samscript18/adaption_defi_wallet_risk_classif_3d6e6540
025
1{2 "best_global_step": null,3 "best_metric": null,4 "best_model_checkpoint": null,5 "epoch": 3.0,6 "eval_steps": 12,7 "global_step": 63,8 "is_hyper_param_search": false,9 "is_local_process_zero": true,10 "is_world_process_zero": true,11 "log_history": [12 {13 "epoch": 0.047619047619047616,14 "grad_norm": 0.39029014110565186,15 "learning_rate": 0.0,16 "loss": 1.74072265625,17 "step": 118 },19 {20 "epoch": 0.09523809523809523,21 "grad_norm": 0.3977895975112915,22 "learning_rate": 2.5e-05,23 "loss": 1.8193359375,24 "step": 225 },26 {27 "epoch": 0.14285714285714285,28 "grad_norm": 0.2535572648048401,29 "learning_rate": 5e-05,30 "loss": 1.408203125,31 "step": 332 },33 {34 "epoch": 0.19047619047619047,35 "grad_norm": 0.21865864098072052,36 "learning_rate": 7.500000000000001e-05,37 "loss": 1.54248046875,38 "step": 439 },40 {41 "epoch": 0.23809523809523808,42 "grad_norm": 0.14044301211833954,43 "learning_rate": 0.0001,44 "loss": 1.357666015625,45 "step": 546 },47 {48 "epoch": 0.2857142857142857,49 "grad_norm": 0.11468952149152756,50 "learning_rate": 9.99362213054543e-05,51 "loss": 1.414794921875,52 "step": 653 },54 {55 "epoch": 0.3333333333333333,56 "grad_norm": 0.15838299691677094,57 "learning_rate": 9.974506600945618e-05,58 "loss": 1.346435546875,59 "step": 760 },61 {62 "epoch": 0.38095238095238093,63 "grad_norm": 0.18158257007598877,64 "learning_rate": 9.942707596246052e-05,65 "loss": 1.345703125,66 "step": 867 },68 {69 "epoch": 0.42857142857142855,70 "grad_norm": 1.183410406112671,71 "learning_rate": 9.8983152541804e-05,72 "loss": 0.7744140625,73 "step": 974 },75 {76 "epoch": 0.47619047619047616,77 "grad_norm": 0.11678645759820938,78 "learning_rate": 9.841455409665323e-05,79 "loss": 1.285888671875,80 "step": 1081 },82 {83 "epoch": 0.5238095238095238,84 "grad_norm": 0.08664851635694504,85 "learning_rate": 9.772289238107715e-05,86 "loss": 1.1910400390625,87 "step": 1188 },89 {90 "epoch": 0.5714285714285714,91 "grad_norm": 0.07821615040302277,92 "learning_rate": 9.691012798535524e-05,93 "loss": 1.207763671875,94 "step": 1295 },96 {97 "epoch": 0.6190476190476191,98 "grad_norm": 0.07503138482570648,99 "learning_rate": 9.59785647784711e-05,100 "loss": 1.2236328125,101 "step": 13102 },103 {104 "epoch": 0.6666666666666666,105 "grad_norm": 0.0798412635922432,106 "learning_rate": 9.493084337754573e-05,107 "loss": 1.28759765625,108 "step": 14109 },110 {111 "epoch": 0.7142857142857143,112 "grad_norm": 0.08004077523946762,113 "learning_rate": 9.376993366272128e-05,114 "loss": 1.1826171875,115 "step": 15116 },117 {118 "epoch": 0.7142857142857143,119 "eval_loss": 1.216796875,120 "eval_runtime": 5.4909,121 "eval_samples_per_second": 0.364,122 "eval_steps_per_second": 0.182,123 "step": 15124 },125 {126 "epoch": 0.7619047619047619,127 "grad_norm": 0.06384038925170898,128 "learning_rate": 9.249912635871318e-05,129 "loss": 1.1982421875,130 "step": 16131 },132 {133 "epoch": 0.8095238095238095,134 "grad_norm": 0.06806826591491699,135 "learning_rate": 9.112202370689337e-05,136 "loss": 1.1966552734375,137 "step": 17138 },139 {140 "epoch": 0.8571428571428571,141 "grad_norm": 0.06299256533384323,142 "learning_rate": 8.964252925434579e-05,143 "loss": 1.2159423828125,144 "step": 18145 },146 {147 "epoch": 0.9047619047619048,148 "grad_norm": 0.04804723337292671,149 "learning_rate": 8.806483678883803e-05,150 "loss": 1.00927734375,151 "step": 19152 },153 {154 "epoch": 0.9523809523809523,155 "grad_norm": 0.060130614787340164,156 "learning_rate": 8.639341845107433e-05,157 "loss": 1.214599609375,158 "step": 20159 },160 {161 "epoch": 1.0,162 "grad_norm": 0.086543969810009,163 "learning_rate": 8.463301205792675e-05,164 "loss": 1.171630859375,165 "step": 21166 },167 {168 "epoch": 1.0476190476190477,169 "grad_norm": 0.0662260353565216,170 "learning_rate": 8.278860767257865e-05,171 "loss": 1.19677734375,172 "step": 22173 },174 {175 "epoch": 1.0952380952380953,176 "grad_norm": 0.05612208694219589,177 "learning_rate": 8.086543345964832e-05,178 "loss": 1.294921875,179 "step": 23180 },181 {182 "epoch": 1.1428571428571428,183 "grad_norm": 0.04827860742807388,184 "learning_rate": 7.886894086538841e-05,185 "loss": 1.03564453125,186 "step": 24187 },188 {189 "epoch": 1.1904761904761905,190 "grad_norm": 0.0562213771045208,191 "learning_rate": 7.680478916496926e-05,192 "loss": 1.2347412109375,193 "step": 25194 },195 {196 "epoch": 1.2380952380952381,197 "grad_norm": 0.05600786581635475,198 "learning_rate": 7.46788294206485e-05,199 "loss": 1.1312255859375,200 "step": 26201 },202 {203 "epoch": 1.2857142857142856,204 "grad_norm": 0.05734400078654289,205 "learning_rate": 7.249708789629944e-05,206 "loss": 1.21484375,207 "step": 27208 },209 {210 "epoch": 1.2857142857142856,211 "eval_loss": 1.169921875,212 "eval_runtime": 5.4973,213 "eval_samples_per_second": 0.364,214 "eval_steps_per_second": 0.182,215 "step": 27216 },217 {218 "epoch": 1.3333333333333333,219 "grad_norm": 0.04520232975482941,220 "learning_rate": 7.026574897531137e-05,221 "loss": 1.19384765625,222 "step": 28223 },224 {225 "epoch": 1.380952380952381,226 "grad_norm": 0.0506543330848217,227 "learning_rate": 6.799113763028295e-05,228 "loss": 1.1943359375,229 "step": 29230 },231 {232 "epoch": 1.4285714285714286,233 "grad_norm": 1.018054485321045,234 "learning_rate": 6.567970149420018e-05,235 "loss": 0.64453125,236 "step": 30237 },238 {239 "epoch": 1.4761904761904763,240 "grad_norm": 0.06562633067369461,241 "learning_rate": 6.333799258392015e-05,242 "loss": 1.1578369140625,243 "step": 31244 },245 {246 "epoch": 1.5238095238095237,247 "grad_norm": 0.04868047311902046,248 "learning_rate": 6.097264872776749e-05,249 "loss": 1.0946044921875,250 "step": 32251 },252 {253 "epoch": 1.5714285714285714,254 "grad_norm": 0.055948544293642044,255 "learning_rate": 5.859037474988874e-05,256 "loss": 1.116943359375,257 "step": 33258 },259 {260 "epoch": 1.619047619047619,261 "grad_norm": 0.050402332097291946,262 "learning_rate": 5.6197923464699875e-05,263 "loss": 1.1273193359375,264 "step": 34265 },266 {267 "epoch": 1.6666666666666665,268 "grad_norm": 0.061322227120399475,269 "learning_rate": 5.3802076535300135e-05,270 "loss": 1.176513671875,271 "step": 35272 },273 {274 "epoch": 1.7142857142857144,275 "grad_norm": 0.05519519001245499,276 "learning_rate": 5.140962525011126e-05,277 "loss": 1.0797119140625,278 "step": 36279 },280 {281 "epoch": 1.7619047619047619,282 "grad_norm": 0.049536723643541336,283 "learning_rate": 4.902735127223252e-05,284 "loss": 1.12481689453125,285 "step": 37286 },287 {288 "epoch": 1.8095238095238095,289 "grad_norm": 0.04784968122839928,290 "learning_rate": 4.6662007416079864e-05,291 "loss": 1.12640380859375,292 "step": 38293 },294 {295 "epoch": 1.8571428571428572,296 "grad_norm": 0.052412062883377075,297 "learning_rate": 4.432029850579983e-05,298 "loss": 1.136962890625,299 "step": 39300 },301 {302 "epoch": 1.8571428571428572,303 "eval_loss": 1.154296875,304 "eval_runtime": 5.4651,305 "eval_samples_per_second": 0.366,306 "eval_steps_per_second": 0.183,307 "step": 39308 },309 {310 "epoch": 1.9047619047619047,311 "grad_norm": 0.03997504711151123,312 "learning_rate": 4.2008862369717074e-05,313 "loss": 0.956787109375,314 "step": 40315 },316 {317 "epoch": 1.9523809523809523,318 "grad_norm": 0.05004314333200455,319 "learning_rate": 3.973425102468864e-05,320 "loss": 1.141845703125,321 "step": 41322 },323 {324 "epoch": 2.0,325 "grad_norm": 0.1028779074549675,326 "learning_rate": 3.750291210370057e-05,327 "loss": 1.067626953125,328 "step": 42329 },330 {331 "epoch": 2.0476190476190474,332 "grad_norm": 0.05941079929471016,333 "learning_rate": 3.5321170579351516e-05,334 "loss": 1.1151123046875,335 "step": 43336 },337 {338 "epoch": 2.0952380952380953,339 "grad_norm": 0.05426109582185745,340 "learning_rate": 3.319521083503075e-05,341 "loss": 1.222900390625,342 "step": 44343 },344 {345 "epoch": 2.142857142857143,346 "grad_norm": 0.08989612013101578,347 "learning_rate": 3.113105913461159e-05,348 "loss": 0.981689453125,349 "step": 45350 },351 {352 "epoch": 2.1904761904761907,353 "grad_norm": 0.048075128346681595,354 "learning_rate": 2.9134566540351693e-05,355 "loss": 1.1778564453125,356 "step": 46357 },358 {359 "epoch": 2.238095238095238,360 "grad_norm": 0.04994767904281616,361 "learning_rate": 2.721139232742137e-05,362 "loss": 1.0787353515625,363 "step": 47364 },365 {366 "epoch": 2.2857142857142856,367 "grad_norm": 0.05583450198173523,368 "learning_rate": 2.5366987942073267e-05,369 "loss": 1.1490478515625,370 "step": 48371 },372 {373 "epoch": 2.3333333333333335,374 "grad_norm": 0.04764580726623535,375 "learning_rate": 2.3606581548925695e-05,376 "loss": 1.14404296875,377 "step": 49378 },379 {380 "epoch": 2.380952380952381,381 "grad_norm": 0.04863216355443001,382 "learning_rate": 2.1935163211161985e-05,383 "loss": 1.1456298828125,384 "step": 50385 },386 {387 "epoch": 2.4285714285714284,388 "grad_norm": 0.02937026508152485,389 "learning_rate": 2.0357470745654214e-05,390 "loss": 0.5966796875,391 "step": 51392 },393 {394 "epoch": 2.4285714285714284,395 "eval_loss": 1.146484375,396 "eval_runtime": 5.4773,397 "eval_samples_per_second": 0.365,398 "eval_steps_per_second": 0.183,399 "step": 51400 },401 {402 "epoch": 2.4761904761904763,403 "grad_norm": 0.06333017349243164,404 "learning_rate": 1.8877976293106646e-05,405 "loss": 1.1070556640625,406 "step": 52407 },408 {409 "epoch": 2.5238095238095237,410 "grad_norm": 0.04482894763350487,411 "learning_rate": 1.7500873641286825e-05,412 "loss": 1.0587158203125,413 "step": 53414 },415 {416 "epoch": 2.571428571428571,417 "grad_norm": 0.040712468326091766,418 "learning_rate": 1.6230066337278724e-05,419 "loss": 1.081787109375,420 "step": 54421 },422 {423 "epoch": 2.619047619047619,424 "grad_norm": 0.08906273543834686,425 "learning_rate": 1.5069156622454287e-05,426 "loss": 1.08984375,427 "step": 55428 },429 {430 "epoch": 2.6666666666666665,431 "grad_norm": 0.061274148523807526,432 "learning_rate": 1.4021435221528906e-05,433 "loss": 1.130615234375,434 "step": 56435 },436 {437 "epoch": 2.7142857142857144,438 "grad_norm": 0.054327286779880524,439 "learning_rate": 1.3089872014644771e-05,440 "loss": 1.0430908203125,441 "step": 57442 },443 {444 "epoch": 2.761904761904762,445 "grad_norm": 0.04634387046098709,446 "learning_rate": 1.2277107618922843e-05,447 "loss": 1.09515380859375,448 "step": 58449 },450 {451 "epoch": 2.8095238095238093,452 "grad_norm": 0.0436868742108345,453 "learning_rate": 1.1585445903346783e-05,454 "loss": 1.10205078125,455 "step": 59456 },457 {458 "epoch": 2.857142857142857,459 "grad_norm": 0.05037835240364075,460 "learning_rate": 1.1016847458196001e-05,461 "loss": 1.1044921875,462 "step": 60463 },464 {465 "epoch": 2.9047619047619047,466 "grad_norm": 0.03902503475546837,467 "learning_rate": 1.0572924037539496e-05,468 "loss": 0.937744140625,469 "step": 61470 },471 {472 "epoch": 2.9523809523809526,473 "grad_norm": 0.050203897058963776,474 "learning_rate": 1.0254933990543831e-05,475 "loss": 1.1142578125,476 "step": 62477 },478 {479 "epoch": 3.0,480 "grad_norm": 0.07041402906179428,481 "learning_rate": 1.0063778694545714e-05,482 "loss": 1.0267333984375,483 "step": 63484 },485 {486 "epoch": 3.0,487 "eval_loss": 1.142578125,488 "eval_runtime": 5.4813,489 "eval_samples_per_second": 0.365,490 "eval_steps_per_second": 0.182,491 "step": 63492 }493 ],494 "logging_steps": 1.0,495 "max_steps": 63,496 "num_input_tokens_seen": 0,497 "num_train_epochs": 3,498 "save_steps": 0,499 "stateful_callbacks": {500 "TrainerControl": {501 "args": {502 "should_epoch_stop": false,503 "should_evaluate": false,504 "should_log": false,505 "should_save": true,506 "should_training_stop": true507 },508 "attributes": {}509 }510 },511 "total_flos": 1.3468496460722995e+18,512 "train_batch_size": 1,513 "trial_name": null,514 "trial_params": null515}516 