File size: 13,592 Bytes
4012ea1 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 | {
"best_metric": null,
"best_model_checkpoint": null,
"epoch": 0.002535004852157725,
"eval_steps": 500,
"global_step": 16,
"is_hyper_param_search": false,
"is_local_process_zero": true,
"is_world_process_zero": true,
"log_history": [
{
"epoch": 0.0001584378032598578,
"grad_norm": 6.941703031770885e-05,
"learning_rate": 2.9998415465061005e-05,
"loss": 0.0,
"loss/policy_avg": 9.534414857625961e-08,
"objective/entropy": 66.7407455444336,
"objective/kl": 0.0,
"objective/rlhf_reward": 1.9394512176513672,
"objective/scores": 1.9393310546875,
"policy/approxkl_avg": 0.0,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.501089334487915,
"step": 1,
"timer/calc_advantages": 1.838552713394165,
"timer/calc_loss": 0.718826174736023,
"timer/get_reward": 0.5057681202888489,
"timer/training_step": 5.023805141448975,
"val/num_eos_tokens": 0.0,
"val/ratio": 1.0,
"val/ratio_var": NaN
},
{
"epoch": 0.0003168756065197156,
"grad_norm": 8.962108612060547,
"learning_rate": 2.999683093012201e-05,
"loss": 0.0079,
"loss/policy_avg": 0.007912077941000462,
"objective/entropy": 58.23992919921875,
"objective/kl": 0.006753697991371155,
"objective/rlhf_reward": 2.58967924118042,
"objective/scores": 2.59033203125,
"policy/approxkl_avg": 0.143972247838974,
"policy/clipfrac_avg": 0.376953125,
"policy/entropy_avg": 0.437211811542511,
"step": 2,
"timer/calc_advantages": 1.693800687789917,
"timer/calc_loss": 0.6173452734947205,
"timer/get_reward": 0.431838721036911,
"timer/training_step": 4.488475322723389,
"val/num_eos_tokens": 0.0,
"val/ratio": 1.0006933212280273,
"val/ratio_var": NaN
},
{
"epoch": 0.0004753134097795734,
"grad_norm": 8.80933666229248,
"learning_rate": 2.9995246395183014e-05,
"loss": 0.0087,
"loss/policy_avg": 0.008671796880662441,
"objective/entropy": 60.747886657714844,
"objective/kl": 0.19268979132175446,
"objective/rlhf_reward": 2.797236442565918,
"objective/scores": 2.8162841796875,
"policy/approxkl_avg": 0.22207684814929962,
"policy/clipfrac_avg": 0.369140625,
"policy/entropy_avg": 0.47127196192741394,
"step": 3,
"timer/calc_advantages": 2.117703437805176,
"timer/calc_loss": 0.8592336773872375,
"timer/get_reward": 0.5989187955856323,
"timer/training_step": 5.804616928100586,
"val/num_eos_tokens": 0.0,
"val/ratio": 0.9997344613075256,
"val/ratio_var": NaN
},
{
"epoch": 0.0006337512130394313,
"grad_norm": 15.397004127502441,
"learning_rate": 2.999366186024402e-05,
"loss": 0.0181,
"loss/policy_avg": 0.018090050667524338,
"objective/entropy": 59.366294860839844,
"objective/kl": 0.16402865946292877,
"objective/rlhf_reward": 2.7970707416534424,
"objective/scores": 2.8134765625,
"policy/approxkl_avg": 0.30915290117263794,
"policy/clipfrac_avg": 0.41796875,
"policy/entropy_avg": 0.4661800265312195,
"step": 4,
"timer/calc_advantages": 2.018901824951172,
"timer/calc_loss": 0.8049512505531311,
"timer/get_reward": 0.5547392964363098,
"timer/training_step": 5.506021022796631,
"val/num_eos_tokens": 0.0,
"val/ratio": 1.0003275871276855,
"val/ratio_var": NaN
},
{
"epoch": 0.000792189016299289,
"grad_norm": 6.137847231002524e-05,
"learning_rate": 2.9992077325305024e-05,
"loss": 0.0,
"loss/policy_avg": 1.0040821507573128e-07,
"objective/entropy": 61.850738525390625,
"objective/kl": 0.0,
"objective/rlhf_reward": 2.3658785820007324,
"objective/scores": 2.36578369140625,
"policy/approxkl_avg": 0.0,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.48403242230415344,
"step": 5,
"timer/calc_advantages": 1.889434576034546,
"timer/calc_loss": 0.7489800453186035,
"timer/get_reward": 0.5218558311462402,
"timer/training_step": 5.19569206237793,
"val/num_eos_tokens": 0.0,
"val/ratio": 1.0,
"val/ratio_var": NaN
},
{
"epoch": 0.0009506268195591468,
"grad_norm": 3.129560947418213,
"learning_rate": 2.9990492790366028e-05,
"loss": 0.002,
"loss/policy_avg": 0.002019597217440605,
"objective/entropy": 64.70206451416016,
"objective/kl": -0.011846143752336502,
"objective/rlhf_reward": 2.4429714679718018,
"objective/scores": 2.44146728515625,
"policy/approxkl_avg": 0.05796036496758461,
"policy/clipfrac_avg": 0.224609375,
"policy/entropy_avg": 0.4975966215133667,
"step": 6,
"timer/calc_advantages": 1.939255714416504,
"timer/calc_loss": 0.7629643082618713,
"timer/get_reward": 0.5286913514137268,
"timer/training_step": 5.261943340301514,
"val/num_eos_tokens": 0.0,
"val/ratio": 1.0005271434783936,
"val/ratio_var": NaN
},
{
"epoch": 0.0011090646228190046,
"grad_norm": 4.28013277053833,
"learning_rate": 2.9988908255427033e-05,
"loss": 0.0037,
"loss/policy_avg": 0.0036908092442899942,
"objective/entropy": 68.02294158935547,
"objective/kl": 0.15080219507217407,
"objective/rlhf_reward": 3.149240016937256,
"objective/scores": 3.1646728515625,
"policy/approxkl_avg": 0.11497651040554047,
"policy/clipfrac_avg": 0.25,
"policy/entropy_avg": 0.5161693692207336,
"step": 7,
"timer/calc_advantages": 2.2946014404296875,
"timer/calc_loss": 0.96665358543396,
"timer/get_reward": 0.6638086438179016,
"timer/training_step": 6.369439125061035,
"val/num_eos_tokens": 0.0,
"val/ratio": 0.99953293800354,
"val/ratio_var": NaN
},
{
"epoch": 0.0012675024260788625,
"grad_norm": 8.370040893554688,
"learning_rate": 2.9987323720488037e-05,
"loss": 0.0078,
"loss/policy_avg": 0.00783943198621273,
"objective/entropy": 66.24089813232422,
"objective/kl": 0.0030039921402931213,
"objective/rlhf_reward": 3.1140072345733643,
"objective/scores": 3.1142501831054688,
"policy/approxkl_avg": 0.13499276340007782,
"policy/clipfrac_avg": 0.345703125,
"policy/entropy_avg": 0.4966353476047516,
"step": 8,
"timer/calc_advantages": 2.4299309253692627,
"timer/calc_loss": 1.046256422996521,
"timer/get_reward": 0.7162653207778931,
"timer/training_step": 6.773098945617676,
"val/num_eos_tokens": 0.0,
"val/ratio": 1.0011930465698242,
"val/ratio_var": NaN
},
{
"epoch": 0.0014259402293387202,
"grad_norm": 5.985812094877474e-05,
"learning_rate": 2.9985739185549042e-05,
"loss": 0.0,
"loss/policy_avg": 1.3085082173347473e-07,
"objective/entropy": 57.91233825683594,
"objective/kl": 0.0,
"objective/rlhf_reward": 2.9793567657470703,
"objective/scores": 2.9791259765625,
"policy/approxkl_avg": 0.0,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.4229130744934082,
"step": 9,
"timer/calc_advantages": 2.2175822257995605,
"timer/calc_loss": 0.940474271774292,
"timer/get_reward": 0.6452049016952515,
"timer/training_step": 6.192902565002441,
"val/num_eos_tokens": 0.0,
"val/ratio": 1.0,
"val/ratio_var": NaN
},
{
"epoch": 0.001584378032598578,
"grad_norm": 2.061887502670288,
"learning_rate": 2.9984154650610046e-05,
"loss": 0.0012,
"loss/policy_avg": 0.00118393381126225,
"objective/entropy": 57.78638458251953,
"objective/kl": 0.016596710309386253,
"objective/rlhf_reward": 2.493161916732788,
"objective/scores": 2.494873046875,
"policy/approxkl_avg": 0.04327381029725075,
"policy/clipfrac_avg": 0.220703125,
"policy/entropy_avg": 0.45412811636924744,
"step": 10,
"timer/calc_advantages": 1.9425208568572998,
"timer/calc_loss": 0.7499428987503052,
"timer/get_reward": 0.5237178802490234,
"timer/training_step": 5.241105556488037,
"val/num_eos_tokens": 0.0,
"val/ratio": 1.0002281665802002,
"val/ratio_var": NaN
},
{
"epoch": 0.0017428158358584358,
"grad_norm": 3.437352180480957,
"learning_rate": 2.998257011567105e-05,
"loss": 0.003,
"loss/policy_avg": 0.003013810608536005,
"objective/entropy": 62.60158920288086,
"objective/kl": 0.04900990426540375,
"objective/rlhf_reward": 3.26180362701416,
"objective/scores": 3.2667236328125,
"policy/approxkl_avg": 0.08036807179450989,
"policy/clipfrac_avg": 0.2734375,
"policy/entropy_avg": 0.49370765686035156,
"step": 11,
"timer/calc_advantages": 2.010678768157959,
"timer/calc_loss": 0.8044981956481934,
"timer/get_reward": 0.5571302771568298,
"timer/training_step": 5.486995220184326,
"val/num_eos_tokens": 0.0,
"val/ratio": 1.0001283884048462,
"val/ratio_var": NaN
},
{
"epoch": 0.0019012536391182935,
"grad_norm": 3.6680727005004883,
"learning_rate": 2.9980985580732055e-05,
"loss": 0.0026,
"loss/policy_avg": 0.002624134300276637,
"objective/entropy": 70.5216064453125,
"objective/kl": 0.039649978280067444,
"objective/rlhf_reward": 1.9454692602157593,
"objective/scores": 1.94927978515625,
"policy/approxkl_avg": 0.07172106206417084,
"policy/clipfrac_avg": 0.34375,
"policy/entropy_avg": 0.5483217239379883,
"step": 12,
"timer/calc_advantages": 1.708000898361206,
"timer/calc_loss": 0.6391787528991699,
"timer/get_reward": 0.4455040693283081,
"timer/training_step": 4.587984085083008,
"val/num_eos_tokens": 0.0,
"val/ratio": 1.0003334283828735,
"val/ratio_var": NaN
},
{
"epoch": 0.0020596914423781513,
"grad_norm": 5.551970753003843e-05,
"learning_rate": 2.997940104579306e-05,
"loss": 0.0,
"loss/policy_avg": 6.216578185558319e-08,
"objective/entropy": 58.013153076171875,
"objective/kl": 0.0,
"objective/rlhf_reward": 1.9868602752685547,
"objective/scores": 1.9864273071289062,
"policy/approxkl_avg": 0.0,
"policy/clipfrac_avg": 0.0,
"policy/entropy_avg": 0.44822999835014343,
"step": 13,
"timer/calc_advantages": 2.0904719829559326,
"timer/calc_loss": 0.8480838537216187,
"timer/get_reward": 0.587557315826416,
"timer/training_step": 5.7307024002075195,
"val/num_eos_tokens": 0.0,
"val/ratio": 1.0,
"val/ratio_var": NaN
},
{
"epoch": 0.002218129245638009,
"grad_norm": 2.7803940773010254,
"learning_rate": 2.9977816510854064e-05,
"loss": 0.0015,
"loss/policy_avg": 0.0014500978868454695,
"objective/entropy": 69.96898651123047,
"objective/kl": 0.06745042651891708,
"objective/rlhf_reward": 2.469928741455078,
"objective/scores": 2.477020263671875,
"policy/approxkl_avg": 0.053128380328416824,
"policy/clipfrac_avg": 0.2734375,
"policy/entropy_avg": 0.5138126015663147,
"step": 14,
"timer/calc_advantages": 1.917910099029541,
"timer/calc_loss": 0.7225819230079651,
"timer/get_reward": 0.5053159594535828,
"timer/training_step": 5.0998334884643555,
"val/num_eos_tokens": 0.0,
"val/ratio": 0.9998296499252319,
"val/ratio_var": NaN
},
{
"epoch": 0.002376567048897867,
"grad_norm": 2.532649278640747,
"learning_rate": 2.997623197591507e-05,
"loss": 0.0018,
"loss/policy_avg": 0.0017822147347033024,
"objective/entropy": 58.39247131347656,
"objective/kl": -0.011621108278632164,
"objective/rlhf_reward": 2.397026538848877,
"objective/scores": 2.39556884765625,
"policy/approxkl_avg": 0.05379398912191391,
"policy/clipfrac_avg": 0.248046875,
"policy/entropy_avg": 0.4527440071105957,
"step": 15,
"timer/calc_advantages": 2.565062999725342,
"timer/calc_loss": 1.1112425327301025,
"timer/get_reward": 0.7521659731864929,
"timer/training_step": 7.156440734863281,
"val/num_eos_tokens": 0.0,
"val/ratio": 1.0004982948303223,
"val/ratio_var": NaN
},
{
"epoch": 0.002535004852157725,
"grad_norm": 5.747053623199463,
"learning_rate": 2.9974647440976074e-05,
"loss": 0.0039,
"loss/policy_avg": 0.0039013477507978678,
"objective/entropy": 58.41150665283203,
"objective/kl": -0.03636426478624344,
"objective/rlhf_reward": 2.3745107650756836,
"objective/scores": 2.370819091796875,
"policy/approxkl_avg": 0.07460636645555496,
"policy/clipfrac_avg": 0.310546875,
"policy/entropy_avg": 0.4354715943336487,
"step": 16,
"timer/calc_advantages": 2.135810136795044,
"timer/calc_loss": 0.8651329278945923,
"timer/get_reward": 0.5979053378105164,
"timer/training_step": 5.837425231933594,
"val/num_eos_tokens": 0.0,
"val/ratio": 1.0008955001831055,
"val/ratio_var": NaN
}
],
"logging_steps": 1,
"max_steps": 18933,
"num_input_tokens_seen": 0,
"num_train_epochs": 3,
"save_steps": 8,
"total_flos": 8202719718014976.0,
"train_batch_size": 4,
"trial_name": null,
"trial_params": null
}
|