Model save
Browse files- README.md +21 -24
- adapter_model.safetensors +1 -1
- all_results.json +2 -17
- train_results.json +2 -2
- trainer_state.json +681 -1497
README.md
CHANGED
|
@@ -2,13 +2,10 @@
|
|
| 2 |
license: apache-2.0
|
| 3 |
library_name: peft
|
| 4 |
tags:
|
| 5 |
-
- alignment-handbook
|
| 6 |
- trl
|
| 7 |
- dpo
|
| 8 |
- generated_from_trainer
|
| 9 |
base_model: mistralai/Mistral-7B-v0.1
|
| 10 |
-
datasets:
|
| 11 |
-
- HuggingFaceH4/ultrafeedback_binarized
|
| 12 |
model-index:
|
| 13 |
- name: zephyr-7b
|
| 14 |
results: []
|
|
@@ -19,19 +16,19 @@ should probably proofread and complete it, then remove this comment. -->
|
|
| 19 |
|
| 20 |
# zephyr-7b
|
| 21 |
|
| 22 |
-
This model is a fine-tuned version of [
|
| 23 |
It achieves the following results on the evaluation set:
|
| 24 |
-
- Loss: 0.
|
| 25 |
-
- Rewards/chosen: -0.
|
| 26 |
-
- Rewards/rejected: -0.
|
| 27 |
-
- Rewards/accuracies: 0.
|
| 28 |
-
- Rewards/margins: 0.
|
| 29 |
-
- Logps/rejected: -
|
| 30 |
-
- Logps/chosen: -
|
| 31 |
-
- Logits/rejected: -
|
| 32 |
-
- Logits/chosen: -
|
| 33 |
-
- Use Label:
|
| 34 |
-
- Pred Label:
|
| 35 |
|
| 36 |
## Model description
|
| 37 |
|
|
@@ -68,15 +65,15 @@ The following hyperparameters were used during training:
|
|
| 68 |
|
| 69 |
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen | Use Label | Pred Label |
|
| 70 |
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|:----------:|:----------:|
|
| 71 |
-
| 0.
|
| 72 |
-
| 0.
|
| 73 |
-
| 0.
|
| 74 |
-
| 0.
|
| 75 |
-
| 0.
|
| 76 |
-
| 0.
|
| 77 |
-
| 0.
|
| 78 |
-
| 0.
|
| 79 |
-
| 0.
|
| 80 |
|
| 81 |
|
| 82 |
### Framework versions
|
|
|
|
| 2 |
license: apache-2.0
|
| 3 |
library_name: peft
|
| 4 |
tags:
|
|
|
|
| 5 |
- trl
|
| 6 |
- dpo
|
| 7 |
- generated_from_trainer
|
| 8 |
base_model: mistralai/Mistral-7B-v0.1
|
|
|
|
|
|
|
| 9 |
model-index:
|
| 10 |
- name: zephyr-7b
|
| 11 |
results: []
|
|
|
|
| 16 |
|
| 17 |
# zephyr-7b
|
| 18 |
|
| 19 |
+
This model is a fine-tuned version of [mistralai/Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) on the None dataset.
|
| 20 |
It achieves the following results on the evaluation set:
|
| 21 |
+
- Loss: 0.6918
|
| 22 |
+
- Rewards/chosen: -0.0863
|
| 23 |
+
- Rewards/rejected: -0.1983
|
| 24 |
+
- Rewards/accuracies: 0.3571
|
| 25 |
+
- Rewards/margins: 0.1120
|
| 26 |
+
- Logps/rejected: -95.2291
|
| 27 |
+
- Logps/chosen: -77.5275
|
| 28 |
+
- Logits/rejected: -1.9113
|
| 29 |
+
- Logits/chosen: -1.9391
|
| 30 |
+
- Use Label: 14335.7139
|
| 31 |
+
- Pred Label: 4352.2856
|
| 32 |
|
| 33 |
## Model description
|
| 34 |
|
|
|
|
| 65 |
|
| 66 |
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen | Use Label | Pred Label |
|
| 67 |
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|:----------:|:----------:|
|
| 68 |
+
| 0.6876 | 0.1 | 100 | 0.6896 | -0.0555 | -0.0989 | 0.3353 | 0.0434 | -85.2883 | -74.4495 | -2.0761 | -2.1076 | 1766.8572 | 89.1429 |
|
| 69 |
+
| 0.6892 | 0.21 | 200 | 0.6894 | -0.0049 | -0.0560 | 0.3492 | 0.0511 | -80.9954 | -69.3876 | -2.0287 | -2.0520 | 3500.8889 | 459.1111 |
|
| 70 |
+
| 0.6904 | 0.31 | 300 | 0.6909 | -0.0625 | -0.1410 | 0.3532 | 0.0785 | -89.5016 | -75.1524 | -1.9943 | -2.0164 | 5140.6826 | 923.3174 |
|
| 71 |
+
| 0.6906 | 0.42 | 400 | 0.6921 | -0.0637 | -0.1541 | 0.3512 | 0.0904 | -90.8064 | -75.2687 | -2.0248 | -2.0481 | 6695.4287 | 1472.5714 |
|
| 72 |
+
| 0.6903 | 0.52 | 500 | 0.6914 | -0.0747 | -0.1726 | 0.3492 | 0.0979 | -92.6561 | -76.3697 | -1.9801 | -2.0071 | 8246.2061 | 2025.7937 |
|
| 73 |
+
| 0.6903 | 0.63 | 600 | 0.6917 | -0.1005 | -0.2047 | 0.3552 | 0.1042 | -95.8670 | -78.9543 | -1.9601 | -1.9870 | 9772.0635 | 2603.9365 |
|
| 74 |
+
| 0.6917 | 0.73 | 700 | 0.6917 | -0.1117 | -0.2224 | 0.3512 | 0.1108 | -97.6411 | -80.0681 | -1.9401 | -1.9659 | 11284.7773 | 3195.2222 |
|
| 75 |
+
| 0.6912 | 0.84 | 800 | 0.6917 | -0.0869 | -0.1981 | 0.3631 | 0.1112 | -95.2089 | -77.5874 | -1.9144 | -1.9422 | 12826.8252 | 3757.1746 |
|
| 76 |
+
| 0.6914 | 0.94 | 900 | 0.6918 | -0.0863 | -0.1983 | 0.3571 | 0.1120 | -95.2291 | -77.5275 | -1.9113 | -1.9391 | 14335.7139 | 4352.2856 |
|
| 77 |
|
| 78 |
|
| 79 |
### Framework versions
|
adapter_model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 671150064
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:37ff211fdd55e89806d0731a069b1a1347270a24364fedc285bf81c9b757d749
|
| 3 |
size 671150064
|
all_results.json
CHANGED
|
@@ -1,22 +1,7 @@
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
-
"
|
| 4 |
-
"
|
| 5 |
-
"eval_logps/chosen": -103.02954864501953,
|
| 6 |
-
"eval_logps/rejected": -131.91891479492188,
|
| 7 |
-
"eval_loss": 0.6906174421310425,
|
| 8 |
-
"eval_pred_label": 4192.14306640625,
|
| 9 |
-
"eval_rewards/accuracies": 0.363095223903656,
|
| 10 |
-
"eval_rewards/chosen": -0.3412899374961853,
|
| 11 |
-
"eval_rewards/margins": 0.22391849756240845,
|
| 12 |
-
"eval_rewards/rejected": -0.5652084350585938,
|
| 13 |
-
"eval_runtime": 247.5585,
|
| 14 |
-
"eval_samples": 2000,
|
| 15 |
-
"eval_samples_per_second": 8.079,
|
| 16 |
-
"eval_steps_per_second": 0.254,
|
| 17 |
-
"eval_use_label": 15879.857421875,
|
| 18 |
-
"train_loss": 0.6880922077838039,
|
| 19 |
-
"train_runtime": 20023.3666,
|
| 20 |
"train_samples": 61135,
|
| 21 |
"train_samples_per_second": 3.053,
|
| 22 |
"train_steps_per_second": 0.048
|
|
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
+
"train_loss": 0.6906769273168754,
|
| 4 |
+
"train_runtime": 20027.4031,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 5 |
"train_samples": 61135,
|
| 6 |
"train_samples_per_second": 3.053,
|
| 7 |
"train_steps_per_second": 0.048
|
train_results.json
CHANGED
|
@@ -1,7 +1,7 @@
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
-
"train_loss": 0.
|
| 4 |
-
"train_runtime":
|
| 5 |
"train_samples": 61135,
|
| 6 |
"train_samples_per_second": 3.053,
|
| 7 |
"train_steps_per_second": 0.048
|
|
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
+
"train_loss": 0.6906769273168754,
|
| 4 |
+
"train_runtime": 20027.4031,
|
| 5 |
"train_samples": 61135,
|
| 6 |
"train_samples_per_second": 3.053,
|
| 7 |
"train_steps_per_second": 0.048
|
trainer_state.json
CHANGED
|
@@ -25,1798 +25,982 @@
|
|
| 25 |
"step": 1,
|
| 26 |
"use_label": 10.0
|
| 27 |
},
|
| 28 |
-
{
|
| 29 |
-
"epoch": 0.01,
|
| 30 |
-
"grad_norm": 0.60546875,
|
| 31 |
-
"learning_rate": 5.208333333333334e-07,
|
| 32 |
-
"logits/chosen": -2.2113068103790283,
|
| 33 |
-
"logits/rejected": -2.2719719409942627,
|
| 34 |
-
"logps/chosen": -57.57659149169922,
|
| 35 |
-
"logps/rejected": -65.19544219970703,
|
| 36 |
-
"loss": 0.693,
|
| 37 |
-
"pred_label": 0.0,
|
| 38 |
-
"rewards/accuracies": 0.2152777761220932,
|
| 39 |
-
"rewards/chosen": 0.001057142741046846,
|
| 40 |
-
"rewards/margins": 3.17241829179693e-05,
|
| 41 |
-
"rewards/rejected": 0.001025418401695788,
|
| 42 |
-
"step": 10,
|
| 43 |
-
"use_label": 90.0
|
| 44 |
-
},
|
| 45 |
{
|
| 46 |
"epoch": 0.02,
|
| 47 |
"grad_norm": 0.6796875,
|
| 48 |
"learning_rate": 1.0416666666666667e-06,
|
| 49 |
-
"logits/chosen": -2.
|
| 50 |
-
"logits/rejected": -2.
|
| 51 |
-
"logps/chosen": -
|
| 52 |
-
"logps/rejected": -
|
| 53 |
-
"loss": 0.
|
| 54 |
"pred_label": 0.0,
|
| 55 |
-
"rewards/accuracies": 0.
|
| 56 |
-
"rewards/chosen": 0.
|
| 57 |
-
"rewards/margins": 0.
|
| 58 |
-
"rewards/rejected": 0.
|
| 59 |
"step": 20,
|
| 60 |
-
"use_label":
|
| 61 |
-
},
|
| 62 |
-
{
|
| 63 |
-
"epoch": 0.03,
|
| 64 |
-
"grad_norm": 0.55078125,
|
| 65 |
-
"learning_rate": 1.5625e-06,
|
| 66 |
-
"logits/chosen": -2.2634024620056152,
|
| 67 |
-
"logits/rejected": -2.2475943565368652,
|
| 68 |
-
"logps/chosen": -53.98667526245117,
|
| 69 |
-
"logps/rejected": -67.89213562011719,
|
| 70 |
-
"loss": 0.692,
|
| 71 |
-
"pred_label": 0.0,
|
| 72 |
-
"rewards/accuracies": 0.2750000059604645,
|
| 73 |
-
"rewards/chosen": 0.01648966409265995,
|
| 74 |
-
"rewards/margins": 0.002599921775981784,
|
| 75 |
-
"rewards/rejected": 0.013889740221202374,
|
| 76 |
-
"step": 30,
|
| 77 |
-
"use_label": 402.0
|
| 78 |
},
|
| 79 |
{
|
| 80 |
"epoch": 0.04,
|
| 81 |
"grad_norm": 0.6328125,
|
| 82 |
"learning_rate": 2.0833333333333334e-06,
|
| 83 |
-
"logits/chosen": -2.
|
| 84 |
-
"logits/rejected": -2.
|
| 85 |
-
"logps/chosen": -
|
| 86 |
-
"logps/rejected": -
|
| 87 |
-
"loss": 0.
|
| 88 |
"pred_label": 0.0,
|
| 89 |
-
"rewards/accuracies": 0.
|
| 90 |
-
"rewards/chosen": 0.
|
| 91 |
-
"rewards/margins": 0.
|
| 92 |
-
"rewards/rejected": 0.
|
| 93 |
"step": 40,
|
| 94 |
-
"use_label":
|
| 95 |
-
},
|
| 96 |
-
{
|
| 97 |
-
"epoch": 0.05,
|
| 98 |
-
"grad_norm": 0.6015625,
|
| 99 |
-
"learning_rate": 2.604166666666667e-06,
|
| 100 |
-
"logits/chosen": -2.3444912433624268,
|
| 101 |
-
"logits/rejected": -2.3341281414031982,
|
| 102 |
-
"logps/chosen": -69.13630676269531,
|
| 103 |
-
"logps/rejected": -84.64376831054688,
|
| 104 |
-
"loss": 0.6889,
|
| 105 |
-
"pred_label": 0.0,
|
| 106 |
-
"rewards/accuracies": 0.2874999940395355,
|
| 107 |
-
"rewards/chosen": 0.02657836303114891,
|
| 108 |
-
"rewards/margins": 0.005359734408557415,
|
| 109 |
-
"rewards/rejected": 0.021218623965978622,
|
| 110 |
-
"step": 50,
|
| 111 |
-
"use_label": 722.0
|
| 112 |
},
|
| 113 |
{
|
| 114 |
"epoch": 0.06,
|
| 115 |
-
"grad_norm": 0.
|
| 116 |
"learning_rate": 3.125e-06,
|
| 117 |
-
"logits/chosen": -2.
|
| 118 |
-
"logits/rejected": -2.
|
| 119 |
-
"logps/chosen": -
|
| 120 |
-
"logps/rejected": -
|
| 121 |
-
"loss": 0.
|
| 122 |
"pred_label": 0.0,
|
| 123 |
-
"rewards/accuracies": 0.
|
| 124 |
-
"rewards/chosen": 0.
|
| 125 |
-
"rewards/margins": 0.
|
| 126 |
-
"rewards/rejected": 0.
|
| 127 |
"step": 60,
|
| 128 |
-
"use_label":
|
| 129 |
-
},
|
| 130 |
-
{
|
| 131 |
-
"epoch": 0.07,
|
| 132 |
-
"grad_norm": 0.79296875,
|
| 133 |
-
"learning_rate": 3.6458333333333333e-06,
|
| 134 |
-
"logits/chosen": -2.344853401184082,
|
| 135 |
-
"logits/rejected": -2.3261306285858154,
|
| 136 |
-
"logps/chosen": -77.20336151123047,
|
| 137 |
-
"logps/rejected": -77.6347885131836,
|
| 138 |
-
"loss": 0.6851,
|
| 139 |
-
"pred_label": 0.0,
|
| 140 |
-
"rewards/accuracies": 0.30000001192092896,
|
| 141 |
-
"rewards/chosen": 0.02531364932656288,
|
| 142 |
-
"rewards/margins": 0.01608472317457199,
|
| 143 |
-
"rewards/rejected": 0.009228924289345741,
|
| 144 |
-
"step": 70,
|
| 145 |
-
"use_label": 1042.0
|
| 146 |
},
|
| 147 |
{
|
| 148 |
"epoch": 0.08,
|
| 149 |
-
"grad_norm": 0.
|
| 150 |
"learning_rate": 4.166666666666667e-06,
|
| 151 |
-
"logits/chosen": -2.
|
| 152 |
-
"logits/rejected": -2.
|
| 153 |
-
"logps/chosen": -
|
| 154 |
-
"logps/rejected": -
|
| 155 |
-
"loss": 0.
|
| 156 |
-
"pred_label":
|
| 157 |
-
"rewards/accuracies": 0.
|
| 158 |
-
"rewards/chosen": 0.
|
| 159 |
-
"rewards/margins": 0.
|
| 160 |
-
"rewards/rejected": -0.
|
| 161 |
"step": 80,
|
| 162 |
-
"use_label":
|
| 163 |
-
},
|
| 164 |
-
{
|
| 165 |
-
"epoch": 0.09,
|
| 166 |
-
"grad_norm": 1.578125,
|
| 167 |
-
"learning_rate": 4.6875000000000004e-06,
|
| 168 |
-
"logits/chosen": -2.1907405853271484,
|
| 169 |
-
"logits/rejected": -2.232959270477295,
|
| 170 |
-
"logps/chosen": -62.31688690185547,
|
| 171 |
-
"logps/rejected": -80.38573455810547,
|
| 172 |
-
"loss": 0.6812,
|
| 173 |
-
"pred_label": 3.0999999046325684,
|
| 174 |
-
"rewards/accuracies": 0.33125001192092896,
|
| 175 |
-
"rewards/chosen": -0.012271342799067497,
|
| 176 |
-
"rewards/margins": 0.04507603123784065,
|
| 177 |
-
"rewards/rejected": -0.0573473684489727,
|
| 178 |
-
"step": 90,
|
| 179 |
-
"use_label": 1358.9000244140625
|
| 180 |
},
|
| 181 |
{
|
| 182 |
"epoch": 0.1,
|
| 183 |
-
"grad_norm": 0.
|
| 184 |
"learning_rate": 4.9997324926814375e-06,
|
| 185 |
-
"logits/chosen": -2.
|
| 186 |
-
"logits/rejected": -2.
|
| 187 |
-
"logps/chosen": -
|
| 188 |
-
"logps/rejected": -
|
| 189 |
-
"loss": 0.
|
| 190 |
-
"pred_label":
|
| 191 |
-
"rewards/accuracies": 0.
|
| 192 |
-
"rewards/chosen": -0.
|
| 193 |
-
"rewards/margins": 0.
|
| 194 |
-
"rewards/rejected": -0.
|
| 195 |
"step": 100,
|
| 196 |
-
"use_label":
|
| 197 |
},
|
| 198 |
{
|
| 199 |
"epoch": 0.1,
|
| 200 |
-
"eval_logits/chosen": -2.
|
| 201 |
-
"eval_logits/rejected": -2.
|
| 202 |
-
"eval_logps/chosen": -
|
| 203 |
-
"eval_logps/rejected": -
|
| 204 |
-
"eval_loss": 0.
|
| 205 |
-
"eval_pred_label":
|
| 206 |
-
"eval_rewards/accuracies": 0.
|
| 207 |
-
"eval_rewards/chosen": -0.
|
| 208 |
-
"eval_rewards/margins": 0.
|
| 209 |
-
"eval_rewards/rejected": -0.
|
| 210 |
-
"eval_runtime":
|
| 211 |
-
"eval_samples_per_second": 8.
|
| 212 |
-
"eval_steps_per_second": 0.
|
| 213 |
-
"eval_use_label":
|
| 214 |
"step": 100
|
| 215 |
},
|
| 216 |
-
{
|
| 217 |
-
"epoch": 0.12,
|
| 218 |
-
"grad_norm": 1.1171875,
|
| 219 |
-
"learning_rate": 4.996723692767927e-06,
|
| 220 |
-
"logits/chosen": -2.114673137664795,
|
| 221 |
-
"logits/rejected": -2.094468355178833,
|
| 222 |
-
"logps/chosen": -63.9236946105957,
|
| 223 |
-
"logps/rejected": -79.44518280029297,
|
| 224 |
-
"loss": 0.6827,
|
| 225 |
-
"pred_label": 34.0,
|
| 226 |
-
"rewards/accuracies": 0.30000001192092896,
|
| 227 |
-
"rewards/chosen": -0.02154584601521492,
|
| 228 |
-
"rewards/margins": 0.04528125748038292,
|
| 229 |
-
"rewards/rejected": -0.06682710349559784,
|
| 230 |
-
"step": 110,
|
| 231 |
-
"use_label": 2152.0
|
| 232 |
-
},
|
| 233 |
{
|
| 234 |
"epoch": 0.13,
|
| 235 |
-
"grad_norm":
|
| 236 |
"learning_rate": 4.9903757462135984e-06,
|
| 237 |
-
"logits/chosen": -2.
|
| 238 |
-
"logits/rejected": -2.
|
| 239 |
-
"logps/chosen": -
|
| 240 |
-
"logps/rejected": -
|
| 241 |
-
"loss": 0.
|
| 242 |
-
"pred_label":
|
| 243 |
-
"rewards/accuracies": 0.
|
| 244 |
-
"rewards/chosen": -0.
|
| 245 |
-
"rewards/margins": 0.
|
| 246 |
-
"rewards/rejected": -0.
|
| 247 |
"step": 120,
|
| 248 |
-
"use_label":
|
| 249 |
-
},
|
| 250 |
-
{
|
| 251 |
-
"epoch": 0.14,
|
| 252 |
-
"grad_norm": 0.546875,
|
| 253 |
-
"learning_rate": 4.980697142834315e-06,
|
| 254 |
-
"logits/chosen": -2.0968613624572754,
|
| 255 |
-
"logits/rejected": -2.1124091148376465,
|
| 256 |
-
"logps/chosen": -66.370849609375,
|
| 257 |
-
"logps/rejected": -77.3319320678711,
|
| 258 |
-
"loss": 0.6845,
|
| 259 |
-
"pred_label": 57.57500076293945,
|
| 260 |
-
"rewards/accuracies": 0.2750000059604645,
|
| 261 |
-
"rewards/chosen": -0.07896758615970612,
|
| 262 |
-
"rewards/margins": 0.04609644412994385,
|
| 263 |
-
"rewards/rejected": -0.12506404519081116,
|
| 264 |
-
"step": 130,
|
| 265 |
-
"use_label": 2448.425048828125
|
| 266 |
},
|
| 267 |
{
|
| 268 |
"epoch": 0.15,
|
| 269 |
-
"grad_norm": 0.
|
| 270 |
"learning_rate": 4.967700826904229e-06,
|
| 271 |
-
"logits/chosen": -2.
|
| 272 |
-
"logits/rejected": -2.
|
| 273 |
-
"logps/chosen": -
|
| 274 |
-
"logps/rejected": -
|
| 275 |
-
"loss": 0.
|
| 276 |
-
"pred_label":
|
| 277 |
-
"rewards/accuracies": 0.
|
| 278 |
-
"rewards/chosen": -0.
|
| 279 |
-
"rewards/margins": 0.
|
| 280 |
-
"rewards/rejected": -0.
|
| 281 |
"step": 140,
|
| 282 |
-
"use_label":
|
| 283 |
-
},
|
| 284 |
-
{
|
| 285 |
-
"epoch": 0.16,
|
| 286 |
-
"grad_norm": 1.1015625,
|
| 287 |
-
"learning_rate": 4.951404179843963e-06,
|
| 288 |
-
"logits/chosen": -2.1765952110290527,
|
| 289 |
-
"logits/rejected": -2.125175714492798,
|
| 290 |
-
"logps/chosen": -54.37804412841797,
|
| 291 |
-
"logps/rejected": -58.982269287109375,
|
| 292 |
-
"loss": 0.6809,
|
| 293 |
-
"pred_label": 91.3499984741211,
|
| 294 |
-
"rewards/accuracies": 0.30000001192092896,
|
| 295 |
-
"rewards/chosen": -0.06883221119642258,
|
| 296 |
-
"rewards/margins": 0.06803621351718903,
|
| 297 |
-
"rewards/rejected": -0.136868417263031,
|
| 298 |
-
"step": 150,
|
| 299 |
-
"use_label": 2734.64990234375
|
| 300 |
},
|
| 301 |
{
|
| 302 |
"epoch": 0.17,
|
| 303 |
-
"grad_norm":
|
| 304 |
"learning_rate": 4.931828996974498e-06,
|
| 305 |
-
"logits/chosen": -2.
|
| 306 |
-
"logits/rejected": -2.
|
| 307 |
-
"logps/chosen": -
|
| 308 |
-
"logps/rejected": -
|
| 309 |
-
"loss": 0.
|
| 310 |
-
"pred_label":
|
| 311 |
-
"rewards/accuracies": 0.
|
| 312 |
-
"rewards/chosen": -0.
|
| 313 |
-
"rewards/margins": 0.
|
| 314 |
-
"rewards/rejected": -0.
|
| 315 |
"step": 160,
|
| 316 |
-
"use_label":
|
| 317 |
-
},
|
| 318 |
-
{
|
| 319 |
-
"epoch": 0.18,
|
| 320 |
-
"grad_norm": 1.1875,
|
| 321 |
-
"learning_rate": 4.909001458367867e-06,
|
| 322 |
-
"logits/chosen": -2.1201233863830566,
|
| 323 |
-
"logits/rejected": -2.0822367668151855,
|
| 324 |
-
"logps/chosen": -75.75311279296875,
|
| 325 |
-
"logps/rejected": -87.55944061279297,
|
| 326 |
-
"loss": 0.6869,
|
| 327 |
-
"pred_label": 141.85000610351562,
|
| 328 |
-
"rewards/accuracies": 0.33125001192092896,
|
| 329 |
-
"rewards/chosen": -0.1179669052362442,
|
| 330 |
-
"rewards/margins": 0.09383226186037064,
|
| 331 |
-
"rewards/rejected": -0.21179917454719543,
|
| 332 |
-
"step": 170,
|
| 333 |
-
"use_label": 3004.14990234375
|
| 334 |
},
|
| 335 |
{
|
| 336 |
"epoch": 0.19,
|
| 337 |
-
"grad_norm":
|
| 338 |
"learning_rate": 4.882952093833628e-06,
|
| 339 |
-
"logits/chosen": -2.
|
| 340 |
-
"logits/rejected": -2.
|
| 341 |
-
"logps/chosen": -
|
| 342 |
-
"logps/rejected": -
|
| 343 |
-
"loss": 0.
|
| 344 |
-
"pred_label":
|
| 345 |
-
"rewards/accuracies": 0.
|
| 346 |
-
"rewards/chosen": -0.
|
| 347 |
-
"rewards/margins": 0.
|
| 348 |
-
"rewards/rejected": -0.
|
| 349 |
"step": 180,
|
| 350 |
-
"use_label":
|
| 351 |
-
},
|
| 352 |
-
{
|
| 353 |
-
"epoch": 0.2,
|
| 354 |
-
"grad_norm": 0.8515625,
|
| 355 |
-
"learning_rate": 4.853715742087947e-06,
|
| 356 |
-
"logits/chosen": -2.1533255577087402,
|
| 357 |
-
"logits/rejected": -2.104222297668457,
|
| 358 |
-
"logps/chosen": -87.3572998046875,
|
| 359 |
-
"logps/rejected": -91.95249938964844,
|
| 360 |
-
"loss": 0.6862,
|
| 361 |
-
"pred_label": 181.39999389648438,
|
| 362 |
-
"rewards/accuracies": 0.375,
|
| 363 |
-
"rewards/chosen": -0.13474301993846893,
|
| 364 |
-
"rewards/margins": 0.08988693356513977,
|
| 365 |
-
"rewards/rejected": -0.2246299535036087,
|
| 366 |
-
"step": 190,
|
| 367 |
-
"use_label": 3284.60009765625
|
| 368 |
},
|
| 369 |
{
|
| 370 |
"epoch": 0.21,
|
| 371 |
-
"grad_norm": 0.
|
| 372 |
"learning_rate": 4.821331504159906e-06,
|
| 373 |
-
"logits/chosen": -2.
|
| 374 |
-
"logits/rejected": -2.
|
| 375 |
-
"logps/chosen": -
|
| 376 |
-
"logps/rejected": -
|
| 377 |
-
"loss": 0.
|
| 378 |
-
"pred_label":
|
| 379 |
-
"rewards/accuracies": 0.
|
| 380 |
-
"rewards/chosen": -0.
|
| 381 |
-
"rewards/margins": 0.
|
| 382 |
-
"rewards/rejected": -0.
|
| 383 |
"step": 200,
|
| 384 |
-
"use_label":
|
| 385 |
},
|
| 386 |
{
|
| 387 |
"epoch": 0.21,
|
| 388 |
-
"eval_logits/chosen": -2.
|
| 389 |
-
"eval_logits/rejected": -
|
| 390 |
-
"eval_logps/chosen": -
|
| 391 |
-
"eval_logps/rejected": -
|
| 392 |
-
"eval_loss": 0.
|
| 393 |
-
"eval_pred_label":
|
| 394 |
-
"eval_rewards/accuracies": 0.
|
| 395 |
-
"eval_rewards/chosen": -0.
|
| 396 |
-
"eval_rewards/margins": 0.
|
| 397 |
-
"eval_rewards/rejected": -0.
|
| 398 |
-
"eval_runtime":
|
| 399 |
-
"eval_samples_per_second": 8.
|
| 400 |
-
"eval_steps_per_second": 0.
|
| 401 |
-
"eval_use_label":
|
| 402 |
"step": 200
|
| 403 |
},
|
| 404 |
-
{
|
| 405 |
-
"epoch": 0.22,
|
| 406 |
-
"grad_norm": 1.1484375,
|
| 407 |
-
"learning_rate": 4.7858426910973435e-06,
|
| 408 |
-
"logits/chosen": -2.1574149131774902,
|
| 409 |
-
"logits/rejected": -2.1307334899902344,
|
| 410 |
-
"logps/chosen": -77.64894104003906,
|
| 411 |
-
"logps/rejected": -89.26710510253906,
|
| 412 |
-
"loss": 0.6828,
|
| 413 |
-
"pred_label": 313.32501220703125,
|
| 414 |
-
"rewards/accuracies": 0.3687500059604645,
|
| 415 |
-
"rewards/chosen": -0.09638272225856781,
|
| 416 |
-
"rewards/margins": 0.12071452289819717,
|
| 417 |
-
"rewards/rejected": -0.2170972377061844,
|
| 418 |
-
"step": 210,
|
| 419 |
-
"use_label": 3976.675048828125
|
| 420 |
-
},
|
| 421 |
{
|
| 422 |
"epoch": 0.23,
|
| 423 |
-
"grad_norm":
|
| 424 |
"learning_rate": 4.747296766042161e-06,
|
| 425 |
-
"logits/chosen": -2.
|
| 426 |
-
"logits/rejected": -2.
|
| 427 |
-
"logps/chosen": -
|
| 428 |
-
"logps/rejected": -
|
| 429 |
-
"loss": 0.
|
| 430 |
-
"pred_label":
|
| 431 |
-
"rewards/accuracies": 0.
|
| 432 |
-
"rewards/chosen": -0.
|
| 433 |
-
"rewards/margins": 0.
|
| 434 |
-
"rewards/rejected": -0.
|
| 435 |
"step": 220,
|
| 436 |
-
"use_label":
|
| 437 |
-
},
|
| 438 |
-
{
|
| 439 |
-
"epoch": 0.24,
|
| 440 |
-
"grad_norm": 1.1484375,
|
| 441 |
-
"learning_rate": 4.705745280752586e-06,
|
| 442 |
-
"logits/chosen": -2.1437509059906006,
|
| 443 |
-
"logits/rejected": -2.084073781967163,
|
| 444 |
-
"logps/chosen": -90.86326599121094,
|
| 445 |
-
"logps/rejected": -96.72235870361328,
|
| 446 |
-
"loss": 0.6875,
|
| 447 |
-
"pred_label": 378.6000061035156,
|
| 448 |
-
"rewards/accuracies": 0.35624998807907104,
|
| 449 |
-
"rewards/chosen": -0.12124122679233551,
|
| 450 |
-
"rewards/margins": 0.11637073755264282,
|
| 451 |
-
"rewards/rejected": -0.23761197924613953,
|
| 452 |
-
"step": 230,
|
| 453 |
-
"use_label": 4231.39990234375
|
| 454 |
},
|
| 455 |
{
|
| 456 |
"epoch": 0.25,
|
| 457 |
-
"grad_norm": 0.
|
| 458 |
"learning_rate": 4.661243806657256e-06,
|
| 459 |
-
"logits/chosen": -2.
|
| 460 |
-
"logits/rejected": -2.
|
| 461 |
-
"logps/chosen": -
|
| 462 |
-
"logps/rejected": -
|
| 463 |
-
"loss": 0.
|
| 464 |
-
"pred_label":
|
| 465 |
-
"rewards/accuracies": 0.
|
| 466 |
-
"rewards/chosen": -0.
|
| 467 |
-
"rewards/margins": 0.
|
| 468 |
-
"rewards/rejected": -0.
|
| 469 |
"step": 240,
|
| 470 |
-
"use_label":
|
| 471 |
-
},
|
| 472 |
-
{
|
| 473 |
-
"epoch": 0.26,
|
| 474 |
-
"grad_norm": 0.890625,
|
| 475 |
-
"learning_rate": 4.613851860533367e-06,
|
| 476 |
-
"logits/chosen": -2.1595332622528076,
|
| 477 |
-
"logits/rejected": -2.183953285217285,
|
| 478 |
-
"logps/chosen": -71.86934661865234,
|
| 479 |
-
"logps/rejected": -80.0597152709961,
|
| 480 |
-
"loss": 0.6844,
|
| 481 |
-
"pred_label": 422.25,
|
| 482 |
-
"rewards/accuracies": 0.3125,
|
| 483 |
-
"rewards/chosen": -0.06741674989461899,
|
| 484 |
-
"rewards/margins": 0.08548234403133392,
|
| 485 |
-
"rewards/rejected": -0.1528991013765335,
|
| 486 |
-
"step": 250,
|
| 487 |
-
"use_label": 4507.75
|
| 488 |
},
|
| 489 |
{
|
| 490 |
"epoch": 0.27,
|
| 491 |
-
"grad_norm":
|
| 492 |
"learning_rate": 4.563632824908252e-06,
|
| 493 |
-
"logits/chosen": -2.
|
| 494 |
-
"logits/rejected": -2.
|
| 495 |
-
"logps/chosen": -
|
| 496 |
-
"logps/rejected": -
|
| 497 |
-
"loss": 0.
|
| 498 |
-
"pred_label":
|
| 499 |
-
"rewards/accuracies": 0.
|
| 500 |
-
"rewards/chosen": -0.
|
| 501 |
-
"rewards/margins": 0.
|
| 502 |
-
"rewards/rejected": -0.
|
| 503 |
"step": 260,
|
| 504 |
-
"use_label":
|
| 505 |
-
},
|
| 506 |
-
{
|
| 507 |
-
"epoch": 0.28,
|
| 508 |
-
"grad_norm": 1.0703125,
|
| 509 |
-
"learning_rate": 4.510653863290871e-06,
|
| 510 |
-
"logits/chosen": -2.1512458324432373,
|
| 511 |
-
"logits/rejected": -2.164412021636963,
|
| 512 |
-
"logps/chosen": -91.74055480957031,
|
| 513 |
-
"logps/rejected": -95.13731384277344,
|
| 514 |
-
"loss": 0.6883,
|
| 515 |
-
"pred_label": 470.04998779296875,
|
| 516 |
-
"rewards/accuracies": 0.3125,
|
| 517 |
-
"rewards/chosen": -0.16311386227607727,
|
| 518 |
-
"rewards/margins": 0.0933571308851242,
|
| 519 |
-
"rewards/rejected": -0.2564709782600403,
|
| 520 |
-
"step": 270,
|
| 521 |
-
"use_label": 4779.9501953125
|
| 522 |
},
|
| 523 |
{
|
| 524 |
"epoch": 0.29,
|
| 525 |
-
"grad_norm": 0.
|
| 526 |
"learning_rate": 4.454985830346574e-06,
|
| 527 |
-
"logits/chosen": -2.
|
| 528 |
-
"logits/rejected": -2.
|
| 529 |
-
"logps/chosen": -
|
| 530 |
-
"logps/rejected": -
|
| 531 |
-
"loss": 0.
|
| 532 |
-
"pred_label":
|
| 533 |
-
"rewards/accuracies": 0.
|
| 534 |
-
"rewards/chosen": -0.
|
| 535 |
-
"rewards/margins": 0.
|
| 536 |
-
"rewards/rejected": -0.
|
| 537 |
"step": 280,
|
| 538 |
-
"use_label":
|
| 539 |
-
},
|
| 540 |
-
{
|
| 541 |
-
"epoch": 0.3,
|
| 542 |
-
"grad_norm": 1.3125,
|
| 543 |
-
"learning_rate": 4.396703177135262e-06,
|
| 544 |
-
"logits/chosen": -1.9870249032974243,
|
| 545 |
-
"logits/rejected": -1.956434965133667,
|
| 546 |
-
"logps/chosen": -89.98160552978516,
|
| 547 |
-
"logps/rejected": -99.75212097167969,
|
| 548 |
-
"loss": 0.6905,
|
| 549 |
-
"pred_label": 527.0499877929688,
|
| 550 |
-
"rewards/accuracies": 0.36250001192092896,
|
| 551 |
-
"rewards/chosen": -0.13706301152706146,
|
| 552 |
-
"rewards/margins": 0.16557420790195465,
|
| 553 |
-
"rewards/rejected": -0.3026372492313385,
|
| 554 |
-
"step": 290,
|
| 555 |
-
"use_label": 5042.9501953125
|
| 556 |
},
|
| 557 |
{
|
| 558 |
"epoch": 0.31,
|
| 559 |
-
"grad_norm":
|
| 560 |
"learning_rate": 4.335883851539693e-06,
|
| 561 |
-
"logits/chosen": -
|
| 562 |
-
"logits/rejected": -
|
| 563 |
-
"logps/chosen": -
|
| 564 |
-
"logps/rejected": -
|
| 565 |
-
"loss": 0.
|
| 566 |
-
"pred_label":
|
| 567 |
-
"rewards/accuracies": 0.
|
| 568 |
-
"rewards/chosen": -0.
|
| 569 |
-
"rewards/margins": 0.
|
| 570 |
-
"rewards/rejected": -0.
|
| 571 |
"step": 300,
|
| 572 |
-
"use_label":
|
| 573 |
},
|
| 574 |
{
|
| 575 |
"epoch": 0.31,
|
| 576 |
-
"eval_logits/chosen": -
|
| 577 |
-
"eval_logits/rejected": -1.
|
| 578 |
-
"eval_logps/chosen": -
|
| 579 |
-
"eval_logps/rejected": -
|
| 580 |
-
"eval_loss": 0.
|
| 581 |
-
"eval_pred_label":
|
| 582 |
-
"eval_rewards/accuracies": 0.
|
| 583 |
-
"eval_rewards/chosen": -0.
|
| 584 |
-
"eval_rewards/margins": 0.
|
| 585 |
-
"eval_rewards/rejected": -0.
|
| 586 |
-
"eval_runtime":
|
| 587 |
-
"eval_samples_per_second": 8.
|
| 588 |
-
"eval_steps_per_second": 0.
|
| 589 |
-
"eval_use_label":
|
| 590 |
"step": 300
|
| 591 |
},
|
| 592 |
-
{
|
| 593 |
-
"epoch": 0.32,
|
| 594 |
-
"grad_norm": 1.5,
|
| 595 |
-
"learning_rate": 4.2726091940171055e-06,
|
| 596 |
-
"logits/chosen": -2.043640613555908,
|
| 597 |
-
"logits/rejected": -2.01674222946167,
|
| 598 |
-
"logps/chosen": -72.24534606933594,
|
| 599 |
-
"logps/rejected": -89.407470703125,
|
| 600 |
-
"loss": 0.6865,
|
| 601 |
-
"pred_label": 688.9500122070312,
|
| 602 |
-
"rewards/accuracies": 0.29374998807907104,
|
| 603 |
-
"rewards/chosen": -0.23255303502082825,
|
| 604 |
-
"rewards/margins": 0.06651856750249863,
|
| 605 |
-
"rewards/rejected": -0.29907160997390747,
|
| 606 |
-
"step": 310,
|
| 607 |
-
"use_label": 5705.0498046875
|
| 608 |
-
},
|
| 609 |
{
|
| 610 |
"epoch": 0.33,
|
| 611 |
-
"grad_norm":
|
| 612 |
"learning_rate": 4.206963828813555e-06,
|
| 613 |
-
"logits/chosen": -
|
| 614 |
-
"logits/rejected": -
|
| 615 |
-
"logps/chosen": -
|
| 616 |
-
"logps/rejected": -
|
| 617 |
-
"loss": 0.
|
| 618 |
-
"pred_label":
|
| 619 |
-
"rewards/accuracies": 0.
|
| 620 |
-
"rewards/chosen": -0.
|
| 621 |
-
"rewards/margins": 0.
|
| 622 |
-
"rewards/rejected": -0.
|
| 623 |
"step": 320,
|
| 624 |
-
"use_label":
|
| 625 |
-
},
|
| 626 |
-
{
|
| 627 |
-
"epoch": 0.35,
|
| 628 |
-
"grad_norm": 0.95703125,
|
| 629 |
-
"learning_rate": 4.139035550786495e-06,
|
| 630 |
-
"logits/chosen": -1.989506483078003,
|
| 631 |
-
"logits/rejected": -1.9580066204071045,
|
| 632 |
-
"logps/chosen": -73.50363159179688,
|
| 633 |
-
"logps/rejected": -87.75289154052734,
|
| 634 |
-
"loss": 0.683,
|
| 635 |
-
"pred_label": 754.4500122070312,
|
| 636 |
-
"rewards/accuracies": 0.36250001192092896,
|
| 637 |
-
"rewards/chosen": -0.1003209576010704,
|
| 638 |
-
"rewards/margins": 0.13466720283031464,
|
| 639 |
-
"rewards/rejected": -0.23498816788196564,
|
| 640 |
-
"step": 330,
|
| 641 |
-
"use_label": 5959.5498046875
|
| 642 |
},
|
| 643 |
{
|
| 644 |
"epoch": 0.36,
|
| 645 |
-
"grad_norm":
|
| 646 |
"learning_rate": 4.068915207986931e-06,
|
| 647 |
-
"logits/chosen": -2.
|
| 648 |
-
"logits/rejected": -
|
| 649 |
-
"logps/chosen": -
|
| 650 |
-
"logps/rejected": -
|
| 651 |
-
"loss": 0.
|
| 652 |
-
"pred_label":
|
| 653 |
"rewards/accuracies": 0.3375000059604645,
|
| 654 |
-
"rewards/chosen": -0.
|
| 655 |
-
"rewards/margins": 0.
|
| 656 |
-
"rewards/rejected": -0.
|
| 657 |
"step": 340,
|
| 658 |
-
"use_label":
|
| 659 |
-
},
|
| 660 |
-
{
|
| 661 |
-
"epoch": 0.37,
|
| 662 |
-
"grad_norm": 0.984375,
|
| 663 |
-
"learning_rate": 3.996696580158211e-06,
|
| 664 |
-
"logits/chosen": -2.0441341400146484,
|
| 665 |
-
"logits/rejected": -2.0229620933532715,
|
| 666 |
-
"logps/chosen": -73.9575424194336,
|
| 667 |
-
"logps/rejected": -86.34129333496094,
|
| 668 |
-
"loss": 0.6869,
|
| 669 |
-
"pred_label": 817.5250244140625,
|
| 670 |
-
"rewards/accuracies": 0.3125,
|
| 671 |
-
"rewards/chosen": -0.133123978972435,
|
| 672 |
-
"rewards/margins": 0.08419892936944962,
|
| 673 |
-
"rewards/rejected": -0.2173229157924652,
|
| 674 |
-
"step": 350,
|
| 675 |
-
"use_label": 6216.47509765625
|
| 676 |
},
|
| 677 |
{
|
| 678 |
"epoch": 0.38,
|
| 679 |
-
"grad_norm":
|
| 680 |
"learning_rate": 3.922476253313921e-06,
|
| 681 |
-
"logits/chosen": -2.
|
| 682 |
-
"logits/rejected": -2.
|
| 683 |
-
"logps/chosen": -
|
| 684 |
-
"logps/rejected": -
|
| 685 |
-
"loss": 0.
|
| 686 |
-
"pred_label":
|
| 687 |
-
"rewards/accuracies": 0.
|
| 688 |
-
"rewards/chosen": -0.
|
| 689 |
-
"rewards/margins": 0.
|
| 690 |
-
"rewards/rejected": -0.
|
| 691 |
"step": 360,
|
| 692 |
-
"use_label":
|
| 693 |
-
},
|
| 694 |
-
{
|
| 695 |
-
"epoch": 0.39,
|
| 696 |
-
"grad_norm": 0.75,
|
| 697 |
-
"learning_rate": 3.846353490562664e-06,
|
| 698 |
-
"logits/chosen": -2.076312780380249,
|
| 699 |
-
"logits/rejected": -1.9995708465576172,
|
| 700 |
-
"logps/chosen": -85.83981323242188,
|
| 701 |
-
"logps/rejected": -95.1656723022461,
|
| 702 |
-
"loss": 0.6844,
|
| 703 |
-
"pred_label": 880.4249877929688,
|
| 704 |
-
"rewards/accuracies": 0.3687500059604645,
|
| 705 |
-
"rewards/chosen": -0.11745607852935791,
|
| 706 |
-
"rewards/margins": 0.14055705070495605,
|
| 707 |
-
"rewards/rejected": -0.2580130994319916,
|
| 708 |
-
"step": 370,
|
| 709 |
-
"use_label": 6473.5751953125
|
| 710 |
},
|
| 711 |
{
|
| 712 |
"epoch": 0.4,
|
| 713 |
-
"grad_norm": 0.
|
| 714 |
"learning_rate": 3.768430099352445e-06,
|
| 715 |
-
"logits/chosen": -2.
|
| 716 |
-
"logits/rejected": -
|
| 717 |
-
"logps/chosen": -
|
| 718 |
-
"logps/rejected": -
|
| 719 |
-
"loss": 0.
|
| 720 |
-
"pred_label":
|
| 721 |
-
"rewards/accuracies": 0.
|
| 722 |
-
"rewards/chosen": -0.
|
| 723 |
-
"rewards/margins": 0.
|
| 724 |
-
"rewards/rejected": -0.
|
| 725 |
"step": 380,
|
| 726 |
-
"use_label":
|
| 727 |
-
},
|
| 728 |
-
{
|
| 729 |
-
"epoch": 0.41,
|
| 730 |
-
"grad_norm": 0.97265625,
|
| 731 |
-
"learning_rate": 3.6888102953122307e-06,
|
| 732 |
-
"logits/chosen": -1.9291635751724243,
|
| 733 |
-
"logits/rejected": -1.914608359336853,
|
| 734 |
-
"logps/chosen": -101.44157409667969,
|
| 735 |
-
"logps/rejected": -96.10136413574219,
|
| 736 |
-
"loss": 0.6878,
|
| 737 |
-
"pred_label": 952.8250122070312,
|
| 738 |
-
"rewards/accuracies": 0.3187499940395355,
|
| 739 |
-
"rewards/chosen": -0.1657881736755371,
|
| 740 |
-
"rewards/margins": 0.12364902347326279,
|
| 741 |
-
"rewards/rejected": -0.2894372344017029,
|
| 742 |
-
"step": 390,
|
| 743 |
-
"use_label": 6721.1748046875
|
| 744 |
},
|
| 745 |
{
|
| 746 |
"epoch": 0.42,
|
| 747 |
-
"grad_norm":
|
| 748 |
"learning_rate": 3.607600562872785e-06,
|
| 749 |
-
"logits/chosen": -
|
| 750 |
-
"logits/rejected": -
|
| 751 |
-
"logps/chosen": -
|
| 752 |
-
"logps/rejected": -
|
| 753 |
-
"loss": 0.
|
| 754 |
-
"pred_label":
|
| 755 |
-
"rewards/accuracies": 0.
|
| 756 |
-
"rewards/chosen": -0.
|
| 757 |
-
"rewards/margins": 0.
|
| 758 |
-
"rewards/rejected": -0.
|
| 759 |
"step": 400,
|
| 760 |
-
"use_label":
|
| 761 |
},
|
| 762 |
{
|
| 763 |
"epoch": 0.42,
|
| 764 |
-
"eval_logits/chosen": -
|
| 765 |
-
"eval_logits/rejected": -
|
| 766 |
-
"eval_logps/chosen": -
|
| 767 |
-
"eval_logps/rejected": -
|
| 768 |
-
"eval_loss": 0.
|
| 769 |
-
"eval_pred_label":
|
| 770 |
-
"eval_rewards/accuracies": 0.
|
| 771 |
-
"eval_rewards/chosen": -0.
|
| 772 |
-
"eval_rewards/margins": 0.
|
| 773 |
-
"eval_rewards/rejected": -0.
|
| 774 |
-
"eval_runtime":
|
| 775 |
-
"eval_samples_per_second": 8.
|
| 776 |
-
"eval_steps_per_second": 0.
|
| 777 |
-
"eval_use_label":
|
| 778 |
"step": 400
|
| 779 |
},
|
| 780 |
-
{
|
| 781 |
-
"epoch": 0.43,
|
| 782 |
-
"grad_norm": 1.28125,
|
| 783 |
-
"learning_rate": 3.5249095128531863e-06,
|
| 784 |
-
"logits/chosen": -1.289879560470581,
|
| 785 |
-
"logits/rejected": -1.4085474014282227,
|
| 786 |
-
"logps/chosen": -85.75054168701172,
|
| 787 |
-
"logps/rejected": -96.24283599853516,
|
| 788 |
-
"loss": 0.6874,
|
| 789 |
-
"pred_label": 1135.699951171875,
|
| 790 |
-
"rewards/accuracies": 0.38749998807907104,
|
| 791 |
-
"rewards/chosen": -0.21242520213127136,
|
| 792 |
-
"rewards/margins": 0.17107079923152924,
|
| 793 |
-
"rewards/rejected": -0.3834960162639618,
|
| 794 |
-
"step": 410,
|
| 795 |
-
"use_label": 7362.2998046875
|
| 796 |
-
},
|
| 797 |
{
|
| 798 |
"epoch": 0.44,
|
| 799 |
-
"grad_norm": 0.
|
| 800 |
"learning_rate": 3.4408477372034743e-06,
|
| 801 |
-
"logits/chosen": -
|
| 802 |
-
"logits/rejected": -
|
| 803 |
-
"logps/chosen": -
|
| 804 |
-
"logps/rejected": -
|
| 805 |
-
"loss": 0.
|
| 806 |
-
"pred_label":
|
| 807 |
-
"rewards/accuracies": 0.
|
| 808 |
-
"rewards/chosen": -0.
|
| 809 |
-
"rewards/margins": 0.
|
| 810 |
-
"rewards/rejected": -0.
|
| 811 |
"step": 420,
|
| 812 |
-
"use_label":
|
| 813 |
-
},
|
| 814 |
-
{
|
| 815 |
-
"epoch": 0.45,
|
| 816 |
-
"grad_norm": 1.1484375,
|
| 817 |
-
"learning_rate": 3.355527661097728e-06,
|
| 818 |
-
"logits/chosen": -1.3129976987838745,
|
| 819 |
-
"logits/rejected": -1.2275488376617432,
|
| 820 |
-
"logps/chosen": -106.88911437988281,
|
| 821 |
-
"logps/rejected": -112.3751449584961,
|
| 822 |
-
"loss": 0.6918,
|
| 823 |
-
"pred_label": 1207.9749755859375,
|
| 824 |
-
"rewards/accuracies": 0.3125,
|
| 825 |
-
"rewards/chosen": -0.3042059540748596,
|
| 826 |
-
"rewards/margins": 0.13597823679447174,
|
| 827 |
-
"rewards/rejected": -0.44018417596817017,
|
| 828 |
-
"step": 430,
|
| 829 |
-
"use_label": 7610.02490234375
|
| 830 |
},
|
| 831 |
{
|
| 832 |
"epoch": 0.46,
|
| 833 |
-
"grad_norm":
|
| 834 |
"learning_rate": 3.269063392575352e-06,
|
| 835 |
-
"logits/chosen": -
|
| 836 |
-
"logits/rejected": -
|
| 837 |
-
"logps/chosen": -
|
| 838 |
-
"logps/rejected": -
|
| 839 |
-
"loss": 0.
|
| 840 |
-
"pred_label":
|
| 841 |
"rewards/accuracies": 0.33125001192092896,
|
| 842 |
-
"rewards/chosen": -0.
|
| 843 |
-
"rewards/margins": 0.
|
| 844 |
-
"rewards/rejected": -0.
|
| 845 |
"step": 440,
|
| 846 |
-
"use_label":
|
| 847 |
-
},
|
| 848 |
-
{
|
| 849 |
-
"epoch": 0.47,
|
| 850 |
-
"grad_norm": 1.375,
|
| 851 |
-
"learning_rate": 3.181570569931697e-06,
|
| 852 |
-
"logits/chosen": -1.4389588832855225,
|
| 853 |
-
"logits/rejected": -1.5265202522277832,
|
| 854 |
-
"logps/chosen": -96.37947845458984,
|
| 855 |
-
"logps/rejected": -113.1718521118164,
|
| 856 |
-
"loss": 0.6951,
|
| 857 |
-
"pred_label": 1281.3499755859375,
|
| 858 |
-
"rewards/accuracies": 0.3375000059604645,
|
| 859 |
-
"rewards/chosen": -0.2355901300907135,
|
| 860 |
-
"rewards/margins": 0.13590970635414124,
|
| 861 |
-
"rewards/rejected": -0.37149983644485474,
|
| 862 |
-
"step": 450,
|
| 863 |
-
"use_label": 7856.64990234375
|
| 864 |
},
|
| 865 |
{
|
| 866 |
"epoch": 0.48,
|
| 867 |
-
"grad_norm":
|
| 868 |
"learning_rate": 3.09316620706208e-06,
|
| 869 |
-
"logits/chosen": -
|
| 870 |
-
"logits/rejected": -
|
| 871 |
-
"logps/chosen": -
|
| 872 |
-
"logps/rejected": -
|
| 873 |
-
"loss": 0.
|
| 874 |
-
"pred_label":
|
| 875 |
-
"rewards/accuracies": 0.
|
| 876 |
-
"rewards/chosen": -0.
|
| 877 |
-
"rewards/margins": 0.
|
| 878 |
-
"rewards/rejected": -0.
|
| 879 |
"step": 460,
|
| 880 |
-
"use_label":
|
| 881 |
-
},
|
| 882 |
-
{
|
| 883 |
-
"epoch": 0.49,
|
| 884 |
-
"grad_norm": 1.1015625,
|
| 885 |
-
"learning_rate": 3.0039685369660785e-06,
|
| 886 |
-
"logits/chosen": -1.175449252128601,
|
| 887 |
-
"logits/rejected": -1.0759943723678589,
|
| 888 |
-
"logps/chosen": -88.91249084472656,
|
| 889 |
-
"logps/rejected": -110.02799987792969,
|
| 890 |
-
"loss": 0.6873,
|
| 891 |
-
"pred_label": 1345.1500244140625,
|
| 892 |
-
"rewards/accuracies": 0.3687500059604645,
|
| 893 |
-
"rewards/chosen": -0.22000393271446228,
|
| 894 |
-
"rewards/margins": 0.1964809000492096,
|
| 895 |
-
"rewards/rejected": -0.4164848327636719,
|
| 896 |
-
"step": 470,
|
| 897 |
-
"use_label": 8112.85009765625
|
| 898 |
},
|
| 899 |
{
|
| 900 |
"epoch": 0.5,
|
| 901 |
-
"grad_norm":
|
| 902 |
"learning_rate": 2.91409685362137e-06,
|
| 903 |
-
"logits/chosen": -
|
| 904 |
-
"logits/rejected": -
|
| 905 |
-
"logps/chosen": -
|
| 906 |
-
"logps/rejected": -
|
| 907 |
-
"loss": 0.
|
| 908 |
-
"pred_label":
|
| 909 |
-
"rewards/accuracies": 0.
|
| 910 |
-
"rewards/chosen": -0.
|
| 911 |
-
"rewards/margins": 0.
|
| 912 |
-
"rewards/rejected": -0.
|
| 913 |
"step": 480,
|
| 914 |
-
"use_label":
|
| 915 |
-
},
|
| 916 |
-
{
|
| 917 |
-
"epoch": 0.51,
|
| 918 |
-
"grad_norm": 1.4375,
|
| 919 |
-
"learning_rate": 2.8236713524386085e-06,
|
| 920 |
-
"logits/chosen": -1.0729541778564453,
|
| 921 |
-
"logits/rejected": -0.9298813939094543,
|
| 922 |
-
"logps/chosen": -88.73147583007812,
|
| 923 |
-
"logps/rejected": -94.53245544433594,
|
| 924 |
-
"loss": 0.6921,
|
| 925 |
-
"pred_label": 1428.9000244140625,
|
| 926 |
-
"rewards/accuracies": 0.26875001192092896,
|
| 927 |
-
"rewards/chosen": -0.22107498347759247,
|
| 928 |
-
"rewards/margins": 0.12524999678134918,
|
| 929 |
-
"rewards/rejected": -0.34632498025894165,
|
| 930 |
-
"step": 490,
|
| 931 |
-
"use_label": 8349.099609375
|
| 932 |
},
|
| 933 |
{
|
| 934 |
"epoch": 0.52,
|
| 935 |
-
"grad_norm":
|
| 936 |
"learning_rate": 2.7328129695107205e-06,
|
| 937 |
-
"logits/chosen": -
|
| 938 |
-
"logits/rejected": -
|
| 939 |
-
"logps/chosen": -
|
| 940 |
-
"logps/rejected": -
|
| 941 |
-
"loss": 0.
|
| 942 |
-
"pred_label":
|
| 943 |
-
"rewards/accuracies": 0.
|
| 944 |
-
"rewards/chosen": -0.
|
| 945 |
-
"rewards/margins": 0.
|
| 946 |
-
"rewards/rejected": -0.
|
| 947 |
"step": 500,
|
| 948 |
-
"use_label":
|
| 949 |
},
|
| 950 |
{
|
| 951 |
"epoch": 0.52,
|
| 952 |
-
"eval_logits/chosen": -
|
| 953 |
-
"eval_logits/rejected": -
|
| 954 |
-
"eval_logps/chosen": -
|
| 955 |
-
"eval_logps/rejected": -
|
| 956 |
-
"eval_loss": 0.
|
| 957 |
-
"eval_pred_label":
|
| 958 |
-
"eval_rewards/accuracies": 0.
|
| 959 |
-
"eval_rewards/chosen": -0.
|
| 960 |
-
"eval_rewards/margins": 0.
|
| 961 |
-
"eval_rewards/rejected": -0.
|
| 962 |
-
"eval_runtime":
|
| 963 |
-
"eval_samples_per_second": 8.
|
| 964 |
-
"eval_steps_per_second": 0.
|
| 965 |
-
"eval_use_label":
|
| 966 |
"step": 500
|
| 967 |
},
|
| 968 |
-
{
|
| 969 |
-
"epoch": 0.53,
|
| 970 |
-
"grad_norm": 1.0078125,
|
| 971 |
-
"learning_rate": 2.641643219871597e-06,
|
| 972 |
-
"logits/chosen": -0.7708507776260376,
|
| 973 |
-
"logits/rejected": -0.882653534412384,
|
| 974 |
-
"logps/chosen": -90.50456237792969,
|
| 975 |
-
"logps/rejected": -116.84162902832031,
|
| 976 |
-
"loss": 0.686,
|
| 977 |
-
"pred_label": 1610.5999755859375,
|
| 978 |
-
"rewards/accuracies": 0.36250001192092896,
|
| 979 |
-
"rewards/chosen": -0.2625977396965027,
|
| 980 |
-
"rewards/margins": 0.20036396384239197,
|
| 981 |
-
"rewards/rejected": -0.4629616141319275,
|
| 982 |
-
"step": 510,
|
| 983 |
-
"use_label": 8991.400390625
|
| 984 |
-
},
|
| 985 |
{
|
| 986 |
"epoch": 0.54,
|
| 987 |
-
"grad_norm":
|
| 988 |
"learning_rate": 2.5502840349805074e-06,
|
| 989 |
-
"logits/chosen": -
|
| 990 |
-
"logits/rejected": -
|
| 991 |
-
"logps/chosen": -
|
| 992 |
-
"logps/rejected": -
|
| 993 |
-
"loss": 0.
|
| 994 |
-
"pred_label":
|
| 995 |
-
"rewards/accuracies": 0.
|
| 996 |
-
"rewards/chosen": -0.
|
| 997 |
-
"rewards/margins": 0.
|
| 998 |
-
"rewards/rejected": -0.
|
| 999 |
"step": 520,
|
| 1000 |
-
"use_label":
|
| 1001 |
-
},
|
| 1002 |
-
{
|
| 1003 |
-
"epoch": 0.55,
|
| 1004 |
-
"grad_norm": 1.3671875,
|
| 1005 |
-
"learning_rate": 2.4588575996495797e-06,
|
| 1006 |
-
"logits/chosen": -0.8304817080497742,
|
| 1007 |
-
"logits/rejected": -0.7847825288772583,
|
| 1008 |
-
"logps/chosen": -105.92545318603516,
|
| 1009 |
-
"logps/rejected": -117.15931701660156,
|
| 1010 |
-
"loss": 0.6895,
|
| 1011 |
-
"pred_label": 1692.175048828125,
|
| 1012 |
-
"rewards/accuracies": 0.36250001192092896,
|
| 1013 |
-
"rewards/chosen": -0.316447913646698,
|
| 1014 |
-
"rewards/margins": 0.17969803512096405,
|
| 1015 |
-
"rewards/rejected": -0.49614596366882324,
|
| 1016 |
-
"step": 530,
|
| 1017 |
-
"use_label": 9229.8251953125
|
| 1018 |
},
|
| 1019 |
{
|
| 1020 |
"epoch": 0.57,
|
| 1021 |
-
"grad_norm":
|
| 1022 |
"learning_rate": 2.367486188632446e-06,
|
| 1023 |
-
"logits/chosen": -
|
| 1024 |
-
"logits/rejected": -
|
| 1025 |
-
"logps/chosen": -
|
| 1026 |
-
"logps/rejected": -
|
| 1027 |
-
"loss": 0.
|
| 1028 |
-
"pred_label":
|
| 1029 |
-
"rewards/accuracies": 0.
|
| 1030 |
-
"rewards/chosen": -0.
|
| 1031 |
-
"rewards/margins": 0.
|
| 1032 |
-
"rewards/rejected": -0.
|
| 1033 |
"step": 540,
|
| 1034 |
-
"use_label":
|
| 1035 |
-
},
|
| 1036 |
-
{
|
| 1037 |
-
"epoch": 0.58,
|
| 1038 |
-
"grad_norm": 1.796875,
|
| 1039 |
-
"learning_rate": 2.276292003092593e-06,
|
| 1040 |
-
"logits/chosen": -0.7944391369819641,
|
| 1041 |
-
"logits/rejected": -0.7596977353096008,
|
| 1042 |
-
"logps/chosen": -107.38740539550781,
|
| 1043 |
-
"logps/rejected": -111.28292083740234,
|
| 1044 |
-
"loss": 0.6887,
|
| 1045 |
-
"pred_label": 1775.7249755859375,
|
| 1046 |
-
"rewards/accuracies": 0.3125,
|
| 1047 |
-
"rewards/chosen": -0.3932684063911438,
|
| 1048 |
-
"rewards/margins": 0.12325477600097656,
|
| 1049 |
-
"rewards/rejected": -0.5165232419967651,
|
| 1050 |
-
"step": 550,
|
| 1051 |
-
"use_label": 9466.275390625
|
| 1052 |
},
|
| 1053 |
{
|
| 1054 |
"epoch": 0.59,
|
| 1055 |
-
"grad_norm":
|
| 1056 |
"learning_rate": 2.1853970071701415e-06,
|
| 1057 |
-
"logits/chosen": -
|
| 1058 |
-
"logits/rejected": -
|
| 1059 |
-
"logps/chosen": -
|
| 1060 |
-
"logps/rejected": -
|
| 1061 |
-
"loss": 0.
|
| 1062 |
-
"pred_label":
|
| 1063 |
-
"rewards/accuracies": 0.
|
| 1064 |
-
"rewards/chosen": -0.
|
| 1065 |
-
"rewards/margins": 0.
|
| 1066 |
-
"rewards/rejected": -0.
|
| 1067 |
"step": 560,
|
| 1068 |
-
"use_label":
|
| 1069 |
-
},
|
| 1070 |
-
{
|
| 1071 |
-
"epoch": 0.6,
|
| 1072 |
-
"grad_norm": 2.125,
|
| 1073 |
-
"learning_rate": 2.0949227648656194e-06,
|
| 1074 |
-
"logits/chosen": -0.925454318523407,
|
| 1075 |
-
"logits/rejected": -0.849765956401825,
|
| 1076 |
-
"logps/chosen": -100.53346252441406,
|
| 1077 |
-
"logps/rejected": -131.70309448242188,
|
| 1078 |
-
"loss": 0.6872,
|
| 1079 |
-
"pred_label": 1852.2249755859375,
|
| 1080 |
-
"rewards/accuracies": 0.375,
|
| 1081 |
-
"rewards/chosen": -0.3393338620662689,
|
| 1082 |
-
"rewards/margins": 0.23398590087890625,
|
| 1083 |
-
"rewards/rejected": -0.5733197927474976,
|
| 1084 |
-
"step": 570,
|
| 1085 |
-
"use_label": 9709.775390625
|
| 1086 |
},
|
| 1087 |
{
|
| 1088 |
"epoch": 0.61,
|
| 1089 |
-
"grad_norm":
|
| 1090 |
"learning_rate": 2.00499027745888e-06,
|
| 1091 |
-
"logits/chosen": -
|
| 1092 |
-
"logits/rejected": -
|
| 1093 |
-
"logps/chosen": -
|
| 1094 |
-
"logps/rejected": -
|
| 1095 |
-
"loss": 0.
|
| 1096 |
-
"pred_label":
|
| 1097 |
-
"rewards/accuracies": 0.
|
| 1098 |
-
"rewards/chosen": -0.
|
| 1099 |
-
"rewards/margins": 0.
|
| 1100 |
-
"rewards/rejected": -0.
|
| 1101 |
"step": 580,
|
| 1102 |
-
"use_label":
|
| 1103 |
-
},
|
| 1104 |
-
{
|
| 1105 |
-
"epoch": 0.62,
|
| 1106 |
-
"grad_norm": 1.1171875,
|
| 1107 |
-
"learning_rate": 1.915719821680624e-06,
|
| 1108 |
-
"logits/chosen": -0.8080962300300598,
|
| 1109 |
-
"logits/rejected": -0.7905328869819641,
|
| 1110 |
-
"logps/chosen": -125.2184066772461,
|
| 1111 |
-
"logps/rejected": -148.79432678222656,
|
| 1112 |
-
"loss": 0.6891,
|
| 1113 |
-
"pred_label": 1939.25,
|
| 1114 |
-
"rewards/accuracies": 0.40625,
|
| 1115 |
-
"rewards/chosen": -0.4552985727787018,
|
| 1116 |
-
"rewards/margins": 0.22290782630443573,
|
| 1117 |
-
"rewards/rejected": -0.6782063245773315,
|
| 1118 |
-
"step": 590,
|
| 1119 |
-
"use_label": 9942.75
|
| 1120 |
},
|
| 1121 |
{
|
| 1122 |
"epoch": 0.63,
|
| 1123 |
-
"grad_norm":
|
| 1124 |
"learning_rate": 1.8272307888529276e-06,
|
| 1125 |
-
"logits/chosen": -
|
| 1126 |
-
"logits/rejected": -
|
| 1127 |
-
"logps/chosen": -
|
| 1128 |
-
"logps/rejected": -
|
| 1129 |
-
"loss": 0.
|
| 1130 |
-
"pred_label":
|
| 1131 |
-
"rewards/accuracies": 0.
|
| 1132 |
-
"rewards/chosen": -0.
|
| 1133 |
-
"rewards/margins": 0.
|
| 1134 |
-
"rewards/rejected": -0.
|
| 1135 |
"step": 600,
|
| 1136 |
-
"use_label":
|
| 1137 |
},
|
| 1138 |
{
|
| 1139 |
"epoch": 0.63,
|
| 1140 |
-
"eval_logits/chosen": -
|
| 1141 |
-
"eval_logits/rejected": -
|
| 1142 |
-
"eval_logps/chosen": -
|
| 1143 |
-
"eval_logps/rejected": -
|
| 1144 |
-
"eval_loss": 0.
|
| 1145 |
-
"eval_pred_label":
|
| 1146 |
-
"eval_rewards/accuracies": 0.
|
| 1147 |
-
"eval_rewards/chosen": -0.
|
| 1148 |
-
"eval_rewards/margins": 0.
|
| 1149 |
-
"eval_rewards/rejected": -0.
|
| 1150 |
-
"eval_runtime":
|
| 1151 |
-
"eval_samples_per_second": 8.
|
| 1152 |
"eval_steps_per_second": 0.254,
|
| 1153 |
-
"eval_use_label":
|
| 1154 |
"step": 600
|
| 1155 |
},
|
| 1156 |
-
{
|
| 1157 |
-
"epoch": 0.64,
|
| 1158 |
-
"grad_norm": 1.515625,
|
| 1159 |
-
"learning_rate": 1.739641525213929e-06,
|
| 1160 |
-
"logits/chosen": -0.572044312953949,
|
| 1161 |
-
"logits/rejected": -0.654716432094574,
|
| 1162 |
-
"logps/chosen": -95.46563720703125,
|
| 1163 |
-
"logps/rejected": -132.0639190673828,
|
| 1164 |
-
"loss": 0.6926,
|
| 1165 |
-
"pred_label": 2185.449951171875,
|
| 1166 |
-
"rewards/accuracies": 0.33125001192092896,
|
| 1167 |
-
"rewards/chosen": -0.3655874729156494,
|
| 1168 |
-
"rewards/margins": 0.21378450095653534,
|
| 1169 |
-
"rewards/rejected": -0.579371988773346,
|
| 1170 |
-
"step": 610,
|
| 1171 |
-
"use_label": 10520.5498046875
|
| 1172 |
-
},
|
| 1173 |
{
|
| 1174 |
"epoch": 0.65,
|
| 1175 |
-
"grad_norm":
|
| 1176 |
"learning_rate": 1.6530691736402317e-06,
|
| 1177 |
-
"logits/chosen": -
|
| 1178 |
-
"logits/rejected": -
|
| 1179 |
-
"logps/chosen": -
|
| 1180 |
-
"logps/rejected": -
|
| 1181 |
-
"loss": 0.
|
| 1182 |
-
"pred_label":
|
| 1183 |
-
"rewards/accuracies": 0.
|
| 1184 |
-
"rewards/chosen": -0.
|
| 1185 |
-
"rewards/margins": 0.
|
| 1186 |
-
"rewards/rejected": -0.
|
| 1187 |
"step": 620,
|
| 1188 |
-
"use_label":
|
| 1189 |
-
},
|
| 1190 |
-
{
|
| 1191 |
-
"epoch": 0.66,
|
| 1192 |
-
"grad_norm": 1.34375,
|
| 1193 |
-
"learning_rate": 1.5676295169786864e-06,
|
| 1194 |
-
"logits/chosen": -0.5626051425933838,
|
| 1195 |
-
"logits/rejected": -0.7373117208480835,
|
| 1196 |
-
"logps/chosen": -109.76419830322266,
|
| 1197 |
-
"logps/rejected": -132.89573669433594,
|
| 1198 |
-
"loss": 0.6861,
|
| 1199 |
-
"pred_label": 2268.074951171875,
|
| 1200 |
-
"rewards/accuracies": 0.35624998807907104,
|
| 1201 |
-
"rewards/chosen": -0.3673921525478363,
|
| 1202 |
-
"rewards/margins": 0.2162620723247528,
|
| 1203 |
-
"rewards/rejected": -0.5836542844772339,
|
| 1204 |
-
"step": 630,
|
| 1205 |
-
"use_label": 10757.9248046875
|
| 1206 |
},
|
| 1207 |
{
|
| 1208 |
"epoch": 0.67,
|
| 1209 |
-
"grad_norm":
|
| 1210 |
"learning_rate": 1.4834368231970922e-06,
|
| 1211 |
-
"logits/chosen": -
|
| 1212 |
-
"logits/rejected": -
|
| 1213 |
-
"logps/chosen": -
|
| 1214 |
-
"logps/rejected": -
|
| 1215 |
-
"loss": 0.
|
| 1216 |
-
"pred_label":
|
| 1217 |
-
"rewards/accuracies": 0.
|
| 1218 |
-
"rewards/chosen": -0.
|
| 1219 |
-
"rewards/margins": 0.
|
| 1220 |
-
"rewards/rejected": -0.
|
| 1221 |
"step": 640,
|
| 1222 |
-
"use_label":
|
| 1223 |
-
},
|
| 1224 |
-
{
|
| 1225 |
-
"epoch": 0.68,
|
| 1226 |
-
"grad_norm": 1.5,
|
| 1227 |
-
"learning_rate": 1.4006036925609245e-06,
|
| 1228 |
-
"logits/chosen": -0.7530516386032104,
|
| 1229 |
-
"logits/rejected": -0.39667490124702454,
|
| 1230 |
-
"logps/chosen": -117.97354888916016,
|
| 1231 |
-
"logps/rejected": -148.5204620361328,
|
| 1232 |
-
"loss": 0.6907,
|
| 1233 |
-
"pred_label": 2364.60009765625,
|
| 1234 |
-
"rewards/accuracies": 0.35624998807907104,
|
| 1235 |
-
"rewards/chosen": -0.4478411078453064,
|
| 1236 |
-
"rewards/margins": 0.25875502824783325,
|
| 1237 |
-
"rewards/rejected": -0.7065961956977844,
|
| 1238 |
-
"step": 650,
|
| 1239 |
-
"use_label": 10981.400390625
|
| 1240 |
},
|
| 1241 |
{
|
| 1242 |
"epoch": 0.69,
|
| 1243 |
-
"grad_norm":
|
| 1244 |
"learning_rate": 1.3192409070404582e-06,
|
| 1245 |
-
"logits/chosen": -
|
| 1246 |
-
"logits/rejected": -
|
| 1247 |
-
"logps/chosen": -
|
| 1248 |
-
"logps/rejected": -
|
| 1249 |
-
"loss": 0.
|
| 1250 |
-
"pred_label":
|
| 1251 |
-
"rewards/accuracies": 0.
|
| 1252 |
-
"rewards/chosen": -0.
|
| 1253 |
-
"rewards/margins": 0.
|
| 1254 |
-
"rewards/rejected": -0.
|
| 1255 |
"step": 660,
|
| 1256 |
-
"use_label":
|
| 1257 |
-
},
|
| 1258 |
-
{
|
| 1259 |
-
"epoch": 0.7,
|
| 1260 |
-
"grad_norm": 1.515625,
|
| 1261 |
-
"learning_rate": 1.2394572821496953e-06,
|
| 1262 |
-
"logits/chosen": -0.9564473032951355,
|
| 1263 |
-
"logits/rejected": -1.0122594833374023,
|
| 1264 |
-
"logps/chosen": -100.20994567871094,
|
| 1265 |
-
"logps/rejected": -121.32554626464844,
|
| 1266 |
-
"loss": 0.6935,
|
| 1267 |
-
"pred_label": 2446.14990234375,
|
| 1268 |
-
"rewards/accuracies": 0.34375,
|
| 1269 |
-
"rewards/chosen": -0.3450331687927246,
|
| 1270 |
-
"rewards/margins": 0.19006122648715973,
|
| 1271 |
-
"rewards/rejected": -0.5350943803787231,
|
| 1272 |
-
"step": 670,
|
| 1273 |
-
"use_label": 11219.849609375
|
| 1274 |
},
|
| 1275 |
{
|
| 1276 |
"epoch": 0.71,
|
| 1277 |
-
"grad_norm":
|
| 1278 |
"learning_rate": 1.1613595214152713e-06,
|
| 1279 |
-
"logits/chosen": -
|
| 1280 |
-
"logits/rejected": -
|
| 1281 |
-
"logps/chosen": -
|
| 1282 |
-
"logps/rejected": -
|
| 1283 |
-
"loss": 0.
|
| 1284 |
-
"pred_label":
|
| 1285 |
-
"rewards/accuracies": 0.
|
| 1286 |
-
"rewards/chosen": -0.
|
| 1287 |
-
"rewards/margins": 0.
|
| 1288 |
-
"rewards/rejected": -0.
|
| 1289 |
"step": 680,
|
| 1290 |
-
"use_label":
|
| 1291 |
-
},
|
| 1292 |
-
{
|
| 1293 |
-
"epoch": 0.72,
|
| 1294 |
-
"grad_norm": 1.578125,
|
| 1295 |
-
"learning_rate": 1.0850520736699362e-06,
|
| 1296 |
-
"logits/chosen": -0.6506579518318176,
|
| 1297 |
-
"logits/rejected": -0.7167869806289673,
|
| 1298 |
-
"logps/chosen": -144.53038024902344,
|
| 1299 |
-
"logps/rejected": -167.38192749023438,
|
| 1300 |
-
"loss": 0.6898,
|
| 1301 |
-
"pred_label": 2534.75,
|
| 1302 |
-
"rewards/accuracies": 0.39375001192092896,
|
| 1303 |
-
"rewards/chosen": -0.42825189232826233,
|
| 1304 |
-
"rewards/margins": 0.28569427132606506,
|
| 1305 |
-
"rewards/rejected": -0.7139460444450378,
|
| 1306 |
-
"step": 690,
|
| 1307 |
-
"use_label": 11451.25
|
| 1308 |
},
|
| 1309 |
{
|
| 1310 |
"epoch": 0.73,
|
| 1311 |
-
"grad_norm":
|
| 1312 |
"learning_rate": 1.0106369933615043e-06,
|
| 1313 |
-
"logits/chosen": -
|
| 1314 |
-
"logits/rejected": -
|
| 1315 |
-
"logps/chosen": -
|
| 1316 |
-
"logps/rejected": -
|
| 1317 |
-
"loss": 0.
|
| 1318 |
-
"pred_label":
|
| 1319 |
-
"rewards/accuracies": 0.
|
| 1320 |
-
"rewards/chosen": -0.
|
| 1321 |
-
"rewards/margins": 0.
|
| 1322 |
-
"rewards/rejected": -0.
|
| 1323 |
"step": 700,
|
| 1324 |
-
"use_label":
|
| 1325 |
},
|
| 1326 |
{
|
| 1327 |
"epoch": 0.73,
|
| 1328 |
-
"eval_logits/chosen": -
|
| 1329 |
-
"eval_logits/rejected": -
|
| 1330 |
-
"eval_logps/chosen": -
|
| 1331 |
-
"eval_logps/rejected": -
|
| 1332 |
-
"eval_loss": 0.
|
| 1333 |
-
"eval_pred_label":
|
| 1334 |
-
"eval_rewards/accuracies": 0.
|
| 1335 |
-
"eval_rewards/chosen": -0.
|
| 1336 |
-
"eval_rewards/margins": 0.
|
| 1337 |
-
"eval_rewards/rejected": -0.
|
| 1338 |
-
"eval_runtime":
|
| 1339 |
-
"eval_samples_per_second": 8.
|
| 1340 |
"eval_steps_per_second": 0.254,
|
| 1341 |
-
"eval_use_label":
|
| 1342 |
"step": 700
|
| 1343 |
},
|
| 1344 |
-
{
|
| 1345 |
-
"epoch": 0.74,
|
| 1346 |
-
"grad_norm": 1.03125,
|
| 1347 |
-
"learning_rate": 9.382138040640714e-07,
|
| 1348 |
-
"logits/chosen": -0.6519032716751099,
|
| 1349 |
-
"logits/rejected": -0.637380063533783,
|
| 1350 |
-
"logps/chosen": -102.23021697998047,
|
| 1351 |
-
"logps/rejected": -127.60137939453125,
|
| 1352 |
-
"loss": 0.6903,
|
| 1353 |
-
"pred_label": 2771.699951171875,
|
| 1354 |
-
"rewards/accuracies": 0.3499999940395355,
|
| 1355 |
-
"rewards/chosen": -0.3915974497795105,
|
| 1356 |
-
"rewards/margins": 0.21561889350414276,
|
| 1357 |
-
"rewards/rejected": -0.6072162985801697,
|
| 1358 |
-
"step": 710,
|
| 1359 |
-
"use_label": 12038.2998046875
|
| 1360 |
-
},
|
| 1361 |
{
|
| 1362 |
"epoch": 0.75,
|
| 1363 |
-
"grad_norm":
|
| 1364 |
"learning_rate": 8.678793653740633e-07,
|
| 1365 |
-
"logits/chosen": -
|
| 1366 |
-
"logits/rejected": -
|
| 1367 |
-
"logps/chosen": -
|
| 1368 |
-
"logps/rejected": -
|
| 1369 |
-
"loss": 0.
|
| 1370 |
-
"pred_label":
|
| 1371 |
-
"rewards/accuracies": 0.
|
| 1372 |
-
"rewards/chosen": -0.
|
| 1373 |
-
"rewards/margins": 0.
|
| 1374 |
-
"rewards/rejected": -0.
|
| 1375 |
"step": 720,
|
| 1376 |
-
"use_label":
|
| 1377 |
-
},
|
| 1378 |
-
{
|
| 1379 |
-
"epoch": 0.76,
|
| 1380 |
-
"grad_norm": 2.21875,
|
| 1381 |
-
"learning_rate": 7.997277433690984e-07,
|
| 1382 |
-
"logits/chosen": -0.6035222411155701,
|
| 1383 |
-
"logits/rejected": -0.65208500623703,
|
| 1384 |
-
"logps/chosen": -100.17440032958984,
|
| 1385 |
-
"logps/rejected": -119.87808990478516,
|
| 1386 |
-
"loss": 0.6865,
|
| 1387 |
-
"pred_label": 2850.0,
|
| 1388 |
-
"rewards/accuracies": 0.35624998807907104,
|
| 1389 |
-
"rewards/chosen": -0.2982019782066345,
|
| 1390 |
-
"rewards/margins": 0.2585477828979492,
|
| 1391 |
-
"rewards/rejected": -0.5567497611045837,
|
| 1392 |
-
"step": 730,
|
| 1393 |
-
"use_label": 12280.0
|
| 1394 |
},
|
| 1395 |
{
|
| 1396 |
"epoch": 0.77,
|
| 1397 |
-
"grad_norm": 0.
|
| 1398 |
"learning_rate": 7.338500848029603e-07,
|
| 1399 |
-
"logits/chosen": -
|
| 1400 |
-
"logits/rejected": -
|
| 1401 |
-
"logps/chosen": -
|
| 1402 |
-
"logps/rejected": -
|
| 1403 |
-
"loss": 0.
|
| 1404 |
-
"pred_label":
|
| 1405 |
-
"rewards/accuracies": 0.
|
| 1406 |
-
"rewards/chosen": -0.
|
| 1407 |
-
"rewards/margins": 0.
|
| 1408 |
-
"rewards/rejected": -0.
|
| 1409 |
"step": 740,
|
| 1410 |
-
"use_label":
|
| 1411 |
-
},
|
| 1412 |
-
{
|
| 1413 |
-
"epoch": 0.79,
|
| 1414 |
-
"grad_norm": 1.1015625,
|
| 1415 |
-
"learning_rate": 6.70334495204884e-07,
|
| 1416 |
-
"logits/chosen": -0.5357509851455688,
|
| 1417 |
-
"logits/rejected": -0.594279408454895,
|
| 1418 |
-
"logps/chosen": -119.76139831542969,
|
| 1419 |
-
"logps/rejected": -145.1709747314453,
|
| 1420 |
-
"loss": 0.6905,
|
| 1421 |
-
"pred_label": 2929.22509765625,
|
| 1422 |
-
"rewards/accuracies": 0.35624998807907104,
|
| 1423 |
-
"rewards/chosen": -0.4223107397556305,
|
| 1424 |
-
"rewards/margins": 0.18705633282661438,
|
| 1425 |
-
"rewards/rejected": -0.6093670725822449,
|
| 1426 |
-
"step": 750,
|
| 1427 |
-
"use_label": 12520.775390625
|
| 1428 |
},
|
| 1429 |
{
|
| 1430 |
"epoch": 0.8,
|
| 1431 |
-
"grad_norm":
|
| 1432 |
"learning_rate": 6.092659210462232e-07,
|
| 1433 |
-
"logits/chosen": -
|
| 1434 |
-
"logits/rejected": -
|
| 1435 |
-
"logps/chosen": -
|
| 1436 |
-
"logps/rejected": -
|
| 1437 |
-
"loss": 0.
|
| 1438 |
-
"pred_label":
|
| 1439 |
-
"rewards/accuracies": 0.
|
| 1440 |
-
"rewards/chosen": -0.
|
| 1441 |
-
"rewards/margins": 0.
|
| 1442 |
-
"rewards/rejected": -0.
|
| 1443 |
"step": 760,
|
| 1444 |
-
"use_label":
|
| 1445 |
-
},
|
| 1446 |
-
{
|
| 1447 |
-
"epoch": 0.81,
|
| 1448 |
-
"grad_norm": 1.4375,
|
| 1449 |
-
"learning_rate": 5.507260361320738e-07,
|
| 1450 |
-
"logits/chosen": -0.6238114833831787,
|
| 1451 |
-
"logits/rejected": -0.6686199307441711,
|
| 1452 |
-
"logps/chosen": -127.0525131225586,
|
| 1453 |
-
"logps/rejected": -142.44747924804688,
|
| 1454 |
-
"loss": 0.689,
|
| 1455 |
-
"pred_label": 3021.85009765625,
|
| 1456 |
-
"rewards/accuracies": 0.39375001192092896,
|
| 1457 |
-
"rewards/chosen": -0.43505221605300903,
|
| 1458 |
-
"rewards/margins": 0.25210094451904297,
|
| 1459 |
-
"rewards/rejected": -0.687153160572052,
|
| 1460 |
-
"step": 770,
|
| 1461 |
-
"use_label": 12748.150390625
|
| 1462 |
},
|
| 1463 |
{
|
| 1464 |
"epoch": 0.82,
|
| 1465 |
-
"grad_norm":
|
| 1466 |
"learning_rate": 4.947931323697983e-07,
|
| 1467 |
-
"logits/chosen": -
|
| 1468 |
-
"logits/rejected": -
|
| 1469 |
-
"logps/chosen": -
|
| 1470 |
-
"logps/rejected": -
|
| 1471 |
-
"loss": 0.
|
| 1472 |
-
"pred_label":
|
| 1473 |
-
"rewards/accuracies": 0.
|
| 1474 |
-
"rewards/chosen": -0.
|
| 1475 |
-
"rewards/margins": 0.
|
| 1476 |
-
"rewards/rejected": -0.
|
| 1477 |
"step": 780,
|
| 1478 |
-
"use_label":
|
| 1479 |
-
},
|
| 1480 |
-
{
|
| 1481 |
-
"epoch": 0.83,
|
| 1482 |
-
"grad_norm": 1.421875,
|
| 1483 |
-
"learning_rate": 4.4154201506053985e-07,
|
| 1484 |
-
"logits/chosen": -0.5256940126419067,
|
| 1485 |
-
"logits/rejected": -0.467402845621109,
|
| 1486 |
-
"logps/chosen": -95.73258209228516,
|
| 1487 |
-
"logps/rejected": -103.3360366821289,
|
| 1488 |
-
"loss": 0.6917,
|
| 1489 |
-
"pred_label": 3123.85009765625,
|
| 1490 |
-
"rewards/accuracies": 0.32499998807907104,
|
| 1491 |
-
"rewards/chosen": -0.30898317694664,
|
| 1492 |
-
"rewards/margins": 0.2029590606689453,
|
| 1493 |
-
"rewards/rejected": -0.5119422674179077,
|
| 1494 |
-
"step": 790,
|
| 1495 |
-
"use_label": 12966.150390625
|
| 1496 |
},
|
| 1497 |
{
|
| 1498 |
"epoch": 0.84,
|
| 1499 |
-
"grad_norm":
|
| 1500 |
"learning_rate": 3.910439028537638e-07,
|
| 1501 |
-
"logits/chosen": -
|
| 1502 |
-
"logits/rejected": -
|
| 1503 |
-
"logps/chosen": -
|
| 1504 |
-
"logps/rejected": -
|
| 1505 |
-
"loss": 0.
|
| 1506 |
-
"pred_label":
|
| 1507 |
-
"rewards/accuracies": 0.
|
| 1508 |
-
"rewards/chosen": -0.
|
| 1509 |
-
"rewards/margins": 0.
|
| 1510 |
-
"rewards/rejected": -0.
|
| 1511 |
"step": 800,
|
| 1512 |
-
"use_label":
|
| 1513 |
},
|
| 1514 |
{
|
| 1515 |
"epoch": 0.84,
|
| 1516 |
-
"eval_logits/chosen": -
|
| 1517 |
-
"eval_logits/rejected": -
|
| 1518 |
-
"eval_logps/chosen": -
|
| 1519 |
-
"eval_logps/rejected": -
|
| 1520 |
-
"eval_loss": 0.
|
| 1521 |
-
"eval_pred_label":
|
| 1522 |
"eval_rewards/accuracies": 0.363095223903656,
|
| 1523 |
-
"eval_rewards/chosen": -0.
|
| 1524 |
-
"eval_rewards/margins": 0.
|
| 1525 |
-
"eval_rewards/rejected": -0.
|
| 1526 |
-
"eval_runtime":
|
| 1527 |
-
"eval_samples_per_second": 8.
|
| 1528 |
"eval_steps_per_second": 0.254,
|
| 1529 |
-
"eval_use_label":
|
| 1530 |
"step": 800
|
| 1531 |
},
|
| 1532 |
-
{
|
| 1533 |
-
"epoch": 0.85,
|
| 1534 |
-
"grad_norm": 1.3828125,
|
| 1535 |
-
"learning_rate": 3.4336633249862084e-07,
|
| 1536 |
-
"logits/chosen": -0.6630854606628418,
|
| 1537 |
-
"logits/rejected": -0.6445407867431641,
|
| 1538 |
-
"logps/chosen": -108.18148040771484,
|
| 1539 |
-
"logps/rejected": -135.99142456054688,
|
| 1540 |
-
"loss": 0.6901,
|
| 1541 |
-
"pred_label": 3350.35009765625,
|
| 1542 |
-
"rewards/accuracies": 0.34375,
|
| 1543 |
-
"rewards/chosen": -0.3832666873931885,
|
| 1544 |
-
"rewards/margins": 0.1908622682094574,
|
| 1545 |
-
"rewards/rejected": -0.5741289258003235,
|
| 1546 |
-
"step": 810,
|
| 1547 |
-
"use_label": 13563.650390625
|
| 1548 |
-
},
|
| 1549 |
{
|
| 1550 |
"epoch": 0.86,
|
| 1551 |
-
"grad_norm":
|
| 1552 |
"learning_rate": 2.98573068519539e-07,
|
| 1553 |
-
"logits/chosen": -
|
| 1554 |
-
"logits/rejected": -
|
| 1555 |
-
"logps/chosen": -
|
| 1556 |
-
"logps/rejected": -
|
| 1557 |
-
"loss": 0.
|
| 1558 |
-
"pred_label":
|
| 1559 |
-
"rewards/accuracies": 0.
|
| 1560 |
-
"rewards/chosen": -0.
|
| 1561 |
-
"rewards/margins": 0.
|
| 1562 |
-
"rewards/rejected": -0.
|
| 1563 |
"step": 820,
|
| 1564 |
-
"use_label":
|
| 1565 |
-
},
|
| 1566 |
-
{
|
| 1567 |
-
"epoch": 0.87,
|
| 1568 |
-
"grad_norm": 1.484375,
|
| 1569 |
-
"learning_rate": 2.5672401793681854e-07,
|
| 1570 |
-
"logits/chosen": -0.5476540923118591,
|
| 1571 |
-
"logits/rejected": -0.43125781416893005,
|
| 1572 |
-
"logps/chosen": -86.91058349609375,
|
| 1573 |
-
"logps/rejected": -110.5887222290039,
|
| 1574 |
-
"loss": 0.6923,
|
| 1575 |
-
"pred_label": 3435.074951171875,
|
| 1576 |
-
"rewards/accuracies": 0.36250001192092896,
|
| 1577 |
-
"rewards/chosen": -0.2886909246444702,
|
| 1578 |
-
"rewards/margins": 0.25071993470191956,
|
| 1579 |
-
"rewards/rejected": -0.5394108295440674,
|
| 1580 |
-
"step": 830,
|
| 1581 |
-
"use_label": 13798.9248046875
|
| 1582 |
},
|
| 1583 |
{
|
| 1584 |
"epoch": 0.88,
|
| 1585 |
-
"grad_norm":
|
| 1586 |
"learning_rate": 2.178751501463036e-07,
|
| 1587 |
-
"logits/chosen": -
|
| 1588 |
-
"logits/rejected": -
|
| 1589 |
-
"logps/chosen": -
|
| 1590 |
-
"logps/rejected": -
|
| 1591 |
"loss": 0.6915,
|
| 1592 |
-
"pred_label":
|
| 1593 |
-
"rewards/accuracies": 0.
|
| 1594 |
-
"rewards/chosen": -0.
|
| 1595 |
-
"rewards/margins": 0.
|
| 1596 |
-
"rewards/rejected": -0.
|
| 1597 |
"step": 840,
|
| 1598 |
-
"use_label":
|
| 1599 |
-
},
|
| 1600 |
-
{
|
| 1601 |
-
"epoch": 0.89,
|
| 1602 |
-
"grad_norm": 1.359375,
|
| 1603 |
-
"learning_rate": 1.820784220652766e-07,
|
| 1604 |
-
"logits/chosen": -0.6778563261032104,
|
| 1605 |
-
"logits/rejected": -0.73534095287323,
|
| 1606 |
-
"logps/chosen": -120.2663345336914,
|
| 1607 |
-
"logps/rejected": -149.02294921875,
|
| 1608 |
-
"loss": 0.6854,
|
| 1609 |
-
"pred_label": 3509.0,
|
| 1610 |
-
"rewards/accuracies": 0.41874998807907104,
|
| 1611 |
-
"rewards/chosen": -0.36049091815948486,
|
| 1612 |
-
"rewards/margins": 0.2984590530395508,
|
| 1613 |
-
"rewards/rejected": -0.6589499711990356,
|
| 1614 |
-
"step": 850,
|
| 1615 |
-
"use_label": 14045.0
|
| 1616 |
},
|
| 1617 |
{
|
| 1618 |
"epoch": 0.9,
|
| 1619 |
-
"grad_norm":
|
| 1620 |
"learning_rate": 1.4938170864468636e-07,
|
| 1621 |
-
"logits/chosen": -
|
| 1622 |
-
"logits/rejected": -
|
| 1623 |
-
"logps/chosen": -
|
| 1624 |
-
"logps/rejected": -
|
| 1625 |
-
"loss": 0.
|
| 1626 |
-
"pred_label":
|
| 1627 |
-
"rewards/accuracies": 0.
|
| 1628 |
-
"rewards/chosen": -0.
|
| 1629 |
-
"rewards/margins": 0.
|
| 1630 |
-
"rewards/rejected": -0.
|
| 1631 |
"step": 860,
|
| 1632 |
-
"use_label":
|
| 1633 |
-
},
|
| 1634 |
-
{
|
| 1635 |
-
"epoch": 0.91,
|
| 1636 |
-
"grad_norm": 1.7578125,
|
| 1637 |
-
"learning_rate": 1.1982873884064466e-07,
|
| 1638 |
-
"logits/chosen": -0.6633087992668152,
|
| 1639 |
-
"logits/rejected": -0.6678288578987122,
|
| 1640 |
-
"logps/chosen": -117.92154693603516,
|
| 1641 |
-
"logps/rejected": -145.3701171875,
|
| 1642 |
-
"loss": 0.6893,
|
| 1643 |
-
"pred_label": 3603.75,
|
| 1644 |
-
"rewards/accuracies": 0.375,
|
| 1645 |
-
"rewards/chosen": -0.3660942316055298,
|
| 1646 |
-
"rewards/margins": 0.2644110918045044,
|
| 1647 |
-
"rewards/rejected": -0.6305053234100342,
|
| 1648 |
-
"step": 870,
|
| 1649 |
-
"use_label": 14270.25
|
| 1650 |
},
|
| 1651 |
{
|
| 1652 |
"epoch": 0.92,
|
| 1653 |
-
"grad_norm": 0.
|
| 1654 |
"learning_rate": 9.345903713082305e-08,
|
| 1655 |
-
"logits/chosen": -
|
| 1656 |
-
"logits/rejected": -
|
| 1657 |
-
"logps/chosen": -
|
| 1658 |
-
"logps/rejected": -
|
| 1659 |
-
"loss": 0.
|
| 1660 |
-
"pred_label":
|
| 1661 |
-
"rewards/accuracies": 0.
|
| 1662 |
-
"rewards/chosen": -0.
|
| 1663 |
-
"rewards/margins": 0.
|
| 1664 |
-
"rewards/rejected": -0.
|
| 1665 |
"step": 880,
|
| 1666 |
-
"use_label":
|
| 1667 |
-
},
|
| 1668 |
-
{
|
| 1669 |
-
"epoch": 0.93,
|
| 1670 |
-
"grad_norm": 1.6484375,
|
| 1671 |
-
"learning_rate": 7.030787065396866e-08,
|
| 1672 |
-
"logits/chosen": -0.5159703493118286,
|
| 1673 |
-
"logits/rejected": -0.5519541501998901,
|
| 1674 |
-
"logps/chosen": -96.9026107788086,
|
| 1675 |
-
"logps/rejected": -120.7626724243164,
|
| 1676 |
-
"loss": 0.693,
|
| 1677 |
-
"pred_label": 3690.675048828125,
|
| 1678 |
-
"rewards/accuracies": 0.28125,
|
| 1679 |
-
"rewards/chosen": -0.3307461142539978,
|
| 1680 |
-
"rewards/margins": 0.1426464170217514,
|
| 1681 |
-
"rewards/rejected": -0.4733925461769104,
|
| 1682 |
-
"step": 890,
|
| 1683 |
-
"use_label": 14503.3251953125
|
| 1684 |
},
|
| 1685 |
{
|
| 1686 |
"epoch": 0.94,
|
| 1687 |
-
"grad_norm":
|
| 1688 |
"learning_rate": 5.0406202043228604e-08,
|
| 1689 |
-
"logits/chosen": -
|
| 1690 |
-
"logits/rejected": -
|
| 1691 |
-
"logps/chosen": -
|
| 1692 |
-
"logps/rejected": -
|
| 1693 |
-
"loss": 0.
|
| 1694 |
-
"pred_label":
|
| 1695 |
-
"rewards/accuracies": 0.
|
| 1696 |
-
"rewards/chosen": -0.
|
| 1697 |
-
"rewards/margins": 0.
|
| 1698 |
-
"rewards/rejected": -0.
|
| 1699 |
"step": 900,
|
| 1700 |
-
"use_label":
|
| 1701 |
},
|
| 1702 |
{
|
| 1703 |
"epoch": 0.94,
|
| 1704 |
-
"eval_logits/chosen": -
|
| 1705 |
-
"eval_logits/rejected": -
|
| 1706 |
-
"eval_logps/chosen": -
|
| 1707 |
-
"eval_logps/rejected": -
|
| 1708 |
-
"eval_loss": 0.
|
| 1709 |
-
"eval_pred_label":
|
| 1710 |
-
"eval_rewards/accuracies": 0.
|
| 1711 |
-
"eval_rewards/chosen": -0.
|
| 1712 |
-
"eval_rewards/margins": 0.
|
| 1713 |
-
"eval_rewards/rejected": -0.
|
| 1714 |
-
"eval_runtime":
|
| 1715 |
-
"eval_samples_per_second": 8.
|
| 1716 |
"eval_steps_per_second": 0.254,
|
| 1717 |
-
"eval_use_label":
|
| 1718 |
"step": 900
|
| 1719 |
},
|
| 1720 |
-
{
|
| 1721 |
-
"epoch": 0.95,
|
| 1722 |
-
"grad_norm": 1.171875,
|
| 1723 |
-
"learning_rate": 3.378064801637687e-08,
|
| 1724 |
-
"logits/chosen": -0.5370496511459351,
|
| 1725 |
-
"logits/rejected": -0.5028234720230103,
|
| 1726 |
-
"logps/chosen": -89.67744445800781,
|
| 1727 |
-
"logps/rejected": -113.96895599365234,
|
| 1728 |
-
"loss": 0.6882,
|
| 1729 |
-
"pred_label": 3916.52490234375,
|
| 1730 |
-
"rewards/accuracies": 0.3187499940395355,
|
| 1731 |
-
"rewards/chosen": -0.2901899218559265,
|
| 1732 |
-
"rewards/margins": 0.2133828103542328,
|
| 1733 |
-
"rewards/rejected": -0.5035727024078369,
|
| 1734 |
-
"step": 910,
|
| 1735 |
-
"use_label": 15101.474609375
|
| 1736 |
-
},
|
| 1737 |
{
|
| 1738 |
"epoch": 0.96,
|
| 1739 |
-
"grad_norm":
|
| 1740 |
"learning_rate": 2.0453443778310766e-08,
|
| 1741 |
-
"logits/chosen": -
|
| 1742 |
-
"logits/rejected": -
|
| 1743 |
-
"logps/chosen": -
|
| 1744 |
-
"logps/rejected": -
|
| 1745 |
-
"loss": 0.
|
| 1746 |
-
"pred_label":
|
| 1747 |
-
"rewards/accuracies": 0.
|
| 1748 |
-
"rewards/chosen": -0.
|
| 1749 |
-
"rewards/margins": 0.
|
| 1750 |
-
"rewards/rejected": -0.
|
| 1751 |
"step": 920,
|
| 1752 |
-
"use_label":
|
| 1753 |
-
},
|
| 1754 |
-
{
|
| 1755 |
-
"epoch": 0.97,
|
| 1756 |
-
"grad_norm": 2.109375,
|
| 1757 |
-
"learning_rate": 1.0442413283435759e-08,
|
| 1758 |
-
"logits/chosen": -0.4513850212097168,
|
| 1759 |
-
"logits/rejected": -0.5099025964736938,
|
| 1760 |
-
"logps/chosen": -92.44239807128906,
|
| 1761 |
-
"logps/rejected": -119.61177062988281,
|
| 1762 |
-
"loss": 0.6878,
|
| 1763 |
-
"pred_label": 3998.60009765625,
|
| 1764 |
-
"rewards/accuracies": 0.33125001192092896,
|
| 1765 |
-
"rewards/chosen": -0.29288578033447266,
|
| 1766 |
-
"rewards/margins": 0.20934204757213593,
|
| 1767 |
-
"rewards/rejected": -0.502227783203125,
|
| 1768 |
-
"step": 930,
|
| 1769 |
-
"use_label": 15339.400390625
|
| 1770 |
},
|
| 1771 |
{
|
| 1772 |
"epoch": 0.98,
|
| 1773 |
-
"grad_norm":
|
| 1774 |
"learning_rate": 3.760945397705828e-09,
|
| 1775 |
-
"logits/chosen": -
|
| 1776 |
-
"logits/rejected": -
|
| 1777 |
-
"logps/chosen": -
|
| 1778 |
-
"logps/rejected": -
|
| 1779 |
-
"loss": 0.
|
| 1780 |
-
"pred_label":
|
| 1781 |
-
"rewards/accuracies": 0.
|
| 1782 |
-
"rewards/chosen": -0.
|
| 1783 |
-
"rewards/margins": 0.
|
| 1784 |
-
"rewards/rejected": -0.
|
| 1785 |
"step": 940,
|
| 1786 |
-
"use_label":
|
| 1787 |
-
},
|
| 1788 |
-
{
|
| 1789 |
-
"epoch": 0.99,
|
| 1790 |
-
"grad_norm": 1.59375,
|
| 1791 |
-
"learning_rate": 4.1797599220405605e-10,
|
| 1792 |
-
"logits/chosen": -0.674268901348114,
|
| 1793 |
-
"logits/rejected": -0.7018919587135315,
|
| 1794 |
-
"logps/chosen": -114.91938781738281,
|
| 1795 |
-
"logps/rejected": -133.3175506591797,
|
| 1796 |
-
"loss": 0.6895,
|
| 1797 |
-
"pred_label": 4082.625,
|
| 1798 |
-
"rewards/accuracies": 0.33125001192092896,
|
| 1799 |
-
"rewards/chosen": -0.3830910325050354,
|
| 1800 |
-
"rewards/margins": 0.1591145098209381,
|
| 1801 |
-
"rewards/rejected": -0.5422054529190063,
|
| 1802 |
-
"step": 950,
|
| 1803 |
-
"use_label": 15575.375
|
| 1804 |
},
|
| 1805 |
{
|
| 1806 |
"epoch": 1.0,
|
| 1807 |
"step": 955,
|
| 1808 |
"total_flos": 0.0,
|
| 1809 |
-
"train_loss": 0.
|
| 1810 |
-
"train_runtime":
|
| 1811 |
"train_samples_per_second": 3.053,
|
| 1812 |
"train_steps_per_second": 0.048
|
| 1813 |
}
|
| 1814 |
],
|
| 1815 |
-
"logging_steps":
|
| 1816 |
"max_steps": 955,
|
| 1817 |
"num_input_tokens_seen": 0,
|
| 1818 |
"num_train_epochs": 1,
|
| 1819 |
-
"save_steps":
|
| 1820 |
"total_flos": 0.0,
|
| 1821 |
"train_batch_size": 4,
|
| 1822 |
"trial_name": null,
|
|
|
|
| 25 |
"step": 1,
|
| 26 |
"use_label": 10.0
|
| 27 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 28 |
{
|
| 29 |
"epoch": 0.02,
|
| 30 |
"grad_norm": 0.6796875,
|
| 31 |
"learning_rate": 1.0416666666666667e-06,
|
| 32 |
+
"logits/chosen": -2.2281553745269775,
|
| 33 |
+
"logits/rejected": -2.276446580886841,
|
| 34 |
+
"logps/chosen": -57.036190032958984,
|
| 35 |
+
"logps/rejected": -66.88007354736328,
|
| 36 |
+
"loss": 0.6927,
|
| 37 |
"pred_label": 0.0,
|
| 38 |
+
"rewards/accuracies": 0.24013157188892365,
|
| 39 |
+
"rewards/chosen": 0.003924594726413488,
|
| 40 |
+
"rewards/margins": 0.0009102027979679406,
|
| 41 |
+
"rewards/rejected": 0.0030143915209919214,
|
| 42 |
"step": 20,
|
| 43 |
+
"use_label": 170.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 44 |
},
|
| 45 |
{
|
| 46 |
"epoch": 0.04,
|
| 47 |
"grad_norm": 0.6328125,
|
| 48 |
"learning_rate": 2.0833333333333334e-06,
|
| 49 |
+
"logits/chosen": -2.2738099098205566,
|
| 50 |
+
"logits/rejected": -2.2623789310455322,
|
| 51 |
+
"logps/chosen": -54.78137969970703,
|
| 52 |
+
"logps/rejected": -67.2437515258789,
|
| 53 |
+
"loss": 0.6914,
|
| 54 |
"pred_label": 0.0,
|
| 55 |
+
"rewards/accuracies": 0.24687500298023224,
|
| 56 |
+
"rewards/chosen": 0.01747792772948742,
|
| 57 |
+
"rewards/margins": 0.001674558618105948,
|
| 58 |
+
"rewards/rejected": 0.015803368762135506,
|
| 59 |
"step": 40,
|
| 60 |
+
"use_label": 482.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 61 |
},
|
| 62 |
{
|
| 63 |
"epoch": 0.06,
|
| 64 |
+
"grad_norm": 0.71875,
|
| 65 |
"learning_rate": 3.125e-06,
|
| 66 |
+
"logits/chosen": -2.3237431049346924,
|
| 67 |
+
"logits/rejected": -2.321906089782715,
|
| 68 |
+
"logps/chosen": -75.5770034790039,
|
| 69 |
+
"logps/rejected": -87.68544006347656,
|
| 70 |
+
"loss": 0.6885,
|
| 71 |
"pred_label": 0.0,
|
| 72 |
+
"rewards/accuracies": 0.3125,
|
| 73 |
+
"rewards/chosen": 0.031676117330789566,
|
| 74 |
+
"rewards/margins": 0.009719676338136196,
|
| 75 |
+
"rewards/rejected": 0.021956440061330795,
|
| 76 |
"step": 60,
|
| 77 |
+
"use_label": 802.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 78 |
},
|
| 79 |
{
|
| 80 |
"epoch": 0.08,
|
| 81 |
+
"grad_norm": 0.73828125,
|
| 82 |
"learning_rate": 4.166666666666667e-06,
|
| 83 |
+
"logits/chosen": -2.2948005199432373,
|
| 84 |
+
"logits/rejected": -2.2623462677001953,
|
| 85 |
+
"logps/chosen": -79.29240417480469,
|
| 86 |
+
"logps/rejected": -83.04844665527344,
|
| 87 |
+
"loss": 0.6876,
|
| 88 |
+
"pred_label": 5.800000190734863,
|
| 89 |
+
"rewards/accuracies": 0.3343749940395355,
|
| 90 |
+
"rewards/chosen": 0.016009245067834854,
|
| 91 |
+
"rewards/margins": 0.018887853249907494,
|
| 92 |
+
"rewards/rejected": -0.0028786074835807085,
|
| 93 |
"step": 80,
|
| 94 |
+
"use_label": 1116.199951171875
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 95 |
},
|
| 96 |
{
|
| 97 |
"epoch": 0.1,
|
| 98 |
+
"grad_norm": 0.6953125,
|
| 99 |
"learning_rate": 4.9997324926814375e-06,
|
| 100 |
+
"logits/chosen": -2.2056884765625,
|
| 101 |
+
"logits/rejected": -2.210036039352417,
|
| 102 |
+
"logps/chosen": -68.87937927246094,
|
| 103 |
+
"logps/rejected": -77.87590026855469,
|
| 104 |
+
"loss": 0.6876,
|
| 105 |
+
"pred_label": 27.537500381469727,
|
| 106 |
+
"rewards/accuracies": 0.34062498807907104,
|
| 107 |
+
"rewards/chosen": -0.010471501387655735,
|
| 108 |
+
"rewards/margins": 0.03584115579724312,
|
| 109 |
+
"rewards/rejected": -0.04631265625357628,
|
| 110 |
"step": 100,
|
| 111 |
+
"use_label": 1414.4625244140625
|
| 112 |
},
|
| 113 |
{
|
| 114 |
"epoch": 0.1,
|
| 115 |
+
"eval_logits/chosen": -2.1076083183288574,
|
| 116 |
+
"eval_logits/rejected": -2.0761499404907227,
|
| 117 |
+
"eval_logps/chosen": -74.44951629638672,
|
| 118 |
+
"eval_logps/rejected": -85.2883071899414,
|
| 119 |
+
"eval_loss": 0.6895647048950195,
|
| 120 |
+
"eval_pred_label": 89.14286041259766,
|
| 121 |
+
"eval_rewards/accuracies": 0.335317462682724,
|
| 122 |
+
"eval_rewards/chosen": -0.05548960343003273,
|
| 123 |
+
"eval_rewards/margins": 0.04341282695531845,
|
| 124 |
+
"eval_rewards/rejected": -0.09890241920948029,
|
| 125 |
+
"eval_runtime": 247.5952,
|
| 126 |
+
"eval_samples_per_second": 8.078,
|
| 127 |
+
"eval_steps_per_second": 0.254,
|
| 128 |
+
"eval_use_label": 1766.857177734375,
|
| 129 |
"step": 100
|
| 130 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 131 |
{
|
| 132 |
"epoch": 0.13,
|
| 133 |
+
"grad_norm": 0.7578125,
|
| 134 |
"learning_rate": 4.9903757462135984e-06,
|
| 135 |
+
"logits/chosen": -2.2542896270751953,
|
| 136 |
+
"logits/rejected": -2.1902401447296143,
|
| 137 |
+
"logps/chosen": -70.2941665649414,
|
| 138 |
+
"logps/rejected": -84.7874755859375,
|
| 139 |
+
"loss": 0.6884,
|
| 140 |
+
"pred_label": 155.6374969482422,
|
| 141 |
+
"rewards/accuracies": 0.3187499940395355,
|
| 142 |
+
"rewards/chosen": -0.023759985342621803,
|
| 143 |
+
"rewards/margins": 0.051492441445589066,
|
| 144 |
+
"rewards/rejected": -0.07525241374969482,
|
| 145 |
"step": 120,
|
| 146 |
+
"use_label": 2110.362548828125
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 147 |
},
|
| 148 |
{
|
| 149 |
"epoch": 0.15,
|
| 150 |
+
"grad_norm": 0.55859375,
|
| 151 |
"learning_rate": 4.967700826904229e-06,
|
| 152 |
+
"logits/chosen": -2.1823272705078125,
|
| 153 |
+
"logits/rejected": -2.210157632827759,
|
| 154 |
+
"logps/chosen": -61.80498504638672,
|
| 155 |
+
"logps/rejected": -76.43424224853516,
|
| 156 |
+
"loss": 0.6907,
|
| 157 |
+
"pred_label": 204.22500610351562,
|
| 158 |
+
"rewards/accuracies": 0.26875001192092896,
|
| 159 |
+
"rewards/chosen": -0.029314354062080383,
|
| 160 |
+
"rewards/margins": 0.036702848970890045,
|
| 161 |
+
"rewards/rejected": -0.06601719558238983,
|
| 162 |
"step": 140,
|
| 163 |
+
"use_label": 2381.77490234375
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 164 |
},
|
| 165 |
{
|
| 166 |
"epoch": 0.17,
|
| 167 |
+
"grad_norm": 0.70703125,
|
| 168 |
"learning_rate": 4.931828996974498e-06,
|
| 169 |
+
"logits/chosen": -2.251568555831909,
|
| 170 |
+
"logits/rejected": -2.220432996749878,
|
| 171 |
+
"logps/chosen": -66.60148620605469,
|
| 172 |
+
"logps/rejected": -71.53702545166016,
|
| 173 |
+
"loss": 0.69,
|
| 174 |
+
"pred_label": 257.2124938964844,
|
| 175 |
+
"rewards/accuracies": 0.3343749940395355,
|
| 176 |
+
"rewards/chosen": -0.020524730905890465,
|
| 177 |
+
"rewards/margins": 0.05932433158159256,
|
| 178 |
+
"rewards/rejected": -0.07984906435012817,
|
| 179 |
"step": 160,
|
| 180 |
+
"use_label": 2648.78759765625
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 181 |
},
|
| 182 |
{
|
| 183 |
"epoch": 0.19,
|
| 184 |
+
"grad_norm": 0.6796875,
|
| 185 |
"learning_rate": 4.882952093833628e-06,
|
| 186 |
+
"logits/chosen": -2.114015817642212,
|
| 187 |
+
"logits/rejected": -2.126950740814209,
|
| 188 |
+
"logps/chosen": -66.40071868896484,
|
| 189 |
+
"logps/rejected": -78.54503631591797,
|
| 190 |
+
"loss": 0.6901,
|
| 191 |
+
"pred_label": 319.9624938964844,
|
| 192 |
+
"rewards/accuracies": 0.328125,
|
| 193 |
+
"rewards/chosen": -0.03171534463763237,
|
| 194 |
+
"rewards/margins": 0.0544399619102478,
|
| 195 |
+
"rewards/rejected": -0.08615531027317047,
|
| 196 |
"step": 180,
|
| 197 |
+
"use_label": 2906.03759765625
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 198 |
},
|
| 199 |
{
|
| 200 |
"epoch": 0.21,
|
| 201 |
+
"grad_norm": 0.9140625,
|
| 202 |
"learning_rate": 4.821331504159906e-06,
|
| 203 |
+
"logits/chosen": -2.138213872909546,
|
| 204 |
+
"logits/rejected": -2.108750343322754,
|
| 205 |
+
"logps/chosen": -77.92289733886719,
|
| 206 |
+
"logps/rejected": -78.32075500488281,
|
| 207 |
+
"loss": 0.6892,
|
| 208 |
+
"pred_label": 383.5249938964844,
|
| 209 |
+
"rewards/accuracies": 0.37812501192092896,
|
| 210 |
+
"rewards/chosen": -0.009543296881020069,
|
| 211 |
+
"rewards/margins": 0.06037301942706108,
|
| 212 |
+
"rewards/rejected": -0.06991632282733917,
|
| 213 |
"step": 200,
|
| 214 |
+
"use_label": 3162.47509765625
|
| 215 |
},
|
| 216 |
{
|
| 217 |
"epoch": 0.21,
|
| 218 |
+
"eval_logits/chosen": -2.051973581314087,
|
| 219 |
+
"eval_logits/rejected": -2.028658390045166,
|
| 220 |
+
"eval_logps/chosen": -69.3875503540039,
|
| 221 |
+
"eval_logps/rejected": -80.99542999267578,
|
| 222 |
+
"eval_loss": 0.6893584132194519,
|
| 223 |
+
"eval_pred_label": 459.1111145019531,
|
| 224 |
+
"eval_rewards/accuracies": 0.3492063581943512,
|
| 225 |
+
"eval_rewards/chosen": -0.0048699695616960526,
|
| 226 |
+
"eval_rewards/margins": 0.05110359564423561,
|
| 227 |
+
"eval_rewards/rejected": -0.05597356706857681,
|
| 228 |
+
"eval_runtime": 247.8689,
|
| 229 |
+
"eval_samples_per_second": 8.069,
|
| 230 |
+
"eval_steps_per_second": 0.254,
|
| 231 |
+
"eval_use_label": 3500.888916015625,
|
| 232 |
"step": 200
|
| 233 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 234 |
{
|
| 235 |
"epoch": 0.23,
|
| 236 |
+
"grad_norm": 0.765625,
|
| 237 |
"learning_rate": 4.747296766042161e-06,
|
| 238 |
+
"logits/chosen": -2.172316074371338,
|
| 239 |
+
"logits/rejected": -2.1599390506744385,
|
| 240 |
+
"logps/chosen": -73.75865173339844,
|
| 241 |
+
"logps/rejected": -76.45826721191406,
|
| 242 |
+
"loss": 0.6906,
|
| 243 |
+
"pred_label": 537.4000244140625,
|
| 244 |
+
"rewards/accuracies": 0.34375,
|
| 245 |
+
"rewards/chosen": -0.017265746369957924,
|
| 246 |
+
"rewards/margins": 0.061459798365831375,
|
| 247 |
+
"rewards/rejected": -0.07872554659843445,
|
| 248 |
"step": 220,
|
| 249 |
+
"use_label": 3832.60009765625
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 250 |
},
|
| 251 |
{
|
| 252 |
"epoch": 0.25,
|
| 253 |
+
"grad_norm": 0.671875,
|
| 254 |
"learning_rate": 4.661243806657256e-06,
|
| 255 |
+
"logits/chosen": -2.1377243995666504,
|
| 256 |
+
"logits/rejected": -2.114131450653076,
|
| 257 |
+
"logps/chosen": -78.08522033691406,
|
| 258 |
+
"logps/rejected": -88.16291809082031,
|
| 259 |
+
"loss": 0.6906,
|
| 260 |
+
"pred_label": 610.8624877929688,
|
| 261 |
+
"rewards/accuracies": 0.3375000059604645,
|
| 262 |
+
"rewards/chosen": -0.06858871877193451,
|
| 263 |
+
"rewards/margins": 0.07855252921581268,
|
| 264 |
+
"rewards/rejected": -0.1471412628889084,
|
| 265 |
"step": 240,
|
| 266 |
+
"use_label": 4079.137451171875
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 267 |
},
|
| 268 |
{
|
| 269 |
"epoch": 0.27,
|
| 270 |
+
"grad_norm": 0.70703125,
|
| 271 |
"learning_rate": 4.563632824908252e-06,
|
| 272 |
+
"logits/chosen": -2.1762757301330566,
|
| 273 |
+
"logits/rejected": -2.173243999481201,
|
| 274 |
+
"logps/chosen": -69.33678436279297,
|
| 275 |
+
"logps/rejected": -82.98787689208984,
|
| 276 |
+
"loss": 0.6907,
|
| 277 |
+
"pred_label": 682.2750244140625,
|
| 278 |
+
"rewards/accuracies": 0.33125001192092896,
|
| 279 |
+
"rewards/chosen": -0.06302420794963837,
|
| 280 |
+
"rewards/margins": 0.0732887014746666,
|
| 281 |
+
"rewards/rejected": -0.13631291687488556,
|
| 282 |
"step": 260,
|
| 283 |
+
"use_label": 4327.72509765625
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 284 |
},
|
| 285 |
{
|
| 286 |
"epoch": 0.29,
|
| 287 |
+
"grad_norm": 0.625,
|
| 288 |
"learning_rate": 4.454985830346574e-06,
|
| 289 |
+
"logits/chosen": -2.16465425491333,
|
| 290 |
+
"logits/rejected": -2.1788923740386963,
|
| 291 |
+
"logps/chosen": -74.41441345214844,
|
| 292 |
+
"logps/rejected": -78.55416870117188,
|
| 293 |
+
"loss": 0.6892,
|
| 294 |
+
"pred_label": 749.125,
|
| 295 |
+
"rewards/accuracies": 0.3062500059604645,
|
| 296 |
+
"rewards/chosen": -0.06083650514483452,
|
| 297 |
+
"rewards/margins": 0.04520425945520401,
|
| 298 |
+
"rewards/rejected": -0.10604077577590942,
|
| 299 |
"step": 280,
|
| 300 |
+
"use_label": 4580.875
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 301 |
},
|
| 302 |
{
|
| 303 |
"epoch": 0.31,
|
| 304 |
+
"grad_norm": 0.65234375,
|
| 305 |
"learning_rate": 4.335883851539693e-06,
|
| 306 |
+
"logits/chosen": -2.0553781986236572,
|
| 307 |
+
"logits/rejected": -2.0573229789733887,
|
| 308 |
+
"logps/chosen": -69.96788024902344,
|
| 309 |
+
"logps/rejected": -80.52223205566406,
|
| 310 |
+
"loss": 0.6904,
|
| 311 |
+
"pred_label": 824.5499877929688,
|
| 312 |
+
"rewards/accuracies": 0.359375,
|
| 313 |
+
"rewards/chosen": -0.04866168648004532,
|
| 314 |
+
"rewards/margins": 0.09801270812749863,
|
| 315 |
+
"rewards/rejected": -0.14667439460754395,
|
| 316 |
"step": 300,
|
| 317 |
+
"use_label": 4825.4501953125
|
| 318 |
},
|
| 319 |
{
|
| 320 |
"epoch": 0.31,
|
| 321 |
+
"eval_logits/chosen": -2.0163989067077637,
|
| 322 |
+
"eval_logits/rejected": -1.9942671060562134,
|
| 323 |
+
"eval_logps/chosen": -75.15243530273438,
|
| 324 |
+
"eval_logps/rejected": -89.50163269042969,
|
| 325 |
+
"eval_loss": 0.6908969879150391,
|
| 326 |
+
"eval_pred_label": 923.3174438476562,
|
| 327 |
+
"eval_rewards/accuracies": 0.3531745970249176,
|
| 328 |
+
"eval_rewards/chosen": -0.06251893937587738,
|
| 329 |
+
"eval_rewards/margins": 0.07851671427488327,
|
| 330 |
+
"eval_rewards/rejected": -0.14103564620018005,
|
| 331 |
+
"eval_runtime": 247.8241,
|
| 332 |
+
"eval_samples_per_second": 8.07,
|
| 333 |
+
"eval_steps_per_second": 0.254,
|
| 334 |
+
"eval_use_label": 5140.6826171875,
|
| 335 |
"step": 300
|
| 336 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 337 |
{
|
| 338 |
"epoch": 0.33,
|
| 339 |
+
"grad_norm": 0.9140625,
|
| 340 |
"learning_rate": 4.206963828813555e-06,
|
| 341 |
+
"logits/chosen": -2.065279483795166,
|
| 342 |
+
"logits/rejected": -2.0684821605682373,
|
| 343 |
+
"logps/chosen": -72.58639526367188,
|
| 344 |
+
"logps/rejected": -89.45655822753906,
|
| 345 |
+
"loss": 0.6899,
|
| 346 |
+
"pred_label": 1033.7874755859375,
|
| 347 |
+
"rewards/accuracies": 0.3125,
|
| 348 |
+
"rewards/chosen": -0.11120834201574326,
|
| 349 |
+
"rewards/margins": 0.0645986869931221,
|
| 350 |
+
"rewards/rejected": -0.17580702900886536,
|
| 351 |
"step": 320,
|
| 352 |
+
"use_label": 5440.21240234375
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 353 |
},
|
| 354 |
{
|
| 355 |
"epoch": 0.36,
|
| 356 |
+
"grad_norm": 0.56640625,
|
| 357 |
"learning_rate": 4.068915207986931e-06,
|
| 358 |
+
"logits/chosen": -2.033398151397705,
|
| 359 |
+
"logits/rejected": -1.991502046585083,
|
| 360 |
+
"logps/chosen": -71.1894760131836,
|
| 361 |
+
"logps/rejected": -84.0774154663086,
|
| 362 |
+
"loss": 0.6917,
|
| 363 |
+
"pred_label": 1122.112548828125,
|
| 364 |
"rewards/accuracies": 0.3375000059604645,
|
| 365 |
+
"rewards/chosen": -0.07950185984373093,
|
| 366 |
+
"rewards/margins": 0.08617939054965973,
|
| 367 |
+
"rewards/rejected": -0.16568127274513245,
|
| 368 |
"step": 340,
|
| 369 |
+
"use_label": 5671.8876953125
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 370 |
},
|
| 371 |
{
|
| 372 |
"epoch": 0.38,
|
| 373 |
+
"grad_norm": 0.84765625,
|
| 374 |
"learning_rate": 3.922476253313921e-06,
|
| 375 |
+
"logits/chosen": -2.0358688831329346,
|
| 376 |
+
"logits/rejected": -2.0224781036376953,
|
| 377 |
+
"logps/chosen": -76.57051849365234,
|
| 378 |
+
"logps/rejected": -84.2589340209961,
|
| 379 |
+
"loss": 0.6914,
|
| 380 |
+
"pred_label": 1204.4124755859375,
|
| 381 |
+
"rewards/accuracies": 0.31562501192092896,
|
| 382 |
+
"rewards/chosen": -0.11715561151504517,
|
| 383 |
+
"rewards/margins": 0.07723374664783478,
|
| 384 |
+
"rewards/rejected": -0.19438934326171875,
|
| 385 |
"step": 360,
|
| 386 |
+
"use_label": 5909.58740234375
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 387 |
},
|
| 388 |
{
|
| 389 |
"epoch": 0.4,
|
| 390 |
+
"grad_norm": 0.55078125,
|
| 391 |
"learning_rate": 3.768430099352445e-06,
|
| 392 |
+
"logits/chosen": -2.12782621383667,
|
| 393 |
+
"logits/rejected": -2.086026430130005,
|
| 394 |
+
"logps/chosen": -74.41622161865234,
|
| 395 |
+
"logps/rejected": -85.17180633544922,
|
| 396 |
+
"loss": 0.6918,
|
| 397 |
+
"pred_label": 1289.9375,
|
| 398 |
+
"rewards/accuracies": 0.3656249940395355,
|
| 399 |
+
"rewards/chosen": -0.07592298835515976,
|
| 400 |
+
"rewards/margins": 0.08457346260547638,
|
| 401 |
+
"rewards/rejected": -0.16049645841121674,
|
| 402 |
"step": 380,
|
| 403 |
+
"use_label": 6144.0625
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 404 |
},
|
| 405 |
{
|
| 406 |
"epoch": 0.42,
|
| 407 |
+
"grad_norm": 0.73046875,
|
| 408 |
"learning_rate": 3.607600562872785e-06,
|
| 409 |
+
"logits/chosen": -2.126784086227417,
|
| 410 |
+
"logits/rejected": -2.1261298656463623,
|
| 411 |
+
"logps/chosen": -83.82131958007812,
|
| 412 |
+
"logps/rejected": -86.00455474853516,
|
| 413 |
+
"loss": 0.6906,
|
| 414 |
+
"pred_label": 1373.137451171875,
|
| 415 |
+
"rewards/accuracies": 0.3375000059604645,
|
| 416 |
+
"rewards/chosen": -0.05874443054199219,
|
| 417 |
+
"rewards/margins": 0.06775099784135818,
|
| 418 |
+
"rewards/rejected": -0.12649545073509216,
|
| 419 |
"step": 400,
|
| 420 |
+
"use_label": 6380.8623046875
|
| 421 |
},
|
| 422 |
{
|
| 423 |
"epoch": 0.42,
|
| 424 |
+
"eval_logits/chosen": -2.0480618476867676,
|
| 425 |
+
"eval_logits/rejected": -2.0248324871063232,
|
| 426 |
+
"eval_logps/chosen": -75.26866149902344,
|
| 427 |
+
"eval_logps/rejected": -90.80635070800781,
|
| 428 |
+
"eval_loss": 0.6920759081840515,
|
| 429 |
+
"eval_pred_label": 1472.5714111328125,
|
| 430 |
+
"eval_rewards/accuracies": 0.3511904776096344,
|
| 431 |
+
"eval_rewards/chosen": -0.06368114054203033,
|
| 432 |
+
"eval_rewards/margins": 0.09040173143148422,
|
| 433 |
+
"eval_rewards/rejected": -0.15408287942409515,
|
| 434 |
+
"eval_runtime": 248.0088,
|
| 435 |
+
"eval_samples_per_second": 8.064,
|
| 436 |
+
"eval_steps_per_second": 0.254,
|
| 437 |
+
"eval_use_label": 6695.4287109375,
|
| 438 |
"step": 400
|
| 439 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 440 |
{
|
| 441 |
"epoch": 0.44,
|
| 442 |
+
"grad_norm": 0.78515625,
|
| 443 |
"learning_rate": 3.4408477372034743e-06,
|
| 444 |
+
"logits/chosen": -2.055358409881592,
|
| 445 |
+
"logits/rejected": -2.068175792694092,
|
| 446 |
+
"logps/chosen": -70.47552490234375,
|
| 447 |
+
"logps/rejected": -79.02010345458984,
|
| 448 |
+
"loss": 0.6903,
|
| 449 |
+
"pred_label": 1589.0374755859375,
|
| 450 |
+
"rewards/accuracies": 0.3656249940395355,
|
| 451 |
+
"rewards/chosen": -0.06399895995855331,
|
| 452 |
+
"rewards/margins": 0.0963120311498642,
|
| 453 |
+
"rewards/rejected": -0.16031098365783691,
|
| 454 |
"step": 420,
|
| 455 |
+
"use_label": 6988.96240234375
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 456 |
},
|
| 457 |
{
|
| 458 |
"epoch": 0.46,
|
| 459 |
+
"grad_norm": 0.95703125,
|
| 460 |
"learning_rate": 3.269063392575352e-06,
|
| 461 |
+
"logits/chosen": -2.0893940925598145,
|
| 462 |
+
"logits/rejected": -2.09212589263916,
|
| 463 |
+
"logps/chosen": -85.68560028076172,
|
| 464 |
+
"logps/rejected": -87.41291809082031,
|
| 465 |
+
"loss": 0.6912,
|
| 466 |
+
"pred_label": 1667.6875,
|
| 467 |
"rewards/accuracies": 0.33125001192092896,
|
| 468 |
+
"rewards/chosen": -0.13728377223014832,
|
| 469 |
+
"rewards/margins": 0.07875251770019531,
|
| 470 |
+
"rewards/rejected": -0.21603628993034363,
|
| 471 |
"step": 440,
|
| 472 |
+
"use_label": 7230.3125
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 473 |
},
|
| 474 |
{
|
| 475 |
"epoch": 0.48,
|
| 476 |
+
"grad_norm": 0.53515625,
|
| 477 |
"learning_rate": 3.09316620706208e-06,
|
| 478 |
+
"logits/chosen": -2.079465389251709,
|
| 479 |
+
"logits/rejected": -2.091001033782959,
|
| 480 |
+
"logps/chosen": -73.67254638671875,
|
| 481 |
+
"logps/rejected": -81.05415344238281,
|
| 482 |
+
"loss": 0.6916,
|
| 483 |
+
"pred_label": 1751.75,
|
| 484 |
+
"rewards/accuracies": 0.30000001192092896,
|
| 485 |
+
"rewards/chosen": -0.0876312330365181,
|
| 486 |
+
"rewards/margins": 0.08376732468605042,
|
| 487 |
+
"rewards/rejected": -0.17139855027198792,
|
| 488 |
"step": 460,
|
| 489 |
+
"use_label": 7466.25
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 490 |
},
|
| 491 |
{
|
| 492 |
"epoch": 0.5,
|
| 493 |
+
"grad_norm": 0.69921875,
|
| 494 |
"learning_rate": 2.91409685362137e-06,
|
| 495 |
+
"logits/chosen": -2.0379364490509033,
|
| 496 |
+
"logits/rejected": -2.0492634773254395,
|
| 497 |
+
"logps/chosen": -77.06828308105469,
|
| 498 |
+
"logps/rejected": -89.38865661621094,
|
| 499 |
+
"loss": 0.6912,
|
| 500 |
+
"pred_label": 1832.6500244140625,
|
| 501 |
+
"rewards/accuracies": 0.36250001192092896,
|
| 502 |
+
"rewards/chosen": -0.06041146069765091,
|
| 503 |
+
"rewards/margins": 0.10216375440359116,
|
| 504 |
+
"rewards/rejected": -0.16257521510124207,
|
| 505 |
"step": 480,
|
| 506 |
+
"use_label": 7705.35009765625
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 507 |
},
|
| 508 |
{
|
| 509 |
"epoch": 0.52,
|
| 510 |
+
"grad_norm": 0.86328125,
|
| 511 |
"learning_rate": 2.7328129695107205e-06,
|
| 512 |
+
"logits/chosen": -2.031346082687378,
|
| 513 |
+
"logits/rejected": -2.0272762775421143,
|
| 514 |
+
"logps/chosen": -79.55888366699219,
|
| 515 |
+
"logps/rejected": -84.47586822509766,
|
| 516 |
+
"loss": 0.6903,
|
| 517 |
+
"pred_label": 1919.5374755859375,
|
| 518 |
+
"rewards/accuracies": 0.36250001192092896,
|
| 519 |
+
"rewards/chosen": -0.08177755773067474,
|
| 520 |
+
"rewards/margins": 0.08017835766077042,
|
| 521 |
+
"rewards/rejected": -0.16195592284202576,
|
| 522 |
"step": 500,
|
| 523 |
+
"use_label": 7938.46240234375
|
| 524 |
},
|
| 525 |
{
|
| 526 |
"epoch": 0.52,
|
| 527 |
+
"eval_logits/chosen": -2.0070507526397705,
|
| 528 |
+
"eval_logits/rejected": -1.9800992012023926,
|
| 529 |
+
"eval_logps/chosen": -76.36968231201172,
|
| 530 |
+
"eval_logps/rejected": -92.65614318847656,
|
| 531 |
+
"eval_loss": 0.6914148926734924,
|
| 532 |
+
"eval_pred_label": 2025.793701171875,
|
| 533 |
+
"eval_rewards/accuracies": 0.3492063581943512,
|
| 534 |
+
"eval_rewards/chosen": -0.07469133287668228,
|
| 535 |
+
"eval_rewards/margins": 0.09788943827152252,
|
| 536 |
+
"eval_rewards/rejected": -0.1725807636976242,
|
| 537 |
+
"eval_runtime": 247.8554,
|
| 538 |
+
"eval_samples_per_second": 8.069,
|
| 539 |
+
"eval_steps_per_second": 0.254,
|
| 540 |
+
"eval_use_label": 8246.2060546875,
|
| 541 |
"step": 500
|
| 542 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 543 |
{
|
| 544 |
"epoch": 0.54,
|
| 545 |
+
"grad_norm": 0.78125,
|
| 546 |
"learning_rate": 2.5502840349805074e-06,
|
| 547 |
+
"logits/chosen": -2.026449203491211,
|
| 548 |
+
"logits/rejected": -2.0701510906219482,
|
| 549 |
+
"logps/chosen": -75.1209487915039,
|
| 550 |
+
"logps/rejected": -88.01356506347656,
|
| 551 |
+
"loss": 0.6913,
|
| 552 |
+
"pred_label": 2148.887451171875,
|
| 553 |
+
"rewards/accuracies": 0.3531250059604645,
|
| 554 |
+
"rewards/chosen": -0.06801941990852356,
|
| 555 |
+
"rewards/margins": 0.09691040217876434,
|
| 556 |
+
"rewards/rejected": -0.1649298369884491,
|
| 557 |
"step": 520,
|
| 558 |
+
"use_label": 8533.1123046875
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 559 |
},
|
| 560 |
{
|
| 561 |
"epoch": 0.57,
|
| 562 |
+
"grad_norm": 1.09375,
|
| 563 |
"learning_rate": 2.367486188632446e-06,
|
| 564 |
+
"logits/chosen": -2.0245327949523926,
|
| 565 |
+
"logits/rejected": -2.0479135513305664,
|
| 566 |
+
"logps/chosen": -84.60169219970703,
|
| 567 |
+
"logps/rejected": -90.6330795288086,
|
| 568 |
+
"loss": 0.692,
|
| 569 |
+
"pred_label": 2235.550048828125,
|
| 570 |
+
"rewards/accuracies": 0.359375,
|
| 571 |
+
"rewards/chosen": -0.09091995656490326,
|
| 572 |
+
"rewards/margins": 0.11123095452785492,
|
| 573 |
+
"rewards/rejected": -0.20215091109275818,
|
| 574 |
"step": 540,
|
| 575 |
+
"use_label": 8766.4501953125
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 576 |
},
|
| 577 |
{
|
| 578 |
"epoch": 0.59,
|
| 579 |
+
"grad_norm": 0.75390625,
|
| 580 |
"learning_rate": 2.1853970071701415e-06,
|
| 581 |
+
"logits/chosen": -2.0177600383758545,
|
| 582 |
+
"logits/rejected": -2.016798257827759,
|
| 583 |
+
"logps/chosen": -78.94650268554688,
|
| 584 |
+
"logps/rejected": -80.36412811279297,
|
| 585 |
+
"loss": 0.6917,
|
| 586 |
+
"pred_label": 2319.53759765625,
|
| 587 |
+
"rewards/accuracies": 0.2874999940395355,
|
| 588 |
+
"rewards/chosen": -0.10138510167598724,
|
| 589 |
+
"rewards/margins": 0.06911652535200119,
|
| 590 |
+
"rewards/rejected": -0.17050163447856903,
|
| 591 |
"step": 560,
|
| 592 |
+
"use_label": 9002.462890625
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 593 |
},
|
| 594 |
{
|
| 595 |
"epoch": 0.61,
|
| 596 |
+
"grad_norm": 0.71875,
|
| 597 |
"learning_rate": 2.00499027745888e-06,
|
| 598 |
+
"logits/chosen": -2.054065704345703,
|
| 599 |
+
"logits/rejected": -2.0555384159088135,
|
| 600 |
+
"logps/chosen": -80.3529281616211,
|
| 601 |
+
"logps/rejected": -95.12947082519531,
|
| 602 |
+
"loss": 0.6919,
|
| 603 |
+
"pred_label": 2401.675048828125,
|
| 604 |
+
"rewards/accuracies": 0.359375,
|
| 605 |
+
"rewards/chosen": -0.09597108513116837,
|
| 606 |
+
"rewards/margins": 0.09131233394145966,
|
| 607 |
+
"rewards/rejected": -0.18728342652320862,
|
| 608 |
"step": 580,
|
| 609 |
+
"use_label": 9240.3251953125
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 610 |
},
|
| 611 |
{
|
| 612 |
"epoch": 0.63,
|
| 613 |
+
"grad_norm": 0.76171875,
|
| 614 |
"learning_rate": 1.8272307888529276e-06,
|
| 615 |
+
"logits/chosen": -2.059126377105713,
|
| 616 |
+
"logits/rejected": -2.099806547164917,
|
| 617 |
+
"logps/chosen": -89.58797454833984,
|
| 618 |
+
"logps/rejected": -108.6166000366211,
|
| 619 |
+
"loss": 0.6903,
|
| 620 |
+
"pred_label": 2492.9375,
|
| 621 |
+
"rewards/accuracies": 0.41874998807907104,
|
| 622 |
+
"rewards/chosen": -0.12580521404743195,
|
| 623 |
+
"rewards/margins": 0.10241512209177017,
|
| 624 |
+
"rewards/rejected": -0.22822031378746033,
|
| 625 |
"step": 600,
|
| 626 |
+
"use_label": 9469.0625
|
| 627 |
},
|
| 628 |
{
|
| 629 |
"epoch": 0.63,
|
| 630 |
+
"eval_logits/chosen": -1.9870026111602783,
|
| 631 |
+
"eval_logits/rejected": -1.960112452507019,
|
| 632 |
+
"eval_logps/chosen": -78.95431518554688,
|
| 633 |
+
"eval_logps/rejected": -95.86695861816406,
|
| 634 |
+
"eval_loss": 0.6917396187782288,
|
| 635 |
+
"eval_pred_label": 2603.9365234375,
|
| 636 |
+
"eval_rewards/accuracies": 0.3551587164402008,
|
| 637 |
+
"eval_rewards/chosen": -0.1005377396941185,
|
| 638 |
+
"eval_rewards/margins": 0.104151152074337,
|
| 639 |
+
"eval_rewards/rejected": -0.2046888917684555,
|
| 640 |
+
"eval_runtime": 247.9642,
|
| 641 |
+
"eval_samples_per_second": 8.066,
|
| 642 |
"eval_steps_per_second": 0.254,
|
| 643 |
+
"eval_use_label": 9772.0634765625,
|
| 644 |
"step": 600
|
| 645 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 646 |
{
|
| 647 |
"epoch": 0.65,
|
| 648 |
+
"grad_norm": 0.5859375,
|
| 649 |
"learning_rate": 1.6530691736402317e-06,
|
| 650 |
+
"logits/chosen": -1.9752880334854126,
|
| 651 |
+
"logits/rejected": -2.011981964111328,
|
| 652 |
+
"logps/chosen": -69.71615600585938,
|
| 653 |
+
"logps/rejected": -95.88337707519531,
|
| 654 |
+
"loss": 0.6918,
|
| 655 |
+
"pred_label": 2726.324951171875,
|
| 656 |
+
"rewards/accuracies": 0.34687501192092896,
|
| 657 |
+
"rewards/chosen": -0.09408678859472275,
|
| 658 |
+
"rewards/margins": 0.09362435340881348,
|
| 659 |
+
"rewards/rejected": -0.18771114945411682,
|
| 660 |
"step": 620,
|
| 661 |
+
"use_label": 10059.6748046875
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 662 |
},
|
| 663 |
{
|
| 664 |
"epoch": 0.67,
|
| 665 |
+
"grad_norm": 0.73046875,
|
| 666 |
"learning_rate": 1.4834368231970922e-06,
|
| 667 |
+
"logits/chosen": -2.0288071632385254,
|
| 668 |
+
"logits/rejected": -2.0409998893737793,
|
| 669 |
+
"logps/chosen": -82.56907653808594,
|
| 670 |
+
"logps/rejected": -90.75765228271484,
|
| 671 |
+
"loss": 0.6894,
|
| 672 |
+
"pred_label": 2805.512451171875,
|
| 673 |
+
"rewards/accuracies": 0.36250001192092896,
|
| 674 |
+
"rewards/chosen": -0.10210500657558441,
|
| 675 |
+
"rewards/margins": 0.10695278644561768,
|
| 676 |
+
"rewards/rejected": -0.2090577781200409,
|
| 677 |
"step": 640,
|
| 678 |
+
"use_label": 10300.4873046875
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 679 |
},
|
| 680 |
{
|
| 681 |
"epoch": 0.69,
|
| 682 |
+
"grad_norm": 0.5625,
|
| 683 |
"learning_rate": 1.3192409070404582e-06,
|
| 684 |
+
"logits/chosen": -2.055405855178833,
|
| 685 |
+
"logits/rejected": -2.0071816444396973,
|
| 686 |
+
"logps/chosen": -77.25361633300781,
|
| 687 |
+
"logps/rejected": -88.34065246582031,
|
| 688 |
+
"loss": 0.6915,
|
| 689 |
+
"pred_label": 2899.9375,
|
| 690 |
+
"rewards/accuracies": 0.34687501192092896,
|
| 691 |
+
"rewards/chosen": -0.11595650017261505,
|
| 692 |
+
"rewards/margins": 0.0952102541923523,
|
| 693 |
+
"rewards/rejected": -0.21116676926612854,
|
| 694 |
"step": 660,
|
| 695 |
+
"use_label": 10526.0625
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 696 |
},
|
| 697 |
{
|
| 698 |
"epoch": 0.71,
|
| 699 |
+
"grad_norm": 0.67578125,
|
| 700 |
"learning_rate": 1.1613595214152713e-06,
|
| 701 |
+
"logits/chosen": -2.056795597076416,
|
| 702 |
+
"logits/rejected": -2.071035861968994,
|
| 703 |
+
"logps/chosen": -88.15283203125,
|
| 704 |
+
"logps/rejected": -96.39839172363281,
|
| 705 |
+
"loss": 0.6918,
|
| 706 |
+
"pred_label": 2978.0625,
|
| 707 |
+
"rewards/accuracies": 0.3499999940395355,
|
| 708 |
+
"rewards/chosen": -0.12273094803094864,
|
| 709 |
+
"rewards/margins": 0.09404005855321884,
|
| 710 |
+
"rewards/rejected": -0.2167709767818451,
|
| 711 |
"step": 680,
|
| 712 |
+
"use_label": 10767.9375
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 713 |
},
|
| 714 |
{
|
| 715 |
"epoch": 0.73,
|
| 716 |
+
"grad_norm": 0.74609375,
|
| 717 |
"learning_rate": 1.0106369933615043e-06,
|
| 718 |
+
"logits/chosen": -2.0782313346862793,
|
| 719 |
+
"logits/rejected": -2.0467371940612793,
|
| 720 |
+
"logps/chosen": -97.93621826171875,
|
| 721 |
+
"logps/rejected": -106.91497802734375,
|
| 722 |
+
"loss": 0.6917,
|
| 723 |
+
"pred_label": 3075.71240234375,
|
| 724 |
+
"rewards/accuracies": 0.3687500059604645,
|
| 725 |
+
"rewards/chosen": -0.1391007900238037,
|
| 726 |
+
"rewards/margins": 0.10766571760177612,
|
| 727 |
+
"rewards/rejected": -0.24676652252674103,
|
| 728 |
"step": 700,
|
| 729 |
+
"use_label": 10990.287109375
|
| 730 |
},
|
| 731 |
{
|
| 732 |
"epoch": 0.73,
|
| 733 |
+
"eval_logits/chosen": -1.9658821821212769,
|
| 734 |
+
"eval_logits/rejected": -1.9401167631149292,
|
| 735 |
+
"eval_logps/chosen": -80.06806182861328,
|
| 736 |
+
"eval_logps/rejected": -97.64107513427734,
|
| 737 |
+
"eval_loss": 0.6917343735694885,
|
| 738 |
+
"eval_pred_label": 3195.22216796875,
|
| 739 |
+
"eval_rewards/accuracies": 0.3511904776096344,
|
| 740 |
+
"eval_rewards/chosen": -0.11167524009943008,
|
| 741 |
+
"eval_rewards/margins": 0.1107548326253891,
|
| 742 |
+
"eval_rewards/rejected": -0.2224300652742386,
|
| 743 |
+
"eval_runtime": 247.943,
|
| 744 |
+
"eval_samples_per_second": 8.066,
|
| 745 |
"eval_steps_per_second": 0.254,
|
| 746 |
+
"eval_use_label": 11284.77734375,
|
| 747 |
"step": 700
|
| 748 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 749 |
{
|
| 750 |
"epoch": 0.75,
|
| 751 |
+
"grad_norm": 0.72265625,
|
| 752 |
"learning_rate": 8.678793653740633e-07,
|
| 753 |
+
"logits/chosen": -2.015249729156494,
|
| 754 |
+
"logits/rejected": -2.0358498096466064,
|
| 755 |
+
"logps/chosen": -70.9017562866211,
|
| 756 |
+
"logps/rejected": -86.4397201538086,
|
| 757 |
+
"loss": 0.6908,
|
| 758 |
+
"pred_label": 3306.39990234375,
|
| 759 |
+
"rewards/accuracies": 0.3187499940395355,
|
| 760 |
+
"rewards/chosen": -0.10931293666362762,
|
| 761 |
+
"rewards/margins": 0.0925455391407013,
|
| 762 |
+
"rewards/rejected": -0.20185847580432892,
|
| 763 |
"step": 720,
|
| 764 |
+
"use_label": 11583.599609375
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 765 |
},
|
| 766 |
{
|
| 767 |
"epoch": 0.77,
|
| 768 |
+
"grad_norm": 0.83203125,
|
| 769 |
"learning_rate": 7.338500848029603e-07,
|
| 770 |
+
"logits/chosen": -2.01334810256958,
|
| 771 |
+
"logits/rejected": -2.0296788215637207,
|
| 772 |
+
"logps/chosen": -74.19635772705078,
|
| 773 |
+
"logps/rejected": -83.99024200439453,
|
| 774 |
+
"loss": 0.6911,
|
| 775 |
+
"pred_label": 3386.16259765625,
|
| 776 |
+
"rewards/accuracies": 0.32499998807907104,
|
| 777 |
+
"rewards/chosen": -0.08706559240818024,
|
| 778 |
+
"rewards/margins": 0.11473299562931061,
|
| 779 |
+
"rewards/rejected": -0.20179858803749084,
|
| 780 |
"step": 740,
|
| 781 |
+
"use_label": 11823.837890625
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 782 |
},
|
| 783 |
{
|
| 784 |
"epoch": 0.8,
|
| 785 |
+
"grad_norm": 0.66015625,
|
| 786 |
"learning_rate": 6.092659210462232e-07,
|
| 787 |
+
"logits/chosen": -2.052433967590332,
|
| 788 |
+
"logits/rejected": -2.060997724533081,
|
| 789 |
+
"logps/chosen": -76.93110656738281,
|
| 790 |
+
"logps/rejected": -97.30107879638672,
|
| 791 |
+
"loss": 0.6904,
|
| 792 |
+
"pred_label": 3466.5,
|
| 793 |
+
"rewards/accuracies": 0.33125001192092896,
|
| 794 |
+
"rewards/chosen": -0.11182014644145966,
|
| 795 |
+
"rewards/margins": 0.07981495559215546,
|
| 796 |
+
"rewards/rejected": -0.1916351020336151,
|
| 797 |
"step": 760,
|
| 798 |
+
"use_label": 12063.5
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 799 |
},
|
| 800 |
{
|
| 801 |
"epoch": 0.82,
|
| 802 |
+
"grad_norm": 0.859375,
|
| 803 |
"learning_rate": 4.947931323697983e-07,
|
| 804 |
+
"logits/chosen": -2.032320737838745,
|
| 805 |
+
"logits/rejected": -2.047227144241333,
|
| 806 |
+
"logps/chosen": -89.46810913085938,
|
| 807 |
+
"logps/rejected": -95.58660125732422,
|
| 808 |
+
"loss": 0.6913,
|
| 809 |
+
"pred_label": 3558.875,
|
| 810 |
+
"rewards/accuracies": 0.375,
|
| 811 |
+
"rewards/chosen": -0.11294672638177872,
|
| 812 |
+
"rewards/margins": 0.11753211170434952,
|
| 813 |
+
"rewards/rejected": -0.23047883808612823,
|
| 814 |
"step": 780,
|
| 815 |
+
"use_label": 12291.125
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 816 |
},
|
| 817 |
{
|
| 818 |
"epoch": 0.84,
|
| 819 |
+
"grad_norm": 0.74609375,
|
| 820 |
"learning_rate": 3.910439028537638e-07,
|
| 821 |
+
"logits/chosen": -2.010045289993286,
|
| 822 |
+
"logits/rejected": -1.989505410194397,
|
| 823 |
+
"logps/chosen": -70.47514343261719,
|
| 824 |
+
"logps/rejected": -75.11082458496094,
|
| 825 |
+
"loss": 0.6912,
|
| 826 |
+
"pred_label": 3649.22509765625,
|
| 827 |
+
"rewards/accuracies": 0.3656249940395355,
|
| 828 |
+
"rewards/chosen": -0.08034199476242065,
|
| 829 |
+
"rewards/margins": 0.0995674580335617,
|
| 830 |
+
"rewards/rejected": -0.17990948259830475,
|
| 831 |
"step": 800,
|
| 832 |
+
"use_label": 12520.775390625
|
| 833 |
},
|
| 834 |
{
|
| 835 |
"epoch": 0.84,
|
| 836 |
+
"eval_logits/chosen": -1.9421576261520386,
|
| 837 |
+
"eval_logits/rejected": -1.9144233465194702,
|
| 838 |
+
"eval_logps/chosen": -77.5874252319336,
|
| 839 |
+
"eval_logps/rejected": -95.20885467529297,
|
| 840 |
+
"eval_loss": 0.6917100548744202,
|
| 841 |
+
"eval_pred_label": 3757.174560546875,
|
| 842 |
"eval_rewards/accuracies": 0.363095223903656,
|
| 843 |
+
"eval_rewards/chosen": -0.08686873316764832,
|
| 844 |
+
"eval_rewards/margins": 0.11123905330896378,
|
| 845 |
+
"eval_rewards/rejected": -0.19810780882835388,
|
| 846 |
+
"eval_runtime": 247.8932,
|
| 847 |
+
"eval_samples_per_second": 8.068,
|
| 848 |
"eval_steps_per_second": 0.254,
|
| 849 |
+
"eval_use_label": 12826.8251953125,
|
| 850 |
"step": 800
|
| 851 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 852 |
{
|
| 853 |
"epoch": 0.86,
|
| 854 |
+
"grad_norm": 0.828125,
|
| 855 |
"learning_rate": 2.98573068519539e-07,
|
| 856 |
+
"logits/chosen": -2.035728931427002,
|
| 857 |
+
"logits/rejected": -2.029679775238037,
|
| 858 |
+
"logps/chosen": -74.97032165527344,
|
| 859 |
+
"logps/rejected": -84.2763900756836,
|
| 860 |
+
"loss": 0.6908,
|
| 861 |
+
"pred_label": 3872.199951171875,
|
| 862 |
+
"rewards/accuracies": 0.3343749940395355,
|
| 863 |
+
"rewards/chosen": -0.1004786491394043,
|
| 864 |
+
"rewards/margins": 0.08142165094614029,
|
| 865 |
+
"rewards/rejected": -0.181900292634964,
|
| 866 |
"step": 820,
|
| 867 |
+
"use_label": 13121.7998046875
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 868 |
},
|
| 869 |
{
|
| 870 |
"epoch": 0.88,
|
| 871 |
+
"grad_norm": 0.6953125,
|
| 872 |
"learning_rate": 2.178751501463036e-07,
|
| 873 |
+
"logits/chosen": -2.0276803970336914,
|
| 874 |
+
"logits/rejected": -2.0149848461151123,
|
| 875 |
+
"logps/chosen": -66.70552062988281,
|
| 876 |
+
"logps/rejected": -70.63726806640625,
|
| 877 |
"loss": 0.6915,
|
| 878 |
+
"pred_label": 3954.60009765625,
|
| 879 |
+
"rewards/accuracies": 0.28437501192092896,
|
| 880 |
+
"rewards/chosen": -0.08035041391849518,
|
| 881 |
+
"rewards/margins": 0.07462439686059952,
|
| 882 |
+
"rewards/rejected": -0.1549748182296753,
|
| 883 |
"step": 840,
|
| 884 |
+
"use_label": 13359.400390625
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 885 |
},
|
| 886 |
{
|
| 887 |
"epoch": 0.9,
|
| 888 |
+
"grad_norm": 0.7578125,
|
| 889 |
"learning_rate": 1.4938170864468636e-07,
|
| 890 |
+
"logits/chosen": -2.048083543777466,
|
| 891 |
+
"logits/rejected": -2.0321922302246094,
|
| 892 |
+
"logps/chosen": -90.8042221069336,
|
| 893 |
+
"logps/rejected": -100.8233413696289,
|
| 894 |
+
"loss": 0.69,
|
| 895 |
+
"pred_label": 4041.72509765625,
|
| 896 |
+
"rewards/accuracies": 0.40625,
|
| 897 |
+
"rewards/chosen": -0.0809466689825058,
|
| 898 |
+
"rewards/margins": 0.1332779824733734,
|
| 899 |
+
"rewards/rejected": -0.2142246663570404,
|
| 900 |
"step": 860,
|
| 901 |
+
"use_label": 13592.275390625
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 902 |
},
|
| 903 |
{
|
| 904 |
"epoch": 0.92,
|
| 905 |
+
"grad_norm": 0.5546875,
|
| 906 |
"learning_rate": 9.345903713082305e-08,
|
| 907 |
+
"logits/chosen": -2.047487735748291,
|
| 908 |
+
"logits/rejected": -2.034466505050659,
|
| 909 |
+
"logps/chosen": -81.69231414794922,
|
| 910 |
+
"logps/rejected": -101.5263442993164,
|
| 911 |
+
"loss": 0.6915,
|
| 912 |
+
"pred_label": 4142.625,
|
| 913 |
+
"rewards/accuracies": 0.38749998807907104,
|
| 914 |
+
"rewards/chosen": -0.09660721570253372,
|
| 915 |
+
"rewards/margins": 0.13364934921264648,
|
| 916 |
+
"rewards/rejected": -0.23025652766227722,
|
| 917 |
"step": 880,
|
| 918 |
+
"use_label": 13811.375
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 919 |
},
|
| 920 |
{
|
| 921 |
"epoch": 0.94,
|
| 922 |
+
"grad_norm": 0.7578125,
|
| 923 |
"learning_rate": 5.0406202043228604e-08,
|
| 924 |
+
"logits/chosen": -1.9304163455963135,
|
| 925 |
+
"logits/rejected": -1.9657026529312134,
|
| 926 |
+
"logps/chosen": -75.30284118652344,
|
| 927 |
+
"logps/rejected": -99.71704864501953,
|
| 928 |
+
"loss": 0.6914,
|
| 929 |
+
"pred_label": 4235.9248046875,
|
| 930 |
+
"rewards/accuracies": 0.3375000059604645,
|
| 931 |
+
"rewards/chosen": -0.08683101832866669,
|
| 932 |
+
"rewards/margins": 0.10066400468349457,
|
| 933 |
+
"rewards/rejected": -0.18749502301216125,
|
| 934 |
"step": 900,
|
| 935 |
+
"use_label": 14038.0751953125
|
| 936 |
},
|
| 937 |
{
|
| 938 |
"epoch": 0.94,
|
| 939 |
+
"eval_logits/chosen": -1.939072847366333,
|
| 940 |
+
"eval_logits/rejected": -1.9112603664398193,
|
| 941 |
+
"eval_logps/chosen": -77.5274658203125,
|
| 942 |
+
"eval_logps/rejected": -95.22908020019531,
|
| 943 |
+
"eval_loss": 0.6917905211448669,
|
| 944 |
+
"eval_pred_label": 4352.28564453125,
|
| 945 |
+
"eval_rewards/accuracies": 0.3571428656578064,
|
| 946 |
+
"eval_rewards/chosen": -0.08626923710107803,
|
| 947 |
+
"eval_rewards/margins": 0.1120409369468689,
|
| 948 |
+
"eval_rewards/rejected": -0.19831016659736633,
|
| 949 |
+
"eval_runtime": 247.7794,
|
| 950 |
+
"eval_samples_per_second": 8.072,
|
| 951 |
"eval_steps_per_second": 0.254,
|
| 952 |
+
"eval_use_label": 14335.7138671875,
|
| 953 |
"step": 900
|
| 954 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 955 |
{
|
| 956 |
"epoch": 0.96,
|
| 957 |
+
"grad_norm": 0.80078125,
|
| 958 |
"learning_rate": 2.0453443778310766e-08,
|
| 959 |
+
"logits/chosen": -1.9801095724105835,
|
| 960 |
+
"logits/rejected": -1.9714418649673462,
|
| 961 |
+
"logps/chosen": -63.8930778503418,
|
| 962 |
+
"logps/rejected": -85.15528869628906,
|
| 963 |
+
"loss": 0.6906,
|
| 964 |
+
"pred_label": 4473.8125,
|
| 965 |
+
"rewards/accuracies": 0.31562501192092896,
|
| 966 |
+
"rewards/chosen": -0.06585933268070221,
|
| 967 |
+
"rewards/margins": 0.11039040982723236,
|
| 968 |
+
"rewards/rejected": -0.17624975740909576,
|
| 969 |
"step": 920,
|
| 970 |
+
"use_label": 14624.1875
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 971 |
},
|
| 972 |
{
|
| 973 |
"epoch": 0.98,
|
| 974 |
+
"grad_norm": 0.8359375,
|
| 975 |
"learning_rate": 3.760945397705828e-09,
|
| 976 |
+
"logits/chosen": -1.9589160680770874,
|
| 977 |
+
"logits/rejected": -1.9971154928207397,
|
| 978 |
+
"logps/chosen": -74.0462646484375,
|
| 979 |
+
"logps/rejected": -91.64708709716797,
|
| 980 |
+
"loss": 0.6913,
|
| 981 |
+
"pred_label": 4558.71240234375,
|
| 982 |
+
"rewards/accuracies": 0.32499998807907104,
|
| 983 |
+
"rewards/chosen": -0.0799408107995987,
|
| 984 |
+
"rewards/margins": 0.10116855055093765,
|
| 985 |
+
"rewards/rejected": -0.18110935389995575,
|
| 986 |
"step": 940,
|
| 987 |
+
"use_label": 14859.287109375
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 988 |
},
|
| 989 |
{
|
| 990 |
"epoch": 1.0,
|
| 991 |
"step": 955,
|
| 992 |
"total_flos": 0.0,
|
| 993 |
+
"train_loss": 0.6906769273168754,
|
| 994 |
+
"train_runtime": 20027.4031,
|
| 995 |
"train_samples_per_second": 3.053,
|
| 996 |
"train_steps_per_second": 0.048
|
| 997 |
}
|
| 998 |
],
|
| 999 |
+
"logging_steps": 20,
|
| 1000 |
"max_steps": 955,
|
| 1001 |
"num_input_tokens_seen": 0,
|
| 1002 |
"num_train_epochs": 1,
|
| 1003 |
+
"save_steps": 100,
|
| 1004 |
"total_flos": 0.0,
|
| 1005 |
"train_batch_size": 4,
|
| 1006 |
"trial_name": null,
|