Model save
Browse files- README.md +1 -1
- all_results.json +10 -4
- eval_results.json +8 -0
- model.safetensors +2 -2
- train_results.json +4 -4
- trainer_state.json +15 -771
- training_args.bin +1 -1
README.md
CHANGED
|
@@ -27,7 +27,7 @@ print(output["generated_text"])
|
|
| 27 |
|
| 28 |
## Training procedure
|
| 29 |
|
| 30 |
-
[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/yuanyige/OpenR1/runs/
|
| 31 |
|
| 32 |
|
| 33 |
This model was trained with GRPO, a method introduced in [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://huggingface.co/papers/2402.03300).
|
|
|
|
| 27 |
|
| 28 |
## Training procedure
|
| 29 |
|
| 30 |
+
[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/yuanyige/OpenR1/runs/uclfg0aq)
|
| 31 |
|
| 32 |
|
| 33 |
This model was trained with GRPO, a method introduced in [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://huggingface.co/papers/2402.03300).
|
all_results.json
CHANGED
|
@@ -1,8 +1,14 @@
|
|
| 1 |
{
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
"total_flos": 0.0,
|
| 3 |
-
"train_loss":
|
| 4 |
-
"train_runtime":
|
| 5 |
"train_samples": 72441,
|
| 6 |
-
"train_samples_per_second":
|
| 7 |
-
"train_steps_per_second": 0.
|
| 8 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"eval_loss": 0.1335800141096115,
|
| 3 |
+
"eval_model_preparation_time": 1.0474,
|
| 4 |
+
"eval_runtime": 40.0309,
|
| 5 |
+
"eval_samples": 99,
|
| 6 |
+
"eval_samples_per_second": 2.473,
|
| 7 |
+
"eval_steps_per_second": 0.025,
|
| 8 |
"total_flos": 0.0,
|
| 9 |
+
"train_loss": 18.18843185901642,
|
| 10 |
+
"train_runtime": 28.8641,
|
| 11 |
"train_samples": 72441,
|
| 12 |
+
"train_samples_per_second": 3.43,
|
| 13 |
+
"train_steps_per_second": 0.035
|
| 14 |
}
|
eval_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"eval_loss": 0.1335800141096115,
|
| 3 |
+
"eval_model_preparation_time": 1.0474,
|
| 4 |
+
"eval_runtime": 40.0309,
|
| 5 |
+
"eval_samples": 99,
|
| 6 |
+
"eval_samples_per_second": 2.473,
|
| 7 |
+
"eval_steps_per_second": 0.025
|
| 8 |
+
}
|
model.safetensors
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:dd924a11b4c220f385b51ffa522daea7c9f3d850e31b162bb5661df483c6d3ee
|
| 3 |
+
size 3087467144
|
train_results.json
CHANGED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"total_flos": 0.0,
|
| 3 |
-
"train_loss":
|
| 4 |
-
"train_runtime":
|
| 5 |
"train_samples": 72441,
|
| 6 |
-
"train_samples_per_second":
|
| 7 |
-
"train_steps_per_second": 0.
|
| 8 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"total_flos": 0.0,
|
| 3 |
+
"train_loss": 18.18843185901642,
|
| 4 |
+
"train_runtime": 28.8641,
|
| 5 |
"train_samples": 72441,
|
| 6 |
+
"train_samples_per_second": 3.43,
|
| 7 |
+
"train_steps_per_second": 0.035
|
| 8 |
}
|
trainer_state.json
CHANGED
|
@@ -1,787 +1,31 @@
|
|
| 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 |
-
"completion_length":
|
| 13 |
-
"epoch": 0
|
| 14 |
-
"
|
| 15 |
-
"
|
| 16 |
-
"
|
| 17 |
-
"
|
| 18 |
-
"
|
| 19 |
-
"
|
| 20 |
-
"rewards/accuracy_reward": 0.1477678638882935,
|
| 21 |
-
"rewards/format_reward": 0.5025669865310192,
|
| 22 |
-
"step": 5
|
| 23 |
-
},
|
| 24 |
-
{
|
| 25 |
-
"completion_length": 255.40380630493163,
|
| 26 |
-
"epoch": 0.0353356890459364,
|
| 27 |
-
"grad_norm": 0.23681369423866272,
|
| 28 |
-
"kl": 0.21269454956054687,
|
| 29 |
-
"learning_rate": 6.896551724137932e-06,
|
| 30 |
-
"loss": 0.1146,
|
| 31 |
-
"reward": 0.9584821842610836,
|
| 32 |
-
"reward_std": 0.2699831190519035,
|
| 33 |
-
"rewards/accuracy_reward": 0.076004467648454,
|
| 34 |
-
"rewards/format_reward": 0.8824777193367481,
|
| 35 |
-
"step": 10
|
| 36 |
-
},
|
| 37 |
-
{
|
| 38 |
-
"completion_length": 210.21340198516845,
|
| 39 |
-
"epoch": 0.053003533568904596,
|
| 40 |
-
"grad_norm": 1.0446707010269165,
|
| 41 |
-
"kl": 0.0361907958984375,
|
| 42 |
-
"learning_rate": 1.0344827586206898e-05,
|
| 43 |
-
"loss": 0.0624,
|
| 44 |
-
"reward": 1.0110491536557675,
|
| 45 |
-
"reward_std": 0.26576729528605936,
|
| 46 |
-
"rewards/accuracy_reward": 0.08939732502913103,
|
| 47 |
-
"rewards/format_reward": 0.9216518297791481,
|
| 48 |
-
"step": 15
|
| 49 |
-
},
|
| 50 |
-
{
|
| 51 |
-
"completion_length": 173.89855737686156,
|
| 52 |
-
"epoch": 0.0706713780918728,
|
| 53 |
-
"grad_norm": 0.2942740023136139,
|
| 54 |
-
"kl": 0.074810791015625,
|
| 55 |
-
"learning_rate": 1.3793103448275863e-05,
|
| 56 |
-
"loss": 0.032,
|
| 57 |
-
"reward": 1.073437547683716,
|
| 58 |
-
"reward_std": 0.23347075609490275,
|
| 59 |
-
"rewards/accuracy_reward": 0.11428571951109916,
|
| 60 |
-
"rewards/format_reward": 0.9591518238186836,
|
| 61 |
-
"step": 20
|
| 62 |
-
},
|
| 63 |
-
{
|
| 64 |
-
"completion_length": 220.5953229904175,
|
| 65 |
-
"epoch": 0.08833922261484099,
|
| 66 |
-
"grad_norm": 3.951831340789795,
|
| 67 |
-
"kl": 0.531695556640625,
|
| 68 |
-
"learning_rate": 1.7241379310344828e-05,
|
| 69 |
-
"loss": 0.0981,
|
| 70 |
-
"reward": 1.0161830835044383,
|
| 71 |
-
"reward_std": 0.3684789047576487,
|
| 72 |
-
"rewards/accuracy_reward": 0.16328125749714673,
|
| 73 |
-
"rewards/format_reward": 0.8529018230736256,
|
| 74 |
-
"step": 25
|
| 75 |
-
},
|
| 76 |
-
{
|
| 77 |
-
"completion_length": 261.87579288482664,
|
| 78 |
-
"epoch": 0.10600706713780919,
|
| 79 |
-
"grad_norm": 11.097806930541992,
|
| 80 |
-
"kl": 2.1247802734375,
|
| 81 |
-
"learning_rate": 1.999923511388017e-05,
|
| 82 |
-
"loss": 0.3122,
|
| 83 |
-
"reward": 0.7962053943425417,
|
| 84 |
-
"reward_std": 0.5473601501435041,
|
| 85 |
-
"rewards/accuracy_reward": 0.17433036593720316,
|
| 86 |
-
"rewards/format_reward": 0.6218750271946192,
|
| 87 |
-
"step": 30
|
| 88 |
-
},
|
| 89 |
-
{
|
| 90 |
-
"completion_length": 151.15904779434203,
|
| 91 |
-
"epoch": 0.12367491166077739,
|
| 92 |
-
"grad_norm": 8.74118709564209,
|
| 93 |
-
"kl": 0.6091796875,
|
| 94 |
-
"learning_rate": 1.9972476383747748e-05,
|
| 95 |
-
"loss": 0.1028,
|
| 96 |
-
"reward": 0.8780134357511997,
|
| 97 |
-
"reward_std": 0.5024180840700865,
|
| 98 |
-
"rewards/accuracy_reward": 0.18404018736910074,
|
| 99 |
-
"rewards/format_reward": 0.6939732424914837,
|
| 100 |
-
"step": 35
|
| 101 |
-
},
|
| 102 |
-
{
|
| 103 |
-
"completion_length": 113.44442443847656,
|
| 104 |
-
"epoch": 0.1413427561837456,
|
| 105 |
-
"grad_norm": 1.2186102867126465,
|
| 106 |
-
"kl": 0.50096435546875,
|
| 107 |
-
"learning_rate": 1.9907590277344582e-05,
|
| 108 |
-
"loss": 0.0013,
|
| 109 |
-
"reward": 1.1537946984171867,
|
| 110 |
-
"reward_std": 0.36896395664662124,
|
| 111 |
-
"rewards/accuracy_reward": 0.23616072488948703,
|
| 112 |
-
"rewards/format_reward": 0.9176339685916901,
|
| 113 |
-
"step": 40
|
| 114 |
-
},
|
| 115 |
-
{
|
| 116 |
-
"completion_length": 156.94252967834473,
|
| 117 |
-
"epoch": 0.15901060070671377,
|
| 118 |
-
"grad_norm": 3.4705989360809326,
|
| 119 |
-
"kl": 1.0330810546875,
|
| 120 |
-
"learning_rate": 1.9804824871166254e-05,
|
| 121 |
-
"loss": 0.0519,
|
| 122 |
-
"reward": 1.2366071969270707,
|
| 123 |
-
"reward_std": 0.28554527759552,
|
| 124 |
-
"rewards/accuracy_reward": 0.28035715613514184,
|
| 125 |
-
"rewards/format_reward": 0.9562500424683094,
|
| 126 |
-
"step": 45
|
| 127 |
-
},
|
| 128 |
-
{
|
| 129 |
-
"completion_length": 140.84342164993285,
|
| 130 |
-
"epoch": 0.17667844522968199,
|
| 131 |
-
"grad_norm": 3.2262566089630127,
|
| 132 |
-
"kl": 0.81142578125,
|
| 133 |
-
"learning_rate": 1.9664573064143604e-05,
|
| 134 |
-
"loss": 0.0412,
|
| 135 |
-
"reward": 1.210714339464903,
|
| 136 |
-
"reward_std": 0.3142393365502357,
|
| 137 |
-
"rewards/accuracy_reward": 0.2649553695693612,
|
| 138 |
-
"rewards/format_reward": 0.9457589693367481,
|
| 139 |
-
"step": 50
|
| 140 |
-
},
|
| 141 |
-
{
|
| 142 |
-
"completion_length": 135.22857761383057,
|
| 143 |
-
"epoch": 0.19434628975265017,
|
| 144 |
-
"grad_norm": 0.37104904651641846,
|
| 145 |
-
"kl": 0.295947265625,
|
| 146 |
-
"learning_rate": 1.948737107548771e-05,
|
| 147 |
-
"loss": 0.0008,
|
| 148 |
-
"reward": 1.2118304073810577,
|
| 149 |
-
"reward_std": 0.31936368104070423,
|
| 150 |
-
"rewards/accuracy_reward": 0.2750000124797225,
|
| 151 |
-
"rewards/format_reward": 0.9368304006755352,
|
| 152 |
-
"step": 55
|
| 153 |
-
},
|
| 154 |
-
{
|
| 155 |
-
"completion_length": 194.0175313949585,
|
| 156 |
-
"epoch": 0.21201413427561838,
|
| 157 |
-
"grad_norm": 79.59091186523438,
|
| 158 |
-
"kl": 2.442626953125,
|
| 159 |
-
"learning_rate": 1.9273896394584103e-05,
|
| 160 |
-
"loss": 0.4489,
|
| 161 |
-
"reward": 1.2225446961820126,
|
| 162 |
-
"reward_std": 0.3489018376916647,
|
| 163 |
-
"rewards/accuracy_reward": 0.2940848357975483,
|
| 164 |
-
"rewards/format_reward": 0.9284598678350449,
|
| 165 |
-
"step": 60
|
| 166 |
-
},
|
| 167 |
-
{
|
| 168 |
-
"completion_length": 299.74309253692627,
|
| 169 |
-
"epoch": 0.22968197879858657,
|
| 170 |
-
"grad_norm": 9.220332145690918,
|
| 171 |
-
"kl": 2.752783203125,
|
| 172 |
-
"learning_rate": 1.9024965190774262e-05,
|
| 173 |
-
"loss": 0.7219,
|
| 174 |
-
"reward": 1.0818080872297287,
|
| 175 |
-
"reward_std": 0.4944979056715965,
|
| 176 |
-
"rewards/accuracy_reward": 0.294196442887187,
|
| 177 |
-
"rewards/format_reward": 0.7876116424798966,
|
| 178 |
-
"step": 65
|
| 179 |
-
},
|
| 180 |
-
{
|
| 181 |
-
"completion_length": 347.3866226196289,
|
| 182 |
-
"epoch": 0.24734982332155478,
|
| 183 |
-
"grad_norm": 8.275785446166992,
|
| 184 |
-
"kl": 5.707421875,
|
| 185 |
-
"learning_rate": 1.8741529192927528e-05,
|
| 186 |
-
"loss": 0.9457,
|
| 187 |
-
"reward": 1.0705357603728771,
|
| 188 |
-
"reward_std": 0.511051194742322,
|
| 189 |
-
"rewards/accuracy_reward": 0.30234376471489666,
|
| 190 |
-
"rewards/format_reward": 0.7681920006871223,
|
| 191 |
-
"step": 70
|
| 192 |
-
},
|
| 193 |
-
{
|
| 194 |
-
"completion_length": 258.911171913147,
|
| 195 |
-
"epoch": 0.26501766784452296,
|
| 196 |
-
"grad_norm": 6.921362400054932,
|
| 197 |
-
"kl": 1.3368896484375,
|
| 198 |
-
"learning_rate": 1.8424672050733577e-05,
|
| 199 |
-
"loss": 0.4684,
|
| 200 |
-
"reward": 1.2554688058793544,
|
| 201 |
-
"reward_std": 0.38661765307188034,
|
| 202 |
-
"rewards/accuracy_reward": 0.3574776967987418,
|
| 203 |
-
"rewards/format_reward": 0.8979911111295223,
|
| 204 |
-
"step": 75
|
| 205 |
-
},
|
| 206 |
-
{
|
| 207 |
-
"completion_length": 274.3730037689209,
|
| 208 |
-
"epoch": 0.2826855123674912,
|
| 209 |
-
"grad_norm": 2.2994608879089355,
|
| 210 |
-
"kl": 0.790625,
|
| 211 |
-
"learning_rate": 1.8075605191627242e-05,
|
| 212 |
-
"loss": 0.4752,
|
| 213 |
-
"reward": 1.2005580842494965,
|
| 214 |
-
"reward_std": 0.4162170600146055,
|
| 215 |
-
"rewards/accuracy_reward": 0.3270089436322451,
|
| 216 |
-
"rewards/format_reward": 0.873549147695303,
|
| 217 |
-
"step": 80
|
| 218 |
-
},
|
| 219 |
-
{
|
| 220 |
-
"completion_length": 333.57389736175537,
|
| 221 |
-
"epoch": 0.3003533568904594,
|
| 222 |
-
"grad_norm": 0.5888619422912598,
|
| 223 |
-
"kl": 0.4383544921875,
|
| 224 |
-
"learning_rate": 1.7695663189185703e-05,
|
| 225 |
-
"loss": 0.6874,
|
| 226 |
-
"reward": 1.0303571939468383,
|
| 227 |
-
"reward_std": 0.513068362697959,
|
| 228 |
-
"rewards/accuracy_reward": 0.2581473340280354,
|
| 229 |
-
"rewards/format_reward": 0.7722098581492901,
|
| 230 |
-
"step": 85
|
| 231 |
-
},
|
| 232 |
-
{
|
| 233 |
-
"completion_length": 218.5093852996826,
|
| 234 |
-
"epoch": 0.31802120141342755,
|
| 235 |
-
"grad_norm": 607.1968994140625,
|
| 236 |
-
"kl": 4.6947509765625,
|
| 237 |
-
"learning_rate": 1.7286298660705877e-05,
|
| 238 |
-
"loss": 0.9755,
|
| 239 |
-
"reward": 1.2218750618398189,
|
| 240 |
-
"reward_std": 0.365594344958663,
|
| 241 |
-
"rewards/accuracy_reward": 0.31350447833538053,
|
| 242 |
-
"rewards/format_reward": 0.9083705767989159,
|
| 243 |
-
"step": 90
|
| 244 |
-
},
|
| 245 |
-
{
|
| 246 |
-
"completion_length": 181.7823745727539,
|
| 247 |
-
"epoch": 0.33568904593639576,
|
| 248 |
-
"grad_norm": 7.599749565124512,
|
| 249 |
-
"kl": 0.5159912109375,
|
| 250 |
-
"learning_rate": 1.6849076713469914e-05,
|
| 251 |
-
"loss": 0.2169,
|
| 252 |
-
"reward": 1.2810268446803093,
|
| 253 |
-
"reward_std": 0.3074988966807723,
|
| 254 |
-
"rewards/accuracy_reward": 0.3281250137835741,
|
| 255 |
-
"rewards/format_reward": 0.952901828289032,
|
| 256 |
-
"step": 95
|
| 257 |
-
},
|
| 258 |
-
{
|
| 259 |
-
"completion_length": 175.87612476348878,
|
| 260 |
-
"epoch": 0.35335689045936397,
|
| 261 |
-
"grad_norm": 1.6138108968734741,
|
| 262 |
-
"kl": 0.40145263671875,
|
| 263 |
-
"learning_rate": 1.6385668960932143e-05,
|
| 264 |
-
"loss": 0.1372,
|
| 265 |
-
"reward": 1.2872768446803093,
|
| 266 |
-
"reward_std": 0.28329721204936503,
|
| 267 |
-
"rewards/accuracy_reward": 0.325446442887187,
|
| 268 |
-
"rewards/format_reward": 0.9618303947150707,
|
| 269 |
-
"step": 100
|
| 270 |
-
},
|
| 271 |
-
{
|
| 272 |
-
"epoch": 0.35335689045936397,
|
| 273 |
-
"eval_completion_length": 172.60817337036133,
|
| 274 |
-
"eval_kl": 2.84375,
|
| 275 |
-
"eval_loss": 0.4997411072254181,
|
| 276 |
-
"eval_reward": 1.262276828289032,
|
| 277 |
-
"eval_reward_std": 0.4010584279894829,
|
| 278 |
-
"eval_rewards/accuracy_reward": 0.3392857238650322,
|
| 279 |
-
"eval_rewards/format_reward": 0.9229911118745804,
|
| 280 |
-
"eval_runtime": 42.145,
|
| 281 |
-
"eval_samples_per_second": 2.349,
|
| 282 |
-
"eval_steps_per_second": 0.024,
|
| 283 |
-
"step": 100
|
| 284 |
-
},
|
| 285 |
-
{
|
| 286 |
-
"completion_length": 202.6909688949585,
|
| 287 |
-
"epoch": 0.3710247349823322,
|
| 288 |
-
"grad_norm": 3.8880608081817627,
|
| 289 |
-
"kl": 1.9408203125,
|
| 290 |
-
"learning_rate": 1.5897847131705194e-05,
|
| 291 |
-
"loss": 0.5156,
|
| 292 |
-
"reward": 1.182031300663948,
|
| 293 |
-
"reward_std": 0.4068946409970522,
|
| 294 |
-
"rewards/accuracy_reward": 0.2914062632247806,
|
| 295 |
-
"rewards/format_reward": 0.8906250417232513,
|
| 296 |
-
"step": 105
|
| 297 |
-
},
|
| 298 |
-
{
|
| 299 |
-
"completion_length": 234.45369472503663,
|
| 300 |
-
"epoch": 0.38869257950530034,
|
| 301 |
-
"grad_norm": 10.51020336151123,
|
| 302 |
-
"kl": 3.070166015625,
|
| 303 |
-
"learning_rate": 1.5387476295779737e-05,
|
| 304 |
-
"loss": 0.6781,
|
| 305 |
-
"reward": 1.1487723655998707,
|
| 306 |
-
"reward_std": 0.4561993557959795,
|
| 307 |
-
"rewards/accuracy_reward": 0.290513406926766,
|
| 308 |
-
"rewards/format_reward": 0.858258967846632,
|
| 309 |
-
"step": 110
|
| 310 |
-
},
|
| 311 |
-
{
|
| 312 |
-
"completion_length": 270.6211072921753,
|
| 313 |
-
"epoch": 0.40636042402826855,
|
| 314 |
-
"grad_norm": 17.52117919921875,
|
| 315 |
-
"kl": 3.108984375,
|
| 316 |
-
"learning_rate": 1.4856507733875837e-05,
|
| 317 |
-
"loss": 0.7116,
|
| 318 |
-
"reward": 1.1513393305242061,
|
| 319 |
-
"reward_std": 0.4505001582205296,
|
| 320 |
-
"rewards/accuracy_reward": 0.28861608654260634,
|
| 321 |
-
"rewards/format_reward": 0.8627232544124126,
|
| 322 |
-
"step": 115
|
| 323 |
-
},
|
| 324 |
-
{
|
| 325 |
-
"completion_length": 293.4304817199707,
|
| 326 |
-
"epoch": 0.42402826855123676,
|
| 327 |
-
"grad_norm": 3.0901668071746826,
|
| 328 |
-
"kl": 1.826953125,
|
| 329 |
-
"learning_rate": 1.4306971477188223e-05,
|
| 330 |
-
"loss": 0.6357,
|
| 331 |
-
"reward": 1.160714340209961,
|
| 332 |
-
"reward_std": 0.4445430090650916,
|
| 333 |
-
"rewards/accuracy_reward": 0.29687501220032575,
|
| 334 |
-
"rewards/format_reward": 0.863839327543974,
|
| 335 |
-
"step": 120
|
| 336 |
-
},
|
| 337 |
-
{
|
| 338 |
-
"completion_length": 293.4064878463745,
|
| 339 |
-
"epoch": 0.4416961130742049,
|
| 340 |
-
"grad_norm": 1.0616240501403809,
|
| 341 |
-
"kl": 0.759033203125,
|
| 342 |
-
"learning_rate": 1.3740968546047935e-05,
|
| 343 |
-
"loss": 0.5418,
|
| 344 |
-
"reward": 1.1334821924567222,
|
| 345 |
-
"reward_std": 0.4580180231481791,
|
| 346 |
-
"rewards/accuracy_reward": 0.2792410831898451,
|
| 347 |
-
"rewards/format_reward": 0.8542411103844643,
|
| 348 |
-
"step": 125
|
| 349 |
-
},
|
| 350 |
-
{
|
| 351 |
-
"completion_length": 187.28449478149415,
|
| 352 |
-
"epoch": 0.45936395759717313,
|
| 353 |
-
"grad_norm": 1.032782793045044,
|
| 354 |
-
"kl": 1.75234375,
|
| 355 |
-
"learning_rate": 1.3160662917174045e-05,
|
| 356 |
-
"loss": 0.2673,
|
| 357 |
-
"reward": 1.285825952887535,
|
| 358 |
-
"reward_std": 0.3237128121778369,
|
| 359 |
-
"rewards/accuracy_reward": 0.32957590855658053,
|
| 360 |
-
"rewards/format_reward": 0.9562500439584255,
|
| 361 |
-
"step": 130
|
| 362 |
-
},
|
| 363 |
-
{
|
| 364 |
-
"completion_length": 249.5670877456665,
|
| 365 |
-
"epoch": 0.47703180212014135,
|
| 366 |
-
"grad_norm": 5.672813892364502,
|
| 367 |
-
"kl": 0.9608154296875,
|
| 368 |
-
"learning_rate": 1.2568273250226681e-05,
|
| 369 |
-
"loss": 0.42,
|
| 370 |
-
"reward": 1.1869420193135738,
|
| 371 |
-
"reward_std": 0.41579487193375825,
|
| 372 |
-
"rewards/accuracy_reward": 0.30133930053561925,
|
| 373 |
-
"rewards/format_reward": 0.8856027238070965,
|
| 374 |
-
"step": 135
|
| 375 |
-
},
|
| 376 |
-
{
|
| 377 |
-
"completion_length": 251.65201930999757,
|
| 378 |
-
"epoch": 0.49469964664310956,
|
| 379 |
-
"grad_norm": 3.7146189212799072,
|
| 380 |
-
"kl": 0.66943359375,
|
| 381 |
-
"learning_rate": 1.1966064405292887e-05,
|
| 382 |
-
"loss": 0.2594,
|
| 383 |
-
"reward": 1.265178633481264,
|
| 384 |
-
"reward_std": 0.32337176529690626,
|
| 385 |
-
"rewards/accuracy_reward": 0.32433037273585796,
|
| 386 |
-
"rewards/format_reward": 0.9408482559025287,
|
| 387 |
-
"step": 140
|
| 388 |
-
},
|
| 389 |
-
{
|
| 390 |
-
"completion_length": 210.5714376449585,
|
| 391 |
-
"epoch": 0.5123674911660777,
|
| 392 |
-
"grad_norm": 2.63114333152771,
|
| 393 |
-
"kl": 0.3202392578125,
|
| 394 |
-
"learning_rate": 1.1356338783736256e-05,
|
| 395 |
-
"loss": 0.0326,
|
| 396 |
-
"reward": 1.3486607745289803,
|
| 397 |
-
"reward_std": 0.23780102925375105,
|
| 398 |
-
"rewards/accuracy_reward": 0.3533482296392322,
|
| 399 |
-
"rewards/format_reward": 0.9953125156462193,
|
| 400 |
-
"step": 145
|
| 401 |
-
},
|
| 402 |
-
{
|
| 403 |
-
"completion_length": 214.0686477661133,
|
| 404 |
-
"epoch": 0.5300353356890459,
|
| 405 |
-
"grad_norm": 1.1995089054107666,
|
| 406 |
-
"kl": 0.38431396484375,
|
| 407 |
-
"learning_rate": 1.0741427525516463e-05,
|
| 408 |
-
"loss": 0.114,
|
| 409 |
-
"reward": 1.2697545185685157,
|
| 410 |
-
"reward_std": 0.3020464600995183,
|
| 411 |
-
"rewards/accuracy_reward": 0.3006696555763483,
|
| 412 |
-
"rewards/format_reward": 0.9690848618745804,
|
| 413 |
-
"step": 150
|
| 414 |
-
},
|
| 415 |
-
{
|
| 416 |
-
"completion_length": 368.81084098815916,
|
| 417 |
-
"epoch": 0.5477031802120141,
|
| 418 |
-
"grad_norm": 6.58673095703125,
|
| 419 |
-
"kl": 1.377197265625,
|
| 420 |
-
"learning_rate": 1.012368159663363e-05,
|
| 421 |
-
"loss": 0.6451,
|
| 422 |
-
"reward": 0.9731027223169804,
|
| 423 |
-
"reward_std": 0.5320311084389686,
|
| 424 |
-
"rewards/accuracy_reward": 0.20669643813744187,
|
| 425 |
-
"rewards/format_reward": 0.7664062835276126,
|
| 426 |
-
"step": 155
|
| 427 |
-
},
|
| 428 |
-
{
|
| 429 |
-
"completion_length": 288.0265756607056,
|
| 430 |
-
"epoch": 0.5653710247349824,
|
| 431 |
-
"grad_norm": 1.0641926527023315,
|
| 432 |
-
"kl": 0.6106201171875,
|
| 433 |
-
"learning_rate": 9.505462800772612e-06,
|
| 434 |
-
"loss": 0.478,
|
| 435 |
-
"reward": 1.054352729767561,
|
| 436 |
-
"reward_std": 0.4532489124685526,
|
| 437 |
-
"rewards/accuracy_reward": 0.20569197335280479,
|
| 438 |
-
"rewards/format_reward": 0.8486607544124126,
|
| 439 |
-
"step": 160
|
| 440 |
-
},
|
| 441 |
-
{
|
| 442 |
-
"completion_length": 286.00916519165037,
|
| 443 |
-
"epoch": 0.5830388692579506,
|
| 444 |
-
"grad_norm": 0.44122907519340515,
|
| 445 |
-
"kl": 0.3580810546875,
|
| 446 |
-
"learning_rate": 8.889134749511956e-06,
|
| 447 |
-
"loss": 0.5047,
|
| 448 |
-
"reward": 1.1090402357280253,
|
| 449 |
-
"reward_std": 0.45567981265485286,
|
| 450 |
-
"rewards/accuracy_reward": 0.24263394130393862,
|
| 451 |
-
"rewards/format_reward": 0.8664062894880772,
|
| 452 |
-
"step": 165
|
| 453 |
-
},
|
| 454 |
-
{
|
| 455 |
-
"completion_length": 248.03639526367186,
|
| 456 |
-
"epoch": 0.6007067137809188,
|
| 457 |
-
"grad_norm": 0.31353357434272766,
|
| 458 |
-
"kl": 0.3409423828125,
|
| 459 |
-
"learning_rate": 8.277053825620836e-06,
|
| 460 |
-
"loss": 0.4022,
|
| 461 |
-
"reward": 1.20323666036129,
|
| 462 |
-
"reward_std": 0.39312053602188823,
|
| 463 |
-
"rewards/accuracy_reward": 0.2875000137835741,
|
| 464 |
-
"rewards/format_reward": 0.9157366491854191,
|
| 465 |
-
"step": 170
|
| 466 |
-
},
|
| 467 |
-
{
|
| 468 |
-
"completion_length": 187.5467725753784,
|
| 469 |
-
"epoch": 0.6183745583038869,
|
| 470 |
-
"grad_norm": 33.15998458862305,
|
| 471 |
-
"kl": 0.42301025390625,
|
| 472 |
-
"learning_rate": 7.671560173993588e-06,
|
| 473 |
-
"loss": 0.1912,
|
| 474 |
-
"reward": 1.291406311094761,
|
| 475 |
-
"reward_std": 0.2932738855481148,
|
| 476 |
-
"rewards/accuracy_reward": 0.3219866223633289,
|
| 477 |
-
"rewards/format_reward": 0.9694196812808513,
|
| 478 |
-
"step": 175
|
| 479 |
-
},
|
| 480 |
-
{
|
| 481 |
-
"completion_length": 185.00536499023437,
|
| 482 |
-
"epoch": 0.6360424028268551,
|
| 483 |
-
"grad_norm": 0.5558907985687256,
|
| 484 |
-
"kl": 0.245166015625,
|
| 485 |
-
"learning_rate": 7.07496875466589e-06,
|
| 486 |
-
"loss": 0.1924,
|
| 487 |
-
"reward": 1.2816964894533158,
|
| 488 |
-
"reward_std": 0.2955435009673238,
|
| 489 |
-
"rewards/accuracy_reward": 0.31372769344598056,
|
| 490 |
-
"rewards/format_reward": 0.9679687887430191,
|
| 491 |
-
"step": 180
|
| 492 |
-
},
|
| 493 |
-
{
|
| 494 |
-
"completion_length": 263.3887405395508,
|
| 495 |
-
"epoch": 0.6537102473498233,
|
| 496 |
-
"grad_norm": 0.36499133706092834,
|
| 497 |
-
"kl": 0.28150634765625,
|
| 498 |
-
"learning_rate": 6.489560492119225e-06,
|
| 499 |
-
"loss": 0.412,
|
| 500 |
-
"reward": 1.1849330902099608,
|
| 501 |
-
"reward_std": 0.40096613820642235,
|
| 502 |
-
"rewards/accuracy_reward": 0.29218751322478054,
|
| 503 |
-
"rewards/format_reward": 0.8927455753087997,
|
| 504 |
-
"step": 185
|
| 505 |
-
},
|
| 506 |
-
{
|
| 507 |
-
"completion_length": 304.0499008178711,
|
| 508 |
-
"epoch": 0.6713780918727915,
|
| 509 |
-
"grad_norm": 0.24227161705493927,
|
| 510 |
-
"kl": 0.26263427734375,
|
| 511 |
-
"learning_rate": 5.9175735547120975e-06,
|
| 512 |
-
"loss": 0.4397,
|
| 513 |
-
"reward": 1.183147370815277,
|
| 514 |
-
"reward_std": 0.44423861615359783,
|
| 515 |
-
"rewards/accuracy_reward": 0.3106026917695999,
|
| 516 |
-
"rewards/format_reward": 0.8725446827709675,
|
| 517 |
-
"step": 190
|
| 518 |
-
},
|
| 519 |
-
{
|
| 520 |
-
"completion_length": 281.2084943771362,
|
| 521 |
-
"epoch": 0.6890459363957597,
|
| 522 |
-
"grad_norm": 0.9491950273513794,
|
| 523 |
-
"kl": 0.2338623046875,
|
| 524 |
-
"learning_rate": 5.361194797579108e-06,
|
| 525 |
-
"loss": 0.2874,
|
| 526 |
-
"reward": 1.2875000551342963,
|
| 527 |
-
"reward_std": 0.3527799446135759,
|
| 528 |
-
"rewards/accuracy_reward": 0.3587053744122386,
|
| 529 |
-
"rewards/format_reward": 0.9287946857511997,
|
| 530 |
-
"step": 195
|
| 531 |
-
},
|
| 532 |
-
{
|
| 533 |
-
"completion_length": 279.403472328186,
|
| 534 |
-
"epoch": 0.7067137809187279,
|
| 535 |
-
"grad_norm": 0.24906866252422333,
|
| 536 |
-
"kl": 0.18758544921875,
|
| 537 |
-
"learning_rate": 4.8225514017138205e-06,
|
| 538 |
-
"loss": 0.226,
|
| 539 |
-
"reward": 1.3204241558909415,
|
| 540 |
-
"reward_std": 0.3341758308932185,
|
| 541 |
-
"rewards/accuracy_reward": 0.37712055407464506,
|
| 542 |
-
"rewards/format_reward": 0.9433036096394062,
|
| 543 |
-
"step": 200
|
| 544 |
-
},
|
| 545 |
-
{
|
| 546 |
-
"epoch": 0.7067137809187279,
|
| 547 |
-
"eval_completion_length": 290.95755767822266,
|
| 548 |
-
"eval_kl": 0.213623046875,
|
| 549 |
-
"eval_loss": 0.34684792160987854,
|
| 550 |
-
"eval_reward": 1.3035714626312256,
|
| 551 |
-
"eval_reward_std": 0.37863466888666153,
|
| 552 |
-
"eval_rewards/accuracy_reward": 0.3816964477300644,
|
| 553 |
-
"eval_rewards/format_reward": 0.9218750447034836,
|
| 554 |
-
"eval_runtime": 52.6337,
|
| 555 |
-
"eval_samples_per_second": 1.881,
|
| 556 |
-
"eval_steps_per_second": 0.019,
|
| 557 |
-
"step": 200
|
| 558 |
-
},
|
| 559 |
-
{
|
| 560 |
-
"completion_length": 314.0254592895508,
|
| 561 |
-
"epoch": 0.7243816254416962,
|
| 562 |
-
"grad_norm": 0.4518482983112335,
|
| 563 |
-
"kl": 0.26624755859375,
|
| 564 |
-
"learning_rate": 4.303702741201431e-06,
|
| 565 |
-
"loss": 0.3849,
|
| 566 |
-
"reward": 1.212611661851406,
|
| 567 |
-
"reward_std": 0.4183976389467716,
|
| 568 |
-
"rewards/accuracy_reward": 0.32354912124574187,
|
| 569 |
-
"rewards/format_reward": 0.8890625379979611,
|
| 570 |
-
"step": 205
|
| 571 |
-
},
|
| 572 |
-
{
|
| 573 |
-
"completion_length": 338.1247905731201,
|
| 574 |
-
"epoch": 0.7420494699646644,
|
| 575 |
-
"grad_norm": 0.29115650057792664,
|
| 576 |
-
"kl": 0.3578857421875,
|
| 577 |
-
"learning_rate": 3.8066325096949153e-06,
|
| 578 |
-
"loss": 0.4956,
|
| 579 |
-
"reward": 1.148772369325161,
|
| 580 |
-
"reward_std": 0.4888919793069363,
|
| 581 |
-
"rewards/accuracy_reward": 0.31272322945296765,
|
| 582 |
-
"rewards/format_reward": 0.8360491462051869,
|
| 583 |
-
"step": 210
|
| 584 |
-
},
|
| 585 |
-
{
|
| 586 |
-
"completion_length": 337.40280590057375,
|
| 587 |
-
"epoch": 0.7597173144876325,
|
| 588 |
-
"grad_norm": 0.4650420546531677,
|
| 589 |
-
"kl": 0.34840087890625,
|
| 590 |
-
"learning_rate": 3.3332411362372063e-06,
|
| 591 |
-
"loss": 0.5048,
|
| 592 |
-
"reward": 1.1222098745405673,
|
| 593 |
-
"reward_std": 0.4806524086743593,
|
| 594 |
-
"rewards/accuracy_reward": 0.29397322703152895,
|
| 595 |
-
"rewards/format_reward": 0.8282366462051869,
|
| 596 |
-
"step": 215
|
| 597 |
-
},
|
| 598 |
-
{
|
| 599 |
-
"completion_length": 303.6270210266113,
|
| 600 |
-
"epoch": 0.7773851590106007,
|
| 601 |
-
"grad_norm": 0.475253701210022,
|
| 602 |
-
"kl": 0.33944091796875,
|
| 603 |
-
"learning_rate": 2.8853385194256677e-06,
|
| 604 |
-
"loss": 0.4854,
|
| 605 |
-
"reward": 1.1648438036441804,
|
| 606 |
-
"reward_std": 0.4518149187788367,
|
| 607 |
-
"rewards/accuracy_reward": 0.3044643012806773,
|
| 608 |
-
"rewards/format_reward": 0.8603795073926449,
|
| 609 |
-
"step": 220
|
| 610 |
-
},
|
| 611 |
-
{
|
| 612 |
-
"completion_length": 294.54075050354004,
|
| 613 |
-
"epoch": 0.7950530035335689,
|
| 614 |
-
"grad_norm": 1.6670526266098022,
|
| 615 |
-
"kl": 0.4580810546875,
|
| 616 |
-
"learning_rate": 2.464637107698046e-06,
|
| 617 |
-
"loss": 0.4903,
|
| 618 |
-
"reward": 1.1580357693135739,
|
| 619 |
-
"reward_std": 0.44512549713253974,
|
| 620 |
-
"rewards/accuracy_reward": 0.2974330489523709,
|
| 621 |
-
"rewards/format_reward": 0.860602717846632,
|
| 622 |
-
"step": 225
|
| 623 |
-
},
|
| 624 |
-
{
|
| 625 |
-
"completion_length": 275.87791481018064,
|
| 626 |
-
"epoch": 0.8127208480565371,
|
| 627 |
-
"grad_norm": 1.4973307847976685,
|
| 628 |
-
"kl": 0.37625732421875,
|
| 629 |
-
"learning_rate": 2.072745352195794e-06,
|
| 630 |
-
"loss": 0.4606,
|
| 631 |
-
"reward": 1.1925223790109158,
|
| 632 |
-
"reward_std": 0.4279437720775604,
|
| 633 |
-
"rewards/accuracy_reward": 0.31104912217706443,
|
| 634 |
-
"rewards/format_reward": 0.8814732596278191,
|
| 635 |
-
"step": 230
|
| 636 |
-
},
|
| 637 |
-
{
|
| 638 |
-
"completion_length": 251.60960941314698,
|
| 639 |
-
"epoch": 0.8303886925795053,
|
| 640 |
-
"grad_norm": 30.014623641967773,
|
| 641 |
-
"kl": 0.34725341796875,
|
| 642 |
-
"learning_rate": 1.7111615572361628e-06,
|
| 643 |
-
"loss": 0.3987,
|
| 644 |
-
"reward": 1.243750061094761,
|
| 645 |
-
"reward_std": 0.41131251212209463,
|
| 646 |
-
"rewards/accuracy_reward": 0.33627233747392893,
|
| 647 |
-
"rewards/format_reward": 0.9074777193367481,
|
| 648 |
-
"step": 235
|
| 649 |
-
},
|
| 650 |
-
{
|
| 651 |
-
"completion_length": 248.69844818115234,
|
| 652 |
-
"epoch": 0.8480565371024735,
|
| 653 |
-
"grad_norm": 54.57441711425781,
|
| 654 |
-
"kl": 1.3016845703125,
|
| 655 |
-
"learning_rate": 1.381268151904298e-06,
|
| 656 |
-
"loss": 0.5182,
|
| 657 |
-
"reward": 1.236607201397419,
|
| 658 |
-
"reward_std": 0.40282299667596816,
|
| 659 |
-
"rewards/accuracy_reward": 0.3315848369151354,
|
| 660 |
-
"rewards/format_reward": 0.9050223655998707,
|
| 661 |
-
"step": 240
|
| 662 |
-
},
|
| 663 |
-
{
|
| 664 |
-
"completion_length": 247.95547924041747,
|
| 665 |
-
"epoch": 0.8657243816254417,
|
| 666 |
-
"grad_norm": 167.22845458984375,
|
| 667 |
-
"kl": 6.20146484375,
|
| 668 |
-
"learning_rate": 1.0843264046665558e-06,
|
| 669 |
-
"loss": 1.1488,
|
| 670 |
-
"reward": 1.2164063028991223,
|
| 671 |
-
"reward_std": 0.3988880608230829,
|
| 672 |
-
"rewards/accuracy_reward": 0.31004465725272895,
|
| 673 |
-
"rewards/format_reward": 0.9063616514205932,
|
| 674 |
-
"step": 245
|
| 675 |
-
},
|
| 676 |
-
{
|
| 677 |
-
"completion_length": 246.49398460388184,
|
| 678 |
-
"epoch": 0.8833922261484098,
|
| 679 |
-
"grad_norm": 48.688541412353516,
|
| 680 |
-
"kl": 1.783447265625,
|
| 681 |
-
"learning_rate": 8.214716012124491e-07,
|
| 682 |
-
"loss": 0.5749,
|
| 683 |
-
"reward": 1.2148438021540642,
|
| 684 |
-
"reward_std": 0.38329584859311583,
|
| 685 |
-
"rewards/accuracy_reward": 0.30535715632140636,
|
| 686 |
-
"rewards/format_reward": 0.9094866514205933,
|
| 687 |
-
"step": 250
|
| 688 |
-
},
|
| 689 |
-
{
|
| 690 |
-
"completion_length": 239.5714382171631,
|
| 691 |
-
"epoch": 0.901060070671378,
|
| 692 |
-
"grad_norm": 47.49518966674805,
|
| 693 |
-
"kl": 0.770556640625,
|
| 694 |
-
"learning_rate": 5.937087039615619e-07,
|
| 695 |
-
"loss": 0.4138,
|
| 696 |
-
"reward": 1.2323661297559738,
|
| 697 |
-
"reward_std": 0.3774874061346054,
|
| 698 |
-
"rewards/accuracy_reward": 0.3184151927940547,
|
| 699 |
-
"rewards/format_reward": 0.9139509372413158,
|
| 700 |
-
"step": 255
|
| 701 |
-
},
|
| 702 |
-
{
|
| 703 |
-
"completion_length": 249.4630693435669,
|
| 704 |
-
"epoch": 0.9187279151943463,
|
| 705 |
-
"grad_norm": 25.54094886779785,
|
| 706 |
-
"kl": 1.382958984375,
|
| 707 |
-
"learning_rate": 4.019085098303077e-07,
|
| 708 |
-
"loss": 0.5233,
|
| 709 |
-
"reward": 1.21964291036129,
|
| 710 |
-
"reward_std": 0.39937015902251005,
|
| 711 |
-
"rewards/accuracy_reward": 0.31763394493609665,
|
| 712 |
-
"rewards/format_reward": 0.9020089708268643,
|
| 713 |
-
"step": 260
|
| 714 |
-
},
|
| 715 |
-
{
|
| 716 |
-
"completion_length": 246.59643955230712,
|
| 717 |
-
"epoch": 0.9363957597173145,
|
| 718 |
-
"grad_norm": 25.444677352905273,
|
| 719 |
-
"kl": 2.867236328125,
|
| 720 |
-
"learning_rate": 2.4680432094837394e-07,
|
| 721 |
-
"loss": 0.6682,
|
| 722 |
-
"reward": 1.2183036252856254,
|
| 723 |
-
"reward_std": 0.40516539234668014,
|
| 724 |
-
"rewards/accuracy_reward": 0.31696430016309024,
|
| 725 |
-
"rewards/format_reward": 0.901339328289032,
|
| 726 |
-
"step": 265
|
| 727 |
-
},
|
| 728 |
-
{
|
| 729 |
-
"completion_length": 249.1561502456665,
|
| 730 |
-
"epoch": 0.9540636042402827,
|
| 731 |
-
"grad_norm": 5.572141647338867,
|
| 732 |
-
"kl": 2.645947265625,
|
| 733 |
-
"learning_rate": 1.289891410535593e-07,
|
| 734 |
-
"loss": 0.6572,
|
| 735 |
-
"reward": 1.2252232685685158,
|
| 736 |
-
"reward_std": 0.40847705118358135,
|
| 737 |
-
"rewards/accuracy_reward": 0.32165180034935476,
|
| 738 |
-
"rewards/format_reward": 0.9035714715719223,
|
| 739 |
-
"step": 270
|
| 740 |
-
},
|
| 741 |
-
{
|
| 742 |
-
"completion_length": 238.4615062713623,
|
| 743 |
-
"epoch": 0.9717314487632509,
|
| 744 |
-
"grad_norm": 17.332447052001953,
|
| 745 |
-
"kl": 1.982080078125,
|
| 746 |
-
"learning_rate": 4.8913408283934874e-08,
|
| 747 |
-
"loss": 0.5867,
|
| 748 |
-
"reward": 1.2314732655882836,
|
| 749 |
-
"reward_std": 0.38298603780567647,
|
| 750 |
-
"rewards/accuracy_reward": 0.32265626490116117,
|
| 751 |
-
"rewards/format_reward": 0.9088170036673546,
|
| 752 |
-
"step": 275
|
| 753 |
-
},
|
| 754 |
-
{
|
| 755 |
-
"completion_length": 242.24800186157228,
|
| 756 |
-
"epoch": 0.9893992932862191,
|
| 757 |
-
"grad_norm": 13.555401802062988,
|
| 758 |
-
"kl": 1.88583984375,
|
| 759 |
-
"learning_rate": 6.883273035447335e-09,
|
| 760 |
-
"loss": 0.5522,
|
| 761 |
-
"reward": 1.2324777357280254,
|
| 762 |
-
"reward_std": 0.3977079253643751,
|
| 763 |
-
"rewards/accuracy_reward": 0.32622769437730315,
|
| 764 |
-
"rewards/format_reward": 0.9062500409781933,
|
| 765 |
-
"step": 280
|
| 766 |
-
},
|
| 767 |
-
{
|
| 768 |
-
"completion_length": 240.61751333872476,
|
| 769 |
-
"epoch": 1.0,
|
| 770 |
-
"kl": 1.9344075520833333,
|
| 771 |
-
"reward": 1.2202381441990535,
|
| 772 |
-
"reward_std": 0.40573689217368764,
|
| 773 |
-
"rewards/accuracy_reward": 0.31733632429192465,
|
| 774 |
-
"rewards/format_reward": 0.9029018295307955,
|
| 775 |
-
"step": 283,
|
| 776 |
"total_flos": 0.0,
|
| 777 |
-
"train_loss":
|
| 778 |
-
"train_runtime":
|
| 779 |
-
"train_samples_per_second":
|
| 780 |
-
"train_steps_per_second": 0.
|
| 781 |
}
|
| 782 |
],
|
| 783 |
"logging_steps": 5,
|
| 784 |
-
"max_steps":
|
| 785 |
"num_input_tokens_seen": 0,
|
| 786 |
"num_train_epochs": 1,
|
| 787 |
"save_steps": 500,
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": null,
|
| 3 |
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 0,
|
| 5 |
"eval_steps": 100,
|
| 6 |
+
"global_step": 0,
|
| 7 |
"is_hyper_param_search": false,
|
| 8 |
"is_local_process_zero": true,
|
| 9 |
"is_world_process_zero": true,
|
| 10 |
"log_history": [
|
| 11 |
{
|
| 12 |
+
"completion_length": 367.53125,
|
| 13 |
+
"epoch": 0,
|
| 14 |
+
"kl": 0.0,
|
| 15 |
+
"reward": 0.6822916666666666,
|
| 16 |
+
"reward_std": 0.45769532521565753,
|
| 17 |
+
"rewards/accuracy_reward": 0.16666666666666666,
|
| 18 |
+
"rewards/format_reward": 0.515625,
|
| 19 |
+
"step": 0,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 20 |
"total_flos": 0.0,
|
| 21 |
+
"train_loss": 18.18843185901642,
|
| 22 |
+
"train_runtime": 28.8641,
|
| 23 |
+
"train_samples_per_second": 3.43,
|
| 24 |
+
"train_steps_per_second": 0.035
|
| 25 |
}
|
| 26 |
],
|
| 27 |
"logging_steps": 5,
|
| 28 |
+
"max_steps": 1,
|
| 29 |
"num_input_tokens_seen": 0,
|
| 30 |
"num_train_epochs": 1,
|
| 31 |
"save_steps": 500,
|
training_args.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 7608
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b4cb960ae5f3824a0620531afc22d0e5240365c7568a5b3c920037d0ad3450e7
|
| 3 |
size 7608
|