DCAgent/a1-stack_phpunit
014
1{2 "best_global_step": null,3 "best_metric": null,4 "best_model_checkpoint": null,5 "epoch": 7.0,6 "eval_steps": 500,7 "global_step": 4627,8 "is_hyper_param_search": false,9 "is_local_process_zero": true,10 "is_world_process_zero": true,11 "log_history": [12 {13 "epoch": 0.007564296520423601,14 "grad_norm": 19.058784422912282,15 "learning_rate": 3.455723542116631e-07,16 "loss": 0.7153,17 "loss_nan_ranks": 0,18 "loss_rank_avg": 0.7505241632461548,19 "step": 5,20 "valid_targets_mean": 3790.7,21 "valid_targets_min": 23922 },23 {24 "epoch": 0.015128593040847202,25 "grad_norm": 17.486795894826894,26 "learning_rate": 7.77537796976242e-07,27 "loss": 0.6813,28 "loss_nan_ranks": 0,29 "loss_rank_avg": 0.6587737798690796,30 "step": 10,31 "valid_targets_mean": 5055.8,32 "valid_targets_min": 94333 },34 {35 "epoch": 0.0226928895612708,36 "grad_norm": 17.853758617180958,37 "learning_rate": 1.209503239740821e-06,38 "loss": 0.71,39 "loss_nan_ranks": 0,40 "loss_rank_avg": 0.7153316736221313,41 "step": 15,42 "valid_targets_mean": 5290.3,43 "valid_targets_min": 80344 },45 {46 "epoch": 0.030257186081694403,47 "grad_norm": 13.456561875333087,48 "learning_rate": 1.6414686825053995e-06,49 "loss": 0.6532,50 "loss_nan_ranks": 0,51 "loss_rank_avg": 0.6719971895217896,52 "step": 20,53 "valid_targets_mean": 4477.1,54 "valid_targets_min": 46155 },56 {57 "epoch": 0.037821482602118005,58 "grad_norm": 8.589481200641513,59 "learning_rate": 2.0734341252699786e-06,60 "loss": 0.5971,61 "loss_nan_ranks": 0,62 "loss_rank_avg": 0.6080985069274902,63 "step": 25,64 "valid_targets_mean": 4065.7,65 "valid_targets_min": 79766 },67 {68 "epoch": 0.0453857791225416,69 "grad_norm": 4.7772774074698585,70 "learning_rate": 2.505399568034557e-06,71 "loss": 0.5155,72 "loss_nan_ranks": 0,73 "loss_rank_avg": 0.4876701831817627,74 "step": 30,75 "valid_targets_mean": 5669.8,76 "valid_targets_min": 72777 },78 {79 "epoch": 0.0529500756429652,80 "grad_norm": 2.9036893563370665,81 "learning_rate": 2.9373650107991366e-06,82 "loss": 0.4905,83 "loss_nan_ranks": 0,84 "loss_rank_avg": 0.44498470425605774,85 "step": 35,86 "valid_targets_mean": 4442.9,87 "valid_targets_min": 79588 },89 {90 "epoch": 0.060514372163388806,91 "grad_norm": 1.5763541463717798,92 "learning_rate": 3.369330453563715e-06,93 "loss": 0.4976,94 "loss_nan_ranks": 0,95 "loss_rank_avg": 0.492475301027298,96 "step": 40,97 "valid_targets_mean": 5954.6,98 "valid_targets_min": 58399 },100 {101 "epoch": 0.0680786686838124,102 "grad_norm": 1.4394911122910266,103 "learning_rate": 3.801295896328294e-06,104 "loss": 0.4597,105 "loss_nan_ranks": 0,106 "loss_rank_avg": 0.5117167234420776,107 "step": 45,108 "valid_targets_mean": 4686.0,109 "valid_targets_min": 610110 },111 {112 "epoch": 0.07564296520423601,113 "grad_norm": 1.1660007728840187,114 "learning_rate": 4.233261339092873e-06,115 "loss": 0.456,116 "loss_nan_ranks": 0,117 "loss_rank_avg": 0.4414580464363098,118 "step": 50,119 "valid_targets_mean": 4432.8,120 "valid_targets_min": 710121 },122 {123 "epoch": 0.0832072617246596,124 "grad_norm": 1.0994270688976955,125 "learning_rate": 4.665226781857452e-06,126 "loss": 0.4252,127 "loss_nan_ranks": 0,128 "loss_rank_avg": 0.5082100629806519,129 "step": 55,130 "valid_targets_mean": 4157.1,131 "valid_targets_min": 620132 },133 {134 "epoch": 0.0907715582450832,135 "grad_norm": 0.8284788849012524,136 "learning_rate": 5.09719222462203e-06,137 "loss": 0.4232,138 "loss_nan_ranks": 0,139 "loss_rank_avg": 0.44552695751190186,140 "step": 60,141 "valid_targets_mean": 4501.6,142 "valid_targets_min": 705143 },144 {145 "epoch": 0.09833585476550681,146 "grad_norm": 0.6310211132039384,147 "learning_rate": 5.52915766738661e-06,148 "loss": 0.3835,149 "loss_nan_ranks": 0,150 "loss_rank_avg": 0.3460736572742462,151 "step": 65,152 "valid_targets_mean": 4682.3,153 "valid_targets_min": 342154 },155 {156 "epoch": 0.1059001512859304,157 "grad_norm": 0.6177628509419143,158 "learning_rate": 5.961123110151188e-06,159 "loss": 0.3921,160 "loss_nan_ranks": 0,161 "loss_rank_avg": 0.38838571310043335,162 "step": 70,163 "valid_targets_mean": 5989.1,164 "valid_targets_min": 1412165 },166 {167 "epoch": 0.11346444780635401,168 "grad_norm": 0.6780059407854138,169 "learning_rate": 6.393088552915767e-06,170 "loss": 0.3815,171 "loss_nan_ranks": 0,172 "loss_rank_avg": 0.3895452618598938,173 "step": 75,174 "valid_targets_mean": 4225.5,175 "valid_targets_min": 391176 },177 {178 "epoch": 0.12102874432677761,179 "grad_norm": 0.5643766880346435,180 "learning_rate": 6.825053995680346e-06,181 "loss": 0.3558,182 "loss_nan_ranks": 0,183 "loss_rank_avg": 0.3587391972541809,184 "step": 80,185 "valid_targets_mean": 5413.6,186 "valid_targets_min": 559187 },188 {189 "epoch": 0.12859304084720122,190 "grad_norm": 0.9090969514772852,191 "learning_rate": 7.257019438444926e-06,192 "loss": 0.3704,193 "loss_nan_ranks": 0,194 "loss_rank_avg": 0.3452695310115814,195 "step": 85,196 "valid_targets_mean": 4190.5,197 "valid_targets_min": 561198 },199 {200 "epoch": 0.1361573373676248,201 "grad_norm": 0.5075113193939108,202 "learning_rate": 7.688984881209504e-06,203 "loss": 0.3446,204 "loss_nan_ranks": 0,205 "loss_rank_avg": 0.3218182325363159,206 "step": 90,207 "valid_targets_mean": 5132.1,208 "valid_targets_min": 821209 },210 {211 "epoch": 0.1437216338880484,212 "grad_norm": 0.4582977864357091,213 "learning_rate": 8.120950323974082e-06,214 "loss": 0.3514,215 "loss_nan_ranks": 0,216 "loss_rank_avg": 0.3215479254722595,217 "step": 95,218 "valid_targets_mean": 6527.9,219 "valid_targets_min": 629220 },221 {222 "epoch": 0.15128593040847202,223 "grad_norm": 0.48765027679444184,224 "learning_rate": 8.552915766738662e-06,225 "loss": 0.3573,226 "loss_nan_ranks": 0,227 "loss_rank_avg": 0.33919212222099304,228 "step": 100,229 "valid_targets_mean": 5894.4,230 "valid_targets_min": 1048231 },232 {233 "epoch": 0.1588502269288956,234 "grad_norm": 0.6167067050085139,235 "learning_rate": 8.98488120950324e-06,236 "loss": 0.347,237 "loss_nan_ranks": 0,238 "loss_rank_avg": 0.3706921935081482,239 "step": 105,240 "valid_targets_mean": 4482.9,241 "valid_targets_min": 571242 },243 {244 "epoch": 0.1664145234493192,245 "grad_norm": 0.5778398632077648,246 "learning_rate": 9.41684665226782e-06,247 "loss": 0.3286,248 "loss_nan_ranks": 0,249 "loss_rank_avg": 0.3193400502204895,250 "step": 110,251 "valid_targets_mean": 5415.0,252 "valid_targets_min": 1665253 },254 {255 "epoch": 0.17397881996974282,256 "grad_norm": 0.5647079800698889,257 "learning_rate": 9.848812095032398e-06,258 "loss": 0.3556,259 "loss_nan_ranks": 0,260 "loss_rank_avg": 0.3541482090950012,261 "step": 115,262 "valid_targets_mean": 5123.8,263 "valid_targets_min": 1269264 },265 {266 "epoch": 0.1815431164901664,267 "grad_norm": 0.5820680453553746,268 "learning_rate": 1.0280777537796978e-05,269 "loss": 0.3176,270 "loss_nan_ranks": 0,271 "loss_rank_avg": 0.30285853147506714,272 "step": 120,273 "valid_targets_mean": 4204.3,274 "valid_targets_min": 556275 },276 {277 "epoch": 0.18910741301059,278 "grad_norm": 0.47973201670838,279 "learning_rate": 1.0712742980561557e-05,280 "loss": 0.3178,281 "loss_nan_ranks": 0,282 "loss_rank_avg": 0.3058018088340759,283 "step": 125,284 "valid_targets_mean": 5260.8,285 "valid_targets_min": 868286 },287 {288 "epoch": 0.19667170953101362,289 "grad_norm": 0.5518953360121684,290 "learning_rate": 1.1144708423326134e-05,291 "loss": 0.3187,292 "loss_nan_ranks": 0,293 "loss_rank_avg": 0.3173016905784607,294 "step": 130,295 "valid_targets_mean": 5620.1,296 "valid_targets_min": 2169297 },298 {299 "epoch": 0.2042360060514372,300 "grad_norm": 0.5717143166242276,301 "learning_rate": 1.1576673866090712e-05,302 "loss": 0.3236,303 "loss_nan_ranks": 0,304 "loss_rank_avg": 0.35504987835884094,305 "step": 135,306 "valid_targets_mean": 5265.2,307 "valid_targets_min": 785308 },309 {310 "epoch": 0.2118003025718608,311 "grad_norm": 0.47865096065818147,312 "learning_rate": 1.2008639308855293e-05,313 "loss": 0.2921,314 "loss_nan_ranks": 0,315 "loss_rank_avg": 0.2680158019065857,316 "step": 140,317 "valid_targets_mean": 5838.5,318 "valid_targets_min": 2352319 },320 {321 "epoch": 0.21936459909228442,322 "grad_norm": 0.5085856003812484,323 "learning_rate": 1.2440604751619871e-05,324 "loss": 0.2999,325 "loss_nan_ranks": 0,326 "loss_rank_avg": 0.3085848391056061,327 "step": 145,328 "valid_targets_mean": 4785.6,329 "valid_targets_min": 653330 },331 {332 "epoch": 0.22692889561270801,333 "grad_norm": 0.5716698549308541,334 "learning_rate": 1.287257019438445e-05,335 "loss": 0.2984,336 "loss_nan_ranks": 0,337 "loss_rank_avg": 0.28182756900787354,338 "step": 150,339 "valid_targets_mean": 4273.5,340 "valid_targets_min": 825341 },342 {343 "epoch": 0.2344931921331316,344 "grad_norm": 0.5245334119291116,345 "learning_rate": 1.330453563714903e-05,346 "loss": 0.2842,347 "loss_nan_ranks": 0,348 "loss_rank_avg": 0.264009952545166,349 "step": 155,350 "valid_targets_mean": 4866.6,351 "valid_targets_min": 467352 },353 {354 "epoch": 0.24205748865355523,355 "grad_norm": 0.4935476283105332,356 "learning_rate": 1.3736501079913609e-05,357 "loss": 0.2967,358 "loss_nan_ranks": 0,359 "loss_rank_avg": 0.28478801250457764,360 "step": 160,361 "valid_targets_mean": 5374.4,362 "valid_targets_min": 2048363 },364 {365 "epoch": 0.24962178517397882,366 "grad_norm": 0.49385239011355125,367 "learning_rate": 1.4168466522678186e-05,368 "loss": 0.2936,369 "loss_nan_ranks": 0,370 "loss_rank_avg": 0.3246859908103943,371 "step": 165,372 "valid_targets_mean": 6078.2,373 "valid_targets_min": 1784374 },375 {376 "epoch": 0.25718608169440244,377 "grad_norm": 0.5031354977672869,378 "learning_rate": 1.4600431965442764e-05,379 "loss": 0.2931,380 "loss_nan_ranks": 0,381 "loss_rank_avg": 0.2915901839733124,382 "step": 170,383 "valid_targets_mean": 5410.8,384 "valid_targets_min": 1914385 },386 {387 "epoch": 0.264750378214826,388 "grad_norm": 0.5743132892962913,389 "learning_rate": 1.5032397408207345e-05,390 "loss": 0.286,391 "loss_nan_ranks": 0,392 "loss_rank_avg": 0.3496958315372467,393 "step": 175,394 "valid_targets_mean": 4596.1,395 "valid_targets_min": 747396 },397 {398 "epoch": 0.2723146747352496,399 "grad_norm": 0.6417430609782164,400 "learning_rate": 1.5464362850971925e-05,401 "loss": 0.2873,402 "loss_nan_ranks": 0,403 "loss_rank_avg": 0.27849793434143066,404 "step": 180,405 "valid_targets_mean": 4129.8,406 "valid_targets_min": 711407 },408 {409 "epoch": 0.27987897125567324,410 "grad_norm": 0.5525787543900736,411 "learning_rate": 1.5896328293736503e-05,412 "loss": 0.2809,413 "loss_nan_ranks": 0,414 "loss_rank_avg": 0.27930599451065063,415 "step": 185,416 "valid_targets_mean": 5218.1,417 "valid_targets_min": 2197418 },419 {420 "epoch": 0.2874432677760968,421 "grad_norm": 0.5887797757799174,422 "learning_rate": 1.6328293736501082e-05,423 "loss": 0.2996,424 "loss_nan_ranks": 0,425 "loss_rank_avg": 0.3260801434516907,426 "step": 190,427 "valid_targets_mean": 4696.6,428 "valid_targets_min": 1128429 },430 {431 "epoch": 0.2950075642965204,432 "grad_norm": 0.5611439547047103,433 "learning_rate": 1.676025917926566e-05,434 "loss": 0.2916,435 "loss_nan_ranks": 0,436 "loss_rank_avg": 0.26734042167663574,437 "step": 195,438 "valid_targets_mean": 5061.5,439 "valid_targets_min": 594440 },441 {442 "epoch": 0.30257186081694404,443 "grad_norm": 0.8142320877951509,444 "learning_rate": 1.719222462203024e-05,445 "loss": 0.2739,446 "loss_nan_ranks": 0,447 "loss_rank_avg": 0.2744123935699463,448 "step": 200,449 "valid_targets_mean": 4981.4,450 "valid_targets_min": 2432451 },452 {453 "epoch": 0.3101361573373676,454 "grad_norm": 0.5453776678521185,455 "learning_rate": 1.7624190064794818e-05,456 "loss": 0.2814,457 "loss_nan_ranks": 0,458 "loss_rank_avg": 0.2693478465080261,459 "step": 205,460 "valid_targets_mean": 4538.8,461 "valid_targets_min": 2627462 },463 {464 "epoch": 0.3177004538577912,465 "grad_norm": 0.572823640761258,466 "learning_rate": 1.8056155507559396e-05,467 "loss": 0.2758,468 "loss_nan_ranks": 0,469 "loss_rank_avg": 0.24936965107917786,470 "step": 210,471 "valid_targets_mean": 4717.1,472 "valid_targets_min": 1796473 },474 {475 "epoch": 0.32526475037821484,476 "grad_norm": 0.48857192086223755,477 "learning_rate": 1.8488120950323975e-05,478 "loss": 0.2703,479 "loss_nan_ranks": 0,480 "loss_rank_avg": 0.27971357107162476,481 "step": 215,482 "valid_targets_mean": 5259.7,483 "valid_targets_min": 2080484 },485 {486 "epoch": 0.3328290468986384,487 "grad_norm": 0.5832326533648584,488 "learning_rate": 1.8920086393088553e-05,489 "loss": 0.2682,490 "loss_nan_ranks": 0,491 "loss_rank_avg": 0.2944541871547699,492 "step": 220,493 "valid_targets_mean": 5435.7,494 "valid_targets_min": 622495 },496 {497 "epoch": 0.340393343419062,498 "grad_norm": 0.5304901414198607,499 "learning_rate": 1.9352051835853135e-05,500 "loss": 0.276,501 "loss_nan_ranks": 0,502 "loss_rank_avg": 0.2581842541694641,503 "step": 225,504 "valid_targets_mean": 5475.9,505 "valid_targets_min": 1850506 },507 {508 "epoch": 0.34795763993948564,509 "grad_norm": 0.5584973891534915,510 "learning_rate": 1.9784017278617714e-05,511 "loss": 0.2647,512 "loss_nan_ranks": 0,513 "loss_rank_avg": 0.2856540381908417,514 "step": 230,515 "valid_targets_mean": 5040.7,516 "valid_targets_min": 837517 },518 {519 "epoch": 0.3555219364599092,520 "grad_norm": 0.7615559789848529,521 "learning_rate": 2.021598272138229e-05,522 "loss": 0.2762,523 "loss_nan_ranks": 0,524 "loss_rank_avg": 0.27540600299835205,525 "step": 235,526 "valid_targets_mean": 4862.9,527 "valid_targets_min": 921528 },529 {530 "epoch": 0.3630862329803328,531 "grad_norm": 0.5494333839539911,532 "learning_rate": 2.064794816414687e-05,533 "loss": 0.2627,534 "loss_nan_ranks": 0,535 "loss_rank_avg": 0.25492146611213684,536 "step": 240,537 "valid_targets_mean": 4764.6,538 "valid_targets_min": 1899539 },540 {541 "epoch": 0.37065052950075644,542 "grad_norm": 0.5675208041808972,543 "learning_rate": 2.107991360691145e-05,544 "loss": 0.2738,545 "loss_nan_ranks": 0,546 "loss_rank_avg": 0.30120372772216797,547 "step": 245,548 "valid_targets_mean": 4816.8,549 "valid_targets_min": 496550 },551 {552 "epoch": 0.37821482602118,553 "grad_norm": 0.5516462624650181,554 "learning_rate": 2.1511879049676025e-05,555 "loss": 0.2668,556 "loss_nan_ranks": 0,557 "loss_rank_avg": 0.2463807463645935,558 "step": 250,559 "valid_targets_mean": 4821.3,560 "valid_targets_min": 892561 },562 {563 "epoch": 0.3857791225416036,564 "grad_norm": 0.5808471847669502,565 "learning_rate": 2.1943844492440607e-05,566 "loss": 0.2605,567 "loss_nan_ranks": 0,568 "loss_rank_avg": 0.25385603308677673,569 "step": 255,570 "valid_targets_mean": 4108.7,571 "valid_targets_min": 697572 },573 {574 "epoch": 0.39334341906202724,575 "grad_norm": 1.3051521187963644,576 "learning_rate": 2.2375809935205186e-05,577 "loss": 0.2557,578 "loss_nan_ranks": 0,579 "loss_rank_avg": 0.2327377051115036,580 "step": 260,581 "valid_targets_mean": 3904.6,582 "valid_targets_min": 526583 },584 {585 "epoch": 0.4009077155824508,586 "grad_norm": 0.540688473321198,587 "learning_rate": 2.2807775377969764e-05,588 "loss": 0.2775,589 "loss_nan_ranks": 0,590 "loss_rank_avg": 0.2616848945617676,591 "step": 265,592 "valid_targets_mean": 5200.8,593 "valid_targets_min": 723594 },595 {596 "epoch": 0.4084720121028744,597 "grad_norm": 0.5572137635803951,598 "learning_rate": 2.3239740820734343e-05,599 "loss": 0.2678,600 "loss_nan_ranks": 0,601 "loss_rank_avg": 0.2894093096256256,602 "step": 270,603 "valid_targets_mean": 5997.7,604 "valid_targets_min": 2741605 },606 {607 "epoch": 0.41603630862329805,608 "grad_norm": 0.5219797727194595,609 "learning_rate": 2.3671706263498925e-05,610 "loss": 0.2585,611 "loss_nan_ranks": 0,612 "loss_rank_avg": 0.23676098883152008,613 "step": 275,614 "valid_targets_mean": 5215.8,615 "valid_targets_min": 372616 },617 {618 "epoch": 0.4236006051437216,619 "grad_norm": 0.7269531089041338,620 "learning_rate": 2.41036717062635e-05,621 "loss": 0.2869,622 "loss_nan_ranks": 0,623 "loss_rank_avg": 0.29276716709136963,624 "step": 280,625 "valid_targets_mean": 5125.2,626 "valid_targets_min": 941627 },628 {629 "epoch": 0.43116490166414523,630 "grad_norm": 0.6854032248788718,631 "learning_rate": 2.453563714902808e-05,632 "loss": 0.2533,633 "loss_nan_ranks": 0,634 "loss_rank_avg": 0.2404939830303192,635 "step": 285,636 "valid_targets_mean": 4140.1,637 "valid_targets_min": 913638 },639 {640 "epoch": 0.43872919818456885,641 "grad_norm": 0.5427787800863983,642 "learning_rate": 2.496760259179266e-05,643 "loss": 0.2733,644 "loss_nan_ranks": 0,645 "loss_rank_avg": 0.29159748554229736,646 "step": 290,647 "valid_targets_mean": 5585.8,648 "valid_targets_min": 738649 },650 {651 "epoch": 0.4462934947049924,652 "grad_norm": 0.6325587088135726,653 "learning_rate": 2.5399568034557236e-05,654 "loss": 0.2709,655 "loss_nan_ranks": 0,656 "loss_rank_avg": 0.24903929233551025,657 "step": 295,658 "valid_targets_mean": 4124.8,659 "valid_targets_min": 760660 },661 {662 "epoch": 0.45385779122541603,663 "grad_norm": 0.6302528172341179,664 "learning_rate": 2.5831533477321818e-05,665 "loss": 0.2546,666 "loss_nan_ranks": 0,667 "loss_rank_avg": 0.2755891680717468,668 "step": 300,669 "valid_targets_mean": 5041.9,670 "valid_targets_min": 765671 },672 {673 "epoch": 0.46142208774583965,674 "grad_norm": 0.6101680849058813,675 "learning_rate": 2.6263498920086393e-05,676 "loss": 0.251,677 "loss_nan_ranks": 0,678 "loss_rank_avg": 0.23466572165489197,679 "step": 305,680 "valid_targets_mean": 4039.0,681 "valid_targets_min": 1489682 },683 {684 "epoch": 0.4689863842662632,685 "grad_norm": 0.5098554824207316,686 "learning_rate": 2.6695464362850975e-05,687 "loss": 0.2503,688 "loss_nan_ranks": 0,689 "loss_rank_avg": 0.24894092977046967,690 "step": 310,691 "valid_targets_mean": 5165.2,692 "valid_targets_min": 543693 },694 {695 "epoch": 0.47655068078668683,696 "grad_norm": 0.5153701765469025,697 "learning_rate": 2.7127429805615553e-05,698 "loss": 0.2987,699 "loss_nan_ranks": 0,700 "loss_rank_avg": 0.24448314309120178,701 "step": 315,702 "valid_targets_mean": 4878.4,703 "valid_targets_min": 1165704 },705 {706 "epoch": 0.48411497730711045,707 "grad_norm": 0.5150020480544099,708 "learning_rate": 2.755939524838013e-05,709 "loss": 0.2643,710 "loss_nan_ranks": 0,711 "loss_rank_avg": 0.24626128375530243,712 "step": 320,713 "valid_targets_mean": 5491.0,714 "valid_targets_min": 2393715 },716 {717 "epoch": 0.491679273827534,718 "grad_norm": 0.44779275422444326,719 "learning_rate": 2.799136069114471e-05,720 "loss": 0.2569,721 "loss_nan_ranks": 0,722 "loss_rank_avg": 0.25347915291786194,723 "step": 325,724 "valid_targets_mean": 6012.2,725 "valid_targets_min": 1728726 },727 {728 "epoch": 0.49924357034795763,729 "grad_norm": 0.5689898293216941,730 "learning_rate": 2.842332613390929e-05,731 "loss": 0.2643,732 "loss_nan_ranks": 0,733 "loss_rank_avg": 0.2442246973514557,734 "step": 330,735 "valid_targets_mean": 5822.6,736 "valid_targets_min": 1755737 },738 {739 "epoch": 0.5068078668683812,740 "grad_norm": 0.4828012134297725,741 "learning_rate": 2.885529157667387e-05,742 "loss": 0.2415,743 "loss_nan_ranks": 0,744 "loss_rank_avg": 0.22166290879249573,745 "step": 335,746 "valid_targets_mean": 5591.9,747 "valid_targets_min": 2595748 },749 {750 "epoch": 0.5143721633888049,751 "grad_norm": 0.5206534739708626,752 "learning_rate": 2.9287257019438446e-05,753 "loss": 0.254,754 "loss_nan_ranks": 0,755 "loss_rank_avg": 0.24545502662658691,756 "step": 340,757 "valid_targets_mean": 4507.7,758 "valid_targets_min": 1277759 },760 {761 "epoch": 0.5219364599092284,762 "grad_norm": 0.5375743058233416,763 "learning_rate": 2.9719222462203028e-05,764 "loss": 0.2387,765 "loss_nan_ranks": 0,766 "loss_rank_avg": 0.24855747818946838,767 "step": 345,768 "valid_targets_mean": 4502.4,769 "valid_targets_min": 706770 },771 {772 "epoch": 0.529500756429652,773 "grad_norm": 0.47062515075329686,774 "learning_rate": 3.0151187904967603e-05,775 "loss": 0.239,776 "loss_nan_ranks": 0,777 "loss_rank_avg": 0.23291119933128357,778 "step": 350,779 "valid_targets_mean": 5619.4,780 "valid_targets_min": 729781 },782 {783 "epoch": 0.5370650529500757,784 "grad_norm": 0.5065710560218952,785 "learning_rate": 3.058315334773218e-05,786 "loss": 0.2257,787 "loss_nan_ranks": 0,788 "loss_rank_avg": 0.23068980872631073,789 "step": 355,790 "valid_targets_mean": 5067.8,791 "valid_targets_min": 800792 },793 {794 "epoch": 0.5446293494704992,795 "grad_norm": 0.6071927129549767,796 "learning_rate": 3.101511879049676e-05,797 "loss": 0.2532,798 "loss_nan_ranks": 0,799 "loss_rank_avg": 0.2714073061943054,800 "step": 360,801 "valid_targets_mean": 4807.8,802 "valid_targets_min": 321803 },804 {805 "epoch": 0.5521936459909228,806 "grad_norm": 0.5932902639247645,807 "learning_rate": 3.144708423326134e-05,808 "loss": 0.267,809 "loss_nan_ranks": 0,810 "loss_rank_avg": 0.2581622004508972,811 "step": 365,812 "valid_targets_mean": 4726.7,813 "valid_targets_min": 263814 },815 {816 "epoch": 0.5597579425113465,817 "grad_norm": 0.5342014246005196,818 "learning_rate": 3.1879049676025925e-05,819 "loss": 0.2527,820 "loss_nan_ranks": 0,821 "loss_rank_avg": 0.28950175642967224,822 "step": 370,823 "valid_targets_mean": 5759.0,824 "valid_targets_min": 736825 },826 {827 "epoch": 0.56732223903177,828 "grad_norm": 0.6598267814370526,829 "learning_rate": 3.23110151187905e-05,830 "loss": 0.2328,831 "loss_nan_ranks": 0,832 "loss_rank_avg": 0.23706093430519104,833 "step": 375,834 "valid_targets_mean": 3705.8,835 "valid_targets_min": 511836 },837 {838 "epoch": 0.5748865355521936,839 "grad_norm": 0.5781352859246836,840 "learning_rate": 3.274298056155508e-05,841 "loss": 0.2694,842 "loss_nan_ranks": 0,843 "loss_rank_avg": 0.2128780633211136,844 "step": 380,845 "valid_targets_mean": 4299.5,846 "valid_targets_min": 565847 },848 {849 "epoch": 0.5824508320726173,850 "grad_norm": 0.5201196925356668,851 "learning_rate": 3.317494600431966e-05,852 "loss": 0.2476,853 "loss_nan_ranks": 0,854 "loss_rank_avg": 0.2847621440887451,855 "step": 385,856 "valid_targets_mean": 5042.4,857 "valid_targets_min": 825858 },859 {860 "epoch": 0.5900151285930408,861 "grad_norm": 0.571209286797819,862 "learning_rate": 3.360691144708423e-05,863 "loss": 0.2476,864 "loss_nan_ranks": 0,865 "loss_rank_avg": 0.261477530002594,866 "step": 390,867 "valid_targets_mean": 4204.5,868 "valid_targets_min": 831869 },870 {871 "epoch": 0.5975794251134644,872 "grad_norm": 0.5703991894869983,873 "learning_rate": 3.4038876889848814e-05,874 "loss": 0.2376,875 "loss_nan_ranks": 0,876 "loss_rank_avg": 0.22247353196144104,877 "step": 395,878 "valid_targets_mean": 4025.1,879 "valid_targets_min": 811880 },881 {882 "epoch": 0.6051437216338881,883 "grad_norm": 0.5694115354716299,884 "learning_rate": 3.447084233261339e-05,885 "loss": 0.2407,886 "loss_nan_ranks": 0,887 "loss_rank_avg": 0.2540602684020996,888 "step": 400,889 "valid_targets_mean": 4897.0,890 "valid_targets_min": 1926891 },892 {893 "epoch": 0.6127080181543116,894 "grad_norm": 0.543873277094909,895 "learning_rate": 3.490280777537797e-05,896 "loss": 0.2417,897 "loss_nan_ranks": 0,898 "loss_rank_avg": 0.2412918210029602,899 "step": 405,900 "valid_targets_mean": 5347.6,901 "valid_targets_min": 2336902 },903 {904 "epoch": 0.6202723146747352,905 "grad_norm": 0.5539147349076818,906 "learning_rate": 3.533477321814255e-05,907 "loss": 0.2529,908 "loss_nan_ranks": 0,909 "loss_rank_avg": 0.27449312806129456,910 "step": 410,911 "valid_targets_mean": 4702.8,912 "valid_targets_min": 641913 },914 {915 "epoch": 0.6278366111951589,916 "grad_norm": 0.5147878532498119,917 "learning_rate": 3.5766738660907135e-05,918 "loss": 0.2482,919 "loss_nan_ranks": 0,920 "loss_rank_avg": 0.21562246978282928,921 "step": 415,922 "valid_targets_mean": 4832.5,923 "valid_targets_min": 2092924 },925 {926 "epoch": 0.6354009077155824,927 "grad_norm": 0.5550659610309338,928 "learning_rate": 3.619870410367171e-05,929 "loss": 0.2372,930 "loss_nan_ranks": 0,931 "loss_rank_avg": 0.23504087328910828,932 "step": 420,933 "valid_targets_mean": 4465.6,934 "valid_targets_min": 643935 },936 {937 "epoch": 0.642965204236006,938 "grad_norm": 0.5881025450085932,939 "learning_rate": 3.6630669546436286e-05,940 "loss": 0.2508,941 "loss_nan_ranks": 0,942 "loss_rank_avg": 0.2751597464084625,943 "step": 425,944 "valid_targets_mean": 5010.4,945 "valid_targets_min": 575946 },947 {948 "epoch": 0.6505295007564297,949 "grad_norm": 0.5850477950512473,950 "learning_rate": 3.706263498920087e-05,951 "loss": 0.2391,952 "loss_nan_ranks": 0,953 "loss_rank_avg": 0.2643757462501526,954 "step": 430,955 "valid_targets_mean": 4947.0,956 "valid_targets_min": 678957 },958 {959 "epoch": 0.6580937972768532,960 "grad_norm": 0.8028527221952423,961 "learning_rate": 3.749460043196544e-05,962 "loss": 0.2512,963 "loss_nan_ranks": 0,964 "loss_rank_avg": 0.23027735948562622,965 "step": 435,966 "valid_targets_mean": 4532.6,967 "valid_targets_min": 632968 },969 {970 "epoch": 0.6656580937972768,971 "grad_norm": 0.5991434210232877,972 "learning_rate": 3.7926565874730025e-05,973 "loss": 0.233,974 "loss_nan_ranks": 0,975 "loss_rank_avg": 0.21481779217720032,976 "step": 440,977 "valid_targets_mean": 4295.4,978 "valid_targets_min": 1930979 },980 {981 "epoch": 0.6732223903177005,982 "grad_norm": 0.601475737977575,983 "learning_rate": 3.83585313174946e-05,984 "loss": 0.2458,985 "loss_nan_ranks": 0,986 "loss_rank_avg": 0.2731437683105469,987 "step": 445,988 "valid_targets_mean": 4081.8,989 "valid_targets_min": 789990 },991 {992 "epoch": 0.680786686838124,993 "grad_norm": 0.544970754754265,994 "learning_rate": 3.879049676025918e-05,995 "loss": 0.248,996 "loss_nan_ranks": 0,997 "loss_rank_avg": 0.2417784333229065,998 "step": 450,999 "valid_targets_mean": 4546.8,1000 "valid_targets_min": 6871001 },1002 {1003 "epoch": 0.6883509833585476,1004 "grad_norm": 0.5772939399178783,1005 "learning_rate": 3.9222462203023764e-05,1006 "loss": 0.2556,1007 "loss_nan_ranks": 0,1008 "loss_rank_avg": 0.23565149307250977,1009 "step": 455,1010 "valid_targets_mean": 4297.6,1011 "valid_targets_min": 6621012 },1013 {1014 "epoch": 0.6959152798789713,1015 "grad_norm": 0.557273760966465,1016 "learning_rate": 3.965442764578834e-05,1017 "loss": 0.2411,1018 "loss_nan_ranks": 0,1019 "loss_rank_avg": 0.23695287108421326,1020 "step": 460,1021 "valid_targets_mean": 5605.1,1022 "valid_targets_min": 4781023 },1024 {1025 "epoch": 0.7034795763993948,1026 "grad_norm": 0.552751474708773,1027 "learning_rate": 3.9999994307824485e-05,1028 "loss": 0.238,1029 "loss_nan_ranks": 0,1030 "loss_rank_avg": 0.23529154062271118,1031 "step": 465,1032 "valid_targets_mean": 4646.6,1033 "valid_targets_min": 8431034 },1035 {1036 "epoch": 0.7110438729198184,1037 "grad_norm": 0.5778857269434653,1038 "learning_rate": 3.9999795082021543e-05,1039 "loss": 0.2698,1040 "loss_nan_ranks": 0,1041 "loss_rank_avg": 0.3176231384277344,1042 "step": 470,1043 "valid_targets_mean": 5449.4,1044 "valid_targets_min": 5381045 },1046 {1047 "epoch": 0.7186081694402421,1048 "grad_norm": 0.5157110157373845,1049 "learning_rate": 3.999931125068276e-05,1050 "loss": 0.2514,1051 "loss_nan_ranks": 0,1052 "loss_rank_avg": 0.2607157230377197,1053 "step": 475,1054 "valid_targets_mean": 5027.3,1055 "valid_targets_min": 9511056 },1057 {1058 "epoch": 0.7261724659606656,1059 "grad_norm": 0.4933102641109027,1060 "learning_rate": 3.9998542820693246e-05,1061 "loss": 0.239,1062 "loss_nan_ranks": 0,1063 "loss_rank_avg": 0.20662082731723785,1064 "step": 480,1065 "valid_targets_mean": 5754.7,1066 "valid_targets_min": 6351067 },1068 {1069 "epoch": 0.7337367624810892,1070 "grad_norm": 0.5426556147022381,1071 "learning_rate": 3.9997489802988096e-05,1072 "loss": 0.2476,1073 "loss_nan_ranks": 0,1074 "loss_rank_avg": 0.23219190537929535,1075 "step": 485,1076 "valid_targets_mean": 4775.2,1077 "valid_targets_min": 7291078 },1079 {1080 "epoch": 0.7413010590015129,1081 "grad_norm": 0.4802976906182571,1082 "learning_rate": 3.9996152212552195e-05,1083 "loss": 0.2301,1084 "loss_nan_ranks": 0,1085 "loss_rank_avg": 0.22934921085834503,1086 "step": 490,1087 "valid_targets_mean": 5057.2,1088 "valid_targets_min": 6391089 },1090 {1091 "epoch": 0.7488653555219364,1092 "grad_norm": 0.6122232574312276,1093 "learning_rate": 3.999453006842002e-05,1094 "loss": 0.2389,1095 "loss_nan_ranks": 0,1096 "loss_rank_avg": 0.244462788105011,1097 "step": 495,1098 "valid_targets_mean": 4329.6,1099 "valid_targets_min": 5741100 },1101 {1102 "epoch": 0.75642965204236,1103 "grad_norm": 0.5032216034650763,1104 "learning_rate": 3.999262339367536e-05,1105 "loss": 0.2514,1106 "loss_nan_ranks": 0,1107 "loss_rank_avg": 0.25049036741256714,1108 "step": 500,1109 "valid_targets_mean": 4904.4,1110 "valid_targets_min": 10071111 },1112 {1113 "epoch": 0.7639939485627837,1114 "grad_norm": 0.6238287961984283,1115 "learning_rate": 3.9990432215451006e-05,1116 "loss": 0.2482,1117 "loss_nan_ranks": 0,1118 "loss_rank_avg": 0.26027706265449524,1119 "step": 505,1120 "valid_targets_mean": 3938.6,1121 "valid_targets_min": 6391122 },1123 {1124 "epoch": 0.7715582450832073,1125 "grad_norm": 0.5535062102674121,1126 "learning_rate": 3.998795656492836e-05,1127 "loss": 0.2352,1128 "loss_nan_ranks": 0,1129 "loss_rank_avg": 0.23881468176841736,1130 "step": 510,1131 "valid_targets_mean": 4196.7,1132 "valid_targets_min": 7441133 },1134 {1135 "epoch": 0.7791225416036308,1136 "grad_norm": 0.4897831464999995,1137 "learning_rate": 3.998519647733696e-05,1138 "loss": 0.2544,1139 "loss_nan_ranks": 0,1140 "loss_rank_avg": 0.2844647169113159,1141 "step": 515,1142 "valid_targets_mean": 6344.6,1143 "valid_targets_min": 6011144 },1145 {1146 "epoch": 0.7866868381240545,1147 "grad_norm": 0.5746692133321736,1148 "learning_rate": 3.998215199195403e-05,1149 "loss": 0.2469,1150 "loss_nan_ranks": 0,1151 "loss_rank_avg": 0.23050951957702637,1152 "step": 520,1153 "valid_targets_mean": 4142.7,1154 "valid_targets_min": 5471155 },1156 {1157 "epoch": 0.794251134644478,1158 "grad_norm": 0.49159881041487485,1159 "learning_rate": 3.997882315210388e-05,1160 "loss": 0.2512,1161 "loss_nan_ranks": 0,1162 "loss_rank_avg": 0.2426537573337555,1163 "step": 525,1164 "valid_targets_mean": 5857.8,1165 "valid_targets_min": 11111166 },1167 {1168 "epoch": 0.8018154311649016,1169 "grad_norm": 0.4965814246543715,1170 "learning_rate": 3.997521000515731e-05,1171 "loss": 0.2327,1172 "loss_nan_ranks": 0,1173 "loss_rank_avg": 0.22411620616912842,1174 "step": 530,1175 "valid_targets_mean": 5881.9,1176 "valid_targets_min": 16611177 },1178 {1179 "epoch": 0.8093797276853253,1180 "grad_norm": 0.5557055978474669,1181 "learning_rate": 3.997131260253092e-05,1182 "loss": 0.2392,1183 "loss_nan_ranks": 0,1184 "loss_rank_avg": 0.25896748900413513,1185 "step": 535,1186 "valid_targets_mean": 4466.9,1187 "valid_targets_min": 5921188 },1189 {1190 "epoch": 0.8169440242057489,1191 "grad_norm": 0.4931361228707487,1192 "learning_rate": 3.9967130999686405e-05,1193 "loss": 0.2356,1194 "loss_nan_ranks": 0,1195 "loss_rank_avg": 0.249818816781044,1196 "step": 540,1197 "valid_targets_mean": 4829.9,1198 "valid_targets_min": 9321199 },1200 {