejected": -2.3613152503967285,
|
| 412 |
+
"logps/chosen": -266.86572265625,
|
| 413 |
+
"logps/rejected": -304.2983093261719,
|
| 414 |
+
"loss": 0.5986,
|
| 415 |
+
"rewards/accuracies": 0.668749988079071,
|
| 416 |
+
"rewards/chosen": -1.2214807271957397,
|
| 417 |
+
"rewards/margins": 0.4553411602973938,
|
| 418 |
+
"rewards/rejected": -1.6768219470977783,
|
| 419 |
"step": 260
|
| 420 |
},
|
| 421 |
{
|
| 422 |
+
"epoch": 0.21,
|
| 423 |
+
"learning_rate": 4.802316912577946e-07,
|
| 424 |
+
"logits/chosen": -2.4108529090881348,
|
| 425 |
+
"logits/rejected": -2.3902478218078613,
|
| 426 |
+
"logps/chosen": -252.7959442138672,
|
| 427 |
+
"logps/rejected": -295.266357421875,
|
| 428 |
+
"loss": 0.5917,
|
| 429 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 430 |
+
"rewards/chosen": -1.0824626684188843,
|
| 431 |
+
"rewards/margins": 0.39643940329551697,
|
| 432 |
+
"rewards/rejected": -1.4789022207260132,
|
| 433 |
"step": 270
|
| 434 |
},
|
| 435 |
{
|
| 436 |
+
"epoch": 0.22,
|
| 437 |
+
"learning_rate": 4.774342398605221e-07,
|
| 438 |
+
"logits/chosen": -2.3505263328552246,
|
| 439 |
+
"logits/rejected": -2.2942967414855957,
|
| 440 |
+
"logps/chosen": -279.871337890625,
|
| 441 |
+
"logps/rejected": -300.4220886230469,
|
| 442 |
+
"loss": 0.5979,
|
| 443 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 444 |
+
"rewards/chosen": -1.1787078380584717,
|
| 445 |
+
"rewards/margins": 0.430286169052124,
|
| 446 |
+
"rewards/rejected": -1.6089938879013062,
|
| 447 |
"step": 280
|
| 448 |
},
|
| 449 |
{
|
| 450 |
+
"epoch": 0.23,
|
| 451 |
+
"learning_rate": 4.744610079482978e-07,
|
| 452 |
+
"logits/chosen": -2.3269264698028564,
|
| 453 |
+
"logits/rejected": -2.2910802364349365,
|
| 454 |
+
"logps/chosen": -255.27706909179688,
|
| 455 |
+
"logps/rejected": -281.60137939453125,
|
| 456 |
+
"loss": 0.5853,
|
| 457 |
+
"rewards/accuracies": 0.6875,
|
| 458 |
+
"rewards/chosen": -1.1753785610198975,
|
| 459 |
+
"rewards/margins": 0.3495523929595947,
|
| 460 |
+
"rewards/rejected": -1.5249310731887817,
|
| 461 |
"step": 290
|
| 462 |
},
|
| 463 |
{
|
| 464 |
+
"epoch": 0.24,
|
| 465 |
+
"learning_rate": 4.713142934875005e-07,
|
| 466 |
+
"logits/chosen": -2.2868428230285645,
|
| 467 |
+
"logits/rejected": -2.2631592750549316,
|
| 468 |
+
"logps/chosen": -284.2200012207031,
|
| 469 |
+
"logps/rejected": -322.45269775390625,
|
| 470 |
+
"loss": 0.5965,
|
| 471 |
"rewards/accuracies": 0.675000011920929,
|
| 472 |
+
"rewards/chosen": -1.336501955986023,
|
| 473 |
+
"rewards/margins": 0.3968800902366638,
|
| 474 |
+
"rewards/rejected": -1.733382225036621,
|
| 475 |
"step": 300
|
| 476 |
},
|
| 477 |
{
|
| 478 |
+
"epoch": 0.24,
|
| 479 |
+
"eval_logits/chosen": -2.265592098236084,
|
| 480 |
+
"eval_logits/rejected": -2.244987964630127,
|
| 481 |
+
"eval_logps/chosen": -282.3620910644531,
|
| 482 |
+
"eval_logps/rejected": -331.2099609375,
|
| 483 |
+
"eval_loss": 0.5907339453697205,
|
| 484 |
+
"eval_rewards/accuracies": 0.6623134613037109,
|
| 485 |
+
"eval_rewards/chosen": -1.3778856992721558,
|
| 486 |
+
"eval_rewards/margins": 0.42287060618400574,
|
| 487 |
+
"eval_rewards/rejected": -1.8007562160491943,
|
| 488 |
+
"eval_runtime": 184.1739,
|
| 489 |
+
"eval_samples_per_second": 46.434,
|
| 490 |
+
"eval_steps_per_second": 0.728,
|
| 491 |
"step": 300
|
| 492 |
},
|
| 493 |
{
|
| 494 |
+
"epoch": 0.25,
|
| 495 |
+
"learning_rate": 4.679965285265706e-07,
|
| 496 |
+
"logits/chosen": -2.2354235649108887,
|
| 497 |
+
"logits/rejected": -2.23685884475708,
|
| 498 |
+
"logps/chosen": -277.09283447265625,
|
| 499 |
+
"logps/rejected": -347.7145080566406,
|
| 500 |
+
"loss": 0.5612,
|
| 501 |
+
"rewards/accuracies": 0.6875,
|
| 502 |
+
"rewards/chosen": -1.3514426946640015,
|
| 503 |
+
"rewards/margins": 0.4907970428466797,
|
| 504 |
+
"rewards/rejected": -1.8422397375106812,
|
| 505 |
"step": 310
|
| 506 |
},
|
| 507 |
{
|
| 508 |
+
"epoch": 0.25,
|
| 509 |
+
"learning_rate": 4.64510277316316e-07,
|
| 510 |
+
"logits/chosen": -2.2262344360351562,
|
| 511 |
+
"logits/rejected": -2.226029634475708,
|
| 512 |
+
"logps/chosen": -271.74212646484375,
|
| 513 |
+
"logps/rejected": -332.5010986328125,
|
| 514 |
+
"loss": 0.5903,
|
| 515 |
+
"rewards/accuracies": 0.612500011920929,
|
| 516 |
+
"rewards/chosen": -1.3337775468826294,
|
| 517 |
+
"rewards/margins": 0.39512914419174194,
|
| 518 |
+
"rewards/rejected": -1.7289068698883057,
|
| 519 |
"step": 320
|
| 520 |
},
|
| 521 |
{
|
| 522 |
+
"epoch": 0.26,
|
| 523 |
+
"learning_rate": 4.6085823432804137e-07,
|
| 524 |
+
"logits/chosen": -2.2451891899108887,
|
| 525 |
+
"logits/rejected": -2.2502384185791016,
|
| 526 |
+
"logps/chosen": -250.6347198486328,
|
| 527 |
+
"logps/rejected": -333.8939208984375,
|
| 528 |
+
"loss": 0.5722,
|
| 529 |
+
"rewards/accuracies": 0.762499988079071,
|
| 530 |
+
"rewards/chosen": -1.1326004266738892,
|
| 531 |
+
"rewards/margins": 0.5066065192222595,
|
| 532 |
+
"rewards/rejected": -1.639206886291504,
|
| 533 |
"step": 330
|
| 534 |
},
|
| 535 |
{
|
| 536 |
+
"epoch": 0.27,
|
| 537 |
+
"learning_rate": 4.570432221710314e-07,
|
| 538 |
+
"logits/chosen": -2.0656931400299072,
|
| 539 |
+
"logits/rejected": -2.0213730335235596,
|
| 540 |
+
"logps/chosen": -318.232177734375,
|
| 541 |
+
"logps/rejected": -369.13311767578125,
|
| 542 |
+
"loss": 0.5766,
|
| 543 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 544 |
+
"rewards/chosen": -1.584176778793335,
|
| 545 |
+
"rewards/margins": 0.5901076197624207,
|
| 546 |
+
"rewards/rejected": -2.1742844581604004,
|
| 547 |
"step": 340
|
| 548 |
},
|
| 549 |
+
{
|
| 550 |
+
"epoch": 0.28,
|
| 551 |
+
"learning_rate": 4.5306818941099866e-07,
|
| 552 |
+
"logits/chosen": -1.9084612131118774,
|
| 553 |
+
"logits/rejected": -1.8514792919158936,
|
| 554 |
+
"logps/chosen": -316.9821472167969,
|
| 555 |
+
"logps/rejected": -352.9412841796875,
|
| 556 |
+
"loss": 0.5825,
|
| 557 |
+
"rewards/accuracies": 0.65625,
|
| 558 |
+
"rewards/chosen": -1.5627154111862183,
|
| 559 |
+
"rewards/margins": 0.5152220726013184,
|
| 560 |
+
"rewards/rejected": -2.077937364578247,
|
| 561 |
+
"step": 350
|
| 562 |
+
},
|
| 563 |
+
{
|
| 564 |
+
"epoch": 0.29,
|
| 565 |
+
"learning_rate": 4.4893620829118124e-07,
|
| 566 |
+
"logits/chosen": -1.8860156536102295,
|
| 567 |
+
"logits/rejected": -1.8301204442977905,
|
| 568 |
+
"logps/chosen": -309.8200378417969,
|
| 569 |
+
"logps/rejected": -362.0408935546875,
|
| 570 |
+
"loss": 0.5755,
|
| 571 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 572 |
+
"rewards/chosen": -1.5869390964508057,
|
| 573 |
+
"rewards/margins": 0.49348369240760803,
|
| 574 |
+
"rewards/rejected": -2.080422878265381,
|
| 575 |
+
"step": 360
|
| 576 |
+
},
|
| 577 |
+
{
|
| 578 |
+
"epoch": 0.29,
|
| 579 |
+
"learning_rate": 4.4465047235785185e-07,
|
| 580 |
+
"logits/chosen": -1.6610889434814453,
|
| 581 |
+
"logits/rejected": -1.585129737854004,
|
| 582 |
+
"logps/chosen": -321.8608703613281,
|
| 583 |
+
"logps/rejected": -380.31036376953125,
|
| 584 |
+
"loss": 0.5697,
|
| 585 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 586 |
+
"rewards/chosen": -1.774713158607483,
|
| 587 |
+
"rewards/margins": 0.6593302488327026,
|
| 588 |
+
"rewards/rejected": -2.4340434074401855,
|
| 589 |
+
"step": 370
|
| 590 |
+
},
|
| 591 |
+
{
|
| 592 |
+
"epoch": 0.3,
|
| 593 |
+
"learning_rate": 4.40214293992074e-07,
|
| 594 |
+
"logits/chosen": -1.385825753211975,
|
| 595 |
+
"logits/rejected": -1.31913161277771,
|
| 596 |
+
"logps/chosen": -377.07269287109375,
|
| 597 |
+
"logps/rejected": -459.5557556152344,
|
| 598 |
+
"loss": 0.5818,
|
| 599 |
+
"rewards/accuracies": 0.706250011920929,
|
| 600 |
+
"rewards/chosen": -2.1890993118286133,
|
| 601 |
+
"rewards/margins": 0.7521292567253113,
|
| 602 |
+
"rewards/rejected": -2.9412286281585693,
|
| 603 |
+
"step": 380
|
| 604 |
+
},
|
| 605 |
+
{
|
| 606 |
+
"epoch": 0.31,
|
| 607 |
+
"learning_rate": 4.3563110184961234e-07,
|
| 608 |
+
"logits/chosen": -1.5089499950408936,
|
| 609 |
+
"logits/rejected": -1.4075387716293335,
|
| 610 |
+
"logps/chosen": -338.3626708984375,
|
| 611 |
+
"logps/rejected": -396.67578125,
|
| 612 |
+
"loss": 0.5584,
|
| 613 |
+
"rewards/accuracies": 0.737500011920929,
|
| 614 |
+
"rewards/chosen": -1.9042552709579468,
|
| 615 |
+
"rewards/margins": 0.5932050347328186,
|
| 616 |
+
"rewards/rejected": -2.49746036529541,
|
| 617 |
+
"step": 390
|
| 618 |
+
},
|
| 619 |
+
{
|
| 620 |
+
"epoch": 0.32,
|
| 621 |
+
"learning_rate": 4.3090443821097566e-07,
|
| 622 |
+
"logits/chosen": -1.2587625980377197,
|
| 623 |
+
"logits/rejected": -1.2017955780029297,
|
| 624 |
+
"logps/chosen": -309.43377685546875,
|
| 625 |
+
"logps/rejected": -372.00531005859375,
|
| 626 |
+
"loss": 0.5729,
|
| 627 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 628 |
+
"rewards/chosen": -1.9031288623809814,
|
| 629 |
+
"rewards/margins": 0.5602144598960876,
|
| 630 |
+
"rewards/rejected": -2.4633431434631348,
|
| 631 |
+
"step": 400
|
| 632 |
+
},
|
| 633 |
+
{
|
| 634 |
+
"epoch": 0.32,
|
| 635 |
+
"eval_logits/chosen": -1.3760210275650024,
|
| 636 |
+
"eval_logits/rejected": -1.2920024394989014,
|
| 637 |
+
"eval_logps/chosen": -312.20635986328125,
|
| 638 |
+
"eval_logps/rejected": -375.1720275878906,
|
| 639 |
+
"eval_loss": 0.5711147785186768,
|
| 640 |
+
"eval_rewards/accuracies": 0.6828358173370361,
|
| 641 |
+
"eval_rewards/chosen": -1.676328182220459,
|
| 642 |
+
"eval_rewards/margins": 0.5640482306480408,
|
| 643 |
+
"eval_rewards/rejected": -2.2403764724731445,
|
| 644 |
+
"eval_runtime": 184.465,
|
| 645 |
+
"eval_samples_per_second": 46.361,
|
| 646 |
+
"eval_steps_per_second": 0.726,
|
| 647 |
+
"step": 400
|
| 648 |
+
},
|
| 649 |
+
{
|
| 650 |
+
"epoch": 0.33,
|
| 651 |
+
"learning_rate": 4.2603795624364195e-07,
|
| 652 |
+
"logits/chosen": -1.2894772291183472,
|
| 653 |
+
"logits/rejected": -1.23129141330719,
|
| 654 |
+
"logps/chosen": -299.457275390625,
|
| 655 |
+
"logps/rejected": -370.8555908203125,
|
| 656 |
+
"loss": 0.5666,
|
| 657 |
+
"rewards/accuracies": 0.6875,
|
| 658 |
+
"rewards/chosen": -1.6054102182388306,
|
| 659 |
+
"rewards/margins": 0.5984233021736145,
|
| 660 |
+
"rewards/rejected": -2.203833818435669,
|
| 661 |
+
"step": 410
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"epoch": 0.33,
|
| 665 |
+
"learning_rate": 4.210354171785795e-07,
|
| 666 |
+
"logits/chosen": -1.022984266281128,
|
| 667 |
+
"logits/rejected": -0.9285897016525269,
|
| 668 |
+
"logps/chosen": -324.4284973144531,
|
| 669 |
+
"logps/rejected": -385.0074157714844,
|
| 670 |
+
"loss": 0.5596,
|
| 671 |
+
"rewards/accuracies": 0.6875,
|
| 672 |
+
"rewards/chosen": -1.921677589416504,
|
| 673 |
+
"rewards/margins": 0.5404387712478638,
|
| 674 |
+
"rewards/rejected": -2.4621164798736572,
|
| 675 |
+
"step": 420
|
| 676 |
+
},
|
| 677 |
+
{
|
| 678 |
+
"epoch": 0.34,
|
| 679 |
+
"learning_rate": 4.15900687403248e-07,
|
| 680 |
+
"logits/chosen": -0.8059805631637573,
|
| 681 |
+
"logits/rejected": -0.7196700572967529,
|
| 682 |
+
"logps/chosen": -353.788330078125,
|
| 683 |
+
"logps/rejected": -411.4853515625,
|
| 684 |
+
"loss": 0.5865,
|
| 685 |
+
"rewards/accuracies": 0.643750011920929,
|
| 686 |
+
"rewards/chosen": -2.1321234703063965,
|
| 687 |
+
"rewards/margins": 0.463266521692276,
|
| 688 |
+
"rewards/rejected": -2.5953898429870605,
|
| 689 |
+
"step": 430
|
| 690 |
+
},
|
| 691 |
+
{
|
| 692 |
+
"epoch": 0.35,
|
| 693 |
+
"learning_rate": 4.1063773547332584e-07,
|
| 694 |
+
"logits/chosen": -0.9645301699638367,
|
| 695 |
+
"logits/rejected": -0.7601315975189209,
|
| 696 |
+
"logps/chosen": -346.8272705078125,
|
| 697 |
+
"logps/rejected": -392.2935791015625,
|
| 698 |
+
"loss": 0.5591,
|
| 699 |
+
"rewards/accuracies": 0.762499988079071,
|
| 700 |
+
"rewards/chosen": -1.9153356552124023,
|
| 701 |
+
"rewards/margins": 0.5891679525375366,
|
| 702 |
+
"rewards/rejected": -2.5045037269592285,
|
| 703 |
+
"step": 440
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 0.36,
|
| 707 |
+
"learning_rate": 4.0525062904547276e-07,
|
| 708 |
+
"logits/chosen": -0.608537495136261,
|
| 709 |
+
"logits/rejected": -0.47767123579978943,
|
| 710 |
+
"logps/chosen": -341.55364990234375,
|
| 711 |
+
"logps/rejected": -434.1073303222656,
|
| 712 |
+
"loss": 0.5687,
|
| 713 |
+
"rewards/accuracies": 0.737500011920929,
|
| 714 |
+
"rewards/chosen": -2.105318069458008,
|
| 715 |
+
"rewards/margins": 0.6994394659996033,
|
| 716 |
+
"rewards/rejected": -2.8047571182250977,
|
| 717 |
+
"step": 450
|
| 718 |
+
},
|
| 719 |
+
{
|
| 720 |
+
"epoch": 0.37,
|
| 721 |
+
"learning_rate": 3.997435317334988e-07,
|
| 722 |
+
"logits/chosen": -0.6356207132339478,
|
| 723 |
+
"logits/rejected": -0.25634175539016724,
|
| 724 |
+
"logps/chosen": -384.43780517578125,
|
| 725 |
+
"logps/rejected": -419.24176025390625,
|
| 726 |
+
"loss": 0.5608,
|
| 727 |
+
"rewards/accuracies": 0.731249988079071,
|
| 728 |
+
"rewards/chosen": -2.2970900535583496,
|
| 729 |
+
"rewards/margins": 0.6535100340843201,
|
| 730 |
+
"rewards/rejected": -2.9506001472473145,
|
| 731 |
+
"step": 460
|
| 732 |
+
},
|
| 733 |
+
{
|
| 734 |
+
"epoch": 0.37,
|
| 735 |
+
"learning_rate": 3.941206998903701e-07,
|
| 736 |
+
"logits/chosen": -1.0318920612335205,
|
| 737 |
+
"logits/rejected": -0.7451022267341614,
|
| 738 |
+
"logps/chosen": -338.9430236816406,
|
| 739 |
+
"logps/rejected": -384.64111328125,
|
| 740 |
+
"loss": 0.5678,
|
| 741 |
+
"rewards/accuracies": 0.737500011920929,
|
| 742 |
+
"rewards/chosen": -1.9643396139144897,
|
| 743 |
+
"rewards/margins": 0.5402536392211914,
|
| 744 |
+
"rewards/rejected": -2.5045928955078125,
|
| 745 |
+
"step": 470
|
| 746 |
+
},
|
| 747 |
+
{
|
| 748 |
+
"epoch": 0.38,
|
| 749 |
+
"learning_rate": 3.8838647931853684e-07,
|
| 750 |
+
"logits/chosen": -0.6847028732299805,
|
| 751 |
+
"logits/rejected": -0.5548251867294312,
|
| 752 |
+
"logps/chosen": -339.61456298828125,
|
| 753 |
+
"logps/rejected": -435.32061767578125,
|
| 754 |
+
"loss": 0.5814,
|
| 755 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 756 |
+
"rewards/chosen": -2.1349122524261475,
|
| 757 |
+
"rewards/margins": 0.7573872804641724,
|
| 758 |
+
"rewards/rejected": -2.8923001289367676,
|
| 759 |
+
"step": 480
|
| 760 |
+
},
|
| 761 |
+
{
|
| 762 |
+
"epoch": 0.39,
|
| 763 |
+
"learning_rate": 3.825453019111281e-07,
|
| 764 |
+
"logits/chosen": -0.5378957986831665,
|
| 765 |
+
"logits/rejected": -0.28533270955085754,
|
| 766 |
+
"logps/chosen": -363.78570556640625,
|
| 767 |
+
"logps/rejected": -430.11749267578125,
|
| 768 |
+
"loss": 0.5327,
|
| 769 |
+
"rewards/accuracies": 0.699999988079071,
|
| 770 |
+
"rewards/chosen": -2.134934425354004,
|
| 771 |
+
"rewards/margins": 0.6089809536933899,
|
| 772 |
+
"rewards/rejected": -2.743915319442749,
|
| 773 |
+
"step": 490
|
| 774 |
+
},
|
| 775 |
+
{
|
| 776 |
+
"epoch": 0.4,
|
| 777 |
+
"learning_rate": 3.7660168222660824e-07,
|
| 778 |
+
"logits/chosen": -0.6318235397338867,
|
| 779 |
+
"logits/rejected": -0.5071814656257629,
|
| 780 |
+
"logps/chosen": -350.5252380371094,
|
| 781 |
+
"logps/rejected": -421.93353271484375,
|
| 782 |
+
"loss": 0.5645,
|
| 783 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 784 |
+
"rewards/chosen": -2.062009572982788,
|
| 785 |
+
"rewards/margins": 0.5333147048950195,
|
| 786 |
+
"rewards/rejected": -2.5953242778778076,
|
| 787 |
+
"step": 500
|
| 788 |
+
},
|
| 789 |
+
{
|
| 790 |
+
"epoch": 0.4,
|
| 791 |
+
"eval_logits/chosen": -0.7860146760940552,
|
| 792 |
+
"eval_logits/rejected": -0.6090859770774841,
|
| 793 |
+
"eval_logps/chosen": -351.7882995605469,
|
| 794 |
+
"eval_logps/rejected": -419.81939697265625,
|
| 795 |
+
"eval_loss": 0.5639454126358032,
|
| 796 |
+
"eval_rewards/accuracies": 0.6986940503120422,
|
| 797 |
+
"eval_rewards/chosen": -2.0721471309661865,
|
| 798 |
+
"eval_rewards/margins": 0.6147031188011169,
|
| 799 |
+
"eval_rewards/rejected": -2.6868505477905273,
|
| 800 |
+
"eval_runtime": 184.5772,
|
| 801 |
+
"eval_samples_per_second": 46.333,
|
| 802 |
+
"eval_steps_per_second": 0.726,
|
| 803 |
+
"step": 500
|
| 804 |
+
},
|
| 805 |
+
{
|
| 806 |
+
"epoch": 0.41,
|
| 807 |
+
"learning_rate": 3.705602139995416e-07,
|
| 808 |
+
"logits/chosen": -0.7258490920066833,
|
| 809 |
+
"logits/rejected": -0.4828409254550934,
|
| 810 |
+
"logps/chosen": -388.1371154785156,
|
| 811 |
+
"logps/rejected": -422.11181640625,
|
| 812 |
+
"loss": 0.574,
|
| 813 |
+
"rewards/accuracies": 0.637499988079071,
|
| 814 |
+
"rewards/chosen": -2.290266513824463,
|
| 815 |
+
"rewards/margins": 0.4104091227054596,
|
| 816 |
+
"rewards/rejected": -2.7006754875183105,
|
| 817 |
+
"step": 510
|
| 818 |
+
},
|
| 819 |
+
{
|
| 820 |
+
"epoch": 0.41,
|
| 821 |
+
"learning_rate": 3.6442556659016475e-07,
|
| 822 |
+
"logits/chosen": -0.5335447192192078,
|
| 823 |
+
"logits/rejected": -0.33706527948379517,
|
| 824 |
+
"logps/chosen": -378.86492919921875,
|
| 825 |
+
"logps/rejected": -429.67724609375,
|
| 826 |
+
"loss": 0.5608,
|
| 827 |
+
"rewards/accuracies": 0.737500011920929,
|
| 828 |
+
"rewards/chosen": -2.236337423324585,
|
| 829 |
+
"rewards/margins": 0.556148886680603,
|
| 830 |
+
"rewards/rejected": -2.7924864292144775,
|
| 831 |
+
"step": 520
|
| 832 |
+
},
|
| 833 |
+
{
|
| 834 |
+
"epoch": 0.42,
|
| 835 |
+
"learning_rate": 3.582024813755076e-07,
|
| 836 |
+
"logits/chosen": -0.39548322558403015,
|
| 837 |
+
"logits/rejected": -0.10662730038166046,
|
| 838 |
+
"logps/chosen": -368.8847961425781,
|
| 839 |
+
"logps/rejected": -473.3500061035156,
|
| 840 |
+
"loss": 0.5485,
|
| 841 |
+
"rewards/accuracies": 0.731249988079071,
|
| 842 |
+
"rewards/chosen": -2.3263449668884277,
|
| 843 |
+
"rewards/margins": 0.8236624598503113,
|
| 844 |
+
"rewards/rejected": -3.150007724761963,
|
| 845 |
+
"step": 530
|
| 846 |
+
},
|
| 847 |
+
{
|
| 848 |
+
"epoch": 0.43,
|
| 849 |
+
"learning_rate": 3.5189576808485404e-07,
|
| 850 |
+
"logits/chosen": 0.15742243826389313,
|
| 851 |
+
"logits/rejected": 0.31491726636886597,
|
| 852 |
+
"logps/chosen": -394.34930419921875,
|
| 853 |
+
"logps/rejected": -492.82232666015625,
|
| 854 |
+
"loss": 0.5478,
|
| 855 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 856 |
+
"rewards/chosen": -2.6109700202941895,
|
| 857 |
+
"rewards/margins": 0.8250136375427246,
|
| 858 |
+
"rewards/rejected": -3.435983657836914,
|
| 859 |
+
"step": 540
|
| 860 |
+
},
|
| 861 |
+
{
|
| 862 |
+
"epoch": 0.44,
|
| 863 |
+
"learning_rate": 3.4551030108237433e-07,
|
| 864 |
+
"logits/chosen": -0.2550584375858307,
|
| 865 |
+
"logits/rejected": -0.06936412304639816,
|
| 866 |
+
"logps/chosen": -406.5508728027344,
|
| 867 |
+
"logps/rejected": -448.47576904296875,
|
| 868 |
+
"loss": 0.5562,
|
| 869 |
+
"rewards/accuracies": 0.6312500238418579,
|
| 870 |
+
"rewards/chosen": -2.5152666568756104,
|
| 871 |
+
"rewards/margins": 0.4819938540458679,
|
| 872 |
+
"rewards/rejected": -2.997260332107544,
|
| 873 |
+
"step": 550
|
| 874 |
+
},
|
| 875 |
+
{
|
| 876 |
+
"epoch": 0.45,
|
| 877 |
+
"learning_rate": 3.390510155998023e-07,
|
| 878 |
+
"logits/chosen": -0.5292027592658997,
|
| 879 |
+
"logits/rejected": -0.2619571387767792,
|
| 880 |
+
"logps/chosen": -371.6798095703125,
|
| 881 |
+
"logps/rejected": -420.7915954589844,
|
| 882 |
+
"loss": 0.5492,
|
| 883 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 884 |
+
"rewards/chosen": -2.1147050857543945,
|
| 885 |
+
"rewards/margins": 0.6524336338043213,
|
| 886 |
+
"rewards/rejected": -2.7671384811401367,
|
| 887 |
+
"step": 560
|
| 888 |
+
},
|
| 889 |
+
{
|
| 890 |
+
"epoch": 0.45,
|
| 891 |
+
"learning_rate": 3.325229039220684e-07,
|
| 892 |
+
"logits/chosen": -0.5881962776184082,
|
| 893 |
+
"logits/rejected": -0.4658876061439514,
|
| 894 |
+
"logps/chosen": -343.7039794921875,
|
| 895 |
+
"logps/rejected": -406.14178466796875,
|
| 896 |
+
"loss": 0.57,
|
| 897 |
+
"rewards/accuracies": 0.6875,
|
| 898 |
+
"rewards/chosen": -2.0860273838043213,
|
| 899 |
+
"rewards/margins": 0.498068630695343,
|
| 900 |
+
"rewards/rejected": -2.5840957164764404,
|
| 901 |
+
"step": 570
|
| 902 |
+
},
|
| 903 |
+
{
|
| 904 |
+
"epoch": 0.46,
|
| 905 |
+
"learning_rate": 3.2593101152883795e-07,
|
| 906 |
+
"logits/chosen": -0.6565806269645691,
|
| 907 |
+
"logits/rejected": -0.2549567222595215,
|
| 908 |
+
"logps/chosen": -374.8047180175781,
|
| 909 |
+
"logps/rejected": -430.33221435546875,
|
| 910 |
+
"loss": 0.5512,
|
| 911 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 912 |
+
"rewards/chosen": -2.2211391925811768,
|
| 913 |
+
"rewards/margins": 0.6813799142837524,
|
| 914 |
+
"rewards/rejected": -2.9025187492370605,
|
| 915 |
+
"step": 580
|
| 916 |
+
},
|
| 917 |
+
{
|
| 918 |
+
"epoch": 0.47,
|
| 919 |
+
"learning_rate": 3.192804331949349e-07,
|
| 920 |
+
"logits/chosen": -0.07184700667858124,
|
| 921 |
+
"logits/rejected": 0.1699156016111374,
|
| 922 |
+
"logps/chosen": -422.27081298828125,
|
| 923 |
+
"logps/rejected": -490.69134521484375,
|
| 924 |
+
"loss": 0.535,
|
| 925 |
+
"rewards/accuracies": 0.731249988079071,
|
| 926 |
+
"rewards/chosen": -2.730973720550537,
|
| 927 |
+
"rewards/margins": 0.7726518511772156,
|
| 928 |
+
"rewards/rejected": -3.5036251544952393,
|
| 929 |
+
"step": 590
|
| 930 |
+
},
|
| 931 |
+
{
|
| 932 |
+
"epoch": 0.48,
|
| 933 |
+
"learning_rate": 3.125763090526674e-07,
|
| 934 |
+
"logits/chosen": -0.029465889558196068,
|
| 935 |
+
"logits/rejected": 0.15842057764530182,
|
| 936 |
+
"logps/chosen": -417.373046875,
|
| 937 |
+
"logps/rejected": -478.73291015625,
|
| 938 |
+
"loss": 0.5513,
|
| 939 |
+
"rewards/accuracies": 0.675000011920929,
|
| 940 |
+
"rewards/chosen": -2.8006317615509033,
|
| 941 |
+
"rewards/margins": 0.6451797485351562,
|
| 942 |
+
"rewards/rejected": -3.4458115100860596,
|
| 943 |
+
"step": 600
|
| 944 |
+
},
|
| 945 |
+
{
|
| 946 |
+
"epoch": 0.48,
|
| 947 |
+
"eval_logits/chosen": -0.10542195290327072,
|
| 948 |
+
"eval_logits/rejected": 0.12242482602596283,
|
| 949 |
+
"eval_logps/chosen": -436.9386291503906,
|
| 950 |
+
"eval_logps/rejected": -505.02227783203125,
|
| 951 |
+
"eval_loss": 0.5582411885261536,
|
| 952 |
+
"eval_rewards/accuracies": 0.7108209133148193,
|
| 953 |
+
"eval_rewards/chosen": -2.9236514568328857,
|
| 954 |
+
"eval_rewards/margins": 0.6152271032333374,
|
| 955 |
+
"eval_rewards/rejected": -3.5388784408569336,
|
| 956 |
+
"eval_runtime": 184.6013,
|
| 957 |
+
"eval_samples_per_second": 46.327,
|
| 958 |
+
"eval_steps_per_second": 0.726,
|
| 959 |
+
"step": 600
|
| 960 |
+
},
|
| 961 |
+
{
|
| 962 |
+
"epoch": 0.49,
|
| 963 |
+
"learning_rate": 3.0582382061909623e-07,
|
| 964 |
+
"logits/chosen": -0.2445104569196701,
|
| 965 |
+
"logits/rejected": -0.018268002197146416,
|
| 966 |
+
"logps/chosen": -441.7857971191406,
|
| 967 |
+
"logps/rejected": -502.60791015625,
|
| 968 |
+
"loss": 0.5594,
|
| 969 |
+
"rewards/accuracies": 0.6875,
|
| 970 |
+
"rewards/chosen": -2.858261823654175,
|
| 971 |
+
"rewards/margins": 0.5510683655738831,
|
| 972 |
+
"rewards/rejected": -3.409330368041992,
|
| 973 |
+
"step": 610
|
| 974 |
+
},
|
| 975 |
+
{
|
| 976 |
+
"epoch": 0.49,
|
| 977 |
+
"learning_rate": 2.9902818679131775e-07,
|
| 978 |
+
"logits/chosen": -0.4190225601196289,
|
| 979 |
+
"logits/rejected": -0.22823679447174072,
|
| 980 |
+
"logps/chosen": -399.03924560546875,
|
| 981 |
+
"logps/rejected": -498.6724548339844,
|
| 982 |
+
"loss": 0.5499,
|
| 983 |
+
"rewards/accuracies": 0.71875,
|
| 984 |
+
"rewards/chosen": -2.650449275970459,
|
| 985 |
+
"rewards/margins": 0.7673205137252808,
|
| 986 |
+
"rewards/rejected": -3.4177703857421875,
|
| 987 |
+
"step": 620
|
| 988 |
+
},
|
| 989 |
+
{
|
| 990 |
+
"epoch": 0.5,
|
| 991 |
+
"learning_rate": 2.921946598128571e-07,
|
| 992 |
+
"logits/chosen": -0.43653860688209534,
|
| 993 |
+
"logits/rejected": -0.20837187767028809,
|
| 994 |
+
"logps/chosen": -402.82781982421875,
|
| 995 |
+
"logps/rejected": -485.4117736816406,
|
| 996 |
+
"loss": 0.5739,
|
| 997 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 998 |
+
"rewards/chosen": -2.573000192642212,
|
| 999 |
+
"rewards/margins": 0.7478531002998352,
|
| 1000 |
+
"rewards/rejected": -3.3208529949188232,
|
| 1001 |
+
"step": 630
|
| 1002 |
+
},
|
| 1003 |
+
{
|
| 1004 |
+
"epoch": 0.51,
|
| 1005 |
+
"learning_rate": 2.8532852121428733e-07,
|
| 1006 |
+
"logits/chosen": -0.43430274724960327,
|
| 1007 |
+
"logits/rejected": -0.13240045309066772,
|
| 1008 |
+
"logps/chosen": -397.2491149902344,
|
| 1009 |
+
"logps/rejected": -442.12384033203125,
|
| 1010 |
+
"loss": 0.5462,
|
| 1011 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 1012 |
+
"rewards/chosen": -2.4125733375549316,
|
| 1013 |
+
"rewards/margins": 0.5821372866630554,
|
| 1014 |
+
"rewards/rejected": -2.9947104454040527,
|
| 1015 |
+
"step": 640
|
| 1016 |
+
},
|
| 1017 |
+
{
|
| 1018 |
+
"epoch": 0.52,
|
| 1019 |
+
"learning_rate": 2.7843507773121414e-07,
|
| 1020 |
+
"logits/chosen": -0.4247920513153076,
|
| 1021 |
+
"logits/rejected": -0.21372787654399872,
|
| 1022 |
+
"logps/chosen": -389.4237976074219,
|
| 1023 |
+
"logps/rejected": -458.3169860839844,
|
| 1024 |
+
"loss": 0.5373,
|
| 1025 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 1026 |
+
"rewards/chosen": -2.444688320159912,
|
| 1027 |
+
"rewards/margins": 0.7236617207527161,
|
| 1028 |
+
"rewards/rejected": -3.1683506965637207,
|
| 1029 |
+
"step": 650
|
| 1030 |
+
},
|
| 1031 |
+
{
|
| 1032 |
+
"epoch": 0.53,
|
| 1033 |
+
"learning_rate": 2.715196572027789e-07,
|
| 1034 |
+
"logits/chosen": -0.6697942614555359,
|
| 1035 |
+
"logits/rejected": -0.4933086931705475,
|
| 1036 |
+
"logps/chosen": -387.529296875,
|
| 1037 |
+
"logps/rejected": -472.73944091796875,
|
| 1038 |
+
"loss": 0.5685,
|
| 1039 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 1040 |
+
"rewards/chosen": -2.3496451377868652,
|
| 1041 |
+
"rewards/margins": 0.7728831171989441,
|
| 1042 |
+
"rewards/rejected": -3.122528314590454,
|
| 1043 |
+
"step": 660
|
| 1044 |
+
},
|
| 1045 |
+
{
|
| 1046 |
+
"epoch": 0.53,
|
| 1047 |
+
"learning_rate": 2.645876044538521e-07,
|
| 1048 |
+
"logits/chosen": -1.0338900089263916,
|
| 1049 |
+
"logits/rejected": -0.8813627362251282,
|
| 1050 |
+
"logps/chosen": -372.53118896484375,
|
| 1051 |
+
"logps/rejected": -426.54241943359375,
|
| 1052 |
+
"loss": 0.5725,
|
| 1053 |
+
"rewards/accuracies": 0.675000011920929,
|
| 1054 |
+
"rewards/chosen": -2.201908588409424,
|
| 1055 |
+
"rewards/margins": 0.5010865926742554,
|
| 1056 |
+
"rewards/rejected": -2.7029950618743896,
|
| 1057 |
+
"step": 670
|
| 1058 |
+
},
|
| 1059 |
+
{
|
| 1060 |
+
"epoch": 0.54,
|
| 1061 |
+
"learning_rate": 2.5764427716409815e-07,
|
| 1062 |
+
"logits/chosen": -0.9278701543807983,
|
| 1063 |
+
"logits/rejected": -0.7282145023345947,
|
| 1064 |
+
"logps/chosen": -347.2828674316406,
|
| 1065 |
+
"logps/rejected": -416.9349060058594,
|
| 1066 |
+
"loss": 0.5479,
|
| 1067 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 1068 |
+
"rewards/chosen": -2.0276436805725098,
|
| 1069 |
+
"rewards/margins": 0.743033230304718,
|
| 1070 |
+
"rewards/rejected": -2.770677089691162,
|
| 1071 |
+
"step": 680
|
| 1072 |
+
},
|
| 1073 |
+
{
|
| 1074 |
+
"epoch": 0.55,
|
| 1075 |
+
"learning_rate": 2.5069504172710494e-07,
|
| 1076 |
+
"logits/chosen": -0.5008482336997986,
|
| 1077 |
+
"logits/rejected": -0.34875133633613586,
|
| 1078 |
+
"logps/chosen": -373.7621154785156,
|
| 1079 |
+
"logps/rejected": -485.12884521484375,
|
| 1080 |
+
"loss": 0.5217,
|
| 1081 |
+
"rewards/accuracies": 0.762499988079071,
|
| 1082 |
+
"rewards/chosen": -2.347053050994873,
|
| 1083 |
+
"rewards/margins": 0.9024646878242493,
|
| 1084 |
+
"rewards/rejected": -3.2495174407958984,
|
| 1085 |
+
"step": 690
|
| 1086 |
+
},
|
| 1087 |
+
{
|
| 1088 |
+
"epoch": 0.56,
|
| 1089 |
+
"learning_rate": 2.4374526910277886e-07,
|
| 1090 |
+
"logits/chosen": 0.06850005686283112,
|
| 1091 |
+
"logits/rejected": 0.41385045647621155,
|
| 1092 |
+
"logps/chosen": -411.46246337890625,
|
| 1093 |
+
"logps/rejected": -476.6162109375,
|
| 1094 |
+
"loss": 0.5571,
|
| 1095 |
+
"rewards/accuracies": 0.706250011920929,
|
| 1096 |
+
"rewards/chosen": -2.6757898330688477,
|
| 1097 |
+
"rewards/margins": 0.8085702657699585,
|
| 1098 |
+
"rewards/rejected": -3.4843602180480957,
|
| 1099 |
+
"step": 700
|
| 1100 |
+
},
|
| 1101 |
+
{
|
| 1102 |
+
"epoch": 0.56,
|
| 1103 |
+
"eval_logits/chosen": 0.035554468631744385,
|
| 1104 |
+
"eval_logits/rejected": 0.2980235815048218,
|
| 1105 |
+
"eval_logps/chosen": -424.2823486328125,
|
| 1106 |
+
"eval_logps/rejected": -505.6960754394531,
|
| 1107 |
+
"eval_loss": 0.5558871626853943,
|
| 1108 |
+
"eval_rewards/accuracies": 0.704291045665741,
|
| 1109 |
+
"eval_rewards/chosen": -2.797088146209717,
|
| 1110 |
+
"eval_rewards/margins": 0.748529314994812,
|
| 1111 |
+
"eval_rewards/rejected": -3.5456173419952393,
|
| 1112 |
+
"eval_runtime": 184.3728,
|
| 1113 |
+
"eval_samples_per_second": 46.384,
|
| 1114 |
+
"eval_steps_per_second": 0.727,
|
| 1115 |
+
"step": 700
|
| 1116 |
+
},
|
| 1117 |
+
{
|
| 1118 |
+
"epoch": 0.57,
|
| 1119 |
+
"learning_rate": 2.368003306662104e-07,
|
| 1120 |
+
"logits/chosen": 0.07857178151607513,
|
| 1121 |
+
"logits/rejected": 0.3302653729915619,
|
| 1122 |
+
"logps/chosen": -413.8836975097656,
|
| 1123 |
+
"logps/rejected": -535.0875244140625,
|
| 1124 |
+
"loss": 0.5287,
|
| 1125 |
+
"rewards/accuracies": 0.762499988079071,
|
| 1126 |
+
"rewards/chosen": -2.7193782329559326,
|
| 1127 |
+
"rewards/margins": 1.0089346170425415,
|
| 1128 |
+
"rewards/rejected": -3.7283127307891846,
|
| 1129 |
+
"step": 710
|
| 1130 |
+
},
|
| 1131 |
+
{
|
| 1132 |
+
"epoch": 0.57,
|
| 1133 |
+
"learning_rate": 2.2986559405621886e-07,
|
| 1134 |
+
"logits/chosen": 0.2789291739463806,
|
| 1135 |
+
"logits/rejected": 0.4242584705352783,
|
| 1136 |
+
"logps/chosen": -422.7801818847656,
|
| 1137 |
+
"logps/rejected": -522.7840576171875,
|
| 1138 |
+
"loss": 0.5551,
|
| 1139 |
+
"rewards/accuracies": 0.731249988079071,
|
| 1140 |
+
"rewards/chosen": -2.925621509552002,
|
| 1141 |
+
"rewards/margins": 0.8043605089187622,
|
| 1142 |
+
"rewards/rejected": -3.729982376098633,
|
| 1143 |
+
"step": 720
|
| 1144 |
+
},
|
| 1145 |
+
{
|
| 1146 |
+
"epoch": 0.58,
|
| 1147 |
+
"learning_rate": 2.2294641902678443e-07,
|
| 1148 |
+
"logits/chosen": -0.19327735900878906,
|
| 1149 |
+
"logits/rejected": 0.043265581130981445,
|
| 1150 |
+
"logps/chosen": -363.1488342285156,
|
| 1151 |
+
"logps/rejected": -470.94970703125,
|
| 1152 |
+
"loss": 0.5284,
|
| 1153 |
+
"rewards/accuracies": 0.78125,
|
| 1154 |
+
"rewards/chosen": -2.39530873298645,
|
| 1155 |
+
"rewards/margins": 0.908363938331604,
|
| 1156 |
+
"rewards/rejected": -3.3036727905273438,
|
| 1157 |
+
"step": 730
|
| 1158 |
+
},
|
| 1159 |
+
{
|
| 1160 |
+
"epoch": 0.59,
|
| 1161 |
+
"learning_rate": 2.160481533045751e-07,
|
| 1162 |
+
"logits/chosen": -0.37412697076797485,
|
| 1163 |
+
"logits/rejected": -0.17320053279399872,
|
| 1164 |
+
"logps/chosen": -390.2896423339844,
|
| 1165 |
+
"logps/rejected": -428.08099365234375,
|
| 1166 |
+
"loss": 0.5572,
|
| 1167 |
+
"rewards/accuracies": 0.675000011920929,
|
| 1168 |
+
"rewards/chosen": -2.3005330562591553,
|
| 1169 |
+
"rewards/margins": 0.48462891578674316,
|
| 1170 |
+
"rewards/rejected": -2.7851624488830566,
|
| 1171 |
+
"step": 740
|
| 1172 |
+
},
|
| 1173 |
+
{
|
| 1174 |
+
"epoch": 0.6,
|
| 1175 |
+
"learning_rate": 2.0917612845576882e-07,
|
| 1176 |
+
"logits/chosen": -0.26352375745773315,
|
| 1177 |
+
"logits/rejected": -0.0010178961092606187,
|
| 1178 |
+
"logps/chosen": -373.3875427246094,
|
| 1179 |
+
"logps/rejected": -440.09442138671875,
|
| 1180 |
+
"loss": 0.5534,
|
| 1181 |
+
"rewards/accuracies": 0.6875,
|
| 1182 |
+
"rewards/chosen": -2.3315823078155518,
|
| 1183 |
+
"rewards/margins": 0.6843063235282898,
|
| 1184 |
+
"rewards/rejected": -3.0158886909484863,
|
| 1185 |
+
"step": 750
|
| 1186 |
+
},
|
| 1187 |
+
{
|
| 1188 |
+
"epoch": 0.6,
|
| 1189 |
+
"learning_rate": 2.0233565576536564e-07,
|
| 1190 |
+
"logits/chosen": -0.3354080021381378,
|
| 1191 |
+
"logits/rejected": -0.006600166670978069,
|
| 1192 |
+
"logps/chosen": -360.56463623046875,
|
| 1193 |
+
"logps/rejected": -440.66961669921875,
|
| 1194 |
+
"loss": 0.5328,
|
| 1195 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 1196 |
+
"rewards/chosen": -2.1626803874969482,
|
| 1197 |
+
"rewards/margins": 0.8473829030990601,
|
| 1198 |
+
"rewards/rejected": -3.010063409805298,
|
| 1199 |
+
"step": 760
|
| 1200 |
+
},
|
| 1201 |
+
{
|
| 1202 |
+
"epoch": 0.61,
|
| 1203 |
+
"learning_rate": 1.9553202213217537e-07,
|
| 1204 |
+
"logits/chosen": -0.021420275792479515,
|
| 1205 |
+
"logits/rejected": 0.19946305453777313,
|
| 1206 |
+
"logps/chosen": -389.1043395996094,
|
| 1207 |
+
"logps/rejected": -448.04998779296875,
|
| 1208 |
+
"loss": 0.5523,
|
| 1209 |
+
"rewards/accuracies": 0.731249988079071,
|
| 1210 |
+
"rewards/chosen": -2.391838788986206,
|
| 1211 |
+
"rewards/margins": 0.6678962707519531,
|
| 1212 |
+
"rewards/rejected": -3.059735059738159,
|
| 1213 |
+
"step": 770
|
| 1214 |
+
},
|
| 1215 |
+
{
|
| 1216 |
+
"epoch": 0.62,
|
| 1217 |
+
"learning_rate": 1.887704859826528e-07,
|
| 1218 |
+
"logits/chosen": -0.15253478288650513,
|
| 1219 |
+
"logits/rejected": -0.00011998042464256287,
|
| 1220 |
+
"logps/chosen": -394.9501953125,
|
| 1221 |
+
"logps/rejected": -462.32843017578125,
|
| 1222 |
+
"loss": 0.5443,
|
| 1223 |
+
"rewards/accuracies": 0.699999988079071,
|
| 1224 |
+
"rewards/chosen": -2.476644992828369,
|
| 1225 |
+
"rewards/margins": 0.566824734210968,
|
| 1226 |
+
"rewards/rejected": -3.0434699058532715,
|
| 1227 |
+
"step": 780
|
| 1228 |
+
},
|
| 1229 |
+
{
|
| 1230 |
+
"epoch": 0.63,
|
| 1231 |
+
"learning_rate": 1.8205627320673836e-07,
|
| 1232 |
+
"logits/chosen": -0.17955633997917175,
|
| 1233 |
+
"logits/rejected": 0.18167546391487122,
|
| 1234 |
+
"logps/chosen": -390.32244873046875,
|
| 1235 |
+
"logps/rejected": -444.895263671875,
|
| 1236 |
+
"loss": 0.5566,
|
| 1237 |
+
"rewards/accuracies": 0.699999988079071,
|
| 1238 |
+
"rewards/chosen": -2.4376220703125,
|
| 1239 |
+
"rewards/margins": 0.7008293271064758,
|
| 1240 |
+
"rewards/rejected": -3.138451099395752,
|
| 1241 |
+
"step": 790
|
| 1242 |
+
},
|
| 1243 |
+
{
|
| 1244 |
+
"epoch": 0.64,
|
| 1245 |
+
"learning_rate": 1.7539457311884675e-07,
|
| 1246 |
+
"logits/chosen": -0.09838727861642838,
|
| 1247 |
+
"logits/rejected": 0.11829495429992676,
|
| 1248 |
+
"logps/chosen": -402.4017333984375,
|
| 1249 |
+
"logps/rejected": -451.49346923828125,
|
| 1250 |
+
"loss": 0.5609,
|
| 1251 |
+
"rewards/accuracies": 0.699999988079071,
|
| 1252 |
+
"rewards/chosen": -2.4441986083984375,
|
| 1253 |
+
"rewards/margins": 0.5067789554595947,
|
| 1254 |
+
"rewards/rejected": -2.9509775638580322,
|
| 1255 |
+
"step": 800
|
| 1256 |
+
},
|
| 1257 |
+
{
|
| 1258 |
+
"epoch": 0.64,
|
| 1259 |
+
"eval_logits/chosen": -0.03116540051996708,
|
| 1260 |
+
"eval_logits/rejected": 0.1922437697649002,
|
| 1261 |
+
"eval_logps/chosen": -387.7091979980469,
|
| 1262 |
+
"eval_logps/rejected": -459.44390869140625,
|
| 1263 |
+
"eval_loss": 0.5468714833259583,
|
| 1264 |
+
"eval_rewards/accuracies": 0.7108209133148193,
|
| 1265 |
+
"eval_rewards/chosen": -2.431356430053711,
|
| 1266 |
+
"eval_rewards/margins": 0.6517390012741089,
|
| 1267 |
+
"eval_rewards/rejected": -3.0830955505371094,
|
| 1268 |
+
"eval_runtime": 184.5893,
|
| 1269 |
+
"eval_samples_per_second": 46.33,
|
| 1270 |
+
"eval_steps_per_second": 0.726,
|
| 1271 |
+
"step": 800
|
| 1272 |
+
},
|
| 1273 |
+
{
|
| 1274 |
+
"epoch": 0.64,
|
| 1275 |
+
"learning_rate": 1.687905344471226e-07,
|
| 1276 |
+
"logits/chosen": 0.07735608518123627,
|
| 1277 |
+
"logits/rejected": 0.3973601460456848,
|
| 1278 |
+
"logps/chosen": -408.05999755859375,
|
| 1279 |
+
"logps/rejected": -459.011474609375,
|
| 1280 |
+
"loss": 0.5384,
|
| 1281 |
+
"rewards/accuracies": 0.706250011920929,
|
| 1282 |
+
"rewards/chosen": -2.5008435249328613,
|
| 1283 |
+
"rewards/margins": 0.6535352468490601,
|
| 1284 |
+
"rewards/rejected": -3.154379367828369,
|
| 1285 |
+
"step": 810
|
| 1286 |
+
},
|
| 1287 |
+
{
|
| 1288 |
+
"epoch": 0.65,
|
| 1289 |
+
"learning_rate": 1.6224926135406693e-07,
|
| 1290 |
+
"logits/chosen": 0.1125444769859314,
|
| 1291 |
+
"logits/rejected": 0.3865428566932678,
|
| 1292 |
+
"logps/chosen": -404.16058349609375,
|
| 1293 |
+
"logps/rejected": -484.68621826171875,
|
| 1294 |
+
"loss": 0.5448,
|
| 1295 |
+
"rewards/accuracies": 0.706250011920929,
|
| 1296 |
+
"rewards/chosen": -2.4405789375305176,
|
| 1297 |
+
"rewards/margins": 0.718208909034729,
|
| 1298 |
+
"rewards/rejected": -3.158787727355957,
|
| 1299 |
+
"step": 820
|
| 1300 |
+
},
|
| 1301 |
+
{
|
| 1302 |
+
"epoch": 0.66,
|
| 1303 |
+
"learning_rate": 1.557758094916053e-07,
|
| 1304 |
+
"logits/chosen": 0.11989516019821167,
|
| 1305 |
+
"logits/rejected": 0.30926594138145447,
|
| 1306 |
+
"logps/chosen": -370.29876708984375,
|
| 1307 |
+
"logps/rejected": -452.27911376953125,
|
| 1308 |
+
"loss": 0.5418,
|
| 1309 |
+
"rewards/accuracies": 0.737500011920929,
|
| 1310 |
+
"rewards/chosen": -2.3860089778900146,
|
| 1311 |
+
"rewards/margins": 0.7260924577713013,
|
| 1312 |
+
"rewards/rejected": -3.1121015548706055,
|
| 1313 |
+
"step": 830
|
| 1314 |
+
},
|
| 1315 |
+
{
|
| 1316 |
+
"epoch": 0.67,
|
| 1317 |
+
"learning_rate": 1.4937518209365108e-07,
|
| 1318 |
+
"logits/chosen": -0.14239154756069183,
|
| 1319 |
+
"logits/rejected": 0.14250756800174713,
|
| 1320 |
+
"logps/chosen": -395.55755615234375,
|
| 1321 |
+
"logps/rejected": -447.6368713378906,
|
| 1322 |
+
"loss": 0.5573,
|
| 1323 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 1324 |
+
"rewards/chosen": -2.378154754638672,
|
| 1325 |
+
"rewards/margins": 0.6160937547683716,
|
| 1326 |
+
"rewards/rejected": -2.994248390197754,
|
| 1327 |
+
"step": 840
|
| 1328 |
+
},
|
| 1329 |
+
{
|
| 1330 |
+
"epoch": 0.68,
|
| 1331 |
+
"learning_rate": 1.4305232610918045e-07,
|
| 1332 |
+
"logits/chosen": -0.16526366770267487,
|
| 1333 |
+
"logits/rejected": 0.16432161629199982,
|
| 1334 |
+
"logps/chosen": -373.45330810546875,
|
| 1335 |
+
"logps/rejected": -436.6773376464844,
|
| 1336 |
+
"loss": 0.5415,
|
| 1337 |
+
"rewards/accuracies": 0.737500011920929,
|
| 1338 |
+
"rewards/chosen": -2.3134028911590576,
|
| 1339 |
+
"rewards/margins": 0.774810791015625,
|
| 1340 |
+
"rewards/rejected": -3.0882136821746826,
|
| 1341 |
+
"step": 850
|
| 1342 |
+
},
|
| 1343 |
+
{
|
| 1344 |
+
"epoch": 0.68,
|
| 1345 |
+
"learning_rate": 1.3681212837880977e-07,
|
| 1346 |
+
"logits/chosen": -0.1321481615304947,
|
| 1347 |
+
"logits/rejected": 0.23287932574748993,
|
| 1348 |
+
"logps/chosen": -364.96990966796875,
|
| 1349 |
+
"logps/rejected": -447.7923278808594,
|
| 1350 |
+
"loss": 0.5396,
|
| 1351 |
+
"rewards/accuracies": 0.7875000238418579,
|
| 1352 |
+
"rewards/chosen": -2.176964282989502,
|
| 1353 |
+
"rewards/margins": 0.8955341577529907,
|
| 1354 |
+
"rewards/rejected": -3.0724985599517822,
|
| 1355 |
+
"step": 860
|
| 1356 |
+
},
|
| 1357 |
+
{
|
| 1358 |
+
"epoch": 0.69,
|
| 1359 |
+
"learning_rate": 1.3065941185782977e-07,
|
| 1360 |
+
"logits/chosen": 0.05437428876757622,
|
| 1361 |
+
"logits/rejected": 0.2819867432117462,
|
| 1362 |
+
"logps/chosen": -383.08599853515625,
|
| 1363 |
+
"logps/rejected": -439.3629455566406,
|
| 1364 |
+
"loss": 0.5505,
|
| 1365 |
+
"rewards/accuracies": 0.637499988079071,
|
| 1366 |
+
"rewards/chosen": -2.405247211456299,
|
| 1367 |
+
"rewards/margins": 0.5403125882148743,
|
| 1368 |
+
"rewards/rejected": -2.9455599784851074,
|
| 1369 |
+
"step": 870
|
| 1370 |
+
},
|
| 1371 |
+
{
|
| 1372 |
+
"epoch": 0.7,
|
| 1373 |
+
"learning_rate": 1.2459893188861613e-07,
|
| 1374 |
+
"logits/chosen": -0.12052659690380096,
|
| 1375 |
+
"logits/rejected": 0.12284734100103378,
|
| 1376 |
+
"logps/chosen": -367.1181640625,
|
| 1377 |
+
"logps/rejected": -468.1044921875,
|
| 1378 |
+
"loss": 0.5185,
|
| 1379 |
+
"rewards/accuracies": 0.75,
|
| 1380 |
+
"rewards/chosen": -2.39152455329895,
|
| 1381 |
+
"rewards/margins": 0.9137696027755737,
|
| 1382 |
+
"rewards/rejected": -3.3052947521209717,
|
| 1383 |
+
"step": 880
|
| 1384 |
+
},
|
| 1385 |
+
{
|
| 1386 |
+
"epoch": 0.71,
|
| 1387 |
+
"learning_rate": 1.1863537252529548e-07,
|
| 1388 |
+
"logits/chosen": 0.14598000049591064,
|
| 1389 |
+
"logits/rejected": 0.38815659284591675,
|
| 1390 |
+
"logps/chosen": -397.891357421875,
|
| 1391 |
+
"logps/rejected": -472.38677978515625,
|
| 1392 |
+
"loss": 0.5323,
|
| 1393 |
+
"rewards/accuracies": 0.731249988079071,
|
| 1394 |
+
"rewards/chosen": -2.512676239013672,
|
| 1395 |
+
"rewards/margins": 0.7713057994842529,
|
| 1396 |
+
"rewards/rejected": -3.2839818000793457,
|
| 1397 |
+
"step": 890
|
| 1398 |
+
},
|
| 1399 |
+
{
|
| 1400 |
+
"epoch": 0.72,
|
| 1401 |
+
"learning_rate": 1.1277334291351145e-07,
|
| 1402 |
+
"logits/chosen": 0.15319526195526123,
|
| 1403 |
+
"logits/rejected": 0.35974830389022827,
|
| 1404 |
+
"logps/chosen": -380.77783203125,
|
| 1405 |
+
"logps/rejected": -449.54315185546875,
|
| 1406 |
+
"loss": 0.5514,
|
| 1407 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 1408 |
+
"rewards/chosen": -2.3706305027008057,
|
| 1409 |
+
"rewards/margins": 0.6724039912223816,
|
| 1410 |
+
"rewards/rejected": -3.043034076690674,
|
| 1411 |
+
"step": 900
|
| 1412 |
+
},
|
| 1413 |
+
{
|
| 1414 |
+
"epoch": 0.72,
|
| 1415 |
+
"eval_logits/chosen": 0.28598034381866455,
|
| 1416 |
+
"eval_logits/rejected": 0.5382024645805359,
|
| 1417 |
+
"eval_logps/chosen": -392.3096008300781,
|
| 1418 |
+
"eval_logps/rejected": -471.95330810546875,
|
| 1419 |
+
"eval_loss": 0.5473664402961731,
|
| 1420 |
+
"eval_rewards/accuracies": 0.6996268630027771,
|
| 1421 |
+
"eval_rewards/chosen": -2.4773612022399902,
|
| 1422 |
+
"eval_rewards/margins": 0.7308279275894165,
|
| 1423 |
+
"eval_rewards/rejected": -3.2081892490386963,
|
| 1424 |
+
"eval_runtime": 184.5275,
|
| 1425 |
+
"eval_samples_per_second": 46.345,
|
| 1426 |
+
"eval_steps_per_second": 0.726,
|
| 1427 |
+
"step": 900
|
| 1428 |
+
},
|
| 1429 |
+
{
|
| 1430 |
+
"epoch": 0.72,
|
| 1431 |
+
"learning_rate": 1.0701737372808431e-07,
|
| 1432 |
+
"logits/chosen": 0.15951867401599884,
|
| 1433 |
+
"logits/rejected": 0.46630391478538513,
|
| 1434 |
+
"logps/chosen": -383.52850341796875,
|
| 1435 |
+
"logps/rejected": -467.2303771972656,
|
| 1436 |
+
"loss": 0.5362,
|
| 1437 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 1438 |
+
"rewards/chosen": -2.2782187461853027,
|
| 1439 |
+
"rewards/margins": 0.8473943471908569,
|
| 1440 |
+
"rewards/rejected": -3.125612735748291,
|
| 1441 |
+
"step": 910
|
| 1442 |
+
},
|
| 1443 |
{
|
| 1444 |
"epoch": 0.73,
|
| 1445 |
+
"learning_rate": 1.0137191367132078e-07,
|
| 1446 |
+
"logits/chosen": 0.2791319191455841,
|
| 1447 |
+
"logits/rejected": 0.45174160599708557,
|
| 1448 |
+
"logps/chosen": -372.1945495605469,
|
| 1449 |
+
"logps/rejected": -446.6507263183594,
|
| 1450 |
+
"loss": 0.5458,
|
| 1451 |
"rewards/accuracies": 0.706250011920929,
|
| 1452 |
+
"rewards/chosen": -2.3160648345947266,
|
| 1453 |
+
"rewards/margins": 0.68004310131073,
|
| 1454 |
+
"rewards/rejected": -2.996107816696167,
|
| 1455 |
+
"step": 920
|
| 1456 |
+
},
|
| 1457 |
+
{
|
| 1458 |
+
"epoch": 0.74,
|
| 1459 |
+
"learning_rate": 9.584132603467827e-08,
|
| 1460 |
+
"logits/chosen": -0.12192128598690033,
|
| 1461 |
+
"logits/rejected": 0.1477951854467392,
|
| 1462 |
+
"logps/chosen": -366.48321533203125,
|
| 1463 |
+
"logps/rejected": -453.130126953125,
|
| 1464 |
+
"loss": 0.5467,
|
| 1465 |
+
"rewards/accuracies": 0.75,
|
| 1466 |
+
"rewards/chosen": -2.200005531311035,
|
| 1467 |
+
"rewards/margins": 0.7978888750076294,
|
| 1468 |
+
"rewards/rejected": -2.997894287109375,
|
| 1469 |
+
"step": 930
|
| 1470 |
},
|
| 1471 |
{
|
| 1472 |
"epoch": 0.75,
|
| 1473 |
+
"learning_rate": 9.042988532644249e-08,
|
| 1474 |
+
"logits/chosen": -0.03106372058391571,
|
| 1475 |
+
"logits/rejected": 0.07721444219350815,
|
| 1476 |
+
"logps/chosen": -344.21270751953125,
|
| 1477 |
+
"logps/rejected": -438.11077880859375,
|
| 1478 |
+
"loss": 0.5161,
|
| 1479 |
+
"rewards/accuracies": 0.71875,
|
| 1480 |
+
"rewards/chosen": -2.149094581604004,
|
| 1481 |
+
"rewards/margins": 0.7353444695472717,
|
| 1482 |
+
"rewards/rejected": -2.884438991546631,
|
| 1483 |
+
"step": 940
|
| 1484 |
+
},
|
| 1485 |
+
{
|
| 1486 |
+
"epoch": 0.76,
|
| 1487 |
+
"learning_rate": 8.514177396802428e-08,
|
| 1488 |
+
"logits/chosen": 0.006801058538258076,
|
| 1489 |
+
"logits/rejected": 0.20282092690467834,
|
| 1490 |
+
"logps/chosen": -358.15167236328125,
|
| 1491 |
+
"logps/rejected": -436.4964294433594,
|
| 1492 |
+
"loss": 0.5385,
|
| 1493 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 1494 |
+
"rewards/chosen": -2.2220425605773926,
|
| 1495 |
+
"rewards/margins": 0.7004804611206055,
|
| 1496 |
+
"rewards/rejected": -2.922523260116577,
|
| 1497 |
+
"step": 950
|
| 1498 |
+
},
|
| 1499 |
+
{
|
| 1500 |
+
"epoch": 0.76,
|
| 1501 |
+
"learning_rate": 7.998107906142839e-08,
|
| 1502 |
+
"logits/chosen": 0.41448846459388733,
|
| 1503 |
+
"logits/rejected": 0.705254852771759,
|
| 1504 |
+
"logps/chosen": -371.27801513671875,
|
| 1505 |
+
"logps/rejected": -434.56866455078125,
|
| 1506 |
+
"loss": 0.5236,
|
| 1507 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 1508 |
+
"rewards/chosen": -2.2166616916656494,
|
| 1509 |
+
"rewards/margins": 0.6714047193527222,
|
| 1510 |
+
"rewards/rejected": -2.888066530227661,
|
| 1511 |
+
"step": 960
|
| 1512 |
},
|
| 1513 |
{
|
| 1514 |
"epoch": 0.77,
|
| 1515 |
+
"learning_rate": 7.495178923039396e-08,
|
| 1516 |
+
"logits/chosen": 0.23847150802612305,
|
| 1517 |
+
"logits/rejected": 0.48661884665489197,
|
| 1518 |
+
"logps/chosen": -366.28179931640625,
|
| 1519 |
+
"logps/rejected": -462.679443359375,
|
| 1520 |
+
"loss": 0.5459,
|
| 1521 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 1522 |
+
"rewards/chosen": -2.1916985511779785,
|
| 1523 |
+
"rewards/margins": 0.8472123146057129,
|
| 1524 |
+
"rewards/rejected": -3.038910388946533,
|
| 1525 |
+
"step": 970
|
| 1526 |
},
|
| 1527 |
{
|
| 1528 |
+
"epoch": 0.78,
|
| 1529 |
+
"learning_rate": 7.005779153764682e-08,
|
| 1530 |
+
"logits/chosen": 0.41438961029052734,
|
| 1531 |
+
"logits/rejected": 0.6912784576416016,
|
| 1532 |
+
"logps/chosen": -382.70123291015625,
|
| 1533 |
+
"logps/rejected": -461.8614807128906,
|
| 1534 |
+
"loss": 0.5453,
|
| 1535 |
"rewards/accuracies": 0.737500011920929,
|
| 1536 |
+
"rewards/chosen": -2.4304287433624268,
|
| 1537 |
+
"rewards/margins": 0.7116767764091492,
|
| 1538 |
+
"rewards/rejected": -3.1421055793762207,
|
| 1539 |
+
"step": 980
|
| 1540 |
+
},
|
| 1541 |
+
{
|
| 1542 |
+
"epoch": 0.79,
|
| 1543 |
+
"learning_rate": 6.530286848064698e-08,
|
| 1544 |
+
"logits/chosen": 0.36573725938796997,
|
| 1545 |
+
"logits/rejected": 0.5834362506866455,
|
| 1546 |
+
"logps/chosen": -384.49749755859375,
|
| 1547 |
+
"logps/rejected": -466.30096435546875,
|
| 1548 |
+
"loss": 0.5528,
|
| 1549 |
+
"rewards/accuracies": 0.71875,
|
| 1550 |
+
"rewards/chosen": -2.5111565589904785,
|
| 1551 |
+
"rewards/margins": 0.7234699130058289,
|
| 1552 |
+
"rewards/rejected": -3.234626054763794,
|
| 1553 |
+
"step": 990
|
| 1554 |
+
},
|
| 1555 |
+
{
|
| 1556 |
+
"epoch": 0.8,
|
| 1557 |
+
"learning_rate": 6.069069506815325e-08,
|
| 1558 |
+
"logits/chosen": 0.45530566573143005,
|
| 1559 |
+
"logits/rejected": 0.5909157991409302,
|
| 1560 |
+
"logps/chosen": -379.1433410644531,
|
| 1561 |
+
"logps/rejected": -468.88458251953125,
|
| 1562 |
+
"loss": 0.527,
|
| 1563 |
+
"rewards/accuracies": 0.762499988079071,
|
| 1564 |
+
"rewards/chosen": -2.5361268520355225,
|
| 1565 |
+
"rewards/margins": 0.7407721281051636,
|
| 1566 |
+
"rewards/rejected": -3.2768986225128174,
|
| 1567 |
+
"step": 1000
|
| 1568 |
+
},
|
| 1569 |
+
{
|
| 1570 |
+
"epoch": 0.8,
|
| 1571 |
+
"eval_logits/chosen": 0.3871051073074341,
|
| 1572 |
+
"eval_logits/rejected": 0.6372014284133911,
|
| 1573 |
+
"eval_logps/chosen": -394.97113037109375,
|
| 1574 |
+
"eval_logps/rejected": -471.8453674316406,
|
| 1575 |
+
"eval_loss": 0.5453863739967346,
|
| 1576 |
+
"eval_rewards/accuracies": 0.70802241563797,
|
| 1577 |
+
"eval_rewards/chosen": -2.503976345062256,
|
| 1578 |
+
"eval_rewards/margins": 0.7031334638595581,
|
| 1579 |
+
"eval_rewards/rejected": -3.2071101665496826,
|
| 1580 |
+
"eval_runtime": 184.4226,
|
| 1581 |
+
"eval_samples_per_second": 46.372,
|
| 1582 |
+
"eval_steps_per_second": 0.727,
|
| 1583 |
+
"step": 1000
|
| 1584 |
+
},
|
| 1585 |
+
{
|
| 1586 |
+
"epoch": 0.8,
|
| 1587 |
+
"learning_rate": 5.6224835979863714e-08,
|
| 1588 |
+
"logits/chosen": 0.31174296140670776,
|
| 1589 |
+
"logits/rejected": 0.6193565130233765,
|
| 1590 |
+
"logps/chosen": -390.387451171875,
|
| 1591 |
+
"logps/rejected": -468.4959411621094,
|
| 1592 |
+
"loss": 0.5568,
|
| 1593 |
+
"rewards/accuracies": 0.71875,
|
| 1594 |
+
"rewards/chosen": -2.32766056060791,
|
| 1595 |
+
"rewards/margins": 0.68747878074646,
|
| 1596 |
+
"rewards/rejected": -3.015139102935791,
|
| 1597 |
+
"step": 1010
|
| 1598 |
+
},
|
| 1599 |
+
{
|
| 1600 |
+
"epoch": 0.81,
|
| 1601 |
+
"learning_rate": 5.190874281132851e-08,
|
| 1602 |
+
"logits/chosen": 0.22277125716209412,
|
| 1603 |
+
"logits/rejected": 0.6487134099006653,
|
| 1604 |
+
"logps/chosen": -402.0958557128906,
|
| 1605 |
+
"logps/rejected": -448.5992736816406,
|
| 1606 |
+
"loss": 0.5408,
|
| 1607 |
+
"rewards/accuracies": 0.71875,
|
| 1608 |
+
"rewards/chosen": -2.359062671661377,
|
| 1609 |
+
"rewards/margins": 0.6533006429672241,
|
| 1610 |
+
"rewards/rejected": -3.0123631954193115,
|
| 1611 |
+
"step": 1020
|
| 1612 |
},
|
| 1613 |
{
|
| 1614 |
"epoch": 0.82,
|
| 1615 |
+
"learning_rate": 4.774575140626316e-08,
|
| 1616 |
+
"logits/chosen": 0.23170511424541473,
|
| 1617 |
+
"logits/rejected": 0.47184085845947266,
|
| 1618 |
+
"logps/chosen": -363.46917724609375,
|
| 1619 |
+
"logps/rejected": -442.47918701171875,
|
| 1620 |
+
"loss": 0.5309,
|
| 1621 |
+
"rewards/accuracies": 0.7437499761581421,
|
| 1622 |
+
"rewards/chosen": -2.142770290374756,
|
| 1623 |
+
"rewards/margins": 0.7513145208358765,
|
| 1624 |
+
"rewards/rejected": -2.894084930419922,
|
| 1625 |
+
"step": 1030
|
| 1626 |
+
},
|
| 1627 |
+
{
|
| 1628 |
+
"epoch": 0.83,
|
| 1629 |
+
"learning_rate": 4.373907927832513e-08,
|
| 1630 |
+
"logits/chosen": 0.07573021948337555,
|
| 1631 |
+
"logits/rejected": 0.32997313141822815,
|
| 1632 |
+
"logps/chosen": -381.45599365234375,
|
| 1633 |
+
"logps/rejected": -443.0684509277344,
|
| 1634 |
+
"loss": 0.5407,
|
| 1635 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 1636 |
+
"rewards/chosen": -2.2033116817474365,
|
| 1637 |
+
"rewards/margins": 0.710732638835907,
|
| 1638 |
+
"rewards/rejected": -2.914044141769409,
|
| 1639 |
+
"step": 1040
|
| 1640 |
},
|
| 1641 |
{
|
| 1642 |
"epoch": 0.84,
|
| 1643 |
+
"learning_rate": 3.9891823124345665e-08,
|
| 1644 |
+
"logits/chosen": 0.23884686827659607,
|
| 1645 |
+
"logits/rejected": 0.6128005385398865,
|
| 1646 |
+
"logps/chosen": -364.00567626953125,
|
| 1647 |
+
"logps/rejected": -433.3273010253906,
|
| 1648 |
+
"loss": 0.5471,
|
| 1649 |
+
"rewards/accuracies": 0.731249988079071,
|
| 1650 |
+
"rewards/chosen": -2.2081527709960938,
|
| 1651 |
+
"rewards/margins": 0.7639263868331909,
|
| 1652 |
+
"rewards/rejected": -2.972079038619995,
|
| 1653 |
+
"step": 1050
|
| 1654 |
},
|
| 1655 |
{
|
| 1656 |
"epoch": 0.84,
|
| 1657 |
+
"learning_rate": 3.620695643093924e-08,
|
| 1658 |
+
"logits/chosen": 0.21963253617286682,
|
| 1659 |
+
"logits/rejected": 0.6894062757492065,
|
| 1660 |
+
"logps/chosen": -399.5767517089844,
|
| 1661 |
+
"logps/rejected": -452.88909912109375,
|
| 1662 |
+
"loss": 0.5154,
|
| 1663 |
+
"rewards/accuracies": 0.71875,
|
| 1664 |
+
"rewards/chosen": -2.3384335041046143,
|
| 1665 |
+
"rewards/margins": 0.7010769844055176,
|
| 1666 |
+
"rewards/rejected": -3.0395102500915527,
|
| 1667 |
+
"step": 1060
|
| 1668 |
+
},
|
| 1669 |
+
{
|
| 1670 |
+
"epoch": 0.85,
|
| 1671 |
+
"learning_rate": 3.268732717634032e-08,
|
| 1672 |
+
"logits/chosen": 0.3474286198616028,
|
| 1673 |
+
"logits/rejected": 0.695271372795105,
|
| 1674 |
+
"logps/chosen": -368.0654602050781,
|
| 1675 |
+
"logps/rejected": -431.47222900390625,
|
| 1676 |
+
"loss": 0.5499,
|
| 1677 |
+
"rewards/accuracies": 0.731249988079071,
|
| 1678 |
+
"rewards/chosen": -2.1910276412963867,
|
| 1679 |
+
"rewards/margins": 0.7267633080482483,
|
| 1680 |
+
"rewards/rejected": -2.9177908897399902,
|
| 1681 |
+
"step": 1070
|
| 1682 |
},
|
| 1683 |
{
|
| 1684 |
"epoch": 0.86,
|
| 1685 |
+
"learning_rate": 2.9335655629243645e-08,
|
| 1686 |
+
"logits/chosen": 0.2347393035888672,
|
| 1687 |
+
"logits/rejected": 0.5894696712493896,
|
| 1688 |
+
"logps/chosen": -388.94757080078125,
|
| 1689 |
+
"logps/rejected": -447.3855895996094,
|
| 1690 |
+
"loss": 0.525,
|
| 1691 |
+
"rewards/accuracies": 0.78125,
|
| 1692 |
+
"rewards/chosen": -2.254683017730713,
|
| 1693 |
+
"rewards/margins": 0.7334609031677246,
|
| 1694 |
+
"rewards/rejected": -2.9881439208984375,
|
| 1695 |
+
"step": 1080
|
| 1696 |
+
},
|
| 1697 |
+
{
|
| 1698 |
+
"epoch": 0.87,
|
| 1699 |
+
"learning_rate": 2.6154532246349476e-08,
|
| 1700 |
+
"logits/chosen": 0.25378522276878357,
|
| 1701 |
+
"logits/rejected": 0.5771256685256958,
|
| 1702 |
+
"logps/chosen": -358.50640869140625,
|
| 1703 |
+
"logps/rejected": -431.145751953125,
|
| 1704 |
+
"loss": 0.5462,
|
| 1705 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 1706 |
+
"rewards/chosen": -2.1666626930236816,
|
| 1707 |
+
"rewards/margins": 0.756801426410675,
|
| 1708 |
+
"rewards/rejected": -2.923464059829712,
|
| 1709 |
+
"step": 1090
|
| 1710 |
},
|
| 1711 |
{
|
| 1712 |
"epoch": 0.88,
|
| 1713 |
+
"learning_rate": 2.31464156702382e-08,
|
| 1714 |
+
"logits/chosen": 0.35370689630508423,
|
| 1715 |
+
"logits/rejected": 0.5671936273574829,
|
| 1716 |
+
"logps/chosen": -363.0,
|
| 1717 |
+
"logps/rejected": -438.209228515625,
|
| 1718 |
+
"loss": 0.5487,
|
| 1719 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 1720 |
+
"rewards/chosen": -2.2130119800567627,
|
| 1721 |
+
"rewards/margins": 0.7499077916145325,
|
| 1722 |
+
"rewards/rejected": -2.9629194736480713,
|
| 1723 |
+
"step": 1100
|
| 1724 |
+
},
|
| 1725 |
+
{
|
| 1726 |
+
"epoch": 0.88,
|
| 1727 |
+
"eval_logits/chosen": 0.1857856959104538,
|
| 1728 |
+
"eval_logits/rejected": 0.43363669514656067,
|
| 1729 |
+
"eval_logps/chosen": -373.08306884765625,
|
| 1730 |
+
"eval_logps/rejected": -450.7598876953125,
|
| 1731 |
+
"eval_loss": 0.5444055199623108,
|
| 1732 |
+
"eval_rewards/accuracies": 0.7089552283287048,
|
| 1733 |
+
"eval_rewards/chosen": -2.285095453262329,
|
| 1734 |
+
"eval_rewards/margins": 0.711159884929657,
|
| 1735 |
+
"eval_rewards/rejected": -2.996255397796631,
|
| 1736 |
+
"eval_runtime": 184.462,
|
| 1737 |
+
"eval_samples_per_second": 46.362,
|
| 1738 |
+
"eval_steps_per_second": 0.726,
|
| 1739 |
+
"step": 1100
|
| 1740 |
+
},
|
| 1741 |
+
{
|
| 1742 |
+
"epoch": 0.88,
|
| 1743 |
+
"learning_rate": 2.031363082912252e-08,
|
| 1744 |
+
"logits/chosen": 0.070524200797081,
|
| 1745 |
+
"logits/rejected": 0.4635602533817291,
|
| 1746 |
+
"logps/chosen": -373.29327392578125,
|
| 1747 |
+
"logps/rejected": -426.85552978515625,
|
| 1748 |
+
"loss": 0.5513,
|
| 1749 |
+
"rewards/accuracies": 0.6937500238418579,
|
| 1750 |
+
"rewards/chosen": -2.2541089057922363,
|
| 1751 |
+
"rewards/margins": 0.6198171973228455,
|
| 1752 |
+
"rewards/rejected": -2.8739261627197266,
|
| 1753 |
+
"step": 1110
|
| 1754 |
+
},
|
| 1755 |
+
{
|
| 1756 |
+
"epoch": 0.89,
|
| 1757 |
+
"learning_rate": 1.7658367139945228e-08,
|
| 1758 |
+
"logits/chosen": 0.2600646913051605,
|
| 1759 |
+
"logits/rejected": 0.5517584681510925,
|
| 1760 |
+
"logps/chosen": -390.8568115234375,
|
| 1761 |
+
"logps/rejected": -462.80828857421875,
|
| 1762 |
+
"loss": 0.5471,
|
| 1763 |
+
"rewards/accuracies": 0.6875,
|
| 1764 |
+
"rewards/chosen": -2.377396821975708,
|
| 1765 |
+
"rewards/margins": 0.6719989776611328,
|
| 1766 |
+
"rewards/rejected": -3.049395799636841,
|
| 1767 |
+
"step": 1120
|
| 1768 |
},
|
| 1769 |
{
|
| 1770 |
"epoch": 0.9,
|
| 1771 |
+
"learning_rate": 1.5182676816211632e-08,
|
| 1772 |
+
"logits/chosen": 0.04413030296564102,
|
| 1773 |
+
"logits/rejected": 0.30151715874671936,
|
| 1774 |
+
"logps/chosen": -382.0662536621094,
|
| 1775 |
+
"logps/rejected": -447.08673095703125,
|
| 1776 |
+
"loss": 0.5431,
|
| 1777 |
+
"rewards/accuracies": 0.7124999761581421,
|
| 1778 |
+
"rewards/chosen": -2.228654384613037,
|
| 1779 |
+
"rewards/margins": 0.6926024556159973,
|
| 1780 |
+
"rewards/rejected": -2.9212570190429688,
|
| 1781 |
+
"step": 1130
|
| 1782 |
+
},
|
| 1783 |
+
{
|
| 1784 |
+
"epoch": 0.91,
|
| 1785 |
+
"learning_rate": 1.2888473281864597e-08,
|
| 1786 |
+
"logits/chosen": 0.14212054014205933,
|
| 1787 |
+
"logits/rejected": 0.47429710626602173,
|
| 1788 |
+
"logps/chosen": -367.8409729003906,
|
| 1789 |
+
"logps/rejected": -435.02764892578125,
|
| 1790 |
+
"loss": 0.5369,
|
| 1791 |
+
"rewards/accuracies": 0.768750011920929,
|
| 1792 |
+
"rewards/chosen": -2.2534115314483643,
|
| 1793 |
+
"rewards/margins": 0.7273036241531372,
|
| 1794 |
+
"rewards/rejected": -2.980715274810791,
|
| 1795 |
+
"step": 1140
|
| 1796 |
},
|
| 1797 |
{
|
| 1798 |
"epoch": 0.92,
|
| 1799 |
+
"learning_rate": 1.0777529692427679e-08,
|
| 1800 |
+
"logits/chosen": 0.04115242511034012,
|
| 1801 |
+
"logits/rejected": 0.28970104455947876,
|
| 1802 |
+
"logps/chosen": -372.7949523925781,
|
| 1803 |
+
"logps/rejected": -456.10675048828125,
|
| 1804 |
+
"loss": 0.5265,
|
| 1805 |
+
"rewards/accuracies": 0.731249988079071,
|
| 1806 |
+
"rewards/chosen": -2.300356388092041,
|
| 1807 |
+
"rewards/margins": 0.8059718012809753,
|
| 1808 |
+
"rewards/rejected": -3.106328248977661,
|
| 1809 |
+
"step": 1150
|
| 1810 |
},
|
| 1811 |
{
|
| 1812 |
+
"epoch": 0.92,
|
| 1813 |
+
"learning_rate": 8.851477564560061e-09,
|
| 1814 |
+
"logits/chosen": 0.0867738351225853,
|
| 1815 |
+
"logits/rejected": 0.4068300127983093,
|
| 1816 |
+
"logps/chosen": -372.08636474609375,
|
| 1817 |
+
"logps/rejected": -426.42388916015625,
|
| 1818 |
+
"loss": 0.5342,
|
| 1819 |
"rewards/accuracies": 0.6937500238418579,
|
| 1820 |
+
"rewards/chosen": -2.331385850906372,
|
| 1821 |
+
"rewards/margins": 0.6490964293479919,
|
| 1822 |
+
"rewards/rejected": -2.9804821014404297,
|
| 1823 |
+
"step": 1160
|
| 1824 |
+
},
|
| 1825 |
+
{
|
| 1826 |
+
"epoch": 0.93,
|
| 1827 |
+
"learning_rate": 7.111805515081531e-09,
|
| 1828 |
+
"logits/chosen": 0.02022993005812168,
|
| 1829 |
+
"logits/rejected": 0.41968393325805664,
|
| 1830 |
+
"logps/chosen": -363.818603515625,
|
| 1831 |
+
"logps/rejected": -447.7919006347656,
|
| 1832 |
+
"loss": 0.5312,
|
| 1833 |
+
"rewards/accuracies": 0.7562500238418579,
|
| 1834 |
+
"rewards/chosen": -2.2372307777404785,
|
| 1835 |
+
"rewards/margins": 0.8540315628051758,
|
| 1836 |
+
"rewards/rejected": -3.0912623405456543,
|
| 1837 |
+
"step": 1170
|
| 1838 |
+
},
|
| 1839 |
+
{
|
| 1840 |
+
"epoch": 0.94,
|
| 1841 |
+
"learning_rate": 5.559858110443016e-09,
|
| 1842 |
+
"logits/chosen": 0.29695388674736023,
|
| 1843 |
+
"logits/rejected": 0.714096188545227,
|
| 1844 |
+
"logps/chosen": -372.5519714355469,
|
| 1845 |
+
"logps/rejected": -442.5354919433594,
|
| 1846 |
+
"loss": 0.5383,
|
| 1847 |
+
"rewards/accuracies": 0.71875,
|
| 1848 |
+
"rewards/chosen": -2.3107995986938477,
|
| 1849 |
+
"rewards/margins": 0.8070123791694641,
|
| 1850 |
+
"rewards/rejected": -3.117811918258667,
|
| 1851 |
+
"step": 1180
|
| 1852 |
+
},
|
| 1853 |
+
{
|
| 1854 |
+
"epoch": 0.95,
|
| 1855 |
+
"learning_rate": 4.196834827531276e-09,
|
| 1856 |
+
"logits/chosen": 0.140055850148201,
|
| 1857 |
+
"logits/rejected": 0.3409932255744934,
|
| 1858 |
+
"logps/chosen": -355.64324951171875,
|
| 1859 |
+
"logps/rejected": -447.585693359375,
|
| 1860 |
+
"loss": 0.5152,
|
| 1861 |
+
"rewards/accuracies": 0.737500011920929,
|
| 1862 |
+
"rewards/chosen": -2.149151563644409,
|
| 1863 |
+
"rewards/margins": 0.7904965877532959,
|
| 1864 |
+
"rewards/rejected": -2.939648151397705,
|
| 1865 |
+
"step": 1190
|
| 1866 |
},
|
| 1867 |
{
|
| 1868 |
"epoch": 0.96,
|
| 1869 |
+
"learning_rate": 3.023789126611137e-09,
|
| 1870 |
+
"logits/chosen": 0.03294936567544937,
|
| 1871 |
+
"logits/rejected": 0.2933207154273987,
|
| 1872 |
+
"logps/chosen": -363.29290771484375,
|
| 1873 |
+
"logps/rejected": -435.640380859375,
|
| 1874 |
+
"loss": 0.5483,
|
| 1875 |
+
"rewards/accuracies": 0.699999988079071,
|
| 1876 |
+
"rewards/chosen": -2.281057357788086,
|
| 1877 |
+
"rewards/margins": 0.7091296911239624,
|
| 1878 |
+
"rewards/rejected": -2.990186929702759,
|
| 1879 |
+
"step": 1200
|
| 1880 |
+
},
|
| 1881 |
+
{
|
| 1882 |
+
"epoch": 0.96,
|
| 1883 |
+
"eval_logits/chosen": 0.07418080419301987,
|
| 1884 |
+
"eval_logits/rejected": 0.32435521483421326,
|
| 1885 |
+
"eval_logps/chosen": -373.978515625,
|
| 1886 |
+
"eval_logps/rejected": -451.6764831542969,
|
| 1887 |
+
"eval_loss": 0.5440130829811096,
|
| 1888 |
+
"eval_rewards/accuracies": 0.7089552283287048,
|
| 1889 |
+
"eval_rewards/chosen": -2.2940499782562256,
|
| 1890 |
+
"eval_rewards/margins": 0.7113713622093201,
|
| 1891 |
+
"eval_rewards/rejected": -3.0054211616516113,
|
| 1892 |
+
"eval_runtime": 184.3589,
|
| 1893 |
+
"eval_samples_per_second": 46.388,
|
| 1894 |
+
"eval_steps_per_second": 0.727,
|
| 1895 |
+
"step": 1200
|
| 1896 |
+
},
|
| 1897 |
+
{
|
| 1898 |
+
"epoch": 0.96,
|
| 1899 |
+
"learning_rate": 2.041627637121929e-09,
|
| 1900 |
+
"logits/chosen": 0.10010697692632675,
|
| 1901 |
+
"logits/rejected": 0.3795483410358429,
|
| 1902 |
+
"logps/chosen": -348.8675231933594,
|
| 1903 |
+
"logps/rejected": -437.20361328125,
|
| 1904 |
+
"loss": 0.5398,
|
| 1905 |
+
"rewards/accuracies": 0.7749999761581421,
|
| 1906 |
+
"rewards/chosen": -2.087364673614502,
|
| 1907 |
+
"rewards/margins": 0.828387439250946,
|
| 1908 |
+
"rewards/rejected": -2.9157521724700928,
|
| 1909 |
+
"step": 1210
|
| 1910 |
+
},
|
| 1911 |
+
{
|
| 1912 |
+
"epoch": 0.97,
|
| 1913 |
+
"learning_rate": 1.2511094569571668e-09,
|
| 1914 |
+
"logits/chosen": 0.09991980344057083,
|
| 1915 |
+
"logits/rejected": 0.4467397630214691,
|
| 1916 |
+
"logps/chosen": -380.14520263671875,
|
| 1917 |
+
"logps/rejected": -440.24658203125,
|
| 1918 |
+
"loss": 0.5345,
|
| 1919 |
+
"rewards/accuracies": 0.737500011920929,
|
| 1920 |
+
"rewards/chosen": -2.253425121307373,
|
| 1921 |
+
"rewards/margins": 0.702509880065918,
|
| 1922 |
+
"rewards/rejected": -2.955935001373291,
|
| 1923 |
+
"step": 1220
|
| 1924 |
},
|
| 1925 |
{
|
| 1926 |
"epoch": 0.98,
|
| 1927 |
+
"learning_rate": 6.528455657691112e-10,
|
| 1928 |
+
"logits/chosen": 0.11626466363668442,
|
| 1929 |
+
"logits/rejected": 0.41348797082901,
|
| 1930 |
+
"logps/chosen": -372.7298889160156,
|
| 1931 |
+
"logps/rejected": -427.22576904296875,
|
| 1932 |
+
"loss": 0.549,
|
| 1933 |
+
"rewards/accuracies": 0.762499988079071,
|
| 1934 |
+
"rewards/chosen": -2.2800345420837402,
|
| 1935 |
+
"rewards/margins": 0.6291176080703735,
|
| 1936 |
+
"rewards/rejected": -2.909151792526245,
|
| 1937 |
+
"step": 1230
|
| 1938 |
+
},
|
| 1939 |
+
{
|
| 1940 |
+
"epoch": 0.99,
|
| 1941 |
+
"learning_rate": 2.4729835275189016e-10,
|
| 1942 |
+
"logits/chosen": 0.06715863198041916,
|
| 1943 |
+
"logits/rejected": 0.29241910576820374,
|
| 1944 |
+
"logps/chosen": -393.8903503417969,
|
| 1945 |
+
"logps/rejected": -477.9420471191406,
|
| 1946 |
+
"loss": 0.5462,
|
| 1947 |
"rewards/accuracies": 0.737500011920929,
|
| 1948 |
+
"rewards/chosen": -2.323488473892212,
|
| 1949 |
+
"rewards/margins": 0.8067766427993774,
|
| 1950 |
+
"rewards/rejected": -3.1302647590637207,
|
| 1951 |
+
"step": 1240
|
| 1952 |
+
},
|
| 1953 |
+
{
|
| 1954 |
+
"epoch": 0.99,
|
| 1955 |
+
"learning_rate": 3.478125926756337e-11,
|
| 1956 |
+
"logits/chosen": 0.25983649492263794,
|
| 1957 |
+
"logits/rejected": 0.4905417561531067,
|
| 1958 |
+
"logps/chosen": -364.73431396484375,
|
| 1959 |
+
"logps/rejected": -443.79296875,
|
| 1960 |
+
"loss": 0.5474,
|
| 1961 |
+
"rewards/accuracies": 0.7250000238418579,
|
| 1962 |
+
"rewards/chosen": -2.2353272438049316,
|
| 1963 |
+
"rewards/margins": 0.771331787109375,
|
| 1964 |
+
"rewards/rejected": -3.0066590309143066,
|
| 1965 |
+
"step": 1250
|
| 1966 |
},
|
| 1967 |
{
|
| 1968 |
"epoch": 1.0,
|
| 1969 |
+
"step": 1256,
|
| 1970 |
"total_flos": 0.0,
|
| 1971 |
+
"train_loss": 0.5712926928784438,
|
| 1972 |
+
"train_runtime": 11573.5451,
|
| 1973 |
+
"train_samples_per_second": 13.894,
|
| 1974 |
+
"train_steps_per_second": 0.109
|
| 1975 |
}
|
| 1976 |
],
|
| 1977 |
"logging_steps": 10,
|
| 1978 |
+
"max_steps": 1256,
|
| 1979 |
"num_train_epochs": 1,
|
| 1980 |
"save_steps": 100,
|
| 1981 |
"total_flos": 0.0,
|
training_args.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 5944
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b0a380877b07afc91f8cd53d271781e6ac2fa02e9d60a2c1c357997bd753b437
|
| 3 |
size 5944
|