Model save
Browse files- README.md +1 -17
- all_results.json +5 -5
- model-00001-of-00003.safetensors +1 -1
- model-00002-of-00003.safetensors +1 -1
- model-00003-of-00003.safetensors +1 -1
- train_results.json +5 -5
- trainer_state.json +85 -719
- training_args.bin +1 -1
README.md
CHANGED
|
@@ -16,16 +16,6 @@ should probably proofread and complete it, then remove this comment. -->
|
|
| 16 |
# zephyr-7b-dpo-full
|
| 17 |
|
| 18 |
This model is a fine-tuned version of [HuggingFaceH4/mistral-7b-sft-beta](https://huggingface.co/HuggingFaceH4/mistral-7b-sft-beta) on the None dataset.
|
| 19 |
-
It achieves the following results on the evaluation set:
|
| 20 |
-
- Loss: 0.0270
|
| 21 |
-
- Rewards/chosen: -1.1958
|
| 22 |
-
- Rewards/rejected: -1.8757
|
| 23 |
-
- Rewards/accuracies: 0.7266
|
| 24 |
-
- Rewards/margins: 0.6799
|
| 25 |
-
- Logps/rejected: -444.9223
|
| 26 |
-
- Logps/chosen: -376.6192
|
| 27 |
-
- Logits/rejected: -2.4111
|
| 28 |
-
- Logits/chosen: -2.4251
|
| 29 |
|
| 30 |
## Model description
|
| 31 |
|
|
@@ -47,7 +37,7 @@ The following hyperparameters were used during training:
|
|
| 47 |
- learning_rate: 5e-07
|
| 48 |
- train_batch_size: 8
|
| 49 |
- eval_batch_size: 8
|
| 50 |
-
- seed:
|
| 51 |
- distributed_type: multi-GPU
|
| 52 |
- num_devices: 8
|
| 53 |
- gradient_accumulation_steps: 2
|
|
@@ -60,12 +50,6 @@ The following hyperparameters were used during training:
|
|
| 60 |
|
| 61 |
### Training results
|
| 62 |
|
| 63 |
-
| Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
|
| 64 |
-
|:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
|
| 65 |
-
| 0.0517 | 0.21 | 100 | 0.0427 | -0.6331 | -1.0207 | 0.7031 | 0.3876 | -359.4206 | -320.3455 | -2.7320 | -2.7523 |
|
| 66 |
-
| 0.0305 | 0.42 | 200 | 0.0297 | -1.0902 | -1.6743 | 0.7227 | 0.5842 | -424.7881 | -366.0565 | -2.5649 | -2.5797 |
|
| 67 |
-
| 0.0258 | 0.63 | 300 | 0.0274 | -1.2031 | -1.8719 | 0.7188 | 0.6688 | -444.5428 | -377.3524 | -2.4247 | -2.4384 |
|
| 68 |
-
| 0.0234 | 0.84 | 400 | 0.0270 | -1.1958 | -1.8757 | 0.7266 | 0.6799 | -444.9223 | -376.6192 | -2.4111 | -2.4251 |
|
| 69 |
|
| 70 |
|
| 71 |
### Framework versions
|
|
|
|
| 16 |
# zephyr-7b-dpo-full
|
| 17 |
|
| 18 |
This model is a fine-tuned version of [HuggingFaceH4/mistral-7b-sft-beta](https://huggingface.co/HuggingFaceH4/mistral-7b-sft-beta) on the None dataset.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 19 |
|
| 20 |
## Model description
|
| 21 |
|
|
|
|
| 37 |
- learning_rate: 5e-07
|
| 38 |
- train_batch_size: 8
|
| 39 |
- eval_batch_size: 8
|
| 40 |
+
- seed: 42
|
| 41 |
- distributed_type: multi-GPU
|
| 42 |
- num_devices: 8
|
| 43 |
- gradient_accumulation_steps: 2
|
|
|
|
| 50 |
|
| 51 |
### Training results
|
| 52 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 53 |
|
| 54 |
|
| 55 |
### 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": 0.3564033879424041,
|
| 4 |
+
"train_runtime": 425.1442,
|
| 5 |
+
"train_samples": 6750,
|
| 6 |
+
"train_samples_per_second": 15.877,
|
| 7 |
+
"train_steps_per_second": 0.125
|
| 8 |
}
|
model-00001-of-00003.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 4943162336
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:91328ad569624c8b14dcdb0ab4fea32f800ac871d8cb5af98a27b4b7452a2c7d
|
| 3 |
size 4943162336
|
model-00002-of-00003.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 4999819336
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8d23390b5bd7cb83885da59402667db713e5665a092d2a0df3084a70f535e0af
|
| 3 |
size 4999819336
|
model-00003-of-00003.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 4540516344
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cba4c24bce3c8d8bf4f70ba942459472c04e1245c651f98e9f1c6a6e2b7a4c6b
|
| 3 |
size 4540516344
|
train_results.json
CHANGED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
-
"train_loss": 0.
|
| 4 |
-
"train_runtime":
|
| 5 |
-
"train_samples":
|
| 6 |
-
"train_samples_per_second":
|
| 7 |
-
"train_steps_per_second": 0.
|
| 8 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"epoch": 1.0,
|
| 3 |
+
"train_loss": 0.3564033879424041,
|
| 4 |
+
"train_runtime": 425.1442,
|
| 5 |
+
"train_samples": 6750,
|
| 6 |
+
"train_samples_per_second": 15.877,
|
| 7 |
+
"train_steps_per_second": 0.125
|
| 8 |
}
|
trainer_state.json
CHANGED
|
@@ -3,19 +3,22 @@
|
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
"epoch": 1.0,
|
| 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 |
-
"
|
| 13 |
-
"
|
| 14 |
-
"
|
| 15 |
-
"
|
| 16 |
-
"
|
| 17 |
-
"
|
| 18 |
-
"
|
|
|
|
|
|
|
|
|
|
| 19 |
"rewards/accuracies": 0.0,
|
| 20 |
"rewards/chosen": 0.0,
|
| 21 |
"rewards/margins": 0.0,
|
|
@@ -23,739 +26,102 @@
|
|
| 23 |
"step": 1
|
| 24 |
},
|
| 25 |
{
|
| 26 |
-
"
|
| 27 |
-
"
|
| 28 |
-
"
|
| 29 |
-
"logits/rejected": -2.7507708072662354,
|
| 30 |
-
"logps/chosen": -275.7712707519531,
|
| 31 |
-
"logps/rejected": -253.4366912841797,
|
| 32 |
-
"loss": 0.1049,
|
| 33 |
-
"rewards/accuracies": 0.4305555522441864,
|
| 34 |
-
"rewards/chosen": -0.00010182532423641533,
|
| 35 |
-
"rewards/margins": -0.0002283406356582418,
|
| 36 |
-
"rewards/rejected": 0.00012651534052565694,
|
| 37 |
-
"step": 10
|
| 38 |
-
},
|
| 39 |
-
{
|
| 40 |
-
"epoch": 0.04,
|
| 41 |
-
"learning_rate": 2.0833333333333333e-07,
|
| 42 |
-
"logits/chosen": -2.7973473072052,
|
| 43 |
-
"logits/rejected": -2.7797961235046387,
|
| 44 |
-
"logps/chosen": -261.93072509765625,
|
| 45 |
-
"logps/rejected": -257.0435791015625,
|
| 46 |
-
"loss": 0.1048,
|
| 47 |
-
"rewards/accuracies": 0.59375,
|
| 48 |
-
"rewards/chosen": 0.00023481599055230618,
|
| 49 |
-
"rewards/margins": 0.0013333772076293826,
|
| 50 |
-
"rewards/rejected": -0.0010985612170770764,
|
| 51 |
-
"step": 20
|
| 52 |
-
},
|
| 53 |
-
{
|
| 54 |
-
"epoch": 0.06,
|
| 55 |
-
"learning_rate": 3.1249999999999997e-07,
|
| 56 |
-
"logits/chosen": -2.7847647666931152,
|
| 57 |
-
"logits/rejected": -2.778301239013672,
|
| 58 |
-
"logps/chosen": -295.0087890625,
|
| 59 |
-
"logps/rejected": -259.0351257324219,
|
| 60 |
-
"loss": 0.1026,
|
| 61 |
-
"rewards/accuracies": 0.6499999761581421,
|
| 62 |
-
"rewards/chosen": 0.0006326682632789016,
|
| 63 |
-
"rewards/margins": 0.008715003728866577,
|
| 64 |
-
"rewards/rejected": -0.008082335814833641,
|
| 65 |
-
"step": 30
|
| 66 |
-
},
|
| 67 |
-
{
|
| 68 |
-
"epoch": 0.08,
|
| 69 |
-
"learning_rate": 4.1666666666666667e-07,
|
| 70 |
-
"logits/chosen": -2.8043212890625,
|
| 71 |
-
"logits/rejected": -2.773319959640503,
|
| 72 |
-
"logps/chosen": -284.64312744140625,
|
| 73 |
-
"logps/rejected": -264.53228759765625,
|
| 74 |
-
"loss": 0.1011,
|
| 75 |
-
"rewards/accuracies": 0.675000011920929,
|
| 76 |
-
"rewards/chosen": -0.007321106735616922,
|
| 77 |
-
"rewards/margins": 0.018408337607979774,
|
| 78 |
-
"rewards/rejected": -0.02572944387793541,
|
| 79 |
-
"step": 40
|
| 80 |
-
},
|
| 81 |
-
{
|
| 82 |
-
"epoch": 0.1,
|
| 83 |
-
"learning_rate": 4.999733114418725e-07,
|
| 84 |
-
"logits/chosen": -2.779865264892578,
|
| 85 |
-
"logits/rejected": -2.7175371646881104,
|
| 86 |
-
"logps/chosen": -283.73773193359375,
|
| 87 |
-
"logps/rejected": -253.7095184326172,
|
| 88 |
-
"loss": 0.0944,
|
| 89 |
-
"rewards/accuracies": 0.737500011920929,
|
| 90 |
-
"rewards/chosen": -0.013570243492722511,
|
| 91 |
-
"rewards/margins": 0.056820280849933624,
|
| 92 |
-
"rewards/rejected": -0.07039052993059158,
|
| 93 |
-
"step": 50
|
| 94 |
-
},
|
| 95 |
-
{
|
| 96 |
-
"epoch": 0.13,
|
| 97 |
-
"learning_rate": 4.990398100856366e-07,
|
| 98 |
-
"logits/chosen": -2.805795192718506,
|
| 99 |
-
"logits/rejected": -2.78430438041687,
|
| 100 |
-
"logps/chosen": -293.5281677246094,
|
| 101 |
-
"logps/rejected": -261.317626953125,
|
| 102 |
-
"loss": 0.0893,
|
| 103 |
-
"rewards/accuracies": 0.612500011920929,
|
| 104 |
-
"rewards/chosen": -0.06878744065761566,
|
| 105 |
-
"rewards/margins": 0.07276646792888641,
|
| 106 |
-
"rewards/rejected": -0.14155389368534088,
|
| 107 |
-
"step": 60
|
| 108 |
-
},
|
| 109 |
-
{
|
| 110 |
-
"epoch": 0.15,
|
| 111 |
-
"learning_rate": 4.967775735898179e-07,
|
| 112 |
-
"logits/chosen": -2.7008605003356934,
|
| 113 |
-
"logits/rejected": -2.6654582023620605,
|
| 114 |
-
"logps/chosen": -272.4254455566406,
|
| 115 |
-
"logps/rejected": -264.991455078125,
|
| 116 |
-
"loss": 0.0818,
|
| 117 |
-
"rewards/accuracies": 0.7124999761581421,
|
| 118 |
-
"rewards/chosen": -0.12376339733600616,
|
| 119 |
-
"rewards/margins": 0.13984574377536774,
|
| 120 |
-
"rewards/rejected": -0.2636091411113739,
|
| 121 |
-
"step": 70
|
| 122 |
-
},
|
| 123 |
-
{
|
| 124 |
-
"epoch": 0.17,
|
| 125 |
-
"learning_rate": 4.931986719649298e-07,
|
| 126 |
-
"logits/chosen": -2.7169277667999268,
|
| 127 |
-
"logits/rejected": -2.700068950653076,
|
| 128 |
-
"logps/chosen": -293.3257751464844,
|
| 129 |
-
"logps/rejected": -285.71270751953125,
|
| 130 |
-
"loss": 0.0738,
|
| 131 |
-
"rewards/accuracies": 0.65625,
|
| 132 |
-
"rewards/chosen": -0.2549767792224884,
|
| 133 |
-
"rewards/margins": 0.13869795203208923,
|
| 134 |
-
"rewards/rejected": -0.39367473125457764,
|
| 135 |
-
"step": 80
|
| 136 |
-
},
|
| 137 |
-
{
|
| 138 |
"epoch": 0.19,
|
| 139 |
-
"learning_rate": 4.
|
| 140 |
-
"logits/chosen": -2.
|
| 141 |
-
"logits/rejected": -2.
|
| 142 |
-
"logps/chosen": -
|
| 143 |
-
"logps/rejected": -
|
| 144 |
-
"loss": 0.
|
| 145 |
-
"rewards/accuracies": 0.
|
| 146 |
-
"rewards/chosen":
|
| 147 |
-
"rewards/margins": 0.
|
| 148 |
-
"rewards/rejected": -0.
|
| 149 |
-
"step":
|
| 150 |
-
},
|
| 151 |
-
{
|
| 152 |
-
"epoch": 0.21,
|
| 153 |
-
"learning_rate": 4.821741763807186e-07,
|
| 154 |
-
"logits/chosen": -2.6386475563049316,
|
| 155 |
-
"logits/rejected": -2.6311068534851074,
|
| 156 |
-
"logps/chosen": -348.7519226074219,
|
| 157 |
-
"logps/rejected": -341.0511169433594,
|
| 158 |
-
"loss": 0.0517,
|
| 159 |
-
"rewards/accuracies": 0.6312500238418579,
|
| 160 |
-
"rewards/chosen": -0.5442075729370117,
|
| 161 |
-
"rewards/margins": 0.2245761901140213,
|
| 162 |
-
"rewards/rejected": -0.7687837481498718,
|
| 163 |
-
"step": 100
|
| 164 |
-
},
|
| 165 |
-
{
|
| 166 |
-
"epoch": 0.21,
|
| 167 |
-
"eval_logits/chosen": -2.752265691757202,
|
| 168 |
-
"eval_logits/rejected": -2.7319908142089844,
|
| 169 |
-
"eval_logps/chosen": -320.34552001953125,
|
| 170 |
-
"eval_logps/rejected": -359.42059326171875,
|
| 171 |
-
"eval_loss": 0.04265438765287399,
|
| 172 |
-
"eval_rewards/accuracies": 0.703125,
|
| 173 |
-
"eval_rewards/chosen": -0.6330589056015015,
|
| 174 |
-
"eval_rewards/margins": 0.38761410117149353,
|
| 175 |
-
"eval_rewards/rejected": -1.0206730365753174,
|
| 176 |
-
"eval_runtime": 53.2642,
|
| 177 |
-
"eval_samples_per_second": 37.549,
|
| 178 |
-
"eval_steps_per_second": 0.601,
|
| 179 |
-
"step": 100
|
| 180 |
-
},
|
| 181 |
-
{
|
| 182 |
-
"epoch": 0.23,
|
| 183 |
-
"learning_rate": 4.747874028753375e-07,
|
| 184 |
-
"logits/chosen": -2.745008945465088,
|
| 185 |
-
"logits/rejected": -2.6902968883514404,
|
| 186 |
-
"logps/chosen": -342.7307434082031,
|
| 187 |
-
"logps/rejected": -338.11322021484375,
|
| 188 |
-
"loss": 0.0397,
|
| 189 |
-
"rewards/accuracies": 0.6875,
|
| 190 |
-
"rewards/chosen": -0.6825590133666992,
|
| 191 |
-
"rewards/margins": 0.41266337037086487,
|
| 192 |
-
"rewards/rejected": -1.0952222347259521,
|
| 193 |
-
"step": 110
|
| 194 |
-
},
|
| 195 |
-
{
|
| 196 |
-
"epoch": 0.25,
|
| 197 |
-
"learning_rate": 4.662012913161997e-07,
|
| 198 |
-
"logits/chosen": -2.6888184547424316,
|
| 199 |
-
"logits/rejected": -2.6683411598205566,
|
| 200 |
-
"logps/chosen": -343.5830078125,
|
| 201 |
-
"logps/rejected": -382.2731628417969,
|
| 202 |
-
"loss": 0.044,
|
| 203 |
-
"rewards/accuracies": 0.6937500238418579,
|
| 204 |
-
"rewards/chosen": -0.8013314008712769,
|
| 205 |
-
"rewards/margins": 0.41239532828330994,
|
| 206 |
-
"rewards/rejected": -1.2137267589569092,
|
| 207 |
-
"step": 120
|
| 208 |
-
},
|
| 209 |
-
{
|
| 210 |
-
"epoch": 0.27,
|
| 211 |
-
"learning_rate": 4.5646165232345103e-07,
|
| 212 |
-
"logits/chosen": -2.690122127532959,
|
| 213 |
-
"logits/rejected": -2.673074245452881,
|
| 214 |
-
"logps/chosen": -349.1435546875,
|
| 215 |
-
"logps/rejected": -378.6903381347656,
|
| 216 |
-
"loss": 0.0367,
|
| 217 |
-
"rewards/accuracies": 0.6937500238418579,
|
| 218 |
-
"rewards/chosen": -0.8628436326980591,
|
| 219 |
-
"rewards/margins": 0.46390500664711,
|
| 220 |
-
"rewards/rejected": -1.3267487287521362,
|
| 221 |
-
"step": 130
|
| 222 |
-
},
|
| 223 |
-
{
|
| 224 |
-
"epoch": 0.29,
|
| 225 |
-
"learning_rate": 4.456204510851956e-07,
|
| 226 |
-
"logits/chosen": -2.6322216987609863,
|
| 227 |
-
"logits/rejected": -2.615206480026245,
|
| 228 |
-
"logps/chosen": -335.52008056640625,
|
| 229 |
-
"logps/rejected": -349.85638427734375,
|
| 230 |
-
"loss": 0.0411,
|
| 231 |
-
"rewards/accuracies": 0.65625,
|
| 232 |
-
"rewards/chosen": -0.8122199773788452,
|
| 233 |
-
"rewards/margins": 0.3598429262638092,
|
| 234 |
-
"rewards/rejected": -1.1720629930496216,
|
| 235 |
-
"step": 140
|
| 236 |
-
},
|
| 237 |
-
{
|
| 238 |
-
"epoch": 0.31,
|
| 239 |
-
"learning_rate": 4.337355301007335e-07,
|
| 240 |
-
"logits/chosen": -2.546915292739868,
|
| 241 |
-
"logits/rejected": -2.5243468284606934,
|
| 242 |
-
"logps/chosen": -353.551025390625,
|
| 243 |
-
"logps/rejected": -377.07989501953125,
|
| 244 |
-
"loss": 0.0368,
|
| 245 |
-
"rewards/accuracies": 0.6625000238418579,
|
| 246 |
-
"rewards/chosen": -0.8309852480888367,
|
| 247 |
-
"rewards/margins": 0.3872799277305603,
|
| 248 |
-
"rewards/rejected": -1.2182650566101074,
|
| 249 |
-
"step": 150
|
| 250 |
-
},
|
| 251 |
-
{
|
| 252 |
-
"epoch": 0.33,
|
| 253 |
-
"learning_rate": 4.2087030056579986e-07,
|
| 254 |
-
"logits/chosen": -2.516842842102051,
|
| 255 |
-
"logits/rejected": -2.509202480316162,
|
| 256 |
-
"logps/chosen": -396.78466796875,
|
| 257 |
-
"logps/rejected": -430.03662109375,
|
| 258 |
-
"loss": 0.0276,
|
| 259 |
-
"rewards/accuracies": 0.637499988079071,
|
| 260 |
-
"rewards/chosen": -1.324944257736206,
|
| 261 |
-
"rewards/margins": 0.2987218201160431,
|
| 262 |
-
"rewards/rejected": -1.6236661672592163,
|
| 263 |
-
"step": 160
|
| 264 |
-
},
|
| 265 |
-
{
|
| 266 |
-
"epoch": 0.36,
|
| 267 |
-
"learning_rate": 4.070934040463998e-07,
|
| 268 |
-
"logits/chosen": -2.486632823944092,
|
| 269 |
-
"logits/rejected": -2.4413046836853027,
|
| 270 |
-
"logps/chosen": -419.40582275390625,
|
| 271 |
-
"logps/rejected": -413.1378479003906,
|
| 272 |
-
"loss": 0.0252,
|
| 273 |
-
"rewards/accuracies": 0.574999988079071,
|
| 274 |
-
"rewards/chosen": -1.4800517559051514,
|
| 275 |
-
"rewards/margins": 0.3067554235458374,
|
| 276 |
-
"rewards/rejected": -1.7868072986602783,
|
| 277 |
-
"step": 170
|
| 278 |
},
|
| 279 |
{
|
|
|
|
|
|
|
|
|
|
| 280 |
"epoch": 0.38,
|
| 281 |
-
"learning_rate": 3.
|
| 282 |
-
"logits/chosen": -2.
|
| 283 |
-
"logits/rejected": -2.
|
| 284 |
-
"logps/chosen": -
|
| 285 |
-
"logps/rejected": -
|
| 286 |
-
"loss": 0.
|
| 287 |
-
"rewards/accuracies": 0.6499999761581421,
|
| 288 |
-
"rewards/chosen": -1.3809983730316162,
|
| 289 |
-
"rewards/margins": 0.39020317792892456,
|
| 290 |
-
"rewards/rejected": -1.771201491355896,
|
| 291 |
-
"step": 180
|
| 292 |
-
},
|
| 293 |
-
{
|
| 294 |
-
"epoch": 0.4,
|
| 295 |
-
"learning_rate": 3.7710310482256523e-07,
|
| 296 |
-
"logits/chosen": -2.5819058418273926,
|
| 297 |
-
"logits/rejected": -2.580766439437866,
|
| 298 |
-
"logps/chosen": -367.2718505859375,
|
| 299 |
-
"logps/rejected": -395.99261474609375,
|
| 300 |
-
"loss": 0.0276,
|
| 301 |
-
"rewards/accuracies": 0.668749988079071,
|
| 302 |
-
"rewards/chosen": -1.1196916103363037,
|
| 303 |
-
"rewards/margins": 0.3849378228187561,
|
| 304 |
-
"rewards/rejected": -1.504629373550415,
|
| 305 |
-
"step": 190
|
| 306 |
-
},
|
| 307 |
-
{
|
| 308 |
-
"epoch": 0.42,
|
| 309 |
-
"learning_rate": 3.610497133404795e-07,
|
| 310 |
-
"logits/chosen": -2.5476508140563965,
|
| 311 |
-
"logits/rejected": -2.5647332668304443,
|
| 312 |
-
"logps/chosen": -385.6960754394531,
|
| 313 |
-
"logps/rejected": -413.998291015625,
|
| 314 |
-
"loss": 0.0305,
|
| 315 |
-
"rewards/accuracies": 0.6187499761581421,
|
| 316 |
-
"rewards/chosen": -1.0930745601654053,
|
| 317 |
-
"rewards/margins": 0.34844768047332764,
|
| 318 |
-
"rewards/rejected": -1.4415223598480225,
|
| 319 |
-
"step": 200
|
| 320 |
-
},
|
| 321 |
-
{
|
| 322 |
-
"epoch": 0.42,
|
| 323 |
-
"eval_logits/chosen": -2.579733371734619,
|
| 324 |
-
"eval_logits/rejected": -2.5648751258850098,
|
| 325 |
-
"eval_logps/chosen": -366.0565490722656,
|
| 326 |
-
"eval_logps/rejected": -424.7881164550781,
|
| 327 |
-
"eval_loss": 0.029730383306741714,
|
| 328 |
-
"eval_rewards/accuracies": 0.72265625,
|
| 329 |
-
"eval_rewards/chosen": -1.0901691913604736,
|
| 330 |
-
"eval_rewards/margins": 0.5841794013977051,
|
| 331 |
-
"eval_rewards/rejected": -1.6743484735488892,
|
| 332 |
-
"eval_runtime": 53.2173,
|
| 333 |
-
"eval_samples_per_second": 37.582,
|
| 334 |
-
"eval_steps_per_second": 0.601,
|
| 335 |
-
"step": 200
|
| 336 |
-
},
|
| 337 |
-
{
|
| 338 |
-
"epoch": 0.44,
|
| 339 |
-
"learning_rate": 3.4440382358952115e-07,
|
| 340 |
-
"logits/chosen": -2.49862003326416,
|
| 341 |
-
"logits/rejected": -2.4916858673095703,
|
| 342 |
-
"logps/chosen": -370.3592834472656,
|
| 343 |
-
"logps/rejected": -407.2472229003906,
|
| 344 |
-
"loss": 0.0299,
|
| 345 |
-
"rewards/accuracies": 0.675000011920929,
|
| 346 |
-
"rewards/chosen": -1.1277533769607544,
|
| 347 |
-
"rewards/margins": 0.5519038438796997,
|
| 348 |
-
"rewards/rejected": -1.679657220840454,
|
| 349 |
-
"step": 210
|
| 350 |
-
},
|
| 351 |
-
{
|
| 352 |
-
"epoch": 0.46,
|
| 353 |
-
"learning_rate": 3.272542485937368e-07,
|
| 354 |
-
"logits/chosen": -2.5778369903564453,
|
| 355 |
-
"logits/rejected": -2.558171510696411,
|
| 356 |
-
"logps/chosen": -380.861328125,
|
| 357 |
-
"logps/rejected": -401.4359130859375,
|
| 358 |
-
"loss": 0.0296,
|
| 359 |
-
"rewards/accuracies": 0.668749988079071,
|
| 360 |
-
"rewards/chosen": -1.1397533416748047,
|
| 361 |
-
"rewards/margins": 0.4481385350227356,
|
| 362 |
-
"rewards/rejected": -1.587891936302185,
|
| 363 |
-
"step": 220
|
| 364 |
-
},
|
| 365 |
-
{
|
| 366 |
-
"epoch": 0.48,
|
| 367 |
-
"learning_rate": 3.096924887558854e-07,
|
| 368 |
-
"logits/chosen": -2.5747199058532715,
|
| 369 |
-
"logits/rejected": -2.535149097442627,
|
| 370 |
-
"logps/chosen": -425.2906188964844,
|
| 371 |
-
"logps/rejected": -415.38677978515625,
|
| 372 |
-
"loss": 0.0278,
|
| 373 |
"rewards/accuracies": 0.6875,
|
| 374 |
-
"rewards/chosen":
|
| 375 |
-
"rewards/margins": 0.
|
| 376 |
-
"rewards/rejected": -
|
| 377 |
-
"step":
|
| 378 |
},
|
| 379 |
{
|
| 380 |
-
"
|
| 381 |
-
"
|
| 382 |
-
"
|
| 383 |
-
"
|
| 384 |
-
"
|
| 385 |
-
"
|
| 386 |
-
"
|
|
|
|
|
|
|
|
|
|
| 387 |
"rewards/accuracies": 0.6937500238418579,
|
| 388 |
-
"rewards/chosen":
|
| 389 |
-
"rewards/margins": 0.
|
| 390 |
-
"rewards/rejected": -
|
| 391 |
-
"step":
|
| 392 |
-
},
|
| 393 |
-
{
|
| 394 |
-
"epoch": 0.52,
|
| 395 |
-
"learning_rate": 2.7370891215954565e-07,
|
| 396 |
-
"logits/chosen": -2.432041645050049,
|
| 397 |
-
"logits/rejected": -2.4213993549346924,
|
| 398 |
-
"logps/chosen": -398.3387451171875,
|
| 399 |
-
"logps/rejected": -416.9798889160156,
|
| 400 |
-
"loss": 0.0264,
|
| 401 |
-
"rewards/accuracies": 0.706250011920929,
|
| 402 |
-
"rewards/chosen": -1.2123913764953613,
|
| 403 |
-
"rewards/margins": 0.5389462113380432,
|
| 404 |
-
"rewards/rejected": -1.7513376474380493,
|
| 405 |
-
"step": 250
|
| 406 |
-
},
|
| 407 |
-
{
|
| 408 |
-
"epoch": 0.54,
|
| 409 |
-
"learning_rate": 2.55479083351317e-07,
|
| 410 |
-
"logits/chosen": -2.477323532104492,
|
| 411 |
-
"logits/rejected": -2.4589943885803223,
|
| 412 |
-
"logps/chosen": -404.10638427734375,
|
| 413 |
-
"logps/rejected": -437.82330322265625,
|
| 414 |
-
"loss": 0.0289,
|
| 415 |
-
"rewards/accuracies": 0.7437499761581421,
|
| 416 |
-
"rewards/chosen": -1.1455386877059937,
|
| 417 |
-
"rewards/margins": 0.6531764268875122,
|
| 418 |
-
"rewards/rejected": -1.7987149953842163,
|
| 419 |
-
"step": 260
|
| 420 |
-
},
|
| 421 |
-
{
|
| 422 |
-
"epoch": 0.56,
|
| 423 |
-
"learning_rate": 2.3722002126275822e-07,
|
| 424 |
-
"logits/chosen": -2.392612934112549,
|
| 425 |
-
"logits/rejected": -2.3851914405822754,
|
| 426 |
-
"logps/chosen": -389.7043151855469,
|
| 427 |
-
"logps/rejected": -410.2840881347656,
|
| 428 |
-
"loss": 0.0276,
|
| 429 |
-
"rewards/accuracies": 0.65625,
|
| 430 |
-
"rewards/chosen": -1.2070677280426025,
|
| 431 |
-
"rewards/margins": 0.4757324755191803,
|
| 432 |
-
"rewards/rejected": -1.68280029296875,
|
| 433 |
-
"step": 270
|
| 434 |
-
},
|
| 435 |
-
{
|
| 436 |
-
"epoch": 0.59,
|
| 437 |
-
"learning_rate": 2.19029145890313e-07,
|
| 438 |
-
"logits/chosen": -2.4097766876220703,
|
| 439 |
-
"logits/rejected": -2.387056350708008,
|
| 440 |
-
"logps/chosen": -399.73663330078125,
|
| 441 |
-
"logps/rejected": -442.6588439941406,
|
| 442 |
-
"loss": 0.0236,
|
| 443 |
-
"rewards/accuracies": 0.7250000238418579,
|
| 444 |
-
"rewards/chosen": -1.4455411434173584,
|
| 445 |
-
"rewards/margins": 0.6014262437820435,
|
| 446 |
-
"rewards/rejected": -2.046967029571533,
|
| 447 |
-
"step": 280
|
| 448 |
-
},
|
| 449 |
-
{
|
| 450 |
-
"epoch": 0.61,
|
| 451 |
-
"learning_rate": 2.0100351342479216e-07,
|
| 452 |
-
"logits/chosen": -2.4165215492248535,
|
| 453 |
-
"logits/rejected": -2.3698172569274902,
|
| 454 |
-
"logps/chosen": -431.76922607421875,
|
| 455 |
-
"logps/rejected": -444.64080810546875,
|
| 456 |
-
"loss": 0.025,
|
| 457 |
-
"rewards/accuracies": 0.706250011920929,
|
| 458 |
-
"rewards/chosen": -1.4643802642822266,
|
| 459 |
-
"rewards/margins": 0.5780037045478821,
|
| 460 |
-
"rewards/rejected": -2.0423836708068848,
|
| 461 |
-
"step": 290
|
| 462 |
-
},
|
| 463 |
-
{
|
| 464 |
-
"epoch": 0.63,
|
| 465 |
-
"learning_rate": 1.8323929841460178e-07,
|
| 466 |
-
"logits/chosen": -2.457998275756836,
|
| 467 |
-
"logits/rejected": -2.436868190765381,
|
| 468 |
-
"logps/chosen": -404.76409912109375,
|
| 469 |
-
"logps/rejected": -460.781005859375,
|
| 470 |
-
"loss": 0.0258,
|
| 471 |
-
"rewards/accuracies": 0.7250000238418579,
|
| 472 |
-
"rewards/chosen": -1.1497269868850708,
|
| 473 |
-
"rewards/margins": 0.6188378930091858,
|
| 474 |
-
"rewards/rejected": -1.7685649394989014,
|
| 475 |
-
"step": 300
|
| 476 |
-
},
|
| 477 |
-
{
|
| 478 |
-
"epoch": 0.63,
|
| 479 |
-
"eval_logits/chosen": -2.438358783721924,
|
| 480 |
-
"eval_logits/rejected": -2.4246902465820312,
|
| 481 |
-
"eval_logps/chosen": -377.3524475097656,
|
| 482 |
-
"eval_logps/rejected": -444.54278564453125,
|
| 483 |
-
"eval_loss": 0.02741866372525692,
|
| 484 |
-
"eval_rewards/accuracies": 0.71875,
|
| 485 |
-
"eval_rewards/chosen": -1.2031283378601074,
|
| 486 |
-
"eval_rewards/margins": 0.668766438961029,
|
| 487 |
-
"eval_rewards/rejected": -1.8718948364257812,
|
| 488 |
-
"eval_runtime": 53.272,
|
| 489 |
-
"eval_samples_per_second": 37.543,
|
| 490 |
-
"eval_steps_per_second": 0.601,
|
| 491 |
-
"step": 300
|
| 492 |
-
},
|
| 493 |
-
{
|
| 494 |
-
"epoch": 0.65,
|
| 495 |
-
"learning_rate": 1.6583128063291573e-07,
|
| 496 |
-
"logits/chosen": -2.4320359230041504,
|
| 497 |
-
"logits/rejected": -2.4065451622009277,
|
| 498 |
-
"logps/chosen": -390.3728332519531,
|
| 499 |
-
"logps/rejected": -426.6919860839844,
|
| 500 |
-
"loss": 0.025,
|
| 501 |
-
"rewards/accuracies": 0.668749988079071,
|
| 502 |
-
"rewards/chosen": -1.2841016054153442,
|
| 503 |
-
"rewards/margins": 0.5093994736671448,
|
| 504 |
-
"rewards/rejected": -1.7935012578964233,
|
| 505 |
-
"step": 310
|
| 506 |
-
},
|
| 507 |
-
{
|
| 508 |
-
"epoch": 0.67,
|
| 509 |
-
"learning_rate": 1.488723393865766e-07,
|
| 510 |
-
"logits/chosen": -2.36087965965271,
|
| 511 |
-
"logits/rejected": -2.3556385040283203,
|
| 512 |
-
"logps/chosen": -418.49493408203125,
|
| 513 |
-
"logps/rejected": -450.12139892578125,
|
| 514 |
-
"loss": 0.0232,
|
| 515 |
-
"rewards/accuracies": 0.675000011920929,
|
| 516 |
-
"rewards/chosen": -1.3573024272918701,
|
| 517 |
-
"rewards/margins": 0.4884418547153473,
|
| 518 |
-
"rewards/rejected": -1.8457443714141846,
|
| 519 |
-
"step": 320
|
| 520 |
-
},
|
| 521 |
-
{
|
| 522 |
-
"epoch": 0.69,
|
| 523 |
-
"learning_rate": 1.3245295796480788e-07,
|
| 524 |
-
"logits/chosen": -2.3685812950134277,
|
| 525 |
-
"logits/rejected": -2.3668787479400635,
|
| 526 |
-
"logps/chosen": -373.538330078125,
|
| 527 |
-
"logps/rejected": -445.90625,
|
| 528 |
-
"loss": 0.025,
|
| 529 |
-
"rewards/accuracies": 0.7250000238418579,
|
| 530 |
-
"rewards/chosen": -1.239685297012329,
|
| 531 |
-
"rewards/margins": 0.6630219221115112,
|
| 532 |
-
"rewards/rejected": -1.9027073383331299,
|
| 533 |
-
"step": 330
|
| 534 |
-
},
|
| 535 |
-
{
|
| 536 |
-
"epoch": 0.71,
|
| 537 |
-
"learning_rate": 1.1666074087171627e-07,
|
| 538 |
-
"logits/chosen": -2.38093900680542,
|
| 539 |
-
"logits/rejected": -2.3856873512268066,
|
| 540 |
-
"logps/chosen": -420.5098571777344,
|
| 541 |
-
"logps/rejected": -464.68243408203125,
|
| 542 |
-
"loss": 0.0268,
|
| 543 |
-
"rewards/accuracies": 0.71875,
|
| 544 |
-
"rewards/chosen": -1.2994890213012695,
|
| 545 |
-
"rewards/margins": 0.645952045917511,
|
| 546 |
-
"rewards/rejected": -1.9454412460327148,
|
| 547 |
-
"step": 340
|
| 548 |
-
},
|
| 549 |
-
{
|
| 550 |
-
"epoch": 0.73,
|
| 551 |
-
"learning_rate": 1.0157994641835734e-07,
|
| 552 |
-
"logits/chosen": -2.4149842262268066,
|
| 553 |
-
"logits/rejected": -2.3791468143463135,
|
| 554 |
-
"logps/chosen": -378.4749450683594,
|
| 555 |
-
"logps/rejected": -417.32305908203125,
|
| 556 |
-
"loss": 0.0256,
|
| 557 |
-
"rewards/accuracies": 0.6812499761581421,
|
| 558 |
-
"rewards/chosen": -1.2759901285171509,
|
| 559 |
-
"rewards/margins": 0.5479945540428162,
|
| 560 |
-
"rewards/rejected": -1.8239845037460327,
|
| 561 |
-
"step": 350
|
| 562 |
},
|
| 563 |
{
|
|
|
|
|
|
|
|
|
|
| 564 |
"epoch": 0.75,
|
| 565 |
-
"learning_rate": 8.
|
| 566 |
-
"logits/chosen": -2.
|
| 567 |
-
"logits/rejected": -2.
|
| 568 |
-
"logps/chosen": -
|
| 569 |
-
"logps/rejected": -
|
| 570 |
-
"loss": 0.
|
| 571 |
-
"rewards/accuracies": 0.
|
| 572 |
-
"rewards/chosen": -
|
| 573 |
-
"rewards/margins": 0.
|
| 574 |
-
"rewards/rejected": -
|
| 575 |
-
"step":
|
| 576 |
-
},
|
| 577 |
-
{
|
| 578 |
-
"epoch": 0.77,
|
| 579 |
-
"learning_rate": 7.387025063449081e-08,
|
| 580 |
-
"logits/chosen": -2.397771120071411,
|
| 581 |
-
"logits/rejected": -2.362600326538086,
|
| 582 |
-
"logps/chosen": -412.84332275390625,
|
| 583 |
-
"logps/rejected": -450.31048583984375,
|
| 584 |
-
"loss": 0.0228,
|
| 585 |
-
"rewards/accuracies": 0.6937500238418579,
|
| 586 |
-
"rewards/chosen": -1.2637946605682373,
|
| 587 |
-
"rewards/margins": 0.5941613912582397,
|
| 588 |
-
"rewards/rejected": -1.8579561710357666,
|
| 589 |
-
"step": 370
|
| 590 |
-
},
|
| 591 |
-
{
|
| 592 |
-
"epoch": 0.79,
|
| 593 |
-
"learning_rate": 6.138919252022435e-08,
|
| 594 |
-
"logits/chosen": -2.4274802207946777,
|
| 595 |
-
"logits/rejected": -2.420586347579956,
|
| 596 |
-
"logps/chosen": -411.314208984375,
|
| 597 |
-
"logps/rejected": -444.90869140625,
|
| 598 |
-
"loss": 0.0251,
|
| 599 |
-
"rewards/accuracies": 0.7250000238418579,
|
| 600 |
-
"rewards/chosen": -1.256594181060791,
|
| 601 |
-
"rewards/margins": 0.5403616428375244,
|
| 602 |
-
"rewards/rejected": -1.7969558238983154,
|
| 603 |
-
"step": 380
|
| 604 |
-
},
|
| 605 |
-
{
|
| 606 |
-
"epoch": 0.82,
|
| 607 |
-
"learning_rate": 4.991445467064689e-08,
|
| 608 |
-
"logits/chosen": -2.3289995193481445,
|
| 609 |
-
"logits/rejected": -2.306408405303955,
|
| 610 |
-
"logps/chosen": -396.445556640625,
|
| 611 |
-
"logps/rejected": -452.8165588378906,
|
| 612 |
-
"loss": 0.024,
|
| 613 |
-
"rewards/accuracies": 0.737500011920929,
|
| 614 |
-
"rewards/chosen": -1.269308090209961,
|
| 615 |
-
"rewards/margins": 0.6829138994216919,
|
| 616 |
-
"rewards/rejected": -1.9522218704223633,
|
| 617 |
-
"step": 390
|
| 618 |
-
},
|
| 619 |
-
{
|
| 620 |
-
"epoch": 0.84,
|
| 621 |
-
"learning_rate": 3.9507259776993954e-08,
|
| 622 |
-
"logits/chosen": -2.41839337348938,
|
| 623 |
-
"logits/rejected": -2.396681547164917,
|
| 624 |
-
"logps/chosen": -400.5920715332031,
|
| 625 |
-
"logps/rejected": -426.2327575683594,
|
| 626 |
-
"loss": 0.0234,
|
| 627 |
-
"rewards/accuracies": 0.65625,
|
| 628 |
-
"rewards/chosen": -1.2493975162506104,
|
| 629 |
-
"rewards/margins": 0.4738248288631439,
|
| 630 |
-
"rewards/rejected": -1.7232223749160767,
|
| 631 |
-
"step": 400
|
| 632 |
-
},
|
| 633 |
-
{
|
| 634 |
-
"epoch": 0.84,
|
| 635 |
-
"eval_logits/chosen": -2.4251461029052734,
|
| 636 |
-
"eval_logits/rejected": -2.4111008644104004,
|
| 637 |
-
"eval_logps/chosen": -376.6192321777344,
|
| 638 |
-
"eval_logps/rejected": -444.92230224609375,
|
| 639 |
-
"eval_loss": 0.027030915021896362,
|
| 640 |
-
"eval_rewards/accuracies": 0.7265625,
|
| 641 |
-
"eval_rewards/chosen": -1.1957957744598389,
|
| 642 |
-
"eval_rewards/margins": 0.6798944473266602,
|
| 643 |
-
"eval_rewards/rejected": -1.8756903409957886,
|
| 644 |
-
"eval_runtime": 53.2021,
|
| 645 |
-
"eval_samples_per_second": 37.593,
|
| 646 |
-
"eval_steps_per_second": 0.601,
|
| 647 |
-
"step": 400
|
| 648 |
-
},
|
| 649 |
-
{
|
| 650 |
-
"epoch": 0.86,
|
| 651 |
-
"learning_rate": 3.022313472693447e-08,
|
| 652 |
-
"logits/chosen": -2.412111759185791,
|
| 653 |
-
"logits/rejected": -2.3663253784179688,
|
| 654 |
-
"logps/chosen": -387.081787109375,
|
| 655 |
-
"logps/rejected": -392.51336669921875,
|
| 656 |
-
"loss": 0.0248,
|
| 657 |
-
"rewards/accuracies": 0.71875,
|
| 658 |
-
"rewards/chosen": -1.1241955757141113,
|
| 659 |
-
"rewards/margins": 0.5859023332595825,
|
| 660 |
-
"rewards/rejected": -1.7100979089736938,
|
| 661 |
-
"step": 410
|
| 662 |
-
},
|
| 663 |
-
{
|
| 664 |
-
"epoch": 0.88,
|
| 665 |
-
"learning_rate": 2.2111614344599684e-08,
|
| 666 |
-
"logits/chosen": -2.420088291168213,
|
| 667 |
-
"logits/rejected": -2.3837363719940186,
|
| 668 |
-
"logps/chosen": -418.43719482421875,
|
| 669 |
-
"logps/rejected": -451.14996337890625,
|
| 670 |
-
"loss": 0.0256,
|
| 671 |
-
"rewards/accuracies": 0.71875,
|
| 672 |
-
"rewards/chosen": -1.2756112813949585,
|
| 673 |
-
"rewards/margins": 0.7357032895088196,
|
| 674 |
-
"rewards/rejected": -2.011314868927002,
|
| 675 |
-
"step": 420
|
| 676 |
-
},
|
| 677 |
-
{
|
| 678 |
-
"epoch": 0.9,
|
| 679 |
-
"learning_rate": 1.521597710086439e-08,
|
| 680 |
-
"logits/chosen": -2.376132011413574,
|
| 681 |
-
"logits/rejected": -2.380525827407837,
|
| 682 |
-
"logps/chosen": -408.94781494140625,
|
| 683 |
-
"logps/rejected": -458.7674865722656,
|
| 684 |
-
"loss": 0.0244,
|
| 685 |
-
"rewards/accuracies": 0.7124999761581421,
|
| 686 |
-
"rewards/chosen": -1.34002685546875,
|
| 687 |
-
"rewards/margins": 0.5449159741401672,
|
| 688 |
-
"rewards/rejected": -1.8849427700042725,
|
| 689 |
-
"step": 430
|
| 690 |
-
},
|
| 691 |
-
{
|
| 692 |
-
"epoch": 0.92,
|
| 693 |
-
"learning_rate": 9.57301420397924e-09,
|
| 694 |
-
"logits/chosen": -2.360464334487915,
|
| 695 |
-
"logits/rejected": -2.3589634895324707,
|
| 696 |
-
"logps/chosen": -394.01226806640625,
|
| 697 |
-
"logps/rejected": -432.554443359375,
|
| 698 |
-
"loss": 0.026,
|
| 699 |
-
"rewards/accuracies": 0.625,
|
| 700 |
-
"rewards/chosen": -1.3313329219818115,
|
| 701 |
-
"rewards/margins": 0.5157779455184937,
|
| 702 |
-
"rewards/rejected": -1.8471107482910156,
|
| 703 |
-
"step": 440
|
| 704 |
},
|
| 705 |
{
|
|
|
|
|
|
|
|
|
|
| 706 |
"epoch": 0.94,
|
| 707 |
-
"learning_rate": 5.
|
| 708 |
-
"logits/chosen": -2.
|
| 709 |
-
"logits/rejected": -2.
|
| 710 |
-
"logps/chosen": -
|
| 711 |
-
"logps/rejected": -
|
| 712 |
-
"loss": 0.
|
| 713 |
-
"rewards/accuracies": 0.675000011920929,
|
| 714 |
-
"rewards/chosen": -1.3699880838394165,
|
| 715 |
-
"rewards/margins": 0.4800630509853363,
|
| 716 |
-
"rewards/rejected": -1.8500511646270752,
|
| 717 |
-
"step": 450
|
| 718 |
-
},
|
| 719 |
-
{
|
| 720 |
-
"epoch": 0.96,
|
| 721 |
-
"learning_rate": 2.158697848236607e-09,
|
| 722 |
-
"logits/chosen": -2.3582141399383545,
|
| 723 |
-
"logits/rejected": -2.3275790214538574,
|
| 724 |
-
"logps/chosen": -391.3001403808594,
|
| 725 |
-
"logps/rejected": -415.3433532714844,
|
| 726 |
-
"loss": 0.0249,
|
| 727 |
-
"rewards/accuracies": 0.65625,
|
| 728 |
-
"rewards/chosen": -1.381425380706787,
|
| 729 |
-
"rewards/margins": 0.4009224474430084,
|
| 730 |
-
"rewards/rejected": -1.7823479175567627,
|
| 731 |
-
"step": 460
|
| 732 |
-
},
|
| 733 |
-
{
|
| 734 |
-
"epoch": 0.98,
|
| 735 |
-
"learning_rate": 4.269029751107489e-10,
|
| 736 |
-
"logits/chosen": -2.3417115211486816,
|
| 737 |
-
"logits/rejected": -2.3047232627868652,
|
| 738 |
-
"logps/chosen": -387.1515197753906,
|
| 739 |
-
"logps/rejected": -449.08837890625,
|
| 740 |
-
"loss": 0.0252,
|
| 741 |
"rewards/accuracies": 0.699999988079071,
|
| 742 |
-
"rewards/chosen":
|
| 743 |
-
"rewards/margins": 0.
|
| 744 |
-
"rewards/rejected": -
|
| 745 |
-
"step":
|
| 746 |
},
|
| 747 |
{
|
| 748 |
"epoch": 1.0,
|
| 749 |
-
"step":
|
| 750 |
"total_flos": 0.0,
|
| 751 |
-
"train_loss": 0.
|
| 752 |
-
"train_runtime":
|
| 753 |
-
"train_samples_per_second":
|
| 754 |
-
"train_steps_per_second": 0.
|
| 755 |
}
|
| 756 |
],
|
| 757 |
"logging_steps": 10,
|
| 758 |
-
"max_steps":
|
| 759 |
"num_train_epochs": 1,
|
| 760 |
"save_steps": 100,
|
| 761 |
"total_flos": 0.0,
|
|
|
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
"epoch": 1.0,
|
| 5 |
"eval_steps": 100,
|
| 6 |
+
"global_step": 53,
|
| 7 |
"is_hyper_param_search": false,
|
| 8 |
"is_local_process_zero": true,
|
| 9 |
"is_world_process_zero": true,
|
| 10 |
"log_history": [
|
| 11 |
{
|
| 12 |
+
"debug/losses": 0.342240571975708,
|
| 13 |
+
"debug/policy_weights": 0.4937487840652466,
|
| 14 |
+
"debug/raw_losses": 0.6931471824645996,
|
| 15 |
+
"epoch": 0.02,
|
| 16 |
+
"learning_rate": 8.333333333333333e-08,
|
| 17 |
+
"logits/chosen": -2.8462421894073486,
|
| 18 |
+
"logits/rejected": -2.8283610343933105,
|
| 19 |
+
"logps/chosen": -274.7393798828125,
|
| 20 |
+
"logps/rejected": -204.42575073242188,
|
| 21 |
+
"loss": 0.3624,
|
| 22 |
"rewards/accuracies": 0.0,
|
| 23 |
"rewards/chosen": 0.0,
|
| 24 |
"rewards/margins": 0.0,
|
|
|
|
| 26 |
"step": 1
|
| 27 |
},
|
| 28 |
{
|
| 29 |
+
"debug/losses": 0.3810771703720093,
|
| 30 |
+
"debug/policy_weights": 0.5504893660545349,
|
| 31 |
+
"debug/raw_losses": 0.6920116543769836,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 32 |
"epoch": 0.19,
|
| 33 |
+
"learning_rate": 4.911172937635942e-07,
|
| 34 |
+
"logits/chosen": -2.852349281311035,
|
| 35 |
+
"logits/rejected": -2.83735990524292,
|
| 36 |
+
"logps/chosen": -306.01458740234375,
|
| 37 |
+
"logps/rejected": -295.93804931640625,
|
| 38 |
+
"loss": 0.3763,
|
| 39 |
+
"rewards/accuracies": 0.4861111044883728,
|
| 40 |
+
"rewards/chosen": 0.0012468346394598484,
|
| 41 |
+
"rewards/margins": 0.0023373025469481945,
|
| 42 |
+
"rewards/rejected": -0.0010904677910730243,
|
| 43 |
+
"step": 10
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 44 |
},
|
| 45 |
{
|
| 46 |
+
"debug/losses": 0.3667130768299103,
|
| 47 |
+
"debug/policy_weights": 0.5492504835128784,
|
| 48 |
+
"debug/raw_losses": 0.6667538285255432,
|
| 49 |
"epoch": 0.38,
|
| 50 |
+
"learning_rate": 3.982949361823388e-07,
|
| 51 |
+
"logits/chosen": -2.8518126010894775,
|
| 52 |
+
"logits/rejected": -2.872077226638794,
|
| 53 |
+
"logps/chosen": -296.2869567871094,
|
| 54 |
+
"logps/rejected": -332.9769592285156,
|
| 55 |
+
"loss": 0.374,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 56 |
"rewards/accuracies": 0.6875,
|
| 57 |
+
"rewards/chosen": 0.017367612570524216,
|
| 58 |
+
"rewards/margins": 0.05825704336166382,
|
| 59 |
+
"rewards/rejected": -0.04088941961526871,
|
| 60 |
+
"step": 20
|
| 61 |
},
|
| 62 |
{
|
| 63 |
+
"debug/losses": 0.35409680008888245,
|
| 64 |
+
"debug/policy_weights": 0.5632873177528381,
|
| 65 |
+
"debug/raw_losses": 0.6232098340988159,
|
| 66 |
+
"epoch": 0.57,
|
| 67 |
+
"learning_rate": 2.416462557480814e-07,
|
| 68 |
+
"logits/chosen": -2.824850559234619,
|
| 69 |
+
"logits/rejected": -2.8103976249694824,
|
| 70 |
+
"logps/chosen": -312.3518981933594,
|
| 71 |
+
"logps/rejected": -323.0265197753906,
|
| 72 |
+
"loss": 0.3551,
|
| 73 |
"rewards/accuracies": 0.6937500238418579,
|
| 74 |
+
"rewards/chosen": 0.014037144370377064,
|
| 75 |
+
"rewards/margins": 0.1827748715877533,
|
| 76 |
+
"rewards/rejected": -0.1687377393245697,
|
| 77 |
+
"step": 30
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 78 |
},
|
| 79 |
{
|
| 80 |
+
"debug/losses": 0.34752941131591797,
|
| 81 |
+
"debug/policy_weights": 0.550245463848114,
|
| 82 |
+
"debug/raw_losses": 0.6313939690589905,
|
| 83 |
"epoch": 0.75,
|
| 84 |
+
"learning_rate": 8.859303711029939e-08,
|
| 85 |
+
"logits/chosen": -2.79345965385437,
|
| 86 |
+
"logits/rejected": -2.797208547592163,
|
| 87 |
+
"logps/chosen": -275.5638122558594,
|
| 88 |
+
"logps/rejected": -348.8089294433594,
|
| 89 |
+
"loss": 0.3454,
|
| 90 |
+
"rewards/accuracies": 0.6812499761581421,
|
| 91 |
+
"rewards/chosen": -0.0044286223128438,
|
| 92 |
+
"rewards/margins": 0.20287349820137024,
|
| 93 |
+
"rewards/rejected": -0.20730212330818176,
|
| 94 |
+
"step": 40
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 95 |
},
|
| 96 |
{
|
| 97 |
+
"debug/losses": 0.3295789361000061,
|
| 98 |
+
"debug/policy_weights": 0.5445331335067749,
|
| 99 |
+
"debug/raw_losses": 0.5927887558937073,
|
| 100 |
"epoch": 0.94,
|
| 101 |
+
"learning_rate": 5.009573740853313e-09,
|
| 102 |
+
"logits/chosen": -2.829876184463501,
|
| 103 |
+
"logits/rejected": -2.8419814109802246,
|
| 104 |
+
"logps/chosen": -307.4209899902344,
|
| 105 |
+
"logps/rejected": -348.12298583984375,
|
| 106 |
+
"loss": 0.3378,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 107 |
"rewards/accuracies": 0.699999988079071,
|
| 108 |
+
"rewards/chosen": 0.0008583018789067864,
|
| 109 |
+
"rewards/margins": 0.31627127528190613,
|
| 110 |
+
"rewards/rejected": -0.3154129683971405,
|
| 111 |
+
"step": 50
|
| 112 |
},
|
| 113 |
{
|
| 114 |
"epoch": 1.0,
|
| 115 |
+
"step": 53,
|
| 116 |
"total_flos": 0.0,
|
| 117 |
+
"train_loss": 0.3564033879424041,
|
| 118 |
+
"train_runtime": 425.1442,
|
| 119 |
+
"train_samples_per_second": 15.877,
|
| 120 |
+
"train_steps_per_second": 0.125
|
| 121 |
}
|
| 122 |
],
|
| 123 |
"logging_steps": 10,
|
| 124 |
+
"max_steps": 53,
|
| 125 |
"num_train_epochs": 1,
|
| 126 |
"save_steps": 100,
|
| 127 |
"total_flos": 0.0,
|
training_args.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 5944
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:badf95e1b28a5ba1204144bfc8bd0b89ed4f9c81844941c1b6c826fe492f8ed6
|
| 3 |
size 5944
|