Model save
Browse files- README.md +18 -13
- all_results.json +5 -5
- model-00001-of-00003.safetensors +1 -1
- model-00002-of-00003.safetensors +1 -1
- model-00003-of-00003.safetensors +1 -1
- runs/May11_01-39-44_n176-082-012/events.out.tfevents.1715363425.n176-082-012.49997.0 +2 -2
- train_results.json +5 -5
- trainer_state.json +1426 -626
- training_args.bin +2 -2
README.md
CHANGED
|
@@ -15,15 +15,15 @@ should probably proofread and complete it, then remove this comment. -->
|
|
| 15 |
|
| 16 |
This model was trained from scratch on the None dataset.
|
| 17 |
It achieves the following results on the evaluation set:
|
| 18 |
-
- Loss:
|
| 19 |
-
- Rewards/chosen: -
|
| 20 |
-
- Rewards/rejected: -
|
| 21 |
-
- Rewards/accuracies: 0.
|
| 22 |
-
- Rewards/margins: 1.
|
| 23 |
-
- Logps/rejected: -
|
| 24 |
-
- Logps/chosen: -
|
| 25 |
-
- Logits/rejected: -0.
|
| 26 |
-
- Logits/chosen: -
|
| 27 |
|
| 28 |
## Model description
|
| 29 |
|
|
@@ -60,10 +60,15 @@ The following hyperparameters were used during training:
|
|
| 60 |
|
| 61 |
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 62 |
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 63 |
-
| 0.
|
| 64 |
-
| 0.
|
| 65 |
-
| 0.
|
| 66 |
-
| 0.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 67 |
|
| 68 |
|
| 69 |
### Framework versions
|
|
|
|
| 15 |
|
| 16 |
This model was trained from scratch on the None dataset.
|
| 17 |
It achieves the following results on the evaluation set:
|
| 18 |
+
- Loss: 2.0232
|
| 19 |
+
- Rewards/chosen: -7.1946
|
| 20 |
+
- Rewards/rejected: -8.7238
|
| 21 |
+
- Rewards/accuracies: 0.6133
|
| 22 |
+
- Rewards/margins: 1.5292
|
| 23 |
+
- Logps/rejected: -1160.1461
|
| 24 |
+
- Logps/chosen: -1001.0963
|
| 25 |
+
- Logits/rejected: -0.4190
|
| 26 |
+
- Logits/chosen: -0.5892
|
| 27 |
|
| 28 |
## Model description
|
| 29 |
|
|
|
|
| 60 |
|
| 61 |
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 62 |
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 63 |
+
| 0.254 | 0.1 | 100 | 1.4761 | -5.3290 | -6.2112 | 0.5898 | 0.8822 | -908.8818 | -814.5385 | -1.4873 | -1.5203 |
|
| 64 |
+
| 0.1844 | 0.21 | 200 | 1.7253 | -6.1555 | -7.4481 | 0.6133 | 1.2926 | -1032.5726 | -897.1824 | -1.4103 | -1.4664 |
|
| 65 |
+
| 0.1635 | 0.31 | 300 | 1.6677 | -6.1768 | -7.3921 | 0.5938 | 1.2153 | -1026.9750 | -899.3143 | -0.6257 | -0.7515 |
|
| 66 |
+
| 0.1606 | 0.42 | 400 | 2.0307 | -7.0774 | -8.4601 | 0.6016 | 1.3827 | -1133.7700 | -989.3718 | -0.4798 | -0.6143 |
|
| 67 |
+
| 0.163 | 0.52 | 500 | 1.8216 | -6.5495 | -8.0368 | 0.5898 | 1.4873 | -1091.4379 | -936.5793 | -0.8136 | -0.9380 |
|
| 68 |
+
| 0.1656 | 0.63 | 600 | 1.8091 | -6.5309 | -7.9593 | 0.625 | 1.4284 | -1083.6920 | -934.7285 | -0.3700 | -0.5360 |
|
| 69 |
+
| 0.1552 | 0.73 | 700 | 2.0767 | -7.6318 | -9.1866 | 0.5977 | 1.5547 | -1206.4197 | -1044.8179 | -0.3588 | -0.5351 |
|
| 70 |
+
| 0.1377 | 0.84 | 800 | 2.0307 | -7.2870 | -8.8043 | 0.6055 | 1.5173 | -1168.1901 | -1010.3356 | -0.4156 | -0.5887 |
|
| 71 |
+
| 0.1462 | 0.94 | 900 | 2.0232 | -7.1946 | -8.7238 | 0.6133 | 1.5292 | -1160.1461 | -1001.0963 | -0.4190 | -0.5892 |
|
| 72 |
|
| 73 |
|
| 74 |
### Framework versions
|
all_results.json
CHANGED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
-
"train_loss": 0.
|
| 4 |
-
"train_runtime":
|
| 5 |
-
"train_samples":
|
| 6 |
-
"train_samples_per_second":
|
| 7 |
-
"train_steps_per_second": 0.
|
| 8 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
+
"train_loss": 0.194384015168195,
|
| 4 |
+
"train_runtime": 15417.2553,
|
| 5 |
+
"train_samples": 122270,
|
| 6 |
+
"train_samples_per_second": 7.931,
|
| 7 |
+
"train_steps_per_second": 0.062
|
| 8 |
}
|
model-00001-of-00003.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 4943178720
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7376a50ac65872ba0f050dddd468c17c88ed63f7db168915cfc56727b44a7986
|
| 3 |
size 4943178720
|
model-00002-of-00003.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 4999819336
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bbb7c97be51922fb040f2d98b5ff5071afea302437ca3fb23a3e471d57e69a9a
|
| 3 |
size 4999819336
|
model-00003-of-00003.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 4540532728
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4f5fe56371c7bc0f27b56b5e0f0eedaa0dedc324888c26a2b74345406bbc0ba6
|
| 3 |
size 4540532728
|
runs/May11_01-39-44_n176-082-012/events.out.tfevents.1715363425.n176-082-012.49997.0
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1bf2a614c5852b700c93b2b9febab8dac728defabac06b64f3a51820dbf4a8ff
|
| 3 |
+
size 77751
|
train_results.json
CHANGED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
-
"train_loss": 0.
|
| 4 |
-
"train_runtime":
|
| 5 |
-
"train_samples":
|
| 6 |
-
"train_samples_per_second":
|
| 7 |
-
"train_steps_per_second": 0.
|
| 8 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
+
"train_loss": 0.194384015168195,
|
| 4 |
+
"train_runtime": 15417.2553,
|
| 5 |
+
"train_samples": 122270,
|
| 6 |
+
"train_samples_per_second": 7.931,
|
| 7 |
+
"train_steps_per_second": 0.062
|
| 8 |
}
|
trainer_state.json
CHANGED
|
@@ -1,21 +1,21 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": null,
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
-
"epoch":
|
| 5 |
"eval_steps": 100,
|
| 6 |
-
"global_step":
|
| 7 |
"is_hyper_param_search": false,
|
| 8 |
"is_local_process_zero": true,
|
| 9 |
"is_world_process_zero": true,
|
| 10 |
"log_history": [
|
| 11 |
{
|
| 12 |
"epoch": 0.0,
|
| 13 |
-
"grad_norm":
|
| 14 |
-
"learning_rate":
|
| 15 |
-
"logits/chosen": -1.
|
| 16 |
-
"logits/rejected": -1.
|
| 17 |
-
"logps/chosen": -
|
| 18 |
-
"logps/rejected": -
|
| 19 |
"loss": 0.6931,
|
| 20 |
"rewards/accuracies": 0.0,
|
| 21 |
"rewards/chosen": 0.0,
|
|
@@ -23,787 +23,1587 @@
|
|
| 23 |
"rewards/rejected": 0.0,
|
| 24 |
"step": 1
|
| 25 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 26 |
{
|
| 27 |
"epoch": 0.02,
|
| 28 |
-
"grad_norm":
|
| 29 |
"learning_rate": 1.0416666666666667e-07,
|
| 30 |
-
"logits/chosen": -1.
|
| 31 |
-
"logits/rejected": -1.
|
| 32 |
-
"logps/chosen": -
|
| 33 |
-
"logps/rejected": -
|
| 34 |
-
"loss": 0.
|
| 35 |
-
"rewards/accuracies": 0.
|
| 36 |
-
"rewards/chosen":
|
| 37 |
-
"rewards/margins":
|
| 38 |
-
"rewards/rejected": -0.
|
| 39 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 40 |
},
|
| 41 |
{
|
| 42 |
"epoch": 0.04,
|
| 43 |
-
"grad_norm":
|
| 44 |
"learning_rate": 2.0833333333333333e-07,
|
| 45 |
-
"logits/chosen": -1.
|
| 46 |
-
"logits/rejected": -1.
|
| 47 |
-
"logps/chosen": -
|
| 48 |
-
"logps/rejected": -
|
| 49 |
-
"loss": 0.
|
| 50 |
-
"rewards/accuracies": 0.
|
| 51 |
-
"rewards/chosen": 0.
|
| 52 |
-
"rewards/margins": 0.
|
| 53 |
-
"rewards/rejected": 0.
|
| 54 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 55 |
},
|
| 56 |
{
|
| 57 |
"epoch": 0.06,
|
| 58 |
-
"grad_norm":
|
| 59 |
"learning_rate": 3.1249999999999997e-07,
|
| 60 |
-
"logits/chosen": -1.
|
| 61 |
-
"logits/rejected": -1.
|
| 62 |
-
"logps/chosen": -
|
| 63 |
-
"logps/rejected": -
|
| 64 |
-
"loss": 0.
|
| 65 |
-
"rewards/accuracies": 0.
|
| 66 |
-
"rewards/chosen": 0.
|
| 67 |
-
"rewards/margins":
|
| 68 |
-
"rewards/rejected":
|
| 69 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 70 |
},
|
| 71 |
{
|
| 72 |
"epoch": 0.08,
|
| 73 |
-
"grad_norm":
|
| 74 |
"learning_rate": 4.1666666666666667e-07,
|
| 75 |
-
"logits/chosen": -1.
|
| 76 |
-
"logits/rejected": -1.
|
| 77 |
-
"logps/chosen": -
|
| 78 |
-
"logps/rejected": -
|
| 79 |
-
"loss": 0.
|
| 80 |
-
"rewards/accuracies": 0.
|
| 81 |
-
"rewards/chosen":
|
| 82 |
-
"rewards/margins":
|
| 83 |
-
"rewards/rejected":
|
| 84 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 85 |
},
|
| 86 |
{
|
| 87 |
"epoch": 0.1,
|
| 88 |
-
"grad_norm":
|
| 89 |
-
"learning_rate": 4.
|
| 90 |
-
"logits/chosen": -1.
|
| 91 |
-
"logits/rejected": -1.
|
| 92 |
-
"logps/chosen": -
|
| 93 |
-
"logps/rejected": -
|
| 94 |
-
"loss": 0.
|
| 95 |
-
"rewards/accuracies": 0.
|
| 96 |
-
"rewards/chosen":
|
| 97 |
-
"rewards/margins":
|
| 98 |
-
"rewards/rejected":
|
| 99 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 100 |
},
|
| 101 |
{
|
| 102 |
"epoch": 0.13,
|
| 103 |
-
"grad_norm":
|
| 104 |
-
"learning_rate": 4.
|
| 105 |
-
"logits/chosen": -1.
|
| 106 |
-
"logits/rejected": -1.
|
| 107 |
-
"logps/chosen": -
|
| 108 |
-
"logps/rejected": -
|
| 109 |
-
"loss": 0.
|
| 110 |
-
"rewards/accuracies": 0.
|
| 111 |
-
"rewards/chosen":
|
| 112 |
-
"rewards/margins":
|
| 113 |
-
"rewards/rejected":
|
| 114 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 115 |
},
|
| 116 |
{
|
| 117 |
"epoch": 0.15,
|
| 118 |
-
"grad_norm":
|
| 119 |
-
"learning_rate": 4.
|
| 120 |
-
"logits/chosen": -1.
|
| 121 |
-
"logits/rejected": -1.
|
| 122 |
-
"logps/chosen": -
|
| 123 |
-
"logps/rejected": -
|
| 124 |
-
"loss": 0.
|
| 125 |
-
"rewards/accuracies": 0.
|
| 126 |
-
"rewards/chosen":
|
| 127 |
-
"rewards/margins":
|
| 128 |
-
"rewards/rejected": -
|
| 129 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 130 |
},
|
| 131 |
{
|
| 132 |
"epoch": 0.17,
|
| 133 |
-
"grad_norm":
|
| 134 |
-
"learning_rate": 4.
|
| 135 |
-
"logits/chosen": -1.
|
| 136 |
-
"logits/rejected": -1.
|
| 137 |
-
"logps/chosen": -
|
| 138 |
-
"logps/rejected": -
|
| 139 |
-
"loss": 0.
|
| 140 |
-
"rewards/accuracies": 0.
|
| 141 |
-
"rewards/chosen": -
|
| 142 |
-
"rewards/margins":
|
| 143 |
-
"rewards/rejected": -
|
| 144 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 145 |
},
|
| 146 |
{
|
| 147 |
"epoch": 0.19,
|
| 148 |
-
"grad_norm":
|
| 149 |
-
"learning_rate": 4.
|
| 150 |
-
"logits/chosen": -1.
|
| 151 |
-
"logits/rejected": -1.
|
| 152 |
-
"logps/chosen": -
|
| 153 |
-
"logps/rejected": -
|
| 154 |
-
"loss": 0.
|
| 155 |
-
"rewards/accuracies": 0.
|
| 156 |
-
"rewards/chosen": -
|
| 157 |
-
"rewards/margins":
|
| 158 |
-
"rewards/rejected": -
|
| 159 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 160 |
},
|
| 161 |
{
|
| 162 |
"epoch": 0.21,
|
| 163 |
-
"grad_norm":
|
| 164 |
-
"learning_rate": 4.
|
| 165 |
-
"logits/chosen": -1.
|
| 166 |
-
"logits/rejected": -1.
|
| 167 |
-
"logps/chosen": -
|
| 168 |
-
"logps/rejected": -
|
| 169 |
-
"loss": 0.
|
| 170 |
-
"rewards/accuracies": 0.
|
| 171 |
-
"rewards/chosen": -
|
| 172 |
-
"rewards/margins":
|
| 173 |
-
"rewards/rejected": -
|
| 174 |
-
"step":
|
| 175 |
},
|
| 176 |
{
|
| 177 |
"epoch": 0.21,
|
| 178 |
-
"eval_logits/chosen": -1.
|
| 179 |
-
"eval_logits/rejected": -1.
|
| 180 |
-
"eval_logps/chosen": -
|
| 181 |
-
"eval_logps/rejected": -
|
| 182 |
-
"eval_loss":
|
| 183 |
-
"eval_rewards/accuracies": 0.
|
| 184 |
-
"eval_rewards/chosen": -
|
| 185 |
-
"eval_rewards/margins":
|
| 186 |
-
"eval_rewards/rejected": -
|
| 187 |
-
"eval_runtime": 97.
|
| 188 |
-
"eval_samples_per_second": 20.
|
| 189 |
"eval_steps_per_second": 0.328,
|
| 190 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 191 |
},
|
| 192 |
{
|
| 193 |
"epoch": 0.23,
|
| 194 |
-
"grad_norm":
|
| 195 |
-
"learning_rate": 4.
|
| 196 |
-
"logits/chosen": -1.
|
| 197 |
-
"logits/rejected": -
|
| 198 |
-
"logps/chosen": -
|
| 199 |
-
"logps/rejected": -
|
| 200 |
-
"loss": 0.
|
| 201 |
-
"rewards/accuracies": 0.
|
| 202 |
-
"rewards/chosen": -
|
| 203 |
-
"rewards/margins":
|
| 204 |
-
"rewards/rejected": -
|
| 205 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 206 |
},
|
| 207 |
{
|
| 208 |
"epoch": 0.25,
|
| 209 |
-
"grad_norm":
|
| 210 |
-
"learning_rate": 4.
|
| 211 |
-
"logits/chosen": -
|
| 212 |
-
"logits/rejected": -
|
| 213 |
-
"logps/chosen": -
|
| 214 |
-
"logps/rejected": -
|
| 215 |
-
"loss": 0.
|
| 216 |
-
"rewards/accuracies": 0.
|
| 217 |
-
"rewards/chosen": -
|
| 218 |
-
"rewards/margins":
|
| 219 |
-
"rewards/rejected": -
|
| 220 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 221 |
},
|
| 222 |
{
|
| 223 |
"epoch": 0.27,
|
| 224 |
-
"grad_norm":
|
| 225 |
-
"learning_rate": 4.
|
| 226 |
-
"logits/chosen": -
|
| 227 |
-
"logits/rejected":
|
| 228 |
-
"logps/chosen": -
|
| 229 |
-
"logps/rejected": -
|
| 230 |
-
"loss": 0.
|
| 231 |
-
"rewards/accuracies": 0.
|
| 232 |
-
"rewards/chosen": -
|
| 233 |
-
"rewards/margins":
|
| 234 |
-
"rewards/rejected": -
|
| 235 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 236 |
},
|
| 237 |
{
|
| 238 |
"epoch": 0.29,
|
| 239 |
-
"grad_norm":
|
| 240 |
-
"learning_rate": 4.
|
| 241 |
-
"logits/chosen": -
|
| 242 |
-
"logits/rejected": -
|
| 243 |
-
"logps/chosen": -
|
| 244 |
-
"logps/rejected": -
|
| 245 |
-
"loss": 0.
|
| 246 |
-
"rewards/accuracies": 0.
|
| 247 |
-
"rewards/chosen": -
|
| 248 |
-
"rewards/margins":
|
| 249 |
-
"rewards/rejected": -
|
| 250 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 251 |
},
|
| 252 |
{
|
| 253 |
"epoch": 0.31,
|
| 254 |
-
"grad_norm":
|
| 255 |
-
"learning_rate": 4.
|
| 256 |
-
"logits/chosen": -
|
| 257 |
-
"logits/rejected": -
|
| 258 |
-
"logps/chosen": -
|
| 259 |
-
"logps/rejected": -
|
| 260 |
-
"loss": 0.
|
| 261 |
-
"rewards/accuracies": 0.
|
| 262 |
-
"rewards/chosen": -
|
| 263 |
-
"rewards/margins":
|
| 264 |
-
"rewards/rejected": -
|
| 265 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 266 |
},
|
| 267 |
{
|
| 268 |
"epoch": 0.33,
|
| 269 |
-
"grad_norm":
|
| 270 |
-
"learning_rate": 4.
|
| 271 |
-
"logits/chosen": -
|
| 272 |
-
"logits/rejected":
|
| 273 |
-
"logps/chosen": -
|
| 274 |
-
"logps/rejected": -
|
| 275 |
-
"loss": 0.
|
| 276 |
-
"rewards/accuracies": 0.
|
| 277 |
-
"rewards/chosen": -
|
| 278 |
-
"rewards/margins":
|
| 279 |
-
"rewards/rejected": -
|
| 280 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 281 |
},
|
| 282 |
{
|
| 283 |
"epoch": 0.36,
|
| 284 |
-
"grad_norm":
|
| 285 |
-
"learning_rate": 4.
|
| 286 |
-
"logits/chosen": -
|
| 287 |
-
"logits/rejected":
|
| 288 |
-
"logps/chosen": -
|
| 289 |
-
"logps/rejected": -
|
| 290 |
-
"loss": 0.
|
| 291 |
-
"rewards/accuracies": 0.
|
| 292 |
-
"rewards/chosen": -
|
| 293 |
-
"rewards/margins":
|
| 294 |
-
"rewards/rejected": -
|
| 295 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 296 |
},
|
| 297 |
{
|
| 298 |
"epoch": 0.38,
|
| 299 |
-
"grad_norm":
|
| 300 |
-
"learning_rate": 3.
|
| 301 |
-
"logits/chosen": -
|
| 302 |
-
"logits/rejected":
|
| 303 |
-
"logps/chosen": -
|
| 304 |
-
"logps/rejected": -
|
| 305 |
-
"loss": 0.
|
| 306 |
-
"rewards/accuracies": 0.
|
| 307 |
-
"rewards/chosen": -
|
| 308 |
-
"rewards/margins":
|
| 309 |
-
"rewards/rejected": -
|
| 310 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 311 |
},
|
| 312 |
{
|
| 313 |
"epoch": 0.4,
|
| 314 |
-
"grad_norm":
|
| 315 |
-
"learning_rate": 3.
|
| 316 |
-
"logits/chosen": -
|
| 317 |
-
"logits/rejected":
|
| 318 |
-
"logps/chosen": -
|
| 319 |
-
"logps/rejected": -
|
| 320 |
-
"loss": 0.
|
| 321 |
-
"rewards/accuracies": 0.
|
| 322 |
-
"rewards/chosen": -
|
| 323 |
-
"rewards/margins":
|
| 324 |
-
"rewards/rejected": -
|
| 325 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 326 |
},
|
| 327 |
{
|
| 328 |
"epoch": 0.42,
|
| 329 |
-
"grad_norm":
|
| 330 |
-
"learning_rate": 3.
|
| 331 |
-
"logits/chosen": -
|
| 332 |
-
"logits/rejected": -
|
| 333 |
-
"logps/chosen": -
|
| 334 |
-
"logps/rejected": -
|
| 335 |
-
"loss": 0.
|
| 336 |
-
"rewards/accuracies": 0.
|
| 337 |
-
"rewards/chosen": -
|
| 338 |
-
"rewards/margins":
|
| 339 |
-
"rewards/rejected": -
|
| 340 |
-
"step":
|
| 341 |
},
|
| 342 |
{
|
| 343 |
"epoch": 0.42,
|
| 344 |
-
"eval_logits/chosen": -
|
| 345 |
-
"eval_logits/rejected": -
|
| 346 |
-
"eval_logps/chosen": -
|
| 347 |
-
"eval_logps/rejected": -
|
| 348 |
-
"eval_loss":
|
| 349 |
-
"eval_rewards/accuracies": 0.
|
| 350 |
-
"eval_rewards/chosen": -
|
| 351 |
-
"eval_rewards/margins":
|
| 352 |
-
"eval_rewards/rejected": -
|
| 353 |
-
"eval_runtime": 97.
|
| 354 |
-
"eval_samples_per_second": 20.
|
| 355 |
-
"eval_steps_per_second": 0.
|
| 356 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 357 |
},
|
| 358 |
{
|
| 359 |
"epoch": 0.44,
|
| 360 |
-
"grad_norm":
|
| 361 |
-
"learning_rate": 3.
|
| 362 |
-
"logits/chosen": -
|
| 363 |
-
"logits/rejected": -
|
| 364 |
-
"logps/chosen": -
|
| 365 |
-
"logps/rejected": -
|
| 366 |
-
"loss": 0.
|
| 367 |
-
"rewards/accuracies": 0.
|
| 368 |
-
"rewards/chosen": -
|
| 369 |
-
"rewards/margins":
|
| 370 |
-
"rewards/rejected": -
|
| 371 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 372 |
},
|
| 373 |
{
|
| 374 |
"epoch": 0.46,
|
| 375 |
-
"grad_norm":
|
| 376 |
-
"learning_rate": 3.
|
| 377 |
-
"logits/chosen": -
|
| 378 |
-
"logits/rejected": -
|
| 379 |
-
"logps/chosen": -
|
| 380 |
-
"logps/rejected": -
|
| 381 |
-
"loss": 0.
|
| 382 |
-
"rewards/accuracies": 0.
|
| 383 |
-
"rewards/chosen": -
|
| 384 |
-
"rewards/margins":
|
| 385 |
-
"rewards/rejected": -
|
| 386 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 387 |
},
|
| 388 |
{
|
| 389 |
"epoch": 0.48,
|
| 390 |
-
"grad_norm":
|
| 391 |
-
"learning_rate": 3.
|
| 392 |
-
"logits/chosen": -
|
| 393 |
-
"logits/rejected": -
|
| 394 |
-
"logps/chosen": -
|
| 395 |
-
"logps/rejected": -
|
| 396 |
-
"loss": 0.
|
| 397 |
-
"rewards/accuracies": 0.
|
| 398 |
-
"rewards/chosen": -
|
| 399 |
-
"rewards/margins":
|
| 400 |
-
"rewards/rejected": -
|
| 401 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 402 |
},
|
| 403 |
{
|
| 404 |
"epoch": 0.5,
|
| 405 |
-
"grad_norm":
|
| 406 |
-
"learning_rate": 2.
|
| 407 |
-
"logits/chosen": -
|
| 408 |
-
"logits/rejected": -
|
| 409 |
-
"logps/chosen": -
|
| 410 |
-
"logps/rejected": -
|
| 411 |
-
"loss": 0.
|
| 412 |
-
"rewards/accuracies": 0.
|
| 413 |
-
"rewards/chosen": -
|
| 414 |
-
"rewards/margins":
|
| 415 |
-
"rewards/rejected": -
|
| 416 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 417 |
},
|
| 418 |
{
|
| 419 |
"epoch": 0.52,
|
| 420 |
-
"grad_norm":
|
| 421 |
-
"learning_rate": 2.
|
| 422 |
-
"logits/chosen": -1.
|
| 423 |
-
"logits/rejected": -
|
| 424 |
-
"logps/chosen": -
|
| 425 |
-
"logps/rejected": -
|
| 426 |
-
"loss": 0.
|
| 427 |
-
"rewards/accuracies": 0.
|
| 428 |
-
"rewards/chosen": -
|
| 429 |
-
"rewards/margins":
|
| 430 |
-
"rewards/rejected": -
|
| 431 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 432 |
},
|
| 433 |
{
|
| 434 |
"epoch": 0.54,
|
| 435 |
-
"grad_norm":
|
| 436 |
-
"learning_rate": 2.
|
| 437 |
-
"logits/chosen": -1.
|
| 438 |
-
"logits/rejected": -
|
| 439 |
-
"logps/chosen": -
|
| 440 |
-
"logps/rejected": -
|
| 441 |
-
"loss": 0.
|
| 442 |
-
"rewards/accuracies": 0.
|
| 443 |
-
"rewards/chosen": -
|
| 444 |
-
"rewards/margins":
|
| 445 |
-
"rewards/rejected": -
|
| 446 |
-
"step":
|
| 447 |
},
|
| 448 |
{
|
| 449 |
-
"epoch": 0.
|
| 450 |
-
"grad_norm":
|
| 451 |
-
"learning_rate": 2.
|
| 452 |
-
"logits/chosen": -1.
|
| 453 |
-
"logits/rejected": -
|
| 454 |
-
"logps/chosen": -
|
| 455 |
-
"logps/rejected": -
|
| 456 |
-
"loss": 0.
|
| 457 |
-
"rewards/accuracies": 0.
|
| 458 |
-
"rewards/chosen": -
|
| 459 |
-
"rewards/margins":
|
| 460 |
-
"rewards/rejected": -
|
| 461 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 462 |
},
|
| 463 |
{
|
| 464 |
"epoch": 0.59,
|
| 465 |
-
"grad_norm":
|
| 466 |
-
"learning_rate": 2.
|
| 467 |
-
"logits/chosen": -
|
| 468 |
-
"logits/rejected": -
|
| 469 |
-
"logps/chosen": -
|
| 470 |
-
"logps/rejected": -
|
| 471 |
-
"loss": 0.
|
| 472 |
-
"rewards/accuracies": 0.
|
| 473 |
-
"rewards/chosen": -
|
| 474 |
-
"rewards/margins":
|
| 475 |
-
"rewards/rejected": -
|
| 476 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 477 |
},
|
| 478 |
{
|
| 479 |
"epoch": 0.61,
|
| 480 |
-
"grad_norm":
|
| 481 |
-
"learning_rate": 2.
|
| 482 |
-
"logits/chosen": -
|
| 483 |
-
"logits/rejected": -
|
| 484 |
-
"logps/chosen": -
|
| 485 |
-
"logps/rejected": -
|
| 486 |
-
"loss": 0.
|
| 487 |
-
"rewards/accuracies": 0.
|
| 488 |
-
"rewards/chosen": -
|
| 489 |
-
"rewards/margins":
|
| 490 |
-
"rewards/rejected": -
|
| 491 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 492 |
},
|
| 493 |
{
|
| 494 |
"epoch": 0.63,
|
| 495 |
-
"grad_norm":
|
| 496 |
-
"learning_rate": 1.
|
| 497 |
-
"logits/chosen": -
|
| 498 |
-
"logits/rejected":
|
| 499 |
-
"logps/chosen": -
|
| 500 |
-
"logps/rejected": -
|
| 501 |
-
"loss": 0.
|
| 502 |
-
"rewards/accuracies": 0.
|
| 503 |
-
"rewards/chosen": -
|
| 504 |
-
"rewards/margins":
|
| 505 |
-
"rewards/rejected": -
|
| 506 |
-
"step":
|
| 507 |
},
|
| 508 |
{
|
| 509 |
"epoch": 0.63,
|
| 510 |
-
"eval_logits/chosen": -
|
| 511 |
-
"eval_logits/rejected": -
|
| 512 |
-
"eval_logps/chosen": -
|
| 513 |
-
"eval_logps/rejected": -
|
| 514 |
-
"eval_loss":
|
| 515 |
-
"eval_rewards/accuracies": 0.
|
| 516 |
-
"eval_rewards/chosen": -
|
| 517 |
-
"eval_rewards/margins": 1.
|
| 518 |
-
"eval_rewards/rejected": -
|
| 519 |
-
"eval_runtime": 97.
|
| 520 |
-
"eval_samples_per_second": 20.
|
| 521 |
"eval_steps_per_second": 0.328,
|
| 522 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 523 |
},
|
| 524 |
{
|
| 525 |
"epoch": 0.65,
|
| 526 |
-
"grad_norm":
|
| 527 |
-
"learning_rate": 1.
|
| 528 |
-
"logits/chosen": -
|
| 529 |
-
"logits/rejected":
|
| 530 |
-
"logps/chosen": -
|
| 531 |
-
"logps/rejected": -
|
| 532 |
-
"loss": 0.
|
| 533 |
-
"rewards/accuracies": 0.
|
| 534 |
-
"rewards/chosen": -
|
| 535 |
-
"rewards/margins":
|
| 536 |
-
"rewards/rejected": -
|
| 537 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 538 |
},
|
| 539 |
{
|
| 540 |
"epoch": 0.67,
|
| 541 |
-
"grad_norm":
|
| 542 |
-
"learning_rate": 1.
|
| 543 |
-
"logits/chosen": -0.
|
| 544 |
-
"logits/rejected":
|
| 545 |
-
"logps/chosen": -
|
| 546 |
-
"logps/rejected": -
|
| 547 |
-
"loss": 0.
|
| 548 |
-
"rewards/accuracies": 0.
|
| 549 |
-
"rewards/chosen": -1.
|
| 550 |
-
"rewards/margins":
|
| 551 |
-
"rewards/rejected": -
|
| 552 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 553 |
},
|
| 554 |
{
|
| 555 |
"epoch": 0.69,
|
| 556 |
-
"grad_norm":
|
| 557 |
-
"learning_rate": 1.
|
| 558 |
-
"logits/chosen": -
|
| 559 |
-
"logits/rejected":
|
| 560 |
-
"logps/chosen": -
|
| 561 |
-
"logps/rejected": -
|
| 562 |
-
"loss": 0.
|
| 563 |
-
"rewards/accuracies": 0.
|
| 564 |
-
"rewards/chosen": -
|
| 565 |
-
"rewards/margins":
|
| 566 |
-
"rewards/rejected": -
|
| 567 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 568 |
},
|
| 569 |
{
|
| 570 |
"epoch": 0.71,
|
| 571 |
-
"grad_norm":
|
| 572 |
-
"learning_rate": 1.
|
| 573 |
-
"logits/chosen": -
|
| 574 |
-
"logits/rejected":
|
| 575 |
-
"logps/chosen": -
|
| 576 |
-
"logps/rejected": -
|
| 577 |
-
"loss": 0.
|
| 578 |
-
"rewards/accuracies": 0.
|
| 579 |
-
"rewards/chosen": -
|
| 580 |
-
"rewards/margins":
|
| 581 |
-
"rewards/rejected": -
|
| 582 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 583 |
},
|
| 584 |
{
|
| 585 |
"epoch": 0.73,
|
| 586 |
-
"grad_norm": 45.
|
| 587 |
-
"learning_rate": 1.
|
| 588 |
-
"logits/chosen": -0.
|
| 589 |
-
"logits/rejected":
|
| 590 |
-
"logps/chosen": -
|
| 591 |
-
"logps/rejected": -
|
| 592 |
-
"loss": 0.
|
| 593 |
-
"rewards/accuracies": 0.
|
| 594 |
-
"rewards/chosen": -
|
| 595 |
-
"rewards/margins":
|
| 596 |
-
"rewards/rejected": -
|
| 597 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 598 |
},
|
| 599 |
{
|
| 600 |
"epoch": 0.75,
|
| 601 |
-
"grad_norm":
|
| 602 |
-
"learning_rate": 8.
|
| 603 |
-
"logits/chosen": -
|
| 604 |
-
"logits/rejected":
|
| 605 |
-
"logps/chosen": -
|
| 606 |
-
"logps/rejected": -
|
| 607 |
-
"loss": 0.
|
| 608 |
-
"rewards/accuracies": 0.
|
| 609 |
-
"rewards/chosen": -
|
| 610 |
-
"rewards/margins":
|
| 611 |
-
"rewards/rejected": -
|
| 612 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 613 |
},
|
| 614 |
{
|
| 615 |
"epoch": 0.77,
|
| 616 |
-
"grad_norm":
|
| 617 |
-
"learning_rate": 7.
|
| 618 |
-
"logits/chosen": -
|
| 619 |
-
"logits/rejected":
|
| 620 |
-
"logps/chosen": -
|
| 621 |
-
"logps/rejected": -
|
| 622 |
-
"loss": 0.
|
| 623 |
-
"rewards/accuracies": 0.
|
| 624 |
-
"rewards/chosen": -
|
| 625 |
-
"rewards/margins":
|
| 626 |
-
"rewards/rejected": -
|
| 627 |
-
"step":
|
| 628 |
},
|
| 629 |
{
|
| 630 |
-
"epoch": 0.
|
| 631 |
-
"grad_norm":
|
| 632 |
-
"learning_rate": 6.
|
| 633 |
-
"logits/chosen": -0.
|
| 634 |
-
"logits/rejected":
|
| 635 |
-
"logps/chosen": -
|
| 636 |
-
"logps/rejected": -
|
| 637 |
-
"loss": 0.
|
| 638 |
-
"rewards/accuracies": 0.
|
| 639 |
-
"rewards/chosen": -
|
| 640 |
-
"rewards/margins":
|
| 641 |
-
"rewards/rejected": -
|
| 642 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 643 |
},
|
| 644 |
{
|
| 645 |
"epoch": 0.82,
|
| 646 |
-
"grad_norm":
|
| 647 |
-
"learning_rate": 4.
|
| 648 |
-
"logits/chosen": -
|
| 649 |
-
"logits/rejected":
|
| 650 |
-
"logps/chosen": -
|
| 651 |
-
"logps/rejected": -
|
| 652 |
-
"loss": 0.
|
| 653 |
-
"rewards/accuracies": 0.
|
| 654 |
-
"rewards/chosen": -
|
| 655 |
-
"rewards/margins":
|
| 656 |
-
"rewards/rejected": -
|
| 657 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 658 |
},
|
| 659 |
{
|
| 660 |
"epoch": 0.84,
|
| 661 |
-
"grad_norm":
|
| 662 |
-
"learning_rate": 3.
|
| 663 |
-
"logits/chosen": -0.
|
| 664 |
-
"logits/rejected":
|
| 665 |
-
"logps/chosen": -
|
| 666 |
-
"logps/rejected": -
|
| 667 |
-
"loss": 0.
|
| 668 |
-
"rewards/accuracies": 0.
|
| 669 |
-
"rewards/chosen": -
|
| 670 |
-
"rewards/margins":
|
| 671 |
-
"rewards/rejected": -
|
| 672 |
-
"step":
|
| 673 |
},
|
| 674 |
{
|
| 675 |
"epoch": 0.84,
|
| 676 |
-
"eval_logits/chosen": -
|
| 677 |
-
"eval_logits/rejected": -0.
|
| 678 |
-
"eval_logps/chosen": -
|
| 679 |
-
"eval_logps/rejected": -
|
| 680 |
-
"eval_loss":
|
| 681 |
-
"eval_rewards/accuracies": 0.
|
| 682 |
-
"eval_rewards/chosen": -
|
| 683 |
-
"eval_rewards/margins": 1.
|
| 684 |
-
"eval_rewards/rejected": -
|
| 685 |
-
"eval_runtime": 97.
|
| 686 |
-
"eval_samples_per_second": 20.
|
| 687 |
-
"eval_steps_per_second": 0.
|
| 688 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 689 |
},
|
| 690 |
{
|
| 691 |
"epoch": 0.86,
|
| 692 |
-
"grad_norm": 44.
|
| 693 |
-
"learning_rate":
|
| 694 |
-
"logits/chosen": -
|
| 695 |
-
"logits/rejected":
|
| 696 |
-
"logps/chosen": -
|
| 697 |
-
"logps/rejected": -
|
| 698 |
-
"loss": 0.
|
| 699 |
-
"rewards/accuracies": 0.
|
| 700 |
-
"rewards/chosen": -
|
| 701 |
-
"rewards/margins":
|
| 702 |
-
"rewards/rejected": -
|
| 703 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 704 |
},
|
| 705 |
{
|
| 706 |
"epoch": 0.88,
|
| 707 |
-
"grad_norm":
|
| 708 |
-
"learning_rate": 2.
|
| 709 |
-
"logits/chosen": -
|
| 710 |
-
"logits/rejected":
|
| 711 |
-
"logps/chosen": -
|
| 712 |
-
"logps/rejected": -
|
| 713 |
-
"loss": 0.
|
| 714 |
-
"rewards/accuracies": 0.
|
| 715 |
-
"rewards/chosen": -
|
| 716 |
-
"rewards/margins":
|
| 717 |
-
"rewards/rejected": -
|
| 718 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 719 |
},
|
| 720 |
{
|
| 721 |
"epoch": 0.9,
|
| 722 |
-
"grad_norm":
|
| 723 |
-
"learning_rate": 1.
|
| 724 |
-
"logits/chosen": -0.
|
| 725 |
-
"logits/rejected":
|
| 726 |
-
"logps/chosen": -
|
| 727 |
-
"logps/rejected": -
|
| 728 |
-
"loss": 0.
|
| 729 |
-
"rewards/accuracies": 0.
|
| 730 |
-
"rewards/chosen": -
|
| 731 |
-
"rewards/margins":
|
| 732 |
-
"rewards/rejected": -
|
| 733 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 734 |
},
|
| 735 |
{
|
| 736 |
"epoch": 0.92,
|
| 737 |
-
"grad_norm":
|
| 738 |
-
"learning_rate": 9.
|
| 739 |
-
"logits/chosen": -0.
|
| 740 |
-
"logits/rejected":
|
| 741 |
-
"logps/chosen": -
|
| 742 |
-
"logps/rejected": -
|
| 743 |
-
"loss": 0.
|
| 744 |
-
"rewards/accuracies": 0.
|
| 745 |
-
"rewards/chosen": -
|
| 746 |
-
"rewards/margins":
|
| 747 |
-
"rewards/rejected": -
|
| 748 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 749 |
},
|
| 750 |
{
|
| 751 |
"epoch": 0.94,
|
| 752 |
-
"grad_norm":
|
| 753 |
-
"learning_rate": 5.
|
| 754 |
-
"logits/chosen": -0.
|
| 755 |
-
"logits/rejected":
|
| 756 |
-
"logps/chosen": -
|
| 757 |
-
"logps/rejected": -
|
| 758 |
-
"loss": 0.
|
| 759 |
-
"rewards/accuracies": 0.
|
| 760 |
-
"rewards/chosen": -1.
|
| 761 |
-
"rewards/margins":
|
| 762 |
-
"rewards/rejected": -
|
| 763 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 764 |
},
|
| 765 |
{
|
| 766 |
"epoch": 0.96,
|
| 767 |
-
"grad_norm":
|
| 768 |
-
"learning_rate": 2.
|
| 769 |
-
"logits/chosen": -0.
|
| 770 |
-
"logits/rejected": -0.
|
| 771 |
-
"logps/chosen": -
|
| 772 |
-
"logps/rejected": -
|
| 773 |
-
"loss": 0.
|
| 774 |
-
"rewards/accuracies": 0.
|
| 775 |
-
"rewards/chosen": -
|
| 776 |
-
"rewards/margins":
|
| 777 |
-
"rewards/rejected": -
|
| 778 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 779 |
},
|
| 780 |
{
|
| 781 |
"epoch": 0.98,
|
| 782 |
-
"grad_norm":
|
| 783 |
-
"learning_rate":
|
| 784 |
-
"logits/chosen": -0.
|
| 785 |
-
"logits/rejected":
|
| 786 |
-
"logps/chosen": -
|
| 787 |
-
"logps/rejected": -
|
| 788 |
-
"loss": 0.
|
| 789 |
-
"rewards/accuracies": 0.
|
| 790 |
-
"rewards/chosen": -
|
| 791 |
-
"rewards/margins":
|
| 792 |
-
"rewards/rejected": -
|
| 793 |
-
"step":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 794 |
},
|
| 795 |
{
|
| 796 |
"epoch": 1.0,
|
| 797 |
-
"step":
|
| 798 |
"total_flos": 0.0,
|
| 799 |
-
"train_loss": 0.
|
| 800 |
-
"train_runtime":
|
| 801 |
-
"train_samples_per_second":
|
| 802 |
-
"train_steps_per_second": 0.
|
| 803 |
}
|
| 804 |
],
|
| 805 |
"logging_steps": 10,
|
| 806 |
-
"max_steps":
|
| 807 |
"num_input_tokens_seen": 0,
|
| 808 |
"num_train_epochs": 1,
|
| 809 |
"save_steps": 100,
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": null,
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0.9994767137624281,
|
| 5 |
"eval_steps": 100,
|
| 6 |
+
"global_step": 955,
|
| 7 |
"is_hyper_param_search": false,
|
| 8 |
"is_local_process_zero": true,
|
| 9 |
"is_world_process_zero": true,
|
| 10 |
"log_history": [
|
| 11 |
{
|
| 12 |
"epoch": 0.0,
|
| 13 |
+
"grad_norm": 50.34250255302009,
|
| 14 |
+
"learning_rate": 5.208333333333333e-09,
|
| 15 |
+
"logits/chosen": -1.8382859230041504,
|
| 16 |
+
"logits/rejected": -1.788834810256958,
|
| 17 |
+
"logps/chosen": -119.0692138671875,
|
| 18 |
+
"logps/rejected": -76.35714721679688,
|
| 19 |
"loss": 0.6931,
|
| 20 |
"rewards/accuracies": 0.0,
|
| 21 |
"rewards/chosen": 0.0,
|
|
|
|
| 23 |
"rewards/rejected": 0.0,
|
| 24 |
"step": 1
|
| 25 |
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 0.01,
|
| 28 |
+
"grad_norm": 46.63620247813633,
|
| 29 |
+
"learning_rate": 5.208333333333333e-08,
|
| 30 |
+
"logits/chosen": -1.6616814136505127,
|
| 31 |
+
"logits/rejected": -1.5508421659469604,
|
| 32 |
+
"logps/chosen": -129.22186279296875,
|
| 33 |
+
"logps/rejected": -82.81047821044922,
|
| 34 |
+
"loss": 0.6931,
|
| 35 |
+
"rewards/accuracies": 0.375,
|
| 36 |
+
"rewards/chosen": 0.0005498434184119105,
|
| 37 |
+
"rewards/margins": 0.0009226154070347548,
|
| 38 |
+
"rewards/rejected": -0.00037277190131135285,
|
| 39 |
+
"step": 10
|
| 40 |
+
},
|
| 41 |
{
|
| 42 |
"epoch": 0.02,
|
| 43 |
+
"grad_norm": 48.137796577108965,
|
| 44 |
"learning_rate": 1.0416666666666667e-07,
|
| 45 |
+
"logits/chosen": -1.6961771249771118,
|
| 46 |
+
"logits/rejected": -1.5263831615447998,
|
| 47 |
+
"logps/chosen": -140.1017608642578,
|
| 48 |
+
"logps/rejected": -81.02197265625,
|
| 49 |
+
"loss": 0.6866,
|
| 50 |
+
"rewards/accuracies": 0.78125,
|
| 51 |
+
"rewards/chosen": 0.013111919164657593,
|
| 52 |
+
"rewards/margins": 0.01593630015850067,
|
| 53 |
+
"rewards/rejected": -0.0028243791311979294,
|
| 54 |
+
"step": 20
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 0.03,
|
| 58 |
+
"grad_norm": 37.06492370153689,
|
| 59 |
+
"learning_rate": 1.5624999999999999e-07,
|
| 60 |
+
"logits/chosen": -1.7120641469955444,
|
| 61 |
+
"logits/rejected": -1.6340694427490234,
|
| 62 |
+
"logps/chosen": -119.171875,
|
| 63 |
+
"logps/rejected": -84.84639739990234,
|
| 64 |
+
"loss": 0.6569,
|
| 65 |
+
"rewards/accuracies": 0.824999988079071,
|
| 66 |
+
"rewards/chosen": 0.05298091098666191,
|
| 67 |
+
"rewards/margins": 0.07273116707801819,
|
| 68 |
+
"rewards/rejected": -0.019750254228711128,
|
| 69 |
+
"step": 30
|
| 70 |
},
|
| 71 |
{
|
| 72 |
"epoch": 0.04,
|
| 73 |
+
"grad_norm": 38.277547537833144,
|
| 74 |
"learning_rate": 2.0833333333333333e-07,
|
| 75 |
+
"logits/chosen": -1.8135931491851807,
|
| 76 |
+
"logits/rejected": -1.7179806232452393,
|
| 77 |
+
"logps/chosen": -130.3672332763672,
|
| 78 |
+
"logps/rejected": -98.59812927246094,
|
| 79 |
+
"loss": 0.5913,
|
| 80 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 81 |
+
"rewards/chosen": 0.1295873373746872,
|
| 82 |
+
"rewards/margins": 0.2865225374698639,
|
| 83 |
+
"rewards/rejected": -0.1569352000951767,
|
| 84 |
+
"step": 40
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 0.05,
|
| 88 |
+
"grad_norm": 25.0126190198897,
|
| 89 |
+
"learning_rate": 2.604166666666667e-07,
|
| 90 |
+
"logits/chosen": -1.67790949344635,
|
| 91 |
+
"logits/rejected": -1.6322219371795654,
|
| 92 |
+
"logps/chosen": -128.07374572753906,
|
| 93 |
+
"logps/rejected": -136.76470947265625,
|
| 94 |
+
"loss": 0.4837,
|
| 95 |
+
"rewards/accuracies": 0.893750011920929,
|
| 96 |
+
"rewards/chosen": 0.012722499668598175,
|
| 97 |
+
"rewards/margins": 0.6054978370666504,
|
| 98 |
+
"rewards/rejected": -0.5927752256393433,
|
| 99 |
+
"step": 50
|
| 100 |
},
|
| 101 |
{
|
| 102 |
"epoch": 0.06,
|
| 103 |
+
"grad_norm": 24.458659174156402,
|
| 104 |
"learning_rate": 3.1249999999999997e-07,
|
| 105 |
+
"logits/chosen": -1.6139692068099976,
|
| 106 |
+
"logits/rejected": -1.611717939376831,
|
| 107 |
+
"logps/chosen": -173.83407592773438,
|
| 108 |
+
"logps/rejected": -218.3336944580078,
|
| 109 |
+
"loss": 0.3944,
|
| 110 |
+
"rewards/accuracies": 0.887499988079071,
|
| 111 |
+
"rewards/chosen": -0.2466244399547577,
|
| 112 |
+
"rewards/margins": 1.035771369934082,
|
| 113 |
+
"rewards/rejected": -1.282395839691162,
|
| 114 |
+
"step": 60
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.07,
|
| 118 |
+
"grad_norm": 32.37910330977062,
|
| 119 |
+
"learning_rate": 3.645833333333333e-07,
|
| 120 |
+
"logits/chosen": -1.4961981773376465,
|
| 121 |
+
"logits/rejected": -1.4705344438552856,
|
| 122 |
+
"logps/chosen": -191.92495727539062,
|
| 123 |
+
"logps/rejected": -297.68841552734375,
|
| 124 |
+
"loss": 0.3475,
|
| 125 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 126 |
+
"rewards/chosen": -0.6499562859535217,
|
| 127 |
+
"rewards/margins": 1.588413953781128,
|
| 128 |
+
"rewards/rejected": -2.238370418548584,
|
| 129 |
+
"step": 70
|
| 130 |
},
|
| 131 |
{
|
| 132 |
"epoch": 0.08,
|
| 133 |
+
"grad_norm": 31.121018022045003,
|
| 134 |
"learning_rate": 4.1666666666666667e-07,
|
| 135 |
+
"logits/chosen": -1.4610167741775513,
|
| 136 |
+
"logits/rejected": -1.3438676595687866,
|
| 137 |
+
"logps/chosen": -249.2657470703125,
|
| 138 |
+
"logps/rejected": -405.2835693359375,
|
| 139 |
+
"loss": 0.2961,
|
| 140 |
+
"rewards/accuracies": 0.887499988079071,
|
| 141 |
+
"rewards/chosen": -1.1774078607559204,
|
| 142 |
+
"rewards/margins": 2.065943717956543,
|
| 143 |
+
"rewards/rejected": -3.243351697921753,
|
| 144 |
+
"step": 80
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 0.09,
|
| 148 |
+
"grad_norm": 34.84421918271163,
|
| 149 |
+
"learning_rate": 4.6874999999999996e-07,
|
| 150 |
+
"logits/chosen": -1.295830488204956,
|
| 151 |
+
"logits/rejected": -1.2198989391326904,
|
| 152 |
+
"logps/chosen": -235.9141082763672,
|
| 153 |
+
"logps/rejected": -426.62005615234375,
|
| 154 |
+
"loss": 0.2782,
|
| 155 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 156 |
+
"rewards/chosen": -1.0462534427642822,
|
| 157 |
+
"rewards/margins": 2.420970916748047,
|
| 158 |
+
"rewards/rejected": -3.467224597930908,
|
| 159 |
+
"step": 90
|
| 160 |
},
|
| 161 |
{
|
| 162 |
"epoch": 0.1,
|
| 163 |
+
"grad_norm": 38.18824950345103,
|
| 164 |
+
"learning_rate": 4.999732492681437e-07,
|
| 165 |
+
"logits/chosen": -1.4374310970306396,
|
| 166 |
+
"logits/rejected": -1.2723599672317505,
|
| 167 |
+
"logps/chosen": -237.91232299804688,
|
| 168 |
+
"logps/rejected": -481.5489807128906,
|
| 169 |
+
"loss": 0.254,
|
| 170 |
+
"rewards/accuracies": 0.887499988079071,
|
| 171 |
+
"rewards/chosen": -1.0713858604431152,
|
| 172 |
+
"rewards/margins": 2.9585039615631104,
|
| 173 |
+
"rewards/rejected": -4.0298895835876465,
|
| 174 |
+
"step": 100
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 0.1,
|
| 178 |
+
"eval_logits/chosen": -1.520336627960205,
|
| 179 |
+
"eval_logits/rejected": -1.4872528314590454,
|
| 180 |
+
"eval_logps/chosen": -814.5384521484375,
|
| 181 |
+
"eval_logps/rejected": -908.8818359375,
|
| 182 |
+
"eval_loss": 1.4761102199554443,
|
| 183 |
+
"eval_rewards/accuracies": 0.58984375,
|
| 184 |
+
"eval_rewards/chosen": -5.329049110412598,
|
| 185 |
+
"eval_rewards/margins": 0.8821536302566528,
|
| 186 |
+
"eval_rewards/rejected": -6.211202144622803,
|
| 187 |
+
"eval_runtime": 97.4159,
|
| 188 |
+
"eval_samples_per_second": 20.531,
|
| 189 |
+
"eval_steps_per_second": 0.328,
|
| 190 |
+
"step": 100
|
| 191 |
+
},
|
| 192 |
+
{
|
| 193 |
+
"epoch": 0.12,
|
| 194 |
+
"grad_norm": 32.08336034758986,
|
| 195 |
+
"learning_rate": 4.996723692767926e-07,
|
| 196 |
+
"logits/chosen": -1.320703148841858,
|
| 197 |
+
"logits/rejected": -1.2061922550201416,
|
| 198 |
+
"logps/chosen": -245.25057983398438,
|
| 199 |
+
"logps/rejected": -540.9496459960938,
|
| 200 |
+
"loss": 0.2321,
|
| 201 |
+
"rewards/accuracies": 0.8812500238418579,
|
| 202 |
+
"rewards/chosen": -1.2422926425933838,
|
| 203 |
+
"rewards/margins": 3.4426703453063965,
|
| 204 |
+
"rewards/rejected": -4.684963226318359,
|
| 205 |
+
"step": 110
|
| 206 |
},
|
| 207 |
{
|
| 208 |
"epoch": 0.13,
|
| 209 |
+
"grad_norm": 42.593718204224416,
|
| 210 |
+
"learning_rate": 4.990375746213598e-07,
|
| 211 |
+
"logits/chosen": -1.4631527662277222,
|
| 212 |
+
"logits/rejected": -1.3092104196548462,
|
| 213 |
+
"logps/chosen": -291.78375244140625,
|
| 214 |
+
"logps/rejected": -656.6674194335938,
|
| 215 |
+
"loss": 0.2292,
|
| 216 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 217 |
+
"rewards/chosen": -1.5867398977279663,
|
| 218 |
+
"rewards/margins": 4.244009017944336,
|
| 219 |
+
"rewards/rejected": -5.830749034881592,
|
| 220 |
+
"step": 120
|
| 221 |
+
},
|
| 222 |
+
{
|
| 223 |
+
"epoch": 0.14,
|
| 224 |
+
"grad_norm": 34.153643966408474,
|
| 225 |
+
"learning_rate": 4.980697142834314e-07,
|
| 226 |
+
"logits/chosen": -1.5025428533554077,
|
| 227 |
+
"logits/rejected": -1.3630611896514893,
|
| 228 |
+
"logps/chosen": -250.1748046875,
|
| 229 |
+
"logps/rejected": -576.2555541992188,
|
| 230 |
+
"loss": 0.2284,
|
| 231 |
+
"rewards/accuracies": 0.8687499761581421,
|
| 232 |
+
"rewards/chosen": -1.1989684104919434,
|
| 233 |
+
"rewards/margins": 3.749941349029541,
|
| 234 |
+
"rewards/rejected": -4.948909759521484,
|
| 235 |
+
"step": 130
|
| 236 |
},
|
| 237 |
{
|
| 238 |
"epoch": 0.15,
|
| 239 |
+
"grad_norm": 38.62955098354431,
|
| 240 |
+
"learning_rate": 4.967700826904229e-07,
|
| 241 |
+
"logits/chosen": -1.5546451807022095,
|
| 242 |
+
"logits/rejected": -1.388270616531372,
|
| 243 |
+
"logps/chosen": -287.029541015625,
|
| 244 |
+
"logps/rejected": -771.34033203125,
|
| 245 |
+
"loss": 0.2221,
|
| 246 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 247 |
+
"rewards/chosen": -1.5651198625564575,
|
| 248 |
+
"rewards/margins": 5.370504379272461,
|
| 249 |
+
"rewards/rejected": -6.935624599456787,
|
| 250 |
+
"step": 140
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"epoch": 0.16,
|
| 254 |
+
"grad_norm": 45.61220950393248,
|
| 255 |
+
"learning_rate": 4.951404179843962e-07,
|
| 256 |
+
"logits/chosen": -1.5795482397079468,
|
| 257 |
+
"logits/rejected": -1.3580058813095093,
|
| 258 |
+
"logps/chosen": -292.9833068847656,
|
| 259 |
+
"logps/rejected": -779.7533569335938,
|
| 260 |
+
"loss": 0.2009,
|
| 261 |
+
"rewards/accuracies": 0.90625,
|
| 262 |
+
"rewards/chosen": -1.4337162971496582,
|
| 263 |
+
"rewards/margins": 5.4721760749816895,
|
| 264 |
+
"rewards/rejected": -6.905892848968506,
|
| 265 |
+
"step": 150
|
| 266 |
},
|
| 267 |
{
|
| 268 |
"epoch": 0.17,
|
| 269 |
+
"grad_norm": 40.221682931286466,
|
| 270 |
+
"learning_rate": 4.931828996974498e-07,
|
| 271 |
+
"logits/chosen": -1.5406692028045654,
|
| 272 |
+
"logits/rejected": -1.3174465894699097,
|
| 273 |
+
"logps/chosen": -287.9991760253906,
|
| 274 |
+
"logps/rejected": -712.6685791015625,
|
| 275 |
+
"loss": 0.2152,
|
| 276 |
+
"rewards/accuracies": 0.875,
|
| 277 |
+
"rewards/chosen": -1.5834462642669678,
|
| 278 |
+
"rewards/margins": 4.822080135345459,
|
| 279 |
+
"rewards/rejected": -6.405526161193848,
|
| 280 |
+
"step": 160
|
| 281 |
+
},
|
| 282 |
+
{
|
| 283 |
+
"epoch": 0.18,
|
| 284 |
+
"grad_norm": 52.489731314855526,
|
| 285 |
+
"learning_rate": 4.909001458367866e-07,
|
| 286 |
+
"logits/chosen": -1.5561904907226562,
|
| 287 |
+
"logits/rejected": -1.3703653812408447,
|
| 288 |
+
"logps/chosen": -361.0309753417969,
|
| 289 |
+
"logps/rejected": -811.6133422851562,
|
| 290 |
+
"loss": 0.1911,
|
| 291 |
+
"rewards/accuracies": 0.887499988079071,
|
| 292 |
+
"rewards/chosen": -2.442492961883545,
|
| 293 |
+
"rewards/margins": 4.990392208099365,
|
| 294 |
+
"rewards/rejected": -7.432885646820068,
|
| 295 |
+
"step": 170
|
| 296 |
},
|
| 297 |
{
|
| 298 |
"epoch": 0.19,
|
| 299 |
+
"grad_norm": 44.88732370136967,
|
| 300 |
+
"learning_rate": 4.882952093833627e-07,
|
| 301 |
+
"logits/chosen": -1.5479228496551514,
|
| 302 |
+
"logits/rejected": -1.402629017829895,
|
| 303 |
+
"logps/chosen": -337.51568603515625,
|
| 304 |
+
"logps/rejected": -855.9022216796875,
|
| 305 |
+
"loss": 0.2064,
|
| 306 |
+
"rewards/accuracies": 0.8812500238418579,
|
| 307 |
+
"rewards/chosen": -2.006166458129883,
|
| 308 |
+
"rewards/margins": 5.707094192504883,
|
| 309 |
+
"rewards/rejected": -7.713259696960449,
|
| 310 |
+
"step": 180
|
| 311 |
+
},
|
| 312 |
+
{
|
| 313 |
+
"epoch": 0.2,
|
| 314 |
+
"grad_norm": 39.308543148166194,
|
| 315 |
+
"learning_rate": 4.853715742087946e-07,
|
| 316 |
+
"logits/chosen": -1.4547832012176514,
|
| 317 |
+
"logits/rejected": -1.2370269298553467,
|
| 318 |
+
"logps/chosen": -310.5637512207031,
|
| 319 |
+
"logps/rejected": -798.4898071289062,
|
| 320 |
+
"loss": 0.1865,
|
| 321 |
+
"rewards/accuracies": 0.887499988079071,
|
| 322 |
+
"rewards/chosen": -1.8440923690795898,
|
| 323 |
+
"rewards/margins": 5.3871355056762695,
|
| 324 |
+
"rewards/rejected": -7.231228828430176,
|
| 325 |
+
"step": 190
|
| 326 |
},
|
| 327 |
{
|
| 328 |
"epoch": 0.21,
|
| 329 |
+
"grad_norm": 37.85441526280708,
|
| 330 |
+
"learning_rate": 4.821331504159906e-07,
|
| 331 |
+
"logits/chosen": -1.3665611743927002,
|
| 332 |
+
"logits/rejected": -1.1807641983032227,
|
| 333 |
+
"logps/chosen": -266.3150939941406,
|
| 334 |
+
"logps/rejected": -686.4867553710938,
|
| 335 |
+
"loss": 0.1844,
|
| 336 |
+
"rewards/accuracies": 0.856249988079071,
|
| 337 |
+
"rewards/chosen": -1.5169986486434937,
|
| 338 |
+
"rewards/margins": 4.59818172454834,
|
| 339 |
+
"rewards/rejected": -6.115180015563965,
|
| 340 |
+
"step": 200
|
| 341 |
},
|
| 342 |
{
|
| 343 |
"epoch": 0.21,
|
| 344 |
+
"eval_logits/chosen": -1.4664057493209839,
|
| 345 |
+
"eval_logits/rejected": -1.4103434085845947,
|
| 346 |
+
"eval_logps/chosen": -897.182373046875,
|
| 347 |
+
"eval_logps/rejected": -1032.5726318359375,
|
| 348 |
+
"eval_loss": 1.7253305912017822,
|
| 349 |
+
"eval_rewards/accuracies": 0.61328125,
|
| 350 |
+
"eval_rewards/chosen": -6.15548849105835,
|
| 351 |
+
"eval_rewards/margins": 1.2926223278045654,
|
| 352 |
+
"eval_rewards/rejected": -7.448111534118652,
|
| 353 |
+
"eval_runtime": 97.4439,
|
| 354 |
+
"eval_samples_per_second": 20.525,
|
| 355 |
"eval_steps_per_second": 0.328,
|
| 356 |
+
"step": 200
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 0.22,
|
| 360 |
+
"grad_norm": 35.05458176508214,
|
| 361 |
+
"learning_rate": 4.785842691097342e-07,
|
| 362 |
+
"logits/chosen": -1.35039484500885,
|
| 363 |
+
"logits/rejected": -1.0796902179718018,
|
| 364 |
+
"logps/chosen": -383.1498107910156,
|
| 365 |
+
"logps/rejected": -1000.6253662109375,
|
| 366 |
+
"loss": 0.1702,
|
| 367 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 368 |
+
"rewards/chosen": -2.5290865898132324,
|
| 369 |
+
"rewards/margins": 6.735090732574463,
|
| 370 |
+
"rewards/rejected": -9.264177322387695,
|
| 371 |
+
"step": 210
|
| 372 |
},
|
| 373 |
{
|
| 374 |
"epoch": 0.23,
|
| 375 |
+
"grad_norm": 48.27429845185424,
|
| 376 |
+
"learning_rate": 4.7472967660421603e-07,
|
| 377 |
+
"logits/chosen": -1.0133236646652222,
|
| 378 |
+
"logits/rejected": -0.6695674657821655,
|
| 379 |
+
"logps/chosen": -290.65142822265625,
|
| 380 |
+
"logps/rejected": -783.7594604492188,
|
| 381 |
+
"loss": 0.1754,
|
| 382 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 383 |
+
"rewards/chosen": -1.5812921524047852,
|
| 384 |
+
"rewards/margins": 5.510292053222656,
|
| 385 |
+
"rewards/rejected": -7.0915846824646,
|
| 386 |
+
"step": 220
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 0.24,
|
| 390 |
+
"grad_norm": 41.00379708475208,
|
| 391 |
+
"learning_rate": 4.705745280752585e-07,
|
| 392 |
+
"logits/chosen": -0.7725291848182678,
|
| 393 |
+
"logits/rejected": -0.4129869341850281,
|
| 394 |
+
"logps/chosen": -351.3033752441406,
|
| 395 |
+
"logps/rejected": -841.9267578125,
|
| 396 |
+
"loss": 0.2017,
|
| 397 |
+
"rewards/accuracies": 0.862500011920929,
|
| 398 |
+
"rewards/chosen": -2.1709389686584473,
|
| 399 |
+
"rewards/margins": 5.4972968101501465,
|
| 400 |
+
"rewards/rejected": -7.668235778808594,
|
| 401 |
+
"step": 230
|
| 402 |
},
|
| 403 |
{
|
| 404 |
"epoch": 0.25,
|
| 405 |
+
"grad_norm": 38.12279956674338,
|
| 406 |
+
"learning_rate": 4.6612438066572555e-07,
|
| 407 |
+
"logits/chosen": -0.4965124726295471,
|
| 408 |
+
"logits/rejected": -0.0673198476433754,
|
| 409 |
+
"logps/chosen": -287.70379638671875,
|
| 410 |
+
"logps/rejected": -758.7793579101562,
|
| 411 |
+
"loss": 0.1843,
|
| 412 |
+
"rewards/accuracies": 0.875,
|
| 413 |
+
"rewards/chosen": -1.6674537658691406,
|
| 414 |
+
"rewards/margins": 5.193233966827393,
|
| 415 |
+
"rewards/rejected": -6.860687255859375,
|
| 416 |
+
"step": 240
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"epoch": 0.26,
|
| 420 |
+
"grad_norm": 41.10699952348945,
|
| 421 |
+
"learning_rate": 4.6138518605333664e-07,
|
| 422 |
+
"logits/chosen": -0.6868407726287842,
|
| 423 |
+
"logits/rejected": -0.13767887651920319,
|
| 424 |
+
"logps/chosen": -342.7250671386719,
|
| 425 |
+
"logps/rejected": -987.6883544921875,
|
| 426 |
+
"loss": 0.1737,
|
| 427 |
+
"rewards/accuracies": 0.918749988079071,
|
| 428 |
+
"rewards/chosen": -2.031644344329834,
|
| 429 |
+
"rewards/margins": 6.998709678649902,
|
| 430 |
+
"rewards/rejected": -9.030354499816895,
|
| 431 |
+
"step": 250
|
| 432 |
},
|
| 433 |
{
|
| 434 |
"epoch": 0.27,
|
| 435 |
+
"grad_norm": 36.89441241984481,
|
| 436 |
+
"learning_rate": 4.5636328249082514e-07,
|
| 437 |
+
"logits/chosen": -0.46804434061050415,
|
| 438 |
+
"logits/rejected": 0.16666679084300995,
|
| 439 |
+
"logps/chosen": -385.1826171875,
|
| 440 |
+
"logps/rejected": -1048.572021484375,
|
| 441 |
+
"loss": 0.1683,
|
| 442 |
+
"rewards/accuracies": 0.8687499761581421,
|
| 443 |
+
"rewards/chosen": -2.4862778186798096,
|
| 444 |
+
"rewards/margins": 7.211634635925293,
|
| 445 |
+
"rewards/rejected": -9.697912216186523,
|
| 446 |
+
"step": 260
|
| 447 |
+
},
|
| 448 |
+
{
|
| 449 |
+
"epoch": 0.28,
|
| 450 |
+
"grad_norm": 33.10469086445788,
|
| 451 |
+
"learning_rate": 4.510653863290871e-07,
|
| 452 |
+
"logits/chosen": -0.6870437860488892,
|
| 453 |
+
"logits/rejected": -0.09238968789577484,
|
| 454 |
+
"logps/chosen": -350.16400146484375,
|
| 455 |
+
"logps/rejected": -1004.6405029296875,
|
| 456 |
+
"loss": 0.1688,
|
| 457 |
+
"rewards/accuracies": 0.887499988079071,
|
| 458 |
+
"rewards/chosen": -2.0535225868225098,
|
| 459 |
+
"rewards/margins": 7.145571708679199,
|
| 460 |
+
"rewards/rejected": -9.199094772338867,
|
| 461 |
+
"step": 270
|
| 462 |
},
|
| 463 |
{
|
| 464 |
"epoch": 0.29,
|
| 465 |
+
"grad_norm": 36.36855269897335,
|
| 466 |
+
"learning_rate": 4.4549858303465737e-07,
|
| 467 |
+
"logits/chosen": -0.6977792382240295,
|
| 468 |
+
"logits/rejected": -0.09684981405735016,
|
| 469 |
+
"logps/chosen": -377.00201416015625,
|
| 470 |
+
"logps/rejected": -1076.9189453125,
|
| 471 |
+
"loss": 0.1621,
|
| 472 |
+
"rewards/accuracies": 0.887499988079071,
|
| 473 |
+
"rewards/chosen": -2.410522222518921,
|
| 474 |
+
"rewards/margins": 7.565877437591553,
|
| 475 |
+
"rewards/rejected": -9.976398468017578,
|
| 476 |
+
"step": 280
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"epoch": 0.3,
|
| 480 |
+
"grad_norm": 31.447925234202767,
|
| 481 |
+
"learning_rate": 4.396703177135261e-07,
|
| 482 |
+
"logits/chosen": -0.9086171984672546,
|
| 483 |
+
"logits/rejected": -0.4163902699947357,
|
| 484 |
+
"logps/chosen": -320.157958984375,
|
| 485 |
+
"logps/rejected": -847.1726684570312,
|
| 486 |
+
"loss": 0.1843,
|
| 487 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 488 |
+
"rewards/chosen": -1.8222324848175049,
|
| 489 |
+
"rewards/margins": 5.811877250671387,
|
| 490 |
+
"rewards/rejected": -7.6341094970703125,
|
| 491 |
+
"step": 290
|
| 492 |
},
|
| 493 |
{
|
| 494 |
"epoch": 0.31,
|
| 495 |
+
"grad_norm": 29.489990961921738,
|
| 496 |
+
"learning_rate": 4.335883851539693e-07,
|
| 497 |
+
"logits/chosen": -0.8754051327705383,
|
| 498 |
+
"logits/rejected": -0.39787793159484863,
|
| 499 |
+
"logps/chosen": -303.1324157714844,
|
| 500 |
+
"logps/rejected": -911.4088745117188,
|
| 501 |
+
"loss": 0.1635,
|
| 502 |
+
"rewards/accuracies": 0.893750011920929,
|
| 503 |
+
"rewards/chosen": -1.7049236297607422,
|
| 504 |
+
"rewards/margins": 6.647863864898682,
|
| 505 |
+
"rewards/rejected": -8.352787017822266,
|
| 506 |
+
"step": 300
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"epoch": 0.31,
|
| 510 |
+
"eval_logits/chosen": -0.7514631748199463,
|
| 511 |
+
"eval_logits/rejected": -0.6256809234619141,
|
| 512 |
+
"eval_logps/chosen": -899.3142700195312,
|
| 513 |
+
"eval_logps/rejected": -1026.9749755859375,
|
| 514 |
+
"eval_loss": 1.6676901578903198,
|
| 515 |
+
"eval_rewards/accuracies": 0.59375,
|
| 516 |
+
"eval_rewards/chosen": -6.176807403564453,
|
| 517 |
+
"eval_rewards/margins": 1.215327262878418,
|
| 518 |
+
"eval_rewards/rejected": -7.392133712768555,
|
| 519 |
+
"eval_runtime": 97.2926,
|
| 520 |
+
"eval_samples_per_second": 20.557,
|
| 521 |
+
"eval_steps_per_second": 0.329,
|
| 522 |
+
"step": 300
|
| 523 |
+
},
|
| 524 |
+
{
|
| 525 |
+
"epoch": 0.32,
|
| 526 |
+
"grad_norm": 36.71426987222159,
|
| 527 |
+
"learning_rate": 4.272609194017105e-07,
|
| 528 |
+
"logits/chosen": -0.7960424423217773,
|
| 529 |
+
"logits/rejected": -0.22463683784008026,
|
| 530 |
+
"logps/chosen": -326.64764404296875,
|
| 531 |
+
"logps/rejected": -895.4781494140625,
|
| 532 |
+
"loss": 0.1774,
|
| 533 |
+
"rewards/accuracies": 0.887499988079071,
|
| 534 |
+
"rewards/chosen": -1.8289636373519897,
|
| 535 |
+
"rewards/margins": 6.345601558685303,
|
| 536 |
+
"rewards/rejected": -8.174566268920898,
|
| 537 |
+
"step": 310
|
| 538 |
},
|
| 539 |
{
|
| 540 |
"epoch": 0.33,
|
| 541 |
+
"grad_norm": 29.01401480971697,
|
| 542 |
+
"learning_rate": 4.2069638288135547e-07,
|
| 543 |
+
"logits/chosen": -0.2821110785007477,
|
| 544 |
+
"logits/rejected": 0.2019030600786209,
|
| 545 |
+
"logps/chosen": -327.86187744140625,
|
| 546 |
+
"logps/rejected": -967.6357421875,
|
| 547 |
+
"loss": 0.1667,
|
| 548 |
+
"rewards/accuracies": 0.90625,
|
| 549 |
+
"rewards/chosen": -2.0269737243652344,
|
| 550 |
+
"rewards/margins": 6.851207733154297,
|
| 551 |
+
"rewards/rejected": -8.878181457519531,
|
| 552 |
+
"step": 320
|
| 553 |
+
},
|
| 554 |
+
{
|
| 555 |
+
"epoch": 0.35,
|
| 556 |
+
"grad_norm": 27.12838321931205,
|
| 557 |
+
"learning_rate": 4.139035550786494e-07,
|
| 558 |
+
"logits/chosen": -0.5025689005851746,
|
| 559 |
+
"logits/rejected": 0.3045334815979004,
|
| 560 |
+
"logps/chosen": -339.73321533203125,
|
| 561 |
+
"logps/rejected": -999.3941650390625,
|
| 562 |
+
"loss": 0.1479,
|
| 563 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 564 |
+
"rewards/chosen": -2.077441453933716,
|
| 565 |
+
"rewards/margins": 7.147647857666016,
|
| 566 |
+
"rewards/rejected": -9.225088119506836,
|
| 567 |
+
"step": 330
|
| 568 |
},
|
| 569 |
{
|
| 570 |
"epoch": 0.36,
|
| 571 |
+
"grad_norm": 38.66513102667531,
|
| 572 |
+
"learning_rate": 4.0689152079869306e-07,
|
| 573 |
+
"logits/chosen": -0.48522940278053284,
|
| 574 |
+
"logits/rejected": 0.2592470049858093,
|
| 575 |
+
"logps/chosen": -318.987060546875,
|
| 576 |
+
"logps/rejected": -1009.4090576171875,
|
| 577 |
+
"loss": 0.1602,
|
| 578 |
+
"rewards/accuracies": 0.918749988079071,
|
| 579 |
+
"rewards/chosen": -1.873308539390564,
|
| 580 |
+
"rewards/margins": 7.443674564361572,
|
| 581 |
+
"rewards/rejected": -9.316983222961426,
|
| 582 |
+
"step": 340
|
| 583 |
+
},
|
| 584 |
+
{
|
| 585 |
+
"epoch": 0.37,
|
| 586 |
+
"grad_norm": 33.031074442796964,
|
| 587 |
+
"learning_rate": 3.99669658015821e-07,
|
| 588 |
+
"logits/chosen": -0.2326478660106659,
|
| 589 |
+
"logits/rejected": 0.3859787583351135,
|
| 590 |
+
"logps/chosen": -323.79840087890625,
|
| 591 |
+
"logps/rejected": -1037.3876953125,
|
| 592 |
+
"loss": 0.1571,
|
| 593 |
+
"rewards/accuracies": 0.9312499761581421,
|
| 594 |
+
"rewards/chosen": -1.968605637550354,
|
| 595 |
+
"rewards/margins": 7.629488468170166,
|
| 596 |
+
"rewards/rejected": -9.59809398651123,
|
| 597 |
+
"step": 350
|
| 598 |
},
|
| 599 |
{
|
| 600 |
"epoch": 0.38,
|
| 601 |
+
"grad_norm": 22.066600953398503,
|
| 602 |
+
"learning_rate": 3.92247625331392e-07,
|
| 603 |
+
"logits/chosen": -0.21493081748485565,
|
| 604 |
+
"logits/rejected": 0.5706053972244263,
|
| 605 |
+
"logps/chosen": -320.01806640625,
|
| 606 |
+
"logps/rejected": -916.8917846679688,
|
| 607 |
+
"loss": 0.1613,
|
| 608 |
+
"rewards/accuracies": 0.918749988079071,
|
| 609 |
+
"rewards/chosen": -1.859118103981018,
|
| 610 |
+
"rewards/margins": 6.541051387786865,
|
| 611 |
+
"rewards/rejected": -8.40017032623291,
|
| 612 |
+
"step": 360
|
| 613 |
+
},
|
| 614 |
+
{
|
| 615 |
+
"epoch": 0.39,
|
| 616 |
+
"grad_norm": 22.111157147625544,
|
| 617 |
+
"learning_rate": 3.846353490562664e-07,
|
| 618 |
+
"logits/chosen": -0.13797323405742645,
|
| 619 |
+
"logits/rejected": 0.591605544090271,
|
| 620 |
+
"logps/chosen": -353.294921875,
|
| 621 |
+
"logps/rejected": -1085.5618896484375,
|
| 622 |
+
"loss": 0.1497,
|
| 623 |
+
"rewards/accuracies": 0.893750011920929,
|
| 624 |
+
"rewards/chosen": -2.195202112197876,
|
| 625 |
+
"rewards/margins": 7.859287261962891,
|
| 626 |
+
"rewards/rejected": -10.054490089416504,
|
| 627 |
+
"step": 370
|
| 628 |
},
|
| 629 |
{
|
| 630 |
"epoch": 0.4,
|
| 631 |
+
"grad_norm": 28.957108905206724,
|
| 632 |
+
"learning_rate": 3.768430099352445e-07,
|
| 633 |
+
"logits/chosen": -0.25790831446647644,
|
| 634 |
+
"logits/rejected": 0.3344423174858093,
|
| 635 |
+
"logps/chosen": -380.2027282714844,
|
| 636 |
+
"logps/rejected": -1103.0699462890625,
|
| 637 |
+
"loss": 0.1628,
|
| 638 |
+
"rewards/accuracies": 0.887499988079071,
|
| 639 |
+
"rewards/chosen": -2.5323684215545654,
|
| 640 |
+
"rewards/margins": 7.7269287109375,
|
| 641 |
+
"rewards/rejected": -10.259297370910645,
|
| 642 |
+
"step": 380
|
| 643 |
+
},
|
| 644 |
+
{
|
| 645 |
+
"epoch": 0.41,
|
| 646 |
+
"grad_norm": 28.445219714861263,
|
| 647 |
+
"learning_rate": 3.6888102953122304e-07,
|
| 648 |
+
"logits/chosen": -0.47374868392944336,
|
| 649 |
+
"logits/rejected": 0.09608779847621918,
|
| 650 |
+
"logps/chosen": -365.1648254394531,
|
| 651 |
+
"logps/rejected": -1052.339111328125,
|
| 652 |
+
"loss": 0.1582,
|
| 653 |
+
"rewards/accuracies": 0.875,
|
| 654 |
+
"rewards/chosen": -2.427854061126709,
|
| 655 |
+
"rewards/margins": 7.350833892822266,
|
| 656 |
+
"rewards/rejected": -9.778688430786133,
|
| 657 |
+
"step": 390
|
| 658 |
},
|
| 659 |
{
|
| 660 |
"epoch": 0.42,
|
| 661 |
+
"grad_norm": 26.329510138760806,
|
| 662 |
+
"learning_rate": 3.607600562872785e-07,
|
| 663 |
+
"logits/chosen": -0.4803605079650879,
|
| 664 |
+
"logits/rejected": -0.029133472591638565,
|
| 665 |
+
"logps/chosen": -345.6582336425781,
|
| 666 |
+
"logps/rejected": -932.9094848632812,
|
| 667 |
+
"loss": 0.1606,
|
| 668 |
+
"rewards/accuracies": 0.893750011920929,
|
| 669 |
+
"rewards/chosen": -2.222365379333496,
|
| 670 |
+
"rewards/margins": 6.369531154632568,
|
| 671 |
+
"rewards/rejected": -8.591897010803223,
|
| 672 |
+
"step": 400
|
| 673 |
},
|
| 674 |
{
|
| 675 |
"epoch": 0.42,
|
| 676 |
+
"eval_logits/chosen": -0.6142900586128235,
|
| 677 |
+
"eval_logits/rejected": -0.4797673225402832,
|
| 678 |
+
"eval_logps/chosen": -989.371826171875,
|
| 679 |
+
"eval_logps/rejected": -1133.77001953125,
|
| 680 |
+
"eval_loss": 2.030726194381714,
|
| 681 |
+
"eval_rewards/accuracies": 0.6015625,
|
| 682 |
+
"eval_rewards/chosen": -7.077382564544678,
|
| 683 |
+
"eval_rewards/margins": 1.3827017545700073,
|
| 684 |
+
"eval_rewards/rejected": -8.460084915161133,
|
| 685 |
+
"eval_runtime": 97.3894,
|
| 686 |
+
"eval_samples_per_second": 20.536,
|
| 687 |
+
"eval_steps_per_second": 0.329,
|
| 688 |
+
"step": 400
|
| 689 |
+
},
|
| 690 |
+
{
|
| 691 |
+
"epoch": 0.43,
|
| 692 |
+
"grad_norm": 30.32229808234622,
|
| 693 |
+
"learning_rate": 3.5249095128531856e-07,
|
| 694 |
+
"logits/chosen": -0.6582576036453247,
|
| 695 |
+
"logits/rejected": -0.05353846028447151,
|
| 696 |
+
"logps/chosen": -382.0814514160156,
|
| 697 |
+
"logps/rejected": -1017.8347778320312,
|
| 698 |
+
"loss": 0.1701,
|
| 699 |
+
"rewards/accuracies": 0.8374999761581421,
|
| 700 |
+
"rewards/chosen": -2.4453189373016357,
|
| 701 |
+
"rewards/margins": 6.919729709625244,
|
| 702 |
+
"rewards/rejected": -9.365047454833984,
|
| 703 |
+
"step": 410
|
| 704 |
},
|
| 705 |
{
|
| 706 |
"epoch": 0.44,
|
| 707 |
+
"grad_norm": 26.65710867508508,
|
| 708 |
+
"learning_rate": 3.4408477372034736e-07,
|
| 709 |
+
"logits/chosen": -0.6557037830352783,
|
| 710 |
+
"logits/rejected": -0.18612375855445862,
|
| 711 |
+
"logps/chosen": -369.58209228515625,
|
| 712 |
+
"logps/rejected": -1136.3447265625,
|
| 713 |
+
"loss": 0.154,
|
| 714 |
+
"rewards/accuracies": 0.90625,
|
| 715 |
+
"rewards/chosen": -2.332486867904663,
|
| 716 |
+
"rewards/margins": 8.186752319335938,
|
| 717 |
+
"rewards/rejected": -10.51923942565918,
|
| 718 |
+
"step": 420
|
| 719 |
+
},
|
| 720 |
+
{
|
| 721 |
+
"epoch": 0.45,
|
| 722 |
+
"grad_norm": 26.24483385496738,
|
| 723 |
+
"learning_rate": 3.3555276610977276e-07,
|
| 724 |
+
"logits/chosen": -0.7492531538009644,
|
| 725 |
+
"logits/rejected": -0.0845475047826767,
|
| 726 |
+
"logps/chosen": -396.35797119140625,
|
| 727 |
+
"logps/rejected": -1169.997802734375,
|
| 728 |
+
"loss": 0.1583,
|
| 729 |
+
"rewards/accuracies": 0.918749988079071,
|
| 730 |
+
"rewards/chosen": -2.655913829803467,
|
| 731 |
+
"rewards/margins": 8.302931785583496,
|
| 732 |
+
"rewards/rejected": -10.958845138549805,
|
| 733 |
+
"step": 430
|
| 734 |
},
|
| 735 |
{
|
| 736 |
"epoch": 0.46,
|
| 737 |
+
"grad_norm": 30.73000357585953,
|
| 738 |
+
"learning_rate": 3.269063392575352e-07,
|
| 739 |
+
"logits/chosen": -0.9397071599960327,
|
| 740 |
+
"logits/rejected": -0.30021554231643677,
|
| 741 |
+
"logps/chosen": -316.59466552734375,
|
| 742 |
+
"logps/rejected": -935.5177001953125,
|
| 743 |
+
"loss": 0.1863,
|
| 744 |
+
"rewards/accuracies": 0.887499988079071,
|
| 745 |
+
"rewards/chosen": -1.8889516592025757,
|
| 746 |
+
"rewards/margins": 6.711340427398682,
|
| 747 |
+
"rewards/rejected": -8.600292205810547,
|
| 748 |
+
"step": 440
|
| 749 |
+
},
|
| 750 |
+
{
|
| 751 |
+
"epoch": 0.47,
|
| 752 |
+
"grad_norm": 37.85434716946158,
|
| 753 |
+
"learning_rate": 3.1815705699316964e-07,
|
| 754 |
+
"logits/chosen": -0.8938137292861938,
|
| 755 |
+
"logits/rejected": -0.3564215898513794,
|
| 756 |
+
"logps/chosen": -316.93023681640625,
|
| 757 |
+
"logps/rejected": -954.3201904296875,
|
| 758 |
+
"loss": 0.1637,
|
| 759 |
+
"rewards/accuracies": 0.8500000238418579,
|
| 760 |
+
"rewards/chosen": -1.857377290725708,
|
| 761 |
+
"rewards/margins": 6.8977460861206055,
|
| 762 |
+
"rewards/rejected": -8.75512409210205,
|
| 763 |
+
"step": 450
|
| 764 |
},
|
| 765 |
{
|
| 766 |
"epoch": 0.48,
|
| 767 |
+
"grad_norm": 47.07671687902545,
|
| 768 |
+
"learning_rate": 3.0931662070620794e-07,
|
| 769 |
+
"logits/chosen": -0.8714531660079956,
|
| 770 |
+
"logits/rejected": -0.3571350872516632,
|
| 771 |
+
"logps/chosen": -362.056396484375,
|
| 772 |
+
"logps/rejected": -1101.175048828125,
|
| 773 |
+
"loss": 0.166,
|
| 774 |
+
"rewards/accuracies": 0.8812500238418579,
|
| 775 |
+
"rewards/chosen": -2.2315657138824463,
|
| 776 |
+
"rewards/margins": 7.899203300476074,
|
| 777 |
+
"rewards/rejected": -10.130769729614258,
|
| 778 |
+
"step": 460
|
| 779 |
+
},
|
| 780 |
+
{
|
| 781 |
+
"epoch": 0.49,
|
| 782 |
+
"grad_norm": 41.83209910846531,
|
| 783 |
+
"learning_rate": 3.003968536966078e-07,
|
| 784 |
+
"logits/chosen": -0.987261176109314,
|
| 785 |
+
"logits/rejected": -0.44738197326660156,
|
| 786 |
+
"logps/chosen": -340.4277648925781,
|
| 787 |
+
"logps/rejected": -1015.7706909179688,
|
| 788 |
+
"loss": 0.1612,
|
| 789 |
+
"rewards/accuracies": 0.90625,
|
| 790 |
+
"rewards/chosen": -2.083724021911621,
|
| 791 |
+
"rewards/margins": 7.344973087310791,
|
| 792 |
+
"rewards/rejected": -9.428696632385254,
|
| 793 |
+
"step": 470
|
| 794 |
},
|
| 795 |
{
|
| 796 |
"epoch": 0.5,
|
| 797 |
+
"grad_norm": 70.09766166369198,
|
| 798 |
+
"learning_rate": 2.9140968536213693e-07,
|
| 799 |
+
"logits/chosen": -0.9282326698303223,
|
| 800 |
+
"logits/rejected": -0.4145265519618988,
|
| 801 |
+
"logps/chosen": -369.92498779296875,
|
| 802 |
+
"logps/rejected": -1071.494384765625,
|
| 803 |
+
"loss": 0.1731,
|
| 804 |
+
"rewards/accuracies": 0.9312499761581421,
|
| 805 |
+
"rewards/chosen": -2.441403865814209,
|
| 806 |
+
"rewards/margins": 7.5186967849731445,
|
| 807 |
+
"rewards/rejected": -9.960100173950195,
|
| 808 |
+
"step": 480
|
| 809 |
+
},
|
| 810 |
+
{
|
| 811 |
+
"epoch": 0.51,
|
| 812 |
+
"grad_norm": 31.072373563077146,
|
| 813 |
+
"learning_rate": 2.823671352438608e-07,
|
| 814 |
+
"logits/chosen": -0.7097938656806946,
|
| 815 |
+
"logits/rejected": -0.2332497388124466,
|
| 816 |
+
"logps/chosen": -376.07647705078125,
|
| 817 |
+
"logps/rejected": -1070.494873046875,
|
| 818 |
+
"loss": 0.1513,
|
| 819 |
+
"rewards/accuracies": 0.918749988079071,
|
| 820 |
+
"rewards/chosen": -2.5956947803497314,
|
| 821 |
+
"rewards/margins": 7.397839546203613,
|
| 822 |
+
"rewards/rejected": -9.993532180786133,
|
| 823 |
+
"step": 490
|
| 824 |
},
|
| 825 |
{
|
| 826 |
"epoch": 0.52,
|
| 827 |
+
"grad_norm": 39.48800882374922,
|
| 828 |
+
"learning_rate": 2.73281296951072e-07,
|
| 829 |
+
"logits/chosen": -1.0332391262054443,
|
| 830 |
+
"logits/rejected": -0.47291478514671326,
|
| 831 |
+
"logps/chosen": -354.30419921875,
|
| 832 |
+
"logps/rejected": -1030.0706787109375,
|
| 833 |
+
"loss": 0.163,
|
| 834 |
+
"rewards/accuracies": 0.9312499761581421,
|
| 835 |
+
"rewards/chosen": -2.254275321960449,
|
| 836 |
+
"rewards/margins": 7.255748748779297,
|
| 837 |
+
"rewards/rejected": -9.51002311706543,
|
| 838 |
+
"step": 500
|
| 839 |
+
},
|
| 840 |
+
{
|
| 841 |
+
"epoch": 0.52,
|
| 842 |
+
"eval_logits/chosen": -0.9380449652671814,
|
| 843 |
+
"eval_logits/rejected": -0.8136107921600342,
|
| 844 |
+
"eval_logps/chosen": -936.579345703125,
|
| 845 |
+
"eval_logps/rejected": -1091.4378662109375,
|
| 846 |
+
"eval_loss": 1.8216179609298706,
|
| 847 |
+
"eval_rewards/accuracies": 0.58984375,
|
| 848 |
+
"eval_rewards/chosen": -6.549457550048828,
|
| 849 |
+
"eval_rewards/margins": 1.487306833267212,
|
| 850 |
+
"eval_rewards/rejected": -8.036764144897461,
|
| 851 |
+
"eval_runtime": 97.3225,
|
| 852 |
+
"eval_samples_per_second": 20.55,
|
| 853 |
+
"eval_steps_per_second": 0.329,
|
| 854 |
+
"step": 500
|
| 855 |
+
},
|
| 856 |
+
{
|
| 857 |
+
"epoch": 0.53,
|
| 858 |
+
"grad_norm": 36.00441662885456,
|
| 859 |
+
"learning_rate": 2.641643219871597e-07,
|
| 860 |
+
"logits/chosen": -1.147862195968628,
|
| 861 |
+
"logits/rejected": -0.4513426721096039,
|
| 862 |
+
"logps/chosen": -346.061767578125,
|
| 863 |
+
"logps/rejected": -974.8058471679688,
|
| 864 |
+
"loss": 0.1448,
|
| 865 |
+
"rewards/accuracies": 0.9312499761581421,
|
| 866 |
+
"rewards/chosen": -1.9859822988510132,
|
| 867 |
+
"rewards/margins": 6.95660924911499,
|
| 868 |
+
"rewards/rejected": -8.942591667175293,
|
| 869 |
+
"step": 510
|
| 870 |
},
|
| 871 |
{
|
| 872 |
"epoch": 0.54,
|
| 873 |
+
"grad_norm": 32.51878519513302,
|
| 874 |
+
"learning_rate": 2.550284034980507e-07,
|
| 875 |
+
"logits/chosen": -1.13883638381958,
|
| 876 |
+
"logits/rejected": -0.4823761582374573,
|
| 877 |
+
"logps/chosen": -372.8052062988281,
|
| 878 |
+
"logps/rejected": -1022.6721801757812,
|
| 879 |
+
"loss": 0.1752,
|
| 880 |
+
"rewards/accuracies": 0.875,
|
| 881 |
+
"rewards/chosen": -2.303035020828247,
|
| 882 |
+
"rewards/margins": 7.056814670562744,
|
| 883 |
+
"rewards/rejected": -9.35984992980957,
|
| 884 |
+
"step": 520
|
| 885 |
},
|
| 886 |
{
|
| 887 |
+
"epoch": 0.55,
|
| 888 |
+
"grad_norm": 22.31604773564865,
|
| 889 |
+
"learning_rate": 2.4588575996495794e-07,
|
| 890 |
+
"logits/chosen": -1.2034275531768799,
|
| 891 |
+
"logits/rejected": -0.5457441210746765,
|
| 892 |
+
"logps/chosen": -309.73455810546875,
|
| 893 |
+
"logps/rejected": -966.7014770507812,
|
| 894 |
+
"loss": 0.1395,
|
| 895 |
+
"rewards/accuracies": 0.9312499761581421,
|
| 896 |
+
"rewards/chosen": -1.8016897439956665,
|
| 897 |
+
"rewards/margins": 7.100803375244141,
|
| 898 |
+
"rewards/rejected": -8.902493476867676,
|
| 899 |
+
"step": 530
|
| 900 |
+
},
|
| 901 |
+
{
|
| 902 |
+
"epoch": 0.57,
|
| 903 |
+
"grad_norm": 25.730552994224812,
|
| 904 |
+
"learning_rate": 2.367486188632446e-07,
|
| 905 |
+
"logits/chosen": -0.8954168558120728,
|
| 906 |
+
"logits/rejected": -0.3190861642360687,
|
| 907 |
+
"logps/chosen": -355.24444580078125,
|
| 908 |
+
"logps/rejected": -993.5911865234375,
|
| 909 |
+
"loss": 0.1404,
|
| 910 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 911 |
+
"rewards/chosen": -2.2296109199523926,
|
| 912 |
+
"rewards/margins": 6.92122745513916,
|
| 913 |
+
"rewards/rejected": -9.150837898254395,
|
| 914 |
+
"step": 540
|
| 915 |
+
},
|
| 916 |
+
{
|
| 917 |
+
"epoch": 0.58,
|
| 918 |
+
"grad_norm": 31.70801335762852,
|
| 919 |
+
"learning_rate": 2.276292003092593e-07,
|
| 920 |
+
"logits/chosen": -0.8856201171875,
|
| 921 |
+
"logits/rejected": -0.2036731243133545,
|
| 922 |
+
"logps/chosen": -384.66900634765625,
|
| 923 |
+
"logps/rejected": -1095.9716796875,
|
| 924 |
+
"loss": 0.1507,
|
| 925 |
+
"rewards/accuracies": 0.90625,
|
| 926 |
+
"rewards/chosen": -2.6011099815368652,
|
| 927 |
+
"rewards/margins": 7.6391730308532715,
|
| 928 |
+
"rewards/rejected": -10.24028205871582,
|
| 929 |
+
"step": 550
|
| 930 |
},
|
| 931 |
{
|
| 932 |
"epoch": 0.59,
|
| 933 |
+
"grad_norm": 33.368612321336464,
|
| 934 |
+
"learning_rate": 2.185397007170141e-07,
|
| 935 |
+
"logits/chosen": -0.6942281126976013,
|
| 936 |
+
"logits/rejected": -0.06129706650972366,
|
| 937 |
+
"logps/chosen": -311.1067810058594,
|
| 938 |
+
"logps/rejected": -868.57275390625,
|
| 939 |
+
"loss": 0.1601,
|
| 940 |
+
"rewards/accuracies": 0.887499988079071,
|
| 941 |
+
"rewards/chosen": -2.0446078777313232,
|
| 942 |
+
"rewards/margins": 6.013201713562012,
|
| 943 |
+
"rewards/rejected": -8.057809829711914,
|
| 944 |
+
"step": 560
|
| 945 |
+
},
|
| 946 |
+
{
|
| 947 |
+
"epoch": 0.6,
|
| 948 |
+
"grad_norm": 37.62845344818987,
|
| 949 |
+
"learning_rate": 2.094922764865619e-07,
|
| 950 |
+
"logits/chosen": -0.7505870461463928,
|
| 951 |
+
"logits/rejected": -0.11658618599176407,
|
| 952 |
+
"logps/chosen": -386.23590087890625,
|
| 953 |
+
"logps/rejected": -1095.0679931640625,
|
| 954 |
+
"loss": 0.1543,
|
| 955 |
+
"rewards/accuracies": 0.887499988079071,
|
| 956 |
+
"rewards/chosen": -2.496262311935425,
|
| 957 |
+
"rewards/margins": 7.5716872215271,
|
| 958 |
+
"rewards/rejected": -10.067950248718262,
|
| 959 |
+
"step": 570
|
| 960 |
},
|
| 961 |
{
|
| 962 |
"epoch": 0.61,
|
| 963 |
+
"grad_norm": 26.026316709252203,
|
| 964 |
+
"learning_rate": 2.0049902774588797e-07,
|
| 965 |
+
"logits/chosen": -0.7590332627296448,
|
| 966 |
+
"logits/rejected": -0.031172871589660645,
|
| 967 |
+
"logps/chosen": -312.64642333984375,
|
| 968 |
+
"logps/rejected": -921.24560546875,
|
| 969 |
+
"loss": 0.1337,
|
| 970 |
+
"rewards/accuracies": 0.887499988079071,
|
| 971 |
+
"rewards/chosen": -1.7880489826202393,
|
| 972 |
+
"rewards/margins": 6.684856414794922,
|
| 973 |
+
"rewards/rejected": -8.472906112670898,
|
| 974 |
+
"step": 580
|
| 975 |
+
},
|
| 976 |
+
{
|
| 977 |
+
"epoch": 0.62,
|
| 978 |
+
"grad_norm": 25.651974922486566,
|
| 979 |
+
"learning_rate": 1.9157198216806238e-07,
|
| 980 |
+
"logits/chosen": -0.49569645524024963,
|
| 981 |
+
"logits/rejected": 0.2907310724258423,
|
| 982 |
+
"logps/chosen": -359.0349426269531,
|
| 983 |
+
"logps/rejected": -1026.7982177734375,
|
| 984 |
+
"loss": 0.1546,
|
| 985 |
+
"rewards/accuracies": 0.887499988079071,
|
| 986 |
+
"rewards/chosen": -2.2977614402770996,
|
| 987 |
+
"rewards/margins": 7.176451206207275,
|
| 988 |
+
"rewards/rejected": -9.474211692810059,
|
| 989 |
+
"step": 590
|
| 990 |
},
|
| 991 |
{
|
| 992 |
"epoch": 0.63,
|
| 993 |
+
"grad_norm": 29.13172550353607,
|
| 994 |
+
"learning_rate": 1.8272307888529274e-07,
|
| 995 |
+
"logits/chosen": -0.5350311398506165,
|
| 996 |
+
"logits/rejected": 0.32648250460624695,
|
| 997 |
+
"logps/chosen": -355.0140380859375,
|
| 998 |
+
"logps/rejected": -1063.1131591796875,
|
| 999 |
+
"loss": 0.1656,
|
| 1000 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1001 |
+
"rewards/chosen": -2.212515354156494,
|
| 1002 |
+
"rewards/margins": 7.633421897888184,
|
| 1003 |
+
"rewards/rejected": -9.845937728881836,
|
| 1004 |
+
"step": 600
|
| 1005 |
},
|
| 1006 |
{
|
| 1007 |
"epoch": 0.63,
|
| 1008 |
+
"eval_logits/chosen": -0.5360411405563354,
|
| 1009 |
+
"eval_logits/rejected": -0.37004101276397705,
|
| 1010 |
+
"eval_logps/chosen": -934.728515625,
|
| 1011 |
+
"eval_logps/rejected": -1083.6920166015625,
|
| 1012 |
+
"eval_loss": 1.8090689182281494,
|
| 1013 |
+
"eval_rewards/accuracies": 0.625,
|
| 1014 |
+
"eval_rewards/chosen": -6.530948162078857,
|
| 1015 |
+
"eval_rewards/margins": 1.4283562898635864,
|
| 1016 |
+
"eval_rewards/rejected": -7.9593048095703125,
|
| 1017 |
+
"eval_runtime": 97.4659,
|
| 1018 |
+
"eval_samples_per_second": 20.52,
|
| 1019 |
"eval_steps_per_second": 0.328,
|
| 1020 |
+
"step": 600
|
| 1021 |
+
},
|
| 1022 |
+
{
|
| 1023 |
+
"epoch": 0.64,
|
| 1024 |
+
"grad_norm": 38.21490518689014,
|
| 1025 |
+
"learning_rate": 1.7396415252139288e-07,
|
| 1026 |
+
"logits/chosen": -0.7486557960510254,
|
| 1027 |
+
"logits/rejected": -0.10511207580566406,
|
| 1028 |
+
"logps/chosen": -321.5223693847656,
|
| 1029 |
+
"logps/rejected": -989.8853759765625,
|
| 1030 |
+
"loss": 0.1515,
|
| 1031 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1032 |
+
"rewards/chosen": -1.8860008716583252,
|
| 1033 |
+
"rewards/margins": 7.177679538726807,
|
| 1034 |
+
"rewards/rejected": -9.063680648803711,
|
| 1035 |
+
"step": 610
|
| 1036 |
},
|
| 1037 |
{
|
| 1038 |
"epoch": 0.65,
|
| 1039 |
+
"grad_norm": 37.606116637861646,
|
| 1040 |
+
"learning_rate": 1.6530691736402316e-07,
|
| 1041 |
+
"logits/chosen": -0.5896056294441223,
|
| 1042 |
+
"logits/rejected": 0.24608202278614044,
|
| 1043 |
+
"logps/chosen": -332.99127197265625,
|
| 1044 |
+
"logps/rejected": -969.2781982421875,
|
| 1045 |
+
"loss": 0.1476,
|
| 1046 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1047 |
+
"rewards/chosen": -2.0645222663879395,
|
| 1048 |
+
"rewards/margins": 6.90951681137085,
|
| 1049 |
+
"rewards/rejected": -8.974038124084473,
|
| 1050 |
+
"step": 620
|
| 1051 |
+
},
|
| 1052 |
+
{
|
| 1053 |
+
"epoch": 0.66,
|
| 1054 |
+
"grad_norm": 29.30177416360563,
|
| 1055 |
+
"learning_rate": 1.5676295169786864e-07,
|
| 1056 |
+
"logits/chosen": -0.5280860662460327,
|
| 1057 |
+
"logits/rejected": 0.11714713275432587,
|
| 1058 |
+
"logps/chosen": -344.58740234375,
|
| 1059 |
+
"logps/rejected": -951.2314453125,
|
| 1060 |
+
"loss": 0.1637,
|
| 1061 |
+
"rewards/accuracies": 0.893750011920929,
|
| 1062 |
+
"rewards/chosen": -2.1917474269866943,
|
| 1063 |
+
"rewards/margins": 6.532890319824219,
|
| 1064 |
+
"rewards/rejected": -8.724637985229492,
|
| 1065 |
+
"step": 630
|
| 1066 |
},
|
| 1067 |
{
|
| 1068 |
"epoch": 0.67,
|
| 1069 |
+
"grad_norm": 20.00745545125351,
|
| 1070 |
+
"learning_rate": 1.483436823197092e-07,
|
| 1071 |
+
"logits/chosen": -0.7493909597396851,
|
| 1072 |
+
"logits/rejected": 0.14372903108596802,
|
| 1073 |
+
"logps/chosen": -323.84112548828125,
|
| 1074 |
+
"logps/rejected": -1045.17333984375,
|
| 1075 |
+
"loss": 0.1427,
|
| 1076 |
+
"rewards/accuracies": 0.949999988079071,
|
| 1077 |
+
"rewards/chosen": -1.9064871072769165,
|
| 1078 |
+
"rewards/margins": 7.77541446685791,
|
| 1079 |
+
"rewards/rejected": -9.681901931762695,
|
| 1080 |
+
"step": 640
|
| 1081 |
+
},
|
| 1082 |
+
{
|
| 1083 |
+
"epoch": 0.68,
|
| 1084 |
+
"grad_norm": 28.980496182255987,
|
| 1085 |
+
"learning_rate": 1.4006036925609243e-07,
|
| 1086 |
+
"logits/chosen": -0.6482175588607788,
|
| 1087 |
+
"logits/rejected": 0.0625019520521164,
|
| 1088 |
+
"logps/chosen": -341.03228759765625,
|
| 1089 |
+
"logps/rejected": -1090.4324951171875,
|
| 1090 |
+
"loss": 0.1431,
|
| 1091 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1092 |
+
"rewards/chosen": -2.016282320022583,
|
| 1093 |
+
"rewards/margins": 8.004495620727539,
|
| 1094 |
+
"rewards/rejected": -10.020777702331543,
|
| 1095 |
+
"step": 650
|
| 1096 |
},
|
| 1097 |
{
|
| 1098 |
"epoch": 0.69,
|
| 1099 |
+
"grad_norm": 28.837458672914796,
|
| 1100 |
+
"learning_rate": 1.319240907040458e-07,
|
| 1101 |
+
"logits/chosen": -0.7208763360977173,
|
| 1102 |
+
"logits/rejected": 0.032781489193439484,
|
| 1103 |
+
"logps/chosen": -354.48773193359375,
|
| 1104 |
+
"logps/rejected": -1058.132080078125,
|
| 1105 |
+
"loss": 0.135,
|
| 1106 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1107 |
+
"rewards/chosen": -2.1903700828552246,
|
| 1108 |
+
"rewards/margins": 7.6033148765563965,
|
| 1109 |
+
"rewards/rejected": -9.793684005737305,
|
| 1110 |
+
"step": 660
|
| 1111 |
+
},
|
| 1112 |
+
{
|
| 1113 |
+
"epoch": 0.7,
|
| 1114 |
+
"grad_norm": 32.56630053751295,
|
| 1115 |
+
"learning_rate": 1.239457282149695e-07,
|
| 1116 |
+
"logits/chosen": -0.6535686254501343,
|
| 1117 |
+
"logits/rejected": 0.021148119121789932,
|
| 1118 |
+
"logps/chosen": -366.6623229980469,
|
| 1119 |
+
"logps/rejected": -1171.2142333984375,
|
| 1120 |
+
"loss": 0.1405,
|
| 1121 |
+
"rewards/accuracies": 0.893750011920929,
|
| 1122 |
+
"rewards/chosen": -2.4802515506744385,
|
| 1123 |
+
"rewards/margins": 8.500134468078613,
|
| 1124 |
+
"rewards/rejected": -10.980386734008789,
|
| 1125 |
+
"step": 670
|
| 1126 |
},
|
| 1127 |
{
|
| 1128 |
"epoch": 0.71,
|
| 1129 |
+
"grad_norm": 36.66974740741856,
|
| 1130 |
+
"learning_rate": 1.1613595214152711e-07,
|
| 1131 |
+
"logits/chosen": -0.7827145457267761,
|
| 1132 |
+
"logits/rejected": 0.08258621394634247,
|
| 1133 |
+
"logps/chosen": -360.87493896484375,
|
| 1134 |
+
"logps/rejected": -1151.84521484375,
|
| 1135 |
+
"loss": 0.153,
|
| 1136 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1137 |
+
"rewards/chosen": -2.3406710624694824,
|
| 1138 |
+
"rewards/margins": 8.442497253417969,
|
| 1139 |
+
"rewards/rejected": -10.783166885375977,
|
| 1140 |
+
"step": 680
|
| 1141 |
+
},
|
| 1142 |
+
{
|
| 1143 |
+
"epoch": 0.72,
|
| 1144 |
+
"grad_norm": 32.97786121900684,
|
| 1145 |
+
"learning_rate": 1.0850520736699362e-07,
|
| 1146 |
+
"logits/chosen": -0.7693961262702942,
|
| 1147 |
+
"logits/rejected": 0.005597646348178387,
|
| 1148 |
+
"logps/chosen": -345.9762268066406,
|
| 1149 |
+
"logps/rejected": -1089.049072265625,
|
| 1150 |
+
"loss": 0.1393,
|
| 1151 |
+
"rewards/accuracies": 0.90625,
|
| 1152 |
+
"rewards/chosen": -2.2439839839935303,
|
| 1153 |
+
"rewards/margins": 7.891903877258301,
|
| 1154 |
+
"rewards/rejected": -10.135889053344727,
|
| 1155 |
+
"step": 690
|
| 1156 |
},
|
| 1157 |
{
|
| 1158 |
"epoch": 0.73,
|
| 1159 |
+
"grad_norm": 45.53691199690897,
|
| 1160 |
+
"learning_rate": 1.0106369933615042e-07,
|
| 1161 |
+
"logits/chosen": -0.47497862577438354,
|
| 1162 |
+
"logits/rejected": 0.2753020226955414,
|
| 1163 |
+
"logps/chosen": -350.84698486328125,
|
| 1164 |
+
"logps/rejected": -1033.717041015625,
|
| 1165 |
+
"loss": 0.1552,
|
| 1166 |
+
"rewards/accuracies": 0.875,
|
| 1167 |
+
"rewards/chosen": -2.2757205963134766,
|
| 1168 |
+
"rewards/margins": 7.295037269592285,
|
| 1169 |
+
"rewards/rejected": -9.570757865905762,
|
| 1170 |
+
"step": 700
|
| 1171 |
+
},
|
| 1172 |
+
{
|
| 1173 |
+
"epoch": 0.73,
|
| 1174 |
+
"eval_logits/chosen": -0.5351493954658508,
|
| 1175 |
+
"eval_logits/rejected": -0.358839213848114,
|
| 1176 |
+
"eval_logps/chosen": -1044.81787109375,
|
| 1177 |
+
"eval_logps/rejected": -1206.419677734375,
|
| 1178 |
+
"eval_loss": 2.0767157077789307,
|
| 1179 |
+
"eval_rewards/accuracies": 0.59765625,
|
| 1180 |
+
"eval_rewards/chosen": -7.631843090057373,
|
| 1181 |
+
"eval_rewards/margins": 1.5547385215759277,
|
| 1182 |
+
"eval_rewards/rejected": -9.186580657958984,
|
| 1183 |
+
"eval_runtime": 97.3655,
|
| 1184 |
+
"eval_samples_per_second": 20.541,
|
| 1185 |
+
"eval_steps_per_second": 0.329,
|
| 1186 |
+
"step": 700
|
| 1187 |
+
},
|
| 1188 |
+
{
|
| 1189 |
+
"epoch": 0.74,
|
| 1190 |
+
"grad_norm": 20.912820446414077,
|
| 1191 |
+
"learning_rate": 9.382138040640714e-08,
|
| 1192 |
+
"logits/chosen": -0.7432624101638794,
|
| 1193 |
+
"logits/rejected": 0.19342878460884094,
|
| 1194 |
+
"logps/chosen": -351.33599853515625,
|
| 1195 |
+
"logps/rejected": -1151.90625,
|
| 1196 |
+
"loss": 0.134,
|
| 1197 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1198 |
+
"rewards/chosen": -2.2001760005950928,
|
| 1199 |
+
"rewards/margins": 8.568601608276367,
|
| 1200 |
+
"rewards/rejected": -10.768776893615723,
|
| 1201 |
+
"step": 710
|
| 1202 |
},
|
| 1203 |
{
|
| 1204 |
"epoch": 0.75,
|
| 1205 |
+
"grad_norm": 43.893306649250164,
|
| 1206 |
+
"learning_rate": 8.678793653740632e-08,
|
| 1207 |
+
"logits/chosen": -0.6166023015975952,
|
| 1208 |
+
"logits/rejected": 0.08570433408021927,
|
| 1209 |
+
"logps/chosen": -365.81402587890625,
|
| 1210 |
+
"logps/rejected": -1052.766357421875,
|
| 1211 |
+
"loss": 0.1483,
|
| 1212 |
+
"rewards/accuracies": 0.875,
|
| 1213 |
+
"rewards/chosen": -2.3786137104034424,
|
| 1214 |
+
"rewards/margins": 7.3965888023376465,
|
| 1215 |
+
"rewards/rejected": -9.775201797485352,
|
| 1216 |
+
"step": 720
|
| 1217 |
+
},
|
| 1218 |
+
{
|
| 1219 |
+
"epoch": 0.76,
|
| 1220 |
+
"grad_norm": 27.044416498643887,
|
| 1221 |
+
"learning_rate": 7.997277433690983e-08,
|
| 1222 |
+
"logits/chosen": -0.8449182510375977,
|
| 1223 |
+
"logits/rejected": -0.04736803472042084,
|
| 1224 |
+
"logps/chosen": -356.64935302734375,
|
| 1225 |
+
"logps/rejected": -1065.8873291015625,
|
| 1226 |
+
"loss": 0.1552,
|
| 1227 |
+
"rewards/accuracies": 0.90625,
|
| 1228 |
+
"rewards/chosen": -2.2181100845336914,
|
| 1229 |
+
"rewards/margins": 7.574668884277344,
|
| 1230 |
+
"rewards/rejected": -9.792778968811035,
|
| 1231 |
+
"step": 730
|
| 1232 |
},
|
| 1233 |
{
|
| 1234 |
"epoch": 0.77,
|
| 1235 |
+
"grad_norm": 44.06234339270125,
|
| 1236 |
+
"learning_rate": 7.338500848029602e-08,
|
| 1237 |
+
"logits/chosen": -0.7003141641616821,
|
| 1238 |
+
"logits/rejected": 0.08363965153694153,
|
| 1239 |
+
"logps/chosen": -341.93157958984375,
|
| 1240 |
+
"logps/rejected": -1166.3671875,
|
| 1241 |
+
"loss": 0.1518,
|
| 1242 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1243 |
+
"rewards/chosen": -2.0371365547180176,
|
| 1244 |
+
"rewards/margins": 8.86846923828125,
|
| 1245 |
+
"rewards/rejected": -10.905606269836426,
|
| 1246 |
+
"step": 740
|
| 1247 |
},
|
| 1248 |
{
|
| 1249 |
+
"epoch": 0.78,
|
| 1250 |
+
"grad_norm": 43.35684912820975,
|
| 1251 |
+
"learning_rate": 6.70334495204884e-08,
|
| 1252 |
+
"logits/chosen": -0.568864643573761,
|
| 1253 |
+
"logits/rejected": 0.3128158748149872,
|
| 1254 |
+
"logps/chosen": -353.63616943359375,
|
| 1255 |
+
"logps/rejected": -1124.0401611328125,
|
| 1256 |
+
"loss": 0.1381,
|
| 1257 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1258 |
+
"rewards/chosen": -2.2875678539276123,
|
| 1259 |
+
"rewards/margins": 8.232528686523438,
|
| 1260 |
+
"rewards/rejected": -10.520097732543945,
|
| 1261 |
+
"step": 750
|
| 1262 |
+
},
|
| 1263 |
+
{
|
| 1264 |
+
"epoch": 0.8,
|
| 1265 |
+
"grad_norm": 26.263047158523968,
|
| 1266 |
+
"learning_rate": 6.092659210462231e-08,
|
| 1267 |
+
"logits/chosen": -0.7811049222946167,
|
| 1268 |
+
"logits/rejected": 0.1209399476647377,
|
| 1269 |
+
"logps/chosen": -373.8915710449219,
|
| 1270 |
+
"logps/rejected": -1145.06396484375,
|
| 1271 |
+
"loss": 0.1361,
|
| 1272 |
+
"rewards/accuracies": 0.8812500238418579,
|
| 1273 |
+
"rewards/chosen": -2.237269878387451,
|
| 1274 |
+
"rewards/margins": 8.342939376831055,
|
| 1275 |
+
"rewards/rejected": -10.580209732055664,
|
| 1276 |
+
"step": 760
|
| 1277 |
+
},
|
| 1278 |
+
{
|
| 1279 |
+
"epoch": 0.81,
|
| 1280 |
+
"grad_norm": 33.38453699325153,
|
| 1281 |
+
"learning_rate": 5.507260361320737e-08,
|
| 1282 |
+
"logits/chosen": -0.6777874231338501,
|
| 1283 |
+
"logits/rejected": 0.20796041190624237,
|
| 1284 |
+
"logps/chosen": -355.58575439453125,
|
| 1285 |
+
"logps/rejected": -1019.7391357421875,
|
| 1286 |
+
"loss": 0.1459,
|
| 1287 |
+
"rewards/accuracies": 0.893750011920929,
|
| 1288 |
+
"rewards/chosen": -2.242889642715454,
|
| 1289 |
+
"rewards/margins": 7.201622009277344,
|
| 1290 |
+
"rewards/rejected": -9.444511413574219,
|
| 1291 |
+
"step": 770
|
| 1292 |
},
|
| 1293 |
{
|
| 1294 |
"epoch": 0.82,
|
| 1295 |
+
"grad_norm": 27.52737122861175,
|
| 1296 |
+
"learning_rate": 4.947931323697982e-08,
|
| 1297 |
+
"logits/chosen": -0.7127518653869629,
|
| 1298 |
+
"logits/rejected": 0.13608665764331818,
|
| 1299 |
+
"logps/chosen": -345.1900939941406,
|
| 1300 |
+
"logps/rejected": -1019.2976684570312,
|
| 1301 |
+
"loss": 0.1525,
|
| 1302 |
+
"rewards/accuracies": 0.8999999761581421,
|
| 1303 |
+
"rewards/chosen": -2.1412816047668457,
|
| 1304 |
+
"rewards/margins": 7.251255989074707,
|
| 1305 |
+
"rewards/rejected": -9.392538070678711,
|
| 1306 |
+
"step": 780
|
| 1307 |
+
},
|
| 1308 |
+
{
|
| 1309 |
+
"epoch": 0.83,
|
| 1310 |
+
"grad_norm": 28.47492588939245,
|
| 1311 |
+
"learning_rate": 4.415420150605398e-08,
|
| 1312 |
+
"logits/chosen": -0.7804344892501831,
|
| 1313 |
+
"logits/rejected": -0.030652623623609543,
|
| 1314 |
+
"logps/chosen": -339.2191162109375,
|
| 1315 |
+
"logps/rejected": -1011.9390869140625,
|
| 1316 |
+
"loss": 0.1619,
|
| 1317 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1318 |
+
"rewards/chosen": -2.1389918327331543,
|
| 1319 |
+
"rewards/margins": 7.208333492279053,
|
| 1320 |
+
"rewards/rejected": -9.347326278686523,
|
| 1321 |
+
"step": 790
|
| 1322 |
},
|
| 1323 |
{
|
| 1324 |
"epoch": 0.84,
|
| 1325 |
+
"grad_norm": 21.149940049259552,
|
| 1326 |
+
"learning_rate": 3.9104390285376374e-08,
|
| 1327 |
+
"logits/chosen": -0.6159350872039795,
|
| 1328 |
+
"logits/rejected": 0.16070613265037537,
|
| 1329 |
+
"logps/chosen": -341.08917236328125,
|
| 1330 |
+
"logps/rejected": -1105.7109375,
|
| 1331 |
+
"loss": 0.1377,
|
| 1332 |
+
"rewards/accuracies": 0.9125000238418579,
|
| 1333 |
+
"rewards/chosen": -2.1167824268341064,
|
| 1334 |
+
"rewards/margins": 8.197529792785645,
|
| 1335 |
+
"rewards/rejected": -10.314311027526855,
|
| 1336 |
+
"step": 800
|
| 1337 |
},
|
| 1338 |
{
|
| 1339 |
"epoch": 0.84,
|
| 1340 |
+
"eval_logits/chosen": -0.588712751865387,
|
| 1341 |
+
"eval_logits/rejected": -0.41556617617607117,
|
| 1342 |
+
"eval_logps/chosen": -1010.3356323242188,
|
| 1343 |
+
"eval_logps/rejected": -1168.1900634765625,
|
| 1344 |
+
"eval_loss": 2.030700206756592,
|
| 1345 |
+
"eval_rewards/accuracies": 0.60546875,
|
| 1346 |
+
"eval_rewards/chosen": -7.287020683288574,
|
| 1347 |
+
"eval_rewards/margins": 1.5172640085220337,
|
| 1348 |
+
"eval_rewards/rejected": -8.804285049438477,
|
| 1349 |
+
"eval_runtime": 97.2124,
|
| 1350 |
+
"eval_samples_per_second": 20.573,
|
| 1351 |
+
"eval_steps_per_second": 0.329,
|
| 1352 |
+
"step": 800
|
| 1353 |
+
},
|
| 1354 |
+
{
|
| 1355 |
+
"epoch": 0.85,
|
| 1356 |
+
"grad_norm": 24.57939572568816,
|
| 1357 |
+
"learning_rate": 3.433663324986208e-08,
|
| 1358 |
+
"logits/chosen": -0.662031352519989,
|
| 1359 |
+
"logits/rejected": 0.13034440577030182,
|
| 1360 |
+
"logps/chosen": -334.8114318847656,
|
| 1361 |
+
"logps/rejected": -1060.559814453125,
|
| 1362 |
+
"loss": 0.158,
|
| 1363 |
+
"rewards/accuracies": 0.918749988079071,
|
| 1364 |
+
"rewards/chosen": -2.030988931655884,
|
| 1365 |
+
"rewards/margins": 7.7772111892700195,
|
| 1366 |
+
"rewards/rejected": -9.808199882507324,
|
| 1367 |
+
"step": 810
|
| 1368 |
},
|
| 1369 |
{
|
| 1370 |
"epoch": 0.86,
|
| 1371 |
+
"grad_norm": 44.23966241882351,
|
| 1372 |
+
"learning_rate": 2.9857306851953897e-08,
|
| 1373 |
+
"logits/chosen": -0.6313251256942749,
|
| 1374 |
+
"logits/rejected": 0.16264604032039642,
|
| 1375 |
+
"logps/chosen": -337.66143798828125,
|
| 1376 |
+
"logps/rejected": -1053.348388671875,
|
| 1377 |
+
"loss": 0.1489,
|
| 1378 |
+
"rewards/accuracies": 0.8812500238418579,
|
| 1379 |
+
"rewards/chosen": -2.0846145153045654,
|
| 1380 |
+
"rewards/margins": 7.6865739822387695,
|
| 1381 |
+
"rewards/rejected": -9.771188735961914,
|
| 1382 |
+
"step": 820
|
| 1383 |
+
},
|
| 1384 |
+
{
|
| 1385 |
+
"epoch": 0.87,
|
| 1386 |
+
"grad_norm": 45.06145430483538,
|
| 1387 |
+
"learning_rate": 2.567240179368185e-08,
|
| 1388 |
+
"logits/chosen": -0.6046188473701477,
|
| 1389 |
+
"logits/rejected": 0.2691899836063385,
|
| 1390 |
+
"logps/chosen": -344.50836181640625,
|
| 1391 |
+
"logps/rejected": -1100.190185546875,
|
| 1392 |
+
"loss": 0.1406,
|
| 1393 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1394 |
+
"rewards/chosen": -2.13590669631958,
|
| 1395 |
+
"rewards/margins": 8.086740493774414,
|
| 1396 |
+
"rewards/rejected": -10.222646713256836,
|
| 1397 |
+
"step": 830
|
| 1398 |
},
|
| 1399 |
{
|
| 1400 |
"epoch": 0.88,
|
| 1401 |
+
"grad_norm": 44.72639985506377,
|
| 1402 |
+
"learning_rate": 2.1787515014630357e-08,
|
| 1403 |
+
"logits/chosen": -0.9460281133651733,
|
| 1404 |
+
"logits/rejected": 0.00803391169756651,
|
| 1405 |
+
"logps/chosen": -338.85980224609375,
|
| 1406 |
+
"logps/rejected": -1107.776611328125,
|
| 1407 |
+
"loss": 0.139,
|
| 1408 |
+
"rewards/accuracies": 0.925000011920929,
|
| 1409 |
+
"rewards/chosen": -1.9058221578598022,
|
| 1410 |
+
"rewards/margins": 8.424005508422852,
|
| 1411 |
+
"rewards/rejected": -10.329828262329102,
|
| 1412 |
+
"step": 840
|
| 1413 |
+
},
|
| 1414 |
+
{
|
| 1415 |
+
"epoch": 0.89,
|
| 1416 |
+
"grad_norm": 33.16503245142525,
|
| 1417 |
+
"learning_rate": 1.820784220652766e-08,
|
| 1418 |
+
"logits/chosen": -0.46998101472854614,
|
| 1419 |
+
"logits/rejected": 0.2729955017566681,
|
| 1420 |
+
"logps/chosen": -349.9148254394531,
|
| 1421 |
+
"logps/rejected": -1000.7681884765625,
|
| 1422 |
+
"loss": 0.1575,
|
| 1423 |
+
"rewards/accuracies": 0.856249988079071,
|
| 1424 |
+
"rewards/chosen": -2.2772185802459717,
|
| 1425 |
+
"rewards/margins": 6.997946262359619,
|
| 1426 |
+
"rewards/rejected": -9.275164604187012,
|
| 1427 |
+
"step": 850
|
| 1428 |
},
|
| 1429 |
{
|
| 1430 |
"epoch": 0.9,
|
| 1431 |
+
"grad_norm": 34.5662684913937,
|
| 1432 |
+
"learning_rate": 1.4938170864468636e-08,
|
| 1433 |
+
"logits/chosen": -0.7951310276985168,
|
| 1434 |
+
"logits/rejected": 0.0224321149289608,
|
| 1435 |
+
"logps/chosen": -330.31744384765625,
|
| 1436 |
+
"logps/rejected": -1079.454345703125,
|
| 1437 |
+
"loss": 0.1483,
|
| 1438 |
+
"rewards/accuracies": 0.9624999761581421,
|
| 1439 |
+
"rewards/chosen": -2.00297212600708,
|
| 1440 |
+
"rewards/margins": 8.02238941192627,
|
| 1441 |
+
"rewards/rejected": -10.025362014770508,
|
| 1442 |
+
"step": 860
|
| 1443 |
+
},
|
| 1444 |
+
{
|
| 1445 |
+
"epoch": 0.91,
|
| 1446 |
+
"grad_norm": 39.90580157192396,
|
| 1447 |
+
"learning_rate": 1.1982873884064465e-08,
|
| 1448 |
+
"logits/chosen": -0.7433925867080688,
|
| 1449 |
+
"logits/rejected": 0.03819055110216141,
|
| 1450 |
+
"logps/chosen": -343.10076904296875,
|
| 1451 |
+
"logps/rejected": -1177.3214111328125,
|
| 1452 |
+
"loss": 0.1454,
|
| 1453 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1454 |
+
"rewards/chosen": -2.095435619354248,
|
| 1455 |
+
"rewards/margins": 8.91866683959961,
|
| 1456 |
+
"rewards/rejected": -11.0141019821167,
|
| 1457 |
+
"step": 870
|
| 1458 |
},
|
| 1459 |
{
|
| 1460 |
"epoch": 0.92,
|
| 1461 |
+
"grad_norm": 33.335658583470845,
|
| 1462 |
+
"learning_rate": 9.345903713082304e-09,
|
| 1463 |
+
"logits/chosen": -0.7445138692855835,
|
| 1464 |
+
"logits/rejected": 0.06519349664449692,
|
| 1465 |
+
"logps/chosen": -360.95751953125,
|
| 1466 |
+
"logps/rejected": -1240.344970703125,
|
| 1467 |
+
"loss": 0.1403,
|
| 1468 |
+
"rewards/accuracies": 0.918749988079071,
|
| 1469 |
+
"rewards/chosen": -2.1640257835388184,
|
| 1470 |
+
"rewards/margins": 9.403306007385254,
|
| 1471 |
+
"rewards/rejected": -11.567331314086914,
|
| 1472 |
+
"step": 880
|
| 1473 |
+
},
|
| 1474 |
+
{
|
| 1475 |
+
"epoch": 0.93,
|
| 1476 |
+
"grad_norm": 34.256779785154734,
|
| 1477 |
+
"learning_rate": 7.030787065396865e-09,
|
| 1478 |
+
"logits/chosen": -0.5979295969009399,
|
| 1479 |
+
"logits/rejected": 0.12500345706939697,
|
| 1480 |
+
"logps/chosen": -324.71563720703125,
|
| 1481 |
+
"logps/rejected": -1090.809814453125,
|
| 1482 |
+
"loss": 0.1529,
|
| 1483 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1484 |
+
"rewards/chosen": -1.9876606464385986,
|
| 1485 |
+
"rewards/margins": 8.175052642822266,
|
| 1486 |
+
"rewards/rejected": -10.162714004516602,
|
| 1487 |
+
"step": 890
|
| 1488 |
},
|
| 1489 |
{
|
| 1490 |
"epoch": 0.94,
|
| 1491 |
+
"grad_norm": 21.350853836162187,
|
| 1492 |
+
"learning_rate": 5.04062020432286e-09,
|
| 1493 |
+
"logits/chosen": -0.6045975685119629,
|
| 1494 |
+
"logits/rejected": 0.057004231959581375,
|
| 1495 |
+
"logps/chosen": -314.7482604980469,
|
| 1496 |
+
"logps/rejected": -1017.1994018554688,
|
| 1497 |
+
"loss": 0.1462,
|
| 1498 |
+
"rewards/accuracies": 0.90625,
|
| 1499 |
+
"rewards/chosen": -1.934511423110962,
|
| 1500 |
+
"rewards/margins": 7.4669060707092285,
|
| 1501 |
+
"rewards/rejected": -9.401416778564453,
|
| 1502 |
+
"step": 900
|
| 1503 |
+
},
|
| 1504 |
+
{
|
| 1505 |
+
"epoch": 0.94,
|
| 1506 |
+
"eval_logits/chosen": -0.5892084240913391,
|
| 1507 |
+
"eval_logits/rejected": -0.4189561903476715,
|
| 1508 |
+
"eval_logps/chosen": -1001.0963134765625,
|
| 1509 |
+
"eval_logps/rejected": -1160.1461181640625,
|
| 1510 |
+
"eval_loss": 2.023190498352051,
|
| 1511 |
+
"eval_rewards/accuracies": 0.61328125,
|
| 1512 |
+
"eval_rewards/chosen": -7.1946282386779785,
|
| 1513 |
+
"eval_rewards/margins": 1.5292174816131592,
|
| 1514 |
+
"eval_rewards/rejected": -8.723845481872559,
|
| 1515 |
+
"eval_runtime": 97.2614,
|
| 1516 |
+
"eval_samples_per_second": 20.563,
|
| 1517 |
+
"eval_steps_per_second": 0.329,
|
| 1518 |
+
"step": 900
|
| 1519 |
+
},
|
| 1520 |
+
{
|
| 1521 |
+
"epoch": 0.95,
|
| 1522 |
+
"grad_norm": 23.95340887933029,
|
| 1523 |
+
"learning_rate": 3.3780648016376866e-09,
|
| 1524 |
+
"logits/chosen": -0.6600544452667236,
|
| 1525 |
+
"logits/rejected": 0.006202346179634333,
|
| 1526 |
+
"logps/chosen": -325.9316101074219,
|
| 1527 |
+
"logps/rejected": -1070.564208984375,
|
| 1528 |
+
"loss": 0.1294,
|
| 1529 |
+
"rewards/accuracies": 0.918749988079071,
|
| 1530 |
+
"rewards/chosen": -2.046983480453491,
|
| 1531 |
+
"rewards/margins": 7.899454593658447,
|
| 1532 |
+
"rewards/rejected": -9.94643783569336,
|
| 1533 |
+
"step": 910
|
| 1534 |
},
|
| 1535 |
{
|
| 1536 |
"epoch": 0.96,
|
| 1537 |
+
"grad_norm": 27.425291781296558,
|
| 1538 |
+
"learning_rate": 2.0453443778310766e-09,
|
| 1539 |
+
"logits/chosen": -0.7382737994194031,
|
| 1540 |
+
"logits/rejected": -0.02898242510855198,
|
| 1541 |
+
"logps/chosen": -355.5832824707031,
|
| 1542 |
+
"logps/rejected": -1111.990234375,
|
| 1543 |
+
"loss": 0.159,
|
| 1544 |
+
"rewards/accuracies": 0.887499988079071,
|
| 1545 |
+
"rewards/chosen": -2.140547275543213,
|
| 1546 |
+
"rewards/margins": 8.157334327697754,
|
| 1547 |
+
"rewards/rejected": -10.297882080078125,
|
| 1548 |
+
"step": 920
|
| 1549 |
+
},
|
| 1550 |
+
{
|
| 1551 |
+
"epoch": 0.97,
|
| 1552 |
+
"grad_norm": 34.03050925733607,
|
| 1553 |
+
"learning_rate": 1.0442413283435758e-09,
|
| 1554 |
+
"logits/chosen": -0.6262997388839722,
|
| 1555 |
+
"logits/rejected": 0.14048056304454803,
|
| 1556 |
+
"logps/chosen": -344.0566101074219,
|
| 1557 |
+
"logps/rejected": -1069.548583984375,
|
| 1558 |
+
"loss": 0.1613,
|
| 1559 |
+
"rewards/accuracies": 0.8812500238418579,
|
| 1560 |
+
"rewards/chosen": -2.1957008838653564,
|
| 1561 |
+
"rewards/margins": 7.760766506195068,
|
| 1562 |
+
"rewards/rejected": -9.956467628479004,
|
| 1563 |
+
"step": 930
|
| 1564 |
},
|
| 1565 |
{
|
| 1566 |
"epoch": 0.98,
|
| 1567 |
+
"grad_norm": 34.773688515138254,
|
| 1568 |
+
"learning_rate": 3.760945397705828e-10,
|
| 1569 |
+
"logits/chosen": -0.7389376163482666,
|
| 1570 |
+
"logits/rejected": 0.0326564684510231,
|
| 1571 |
+
"logps/chosen": -334.1388854980469,
|
| 1572 |
+
"logps/rejected": -1029.4984130859375,
|
| 1573 |
+
"loss": 0.1497,
|
| 1574 |
+
"rewards/accuracies": 0.875,
|
| 1575 |
+
"rewards/chosen": -2.1344165802001953,
|
| 1576 |
+
"rewards/margins": 7.39968204498291,
|
| 1577 |
+
"rewards/rejected": -9.534098625183105,
|
| 1578 |
+
"step": 940
|
| 1579 |
+
},
|
| 1580 |
+
{
|
| 1581 |
+
"epoch": 0.99,
|
| 1582 |
+
"grad_norm": 25.746566103316148,
|
| 1583 |
+
"learning_rate": 4.17975992204056e-11,
|
| 1584 |
+
"logits/chosen": -0.5864154696464539,
|
| 1585 |
+
"logits/rejected": 0.14644786715507507,
|
| 1586 |
+
"logps/chosen": -315.15032958984375,
|
| 1587 |
+
"logps/rejected": -1021.39453125,
|
| 1588 |
+
"loss": 0.1545,
|
| 1589 |
+
"rewards/accuracies": 0.918749988079071,
|
| 1590 |
+
"rewards/chosen": -1.8772588968276978,
|
| 1591 |
+
"rewards/margins": 7.5670928955078125,
|
| 1592 |
+
"rewards/rejected": -9.444352149963379,
|
| 1593 |
+
"step": 950
|
| 1594 |
},
|
| 1595 |
{
|
| 1596 |
"epoch": 1.0,
|
| 1597 |
+
"step": 955,
|
| 1598 |
"total_flos": 0.0,
|
| 1599 |
+
"train_loss": 0.194384015168195,
|
| 1600 |
+
"train_runtime": 15417.2553,
|
| 1601 |
+
"train_samples_per_second": 7.931,
|
| 1602 |
+
"train_steps_per_second": 0.062
|
| 1603 |
}
|
| 1604 |
],
|
| 1605 |
"logging_steps": 10,
|
| 1606 |
+
"max_steps": 955,
|
| 1607 |
"num_input_tokens_seen": 0,
|
| 1608 |
"num_train_epochs": 1,
|
| 1609 |
"save_steps": 100,
|
training_args.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c0584d705e1910899a2cefd1070bfd070f2e8f13e53448281c5e88cfb5c6a342
|
| 3 |
+
size 6264
|