diff --git "a/trainer_state.json" "b/trainer_state.json" new file mode 100644--- /dev/null +++ "b/trainer_state.json" @@ -0,0 +1,21211 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 7.0, + "eval_steps": 500, + "global_step": 9624, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.0036376864314296106, + "grad_norm": 10.114328737598813, + "learning_rate": 1.6614745586708205e-07, + "loss": 0.5974, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24675223231315613, + "step": 5, + "valid_targets_mean": 2722.6, + "valid_targets_min": 458 + }, + { + "epoch": 0.007275372862859221, + "grad_norm": 10.836277666677635, + "learning_rate": 3.7383177570093457e-07, + "loss": 0.5133, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25949111580848694, + "step": 10, + "valid_targets_mean": 3113.1, + "valid_targets_min": 2461 + }, + { + "epoch": 0.010913059294288832, + "grad_norm": 9.10754325239007, + "learning_rate": 5.815160955347872e-07, + "loss": 0.4607, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26836875081062317, + "step": 15, + "valid_targets_mean": 3648.1, + "valid_targets_min": 2068 + }, + { + "epoch": 0.014550745725718442, + "grad_norm": 9.254340078421166, + "learning_rate": 7.892004153686398e-07, + "loss": 0.5163, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2516695559024811, + "step": 20, + "valid_targets_mean": 3423.6, + "valid_targets_min": 2425 + }, + { + "epoch": 0.018188432157148052, + "grad_norm": 8.405271823744352, + "learning_rate": 9.968847352024923e-07, + "loss": 0.4674, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2306375950574875, + "step": 25, + "valid_targets_mean": 2724.1, + "valid_targets_min": 1198 + }, + { + "epoch": 0.021826118588577664, + "grad_norm": 11.75361526037543, + "learning_rate": 1.2045690550363449e-06, + "loss": 0.5499, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.37609297037124634, + "step": 30, + "valid_targets_mean": 1232.5, + "valid_targets_min": 807 + }, + { + "epoch": 0.025463805020007276, + "grad_norm": 5.397556970258138, + "learning_rate": 1.4122533748701975e-06, + "loss": 0.7277, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26106715202331543, + "step": 35, + "valid_targets_mean": 3487.4, + "valid_targets_min": 2351 + }, + { + "epoch": 0.029101491451436885, + "grad_norm": 5.461486199227446, + "learning_rate": 1.6199376947040499e-06, + "loss": 0.4336, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.316693514585495, + "step": 40, + "valid_targets_mean": 3126.4, + "valid_targets_min": 781 + }, + { + "epoch": 0.032739177882866496, + "grad_norm": 3.0452736930127715, + "learning_rate": 1.8276220145379025e-06, + "loss": 0.4451, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19885537028312683, + "step": 45, + "valid_targets_mean": 3474.8, + "valid_targets_min": 799 + }, + { + "epoch": 0.036376864314296105, + "grad_norm": 2.43914221684919, + "learning_rate": 2.0353063343717553e-06, + "loss": 0.3639, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16437993943691254, + "step": 50, + "valid_targets_mean": 5026.1, + "valid_targets_min": 2802 + }, + { + "epoch": 0.04001455074572572, + "grad_norm": 2.7286518797586004, + "learning_rate": 2.2429906542056077e-06, + "loss": 0.3155, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13479313254356384, + "step": 55, + "valid_targets_mean": 2611.0, + "valid_targets_min": 574 + }, + { + "epoch": 0.04365223717715533, + "grad_norm": 2.243509024335104, + "learning_rate": 2.4506749740394605e-06, + "loss": 0.3299, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15883654356002808, + "step": 60, + "valid_targets_mean": 4036.5, + "valid_targets_min": 2642 + }, + { + "epoch": 0.04728992360858494, + "grad_norm": 2.0692055857970235, + "learning_rate": 2.658359293873313e-06, + "loss": 0.3199, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20522472262382507, + "step": 65, + "valid_targets_mean": 1960.5, + "valid_targets_min": 593 + }, + { + "epoch": 0.05092761004001455, + "grad_norm": 1.0934254810603672, + "learning_rate": 2.8660436137071652e-06, + "loss": 0.3409, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12179429829120636, + "step": 70, + "valid_targets_mean": 3354.2, + "valid_targets_min": 1353 + }, + { + "epoch": 0.05456529647144416, + "grad_norm": 1.2937983373809205, + "learning_rate": 3.073727933541018e-06, + "loss": 0.2765, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1612856388092041, + "step": 75, + "valid_targets_mean": 2300.9, + "valid_targets_min": 818 + }, + { + "epoch": 0.05820298290287377, + "grad_norm": 0.8828807162603297, + "learning_rate": 3.2814122533748704e-06, + "loss": 0.2752, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0998382493853569, + "step": 80, + "valid_targets_mean": 2295.8, + "valid_targets_min": 934 + }, + { + "epoch": 0.061840669334303384, + "grad_norm": 0.7223368707924613, + "learning_rate": 3.489096573208723e-06, + "loss": 0.2332, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12098997831344604, + "step": 85, + "valid_targets_mean": 3524.4, + "valid_targets_min": 2798 + }, + { + "epoch": 0.06547835576573299, + "grad_norm": 0.8846476611659986, + "learning_rate": 3.6967808930425756e-06, + "loss": 0.2845, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11931869387626648, + "step": 90, + "valid_targets_mean": 2810.1, + "valid_targets_min": 471 + }, + { + "epoch": 0.06911604219716261, + "grad_norm": 0.9070141289582611, + "learning_rate": 3.9044652128764284e-06, + "loss": 0.2703, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14734545350074768, + "step": 95, + "valid_targets_mean": 3042.8, + "valid_targets_min": 903 + }, + { + "epoch": 0.07275372862859221, + "grad_norm": 0.7222609127376945, + "learning_rate": 4.112149532710281e-06, + "loss": 0.2558, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09504024684429169, + "step": 100, + "valid_targets_mean": 3081.6, + "valid_targets_min": 474 + }, + { + "epoch": 0.07639141506002183, + "grad_norm": 0.6756628201226269, + "learning_rate": 4.319833852544133e-06, + "loss": 0.2499, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11739324033260345, + "step": 105, + "valid_targets_mean": 3468.0, + "valid_targets_min": 640 + }, + { + "epoch": 0.08002910149145144, + "grad_norm": 0.6263014479805286, + "learning_rate": 4.527518172377986e-06, + "loss": 0.2828, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08657917380332947, + "step": 110, + "valid_targets_mean": 3438.4, + "valid_targets_min": 902 + }, + { + "epoch": 0.08366678792288104, + "grad_norm": 0.6112140605891385, + "learning_rate": 4.735202492211838e-06, + "loss": 0.2097, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10475238412618637, + "step": 115, + "valid_targets_mean": 3212.8, + "valid_targets_min": 1837 + }, + { + "epoch": 0.08730447435431066, + "grad_norm": 2.491869329480981, + "learning_rate": 4.94288681204569e-06, + "loss": 0.2951, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22906696796417236, + "step": 120, + "valid_targets_mean": 1358.8, + "valid_targets_min": 756 + }, + { + "epoch": 0.09094216078574027, + "grad_norm": 0.8509431118777082, + "learning_rate": 5.150571131879544e-06, + "loss": 0.2639, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17456746101379395, + "step": 125, + "valid_targets_mean": 2896.2, + "valid_targets_min": 1072 + }, + { + "epoch": 0.09457984721716987, + "grad_norm": 0.6428172967471864, + "learning_rate": 5.358255451713395e-06, + "loss": 0.2073, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06719799339771271, + "step": 130, + "valid_targets_mean": 2908.1, + "valid_targets_min": 895 + }, + { + "epoch": 0.09821753364859949, + "grad_norm": 1.2099121819100576, + "learning_rate": 5.565939771547248e-06, + "loss": 0.3783, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2754223346710205, + "step": 135, + "valid_targets_mean": 2562.2, + "valid_targets_min": 878 + }, + { + "epoch": 0.1018552200800291, + "grad_norm": 0.6473027480792377, + "learning_rate": 5.773624091381102e-06, + "loss": 0.3339, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08653141558170319, + "step": 140, + "valid_targets_mean": 2087.6, + "valid_targets_min": 756 + }, + { + "epoch": 0.1054929065114587, + "grad_norm": 0.5586054598175749, + "learning_rate": 5.981308411214953e-06, + "loss": 0.2228, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08340767025947571, + "step": 145, + "valid_targets_mean": 3421.8, + "valid_targets_min": 2963 + }, + { + "epoch": 0.10913059294288832, + "grad_norm": 2.2230690848163124, + "learning_rate": 6.188992731048806e-06, + "loss": 0.3044, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3192937970161438, + "step": 150, + "valid_targets_mean": 2375.2, + "valid_targets_min": 1153 + }, + { + "epoch": 0.11276827937431794, + "grad_norm": 0.4283495213601293, + "learning_rate": 6.39667705088266e-06, + "loss": 0.215, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07745062559843063, + "step": 155, + "valid_targets_mean": 4886.2, + "valid_targets_min": 1700 + }, + { + "epoch": 0.11640596580574754, + "grad_norm": 0.4897837800264012, + "learning_rate": 6.604361370716511e-06, + "loss": 0.1815, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.060412921011447906, + "step": 160, + "valid_targets_mean": 3965.1, + "valid_targets_min": 789 + }, + { + "epoch": 0.12004365223717715, + "grad_norm": 0.4853922699745468, + "learning_rate": 6.812045690550364e-06, + "loss": 0.1964, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10330147296190262, + "step": 165, + "valid_targets_mean": 3771.5, + "valid_targets_min": 995 + }, + { + "epoch": 0.12368133866860677, + "grad_norm": 0.5472148422449984, + "learning_rate": 7.019730010384217e-06, + "loss": 0.1819, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06253156065940857, + "step": 170, + "valid_targets_mean": 2220.8, + "valid_targets_min": 488 + }, + { + "epoch": 0.12731902510003637, + "grad_norm": 0.5551647080715167, + "learning_rate": 7.227414330218069e-06, + "loss": 0.1891, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10403120517730713, + "step": 175, + "valid_targets_mean": 3635.4, + "valid_targets_min": 2437 + }, + { + "epoch": 0.13095671153146599, + "grad_norm": 0.7326366222737842, + "learning_rate": 7.435098650051922e-06, + "loss": 0.1764, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07533711194992065, + "step": 180, + "valid_targets_mean": 1752.1, + "valid_targets_min": 712 + }, + { + "epoch": 0.1345943979628956, + "grad_norm": 0.5421761110532238, + "learning_rate": 7.642782969885774e-06, + "loss": 0.1755, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08784523606300354, + "step": 185, + "valid_targets_mean": 3713.4, + "valid_targets_min": 2422 + }, + { + "epoch": 0.13823208439432522, + "grad_norm": 0.5250650561345698, + "learning_rate": 7.850467289719627e-06, + "loss": 0.1566, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08170074224472046, + "step": 190, + "valid_targets_mean": 3998.5, + "valid_targets_min": 3643 + }, + { + "epoch": 0.14186977082575483, + "grad_norm": 0.5114475168047639, + "learning_rate": 8.058151609553479e-06, + "loss": 0.2449, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.061784546822309494, + "step": 195, + "valid_targets_mean": 3319.0, + "valid_targets_min": 772 + }, + { + "epoch": 0.14550745725718442, + "grad_norm": 0.5536542807623532, + "learning_rate": 8.265835929387332e-06, + "loss": 0.1485, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06559751182794571, + "step": 200, + "valid_targets_mean": 2934.8, + "valid_targets_min": 722 + }, + { + "epoch": 0.14914514368861403, + "grad_norm": 0.43858622334489056, + "learning_rate": 8.473520249221184e-06, + "loss": 0.1607, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06224973872303963, + "step": 205, + "valid_targets_mean": 3656.4, + "valid_targets_min": 2788 + }, + { + "epoch": 0.15278283012004365, + "grad_norm": 0.593916700229624, + "learning_rate": 8.681204569055037e-06, + "loss": 0.1614, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08881933987140656, + "step": 210, + "valid_targets_mean": 3202.5, + "valid_targets_min": 948 + }, + { + "epoch": 0.15642051655147327, + "grad_norm": 0.9989629426392911, + "learning_rate": 8.888888888888888e-06, + "loss": 0.1875, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0891784280538559, + "step": 215, + "valid_targets_mean": 1002.0, + "valid_targets_min": 490 + }, + { + "epoch": 0.16005820298290288, + "grad_norm": 0.48675282889558635, + "learning_rate": 9.096573208722742e-06, + "loss": 0.2075, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09112532436847687, + "step": 220, + "valid_targets_mean": 3751.9, + "valid_targets_min": 2800 + }, + { + "epoch": 0.1636958894143325, + "grad_norm": 0.8125467253710073, + "learning_rate": 9.304257528556595e-06, + "loss": 0.219, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14939936995506287, + "step": 225, + "valid_targets_mean": 2641.8, + "valid_targets_min": 896 + }, + { + "epoch": 0.16733357584576208, + "grad_norm": 0.6873661618942163, + "learning_rate": 9.511941848390446e-06, + "loss": 0.1865, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11107177287340164, + "step": 230, + "valid_targets_mean": 3199.0, + "valid_targets_min": 1333 + }, + { + "epoch": 0.1709712622771917, + "grad_norm": 0.6540772425865923, + "learning_rate": 9.7196261682243e-06, + "loss": 0.2094, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09859389066696167, + "step": 235, + "valid_targets_mean": 2548.1, + "valid_targets_min": 693 + }, + { + "epoch": 0.17460894870862131, + "grad_norm": 0.6982550143341367, + "learning_rate": 9.927310488058153e-06, + "loss": 0.2176, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06837408244609833, + "step": 240, + "valid_targets_mean": 2482.9, + "valid_targets_min": 803 + }, + { + "epoch": 0.17824663514005093, + "grad_norm": 0.6289640785578267, + "learning_rate": 1.0134994807892004e-05, + "loss": 0.1735, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08963456749916077, + "step": 245, + "valid_targets_mean": 3239.2, + "valid_targets_min": 2525 + }, + { + "epoch": 0.18188432157148055, + "grad_norm": 0.8337454593805554, + "learning_rate": 1.0342679127725856e-05, + "loss": 0.2038, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07738061994314194, + "step": 250, + "valid_targets_mean": 3806.1, + "valid_targets_min": 2393 + }, + { + "epoch": 0.18552200800291016, + "grad_norm": 0.5551993694092996, + "learning_rate": 1.055036344755971e-05, + "loss": 0.1838, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07910811901092529, + "step": 255, + "valid_targets_mean": 2990.9, + "valid_targets_min": 804 + }, + { + "epoch": 0.18915969443433975, + "grad_norm": 0.5294640422904988, + "learning_rate": 1.0758047767393562e-05, + "loss": 0.1607, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08286316692829132, + "step": 260, + "valid_targets_mean": 4885.8, + "valid_targets_min": 3695 + }, + { + "epoch": 0.19279738086576936, + "grad_norm": 0.6498581304855479, + "learning_rate": 1.0965732087227414e-05, + "loss": 0.1512, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09251941740512848, + "step": 265, + "valid_targets_mean": 3128.8, + "valid_targets_min": 596 + }, + { + "epoch": 0.19643506729719898, + "grad_norm": 0.4027341733641733, + "learning_rate": 1.1173416407061269e-05, + "loss": 0.1438, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05383671820163727, + "step": 270, + "valid_targets_mean": 4567.8, + "valid_targets_min": 3804 + }, + { + "epoch": 0.2000727537286286, + "grad_norm": 0.6581359254080384, + "learning_rate": 1.138110072689512e-05, + "loss": 0.1563, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.061755333095788956, + "step": 275, + "valid_targets_mean": 3853.1, + "valid_targets_min": 2862 + }, + { + "epoch": 0.2037104401600582, + "grad_norm": 0.8006930209804636, + "learning_rate": 1.1588785046728972e-05, + "loss": 0.1897, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10646345466375351, + "step": 280, + "valid_targets_mean": 1072.6, + "valid_targets_min": 497 + }, + { + "epoch": 0.20734812659148782, + "grad_norm": 0.576942392061571, + "learning_rate": 1.1796469366562827e-05, + "loss": 0.1593, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08792576193809509, + "step": 285, + "valid_targets_mean": 4191.0, + "valid_targets_min": 3918 + }, + { + "epoch": 0.2109858130229174, + "grad_norm": 0.9752161257099753, + "learning_rate": 1.2004153686396678e-05, + "loss": 0.18, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0794653445482254, + "step": 290, + "valid_targets_mean": 2121.9, + "valid_targets_min": 785 + }, + { + "epoch": 0.21462349945434703, + "grad_norm": 0.4943449412606904, + "learning_rate": 1.221183800623053e-05, + "loss": 0.1312, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06692871451377869, + "step": 295, + "valid_targets_mean": 3451.6, + "valid_targets_min": 905 + }, + { + "epoch": 0.21826118588577664, + "grad_norm": 0.6312051455685184, + "learning_rate": 1.2419522326064383e-05, + "loss": 0.169, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12060990929603577, + "step": 300, + "valid_targets_mean": 2929.2, + "valid_targets_min": 657 + }, + { + "epoch": 0.22189887231720626, + "grad_norm": 0.5156105269687256, + "learning_rate": 1.2627206645898236e-05, + "loss": 0.1597, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0803900957107544, + "step": 305, + "valid_targets_mean": 3798.9, + "valid_targets_min": 3184 + }, + { + "epoch": 0.22553655874863587, + "grad_norm": 0.7439332656900396, + "learning_rate": 1.2834890965732088e-05, + "loss": 0.1411, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0830249935388565, + "step": 310, + "valid_targets_mean": 1463.9, + "valid_targets_min": 627 + }, + { + "epoch": 0.2291742451800655, + "grad_norm": 0.6995751620104704, + "learning_rate": 1.3042575285565941e-05, + "loss": 0.1636, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07560700178146362, + "step": 315, + "valid_targets_mean": 1701.5, + "valid_targets_min": 827 + }, + { + "epoch": 0.23281193161149508, + "grad_norm": 0.6884303295081623, + "learning_rate": 1.3250259605399793e-05, + "loss": 0.1654, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07281231880187988, + "step": 320, + "valid_targets_mean": 1789.9, + "valid_targets_min": 940 + }, + { + "epoch": 0.2364496180429247, + "grad_norm": 0.5830219671100422, + "learning_rate": 1.3457943925233646e-05, + "loss": 0.1854, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06614746153354645, + "step": 325, + "valid_targets_mean": 2943.8, + "valid_targets_min": 958 + }, + { + "epoch": 0.2400873044743543, + "grad_norm": 0.6239882002283308, + "learning_rate": 1.3665628245067499e-05, + "loss": 0.1547, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06254483759403229, + "step": 330, + "valid_targets_mean": 1859.2, + "valid_targets_min": 719 + }, + { + "epoch": 0.24372499090578392, + "grad_norm": 0.6486043261616687, + "learning_rate": 1.387331256490135e-05, + "loss": 0.1878, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.055449459701776505, + "step": 335, + "valid_targets_mean": 1703.9, + "valid_targets_min": 759 + }, + { + "epoch": 0.24736267733721354, + "grad_norm": 1.0718707500156668, + "learning_rate": 1.4080996884735202e-05, + "loss": 0.1851, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1858227699995041, + "step": 340, + "valid_targets_mean": 1556.6, + "valid_targets_min": 521 + }, + { + "epoch": 0.25100036376864315, + "grad_norm": 0.5349096962984586, + "learning_rate": 1.4288681204569057e-05, + "loss": 0.152, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09451160579919815, + "step": 345, + "valid_targets_mean": 3267.6, + "valid_targets_min": 1032 + }, + { + "epoch": 0.25463805020007274, + "grad_norm": 0.47475494896880566, + "learning_rate": 1.4496365524402909e-05, + "loss": 0.1313, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05547623336315155, + "step": 350, + "valid_targets_mean": 3609.0, + "valid_targets_min": 2741 + }, + { + "epoch": 0.2582757366315024, + "grad_norm": 0.5691991220973869, + "learning_rate": 1.470404984423676e-05, + "loss": 0.1508, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06443467736244202, + "step": 355, + "valid_targets_mean": 2735.8, + "valid_targets_min": 1692 + }, + { + "epoch": 0.26191342306293197, + "grad_norm": 0.5894467633231665, + "learning_rate": 1.4911734164070615e-05, + "loss": 0.1949, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0876677930355072, + "step": 360, + "valid_targets_mean": 3693.4, + "valid_targets_min": 2215 + }, + { + "epoch": 0.26555110949436156, + "grad_norm": 0.5107978377287441, + "learning_rate": 1.5119418483904467e-05, + "loss": 0.1267, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05755075439810753, + "step": 365, + "valid_targets_mean": 3109.9, + "valid_targets_min": 723 + }, + { + "epoch": 0.2691887959257912, + "grad_norm": 0.4530157770042509, + "learning_rate": 1.5327102803738318e-05, + "loss": 0.1494, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.062046118080616, + "step": 370, + "valid_targets_mean": 4984.2, + "valid_targets_min": 3282 + }, + { + "epoch": 0.2728264823572208, + "grad_norm": 0.46100374867462834, + "learning_rate": 1.5534787123572173e-05, + "loss": 0.1425, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.062385547906160355, + "step": 375, + "valid_targets_mean": 3863.4, + "valid_targets_min": 995 + }, + { + "epoch": 0.27646416878865043, + "grad_norm": 0.4496604042412126, + "learning_rate": 1.5742471443406025e-05, + "loss": 0.1248, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06038970872759819, + "step": 380, + "valid_targets_mean": 4201.0, + "valid_targets_min": 949 + }, + { + "epoch": 0.28010185522008, + "grad_norm": 0.5358921120993997, + "learning_rate": 1.5950155763239876e-05, + "loss": 0.1579, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09568338096141815, + "step": 385, + "valid_targets_mean": 4321.6, + "valid_targets_min": 2724 + }, + { + "epoch": 0.28373954165150966, + "grad_norm": 0.47351649695689224, + "learning_rate": 1.615784008307373e-05, + "loss": 0.1333, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05034826695919037, + "step": 390, + "valid_targets_mean": 3384.5, + "valid_targets_min": 424 + }, + { + "epoch": 0.28737722808293925, + "grad_norm": 0.5889980549969783, + "learning_rate": 1.6365524402907583e-05, + "loss": 0.1785, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06011715531349182, + "step": 395, + "valid_targets_mean": 2987.2, + "valid_targets_min": 531 + }, + { + "epoch": 0.29101491451436884, + "grad_norm": 0.5721381524879638, + "learning_rate": 1.6573208722741434e-05, + "loss": 0.1676, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.059478964656591415, + "step": 400, + "valid_targets_mean": 3469.5, + "valid_targets_min": 1195 + }, + { + "epoch": 0.2946526009457985, + "grad_norm": 0.5230301284684931, + "learning_rate": 1.6780893042575286e-05, + "loss": 0.127, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10258873552083969, + "step": 405, + "valid_targets_mean": 2682.8, + "valid_targets_min": 722 + }, + { + "epoch": 0.29829028737722807, + "grad_norm": 0.452219020884573, + "learning_rate": 1.698857736240914e-05, + "loss": 0.1708, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06418130546808243, + "step": 410, + "valid_targets_mean": 3496.9, + "valid_targets_min": 1251 + }, + { + "epoch": 0.3019279738086577, + "grad_norm": 0.6667040744334622, + "learning_rate": 1.7196261682242992e-05, + "loss": 0.141, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07812710106372833, + "step": 415, + "valid_targets_mean": 2912.5, + "valid_targets_min": 970 + }, + { + "epoch": 0.3055656602400873, + "grad_norm": 0.9434029787254077, + "learning_rate": 1.7403946002076844e-05, + "loss": 0.1652, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07562769949436188, + "step": 420, + "valid_targets_mean": 1423.6, + "valid_targets_min": 634 + }, + { + "epoch": 0.3092033466715169, + "grad_norm": 0.8623506268625595, + "learning_rate": 1.76116303219107e-05, + "loss": 0.179, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1142093613743782, + "step": 425, + "valid_targets_mean": 1423.9, + "valid_targets_min": 685 + }, + { + "epoch": 0.31284103310294653, + "grad_norm": 0.5453696038056621, + "learning_rate": 1.781931464174455e-05, + "loss": 0.191, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05685334652662277, + "step": 430, + "valid_targets_mean": 2932.5, + "valid_targets_min": 593 + }, + { + "epoch": 0.3164787195343761, + "grad_norm": 0.48323269434561383, + "learning_rate": 1.8026998961578402e-05, + "loss": 0.1501, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05961580574512482, + "step": 435, + "valid_targets_mean": 3945.8, + "valid_targets_min": 3129 + }, + { + "epoch": 0.32011640596580576, + "grad_norm": 0.5767639741775008, + "learning_rate": 1.8234683281412257e-05, + "loss": 0.1575, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08605828881263733, + "step": 440, + "valid_targets_mean": 3621.1, + "valid_targets_min": 859 + }, + { + "epoch": 0.32375409239723535, + "grad_norm": 0.6173773789951866, + "learning_rate": 1.8442367601246108e-05, + "loss": 0.1417, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07626128196716309, + "step": 445, + "valid_targets_mean": 2553.0, + "valid_targets_min": 616 + }, + { + "epoch": 0.327391778828665, + "grad_norm": 0.5504191023648846, + "learning_rate": 1.865005192107996e-05, + "loss": 0.1413, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0625932514667511, + "step": 450, + "valid_targets_mean": 2690.5, + "valid_targets_min": 541 + }, + { + "epoch": 0.3310294652600946, + "grad_norm": 1.063375569665897, + "learning_rate": 1.8857736240913815e-05, + "loss": 0.1364, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08778353035449982, + "step": 455, + "valid_targets_mean": 906.4, + "valid_targets_min": 598 + }, + { + "epoch": 0.33466715169152417, + "grad_norm": 0.5190304013797349, + "learning_rate": 1.9065420560747666e-05, + "loss": 0.1666, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06929586082696915, + "step": 460, + "valid_targets_mean": 3147.2, + "valid_targets_min": 1101 + }, + { + "epoch": 0.3383048381229538, + "grad_norm": 0.6112493851638771, + "learning_rate": 1.9273104880581518e-05, + "loss": 0.1282, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05838174745440483, + "step": 465, + "valid_targets_mean": 2331.8, + "valid_targets_min": 680 + }, + { + "epoch": 0.3419425245543834, + "grad_norm": 0.5455386806146095, + "learning_rate": 1.948078920041537e-05, + "loss": 0.1372, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07562888413667679, + "step": 470, + "valid_targets_mean": 2577.0, + "valid_targets_min": 774 + }, + { + "epoch": 0.34558021098581304, + "grad_norm": 0.5048148760447928, + "learning_rate": 1.9688473520249224e-05, + "loss": 0.1574, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07996127754449844, + "step": 475, + "valid_targets_mean": 4140.1, + "valid_targets_min": 1474 + }, + { + "epoch": 0.34921789741724263, + "grad_norm": 0.34697637578113577, + "learning_rate": 1.9896157840083076e-05, + "loss": 0.1215, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.042637646198272705, + "step": 480, + "valid_targets_mean": 5498.9, + "valid_targets_min": 3677 + }, + { + "epoch": 0.35285558384867227, + "grad_norm": 0.5177552414939285, + "learning_rate": 2.0103842159916927e-05, + "loss": 0.1361, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08700190484523773, + "step": 485, + "valid_targets_mean": 3352.5, + "valid_targets_min": 675 + }, + { + "epoch": 0.35649327028010186, + "grad_norm": 0.46843393532173, + "learning_rate": 2.031152647975078e-05, + "loss": 0.1298, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05084184184670448, + "step": 490, + "valid_targets_mean": 3902.5, + "valid_targets_min": 795 + }, + { + "epoch": 0.36013095671153145, + "grad_norm": 0.7128740152584873, + "learning_rate": 2.0519210799584634e-05, + "loss": 0.2679, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11826781928539276, + "step": 495, + "valid_targets_mean": 2495.0, + "valid_targets_min": 605 + }, + { + "epoch": 0.3637686431429611, + "grad_norm": 0.5193710275728378, + "learning_rate": 2.0726895119418485e-05, + "loss": 0.1277, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06751289963722229, + "step": 500, + "valid_targets_mean": 3298.6, + "valid_targets_min": 985 + }, + { + "epoch": 0.3674063295743907, + "grad_norm": 0.5586703740428514, + "learning_rate": 2.0934579439252337e-05, + "loss": 0.1469, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07501952350139618, + "step": 505, + "valid_targets_mean": 3505.2, + "valid_targets_min": 878 + }, + { + "epoch": 0.3710440160058203, + "grad_norm": 0.8798679177045897, + "learning_rate": 2.114226375908619e-05, + "loss": 0.1717, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16498541831970215, + "step": 510, + "valid_targets_mean": 2289.5, + "valid_targets_min": 1214 + }, + { + "epoch": 0.3746817024372499, + "grad_norm": 0.6458974575613675, + "learning_rate": 2.1349948078920043e-05, + "loss": 0.3094, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21044906973838806, + "step": 515, + "valid_targets_mean": 8496.5, + "valid_targets_min": 6565 + }, + { + "epoch": 0.3783193888686795, + "grad_norm": 0.5802201960909017, + "learning_rate": 2.1557632398753895e-05, + "loss": 0.3286, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14753101766109467, + "step": 520, + "valid_targets_mean": 5663.5, + "valid_targets_min": 3833 + }, + { + "epoch": 0.38195707530010914, + "grad_norm": 0.5549685950413871, + "learning_rate": 2.1765316718587747e-05, + "loss": 0.316, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14888204634189606, + "step": 525, + "valid_targets_mean": 6654.1, + "valid_targets_min": 5324 + }, + { + "epoch": 0.3855947617315387, + "grad_norm": 0.4836519954981779, + "learning_rate": 2.19730010384216e-05, + "loss": 0.2846, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1406727433204651, + "step": 530, + "valid_targets_mean": 7635.1, + "valid_targets_min": 5672 + }, + { + "epoch": 0.38923244816296837, + "grad_norm": 0.4153837667667638, + "learning_rate": 2.2180685358255453e-05, + "loss": 0.2896, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1318293809890747, + "step": 535, + "valid_targets_mean": 8517.0, + "valid_targets_min": 5045 + }, + { + "epoch": 0.39287013459439796, + "grad_norm": 0.4632888651056554, + "learning_rate": 2.2388369678089305e-05, + "loss": 0.2887, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14035508036613464, + "step": 540, + "valid_targets_mean": 6759.1, + "valid_targets_min": 3152 + }, + { + "epoch": 0.3965078210258276, + "grad_norm": 0.44324252000921394, + "learning_rate": 2.259605399792316e-05, + "loss": 0.3005, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13268306851387024, + "step": 545, + "valid_targets_mean": 7822.8, + "valid_targets_min": 4820 + }, + { + "epoch": 0.4001455074572572, + "grad_norm": 0.4661949582387517, + "learning_rate": 2.280373831775701e-05, + "loss": 0.2843, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11786442995071411, + "step": 550, + "valid_targets_mean": 6852.2, + "valid_targets_min": 4135 + }, + { + "epoch": 0.4037831938886868, + "grad_norm": 0.5854103643675843, + "learning_rate": 2.3011422637590863e-05, + "loss": 0.2825, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12765797972679138, + "step": 555, + "valid_targets_mean": 4794.2, + "valid_targets_min": 3230 + }, + { + "epoch": 0.4074208803201164, + "grad_norm": 0.42267593920092394, + "learning_rate": 2.3219106957424717e-05, + "loss": 0.2664, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14371243119239807, + "step": 560, + "valid_targets_mean": 7691.0, + "valid_targets_min": 5900 + }, + { + "epoch": 0.411058566751546, + "grad_norm": 0.5060591256008664, + "learning_rate": 2.3426791277258566e-05, + "loss": 0.2782, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1315077841281891, + "step": 565, + "valid_targets_mean": 6510.9, + "valid_targets_min": 5639 + }, + { + "epoch": 0.41469625318297565, + "grad_norm": 0.4732618707051439, + "learning_rate": 2.363447559709242e-05, + "loss": 0.2624, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13586567342281342, + "step": 570, + "valid_targets_mean": 6096.8, + "valid_targets_min": 4669 + }, + { + "epoch": 0.41833393961440524, + "grad_norm": 0.4744321534333235, + "learning_rate": 2.3842159916926275e-05, + "loss": 0.2637, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1362125426530838, + "step": 575, + "valid_targets_mean": 6812.2, + "valid_targets_min": 5160 + }, + { + "epoch": 0.4219716260458348, + "grad_norm": 0.46235630753899015, + "learning_rate": 2.4049844236760124e-05, + "loss": 0.2554, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11827521026134491, + "step": 580, + "valid_targets_mean": 6980.4, + "valid_targets_min": 5151 + }, + { + "epoch": 0.42560931247726447, + "grad_norm": 0.4486251033038335, + "learning_rate": 2.425752855659398e-05, + "loss": 0.2594, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1372971385717392, + "step": 585, + "valid_targets_mean": 7417.1, + "valid_targets_min": 5932 + }, + { + "epoch": 0.42924699890869406, + "grad_norm": 0.47575377239899874, + "learning_rate": 2.4465212876427833e-05, + "loss": 0.2632, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12701694667339325, + "step": 590, + "valid_targets_mean": 6258.5, + "valid_targets_min": 5569 + }, + { + "epoch": 0.4328846853401237, + "grad_norm": 0.49592751895065207, + "learning_rate": 2.467289719626168e-05, + "loss": 0.2529, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14206691086292267, + "step": 595, + "valid_targets_mean": 7689.1, + "valid_targets_min": 6223 + }, + { + "epoch": 0.4365223717715533, + "grad_norm": 0.6471209264214034, + "learning_rate": 2.4880581516095537e-05, + "loss": 0.265, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15987734496593475, + "step": 600, + "valid_targets_mean": 4371.9, + "valid_targets_min": 1172 + }, + { + "epoch": 0.44016005820298293, + "grad_norm": 0.47786105298787274, + "learning_rate": 2.508826583592939e-05, + "loss": 0.277, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1452629715204239, + "step": 605, + "valid_targets_mean": 6936.2, + "valid_targets_min": 5619 + }, + { + "epoch": 0.4437977446344125, + "grad_norm": 0.7950312960024684, + "learning_rate": 2.529595015576324e-05, + "loss": 0.2631, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20515002310276031, + "step": 610, + "valid_targets_mean": 3273.2, + "valid_targets_min": 442 + }, + { + "epoch": 0.4474354310658421, + "grad_norm": 0.43379463584417743, + "learning_rate": 2.5503634475597095e-05, + "loss": 0.2398, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12256474792957306, + "step": 615, + "valid_targets_mean": 8141.1, + "valid_targets_min": 5061 + }, + { + "epoch": 0.45107311749727175, + "grad_norm": 0.4791452581941102, + "learning_rate": 2.571131879543095e-05, + "loss": 0.2522, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12769924104213715, + "step": 620, + "valid_targets_mean": 7371.1, + "valid_targets_min": 5112 + }, + { + "epoch": 0.45471080392870133, + "grad_norm": 0.5015500009570756, + "learning_rate": 2.5919003115264798e-05, + "loss": 0.252, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11168272793292999, + "step": 625, + "valid_targets_mean": 5659.4, + "valid_targets_min": 4626 + }, + { + "epoch": 0.458348490360131, + "grad_norm": 0.45226968241528215, + "learning_rate": 2.6126687435098653e-05, + "loss": 0.2592, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11406582593917847, + "step": 630, + "valid_targets_mean": 7801.8, + "valid_targets_min": 6172 + }, + { + "epoch": 0.46198617679156057, + "grad_norm": 0.45719440616058576, + "learning_rate": 2.6334371754932508e-05, + "loss": 0.2496, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12513884902000427, + "step": 635, + "valid_targets_mean": 7547.4, + "valid_targets_min": 5804 + }, + { + "epoch": 0.46562386322299015, + "grad_norm": 0.46382096735096817, + "learning_rate": 2.6542056074766356e-05, + "loss": 0.2363, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1135328859090805, + "step": 640, + "valid_targets_mean": 6474.2, + "valid_targets_min": 5050 + }, + { + "epoch": 0.4692615496544198, + "grad_norm": 0.425899925981405, + "learning_rate": 2.674974039460021e-05, + "loss": 0.2184, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11356712132692337, + "step": 645, + "valid_targets_mean": 7225.4, + "valid_targets_min": 5298 + }, + { + "epoch": 0.4728992360858494, + "grad_norm": 0.5150548435061688, + "learning_rate": 2.6957424714434066e-05, + "loss": 0.2188, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10201641917228699, + "step": 650, + "valid_targets_mean": 6442.9, + "valid_targets_min": 4679 + }, + { + "epoch": 0.476536922517279, + "grad_norm": 0.569601064417518, + "learning_rate": 2.7165109034267914e-05, + "loss": 0.2608, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10547998547554016, + "step": 655, + "valid_targets_mean": 3651.9, + "valid_targets_min": 1801 + }, + { + "epoch": 0.4801746089487086, + "grad_norm": 0.5907452340285829, + "learning_rate": 2.737279335410177e-05, + "loss": 0.2663, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1430007815361023, + "step": 660, + "valid_targets_mean": 7512.8, + "valid_targets_min": 5331 + }, + { + "epoch": 0.48381229538013826, + "grad_norm": 0.46561675238097866, + "learning_rate": 2.758047767393562e-05, + "loss": 0.2575, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13377757370471954, + "step": 665, + "valid_targets_mean": 7108.8, + "valid_targets_min": 4822 + }, + { + "epoch": 0.48744998181156785, + "grad_norm": 0.4759174559916989, + "learning_rate": 2.778816199376947e-05, + "loss": 0.2667, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11889410018920898, + "step": 670, + "valid_targets_mean": 7125.8, + "valid_targets_min": 4623 + }, + { + "epoch": 0.49108766824299743, + "grad_norm": 0.44543953356079957, + "learning_rate": 2.7995846313603327e-05, + "loss": 0.2619, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11849747598171234, + "step": 675, + "valid_targets_mean": 6269.2, + "valid_targets_min": 5458 + }, + { + "epoch": 0.4947253546744271, + "grad_norm": 0.46117082801844766, + "learning_rate": 2.8203530633437178e-05, + "loss": 0.2471, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1260770857334137, + "step": 680, + "valid_targets_mean": 7455.1, + "valid_targets_min": 6008 + }, + { + "epoch": 0.49836304110585666, + "grad_norm": 0.48696308104720215, + "learning_rate": 2.841121495327103e-05, + "loss": 0.255, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12613046169281006, + "step": 685, + "valid_targets_mean": 7104.4, + "valid_targets_min": 5386 + }, + { + "epoch": 0.5020007275372863, + "grad_norm": 0.4673451634540958, + "learning_rate": 2.8618899273104885e-05, + "loss": 0.2451, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12733058631420135, + "step": 690, + "valid_targets_mean": 6532.6, + "valid_targets_min": 4855 + }, + { + "epoch": 0.5056384139687159, + "grad_norm": 0.457752166250232, + "learning_rate": 2.8826583592938736e-05, + "loss": 0.2513, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12072950601577759, + "step": 695, + "valid_targets_mean": 6743.9, + "valid_targets_min": 5613 + }, + { + "epoch": 0.5092761004001455, + "grad_norm": 0.4696126068572868, + "learning_rate": 2.9034267912772588e-05, + "loss": 0.2494, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12563422322273254, + "step": 700, + "valid_targets_mean": 6373.6, + "valid_targets_min": 4961 + }, + { + "epoch": 0.5129137868315751, + "grad_norm": 0.5127191158706937, + "learning_rate": 2.924195223260644e-05, + "loss": 0.2539, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13112369179725647, + "step": 705, + "valid_targets_mean": 6461.2, + "valid_targets_min": 5286 + }, + { + "epoch": 0.5165514732630048, + "grad_norm": 0.48857734236249845, + "learning_rate": 2.9449636552440294e-05, + "loss": 0.2528, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13163086771965027, + "step": 710, + "valid_targets_mean": 6780.5, + "valid_targets_min": 5201 + }, + { + "epoch": 0.5201891596944344, + "grad_norm": 0.526601905395271, + "learning_rate": 2.9657320872274146e-05, + "loss": 0.2507, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12659509479999542, + "step": 715, + "valid_targets_mean": 5710.8, + "valid_targets_min": 4838 + }, + { + "epoch": 0.5238268461258639, + "grad_norm": 0.8072315324082411, + "learning_rate": 2.9865005192107997e-05, + "loss": 0.2356, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20060086250305176, + "step": 720, + "valid_targets_mean": 3235.6, + "valid_targets_min": 176 + }, + { + "epoch": 0.5274645325572935, + "grad_norm": 0.4908360570855165, + "learning_rate": 3.0072689511941852e-05, + "loss": 0.2486, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11582855135202408, + "step": 725, + "valid_targets_mean": 6674.8, + "valid_targets_min": 4582 + }, + { + "epoch": 0.5311022189887231, + "grad_norm": 0.4812844302279063, + "learning_rate": 3.0280373831775704e-05, + "loss": 0.2414, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11961640417575836, + "step": 730, + "valid_targets_mean": 6157.8, + "valid_targets_min": 5385 + }, + { + "epoch": 0.5347399054201528, + "grad_norm": 0.5099624490032726, + "learning_rate": 3.0488058151609555e-05, + "loss": 0.246, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11810891330242157, + "step": 735, + "valid_targets_mean": 6006.2, + "valid_targets_min": 4478 + }, + { + "epoch": 0.5383775918515824, + "grad_norm": 0.4738760724299624, + "learning_rate": 3.0695742471443414e-05, + "loss": 0.2406, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12052886188030243, + "step": 740, + "valid_targets_mean": 5893.4, + "valid_targets_min": 3911 + }, + { + "epoch": 0.542015278283012, + "grad_norm": 0.49294594912111095, + "learning_rate": 3.090342679127726e-05, + "loss": 0.2389, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11911094188690186, + "step": 745, + "valid_targets_mean": 6134.2, + "valid_targets_min": 5318 + }, + { + "epoch": 0.5456529647144416, + "grad_norm": 0.5443170969822231, + "learning_rate": 3.111111111111112e-05, + "loss": 0.2488, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12518388032913208, + "step": 750, + "valid_targets_mean": 5474.4, + "valid_targets_min": 4959 + }, + { + "epoch": 0.5492906511458713, + "grad_norm": 0.875863087414027, + "learning_rate": 3.131879543094497e-05, + "loss": 0.2437, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08180184662342072, + "step": 755, + "valid_targets_mean": 1804.1, + "valid_targets_min": 842 + }, + { + "epoch": 0.5529283375773009, + "grad_norm": 0.4811814006766875, + "learning_rate": 3.152647975077882e-05, + "loss": 0.2191, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11149038374423981, + "step": 760, + "valid_targets_mean": 6118.8, + "valid_targets_min": 4899 + }, + { + "epoch": 0.5565660240087305, + "grad_norm": 0.49671690646502353, + "learning_rate": 3.173416407061267e-05, + "loss": 0.2358, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11093097180128098, + "step": 765, + "valid_targets_mean": 6773.8, + "valid_targets_min": 5046 + }, + { + "epoch": 0.56020371044016, + "grad_norm": 0.49430500829651336, + "learning_rate": 3.194184839044652e-05, + "loss": 0.2441, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1279737949371338, + "step": 770, + "valid_targets_mean": 6633.6, + "valid_targets_min": 5284 + }, + { + "epoch": 0.5638413968715896, + "grad_norm": 0.4886940177738384, + "learning_rate": 3.2149532710280374e-05, + "loss": 0.2367, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11924019455909729, + "step": 775, + "valid_targets_mean": 6055.2, + "valid_targets_min": 4928 + }, + { + "epoch": 0.5674790833030193, + "grad_norm": 0.5077665170673344, + "learning_rate": 3.235721703011423e-05, + "loss": 0.2312, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12405657768249512, + "step": 780, + "valid_targets_mean": 7259.4, + "valid_targets_min": 5758 + }, + { + "epoch": 0.5711167697344489, + "grad_norm": 0.5255779549241664, + "learning_rate": 3.2564901349948084e-05, + "loss": 0.2354, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12088797241449356, + "step": 785, + "valid_targets_mean": 5731.8, + "valid_targets_min": 4629 + }, + { + "epoch": 0.5747544561658785, + "grad_norm": 1.1927697524021514, + "learning_rate": 3.2772585669781936e-05, + "loss": 0.289, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29794391989707947, + "step": 790, + "valid_targets_mean": 5360.9, + "valid_targets_min": 2317 + }, + { + "epoch": 0.5783921425973081, + "grad_norm": 0.6916666593769295, + "learning_rate": 3.298026998961579e-05, + "loss": 0.4046, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21879465878009796, + "step": 795, + "valid_targets_mean": 5303.1, + "valid_targets_min": 533 + }, + { + "epoch": 0.5820298290287377, + "grad_norm": 0.7009271611212158, + "learning_rate": 3.318795430944964e-05, + "loss": 0.3911, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20250442624092102, + "step": 800, + "valid_targets_mean": 4545.8, + "valid_targets_min": 835 + }, + { + "epoch": 0.5856675154601674, + "grad_norm": 0.7221739606791518, + "learning_rate": 3.339563862928349e-05, + "loss": 0.3821, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.216229647397995, + "step": 805, + "valid_targets_mean": 3618.9, + "valid_targets_min": 1201 + }, + { + "epoch": 0.589305201891597, + "grad_norm": 0.663334096587329, + "learning_rate": 3.360332294911734e-05, + "loss": 0.3609, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1709412932395935, + "step": 810, + "valid_targets_mean": 4253.5, + "valid_targets_min": 1880 + }, + { + "epoch": 0.5929428883230266, + "grad_norm": 0.6039335155140695, + "learning_rate": 3.3811007268951193e-05, + "loss": 0.3898, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26600396633148193, + "step": 815, + "valid_targets_mean": 6610.1, + "valid_targets_min": 2171 + }, + { + "epoch": 0.5965805747544561, + "grad_norm": 0.5500413661085425, + "learning_rate": 3.401869158878505e-05, + "loss": 0.3571, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20335346460342407, + "step": 820, + "valid_targets_mean": 7171.4, + "valid_targets_min": 4500 + }, + { + "epoch": 0.6002182611858857, + "grad_norm": 0.5158405862398696, + "learning_rate": 3.42263759086189e-05, + "loss": 0.3635, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18620151281356812, + "step": 825, + "valid_targets_mean": 6207.4, + "valid_targets_min": 1855 + }, + { + "epoch": 0.6038559476173154, + "grad_norm": 0.609054667686504, + "learning_rate": 3.4434060228452755e-05, + "loss": 0.3652, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19188907742500305, + "step": 830, + "valid_targets_mean": 4075.9, + "valid_targets_min": 1024 + }, + { + "epoch": 0.607493634048745, + "grad_norm": 0.6234712256414725, + "learning_rate": 3.4641744548286606e-05, + "loss": 0.3697, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15819114446640015, + "step": 835, + "valid_targets_mean": 4767.1, + "valid_targets_min": 2107 + }, + { + "epoch": 0.6111313204801746, + "grad_norm": 0.6885460593445002, + "learning_rate": 3.484942886812046e-05, + "loss": 0.3685, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21019208431243896, + "step": 840, + "valid_targets_mean": 4211.4, + "valid_targets_min": 802 + }, + { + "epoch": 0.6147690069116042, + "grad_norm": 0.49073614618696165, + "learning_rate": 3.505711318795431e-05, + "loss": 0.3491, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18104946613311768, + "step": 845, + "valid_targets_mean": 5345.1, + "valid_targets_min": 1597 + }, + { + "epoch": 0.6184066933430338, + "grad_norm": 1.0328278425308346, + "learning_rate": 3.526479750778816e-05, + "loss": 0.3652, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1876482218503952, + "step": 850, + "valid_targets_mean": 3671.4, + "valid_targets_min": 1650 + }, + { + "epoch": 0.6220443797744635, + "grad_norm": 0.6041569049820559, + "learning_rate": 3.547248182762202e-05, + "loss": 0.3552, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19097556173801422, + "step": 855, + "valid_targets_mean": 4799.1, + "valid_targets_min": 1385 + }, + { + "epoch": 0.6256820662058931, + "grad_norm": 0.6656085197271374, + "learning_rate": 3.568016614745587e-05, + "loss": 0.3623, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2004096508026123, + "step": 860, + "valid_targets_mean": 4023.5, + "valid_targets_min": 2908 + }, + { + "epoch": 0.6293197526373226, + "grad_norm": 0.5603623922767227, + "learning_rate": 3.588785046728972e-05, + "loss": 0.3692, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16860738396644592, + "step": 865, + "valid_targets_mean": 5064.6, + "valid_targets_min": 2785 + }, + { + "epoch": 0.6329574390687522, + "grad_norm": 0.6785084293942854, + "learning_rate": 3.6095534787123574e-05, + "loss": 0.3592, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17804642021656036, + "step": 870, + "valid_targets_mean": 3886.4, + "valid_targets_min": 1210 + }, + { + "epoch": 0.6365951255001819, + "grad_norm": 0.6515652121612323, + "learning_rate": 3.6303219106957426e-05, + "loss": 0.3707, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.205357164144516, + "step": 875, + "valid_targets_mean": 3708.8, + "valid_targets_min": 1064 + }, + { + "epoch": 0.6402328119316115, + "grad_norm": 0.8631266636797919, + "learning_rate": 3.651090342679128e-05, + "loss": 0.3778, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1885777711868286, + "step": 880, + "valid_targets_mean": 3276.2, + "valid_targets_min": 1678 + }, + { + "epoch": 0.6438704983630411, + "grad_norm": 0.6534275536962839, + "learning_rate": 3.6718587746625135e-05, + "loss": 0.3736, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1933945119380951, + "step": 885, + "valid_targets_mean": 3520.1, + "valid_targets_min": 671 + }, + { + "epoch": 0.6475081847944707, + "grad_norm": 0.6957210503826712, + "learning_rate": 3.692627206645899e-05, + "loss": 0.3601, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17438723146915436, + "step": 890, + "valid_targets_mean": 3488.5, + "valid_targets_min": 2000 + }, + { + "epoch": 0.6511458712259003, + "grad_norm": 0.6793387332715758, + "learning_rate": 3.713395638629284e-05, + "loss": 0.3474, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21124738454818726, + "step": 895, + "valid_targets_mean": 4402.6, + "valid_targets_min": 1532 + }, + { + "epoch": 0.65478355765733, + "grad_norm": 0.643009367537289, + "learning_rate": 3.734164070612669e-05, + "loss": 0.35, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13972952961921692, + "step": 900, + "valid_targets_mean": 3749.4, + "valid_targets_min": 1051 + }, + { + "epoch": 0.6584212440887596, + "grad_norm": 0.6459988651401798, + "learning_rate": 3.754932502596054e-05, + "loss": 0.3419, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16900065541267395, + "step": 905, + "valid_targets_mean": 3368.1, + "valid_targets_min": 1345 + }, + { + "epoch": 0.6620589305201892, + "grad_norm": 0.5925500708543067, + "learning_rate": 3.775700934579439e-05, + "loss": 0.363, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17137889564037323, + "step": 910, + "valid_targets_mean": 4128.9, + "valid_targets_min": 2231 + }, + { + "epoch": 0.6656966169516187, + "grad_norm": 0.6680369269614936, + "learning_rate": 3.796469366562825e-05, + "loss": 0.3619, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15056374669075012, + "step": 915, + "valid_targets_mean": 3672.6, + "valid_targets_min": 2127 + }, + { + "epoch": 0.6693343033830483, + "grad_norm": 0.8809109369729928, + "learning_rate": 3.8172377985462096e-05, + "loss": 0.3449, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1554284691810608, + "step": 920, + "valid_targets_mean": 3006.8, + "valid_targets_min": 1265 + }, + { + "epoch": 0.672971989814478, + "grad_norm": 1.1584429464503139, + "learning_rate": 3.8380062305295954e-05, + "loss": 0.3467, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1504942774772644, + "step": 925, + "valid_targets_mean": 2981.9, + "valid_targets_min": 908 + }, + { + "epoch": 0.6766096762459076, + "grad_norm": 0.6853531973745671, + "learning_rate": 3.8587746625129806e-05, + "loss": 0.3474, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20189368724822998, + "step": 930, + "valid_targets_mean": 3147.5, + "valid_targets_min": 1264 + }, + { + "epoch": 0.6802473626773372, + "grad_norm": 0.724690761810996, + "learning_rate": 3.879543094496366e-05, + "loss": 0.3482, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1616716831922531, + "step": 935, + "valid_targets_mean": 2517.9, + "valid_targets_min": 926 + }, + { + "epoch": 0.6838850491087668, + "grad_norm": 0.6713789936152544, + "learning_rate": 3.900311526479751e-05, + "loss": 0.3504, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1724017858505249, + "step": 940, + "valid_targets_mean": 3204.5, + "valid_targets_min": 1016 + }, + { + "epoch": 0.6875227355401964, + "grad_norm": 0.678340338765401, + "learning_rate": 3.921079958463137e-05, + "loss": 0.338, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17159214615821838, + "step": 945, + "valid_targets_mean": 3525.0, + "valid_targets_min": 2019 + }, + { + "epoch": 0.6911604219716261, + "grad_norm": 0.6156917804759984, + "learning_rate": 3.941848390446521e-05, + "loss": 0.3529, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19824552536010742, + "step": 950, + "valid_targets_mean": 4745.4, + "valid_targets_min": 3023 + }, + { + "epoch": 0.6947981084030557, + "grad_norm": 0.8036976450483508, + "learning_rate": 3.962616822429907e-05, + "loss": 0.3409, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15191900730133057, + "step": 955, + "valid_targets_mean": 3503.6, + "valid_targets_min": 1110 + }, + { + "epoch": 0.6984357948344853, + "grad_norm": 0.637659985164992, + "learning_rate": 3.983385254413292e-05, + "loss": 0.3491, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20231162011623383, + "step": 960, + "valid_targets_mean": 4206.5, + "valid_targets_min": 1460 + }, + { + "epoch": 0.7020734812659148, + "grad_norm": 0.657532785386383, + "learning_rate": 3.999999868458321e-05, + "loss": 0.3479, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18427856266498566, + "step": 965, + "valid_targets_mean": 3938.6, + "valid_targets_min": 2558 + }, + { + "epoch": 0.7057111676973445, + "grad_norm": 0.6508185681652344, + "learning_rate": 3.999995264501363e-05, + "loss": 0.358, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16197270154953003, + "step": 970, + "valid_targets_mean": 3421.1, + "valid_targets_min": 1673 + }, + { + "epoch": 0.7093488541287741, + "grad_norm": 0.6173923685788815, + "learning_rate": 3.9999840834777436e-05, + "loss": 0.3389, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16032695770263672, + "step": 975, + "valid_targets_mean": 3549.6, + "valid_targets_min": 1860 + }, + { + "epoch": 0.7129865405602037, + "grad_norm": 0.5683356466976641, + "learning_rate": 3.999966325424233e-05, + "loss": 0.3441, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15459167957305908, + "step": 980, + "valid_targets_mean": 4414.0, + "valid_targets_min": 2495 + }, + { + "epoch": 0.7166242269916333, + "grad_norm": 0.6760392355089757, + "learning_rate": 3.999941990399228e-05, + "loss": 0.336, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1955871880054474, + "step": 985, + "valid_targets_mean": 3861.9, + "valid_targets_min": 1945 + }, + { + "epoch": 0.7202619134230629, + "grad_norm": 0.6474855855494979, + "learning_rate": 3.999911078482757e-05, + "loss": 0.3522, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17476028203964233, + "step": 990, + "valid_targets_mean": 4488.5, + "valid_targets_min": 1908 + }, + { + "epoch": 0.7238995998544926, + "grad_norm": 0.578406454783423, + "learning_rate": 3.9998735897764735e-05, + "loss": 0.3466, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17575867474079132, + "step": 995, + "valid_targets_mean": 3507.5, + "valid_targets_min": 1880 + }, + { + "epoch": 0.7275372862859222, + "grad_norm": 0.5809996191964969, + "learning_rate": 3.9998295244036615e-05, + "loss": 0.3383, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17679500579833984, + "step": 1000, + "valid_targets_mean": 4491.5, + "valid_targets_min": 2492 + }, + { + "epoch": 0.7311749727173518, + "grad_norm": 0.6139357782885588, + "learning_rate": 3.999778882509231e-05, + "loss": 0.3408, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17411646246910095, + "step": 1005, + "valid_targets_mean": 4228.2, + "valid_targets_min": 3004 + }, + { + "epoch": 0.7348126591487814, + "grad_norm": 0.6416262420677474, + "learning_rate": 3.999721664259721e-05, + "loss": 0.3328, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17080555856227875, + "step": 1010, + "valid_targets_mean": 3899.9, + "valid_targets_min": 1833 + }, + { + "epoch": 0.7384503455802109, + "grad_norm": 0.6853139429981882, + "learning_rate": 3.999657869843295e-05, + "loss": 0.3432, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1804819405078888, + "step": 1015, + "valid_targets_mean": 3114.2, + "valid_targets_min": 1182 + }, + { + "epoch": 0.7420880320116406, + "grad_norm": 0.9888359541116599, + "learning_rate": 3.999587499469746e-05, + "loss": 0.3331, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16281211376190186, + "step": 1020, + "valid_targets_mean": 4926.9, + "valid_targets_min": 1869 + }, + { + "epoch": 0.7457257184430702, + "grad_norm": 0.6318118856129171, + "learning_rate": 3.999510553370487e-05, + "loss": 0.3433, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2049168050289154, + "step": 1025, + "valid_targets_mean": 4382.4, + "valid_targets_min": 2498 + }, + { + "epoch": 0.7493634048744998, + "grad_norm": 0.6522475950917355, + "learning_rate": 3.999427031798558e-05, + "loss": 0.3627, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17029546201229095, + "step": 1030, + "valid_targets_mean": 3098.8, + "valid_targets_min": 988 + }, + { + "epoch": 0.7530010913059294, + "grad_norm": 0.5866623964266694, + "learning_rate": 3.9993369350286265e-05, + "loss": 0.3337, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17328611016273499, + "step": 1035, + "valid_targets_mean": 3733.5, + "valid_targets_min": 1534 + }, + { + "epoch": 0.756638777737359, + "grad_norm": 0.6012801907183116, + "learning_rate": 3.999240263356977e-05, + "loss": 0.338, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15440979599952698, + "step": 1040, + "valid_targets_mean": 4108.1, + "valid_targets_min": 2252 + }, + { + "epoch": 0.7602764641687887, + "grad_norm": 0.7292669815240751, + "learning_rate": 3.999137017101518e-05, + "loss": 0.3431, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16815641522407532, + "step": 1045, + "valid_targets_mean": 3446.6, + "valid_targets_min": 1436 + }, + { + "epoch": 0.7639141506002183, + "grad_norm": 0.6624632193814274, + "learning_rate": 3.999027196601781e-05, + "loss": 0.339, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1592995822429657, + "step": 1050, + "valid_targets_mean": 3819.2, + "valid_targets_min": 2333 + }, + { + "epoch": 0.7675518370316479, + "grad_norm": 0.5831565821697107, + "learning_rate": 3.9989108022189134e-05, + "loss": 0.3276, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1494516134262085, + "step": 1055, + "valid_targets_mean": 3889.8, + "valid_targets_min": 1272 + }, + { + "epoch": 0.7711895234630775, + "grad_norm": 0.6773162074311764, + "learning_rate": 3.998787834335684e-05, + "loss": 0.327, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16385817527770996, + "step": 1060, + "valid_targets_mean": 3233.0, + "valid_targets_min": 1319 + }, + { + "epoch": 0.774827209894507, + "grad_norm": 0.5587288746412141, + "learning_rate": 3.998658293356477e-05, + "loss": 0.3117, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13708826899528503, + "step": 1065, + "valid_targets_mean": 3186.6, + "valid_targets_min": 1194 + }, + { + "epoch": 0.7784648963259367, + "grad_norm": 0.7661568217219711, + "learning_rate": 3.9985221797072934e-05, + "loss": 0.3541, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14759965240955353, + "step": 1070, + "valid_targets_mean": 2710.6, + "valid_targets_min": 1453 + }, + { + "epoch": 0.7821025827573663, + "grad_norm": 0.5378818312087793, + "learning_rate": 3.998379493835749e-05, + "loss": 0.3246, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13401705026626587, + "step": 1075, + "valid_targets_mean": 5344.5, + "valid_targets_min": 3603 + }, + { + "epoch": 0.7857402691887959, + "grad_norm": 0.6058497529333309, + "learning_rate": 3.9982302362110716e-05, + "loss": 0.3434, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.152265727519989, + "step": 1080, + "valid_targets_mean": 3067.5, + "valid_targets_min": 1256 + }, + { + "epoch": 0.7893779556202255, + "grad_norm": 0.6761151400007801, + "learning_rate": 3.9980744073241016e-05, + "loss": 0.337, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22401230037212372, + "step": 1085, + "valid_targets_mean": 4443.9, + "valid_targets_min": 1336 + }, + { + "epoch": 0.7930156420516552, + "grad_norm": 0.660623663818504, + "learning_rate": 3.997912007687288e-05, + "loss": 0.3472, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17278507351875305, + "step": 1090, + "valid_targets_mean": 3581.8, + "valid_targets_min": 1275 + }, + { + "epoch": 0.7966533284830848, + "grad_norm": 0.685978446776929, + "learning_rate": 3.997743037834689e-05, + "loss": 0.3331, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1888178288936615, + "step": 1095, + "valid_targets_mean": 3770.1, + "valid_targets_min": 2202 + }, + { + "epoch": 0.8002910149145144, + "grad_norm": 0.6187666950384704, + "learning_rate": 3.99756749832197e-05, + "loss": 0.3307, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19107113778591156, + "step": 1100, + "valid_targets_mean": 4112.8, + "valid_targets_min": 3332 + }, + { + "epoch": 0.803928701345944, + "grad_norm": 0.7543573492952697, + "learning_rate": 3.997385389726398e-05, + "loss": 0.3258, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1767590045928955, + "step": 1105, + "valid_targets_mean": 3260.0, + "valid_targets_min": 2195 + }, + { + "epoch": 0.8075663877773736, + "grad_norm": 0.6339483316384431, + "learning_rate": 3.997196712646845e-05, + "loss": 0.3277, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16353940963745117, + "step": 1110, + "valid_targets_mean": 4039.6, + "valid_targets_min": 2343 + }, + { + "epoch": 0.8112040742088032, + "grad_norm": 0.6460202253848272, + "learning_rate": 3.997001467703784e-05, + "loss": 0.3203, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15101701021194458, + "step": 1115, + "valid_targets_mean": 2678.8, + "valid_targets_min": 1026 + }, + { + "epoch": 0.8148417606402328, + "grad_norm": 0.6770982822459237, + "learning_rate": 3.9967996555392866e-05, + "loss": 0.3183, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16013629734516144, + "step": 1120, + "valid_targets_mean": 3754.5, + "valid_targets_min": 1042 + }, + { + "epoch": 0.8184794470716624, + "grad_norm": 0.6554031814224496, + "learning_rate": 3.996591276817019e-05, + "loss": 0.3134, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1602095365524292, + "step": 1125, + "valid_targets_mean": 3301.6, + "valid_targets_min": 1274 + }, + { + "epoch": 0.822117133503092, + "grad_norm": 0.6316363854861172, + "learning_rate": 3.9963763322222443e-05, + "loss": 0.3282, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1596628576517105, + "step": 1130, + "valid_targets_mean": 3632.2, + "valid_targets_min": 914 + }, + { + "epoch": 0.8257548199345216, + "grad_norm": 0.5693198775342707, + "learning_rate": 3.9961548224618153e-05, + "loss": 0.3555, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14918836951255798, + "step": 1135, + "valid_targets_mean": 4342.5, + "valid_targets_min": 2093 + }, + { + "epoch": 0.8293925063659513, + "grad_norm": 0.718283229657347, + "learning_rate": 3.995926748264178e-05, + "loss": 0.3278, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17033600807189941, + "step": 1140, + "valid_targets_mean": 3609.6, + "valid_targets_min": 1683 + }, + { + "epoch": 0.8330301927973809, + "grad_norm": 0.6227754865681429, + "learning_rate": 3.995692110379362e-05, + "loss": 0.3375, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15690159797668457, + "step": 1145, + "valid_targets_mean": 4096.2, + "valid_targets_min": 2253 + }, + { + "epoch": 0.8366678792288105, + "grad_norm": 0.5631873230982087, + "learning_rate": 3.995450909578984e-05, + "loss": 0.3204, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15026238560676575, + "step": 1150, + "valid_targets_mean": 3794.4, + "valid_targets_min": 1789 + }, + { + "epoch": 0.8403055656602401, + "grad_norm": 0.6299268840476133, + "learning_rate": 3.995203146656243e-05, + "loss": 0.3309, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16208410263061523, + "step": 1155, + "valid_targets_mean": 3753.8, + "valid_targets_min": 1586 + }, + { + "epoch": 0.8439432520916696, + "grad_norm": 0.6556686634431741, + "learning_rate": 3.994948822425918e-05, + "loss": 0.3241, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1465681493282318, + "step": 1160, + "valid_targets_mean": 2995.5, + "valid_targets_min": 2033 + }, + { + "epoch": 0.8475809385230993, + "grad_norm": 0.5613797827611839, + "learning_rate": 3.9946879377243644e-05, + "loss": 0.3109, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1484929621219635, + "step": 1165, + "valid_targets_mean": 3690.5, + "valid_targets_min": 1616 + }, + { + "epoch": 0.8512186249545289, + "grad_norm": 0.5598876225010666, + "learning_rate": 3.9944204934095114e-05, + "loss": 0.3329, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14305289089679718, + "step": 1170, + "valid_targets_mean": 3303.9, + "valid_targets_min": 2000 + }, + { + "epoch": 0.8548563113859585, + "grad_norm": 0.6815669845391257, + "learning_rate": 3.9941464903608614e-05, + "loss": 0.3214, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18238401412963867, + "step": 1175, + "valid_targets_mean": 3788.9, + "valid_targets_min": 1750 + }, + { + "epoch": 0.8584939978173881, + "grad_norm": 0.6646858482826106, + "learning_rate": 3.9938659294794854e-05, + "loss": 0.3326, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1841597557067871, + "step": 1180, + "valid_targets_mean": 3898.2, + "valid_targets_min": 2126 + }, + { + "epoch": 0.8621316842488177, + "grad_norm": 0.6102849578898638, + "learning_rate": 3.993578811688018e-05, + "loss": 0.3413, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17783446609973907, + "step": 1185, + "valid_targets_mean": 3923.1, + "valid_targets_min": 3049 + }, + { + "epoch": 0.8657693706802474, + "grad_norm": 0.6012551113248058, + "learning_rate": 3.993285137930658e-05, + "loss": 0.331, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18684765696525574, + "step": 1190, + "valid_targets_mean": 4337.1, + "valid_targets_min": 1571 + }, + { + "epoch": 0.869407057111677, + "grad_norm": 0.6459728415124562, + "learning_rate": 3.992984909173165e-05, + "loss": 0.3514, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15876203775405884, + "step": 1195, + "valid_targets_mean": 2998.0, + "valid_targets_min": 1487 + }, + { + "epoch": 0.8730447435431066, + "grad_norm": 0.5693327883618817, + "learning_rate": 3.9926781264028525e-05, + "loss": 0.3358, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1383555680513382, + "step": 1200, + "valid_targets_mean": 3627.1, + "valid_targets_min": 1970 + }, + { + "epoch": 0.8766824299745362, + "grad_norm": 0.5862749162816423, + "learning_rate": 3.9923647906285884e-05, + "loss": 0.3404, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14611968398094177, + "step": 1205, + "valid_targets_mean": 4038.0, + "valid_targets_min": 2492 + }, + { + "epoch": 0.8803201164059659, + "grad_norm": 0.640446717159042, + "learning_rate": 3.99204490288079e-05, + "loss": 0.3403, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14502911269664764, + "step": 1210, + "valid_targets_mean": 3211.2, + "valid_targets_min": 1066 + }, + { + "epoch": 0.8839578028373954, + "grad_norm": 0.6228864870312607, + "learning_rate": 3.991718464211421e-05, + "loss": 0.3431, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15157559514045715, + "step": 1215, + "valid_targets_mean": 3401.9, + "valid_targets_min": 2281 + }, + { + "epoch": 0.887595489268825, + "grad_norm": 0.6666026795803947, + "learning_rate": 3.9913854756939906e-05, + "loss": 0.3326, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.165190190076828, + "step": 1220, + "valid_targets_mean": 3662.2, + "valid_targets_min": 1969 + }, + { + "epoch": 0.8912331757002546, + "grad_norm": 0.6432432389038208, + "learning_rate": 3.991045938423543e-05, + "loss": 0.3365, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1464964747428894, + "step": 1225, + "valid_targets_mean": 3189.6, + "valid_targets_min": 1005 + }, + { + "epoch": 0.8948708621316842, + "grad_norm": 0.6482497816184194, + "learning_rate": 3.990699853516661e-05, + "loss": 0.3353, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.149469792842865, + "step": 1230, + "valid_targets_mean": 3546.5, + "valid_targets_min": 1980 + }, + { + "epoch": 0.8985085485631139, + "grad_norm": 0.5971579522758298, + "learning_rate": 3.9903472221114595e-05, + "loss": 0.3341, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19022831320762634, + "step": 1235, + "valid_targets_mean": 4803.1, + "valid_targets_min": 1257 + }, + { + "epoch": 0.9021462349945435, + "grad_norm": 0.6284811025410344, + "learning_rate": 3.9899880453675806e-05, + "loss": 0.3275, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17231744527816772, + "step": 1240, + "valid_targets_mean": 3118.1, + "valid_targets_min": 1357 + }, + { + "epoch": 0.9057839214259731, + "grad_norm": 0.6095777977349045, + "learning_rate": 3.9896223244661925e-05, + "loss": 0.3094, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13658513128757477, + "step": 1245, + "valid_targets_mean": 4463.8, + "valid_targets_min": 1410 + }, + { + "epoch": 0.9094216078574027, + "grad_norm": 0.6694691525383483, + "learning_rate": 3.989250060609983e-05, + "loss": 0.3337, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1531490981578827, + "step": 1250, + "valid_targets_mean": 3429.8, + "valid_targets_min": 1390 + }, + { + "epoch": 0.9130592942888323, + "grad_norm": 0.6385482663399477, + "learning_rate": 3.988871255023158e-05, + "loss": 0.315, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14360493421554565, + "step": 1255, + "valid_targets_mean": 3173.0, + "valid_targets_min": 790 + }, + { + "epoch": 0.916696980720262, + "grad_norm": 0.6184009449502986, + "learning_rate": 3.9884859089514336e-05, + "loss": 0.3392, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1692851483821869, + "step": 1260, + "valid_targets_mean": 3788.8, + "valid_targets_min": 1901 + }, + { + "epoch": 0.9203346671516915, + "grad_norm": 0.774749732585352, + "learning_rate": 3.988094023662038e-05, + "loss": 0.327, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15064449608325958, + "step": 1265, + "valid_targets_mean": 2697.1, + "valid_targets_min": 1460 + }, + { + "epoch": 0.9239723535831211, + "grad_norm": 0.5672626353349095, + "learning_rate": 3.987695600443701e-05, + "loss": 0.3244, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18411891162395477, + "step": 1270, + "valid_targets_mean": 5088.8, + "valid_targets_min": 952 + }, + { + "epoch": 0.9276100400145507, + "grad_norm": 0.7223638419395841, + "learning_rate": 3.987290640606653e-05, + "loss": 0.3284, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17547424137592316, + "step": 1275, + "valid_targets_mean": 3070.5, + "valid_targets_min": 1768 + }, + { + "epoch": 0.9312477264459803, + "grad_norm": 0.570083218096675, + "learning_rate": 3.986879145482623e-05, + "loss": 0.3256, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1287434846162796, + "step": 1280, + "valid_targets_mean": 3585.0, + "valid_targets_min": 1931 + }, + { + "epoch": 0.93488541287741, + "grad_norm": 0.48781835536342566, + "learning_rate": 3.986461116424829e-05, + "loss": 0.3164, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14737536013126373, + "step": 1285, + "valid_targets_mean": 5915.4, + "valid_targets_min": 1233 + }, + { + "epoch": 0.9385230993088396, + "grad_norm": 0.5421621631363923, + "learning_rate": 3.9860365548079756e-05, + "loss": 0.3135, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15485887229442596, + "step": 1290, + "valid_targets_mean": 5310.6, + "valid_targets_min": 1741 + }, + { + "epoch": 0.9421607857402692, + "grad_norm": 0.5528766464894317, + "learning_rate": 3.985605462028252e-05, + "loss": 0.3141, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14914925396442413, + "step": 1295, + "valid_targets_mean": 3884.8, + "valid_targets_min": 856 + }, + { + "epoch": 0.9457984721716988, + "grad_norm": 0.596523651779252, + "learning_rate": 3.985167839503324e-05, + "loss": 0.3182, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17303654551506042, + "step": 1300, + "valid_targets_mean": 3649.5, + "valid_targets_min": 1062 + }, + { + "epoch": 0.9494361586031284, + "grad_norm": 0.5226596389170851, + "learning_rate": 3.9847236886723324e-05, + "loss": 0.3106, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15475639700889587, + "step": 1305, + "valid_targets_mean": 4412.6, + "valid_targets_min": 974 + }, + { + "epoch": 0.953073845034558, + "grad_norm": 0.6432401354162569, + "learning_rate": 3.984273010995884e-05, + "loss": 0.3171, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1688249111175537, + "step": 1310, + "valid_targets_mean": 3077.5, + "valid_targets_min": 1447 + }, + { + "epoch": 0.9567115314659876, + "grad_norm": 0.5663195874886476, + "learning_rate": 3.983815807956054e-05, + "loss": 0.313, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1851254254579544, + "step": 1315, + "valid_targets_mean": 4663.0, + "valid_targets_min": 2144 + }, + { + "epoch": 0.9603492178974172, + "grad_norm": 0.574791636205698, + "learning_rate": 3.9833520810563685e-05, + "loss": 0.3131, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17638467252254486, + "step": 1320, + "valid_targets_mean": 4543.1, + "valid_targets_min": 1515 + }, + { + "epoch": 0.9639869043288468, + "grad_norm": 0.7370533352367081, + "learning_rate": 3.982881831821816e-05, + "loss": 0.3328, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1595374345779419, + "step": 1325, + "valid_targets_mean": 3034.9, + "valid_targets_min": 1349 + }, + { + "epoch": 0.9676245907602765, + "grad_norm": 0.6337620564211838, + "learning_rate": 3.982405061798829e-05, + "loss": 0.3345, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16443896293640137, + "step": 1330, + "valid_targets_mean": 3150.4, + "valid_targets_min": 1906 + }, + { + "epoch": 0.9712622771917061, + "grad_norm": 0.6194407128308724, + "learning_rate": 3.9819217725552864e-05, + "loss": 0.3206, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16991394758224487, + "step": 1335, + "valid_targets_mean": 3655.1, + "valid_targets_min": 2076 + }, + { + "epoch": 0.9748999636231357, + "grad_norm": 0.5872929037750567, + "learning_rate": 3.981431965680504e-05, + "loss": 0.3114, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13184162974357605, + "step": 1340, + "valid_targets_mean": 2994.0, + "valid_targets_min": 1021 + }, + { + "epoch": 0.9785376500545653, + "grad_norm": 0.6675980876687962, + "learning_rate": 3.9809356427852325e-05, + "loss": 0.3228, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16666871309280396, + "step": 1345, + "valid_targets_mean": 3226.9, + "valid_targets_min": 654 + }, + { + "epoch": 0.9821753364859949, + "grad_norm": 0.6112702224725588, + "learning_rate": 3.9804328055016495e-05, + "loss": 0.3065, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1803649365901947, + "step": 1350, + "valid_targets_mean": 4790.8, + "valid_targets_min": 3074 + }, + { + "epoch": 0.9858130229174246, + "grad_norm": 0.7511025741536864, + "learning_rate": 3.979923455483356e-05, + "loss": 0.3374, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1684199571609497, + "step": 1355, + "valid_targets_mean": 2604.0, + "valid_targets_min": 815 + }, + { + "epoch": 0.9894507093488542, + "grad_norm": 0.604866503549453, + "learning_rate": 3.9794075944053704e-05, + "loss": 0.3197, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14648959040641785, + "step": 1360, + "valid_targets_mean": 3560.0, + "valid_targets_min": 2205 + }, + { + "epoch": 0.9930883957802837, + "grad_norm": 0.5862921567480233, + "learning_rate": 3.9788852239641237e-05, + "loss": 0.3256, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17271004617214203, + "step": 1365, + "valid_targets_mean": 4216.0, + "valid_targets_min": 1793 + }, + { + "epoch": 0.9967260822117133, + "grad_norm": 0.656143296522615, + "learning_rate": 3.978356345877452e-05, + "loss": 0.326, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1504444181919098, + "step": 1370, + "valid_targets_mean": 3134.5, + "valid_targets_min": 1217 + }, + { + "epoch": 1.0, + "grad_norm": 1.0655296891300852, + "learning_rate": 3.977820961884593e-05, + "loss": 0.3023, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2132376730442047, + "step": 1375, + "valid_targets_mean": 1910.0, + "valid_targets_min": 393 + }, + { + "epoch": 1.0036376864314296, + "grad_norm": 0.5593247713246968, + "learning_rate": 3.977279073746179e-05, + "loss": 0.1843, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07413534075021744, + "step": 1380, + "valid_targets_mean": 2722.6, + "valid_targets_min": 458 + }, + { + "epoch": 1.0072753728628592, + "grad_norm": 0.5940305736285985, + "learning_rate": 3.9767306832442314e-05, + "loss": 0.1439, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07445372641086578, + "step": 1385, + "valid_targets_mean": 3113.1, + "valid_targets_min": 2461 + }, + { + "epoch": 1.0109130592942888, + "grad_norm": 0.415561728505045, + "learning_rate": 3.9761757921821544e-05, + "loss": 0.1291, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07941805571317673, + "step": 1390, + "valid_targets_mean": 3648.1, + "valid_targets_min": 2068 + }, + { + "epoch": 1.0145507457257184, + "grad_norm": 0.4778099380213266, + "learning_rate": 3.975614402384731e-05, + "loss": 0.1555, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07335877418518066, + "step": 1395, + "valid_targets_mean": 3423.6, + "valid_targets_min": 2425 + }, + { + "epoch": 1.0181884321571482, + "grad_norm": 0.5000812567460028, + "learning_rate": 3.975046515698114e-05, + "loss": 0.1263, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.061342209577560425, + "step": 1400, + "valid_targets_mean": 2724.1, + "valid_targets_min": 1198 + }, + { + "epoch": 1.0218261185885777, + "grad_norm": 0.8131398497931814, + "learning_rate": 3.974472133989822e-05, + "loss": 0.1496, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09955982118844986, + "step": 1405, + "valid_targets_mean": 1232.5, + "valid_targets_min": 807 + }, + { + "epoch": 1.0254638050200073, + "grad_norm": 0.4744328729314133, + "learning_rate": 3.973891259148733e-05, + "loss": 0.2002, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08083701133728027, + "step": 1410, + "valid_targets_mean": 3487.4, + "valid_targets_min": 2351 + }, + { + "epoch": 1.029101491451437, + "grad_norm": 0.5379691893298337, + "learning_rate": 3.973303893085078e-05, + "loss": 0.138, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10664110630750656, + "step": 1415, + "valid_targets_mean": 3126.4, + "valid_targets_min": 781 + }, + { + "epoch": 1.0327391778828665, + "grad_norm": 0.41708102646886025, + "learning_rate": 3.972710037730434e-05, + "loss": 0.1534, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07369478791952133, + "step": 1420, + "valid_targets_mean": 3474.8, + "valid_targets_min": 799 + }, + { + "epoch": 1.036376864314296, + "grad_norm": 0.3484076314686727, + "learning_rate": 3.97210969503772e-05, + "loss": 0.1555, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06519393622875214, + "step": 1425, + "valid_targets_mean": 5026.1, + "valid_targets_min": 2802 + }, + { + "epoch": 1.0400145507457257, + "grad_norm": 0.4355395177969458, + "learning_rate": 3.971502866981186e-05, + "loss": 0.1222, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04783561825752258, + "step": 1430, + "valid_targets_mean": 2611.0, + "valid_targets_min": 574 + }, + { + "epoch": 1.0436522371771553, + "grad_norm": 0.423119289673439, + "learning_rate": 3.9708895555564114e-05, + "loss": 0.1323, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06382015347480774, + "step": 1435, + "valid_targets_mean": 4036.5, + "valid_targets_min": 2642 + }, + { + "epoch": 1.0472899236085849, + "grad_norm": 0.6567700835927867, + "learning_rate": 3.970269762780297e-05, + "loss": 0.1382, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09206438064575195, + "step": 1440, + "valid_targets_mean": 1960.5, + "valid_targets_min": 593 + }, + { + "epoch": 1.0509276100400144, + "grad_norm": 0.490408920973317, + "learning_rate": 3.969643490691057e-05, + "loss": 0.1658, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.055813368409872055, + "step": 1445, + "valid_targets_mean": 3354.2, + "valid_targets_min": 1353 + }, + { + "epoch": 1.0545652964714443, + "grad_norm": 0.6049513405025357, + "learning_rate": 3.969010741348211e-05, + "loss": 0.128, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07597864419221878, + "step": 1450, + "valid_targets_mean": 2300.9, + "valid_targets_min": 818 + }, + { + "epoch": 1.0582029829028738, + "grad_norm": 0.40521282816418985, + "learning_rate": 3.968371516832582e-05, + "loss": 0.1282, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04739393666386604, + "step": 1455, + "valid_targets_mean": 2295.8, + "valid_targets_min": 934 + }, + { + "epoch": 1.0618406693343034, + "grad_norm": 0.3830802293281381, + "learning_rate": 3.9677258192462866e-05, + "loss": 0.1165, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06394504010677338, + "step": 1460, + "valid_targets_mean": 3524.4, + "valid_targets_min": 2798 + }, + { + "epoch": 1.065478355765733, + "grad_norm": 0.5076301389516883, + "learning_rate": 3.9670736507127284e-05, + "loss": 0.1447, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.059483226388692856, + "step": 1465, + "valid_targets_mean": 2810.1, + "valid_targets_min": 471 + }, + { + "epoch": 1.0691160421971626, + "grad_norm": 0.4876962322944028, + "learning_rate": 3.96641501337659e-05, + "loss": 0.1457, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09097189456224442, + "step": 1470, + "valid_targets_mean": 3042.8, + "valid_targets_min": 903 + }, + { + "epoch": 1.0727537286285922, + "grad_norm": 0.44000542208841753, + "learning_rate": 3.9657499094038264e-05, + "loss": 0.1454, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05602867156267166, + "step": 1475, + "valid_targets_mean": 3081.6, + "valid_targets_min": 474 + }, + { + "epoch": 1.0763914150600218, + "grad_norm": 0.48762092907665755, + "learning_rate": 3.96507834098166e-05, + "loss": 0.1454, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07145896553993225, + "step": 1480, + "valid_targets_mean": 3468.0, + "valid_targets_min": 640 + }, + { + "epoch": 1.0800291014914514, + "grad_norm": 0.3788325715710137, + "learning_rate": 3.964400310318571e-05, + "loss": 0.1847, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047245852649211884, + "step": 1485, + "valid_targets_mean": 3438.4, + "valid_targets_min": 902 + }, + { + "epoch": 1.083666787922881, + "grad_norm": 0.4613628273270764, + "learning_rate": 3.9637158196442925e-05, + "loss": 0.1233, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06098242849111557, + "step": 1490, + "valid_targets_mean": 3212.8, + "valid_targets_min": 1837 + }, + { + "epoch": 1.0873044743543105, + "grad_norm": 0.8871883753569471, + "learning_rate": 3.963024871209798e-05, + "loss": 0.1683, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1379985511302948, + "step": 1495, + "valid_targets_mean": 1358.8, + "valid_targets_min": 756 + }, + { + "epoch": 1.0909421607857404, + "grad_norm": 0.5905671199614322, + "learning_rate": 3.962327467287302e-05, + "loss": 0.1662, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11513521522283554, + "step": 1500, + "valid_targets_mean": 2896.2, + "valid_targets_min": 1072 + }, + { + "epoch": 1.09457984721717, + "grad_norm": 0.4221539879664716, + "learning_rate": 3.961623610170244e-05, + "loss": 0.1305, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04142339900135994, + "step": 1505, + "valid_targets_mean": 2908.1, + "valid_targets_min": 895 + }, + { + "epoch": 1.0982175336485995, + "grad_norm": 0.660456152294959, + "learning_rate": 3.9609133021732884e-05, + "loss": 0.2711, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22519733011722565, + "step": 1510, + "valid_targets_mean": 2562.2, + "valid_targets_min": 878 + }, + { + "epoch": 1.1018552200800291, + "grad_norm": 0.49650624305471747, + "learning_rate": 3.960196545632311e-05, + "loss": 0.2456, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06043098494410515, + "step": 1515, + "valid_targets_mean": 2087.6, + "valid_targets_min": 756 + }, + { + "epoch": 1.1054929065114587, + "grad_norm": 0.40687056975770974, + "learning_rate": 3.9594733429043966e-05, + "loss": 0.1449, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05202292650938034, + "step": 1520, + "valid_targets_mean": 3421.8, + "valid_targets_min": 2963 + }, + { + "epoch": 1.1091305929428883, + "grad_norm": 0.8421152814316041, + "learning_rate": 3.9587436963678266e-05, + "loss": 0.2234, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2647666335105896, + "step": 1525, + "valid_targets_mean": 2375.2, + "valid_targets_min": 1153 + }, + { + "epoch": 1.1127682793743179, + "grad_norm": 0.297723898249341, + "learning_rate": 3.9580076084220735e-05, + "loss": 0.1525, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05135089159011841, + "step": 1530, + "valid_targets_mean": 4886.2, + "valid_targets_min": 1700 + }, + { + "epoch": 1.1164059658057475, + "grad_norm": 0.3337360603423595, + "learning_rate": 3.957265081487792e-05, + "loss": 0.1247, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04302198812365532, + "step": 1535, + "valid_targets_mean": 3965.1, + "valid_targets_min": 789 + }, + { + "epoch": 1.120043652237177, + "grad_norm": 0.39018646821428, + "learning_rate": 3.9565161180068144e-05, + "loss": 0.1388, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06908415257930756, + "step": 1540, + "valid_targets_mean": 3771.5, + "valid_targets_min": 995 + }, + { + "epoch": 1.1236813386686069, + "grad_norm": 0.44167112147559084, + "learning_rate": 3.955760720442136e-05, + "loss": 0.1244, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04099872708320618, + "step": 1545, + "valid_targets_mean": 2220.8, + "valid_targets_min": 488 + }, + { + "epoch": 1.1273190251000365, + "grad_norm": 0.43853376189675153, + "learning_rate": 3.9549988912779136e-05, + "loss": 0.1287, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0733383297920227, + "step": 1550, + "valid_targets_mean": 3635.4, + "valid_targets_min": 2437 + }, + { + "epoch": 1.130956711531466, + "grad_norm": 0.5295476431527381, + "learning_rate": 3.954230633019454e-05, + "loss": 0.1222, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05046086013317108, + "step": 1555, + "valid_targets_mean": 1752.1, + "valid_targets_min": 712 + }, + { + "epoch": 1.1345943979628956, + "grad_norm": 0.47259865845187454, + "learning_rate": 3.9534559481932054e-05, + "loss": 0.1215, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06148086488246918, + "step": 1560, + "valid_targets_mean": 3713.4, + "valid_targets_min": 2422 + }, + { + "epoch": 1.1382320843943252, + "grad_norm": 0.3928650631908166, + "learning_rate": 3.952674839346751e-05, + "loss": 0.1094, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05788934975862503, + "step": 1565, + "valid_targets_mean": 3998.5, + "valid_targets_min": 3643 + }, + { + "epoch": 1.1418697708257548, + "grad_norm": 0.3621522517175758, + "learning_rate": 3.951887309048799e-05, + "loss": 0.1847, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04319798946380615, + "step": 1570, + "valid_targets_mean": 3319.0, + "valid_targets_min": 772 + }, + { + "epoch": 1.1455074572571844, + "grad_norm": 0.40399678999022703, + "learning_rate": 3.951093359889177e-05, + "loss": 0.1038, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04470772668719292, + "step": 1575, + "valid_targets_mean": 2934.8, + "valid_targets_min": 722 + }, + { + "epoch": 1.149145143688614, + "grad_norm": 0.382103863197947, + "learning_rate": 3.950292994478816e-05, + "loss": 0.1146, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.042668674141168594, + "step": 1580, + "valid_targets_mean": 3656.4, + "valid_targets_min": 2788 + }, + { + "epoch": 1.1527828301200436, + "grad_norm": 0.3857735027250339, + "learning_rate": 3.949486215449754e-05, + "loss": 0.1145, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05995999276638031, + "step": 1585, + "valid_targets_mean": 3202.5, + "valid_targets_min": 948 + }, + { + "epoch": 1.1564205165514734, + "grad_norm": 0.5992881148233499, + "learning_rate": 3.948673025455115e-05, + "loss": 0.1397, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06784547865390778, + "step": 1590, + "valid_targets_mean": 1002.0, + "valid_targets_min": 490 + }, + { + "epoch": 1.160058202982903, + "grad_norm": 0.348243613533693, + "learning_rate": 3.947853427169108e-05, + "loss": 0.1552, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06678116321563721, + "step": 1595, + "valid_targets_mean": 3751.9, + "valid_targets_min": 2800 + }, + { + "epoch": 1.1636958894143326, + "grad_norm": 0.5221959135535658, + "learning_rate": 3.947027423287017e-05, + "loss": 0.1658, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11157066375017166, + "step": 1600, + "valid_targets_mean": 2641.8, + "valid_targets_min": 896 + }, + { + "epoch": 1.1673335758457621, + "grad_norm": 0.5066636839657876, + "learning_rate": 3.946195016525187e-05, + "loss": 0.1399, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08107417821884155, + "step": 1605, + "valid_targets_mean": 3199.0, + "valid_targets_min": 1333 + }, + { + "epoch": 1.1709712622771917, + "grad_norm": 0.49060182909541805, + "learning_rate": 3.945356209621025e-05, + "loss": 0.1574, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0716705247759819, + "step": 1610, + "valid_targets_mean": 2548.1, + "valid_targets_min": 693 + }, + { + "epoch": 1.1746089487086213, + "grad_norm": 0.5592325072387145, + "learning_rate": 3.94451100533298e-05, + "loss": 0.1695, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05087967962026596, + "step": 1615, + "valid_targets_mean": 2482.9, + "valid_targets_min": 803 + }, + { + "epoch": 1.178246635140051, + "grad_norm": 0.3976158090775834, + "learning_rate": 3.9436594064405415e-05, + "loss": 0.1301, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06798574328422546, + "step": 1620, + "valid_targets_mean": 3239.2, + "valid_targets_min": 2525 + }, + { + "epoch": 1.1818843215714805, + "grad_norm": 0.5064279924801413, + "learning_rate": 3.942801415744228e-05, + "loss": 0.1589, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.058664318174123764, + "step": 1625, + "valid_targets_mean": 3806.1, + "valid_targets_min": 2393 + }, + { + "epoch": 1.18552200800291, + "grad_norm": 0.42216596792640404, + "learning_rate": 3.941937036065576e-05, + "loss": 0.142, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.060738176107406616, + "step": 1630, + "valid_targets_mean": 2990.9, + "valid_targets_min": 804 + }, + { + "epoch": 1.1891596944343397, + "grad_norm": 0.3500232514724261, + "learning_rate": 3.941066270247135e-05, + "loss": 0.1238, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0598750114440918, + "step": 1635, + "valid_targets_mean": 4885.8, + "valid_targets_min": 3695 + }, + { + "epoch": 1.1927973808657693, + "grad_norm": 0.4449168109852547, + "learning_rate": 3.940189121152453e-05, + "loss": 0.1146, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07115556299686432, + "step": 1640, + "valid_targets_mean": 3128.8, + "valid_targets_min": 596 + }, + { + "epoch": 1.196435067297199, + "grad_norm": 0.2838419572742785, + "learning_rate": 3.939305591666072e-05, + "loss": 0.1105, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0412953719496727, + "step": 1645, + "valid_targets_mean": 4567.8, + "valid_targets_min": 3804 + }, + { + "epoch": 1.2000727537286286, + "grad_norm": 0.35158412392355237, + "learning_rate": 3.938415684693514e-05, + "loss": 0.1183, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04582654684782028, + "step": 1650, + "valid_targets_mean": 3853.1, + "valid_targets_min": 2862 + }, + { + "epoch": 1.2037104401600582, + "grad_norm": 0.616786712642024, + "learning_rate": 3.937519403161275e-05, + "loss": 0.1468, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08316540718078613, + "step": 1655, + "valid_targets_mean": 1072.6, + "valid_targets_min": 497 + }, + { + "epoch": 1.2073481265914878, + "grad_norm": 0.37852412618129555, + "learning_rate": 3.936616750016814e-05, + "loss": 0.1237, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06877662986516953, + "step": 1660, + "valid_targets_mean": 4191.0, + "valid_targets_min": 3918 + }, + { + "epoch": 1.2109858130229174, + "grad_norm": 0.5424583041507227, + "learning_rate": 3.935707728228543e-05, + "loss": 0.1383, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0625147670507431, + "step": 1665, + "valid_targets_mean": 2121.9, + "valid_targets_min": 785 + }, + { + "epoch": 1.214623499454347, + "grad_norm": 0.3821254324571586, + "learning_rate": 3.9347923407858175e-05, + "loss": 0.102, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05249269679188728, + "step": 1670, + "valid_targets_mean": 3451.6, + "valid_targets_min": 905 + }, + { + "epoch": 1.2182611858857766, + "grad_norm": 0.4229032745545744, + "learning_rate": 3.933870590698926e-05, + "loss": 0.1335, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09667232632637024, + "step": 1675, + "valid_targets_mean": 2929.2, + "valid_targets_min": 657 + }, + { + "epoch": 1.2218988723172062, + "grad_norm": 0.3687504980982325, + "learning_rate": 3.932942480999083e-05, + "loss": 0.1256, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06398534774780273, + "step": 1680, + "valid_targets_mean": 3798.9, + "valid_targets_min": 3184 + }, + { + "epoch": 1.2255365587486358, + "grad_norm": 0.470829754135503, + "learning_rate": 3.932008014738414e-05, + "loss": 0.1105, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06568412482738495, + "step": 1685, + "valid_targets_mean": 1463.9, + "valid_targets_min": 627 + }, + { + "epoch": 1.2291742451800656, + "grad_norm": 0.4848119818214479, + "learning_rate": 3.93106719498995e-05, + "loss": 0.1279, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06020846962928772, + "step": 1690, + "valid_targets_mean": 1701.5, + "valid_targets_min": 827 + }, + { + "epoch": 1.2328119316114952, + "grad_norm": 0.4901774946359283, + "learning_rate": 3.930120024847616e-05, + "loss": 0.131, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05775256082415581, + "step": 1695, + "valid_targets_mean": 1789.9, + "valid_targets_min": 940 + }, + { + "epoch": 1.2364496180429247, + "grad_norm": 0.4064460160666167, + "learning_rate": 3.929166507426219e-05, + "loss": 0.1484, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.051577627658843994, + "step": 1700, + "valid_targets_mean": 2943.8, + "valid_targets_min": 958 + }, + { + "epoch": 1.2400873044743543, + "grad_norm": 0.44263662204705556, + "learning_rate": 3.928206645861442e-05, + "loss": 0.1247, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04974406585097313, + "step": 1705, + "valid_targets_mean": 1859.2, + "valid_targets_min": 719 + }, + { + "epoch": 1.243724990905784, + "grad_norm": 0.5130266021054247, + "learning_rate": 3.927240443309827e-05, + "loss": 0.152, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04486433416604996, + "step": 1710, + "valid_targets_mean": 1703.9, + "valid_targets_min": 759 + }, + { + "epoch": 1.2473626773372135, + "grad_norm": 0.6835784441996091, + "learning_rate": 3.926267902948772e-05, + "loss": 0.1495, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1515578031539917, + "step": 1715, + "valid_targets_mean": 1556.6, + "valid_targets_min": 521 + }, + { + "epoch": 1.251000363768643, + "grad_norm": 0.4337592663331354, + "learning_rate": 3.925289027976517e-05, + "loss": 0.122, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07548113167285919, + "step": 1720, + "valid_targets_mean": 3267.6, + "valid_targets_min": 1032 + }, + { + "epoch": 1.2546380502000727, + "grad_norm": 0.33386663527128274, + "learning_rate": 3.9243038216121304e-05, + "loss": 0.1052, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04381087049841881, + "step": 1725, + "valid_targets_mean": 3609.0, + "valid_targets_min": 2741 + }, + { + "epoch": 1.2582757366315023, + "grad_norm": 0.36429203050037634, + "learning_rate": 3.923312287095505e-05, + "loss": 0.1208, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05206148326396942, + "step": 1730, + "valid_targets_mean": 2735.8, + "valid_targets_min": 1692 + }, + { + "epoch": 1.261913423062932, + "grad_norm": 0.39961502619895306, + "learning_rate": 3.922314427687343e-05, + "loss": 0.1584, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06928849220275879, + "step": 1735, + "valid_targets_mean": 3693.4, + "valid_targets_min": 2215 + }, + { + "epoch": 1.2655511094943614, + "grad_norm": 0.37310032152066136, + "learning_rate": 3.921310246669146e-05, + "loss": 0.1037, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04668722301721573, + "step": 1740, + "valid_targets_mean": 3109.9, + "valid_targets_min": 723 + }, + { + "epoch": 1.2691887959257913, + "grad_norm": 0.33785327253944564, + "learning_rate": 3.920299747343204e-05, + "loss": 0.1205, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.050171785056591034, + "step": 1745, + "valid_targets_mean": 4984.2, + "valid_targets_min": 3282 + }, + { + "epoch": 1.2728264823572208, + "grad_norm": 0.3410801462129656, + "learning_rate": 3.9192829330325856e-05, + "loss": 0.1158, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04932911694049835, + "step": 1750, + "valid_targets_mean": 3863.4, + "valid_targets_min": 995 + }, + { + "epoch": 1.2764641687886504, + "grad_norm": 0.2976848235359673, + "learning_rate": 3.918259807081128e-05, + "loss": 0.0988, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047590941190719604, + "step": 1755, + "valid_targets_mean": 4201.0, + "valid_targets_min": 949 + }, + { + "epoch": 1.28010185522008, + "grad_norm": 0.41671320111229176, + "learning_rate": 3.917230372853421e-05, + "loss": 0.1275, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07665795087814331, + "step": 1760, + "valid_targets_mean": 4321.6, + "valid_targets_min": 2724 + }, + { + "epoch": 1.2837395416515096, + "grad_norm": 0.36332573044733035, + "learning_rate": 3.916194633734803e-05, + "loss": 0.1073, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03910285234451294, + "step": 1765, + "valid_targets_mean": 3384.5, + "valid_targets_min": 424 + }, + { + "epoch": 1.2873772280829392, + "grad_norm": 0.40517497696076143, + "learning_rate": 3.9151525931313425e-05, + "loss": 0.1474, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04973211511969566, + "step": 1770, + "valid_targets_mean": 2987.2, + "valid_targets_min": 531 + }, + { + "epoch": 1.2910149145143688, + "grad_norm": 0.35699038759679463, + "learning_rate": 3.914104254469835e-05, + "loss": 0.1397, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047396156936883926, + "step": 1775, + "valid_targets_mean": 3469.5, + "valid_targets_min": 1195 + }, + { + "epoch": 1.2946526009457986, + "grad_norm": 0.3761496609987071, + "learning_rate": 3.9130496211977845e-05, + "loss": 0.105, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08888816833496094, + "step": 1780, + "valid_targets_mean": 2682.8, + "valid_targets_min": 722 + }, + { + "epoch": 1.298290287377228, + "grad_norm": 0.3335474071093571, + "learning_rate": 3.911988696783396e-05, + "loss": 0.1426, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05256030708551407, + "step": 1785, + "valid_targets_mean": 3496.9, + "valid_targets_min": 1251 + }, + { + "epoch": 1.3019279738086578, + "grad_norm": 0.39565935327840374, + "learning_rate": 3.910921484715563e-05, + "loss": 0.1154, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06513626873493195, + "step": 1790, + "valid_targets_mean": 2912.5, + "valid_targets_min": 970 + }, + { + "epoch": 1.3055656602400874, + "grad_norm": 0.630978373794118, + "learning_rate": 3.909847988503856e-05, + "loss": 0.1356, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06392647325992584, + "step": 1795, + "valid_targets_mean": 1423.6, + "valid_targets_min": 634 + }, + { + "epoch": 1.309203346671517, + "grad_norm": 0.5607460216522235, + "learning_rate": 3.9087682116785114e-05, + "loss": 0.1481, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09758636355400085, + "step": 1800, + "valid_targets_mean": 1423.9, + "valid_targets_min": 685 + }, + { + "epoch": 1.3128410331029465, + "grad_norm": 0.3962199571413524, + "learning_rate": 3.907682157790421e-05, + "loss": 0.1619, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047784969210624695, + "step": 1805, + "valid_targets_mean": 2932.5, + "valid_targets_min": 593 + }, + { + "epoch": 1.3164787195343761, + "grad_norm": 0.3273527045816846, + "learning_rate": 3.906589830411116e-05, + "loss": 0.1249, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04888863116502762, + "step": 1810, + "valid_targets_mean": 3945.8, + "valid_targets_min": 3129 + }, + { + "epoch": 1.3201164059658057, + "grad_norm": 0.4053243654066863, + "learning_rate": 3.905491233132761e-05, + "loss": 0.13, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07129421830177307, + "step": 1815, + "valid_targets_mean": 3621.1, + "valid_targets_min": 859 + }, + { + "epoch": 1.3237540923972353, + "grad_norm": 0.4937708277623076, + "learning_rate": 3.904386369568137e-05, + "loss": 0.1173, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06353890895843506, + "step": 1820, + "valid_targets_mean": 2553.0, + "valid_targets_min": 616 + }, + { + "epoch": 1.327391778828665, + "grad_norm": 0.4052223477946479, + "learning_rate": 3.903275243350634e-05, + "loss": 0.1165, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05259866639971733, + "step": 1825, + "valid_targets_mean": 2690.5, + "valid_targets_min": 541 + }, + { + "epoch": 1.3310294652600945, + "grad_norm": 0.7019487094309289, + "learning_rate": 3.9021578581342373e-05, + "loss": 0.1137, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07280460745096207, + "step": 1830, + "valid_targets_mean": 906.4, + "valid_targets_min": 598 + }, + { + "epoch": 1.3346671516915243, + "grad_norm": 0.41790422637417646, + "learning_rate": 3.901034217593512e-05, + "loss": 0.1371, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05878820642828941, + "step": 1835, + "valid_targets_mean": 3147.2, + "valid_targets_min": 1101 + }, + { + "epoch": 1.3383048381229539, + "grad_norm": 0.42950310159547084, + "learning_rate": 3.899904325423599e-05, + "loss": 0.107, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04789833724498749, + "step": 1840, + "valid_targets_mean": 2331.8, + "valid_targets_min": 680 + }, + { + "epoch": 1.3419425245543835, + "grad_norm": 0.36222847550945714, + "learning_rate": 3.8987681853401924e-05, + "loss": 0.114, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06246612221002579, + "step": 1845, + "valid_targets_mean": 2577.0, + "valid_targets_min": 774 + }, + { + "epoch": 1.345580210985813, + "grad_norm": 0.36603208730651215, + "learning_rate": 3.897625801079537e-05, + "loss": 0.1314, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06727999448776245, + "step": 1850, + "valid_targets_mean": 4140.1, + "valid_targets_min": 1474 + }, + { + "epoch": 1.3492178974172426, + "grad_norm": 0.2476294503009053, + "learning_rate": 3.89647717639841e-05, + "loss": 0.1018, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03605909273028374, + "step": 1855, + "valid_targets_mean": 5498.9, + "valid_targets_min": 3677 + }, + { + "epoch": 1.3528555838486722, + "grad_norm": 0.4065596446707776, + "learning_rate": 3.8953223150741115e-05, + "loss": 0.1134, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07172653079032898, + "step": 1860, + "valid_targets_mean": 3352.5, + "valid_targets_min": 675 + }, + { + "epoch": 1.3564932702801018, + "grad_norm": 0.3380381733697883, + "learning_rate": 3.894161220904449e-05, + "loss": 0.1086, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04277428239583969, + "step": 1865, + "valid_targets_mean": 3902.5, + "valid_targets_min": 795 + }, + { + "epoch": 1.3601309567115314, + "grad_norm": 0.5257827772615352, + "learning_rate": 3.8929938977077304e-05, + "loss": 0.2286, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09965059906244278, + "step": 1870, + "valid_targets_mean": 2495.0, + "valid_targets_min": 605 + }, + { + "epoch": 1.363768643142961, + "grad_norm": 0.35313115845866094, + "learning_rate": 3.8918203493227445e-05, + "loss": 0.107, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05652537941932678, + "step": 1875, + "valid_targets_mean": 3298.6, + "valid_targets_min": 985 + }, + { + "epoch": 1.3674063295743908, + "grad_norm": 0.4147914431166218, + "learning_rate": 3.8906405796087545e-05, + "loss": 0.1237, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06428414583206177, + "step": 1880, + "valid_targets_mean": 3505.2, + "valid_targets_min": 878 + }, + { + "epoch": 1.3710440160058204, + "grad_norm": 0.6037824511800534, + "learning_rate": 3.889454592445481e-05, + "loss": 0.1462, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14526638388633728, + "step": 1885, + "valid_targets_mean": 2289.5, + "valid_targets_min": 1214 + }, + { + "epoch": 1.37468170243725, + "grad_norm": 0.41928804792614427, + "learning_rate": 3.888262391733091e-05, + "loss": 0.2333, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1583964228630066, + "step": 1890, + "valid_targets_mean": 8496.5, + "valid_targets_min": 6565 + }, + { + "epoch": 1.3783193888686796, + "grad_norm": 0.4193595505467731, + "learning_rate": 3.887063981392187e-05, + "loss": 0.2484, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11108408868312836, + "step": 1895, + "valid_targets_mean": 5663.5, + "valid_targets_min": 3833 + }, + { + "epoch": 1.3819570753001091, + "grad_norm": 0.42172758381716413, + "learning_rate": 3.885859365363789e-05, + "loss": 0.2443, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11481094360351562, + "step": 1900, + "valid_targets_mean": 6654.1, + "valid_targets_min": 5324 + }, + { + "epoch": 1.3855947617315387, + "grad_norm": 0.3917059828127558, + "learning_rate": 3.884648547609328e-05, + "loss": 0.2242, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11099789291620255, + "step": 1905, + "valid_targets_mean": 7635.1, + "valid_targets_min": 5672 + }, + { + "epoch": 1.3892324481629683, + "grad_norm": 0.37278040660522255, + "learning_rate": 3.883431532110627e-05, + "loss": 0.2308, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10405522584915161, + "step": 1910, + "valid_targets_mean": 8517.0, + "valid_targets_min": 5045 + }, + { + "epoch": 1.392870134594398, + "grad_norm": 0.4313172087446921, + "learning_rate": 3.882208322869891e-05, + "loss": 0.2334, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11371403932571411, + "step": 1915, + "valid_targets_mean": 6759.1, + "valid_targets_min": 3152 + }, + { + "epoch": 1.3965078210258275, + "grad_norm": 0.3801025113524813, + "learning_rate": 3.8809789239096956e-05, + "loss": 0.243, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10722364485263824, + "step": 1920, + "valid_targets_mean": 7822.8, + "valid_targets_min": 4820 + }, + { + "epoch": 1.4001455074572573, + "grad_norm": 0.37175366193597575, + "learning_rate": 3.879743339272968e-05, + "loss": 0.2298, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09437688440084457, + "step": 1925, + "valid_targets_mean": 6852.2, + "valid_targets_min": 4135 + }, + { + "epoch": 1.4037831938886867, + "grad_norm": 0.4606726479552531, + "learning_rate": 3.87850157302298e-05, + "loss": 0.2277, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10303887724876404, + "step": 1930, + "valid_targets_mean": 4794.2, + "valid_targets_min": 3230 + }, + { + "epoch": 1.4074208803201165, + "grad_norm": 0.3851704681376582, + "learning_rate": 3.8772536292433314e-05, + "loss": 0.2143, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11796558648347855, + "step": 1935, + "valid_targets_mean": 7691.0, + "valid_targets_min": 5900 + }, + { + "epoch": 1.411058566751546, + "grad_norm": 0.43556090205597964, + "learning_rate": 3.8759995120379355e-05, + "loss": 0.2284, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10779024660587311, + "step": 1940, + "valid_targets_mean": 6510.9, + "valid_targets_min": 5639 + }, + { + "epoch": 1.4146962531829756, + "grad_norm": 0.43088359079148214, + "learning_rate": 3.874739225531009e-05, + "loss": 0.2155, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11013513803482056, + "step": 1945, + "valid_targets_mean": 6096.8, + "valid_targets_min": 4669 + }, + { + "epoch": 1.4183339396144052, + "grad_norm": 0.3882820932878808, + "learning_rate": 3.873472773867056e-05, + "loss": 0.2203, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11501550674438477, + "step": 1950, + "valid_targets_mean": 6812.2, + "valid_targets_min": 5160 + }, + { + "epoch": 1.4219716260458348, + "grad_norm": 0.4585136223900485, + "learning_rate": 3.8722001612108545e-05, + "loss": 0.2129, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09822295606136322, + "step": 1955, + "valid_targets_mean": 6980.4, + "valid_targets_min": 5151 + }, + { + "epoch": 1.4256093124772644, + "grad_norm": 0.4803546965450771, + "learning_rate": 3.870921391747443e-05, + "loss": 0.2169, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1156778633594513, + "step": 1960, + "valid_targets_mean": 7417.1, + "valid_targets_min": 5932 + }, + { + "epoch": 1.429246998908694, + "grad_norm": 0.42969343124021026, + "learning_rate": 3.869636469682109e-05, + "loss": 0.2215, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10672380775213242, + "step": 1965, + "valid_targets_mean": 6258.5, + "valid_targets_min": 5569 + }, + { + "epoch": 1.4328846853401238, + "grad_norm": 0.44099713115240163, + "learning_rate": 3.8683453992403704e-05, + "loss": 0.2138, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12096798419952393, + "step": 1970, + "valid_targets_mean": 7689.1, + "valid_targets_min": 6223 + }, + { + "epoch": 1.4365223717715532, + "grad_norm": 0.5541044047259974, + "learning_rate": 3.8670481846679656e-05, + "loss": 0.2242, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13378627598285675, + "step": 1975, + "valid_targets_mean": 4371.9, + "valid_targets_min": 1172 + }, + { + "epoch": 1.440160058202983, + "grad_norm": 0.4150780656222367, + "learning_rate": 3.865744830230838e-05, + "loss": 0.2358, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12411895394325256, + "step": 1980, + "valid_targets_mean": 6936.2, + "valid_targets_min": 5619 + }, + { + "epoch": 1.4437977446344126, + "grad_norm": 0.6807365899255924, + "learning_rate": 3.864435340215124e-05, + "loss": 0.2194, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16629862785339355, + "step": 1985, + "valid_targets_mean": 3273.2, + "valid_targets_min": 442 + }, + { + "epoch": 1.4474354310658422, + "grad_norm": 0.3803897060258293, + "learning_rate": 3.863119718927132e-05, + "loss": 0.2034, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10577964782714844, + "step": 1990, + "valid_targets_mean": 8141.1, + "valid_targets_min": 5061 + }, + { + "epoch": 1.4510731174972717, + "grad_norm": 0.4255077471298386, + "learning_rate": 3.8617979706933384e-05, + "loss": 0.2153, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10841190814971924, + "step": 1995, + "valid_targets_mean": 7371.1, + "valid_targets_min": 5112 + }, + { + "epoch": 1.4547108039287013, + "grad_norm": 0.4099829710278214, + "learning_rate": 3.860470099860368e-05, + "loss": 0.2151, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0953240841627121, + "step": 2000, + "valid_targets_mean": 5659.4, + "valid_targets_min": 4626 + }, + { + "epoch": 1.458348490360131, + "grad_norm": 0.3630859887444452, + "learning_rate": 3.8591361107949766e-05, + "loss": 0.2223, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09767553210258484, + "step": 2005, + "valid_targets_mean": 7801.8, + "valid_targets_min": 6172 + }, + { + "epoch": 1.4619861767915605, + "grad_norm": 0.3946760858160747, + "learning_rate": 3.8577960078840424e-05, + "loss": 0.2139, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10718953609466553, + "step": 2010, + "valid_targets_mean": 7547.4, + "valid_targets_min": 5804 + }, + { + "epoch": 1.46562386322299, + "grad_norm": 0.3843749466132778, + "learning_rate": 3.85644979553455e-05, + "loss": 0.2027, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09711812436580658, + "step": 2015, + "valid_targets_mean": 6474.2, + "valid_targets_min": 5050 + }, + { + "epoch": 1.4692615496544197, + "grad_norm": 0.3757265226392254, + "learning_rate": 3.855097478173574e-05, + "loss": 0.1879, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09788639843463898, + "step": 2020, + "valid_targets_mean": 7225.4, + "valid_targets_min": 5298 + }, + { + "epoch": 1.4728992360858495, + "grad_norm": 0.38152926551545313, + "learning_rate": 3.8537390602482644e-05, + "loss": 0.1886, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08840848505496979, + "step": 2025, + "valid_targets_mean": 6442.9, + "valid_targets_min": 4679 + }, + { + "epoch": 1.476536922517279, + "grad_norm": 0.5400123578605328, + "learning_rate": 3.8523745462258346e-05, + "loss": 0.2245, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09060432016849518, + "step": 2030, + "valid_targets_mean": 3651.9, + "valid_targets_min": 1801 + }, + { + "epoch": 1.4801746089487087, + "grad_norm": 0.42006373311235884, + "learning_rate": 3.851003940593546e-05, + "loss": 0.2263, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12435144186019897, + "step": 2035, + "valid_targets_mean": 7512.8, + "valid_targets_min": 5331 + }, + { + "epoch": 1.4838122953801383, + "grad_norm": 0.4005432643435411, + "learning_rate": 3.849627247858691e-05, + "loss": 0.2231, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11559196561574936, + "step": 2040, + "valid_targets_mean": 7108.8, + "valid_targets_min": 4822 + }, + { + "epoch": 1.4874499818115678, + "grad_norm": 0.4021293982701474, + "learning_rate": 3.848244472548581e-05, + "loss": 0.2292, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10219310224056244, + "step": 2045, + "valid_targets_mean": 7125.8, + "valid_targets_min": 4623 + }, + { + "epoch": 1.4910876682429974, + "grad_norm": 0.39810224184312526, + "learning_rate": 3.846855619210528e-05, + "loss": 0.2255, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1013958752155304, + "step": 2050, + "valid_targets_mean": 6269.2, + "valid_targets_min": 5458 + }, + { + "epoch": 1.494725354674427, + "grad_norm": 0.4363997646259209, + "learning_rate": 3.8454606924118345e-05, + "loss": 0.2136, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10935162752866745, + "step": 2055, + "valid_targets_mean": 7455.1, + "valid_targets_min": 6008 + }, + { + "epoch": 1.4983630411058566, + "grad_norm": 0.39347252850260633, + "learning_rate": 3.844059696739774e-05, + "loss": 0.222, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11030527949333191, + "step": 2060, + "valid_targets_mean": 7104.4, + "valid_targets_min": 5386 + }, + { + "epoch": 1.5020007275372862, + "grad_norm": 0.40555869460424326, + "learning_rate": 3.842652636801579e-05, + "loss": 0.2136, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11040173470973969, + "step": 2065, + "valid_targets_mean": 6532.6, + "valid_targets_min": 4855 + }, + { + "epoch": 1.505638413968716, + "grad_norm": 0.3965617506847021, + "learning_rate": 3.8412395172244236e-05, + "loss": 0.2194, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10636870563030243, + "step": 2070, + "valid_targets_mean": 6743.9, + "valid_targets_min": 5613 + }, + { + "epoch": 1.5092761004001454, + "grad_norm": 0.4057849645863691, + "learning_rate": 3.83982034265541e-05, + "loss": 0.2184, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10976351797580719, + "step": 2075, + "valid_targets_mean": 6373.6, + "valid_targets_min": 4961 + }, + { + "epoch": 1.5129137868315752, + "grad_norm": 0.45047356634927255, + "learning_rate": 3.838395117761551e-05, + "loss": 0.2221, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11458024382591248, + "step": 2080, + "valid_targets_mean": 6461.2, + "valid_targets_min": 5286 + }, + { + "epoch": 1.5165514732630048, + "grad_norm": 0.4191891542188839, + "learning_rate": 3.836963847229758e-05, + "loss": 0.219, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11502288281917572, + "step": 2085, + "valid_targets_mean": 6780.5, + "valid_targets_min": 5201 + }, + { + "epoch": 1.5201891596944344, + "grad_norm": 0.45399354984270424, + "learning_rate": 3.8355265357668233e-05, + "loss": 0.2172, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1091507077217102, + "step": 2090, + "valid_targets_mean": 5710.8, + "valid_targets_min": 4838 + }, + { + "epoch": 1.523826846125864, + "grad_norm": 0.74102966195753, + "learning_rate": 3.8340831880994044e-05, + "loss": 0.1986, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17269501090049744, + "step": 2095, + "valid_targets_mean": 3235.6, + "valid_targets_min": 176 + }, + { + "epoch": 1.5274645325572935, + "grad_norm": 0.43458743400041583, + "learning_rate": 3.83263380897401e-05, + "loss": 0.2179, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10139960795640945, + "step": 2100, + "valid_targets_mean": 6674.8, + "valid_targets_min": 4582 + }, + { + "epoch": 1.5311022189887231, + "grad_norm": 0.4992921675728047, + "learning_rate": 3.831178403156982e-05, + "loss": 0.2118, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10459695011377335, + "step": 2105, + "valid_targets_mean": 6157.8, + "valid_targets_min": 5385 + }, + { + "epoch": 1.5347399054201527, + "grad_norm": 0.40982834976439947, + "learning_rate": 3.829716975434483e-05, + "loss": 0.2155, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10296086221933365, + "step": 2110, + "valid_targets_mean": 6006.2, + "valid_targets_min": 4478 + }, + { + "epoch": 1.5383775918515825, + "grad_norm": 0.4101728628907703, + "learning_rate": 3.828249530612477e-05, + "loss": 0.2113, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10560064017772675, + "step": 2115, + "valid_targets_mean": 5893.4, + "valid_targets_min": 3911 + }, + { + "epoch": 1.5420152782830119, + "grad_norm": 0.43652944366229435, + "learning_rate": 3.826776073516719e-05, + "loss": 0.2102, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10414370894432068, + "step": 2120, + "valid_targets_mean": 6134.2, + "valid_targets_min": 5318 + }, + { + "epoch": 1.5456529647144417, + "grad_norm": 0.4603205779508088, + "learning_rate": 3.825296608992731e-05, + "loss": 0.2195, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11063805967569351, + "step": 2125, + "valid_targets_mean": 5474.4, + "valid_targets_min": 4959 + }, + { + "epoch": 1.5492906511458713, + "grad_norm": 0.6186562950731753, + "learning_rate": 3.8238111419057935e-05, + "loss": 0.2142, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07158908247947693, + "step": 2130, + "valid_targets_mean": 1804.1, + "valid_targets_min": 842 + }, + { + "epoch": 1.5529283375773009, + "grad_norm": 0.42354760509429584, + "learning_rate": 3.822319677140926e-05, + "loss": 0.1898, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09750007092952728, + "step": 2135, + "valid_targets_mean": 6118.8, + "valid_targets_min": 4899 + }, + { + "epoch": 1.5565660240087305, + "grad_norm": 0.3947810431071811, + "learning_rate": 3.820822219602873e-05, + "loss": 0.2072, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09812428057193756, + "step": 2140, + "valid_targets_mean": 6773.8, + "valid_targets_min": 5046 + }, + { + "epoch": 1.56020371044016, + "grad_norm": 0.4700186955007303, + "learning_rate": 3.819318774216083e-05, + "loss": 0.2164, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1122770756483078, + "step": 2145, + "valid_targets_mean": 6633.6, + "valid_targets_min": 5284 + }, + { + "epoch": 1.5638413968715896, + "grad_norm": 0.417748301130448, + "learning_rate": 3.8178093459247e-05, + "loss": 0.2098, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1063690334558487, + "step": 2150, + "valid_targets_mean": 6055.2, + "valid_targets_min": 4928 + }, + { + "epoch": 1.5674790833030192, + "grad_norm": 0.3917722365168458, + "learning_rate": 3.81629393969254e-05, + "loss": 0.2048, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1107511892914772, + "step": 2155, + "valid_targets_mean": 7259.4, + "valid_targets_min": 5758 + }, + { + "epoch": 1.571116769734449, + "grad_norm": 0.44529477749763946, + "learning_rate": 3.8147725605030785e-05, + "loss": 0.2088, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10766759514808655, + "step": 2160, + "valid_targets_mean": 5731.8, + "valid_targets_min": 4629 + }, + { + "epoch": 1.5747544561658784, + "grad_norm": 0.836146494147364, + "learning_rate": 3.8132452133594345e-05, + "loss": 0.244, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2377433478832245, + "step": 2165, + "valid_targets_mean": 5360.9, + "valid_targets_min": 2317 + }, + { + "epoch": 1.5783921425973082, + "grad_norm": 0.6013295237186254, + "learning_rate": 3.8117119032843506e-05, + "loss": 0.3135, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17129886150360107, + "step": 2170, + "valid_targets_mean": 5303.1, + "valid_targets_min": 533 + }, + { + "epoch": 1.5820298290287376, + "grad_norm": 0.5644731840203969, + "learning_rate": 3.810172635320181e-05, + "loss": 0.3079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16112032532691956, + "step": 2175, + "valid_targets_mean": 4545.8, + "valid_targets_min": 835 + }, + { + "epoch": 1.5856675154601674, + "grad_norm": 0.6376116629142587, + "learning_rate": 3.808627414528871e-05, + "loss": 0.3044, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16823893785476685, + "step": 2180, + "valid_targets_mean": 3618.9, + "valid_targets_min": 1201 + }, + { + "epoch": 1.589305201891597, + "grad_norm": 0.5437508091090669, + "learning_rate": 3.807076245991943e-05, + "loss": 0.2893, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13703912496566772, + "step": 2185, + "valid_targets_mean": 4253.5, + "valid_targets_min": 1880 + }, + { + "epoch": 1.5929428883230266, + "grad_norm": 0.5356380292276516, + "learning_rate": 3.8055191348104795e-05, + "loss": 0.3188, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22360485792160034, + "step": 2190, + "valid_targets_mean": 6610.1, + "valid_targets_min": 2171 + }, + { + "epoch": 1.5965805747544561, + "grad_norm": 0.46692160891098455, + "learning_rate": 3.8039560861051036e-05, + "loss": 0.2929, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1661079227924347, + "step": 2195, + "valid_targets_mean": 7171.4, + "valid_targets_min": 4500 + }, + { + "epoch": 1.6002182611858857, + "grad_norm": 0.4846959313788492, + "learning_rate": 3.802387105015964e-05, + "loss": 0.3002, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15390652418136597, + "step": 2200, + "valid_targets_mean": 6207.4, + "valid_targets_min": 1855 + }, + { + "epoch": 1.6038559476173155, + "grad_norm": 0.6389043800781069, + "learning_rate": 3.800812196702722e-05, + "loss": 0.2998, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15574614703655243, + "step": 2205, + "valid_targets_mean": 4075.9, + "valid_targets_min": 1024 + }, + { + "epoch": 1.607493634048745, + "grad_norm": 0.5049544137792555, + "learning_rate": 3.799231366344527e-05, + "loss": 0.3078, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1319126933813095, + "step": 2210, + "valid_targets_mean": 4767.1, + "valid_targets_min": 2107 + }, + { + "epoch": 1.6111313204801747, + "grad_norm": 0.5808448306054123, + "learning_rate": 3.797644619140005e-05, + "loss": 0.3071, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17311209440231323, + "step": 2215, + "valid_targets_mean": 4211.4, + "valid_targets_min": 802 + }, + { + "epoch": 1.614769006911604, + "grad_norm": 0.6678161853165288, + "learning_rate": 3.796051960307239e-05, + "loss": 0.2918, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14996668696403503, + "step": 2220, + "valid_targets_mean": 5345.1, + "valid_targets_min": 1597 + }, + { + "epoch": 1.6184066933430339, + "grad_norm": 0.6150080452957647, + "learning_rate": 3.794453395083753e-05, + "loss": 0.3078, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15917754173278809, + "step": 2225, + "valid_targets_mean": 3671.4, + "valid_targets_min": 1650 + }, + { + "epoch": 1.6220443797744635, + "grad_norm": 0.5770637783183372, + "learning_rate": 3.792848928726496e-05, + "loss": 0.2978, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15789106488227844, + "step": 2230, + "valid_targets_mean": 4799.1, + "valid_targets_min": 1385 + }, + { + "epoch": 1.625682066205893, + "grad_norm": 0.5818263289019308, + "learning_rate": 3.7912385665118194e-05, + "loss": 0.3014, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16610342264175415, + "step": 2235, + "valid_targets_mean": 4023.5, + "valid_targets_min": 2908 + }, + { + "epoch": 1.6293197526373226, + "grad_norm": 0.5142686853683419, + "learning_rate": 3.789622313735467e-05, + "loss": 0.3075, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14128988981246948, + "step": 2240, + "valid_targets_mean": 5064.6, + "valid_targets_min": 2785 + }, + { + "epoch": 1.6329574390687522, + "grad_norm": 0.5805898038328478, + "learning_rate": 3.7880001757125515e-05, + "loss": 0.298, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1464257538318634, + "step": 2245, + "valid_targets_mean": 3886.4, + "valid_targets_min": 1210 + }, + { + "epoch": 1.636595125500182, + "grad_norm": 0.6014615898979684, + "learning_rate": 3.7863721577775416e-05, + "loss": 0.3096, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17074057459831238, + "step": 2250, + "valid_targets_mean": 3708.8, + "valid_targets_min": 1064 + }, + { + "epoch": 1.6402328119316114, + "grad_norm": 0.6421611626106957, + "learning_rate": 3.784738265284241e-05, + "loss": 0.3195, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15891432762145996, + "step": 2255, + "valid_targets_mean": 3276.2, + "valid_targets_min": 1678 + }, + { + "epoch": 1.6438704983630412, + "grad_norm": 0.6009916372375798, + "learning_rate": 3.783098503605771e-05, + "loss": 0.3133, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1621539145708084, + "step": 2260, + "valid_targets_mean": 3520.1, + "valid_targets_min": 671 + }, + { + "epoch": 1.6475081847944706, + "grad_norm": 0.6052474520830297, + "learning_rate": 3.781452878134557e-05, + "loss": 0.3026, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14615970849990845, + "step": 2265, + "valid_targets_mean": 3488.5, + "valid_targets_min": 2000 + }, + { + "epoch": 1.6511458712259004, + "grad_norm": 0.6053016945668253, + "learning_rate": 3.779801394282305e-05, + "loss": 0.2918, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17980742454528809, + "step": 2270, + "valid_targets_mean": 4402.6, + "valid_targets_min": 1532 + }, + { + "epoch": 1.65478355765733, + "grad_norm": 0.5907605481326663, + "learning_rate": 3.778144057479988e-05, + "loss": 0.294, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11638141423463821, + "step": 2275, + "valid_targets_mean": 3749.4, + "valid_targets_min": 1051 + }, + { + "epoch": 1.6584212440887596, + "grad_norm": 0.5630413780573581, + "learning_rate": 3.7764808731778255e-05, + "loss": 0.2884, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14335747063159943, + "step": 2280, + "valid_targets_mean": 3368.1, + "valid_targets_min": 1345 + }, + { + "epoch": 1.6620589305201892, + "grad_norm": 0.5605317904953371, + "learning_rate": 3.77481184684527e-05, + "loss": 0.3069, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1450687199831009, + "step": 2285, + "valid_targets_mean": 4128.9, + "valid_targets_min": 2231 + }, + { + "epoch": 1.6656966169516187, + "grad_norm": 0.6216291924240485, + "learning_rate": 3.773136983970979e-05, + "loss": 0.3081, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12899094820022583, + "step": 2290, + "valid_targets_mean": 3672.6, + "valid_targets_min": 2127 + }, + { + "epoch": 1.6693343033830483, + "grad_norm": 0.6508518380196745, + "learning_rate": 3.771456290062811e-05, + "loss": 0.2917, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13390398025512695, + "step": 2295, + "valid_targets_mean": 3006.8, + "valid_targets_min": 1265 + }, + { + "epoch": 1.672971989814478, + "grad_norm": 0.6461042399489001, + "learning_rate": 3.769769770647797e-05, + "loss": 0.2947, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13026036322116852, + "step": 2300, + "valid_targets_mean": 2981.9, + "valid_targets_min": 908 + }, + { + "epoch": 1.6766096762459077, + "grad_norm": 0.6781709900425255, + "learning_rate": 3.768077431272124e-05, + "loss": 0.2927, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16859179735183716, + "step": 2305, + "valid_targets_mean": 3147.5, + "valid_targets_min": 1264 + }, + { + "epoch": 1.680247362677337, + "grad_norm": 0.628002382718009, + "learning_rate": 3.76637927750112e-05, + "loss": 0.2933, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13640105724334717, + "step": 2310, + "valid_targets_mean": 2517.9, + "valid_targets_min": 926 + }, + { + "epoch": 1.683885049108767, + "grad_norm": 0.6447676032002161, + "learning_rate": 3.764675314919234e-05, + "loss": 0.2975, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14538241922855377, + "step": 2315, + "valid_targets_mean": 3204.5, + "valid_targets_min": 1016 + }, + { + "epoch": 1.6875227355401963, + "grad_norm": 0.586259989951479, + "learning_rate": 3.762965549130018e-05, + "loss": 0.2861, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14411848783493042, + "step": 2320, + "valid_targets_mean": 3525.0, + "valid_targets_min": 2019 + }, + { + "epoch": 1.691160421971626, + "grad_norm": 0.5932117849897315, + "learning_rate": 3.761249985756104e-05, + "loss": 0.2985, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1695472002029419, + "step": 2325, + "valid_targets_mean": 4745.4, + "valid_targets_min": 3023 + }, + { + "epoch": 1.6947981084030557, + "grad_norm": 0.6089333033456126, + "learning_rate": 3.759528630439196e-05, + "loss": 0.2893, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13138367235660553, + "step": 2330, + "valid_targets_mean": 3503.6, + "valid_targets_min": 1110 + }, + { + "epoch": 1.6984357948344853, + "grad_norm": 0.7468568990783863, + "learning_rate": 3.757801488840039e-05, + "loss": 0.2941, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17060303688049316, + "step": 2335, + "valid_targets_mean": 4206.5, + "valid_targets_min": 1460 + }, + { + "epoch": 1.7020734812659148, + "grad_norm": 0.6251899634927023, + "learning_rate": 3.756068566638412e-05, + "loss": 0.2938, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15705326199531555, + "step": 2340, + "valid_targets_mean": 3938.6, + "valid_targets_min": 2558 + }, + { + "epoch": 1.7057111676973444, + "grad_norm": 0.5810193744276662, + "learning_rate": 3.754329869533099e-05, + "loss": 0.3052, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13834260404109955, + "step": 2345, + "valid_targets_mean": 3421.1, + "valid_targets_min": 1673 + }, + { + "epoch": 1.7093488541287742, + "grad_norm": 0.563880136879837, + "learning_rate": 3.752585403241877e-05, + "loss": 0.2896, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13693152368068695, + "step": 2350, + "valid_targets_mean": 3549.6, + "valid_targets_min": 1860 + }, + { + "epoch": 1.7129865405602036, + "grad_norm": 0.5533740474067155, + "learning_rate": 3.7508351735014955e-05, + "loss": 0.2915, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13213478028774261, + "step": 2355, + "valid_targets_mean": 4414.0, + "valid_targets_min": 2495 + }, + { + "epoch": 1.7166242269916334, + "grad_norm": 0.6234130899745299, + "learning_rate": 3.749079186067657e-05, + "loss": 0.2867, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16662061214447021, + "step": 2360, + "valid_targets_mean": 3861.9, + "valid_targets_min": 1945 + }, + { + "epoch": 1.7202619134230628, + "grad_norm": 0.5994023668013013, + "learning_rate": 3.747317446714998e-05, + "loss": 0.3014, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14931488037109375, + "step": 2365, + "valid_targets_mean": 4488.5, + "valid_targets_min": 1908 + }, + { + "epoch": 1.7238995998544926, + "grad_norm": 0.5688381495346426, + "learning_rate": 3.745549961237072e-05, + "loss": 0.2943, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15035295486450195, + "step": 2370, + "valid_targets_mean": 3507.5, + "valid_targets_min": 1880 + }, + { + "epoch": 1.7275372862859222, + "grad_norm": 0.5540436899957777, + "learning_rate": 3.743776735446326e-05, + "loss": 0.2882, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14924313127994537, + "step": 2375, + "valid_targets_mean": 4491.5, + "valid_targets_min": 2492 + }, + { + "epoch": 1.7311749727173518, + "grad_norm": 0.5896850175905458, + "learning_rate": 3.741997775174087e-05, + "loss": 0.2905, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14745959639549255, + "step": 2380, + "valid_targets_mean": 4228.2, + "valid_targets_min": 3004 + }, + { + "epoch": 1.7348126591487814, + "grad_norm": 0.608785671651594, + "learning_rate": 3.740213086270538e-05, + "loss": 0.2839, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14578546583652496, + "step": 2385, + "valid_targets_mean": 3899.9, + "valid_targets_min": 1833 + }, + { + "epoch": 1.738450345580211, + "grad_norm": 0.6329594587621077, + "learning_rate": 3.7384226746047035e-05, + "loss": 0.2919, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15287545323371887, + "step": 2390, + "valid_targets_mean": 3114.2, + "valid_targets_min": 1182 + }, + { + "epoch": 1.7420880320116408, + "grad_norm": 0.5216313071867097, + "learning_rate": 3.7366265460644244e-05, + "loss": 0.289, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1445210725069046, + "step": 2395, + "valid_targets_mean": 4926.9, + "valid_targets_min": 1869 + }, + { + "epoch": 1.7457257184430701, + "grad_norm": 0.6116847254068266, + "learning_rate": 3.734824706556344e-05, + "loss": 0.2928, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17354264855384827, + "step": 2400, + "valid_targets_mean": 4382.4, + "valid_targets_min": 2498 + }, + { + "epoch": 1.7493634048745, + "grad_norm": 0.6252057289285877, + "learning_rate": 3.7330171620058846e-05, + "loss": 0.3109, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14632268249988556, + "step": 2405, + "valid_targets_mean": 3098.8, + "valid_targets_min": 988 + }, + { + "epoch": 1.7530010913059293, + "grad_norm": 0.5431790150258871, + "learning_rate": 3.7312039183572326e-05, + "loss": 0.2861, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14950695633888245, + "step": 2410, + "valid_targets_mean": 3733.5, + "valid_targets_min": 1534 + }, + { + "epoch": 1.756638777737359, + "grad_norm": 0.5321228177303454, + "learning_rate": 3.7293849815733134e-05, + "loss": 0.2915, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1323934644460678, + "step": 2415, + "valid_targets_mean": 4108.1, + "valid_targets_min": 2252 + }, + { + "epoch": 1.7602764641687887, + "grad_norm": 0.6923928289921933, + "learning_rate": 3.727560357635774e-05, + "loss": 0.2961, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14615699648857117, + "step": 2420, + "valid_targets_mean": 3446.6, + "valid_targets_min": 1436 + }, + { + "epoch": 1.7639141506002183, + "grad_norm": 0.5682707402696403, + "learning_rate": 3.725730052544967e-05, + "loss": 0.2934, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13688966631889343, + "step": 2425, + "valid_targets_mean": 3819.2, + "valid_targets_min": 2333 + }, + { + "epoch": 1.7675518370316479, + "grad_norm": 0.5736314237790495, + "learning_rate": 3.723894072319926e-05, + "loss": 0.2831, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.128215491771698, + "step": 2430, + "valid_targets_mean": 3889.8, + "valid_targets_min": 1272 + }, + { + "epoch": 1.7711895234630775, + "grad_norm": 0.6008646951344413, + "learning_rate": 3.7220524229983465e-05, + "loss": 0.2821, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14260374009609222, + "step": 2435, + "valid_targets_mean": 3233.0, + "valid_targets_min": 1319 + }, + { + "epoch": 1.774827209894507, + "grad_norm": 0.5215414087685304, + "learning_rate": 3.720205110636569e-05, + "loss": 0.2673, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11734236776828766, + "step": 2440, + "valid_targets_mean": 3186.6, + "valid_targets_min": 1194 + }, + { + "epoch": 1.7784648963259366, + "grad_norm": 0.6439919582434681, + "learning_rate": 3.718352141309554e-05, + "loss": 0.307, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12737905979156494, + "step": 2445, + "valid_targets_mean": 2710.6, + "valid_targets_min": 1453 + }, + { + "epoch": 1.7821025827573664, + "grad_norm": 0.5262608951133529, + "learning_rate": 3.71649352111087e-05, + "loss": 0.2804, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1160137802362442, + "step": 2450, + "valid_targets_mean": 5344.5, + "valid_targets_min": 3603 + }, + { + "epoch": 1.7857402691887958, + "grad_norm": 0.5681624019239476, + "learning_rate": 3.7146292561526654e-05, + "loss": 0.2972, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13117824494838715, + "step": 2455, + "valid_targets_mean": 3067.5, + "valid_targets_min": 1256 + }, + { + "epoch": 1.7893779556202256, + "grad_norm": 0.6757011523004973, + "learning_rate": 3.7127593525656505e-05, + "loss": 0.2923, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1938197910785675, + "step": 2460, + "valid_targets_mean": 4443.9, + "valid_targets_min": 1336 + }, + { + "epoch": 1.7930156420516552, + "grad_norm": 0.6460725725454973, + "learning_rate": 3.710883816499082e-05, + "loss": 0.3003, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14912033081054688, + "step": 2465, + "valid_targets_mean": 3581.8, + "valid_targets_min": 1275 + }, + { + "epoch": 1.7966533284830848, + "grad_norm": 0.6309523128403813, + "learning_rate": 3.709002654120736e-05, + "loss": 0.2884, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16325312852859497, + "step": 2470, + "valid_targets_mean": 3770.1, + "valid_targets_min": 2202 + }, + { + "epoch": 1.8002910149145144, + "grad_norm": 0.6266569787630676, + "learning_rate": 3.707115871616893e-05, + "loss": 0.2839, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16523504257202148, + "step": 2475, + "valid_targets_mean": 4112.8, + "valid_targets_min": 3332 + }, + { + "epoch": 1.803928701345944, + "grad_norm": 0.6070853498255162, + "learning_rate": 3.705223475192315e-05, + "loss": 0.2824, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15209759771823883, + "step": 2480, + "valid_targets_mean": 3260.0, + "valid_targets_min": 2195 + }, + { + "epoch": 1.8075663877773736, + "grad_norm": 0.5870749881706073, + "learning_rate": 3.703325471070225e-05, + "loss": 0.2818, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14166031777858734, + "step": 2485, + "valid_targets_mean": 4039.6, + "valid_targets_min": 2343 + }, + { + "epoch": 1.8112040742088031, + "grad_norm": 0.5915732292302752, + "learning_rate": 3.7014218654922884e-05, + "loss": 0.2755, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12826409935951233, + "step": 2490, + "valid_targets_mean": 2678.8, + "valid_targets_min": 1026 + }, + { + "epoch": 1.814841760640233, + "grad_norm": 0.6232498047236771, + "learning_rate": 3.699512664718589e-05, + "loss": 0.2772, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13827820122241974, + "step": 2495, + "valid_targets_mean": 3754.5, + "valid_targets_min": 1042 + }, + { + "epoch": 1.8184794470716623, + "grad_norm": 0.6321492068283825, + "learning_rate": 3.6975978750276136e-05, + "loss": 0.2677, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13580133020877838, + "step": 2500, + "valid_targets_mean": 3301.6, + "valid_targets_min": 1274 + }, + { + "epoch": 1.8221171335030921, + "grad_norm": 0.593507872700842, + "learning_rate": 3.695677502716226e-05, + "loss": 0.2832, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13837912678718567, + "step": 2505, + "valid_targets_mean": 3632.2, + "valid_targets_min": 914 + }, + { + "epoch": 1.8257548199345215, + "grad_norm": 0.5341554227747026, + "learning_rate": 3.69375155409965e-05, + "loss": 0.3099, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12933139503002167, + "step": 2510, + "valid_targets_mean": 4342.5, + "valid_targets_min": 2093 + }, + { + "epoch": 1.8293925063659513, + "grad_norm": 0.6663177728420913, + "learning_rate": 3.691820035511446e-05, + "loss": 0.285, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14841708540916443, + "step": 2515, + "valid_targets_mean": 3609.6, + "valid_targets_min": 1683 + }, + { + "epoch": 1.8330301927973809, + "grad_norm": 0.5505921345744846, + "learning_rate": 3.6898829533034926e-05, + "loss": 0.2919, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13663962483406067, + "step": 2520, + "valid_targets_mean": 4096.2, + "valid_targets_min": 2253 + }, + { + "epoch": 1.8366678792288105, + "grad_norm": 0.5803891170668133, + "learning_rate": 3.687940313845964e-05, + "loss": 0.2795, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13058030605316162, + "step": 2525, + "valid_targets_mean": 3794.4, + "valid_targets_min": 1789 + }, + { + "epoch": 1.84030556566024, + "grad_norm": 0.590053412570449, + "learning_rate": 3.685992123527311e-05, + "loss": 0.2882, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14132004976272583, + "step": 2530, + "valid_targets_mean": 3753.8, + "valid_targets_min": 1586 + }, + { + "epoch": 1.8439432520916696, + "grad_norm": 0.6091203491528314, + "learning_rate": 3.6840383887542366e-05, + "loss": 0.2833, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1279279589653015, + "step": 2535, + "valid_targets_mean": 2995.5, + "valid_targets_min": 2033 + }, + { + "epoch": 1.8475809385230995, + "grad_norm": 0.5424579702521947, + "learning_rate": 3.682079115951679e-05, + "loss": 0.2695, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12945051491260529, + "step": 2540, + "valid_targets_mean": 3690.5, + "valid_targets_min": 1616 + }, + { + "epoch": 1.8512186249545288, + "grad_norm": 0.539304016555912, + "learning_rate": 3.680114311562785e-05, + "loss": 0.2891, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12396745383739471, + "step": 2545, + "valid_targets_mean": 3303.9, + "valid_targets_min": 2000 + }, + { + "epoch": 1.8548563113859586, + "grad_norm": 0.58405822163949, + "learning_rate": 3.6781439820488974e-05, + "loss": 0.2783, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15788614749908447, + "step": 2550, + "valid_targets_mean": 3788.9, + "valid_targets_min": 1750 + }, + { + "epoch": 1.858493997817388, + "grad_norm": 0.6074785950522859, + "learning_rate": 3.6761681338895245e-05, + "loss": 0.2898, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16149112582206726, + "step": 2555, + "valid_targets_mean": 3898.2, + "valid_targets_min": 2126 + }, + { + "epoch": 1.8621316842488178, + "grad_norm": 0.5835907835314927, + "learning_rate": 3.6741867735823245e-05, + "loss": 0.2985, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15480536222457886, + "step": 2560, + "valid_targets_mean": 3923.1, + "valid_targets_min": 3049 + }, + { + "epoch": 1.8657693706802474, + "grad_norm": 0.5800526509989581, + "learning_rate": 3.672199907643082e-05, + "loss": 0.2875, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16368468105793, + "step": 2565, + "valid_targets_mean": 4337.1, + "valid_targets_min": 1571 + }, + { + "epoch": 1.869407057111677, + "grad_norm": 0.6199866221253162, + "learning_rate": 3.6702075426056875e-05, + "loss": 0.3052, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13748183846473694, + "step": 2570, + "valid_targets_mean": 2998.0, + "valid_targets_min": 1487 + }, + { + "epoch": 1.8730447435431066, + "grad_norm": 0.5850717564336533, + "learning_rate": 3.6682096850221154e-05, + "loss": 0.2924, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11947401612997055, + "step": 2575, + "valid_targets_mean": 3627.1, + "valid_targets_min": 1970 + }, + { + "epoch": 1.8766824299745362, + "grad_norm": 0.5583004251910663, + "learning_rate": 3.666206341462402e-05, + "loss": 0.2986, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.128853440284729, + "step": 2580, + "valid_targets_mean": 4038.0, + "valid_targets_min": 2492 + }, + { + "epoch": 1.880320116405966, + "grad_norm": 0.5902407794373904, + "learning_rate": 3.664197518514626e-05, + "loss": 0.2993, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1261817216873169, + "step": 2585, + "valid_targets_mean": 3211.2, + "valid_targets_min": 1066 + }, + { + "epoch": 1.8839578028373953, + "grad_norm": 0.5713311447831886, + "learning_rate": 3.662183222784883e-05, + "loss": 0.2988, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13162414729595184, + "step": 2590, + "valid_targets_mean": 3401.9, + "valid_targets_min": 2281 + }, + { + "epoch": 1.8875954892688251, + "grad_norm": 0.6203701969103899, + "learning_rate": 3.660163460897268e-05, + "loss": 0.2935, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14571444690227509, + "step": 2595, + "valid_targets_mean": 3662.2, + "valid_targets_min": 1969 + }, + { + "epoch": 1.8912331757002545, + "grad_norm": 0.5899995925255076, + "learning_rate": 3.658138239493851e-05, + "loss": 0.2938, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12907736003398895, + "step": 2600, + "valid_targets_mean": 3189.6, + "valid_targets_min": 1005 + }, + { + "epoch": 1.8948708621316843, + "grad_norm": 0.5984815488543777, + "learning_rate": 3.656107565234656e-05, + "loss": 0.2935, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13155272603034973, + "step": 2605, + "valid_targets_mean": 3546.5, + "valid_targets_min": 1980 + }, + { + "epoch": 1.898508548563114, + "grad_norm": 0.5542486166646338, + "learning_rate": 3.6540714447976385e-05, + "loss": 0.2925, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16652345657348633, + "step": 2610, + "valid_targets_mean": 4803.1, + "valid_targets_min": 1257 + }, + { + "epoch": 1.9021462349945435, + "grad_norm": 0.6234214844496144, + "learning_rate": 3.652029884878665e-05, + "loss": 0.2856, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1500328630208969, + "step": 2615, + "valid_targets_mean": 3118.1, + "valid_targets_min": 1357 + }, + { + "epoch": 1.905783921425973, + "grad_norm": 0.5558923229052543, + "learning_rate": 3.649982892191488e-05, + "loss": 0.2705, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12016848474740982, + "step": 2620, + "valid_targets_mean": 4463.8, + "valid_targets_min": 1410 + }, + { + "epoch": 1.9094216078574027, + "grad_norm": 0.5927565835803827, + "learning_rate": 3.647930473467728e-05, + "loss": 0.292, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1333162933588028, + "step": 2625, + "valid_targets_mean": 3429.8, + "valid_targets_min": 1390 + }, + { + "epoch": 1.9130592942888323, + "grad_norm": 0.5966013376643061, + "learning_rate": 3.6458726354568474e-05, + "loss": 0.2759, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12504485249519348, + "step": 2630, + "valid_targets_mean": 3173.0, + "valid_targets_min": 790 + }, + { + "epoch": 1.9166969807202618, + "grad_norm": 0.5758583039279905, + "learning_rate": 3.643809384926133e-05, + "loss": 0.2968, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14954900741577148, + "step": 2635, + "valid_targets_mean": 3788.8, + "valid_targets_min": 1901 + }, + { + "epoch": 1.9203346671516917, + "grad_norm": 0.6426947379141987, + "learning_rate": 3.641740728660667e-05, + "loss": 0.287, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13260552287101746, + "step": 2640, + "valid_targets_mean": 2697.1, + "valid_targets_min": 1460 + }, + { + "epoch": 1.923972353583121, + "grad_norm": 0.6371656585918946, + "learning_rate": 3.6396666734633115e-05, + "loss": 0.2845, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16277393698692322, + "step": 2645, + "valid_targets_mean": 5088.8, + "valid_targets_min": 952 + }, + { + "epoch": 1.9276100400145508, + "grad_norm": 0.6819385277108153, + "learning_rate": 3.637587226154683e-05, + "loss": 0.2884, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15208669006824493, + "step": 2650, + "valid_targets_mean": 3070.5, + "valid_targets_min": 1768 + }, + { + "epoch": 1.9312477264459802, + "grad_norm": 0.8346096868294602, + "learning_rate": 3.635502393573127e-05, + "loss": 0.2863, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1127871498465538, + "step": 2655, + "valid_targets_mean": 3585.0, + "valid_targets_min": 1931 + }, + { + "epoch": 1.93488541287741, + "grad_norm": 0.49659462916093533, + "learning_rate": 3.633412182574704e-05, + "loss": 0.2784, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13027313351631165, + "step": 2660, + "valid_targets_mean": 5915.4, + "valid_targets_min": 1233 + }, + { + "epoch": 1.9385230993088396, + "grad_norm": 0.5201387573873989, + "learning_rate": 3.631316600033158e-05, + "loss": 0.274, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1369105875492096, + "step": 2665, + "valid_targets_mean": 5310.6, + "valid_targets_min": 1741 + }, + { + "epoch": 1.9421607857402692, + "grad_norm": 0.5837547063752329, + "learning_rate": 3.629215652839898e-05, + "loss": 0.275, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1305580735206604, + "step": 2670, + "valid_targets_mean": 3884.8, + "valid_targets_min": 856 + }, + { + "epoch": 1.9457984721716988, + "grad_norm": 0.5582339581567916, + "learning_rate": 3.627109347903974e-05, + "loss": 0.2783, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15124498307704926, + "step": 2675, + "valid_targets_mean": 3649.5, + "valid_targets_min": 1062 + }, + { + "epoch": 1.9494361586031284, + "grad_norm": 0.5223300927375242, + "learning_rate": 3.624997692152059e-05, + "loss": 0.2709, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13489292562007904, + "step": 2680, + "valid_targets_mean": 4412.6, + "valid_targets_min": 974 + }, + { + "epoch": 1.9530738450345582, + "grad_norm": 0.6117851569720596, + "learning_rate": 3.622880692528417e-05, + "loss": 0.2763, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14685887098312378, + "step": 2685, + "valid_targets_mean": 3077.5, + "valid_targets_min": 1447 + }, + { + "epoch": 1.9567115314659875, + "grad_norm": 0.5630035366421878, + "learning_rate": 3.620758355994891e-05, + "loss": 0.2721, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16126714646816254, + "step": 2690, + "valid_targets_mean": 4663.0, + "valid_targets_min": 2144 + }, + { + "epoch": 1.9603492178974173, + "grad_norm": 0.5627085548172075, + "learning_rate": 3.61863068953087e-05, + "loss": 0.2735, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15522009134292603, + "step": 2695, + "valid_targets_mean": 4543.1, + "valid_targets_min": 1515 + }, + { + "epoch": 1.9639869043288467, + "grad_norm": 0.6860784319348854, + "learning_rate": 3.616497700133274e-05, + "loss": 0.2915, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14012083411216736, + "step": 2700, + "valid_targets_mean": 3034.9, + "valid_targets_min": 1349 + }, + { + "epoch": 1.9676245907602765, + "grad_norm": 0.67406721995344, + "learning_rate": 3.614359394816526e-05, + "loss": 0.2923, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14292705059051514, + "step": 2705, + "valid_targets_mean": 3150.4, + "valid_targets_min": 1906 + }, + { + "epoch": 1.971262277191706, + "grad_norm": 0.5942684865116854, + "learning_rate": 3.612215780612529e-05, + "loss": 0.2796, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14757180213928223, + "step": 2710, + "valid_targets_mean": 3655.1, + "valid_targets_min": 2076 + }, + { + "epoch": 1.9748999636231357, + "grad_norm": 0.5946444097414159, + "learning_rate": 3.610066864570649e-05, + "loss": 0.2728, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11566498875617981, + "step": 2715, + "valid_targets_mean": 2994.0, + "valid_targets_min": 1021 + }, + { + "epoch": 1.9785376500545653, + "grad_norm": 0.6196009319984825, + "learning_rate": 3.6079126537576835e-05, + "loss": 0.2851, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1436595618724823, + "step": 2720, + "valid_targets_mean": 3226.9, + "valid_targets_min": 654 + }, + { + "epoch": 1.9821753364859949, + "grad_norm": 0.5672780789663087, + "learning_rate": 3.605753155257844e-05, + "loss": 0.2716, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1594313681125641, + "step": 2725, + "valid_targets_mean": 4790.8, + "valid_targets_min": 3074 + }, + { + "epoch": 1.9858130229174247, + "grad_norm": 0.6880926235835337, + "learning_rate": 3.603588376172729e-05, + "loss": 0.2961, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14670860767364502, + "step": 2730, + "valid_targets_mean": 2604.0, + "valid_targets_min": 815 + }, + { + "epoch": 1.989450709348854, + "grad_norm": 0.5662385564660654, + "learning_rate": 3.601418323621305e-05, + "loss": 0.2813, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1282183825969696, + "step": 2735, + "valid_targets_mean": 3560.0, + "valid_targets_min": 2205 + }, + { + "epoch": 1.9930883957802839, + "grad_norm": 0.5650672128215765, + "learning_rate": 3.5992430047398785e-05, + "loss": 0.2865, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1532011181116104, + "step": 2740, + "valid_targets_mean": 4216.0, + "valid_targets_min": 1793 + }, + { + "epoch": 1.9967260822117132, + "grad_norm": 0.5915622486159243, + "learning_rate": 3.597062426682074e-05, + "loss": 0.2885, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13346654176712036, + "step": 2745, + "valid_targets_mean": 3134.5, + "valid_targets_min": 1217 + }, + { + "epoch": 2.0, + "grad_norm": 0.8083546393606036, + "learning_rate": 3.5948765966188136e-05, + "loss": 0.2622, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16324765980243683, + "step": 2750, + "valid_targets_mean": 1910.0, + "valid_targets_min": 393 + }, + { + "epoch": 2.00363768643143, + "grad_norm": 0.48877741768238864, + "learning_rate": 3.592685521738289e-05, + "loss": 0.1482, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06046832352876663, + "step": 2755, + "valid_targets_mean": 2722.6, + "valid_targets_min": 458 + }, + { + "epoch": 2.007275372862859, + "grad_norm": 0.47743291162182705, + "learning_rate": 3.59048920924594e-05, + "loss": 0.1166, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05880669504404068, + "step": 2760, + "valid_targets_mean": 3113.1, + "valid_targets_min": 2461 + }, + { + "epoch": 2.010913059294289, + "grad_norm": 0.36714858325078414, + "learning_rate": 3.588287666364429e-05, + "loss": 0.1058, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06400761753320694, + "step": 2765, + "valid_targets_mean": 3648.1, + "valid_targets_min": 2068 + }, + { + "epoch": 2.0145507457257184, + "grad_norm": 0.4201960619632501, + "learning_rate": 3.586080900333623e-05, + "loss": 0.1296, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06039399653673172, + "step": 2770, + "valid_targets_mean": 3423.6, + "valid_targets_min": 2425 + }, + { + "epoch": 2.018188432157148, + "grad_norm": 0.43723437195874515, + "learning_rate": 3.583868918410561e-05, + "loss": 0.1043, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05166186764836311, + "step": 2775, + "valid_targets_mean": 2724.1, + "valid_targets_min": 1198 + }, + { + "epoch": 2.0218261185885775, + "grad_norm": 0.678843501818486, + "learning_rate": 3.581651727869436e-05, + "loss": 0.1237, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08352436125278473, + "step": 2780, + "valid_targets_mean": 1232.5, + "valid_targets_min": 807 + }, + { + "epoch": 2.0254638050200073, + "grad_norm": 0.4825739095815928, + "learning_rate": 3.5794293360015726e-05, + "loss": 0.1662, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06753218919038773, + "step": 2785, + "valid_targets_mean": 3487.4, + "valid_targets_min": 2351 + }, + { + "epoch": 2.0291014914514367, + "grad_norm": 0.466138907729386, + "learning_rate": 3.577201750115396e-05, + "loss": 0.1157, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08982378244400024, + "step": 2790, + "valid_targets_mean": 3126.4, + "valid_targets_min": 781 + }, + { + "epoch": 2.0327391778828665, + "grad_norm": 0.3467328138475409, + "learning_rate": 3.574968977536414e-05, + "loss": 0.1279, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05967942625284195, + "step": 2795, + "valid_targets_mean": 3474.8, + "valid_targets_min": 799 + }, + { + "epoch": 2.0363768643142963, + "grad_norm": 0.3119279393122668, + "learning_rate": 3.572731025607192e-05, + "loss": 0.1316, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.053353969007730484, + "step": 2800, + "valid_targets_mean": 5026.1, + "valid_targets_min": 2802 + }, + { + "epoch": 2.0400145507457257, + "grad_norm": 0.34006324847202213, + "learning_rate": 3.5704879016873256e-05, + "loss": 0.1023, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04003140702843666, + "step": 2805, + "valid_targets_mean": 2611.0, + "valid_targets_min": 574 + }, + { + "epoch": 2.0436522371771555, + "grad_norm": 0.3704314671864696, + "learning_rate": 3.568239613153421e-05, + "loss": 0.1116, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.053696949034929276, + "step": 2810, + "valid_targets_mean": 4036.5, + "valid_targets_min": 2642 + }, + { + "epoch": 2.047289923608585, + "grad_norm": 0.5452756534730892, + "learning_rate": 3.5659861673990676e-05, + "loss": 0.116, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0762016773223877, + "step": 2815, + "valid_targets_mean": 1960.5, + "valid_targets_min": 593 + }, + { + "epoch": 2.0509276100400147, + "grad_norm": 0.3902525806254664, + "learning_rate": 3.563727571834814e-05, + "loss": 0.1433, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04666804522275925, + "step": 2820, + "valid_targets_mean": 3354.2, + "valid_targets_min": 1353 + }, + { + "epoch": 2.054565296471444, + "grad_norm": 0.5015425679939255, + "learning_rate": 3.561463833888145e-05, + "loss": 0.1077, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06352075189352036, + "step": 2825, + "valid_targets_mean": 2300.9, + "valid_targets_min": 818 + }, + { + "epoch": 2.058202982902874, + "grad_norm": 0.3493294924730744, + "learning_rate": 3.559194961003455e-05, + "loss": 0.1079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03910066932439804, + "step": 2830, + "valid_targets_mean": 2295.8, + "valid_targets_min": 934 + }, + { + "epoch": 2.061840669334303, + "grad_norm": 0.3535910557282145, + "learning_rate": 3.556920960642027e-05, + "loss": 0.0989, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.054950203746557236, + "step": 2835, + "valid_targets_mean": 3524.4, + "valid_targets_min": 2798 + }, + { + "epoch": 2.065478355765733, + "grad_norm": 0.46359763626155465, + "learning_rate": 3.554641840282004e-05, + "loss": 0.1233, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0501517578959465, + "step": 2840, + "valid_targets_mean": 2810.1, + "valid_targets_min": 471 + }, + { + "epoch": 2.0691160421971624, + "grad_norm": 0.4306839969553911, + "learning_rate": 3.5523576074183675e-05, + "loss": 0.1246, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07864253222942352, + "step": 2845, + "valid_targets_mean": 3042.8, + "valid_targets_min": 903 + }, + { + "epoch": 2.072753728628592, + "grad_norm": 0.40951537503176416, + "learning_rate": 3.5500682695629096e-05, + "loss": 0.1239, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.048468418419361115, + "step": 2850, + "valid_targets_mean": 3081.6, + "valid_targets_min": 474 + }, + { + "epoch": 2.076391415060022, + "grad_norm": 0.5068986008739655, + "learning_rate": 3.547773834244214e-05, + "loss": 0.1252, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.060791343450546265, + "step": 2855, + "valid_targets_mean": 3468.0, + "valid_targets_min": 640 + }, + { + "epoch": 2.0800291014914514, + "grad_norm": 0.3372411707634332, + "learning_rate": 3.5454743090076244e-05, + "loss": 0.1605, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.039850834757089615, + "step": 2860, + "valid_targets_mean": 3438.4, + "valid_targets_min": 902 + }, + { + "epoch": 2.083666787922881, + "grad_norm": 0.40871987354280637, + "learning_rate": 3.543169701415225e-05, + "loss": 0.1059, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05288071557879448, + "step": 2865, + "valid_targets_mean": 3212.8, + "valid_targets_min": 1837 + }, + { + "epoch": 2.0873044743543105, + "grad_norm": 0.7524064180339942, + "learning_rate": 3.5408600190458115e-05, + "loss": 0.1429, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11426818370819092, + "step": 2870, + "valid_targets_mean": 1358.8, + "valid_targets_min": 756 + }, + { + "epoch": 2.0909421607857404, + "grad_norm": 0.5826929123282547, + "learning_rate": 3.5385452694948694e-05, + "loss": 0.1436, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09977076202630997, + "step": 2875, + "valid_targets_mean": 2896.2, + "valid_targets_min": 1072 + }, + { + "epoch": 2.0945798472171697, + "grad_norm": 0.38440809464150555, + "learning_rate": 3.5362254603745494e-05, + "loss": 0.1121, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.036122795194387436, + "step": 2880, + "valid_targets_mean": 2908.1, + "valid_targets_min": 895 + }, + { + "epoch": 2.0982175336485995, + "grad_norm": 0.6506488522662732, + "learning_rate": 3.5339005993136375e-05, + "loss": 0.2414, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2070414423942566, + "step": 2885, + "valid_targets_mean": 2562.2, + "valid_targets_min": 878 + }, + { + "epoch": 2.101855220080029, + "grad_norm": 0.439691448847747, + "learning_rate": 3.531570693957536e-05, + "loss": 0.2202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05212794616818428, + "step": 2890, + "valid_targets_mean": 2087.6, + "valid_targets_min": 756 + }, + { + "epoch": 2.1054929065114587, + "grad_norm": 0.32916444275103013, + "learning_rate": 3.529235751968233e-05, + "loss": 0.1258, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04458755999803543, + "step": 2895, + "valid_targets_mean": 3421.8, + "valid_targets_min": 2963 + }, + { + "epoch": 2.1091305929428885, + "grad_norm": 0.7309167612072405, + "learning_rate": 3.526895781024282e-05, + "loss": 0.2, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24805089831352234, + "step": 2900, + "valid_targets_mean": 2375.2, + "valid_targets_min": 1153 + }, + { + "epoch": 2.112768279374318, + "grad_norm": 0.28666814907371146, + "learning_rate": 3.5245507888207746e-05, + "loss": 0.1338, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04441724345088005, + "step": 2905, + "valid_targets_mean": 4886.2, + "valid_targets_min": 1700 + }, + { + "epoch": 2.1164059658057477, + "grad_norm": 0.322337168111202, + "learning_rate": 3.522200783069313e-05, + "loss": 0.1088, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03706035017967224, + "step": 2910, + "valid_targets_mean": 3965.1, + "valid_targets_min": 789 + }, + { + "epoch": 2.120043652237177, + "grad_norm": 0.38197397813000467, + "learning_rate": 3.5198457714979875e-05, + "loss": 0.1199, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0603359118103981, + "step": 2915, + "valid_targets_mean": 3771.5, + "valid_targets_min": 995 + }, + { + "epoch": 2.123681338668607, + "grad_norm": 0.40412560723057667, + "learning_rate": 3.517485761851352e-05, + "loss": 0.1078, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03604806587100029, + "step": 2920, + "valid_targets_mean": 2220.8, + "valid_targets_min": 488 + }, + { + "epoch": 2.1273190251000362, + "grad_norm": 0.4368204982701886, + "learning_rate": 3.515120761890393e-05, + "loss": 0.1097, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06320638954639435, + "step": 2925, + "valid_targets_mean": 3635.4, + "valid_targets_min": 2437 + }, + { + "epoch": 2.130956711531466, + "grad_norm": 0.5073561036149132, + "learning_rate": 3.5127507793925124e-05, + "loss": 0.105, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0446467325091362, + "step": 2930, + "valid_targets_mean": 1752.1, + "valid_targets_min": 712 + }, + { + "epoch": 2.1345943979628954, + "grad_norm": 0.3640595873650366, + "learning_rate": 3.510375822151494e-05, + "loss": 0.1045, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05360904335975647, + "step": 2935, + "valid_targets_mean": 3713.4, + "valid_targets_min": 2422 + }, + { + "epoch": 2.138232084394325, + "grad_norm": 0.3868439652976722, + "learning_rate": 3.5079958979774816e-05, + "loss": 0.0944, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05026811733841896, + "step": 2940, + "valid_targets_mean": 3998.5, + "valid_targets_min": 3643 + }, + { + "epoch": 2.141869770825755, + "grad_norm": 0.3045786934315857, + "learning_rate": 3.505611014696955e-05, + "loss": 0.1611, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.036718808114528656, + "step": 2945, + "valid_targets_mean": 3319.0, + "valid_targets_min": 772 + }, + { + "epoch": 2.1455074572571844, + "grad_norm": 0.3609787195208623, + "learning_rate": 3.503221180152701e-05, + "loss": 0.088, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.037231795489788055, + "step": 2950, + "valid_targets_mean": 2934.8, + "valid_targets_min": 722 + }, + { + "epoch": 2.149145143688614, + "grad_norm": 0.3411383793499024, + "learning_rate": 3.5008264022037876e-05, + "loss": 0.0981, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.037633031606674194, + "step": 2955, + "valid_targets_mean": 3656.4, + "valid_targets_min": 2788 + }, + { + "epoch": 2.1527828301200436, + "grad_norm": 0.368237750277764, + "learning_rate": 3.4984266887255415e-05, + "loss": 0.0989, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05114630609750748, + "step": 2960, + "valid_targets_mean": 3202.5, + "valid_targets_min": 948 + }, + { + "epoch": 2.1564205165514734, + "grad_norm": 0.5452753320982576, + "learning_rate": 3.496022047609518e-05, + "loss": 0.1216, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05943308770656586, + "step": 2965, + "valid_targets_mean": 1002.0, + "valid_targets_min": 490 + }, + { + "epoch": 2.1600582029829027, + "grad_norm": 0.33242977746250935, + "learning_rate": 3.493612486763479e-05, + "loss": 0.1355, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05804983153939247, + "step": 2970, + "valid_targets_mean": 3751.9, + "valid_targets_min": 2800 + }, + { + "epoch": 2.1636958894143326, + "grad_norm": 0.4898617278456878, + "learning_rate": 3.4911980141113655e-05, + "loss": 0.1446, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09501232951879501, + "step": 2975, + "valid_targets_mean": 2641.8, + "valid_targets_min": 896 + }, + { + "epoch": 2.167333575845762, + "grad_norm": 0.4577435486320592, + "learning_rate": 3.488778637593269e-05, + "loss": 0.1202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0689048320055008, + "step": 2980, + "valid_targets_mean": 3199.0, + "valid_targets_min": 1333 + }, + { + "epoch": 2.1709712622771917, + "grad_norm": 0.4597882379786227, + "learning_rate": 3.486354365165408e-05, + "loss": 0.1364, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06140146777033806, + "step": 2985, + "valid_targets_mean": 2548.1, + "valid_targets_min": 693 + }, + { + "epoch": 2.174608948708621, + "grad_norm": 0.37634851223301274, + "learning_rate": 3.4839252048001024e-05, + "loss": 0.1497, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0453992635011673, + "step": 2990, + "valid_targets_mean": 2482.9, + "valid_targets_min": 803 + }, + { + "epoch": 2.178246635140051, + "grad_norm": 0.3902061564036622, + "learning_rate": 3.481491164485747e-05, + "loss": 0.1132, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05945345014333725, + "step": 2995, + "valid_targets_mean": 3239.2, + "valid_targets_min": 2525 + }, + { + "epoch": 2.1818843215714807, + "grad_norm": 0.4255019873826585, + "learning_rate": 3.479052252226783e-05, + "loss": 0.1378, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04884905368089676, + "step": 3000, + "valid_targets_mean": 3806.1, + "valid_targets_min": 2393 + }, + { + "epoch": 2.18552200800291, + "grad_norm": 0.41520235459088456, + "learning_rate": 3.476608476043674e-05, + "loss": 0.1217, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05218043178319931, + "step": 3005, + "valid_targets_mean": 2990.9, + "valid_targets_min": 804 + }, + { + "epoch": 2.18915969443434, + "grad_norm": 0.35768118664784165, + "learning_rate": 3.474159843972877e-05, + "loss": 0.1077, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05171472951769829, + "step": 3010, + "valid_targets_mean": 4885.8, + "valid_targets_min": 3695 + }, + { + "epoch": 2.1927973808657693, + "grad_norm": 0.4309006562177706, + "learning_rate": 3.4717063640668195e-05, + "loss": 0.0988, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06148819625377655, + "step": 3015, + "valid_targets_mean": 3128.8, + "valid_targets_min": 596 + }, + { + "epoch": 2.196435067297199, + "grad_norm": 0.2913973437331078, + "learning_rate": 3.469248044393873e-05, + "loss": 0.0962, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03654723986983299, + "step": 3020, + "valid_targets_mean": 4567.8, + "valid_targets_min": 3804 + }, + { + "epoch": 2.2000727537286284, + "grad_norm": 0.3273830422311285, + "learning_rate": 3.4667848930383203e-05, + "loss": 0.1034, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04058642312884331, + "step": 3025, + "valid_targets_mean": 3853.1, + "valid_targets_min": 2862 + }, + { + "epoch": 2.2037104401600582, + "grad_norm": 0.5782213176457369, + "learning_rate": 3.464316918100338e-05, + "loss": 0.1284, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07317621260881424, + "step": 3030, + "valid_targets_mean": 1072.6, + "valid_targets_min": 497 + }, + { + "epoch": 2.207348126591488, + "grad_norm": 0.3415104723654016, + "learning_rate": 3.461844127695962e-05, + "loss": 0.1077, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06036854535341263, + "step": 3035, + "valid_targets_mean": 4191.0, + "valid_targets_min": 3918 + }, + { + "epoch": 2.2109858130229174, + "grad_norm": 0.5251908166979228, + "learning_rate": 3.4593665299570656e-05, + "loss": 0.1218, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05495373159646988, + "step": 3040, + "valid_targets_mean": 2121.9, + "valid_targets_min": 785 + }, + { + "epoch": 2.214623499454347, + "grad_norm": 0.3176026902272693, + "learning_rate": 3.456884133031331e-05, + "loss": 0.0886, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04604320228099823, + "step": 3045, + "valid_targets_mean": 3451.6, + "valid_targets_min": 905 + }, + { + "epoch": 2.2182611858857766, + "grad_norm": 0.3817545415433723, + "learning_rate": 3.4543969450822216e-05, + "loss": 0.1166, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0843544751405716, + "step": 3050, + "valid_targets_mean": 2929.2, + "valid_targets_min": 657 + }, + { + "epoch": 2.2218988723172064, + "grad_norm": 0.3614669083194947, + "learning_rate": 3.451904974288958e-05, + "loss": 0.1086, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05587051808834076, + "step": 3055, + "valid_targets_mean": 3798.9, + "valid_targets_min": 3184 + }, + { + "epoch": 2.2255365587486358, + "grad_norm": 0.46611145073456844, + "learning_rate": 3.4494082288464886e-05, + "loss": 0.0963, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05795666202902794, + "step": 3060, + "valid_targets_mean": 1463.9, + "valid_targets_min": 627 + }, + { + "epoch": 2.2291742451800656, + "grad_norm": 0.46043041108726945, + "learning_rate": 3.4469067169654636e-05, + "loss": 0.1104, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05231434851884842, + "step": 3065, + "valid_targets_mean": 1701.5, + "valid_targets_min": 827 + }, + { + "epoch": 2.232811931611495, + "grad_norm": 0.47537101008718463, + "learning_rate": 3.444400446872208e-05, + "loss": 0.1136, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05114983767271042, + "step": 3070, + "valid_targets_mean": 1789.9, + "valid_targets_min": 940 + }, + { + "epoch": 2.2364496180429247, + "grad_norm": 0.39919004972361094, + "learning_rate": 3.4418894268086924e-05, + "loss": 0.1304, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04644160717725754, + "step": 3075, + "valid_targets_mean": 2943.8, + "valid_targets_min": 958 + }, + { + "epoch": 2.240087304474354, + "grad_norm": 0.39888403661777105, + "learning_rate": 3.439373665032511e-05, + "loss": 0.1087, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04403233528137207, + "step": 3080, + "valid_targets_mean": 1859.2, + "valid_targets_min": 719 + }, + { + "epoch": 2.243724990905784, + "grad_norm": 0.5230497624385626, + "learning_rate": 3.43685316981685e-05, + "loss": 0.1333, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04043084383010864, + "step": 3085, + "valid_targets_mean": 1703.9, + "valid_targets_min": 759 + }, + { + "epoch": 2.2473626773372137, + "grad_norm": 0.6677232070125722, + "learning_rate": 3.434327949450461e-05, + "loss": 0.1323, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1349942535161972, + "step": 3090, + "valid_targets_mean": 1556.6, + "valid_targets_min": 521 + }, + { + "epoch": 2.251000363768643, + "grad_norm": 0.38268714114632857, + "learning_rate": 3.431798012237634e-05, + "loss": 0.1071, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06546276807785034, + "step": 3095, + "valid_targets_mean": 3267.6, + "valid_targets_min": 1032 + }, + { + "epoch": 2.254638050200073, + "grad_norm": 0.30947618810076505, + "learning_rate": 3.429263366498173e-05, + "loss": 0.0926, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03844860941171646, + "step": 3100, + "valid_targets_mean": 3609.0, + "valid_targets_min": 2741 + }, + { + "epoch": 2.2582757366315023, + "grad_norm": 0.6793645894931464, + "learning_rate": 3.426724020567365e-05, + "loss": 0.1057, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04558804631233215, + "step": 3105, + "valid_targets_mean": 2735.8, + "valid_targets_min": 1692 + }, + { + "epoch": 2.261913423062932, + "grad_norm": 0.3686652078189271, + "learning_rate": 3.424179982795952e-05, + "loss": 0.1402, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05999976024031639, + "step": 3110, + "valid_targets_mean": 3693.4, + "valid_targets_min": 2215 + }, + { + "epoch": 2.2655511094943614, + "grad_norm": 0.3288904285517794, + "learning_rate": 3.4216312615501085e-05, + "loss": 0.0903, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.040833331644535065, + "step": 3115, + "valid_targets_mean": 3109.9, + "valid_targets_min": 723 + }, + { + "epoch": 2.2691887959257913, + "grad_norm": 0.30064474059360735, + "learning_rate": 3.4190778652114076e-05, + "loss": 0.1065, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04496636241674423, + "step": 3120, + "valid_targets_mean": 4984.2, + "valid_targets_min": 3282 + }, + { + "epoch": 2.2728264823572206, + "grad_norm": 0.32155087133154403, + "learning_rate": 3.416519802176799e-05, + "loss": 0.1022, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04240599647164345, + "step": 3125, + "valid_targets_mean": 3863.4, + "valid_targets_min": 995 + }, + { + "epoch": 2.2764641687886504, + "grad_norm": 0.28673008511338594, + "learning_rate": 3.4139570808585775e-05, + "loss": 0.0869, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.041331663727760315, + "step": 3130, + "valid_targets_mean": 4201.0, + "valid_targets_min": 949 + }, + { + "epoch": 2.28010185522008, + "grad_norm": 0.3799547762585434, + "learning_rate": 3.4113897096843585e-05, + "loss": 0.1125, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06887679547071457, + "step": 3135, + "valid_targets_mean": 4321.6, + "valid_targets_min": 2724 + }, + { + "epoch": 2.2837395416515096, + "grad_norm": 0.31793383437234235, + "learning_rate": 3.408817697097047e-05, + "loss": 0.0946, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03347257152199745, + "step": 3140, + "valid_targets_mean": 3384.5, + "valid_targets_min": 424 + }, + { + "epoch": 2.2873772280829394, + "grad_norm": 0.37780448828278534, + "learning_rate": 3.406241051554813e-05, + "loss": 0.1326, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04301265999674797, + "step": 3145, + "valid_targets_mean": 2987.2, + "valid_targets_min": 531 + }, + { + "epoch": 2.291014914514369, + "grad_norm": 0.36627523276530444, + "learning_rate": 3.403659781531059e-05, + "loss": 0.1254, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0419604554772377, + "step": 3150, + "valid_targets_mean": 3469.5, + "valid_targets_min": 1195 + }, + { + "epoch": 2.2946526009457986, + "grad_norm": 0.3494887071864648, + "learning_rate": 3.401073895514401e-05, + "loss": 0.0916, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07906490564346313, + "step": 3155, + "valid_targets_mean": 2682.8, + "valid_targets_min": 722 + }, + { + "epoch": 2.298290287377228, + "grad_norm": 0.32830089514040195, + "learning_rate": 3.398483402008629e-05, + "loss": 0.1256, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04559716582298279, + "step": 3160, + "valid_targets_mean": 3496.9, + "valid_targets_min": 1251 + }, + { + "epoch": 2.3019279738086578, + "grad_norm": 0.3847983779953842, + "learning_rate": 3.395888309532687e-05, + "loss": 0.1017, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05734722688794136, + "step": 3165, + "valid_targets_mean": 2912.5, + "valid_targets_min": 970 + }, + { + "epoch": 2.305565660240087, + "grad_norm": 0.6160948054389984, + "learning_rate": 3.393288626620645e-05, + "loss": 0.1191, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.056754209101200104, + "step": 3170, + "valid_targets_mean": 1423.6, + "valid_targets_min": 634 + }, + { + "epoch": 2.309203346671517, + "grad_norm": 0.5387324806389089, + "learning_rate": 3.3906843618216676e-05, + "loss": 0.1312, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08697140216827393, + "step": 3175, + "valid_targets_mean": 1423.9, + "valid_targets_min": 685 + }, + { + "epoch": 2.3128410331029468, + "grad_norm": 0.367336136030312, + "learning_rate": 3.388075523699985e-05, + "loss": 0.1435, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04181642830371857, + "step": 3180, + "valid_targets_mean": 2932.5, + "valid_targets_min": 593 + }, + { + "epoch": 2.316478719534376, + "grad_norm": 0.31331712749900903, + "learning_rate": 3.3854621208348696e-05, + "loss": 0.1108, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04302429407835007, + "step": 3185, + "valid_targets_mean": 3945.8, + "valid_targets_min": 3129 + }, + { + "epoch": 2.320116405965806, + "grad_norm": 0.4017100942331289, + "learning_rate": 3.3828441618206036e-05, + "loss": 0.1146, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06286880373954773, + "step": 3190, + "valid_targets_mean": 3621.1, + "valid_targets_min": 859 + }, + { + "epoch": 2.3237540923972353, + "grad_norm": 0.443106939348074, + "learning_rate": 3.3802216552664526e-05, + "loss": 0.1036, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05574142560362816, + "step": 3195, + "valid_targets_mean": 2553.0, + "valid_targets_min": 616 + }, + { + "epoch": 2.327391778828665, + "grad_norm": 0.39754449378610746, + "learning_rate": 3.377594609796638e-05, + "loss": 0.1025, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04675203561782837, + "step": 3200, + "valid_targets_mean": 2690.5, + "valid_targets_min": 541 + }, + { + "epoch": 2.3310294652600945, + "grad_norm": 0.6559374197603942, + "learning_rate": 3.374963034050307e-05, + "loss": 0.0989, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06261812150478363, + "step": 3205, + "valid_targets_mean": 906.4, + "valid_targets_min": 598 + }, + { + "epoch": 2.3346671516915243, + "grad_norm": 0.3919958698786588, + "learning_rate": 3.372326936681504e-05, + "loss": 0.1191, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05139993876218796, + "step": 3210, + "valid_targets_mean": 3147.2, + "valid_targets_min": 1101 + }, + { + "epoch": 2.3383048381229536, + "grad_norm": 0.42016807041890075, + "learning_rate": 3.369686326359144e-05, + "loss": 0.0957, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04272108152508736, + "step": 3215, + "valid_targets_mean": 2331.8, + "valid_targets_min": 680 + }, + { + "epoch": 2.3419425245543835, + "grad_norm": 0.3500402856011947, + "learning_rate": 3.3670412117669825e-05, + "loss": 0.1001, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05423419177532196, + "step": 3220, + "valid_targets_mean": 2577.0, + "valid_targets_min": 774 + }, + { + "epoch": 2.345580210985813, + "grad_norm": 0.5814801166329157, + "learning_rate": 3.3643916016035874e-05, + "loss": 0.116, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0598844438791275, + "step": 3225, + "valid_targets_mean": 4140.1, + "valid_targets_min": 1474 + }, + { + "epoch": 2.3492178974172426, + "grad_norm": 0.22676603492948053, + "learning_rate": 3.36173750458231e-05, + "loss": 0.0903, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03204575181007385, + "step": 3230, + "valid_targets_mean": 5498.9, + "valid_targets_min": 3677 + }, + { + "epoch": 2.3528555838486724, + "grad_norm": 0.35586099180450875, + "learning_rate": 3.359078929431259e-05, + "loss": 0.1, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06229211390018463, + "step": 3235, + "valid_targets_mean": 3352.5, + "valid_targets_min": 675 + }, + { + "epoch": 2.356493270280102, + "grad_norm": 0.32461672066665076, + "learning_rate": 3.356415884893268e-05, + "loss": 0.0957, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.038497574627399445, + "step": 3240, + "valid_targets_mean": 3902.5, + "valid_targets_min": 795 + }, + { + "epoch": 2.3601309567115316, + "grad_norm": 0.4895848203574955, + "learning_rate": 3.3537483797258664e-05, + "loss": 0.208, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09083686769008636, + "step": 3245, + "valid_targets_mean": 2495.0, + "valid_targets_min": 605 + }, + { + "epoch": 2.363768643142961, + "grad_norm": 0.32671070379967243, + "learning_rate": 3.3510764227012574e-05, + "loss": 0.0952, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04977714642882347, + "step": 3250, + "valid_targets_mean": 3298.6, + "valid_targets_min": 985 + }, + { + "epoch": 2.367406329574391, + "grad_norm": 0.38202313751994293, + "learning_rate": 3.3484000226062794e-05, + "loss": 0.1096, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05664435029029846, + "step": 3255, + "valid_targets_mean": 3505.2, + "valid_targets_min": 878 + }, + { + "epoch": 2.37104401600582, + "grad_norm": 0.5809006989264324, + "learning_rate": 3.345719188242385e-05, + "loss": 0.1316, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13489645719528198, + "step": 3260, + "valid_targets_mean": 2289.5, + "valid_targets_min": 1214 + }, + { + "epoch": 2.37468170243725, + "grad_norm": 0.3892548073353111, + "learning_rate": 3.343033928425608e-05, + "loss": 0.2042, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13708138465881348, + "step": 3265, + "valid_targets_mean": 8496.5, + "valid_targets_min": 6565 + }, + { + "epoch": 2.3783193888686793, + "grad_norm": 0.4157579197200859, + "learning_rate": 3.340344251986537e-05, + "loss": 0.2189, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09719507396221161, + "step": 3270, + "valid_targets_mean": 5663.5, + "valid_targets_min": 3833 + }, + { + "epoch": 2.381957075300109, + "grad_norm": 0.5472286652980553, + "learning_rate": 3.337650167770281e-05, + "loss": 0.2127, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10218753665685654, + "step": 3275, + "valid_targets_mean": 6654.1, + "valid_targets_min": 5324 + }, + { + "epoch": 2.3855947617315385, + "grad_norm": 0.3683728522432601, + "learning_rate": 3.3349516846364496e-05, + "loss": 0.1999, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09909018874168396, + "step": 3280, + "valid_targets_mean": 7635.1, + "valid_targets_min": 5672 + }, + { + "epoch": 2.3892324481629683, + "grad_norm": 0.3240162609659728, + "learning_rate": 3.332248811459113e-05, + "loss": 0.2061, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09315121173858643, + "step": 3285, + "valid_targets_mean": 8517.0, + "valid_targets_min": 5045 + }, + { + "epoch": 2.392870134594398, + "grad_norm": 0.3855436396146591, + "learning_rate": 3.329541557126782e-05, + "loss": 0.2094, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10226185619831085, + "step": 3290, + "valid_targets_mean": 6759.1, + "valid_targets_min": 3152 + }, + { + "epoch": 2.3965078210258275, + "grad_norm": 0.36467200883821904, + "learning_rate": 3.326829930542374e-05, + "loss": 0.2173, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09596486389636993, + "step": 3295, + "valid_targets_mean": 7822.8, + "valid_targets_min": 4820 + }, + { + "epoch": 2.4001455074572573, + "grad_norm": 0.350294089943631, + "learning_rate": 3.324113940623185e-05, + "loss": 0.2057, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08340615779161453, + "step": 3300, + "valid_targets_mean": 6852.2, + "valid_targets_min": 4135 + }, + { + "epoch": 2.4037831938886867, + "grad_norm": 0.40276674192928547, + "learning_rate": 3.3213935963008576e-05, + "loss": 0.203, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09053843468427658, + "step": 3305, + "valid_targets_mean": 4794.2, + "valid_targets_min": 3230 + }, + { + "epoch": 2.4074208803201165, + "grad_norm": 0.3790952773712523, + "learning_rate": 3.318668906521358e-05, + "loss": 0.1869, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10481509566307068, + "step": 3310, + "valid_targets_mean": 7691.0, + "valid_targets_min": 5900 + }, + { + "epoch": 2.411058566751546, + "grad_norm": 0.3860326747770965, + "learning_rate": 3.315939880244939e-05, + "loss": 0.2045, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09641649574041367, + "step": 3315, + "valid_targets_mean": 6510.9, + "valid_targets_min": 5639 + }, + { + "epoch": 2.4146962531829756, + "grad_norm": 0.482594052193025, + "learning_rate": 3.3132065264461154e-05, + "loss": 0.1929, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09887497127056122, + "step": 3320, + "valid_targets_mean": 6096.8, + "valid_targets_min": 4669 + }, + { + "epoch": 2.4183339396144055, + "grad_norm": 0.39100732400948796, + "learning_rate": 3.310468854113636e-05, + "loss": 0.2002, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10299715399742126, + "step": 3325, + "valid_targets_mean": 6812.2, + "valid_targets_min": 5160 + }, + { + "epoch": 2.421971626045835, + "grad_norm": 0.34243086971596, + "learning_rate": 3.307726872250447e-05, + "loss": 0.1915, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08936551958322525, + "step": 3330, + "valid_targets_mean": 6980.4, + "valid_targets_min": 5151 + }, + { + "epoch": 2.4256093124772646, + "grad_norm": 0.3857770446354423, + "learning_rate": 3.304980589873669e-05, + "loss": 0.1951, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10521677136421204, + "step": 3335, + "valid_targets_mean": 7417.1, + "valid_targets_min": 5932 + }, + { + "epoch": 2.429246998908694, + "grad_norm": 0.38439573165748214, + "learning_rate": 3.302230016014564e-05, + "loss": 0.2007, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09663460403680801, + "step": 3340, + "valid_targets_mean": 6258.5, + "valid_targets_min": 5569 + }, + { + "epoch": 2.432884685340124, + "grad_norm": 0.37286396934913496, + "learning_rate": 3.299475159718508e-05, + "loss": 0.1935, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1097746267914772, + "step": 3345, + "valid_targets_mean": 7689.1, + "valid_targets_min": 6223 + }, + { + "epoch": 2.436522371771553, + "grad_norm": 0.5110613646584101, + "learning_rate": 3.296716030044958e-05, + "loss": 0.2044, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11882121860980988, + "step": 3350, + "valid_targets_mean": 4371.9, + "valid_targets_min": 1172 + }, + { + "epoch": 2.440160058202983, + "grad_norm": 0.5768193659137435, + "learning_rate": 3.293952636067426e-05, + "loss": 0.2138, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11187891662120819, + "step": 3355, + "valid_targets_mean": 6936.2, + "valid_targets_min": 5619 + }, + { + "epoch": 2.4437977446344123, + "grad_norm": 0.6251777213468406, + "learning_rate": 3.2911849868734487e-05, + "loss": 0.1924, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13742980360984802, + "step": 3360, + "valid_targets_mean": 3273.2, + "valid_targets_min": 442 + }, + { + "epoch": 2.447435431065842, + "grad_norm": 0.3545978699999081, + "learning_rate": 3.288413091564551e-05, + "loss": 0.1843, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09604652225971222, + "step": 3365, + "valid_targets_mean": 8141.1, + "valid_targets_min": 5061 + }, + { + "epoch": 2.4510731174972715, + "grad_norm": 0.38633173888563, + "learning_rate": 3.285636959256227e-05, + "loss": 0.1947, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0976007729768753, + "step": 3370, + "valid_targets_mean": 7371.1, + "valid_targets_min": 5112 + }, + { + "epoch": 2.4547108039287013, + "grad_norm": 0.3700387318315853, + "learning_rate": 3.2828565990779e-05, + "loss": 0.1939, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08562079817056656, + "step": 3375, + "valid_targets_mean": 5659.4, + "valid_targets_min": 4626 + }, + { + "epoch": 2.458348490360131, + "grad_norm": 0.34094506456604434, + "learning_rate": 3.280072020172901e-05, + "loss": 0.2009, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08826494216918945, + "step": 3380, + "valid_targets_mean": 7801.8, + "valid_targets_min": 6172 + }, + { + "epoch": 2.4619861767915605, + "grad_norm": 0.36804129248113576, + "learning_rate": 3.277283231698429e-05, + "loss": 0.1927, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09703350812196732, + "step": 3385, + "valid_targets_mean": 7547.4, + "valid_targets_min": 5804 + }, + { + "epoch": 2.4656238632229903, + "grad_norm": 0.41163414064698906, + "learning_rate": 3.2744902428255336e-05, + "loss": 0.1826, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08697488158941269, + "step": 3390, + "valid_targets_mean": 6474.2, + "valid_targets_min": 5050 + }, + { + "epoch": 2.4692615496544197, + "grad_norm": 0.35382157484830357, + "learning_rate": 3.2716930627390704e-05, + "loss": 0.17, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08925110101699829, + "step": 3395, + "valid_targets_mean": 7225.4, + "valid_targets_min": 5298 + }, + { + "epoch": 2.4728992360858495, + "grad_norm": 0.37983856603151583, + "learning_rate": 3.268891700637683e-05, + "loss": 0.1708, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08025212585926056, + "step": 3400, + "valid_targets_mean": 6442.9, + "valid_targets_min": 4679 + }, + { + "epoch": 2.476536922517279, + "grad_norm": 0.4680715483679767, + "learning_rate": 3.2660861657337646e-05, + "loss": 0.2028, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08207740634679794, + "step": 3405, + "valid_targets_mean": 3651.9, + "valid_targets_min": 1801 + }, + { + "epoch": 2.4801746089487087, + "grad_norm": 0.42519569746788316, + "learning_rate": 3.263276467253432e-05, + "loss": 0.1977, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11332453042268753, + "step": 3410, + "valid_targets_mean": 7512.8, + "valid_targets_min": 5331 + }, + { + "epoch": 2.4838122953801385, + "grad_norm": 0.4039073727920973, + "learning_rate": 3.2604626144364955e-05, + "loss": 0.2022, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10546555370092392, + "step": 3415, + "valid_targets_mean": 7108.8, + "valid_targets_min": 4822 + }, + { + "epoch": 2.487449981811568, + "grad_norm": 0.37550084513501947, + "learning_rate": 3.257644616536425e-05, + "loss": 0.2078, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09174621105194092, + "step": 3420, + "valid_targets_mean": 7125.8, + "valid_targets_min": 4623 + }, + { + "epoch": 2.491087668242997, + "grad_norm": 0.3720051435809319, + "learning_rate": 3.254822482820322e-05, + "loss": 0.2037, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09186109900474548, + "step": 3425, + "valid_targets_mean": 6269.2, + "valid_targets_min": 5458 + }, + { + "epoch": 2.494725354674427, + "grad_norm": 0.3901355274317811, + "learning_rate": 3.251996222568891e-05, + "loss": 0.1941, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0994078665971756, + "step": 3430, + "valid_targets_mean": 7455.1, + "valid_targets_min": 6008 + }, + { + "epoch": 2.498363041105857, + "grad_norm": 0.39703585776935474, + "learning_rate": 3.2491658450764024e-05, + "loss": 0.203, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1002398431301117, + "step": 3435, + "valid_targets_mean": 7104.4, + "valid_targets_min": 5386 + }, + { + "epoch": 2.502000727537286, + "grad_norm": 0.3591710149581235, + "learning_rate": 3.246331359650671e-05, + "loss": 0.1943, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10005123913288116, + "step": 3440, + "valid_targets_mean": 6532.6, + "valid_targets_min": 4855 + }, + { + "epoch": 2.505638413968716, + "grad_norm": 0.38497760290366395, + "learning_rate": 3.243492775613018e-05, + "loss": 0.2002, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0973723828792572, + "step": 3445, + "valid_targets_mean": 6743.9, + "valid_targets_min": 5613 + }, + { + "epoch": 2.5092761004001454, + "grad_norm": 0.39826715591883977, + "learning_rate": 3.240650102298244e-05, + "loss": 0.1989, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1001736968755722, + "step": 3450, + "valid_targets_mean": 6373.6, + "valid_targets_min": 4961 + }, + { + "epoch": 2.512913786831575, + "grad_norm": 0.39061216603308635, + "learning_rate": 3.237803349054596e-05, + "loss": 0.2027, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1046798899769783, + "step": 3455, + "valid_targets_mean": 6461.2, + "valid_targets_min": 5286 + }, + { + "epoch": 2.5165514732630045, + "grad_norm": 0.37880436756312397, + "learning_rate": 3.234952525243741e-05, + "loss": 0.1985, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10456463694572449, + "step": 3460, + "valid_targets_mean": 6780.5, + "valid_targets_min": 5201 + }, + { + "epoch": 2.5201891596944344, + "grad_norm": 0.41194772333977125, + "learning_rate": 3.2320976402407286e-05, + "loss": 0.1978, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09901571273803711, + "step": 3465, + "valid_targets_mean": 5710.8, + "valid_targets_min": 4838 + }, + { + "epoch": 2.523826846125864, + "grad_norm": 0.6570207459245024, + "learning_rate": 3.229238703433966e-05, + "loss": 0.1722, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15209351480007172, + "step": 3470, + "valid_targets_mean": 3235.6, + "valid_targets_min": 176 + }, + { + "epoch": 2.5274645325572935, + "grad_norm": 0.41028658723668604, + "learning_rate": 3.2263757242251846e-05, + "loss": 0.199, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09187071025371552, + "step": 3475, + "valid_targets_mean": 6674.8, + "valid_targets_min": 4582 + }, + { + "epoch": 2.531102218988723, + "grad_norm": 0.4011527121623326, + "learning_rate": 3.2235087120294105e-05, + "loss": 0.1933, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09647966921329498, + "step": 3480, + "valid_targets_mean": 6157.8, + "valid_targets_min": 5385 + }, + { + "epoch": 2.5347399054201527, + "grad_norm": 0.409071675914537, + "learning_rate": 3.2206376762749284e-05, + "loss": 0.1966, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09314088523387909, + "step": 3485, + "valid_targets_mean": 6006.2, + "valid_targets_min": 4478 + }, + { + "epoch": 2.5383775918515825, + "grad_norm": 0.4308197508305145, + "learning_rate": 3.21776262640326e-05, + "loss": 0.1925, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09620235115289688, + "step": 3490, + "valid_targets_mean": 5893.4, + "valid_targets_min": 3911 + }, + { + "epoch": 2.542015278283012, + "grad_norm": 0.4182218961589392, + "learning_rate": 3.2148835718691235e-05, + "loss": 0.192, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09501022845506668, + "step": 3495, + "valid_targets_mean": 6134.2, + "valid_targets_min": 5318 + }, + { + "epoch": 2.5456529647144417, + "grad_norm": 0.4037376244318148, + "learning_rate": 3.212000522140408e-05, + "loss": 0.2012, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10076956450939178, + "step": 3500, + "valid_targets_mean": 5474.4, + "valid_targets_min": 4959 + }, + { + "epoch": 2.5492906511458715, + "grad_norm": 0.5507240736751244, + "learning_rate": 3.209113486698143e-05, + "loss": 0.1939, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06274239718914032, + "step": 3505, + "valid_targets_mean": 1804.1, + "valid_targets_min": 842 + }, + { + "epoch": 2.552928337577301, + "grad_norm": 0.4448698565321103, + "learning_rate": 3.206222475036461e-05, + "loss": 0.1687, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08868709206581116, + "step": 3510, + "valid_targets_mean": 6118.8, + "valid_targets_min": 4899 + }, + { + "epoch": 2.5565660240087302, + "grad_norm": 0.3804590902540065, + "learning_rate": 3.203327496662574e-05, + "loss": 0.1893, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08951560407876968, + "step": 3515, + "valid_targets_mean": 6773.8, + "valid_targets_min": 5046 + }, + { + "epoch": 2.56020371044016, + "grad_norm": 0.4122126221346079, + "learning_rate": 3.200428561096737e-05, + "loss": 0.2006, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10286575555801392, + "step": 3520, + "valid_targets_mean": 6633.6, + "valid_targets_min": 5284 + }, + { + "epoch": 2.56384139687159, + "grad_norm": 0.40724496414861533, + "learning_rate": 3.197525677872219e-05, + "loss": 0.1917, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09742819517850876, + "step": 3525, + "valid_targets_mean": 6055.2, + "valid_targets_min": 4928 + }, + { + "epoch": 2.567479083303019, + "grad_norm": 0.3866253278633049, + "learning_rate": 3.19461885653527e-05, + "loss": 0.1873, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10094960778951645, + "step": 3530, + "valid_targets_mean": 7259.4, + "valid_targets_min": 5758 + }, + { + "epoch": 2.571116769734449, + "grad_norm": 0.4229160766136477, + "learning_rate": 3.191708106645092e-05, + "loss": 0.1914, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09919090569019318, + "step": 3535, + "valid_targets_mean": 5731.8, + "valid_targets_min": 4629 + }, + { + "epoch": 2.5747544561658784, + "grad_norm": 0.7464748039718809, + "learning_rate": 3.188793437773807e-05, + "loss": 0.2209, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21008995175361633, + "step": 3540, + "valid_targets_mean": 5360.9, + "valid_targets_min": 2317 + }, + { + "epoch": 2.578392142597308, + "grad_norm": 0.5146559740100234, + "learning_rate": 3.1858748595064225e-05, + "loss": 0.2787, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15214799344539642, + "step": 3545, + "valid_targets_mean": 5303.1, + "valid_targets_min": 533 + }, + { + "epoch": 2.5820298290287376, + "grad_norm": 0.5473989779006928, + "learning_rate": 3.1829523814408024e-05, + "loss": 0.2729, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14228889346122742, + "step": 3550, + "valid_targets_mean": 4545.8, + "valid_targets_min": 835 + }, + { + "epoch": 2.5856675154601674, + "grad_norm": 0.6154152869609658, + "learning_rate": 3.180026013187636e-05, + "loss": 0.269, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1482471227645874, + "step": 3555, + "valid_targets_mean": 3618.9, + "valid_targets_min": 1201 + }, + { + "epoch": 2.589305201891597, + "grad_norm": 0.5412232164787071, + "learning_rate": 3.177095764370407e-05, + "loss": 0.2562, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12054409086704254, + "step": 3560, + "valid_targets_mean": 4253.5, + "valid_targets_min": 1880 + }, + { + "epoch": 2.5929428883230266, + "grad_norm": 0.5002995233831748, + "learning_rate": 3.1741616446253574e-05, + "loss": 0.283, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19970285892486572, + "step": 3565, + "valid_targets_mean": 6610.1, + "valid_targets_min": 2171 + }, + { + "epoch": 2.596580574754456, + "grad_norm": 0.47078382667139346, + "learning_rate": 3.171223663601463e-05, + "loss": 0.2592, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14655911922454834, + "step": 3570, + "valid_targets_mean": 7171.4, + "valid_targets_min": 4500 + }, + { + "epoch": 2.6002182611858857, + "grad_norm": 0.450469960509592, + "learning_rate": 3.1682818309603915e-05, + "loss": 0.2662, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13849776983261108, + "step": 3575, + "valid_targets_mean": 6207.4, + "valid_targets_min": 1855 + }, + { + "epoch": 2.6038559476173155, + "grad_norm": 0.5835136970113094, + "learning_rate": 3.1653361563764846e-05, + "loss": 0.2652, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1379612386226654, + "step": 3580, + "valid_targets_mean": 4075.9, + "valid_targets_min": 1024 + }, + { + "epoch": 2.607493634048745, + "grad_norm": 0.49465725052391746, + "learning_rate": 3.1623866495367115e-05, + "loss": 0.2744, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11742550879716873, + "step": 3585, + "valid_targets_mean": 4767.1, + "valid_targets_min": 2107 + }, + { + "epoch": 2.6111313204801747, + "grad_norm": 0.5602290740560552, + "learning_rate": 3.159433320140648e-05, + "loss": 0.2725, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1518881916999817, + "step": 3590, + "valid_targets_mean": 4211.4, + "valid_targets_min": 802 + }, + { + "epoch": 2.614769006911604, + "grad_norm": 0.4515256860731421, + "learning_rate": 3.156476177900438e-05, + "loss": 0.2599, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1362452358007431, + "step": 3595, + "valid_targets_mean": 5345.1, + "valid_targets_min": 1597 + }, + { + "epoch": 2.618406693343034, + "grad_norm": 0.5926779925023133, + "learning_rate": 3.153515232540767e-05, + "loss": 0.2749, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1411522775888443, + "step": 3600, + "valid_targets_mean": 3671.4, + "valid_targets_min": 1650 + }, + { + "epoch": 2.6220443797744633, + "grad_norm": 0.5203705391832635, + "learning_rate": 3.150550493798824e-05, + "loss": 0.2643, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14013808965682983, + "step": 3605, + "valid_targets_mean": 4799.1, + "valid_targets_min": 1385 + }, + { + "epoch": 2.625682066205893, + "grad_norm": 0.5983532634033676, + "learning_rate": 3.147581971424276e-05, + "loss": 0.2643, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1456187665462494, + "step": 3610, + "valid_targets_mean": 4023.5, + "valid_targets_min": 2908 + }, + { + "epoch": 2.629319752637323, + "grad_norm": 0.5219496611958448, + "learning_rate": 3.144609675179229e-05, + "loss": 0.2729, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12335389107465744, + "step": 3615, + "valid_targets_mean": 5064.6, + "valid_targets_min": 2785 + }, + { + "epoch": 2.6329574390687522, + "grad_norm": 0.767524288291868, + "learning_rate": 3.141633614838203e-05, + "loss": 0.2616, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12773944437503815, + "step": 3620, + "valid_targets_mean": 3886.4, + "valid_targets_min": 1210 + }, + { + "epoch": 2.636595125500182, + "grad_norm": 0.6013743230966719, + "learning_rate": 3.1386538001880934e-05, + "loss": 0.2714, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1508198082447052, + "step": 3625, + "valid_targets_mean": 3708.8, + "valid_targets_min": 1064 + }, + { + "epoch": 2.6402328119316114, + "grad_norm": 0.7654802246542894, + "learning_rate": 3.135670241028145e-05, + "loss": 0.2813, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14045880734920502, + "step": 3630, + "valid_targets_mean": 3276.2, + "valid_targets_min": 1678 + }, + { + "epoch": 2.643870498363041, + "grad_norm": 0.5730779650432167, + "learning_rate": 3.132682947169913e-05, + "loss": 0.2744, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.143113374710083, + "step": 3635, + "valid_targets_mean": 3520.1, + "valid_targets_min": 671 + }, + { + "epoch": 2.6475081847944706, + "grad_norm": 0.7894258962392221, + "learning_rate": 3.1296919284372366e-05, + "loss": 0.267, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1275748312473297, + "step": 3640, + "valid_targets_mean": 3488.5, + "valid_targets_min": 2000 + }, + { + "epoch": 2.6511458712259004, + "grad_norm": 0.5989467856982369, + "learning_rate": 3.1266971946662045e-05, + "loss": 0.255, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15826180577278137, + "step": 3645, + "valid_targets_mean": 4402.6, + "valid_targets_min": 1532 + }, + { + "epoch": 2.65478355765733, + "grad_norm": 0.5561475975512531, + "learning_rate": 3.123698755705121e-05, + "loss": 0.2593, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10212226957082748, + "step": 3650, + "valid_targets_mean": 3749.4, + "valid_targets_min": 1051 + }, + { + "epoch": 2.6584212440887596, + "grad_norm": 0.573303806632719, + "learning_rate": 3.1206966214144764e-05, + "loss": 0.2521, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12638968229293823, + "step": 3655, + "valid_targets_mean": 3368.1, + "valid_targets_min": 1345 + }, + { + "epoch": 2.662058930520189, + "grad_norm": 0.6142761803711501, + "learning_rate": 3.1176908016669125e-05, + "loss": 0.2701, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12860994040966034, + "step": 3660, + "valid_targets_mean": 4128.9, + "valid_targets_min": 2231 + }, + { + "epoch": 2.6656966169516187, + "grad_norm": 0.5954187203821869, + "learning_rate": 3.114681306347191e-05, + "loss": 0.2716, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11597830057144165, + "step": 3665, + "valid_targets_mean": 3672.6, + "valid_targets_min": 2127 + }, + { + "epoch": 2.6693343033830486, + "grad_norm": 0.6983419478600699, + "learning_rate": 3.111668145352163e-05, + "loss": 0.2572, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11871923506259918, + "step": 3670, + "valid_targets_mean": 3006.8, + "valid_targets_min": 1265 + }, + { + "epoch": 2.672971989814478, + "grad_norm": 0.6012291230829881, + "learning_rate": 3.108651328590728e-05, + "loss": 0.2595, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1137273982167244, + "step": 3675, + "valid_targets_mean": 2981.9, + "valid_targets_min": 908 + }, + { + "epoch": 2.6766096762459077, + "grad_norm": 0.6742498742851909, + "learning_rate": 3.105630865983816e-05, + "loss": 0.2593, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1480099856853485, + "step": 3680, + "valid_targets_mean": 3147.5, + "valid_targets_min": 1264 + }, + { + "epoch": 2.680247362677337, + "grad_norm": 0.6078532035060964, + "learning_rate": 3.102606767464341e-05, + "loss": 0.2593, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11976810544729233, + "step": 3685, + "valid_targets_mean": 2517.9, + "valid_targets_min": 926 + }, + { + "epoch": 2.683885049108767, + "grad_norm": 0.6589916974796366, + "learning_rate": 3.0995790429771756e-05, + "loss": 0.2627, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12835237383842468, + "step": 3690, + "valid_targets_mean": 3204.5, + "valid_targets_min": 1016 + }, + { + "epoch": 2.6875227355401963, + "grad_norm": 0.6591258140392785, + "learning_rate": 3.0965477024791154e-05, + "loss": 0.2529, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12616384029388428, + "step": 3695, + "valid_targets_mean": 3525.0, + "valid_targets_min": 2019 + }, + { + "epoch": 2.691160421971626, + "grad_norm": 0.5840760950845464, + "learning_rate": 3.093512755938849e-05, + "loss": 0.2619, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1528211236000061, + "step": 3700, + "valid_targets_mean": 4745.4, + "valid_targets_min": 3023 + }, + { + "epoch": 2.694798108403056, + "grad_norm": 0.6025263201369219, + "learning_rate": 3.090474213336923e-05, + "loss": 0.254, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1161726787686348, + "step": 3705, + "valid_targets_mean": 3503.6, + "valid_targets_min": 1110 + }, + { + "epoch": 2.6984357948344853, + "grad_norm": 0.6140152451914113, + "learning_rate": 3.087432084665709e-05, + "loss": 0.2597, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15257856249809265, + "step": 3710, + "valid_targets_mean": 4206.5, + "valid_targets_min": 1460 + }, + { + "epoch": 2.7020734812659146, + "grad_norm": 0.5597374766875063, + "learning_rate": 3.084386379929373e-05, + "loss": 0.2578, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1400504857301712, + "step": 3715, + "valid_targets_mean": 3938.6, + "valid_targets_min": 2558 + }, + { + "epoch": 2.7057111676973444, + "grad_norm": 0.5645384854413896, + "learning_rate": 3.0813371091438406e-05, + "loss": 0.2682, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12094873189926147, + "step": 3720, + "valid_targets_mean": 3421.1, + "valid_targets_min": 1673 + }, + { + "epoch": 2.7093488541287742, + "grad_norm": 0.5709923108558784, + "learning_rate": 3.0782842823367634e-05, + "loss": 0.2567, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12012352794408798, + "step": 3725, + "valid_targets_mean": 3549.6, + "valid_targets_min": 1860 + }, + { + "epoch": 2.7129865405602036, + "grad_norm": 0.5091379334827317, + "learning_rate": 3.0752279095474867e-05, + "loss": 0.257, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11913782358169556, + "step": 3730, + "valid_targets_mean": 4414.0, + "valid_targets_min": 2495 + }, + { + "epoch": 2.7166242269916334, + "grad_norm": 0.608377934530034, + "learning_rate": 3.0721680008270206e-05, + "loss": 0.2558, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14623546600341797, + "step": 3735, + "valid_targets_mean": 3861.9, + "valid_targets_min": 1945 + }, + { + "epoch": 2.720261913423063, + "grad_norm": 0.579851620870342, + "learning_rate": 3.0691045662379986e-05, + "loss": 0.2663, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13145741820335388, + "step": 3740, + "valid_targets_mean": 4488.5, + "valid_targets_min": 1908 + }, + { + "epoch": 2.7238995998544926, + "grad_norm": 0.5570030055185491, + "learning_rate": 3.0660376158546526e-05, + "loss": 0.2582, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13337917625904083, + "step": 3745, + "valid_targets_mean": 3507.5, + "valid_targets_min": 1880 + }, + { + "epoch": 2.727537286285922, + "grad_norm": 0.5656205441195851, + "learning_rate": 3.0629671597627745e-05, + "loss": 0.2539, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1298210471868515, + "step": 3750, + "valid_targets_mean": 4491.5, + "valid_targets_min": 2492 + }, + { + "epoch": 2.7311749727173518, + "grad_norm": 0.556409024489974, + "learning_rate": 3.059893208059686e-05, + "loss": 0.2565, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12925368547439575, + "step": 3755, + "valid_targets_mean": 4228.2, + "valid_targets_min": 3004 + }, + { + "epoch": 2.7348126591487816, + "grad_norm": 0.5847909593221858, + "learning_rate": 3.056815770854203e-05, + "loss": 0.2504, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12889792025089264, + "step": 3760, + "valid_targets_mean": 3899.9, + "valid_targets_min": 1833 + }, + { + "epoch": 2.738450345580211, + "grad_norm": 0.6210214552734158, + "learning_rate": 3.0537348582666044e-05, + "loss": 0.2557, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13237982988357544, + "step": 3765, + "valid_targets_mean": 3114.2, + "valid_targets_min": 1182 + }, + { + "epoch": 2.7420880320116408, + "grad_norm": 0.520286375988191, + "learning_rate": 3.0506504804285977e-05, + "loss": 0.2558, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1275361180305481, + "step": 3770, + "valid_targets_mean": 4926.9, + "valid_targets_min": 1869 + }, + { + "epoch": 2.74572571844307, + "grad_norm": 0.6011879368634894, + "learning_rate": 3.0475626474832868e-05, + "loss": 0.2578, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15222689509391785, + "step": 3775, + "valid_targets_mean": 4382.4, + "valid_targets_min": 2498 + }, + { + "epoch": 2.7493634048745, + "grad_norm": 0.612325344187649, + "learning_rate": 3.0444713695851375e-05, + "loss": 0.2737, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1281633973121643, + "step": 3780, + "valid_targets_mean": 3098.8, + "valid_targets_min": 988 + }, + { + "epoch": 2.7530010913059293, + "grad_norm": 0.6214703276508315, + "learning_rate": 3.041376656899943e-05, + "loss": 0.252, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1321743130683899, + "step": 3785, + "valid_targets_mean": 3733.5, + "valid_targets_min": 1534 + }, + { + "epoch": 2.756638777737359, + "grad_norm": 0.5390450846696037, + "learning_rate": 3.038278519604795e-05, + "loss": 0.2587, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1150560975074768, + "step": 3790, + "valid_targets_mean": 4108.1, + "valid_targets_min": 2252 + }, + { + "epoch": 2.760276464168789, + "grad_norm": 0.6991861566138452, + "learning_rate": 3.035176967888044e-05, + "loss": 0.2612, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13178548216819763, + "step": 3795, + "valid_targets_mean": 3446.6, + "valid_targets_min": 1436 + }, + { + "epoch": 2.7639141506002183, + "grad_norm": 0.5813838061438311, + "learning_rate": 3.0320720119492707e-05, + "loss": 0.2588, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11965377628803253, + "step": 3800, + "valid_targets_mean": 3819.2, + "valid_targets_min": 2333 + }, + { + "epoch": 2.7675518370316476, + "grad_norm": 0.5820419855708779, + "learning_rate": 3.0289636619992505e-05, + "loss": 0.2496, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11419196426868439, + "step": 3805, + "valid_targets_mean": 3889.8, + "valid_targets_min": 1272 + }, + { + "epoch": 2.7711895234630775, + "grad_norm": 0.5621078206894301, + "learning_rate": 3.02585192825992e-05, + "loss": 0.2489, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12710028886795044, + "step": 3810, + "valid_targets_mean": 3233.0, + "valid_targets_min": 1319 + }, + { + "epoch": 2.7748272098945073, + "grad_norm": 0.5333921351286529, + "learning_rate": 3.0227368209643437e-05, + "loss": 0.2346, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10305584967136383, + "step": 3815, + "valid_targets_mean": 3186.6, + "valid_targets_min": 1194 + }, + { + "epoch": 2.7784648963259366, + "grad_norm": 0.6065180915635606, + "learning_rate": 3.0196183503566794e-05, + "loss": 0.2721, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11296145617961884, + "step": 3820, + "valid_targets_mean": 2710.6, + "valid_targets_min": 1453 + }, + { + "epoch": 2.7821025827573664, + "grad_norm": 0.4822542983016036, + "learning_rate": 3.0164965266921462e-05, + "loss": 0.2486, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10369212925434113, + "step": 3825, + "valid_targets_mean": 5344.5, + "valid_targets_min": 3603 + }, + { + "epoch": 2.785740269188796, + "grad_norm": 0.5620612961248432, + "learning_rate": 3.01337136023699e-05, + "loss": 0.2639, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11515491455793381, + "step": 3830, + "valid_targets_mean": 3067.5, + "valid_targets_min": 1256 + }, + { + "epoch": 2.7893779556202256, + "grad_norm": 0.648450454997233, + "learning_rate": 3.0102428612684486e-05, + "loss": 0.2582, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17175883054733276, + "step": 3835, + "valid_targets_mean": 4443.9, + "valid_targets_min": 1336 + }, + { + "epoch": 2.793015642051655, + "grad_norm": 0.6857976647562561, + "learning_rate": 3.0071110400747206e-05, + "loss": 0.2664, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1323315054178238, + "step": 3840, + "valid_targets_mean": 3581.8, + "valid_targets_min": 1275 + }, + { + "epoch": 2.796653328483085, + "grad_norm": 0.6201774621906035, + "learning_rate": 3.003975906954928e-05, + "loss": 0.2548, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1445835828781128, + "step": 3845, + "valid_targets_mean": 3770.1, + "valid_targets_min": 2202 + }, + { + "epoch": 2.8002910149145146, + "grad_norm": 0.5980420777205645, + "learning_rate": 3.0008374722190843e-05, + "loss": 0.2496, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14765512943267822, + "step": 3850, + "valid_targets_mean": 4112.8, + "valid_targets_min": 3332 + }, + { + "epoch": 2.803928701345944, + "grad_norm": 0.6146780984239388, + "learning_rate": 2.9976957461880615e-05, + "loss": 0.2487, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13182848691940308, + "step": 3855, + "valid_targets_mean": 3260.0, + "valid_targets_min": 2195 + }, + { + "epoch": 2.8075663877773733, + "grad_norm": 0.5738562828932966, + "learning_rate": 2.9945507391935557e-05, + "loss": 0.2481, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1248004361987114, + "step": 3860, + "valid_targets_mean": 4039.6, + "valid_targets_min": 2343 + }, + { + "epoch": 2.811204074208803, + "grad_norm": 0.5743057452880482, + "learning_rate": 2.9914024615780516e-05, + "loss": 0.243, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11202243715524673, + "step": 3865, + "valid_targets_mean": 2678.8, + "valid_targets_min": 1026 + }, + { + "epoch": 2.814841760640233, + "grad_norm": 0.6121492610333903, + "learning_rate": 2.9882509236947887e-05, + "loss": 0.2453, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12265359610319138, + "step": 3870, + "valid_targets_mean": 3754.5, + "valid_targets_min": 1042 + }, + { + "epoch": 2.8184794470716623, + "grad_norm": 0.6117638378717719, + "learning_rate": 2.9850961359077293e-05, + "loss": 0.2334, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11896864324808121, + "step": 3875, + "valid_targets_mean": 3301.6, + "valid_targets_min": 1274 + }, + { + "epoch": 2.822117133503092, + "grad_norm": 0.9238150051048771, + "learning_rate": 2.9819381085915225e-05, + "loss": 0.2495, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12131185829639435, + "step": 3880, + "valid_targets_mean": 3632.2, + "valid_targets_min": 914 + }, + { + "epoch": 2.8257548199345215, + "grad_norm": 0.5373603562964315, + "learning_rate": 2.9787768521314717e-05, + "loss": 0.2747, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11418972909450531, + "step": 3885, + "valid_targets_mean": 4342.5, + "valid_targets_min": 2093 + }, + { + "epoch": 2.8293925063659513, + "grad_norm": 0.6330877618504857, + "learning_rate": 2.9756123769234978e-05, + "loss": 0.2535, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13179397583007812, + "step": 3890, + "valid_targets_mean": 3609.6, + "valid_targets_min": 1683 + }, + { + "epoch": 2.8330301927973807, + "grad_norm": 0.5261216759536143, + "learning_rate": 2.972444693374109e-05, + "loss": 0.2602, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12093304097652435, + "step": 3895, + "valid_targets_mean": 4096.2, + "valid_targets_min": 2253 + }, + { + "epoch": 2.8366678792288105, + "grad_norm": 0.5830020749156832, + "learning_rate": 2.9692738119003618e-05, + "loss": 0.2515, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11777321994304657, + "step": 3900, + "valid_targets_mean": 3794.4, + "valid_targets_min": 1789 + }, + { + "epoch": 2.8403055656602403, + "grad_norm": 0.6483066611599678, + "learning_rate": 2.9660997429298305e-05, + "loss": 0.2554, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12602590024471283, + "step": 3905, + "valid_targets_mean": 3753.8, + "valid_targets_min": 1586 + }, + { + "epoch": 2.8439432520916696, + "grad_norm": 0.6099131743173286, + "learning_rate": 2.962922496900572e-05, + "loss": 0.2512, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11322356760501862, + "step": 3910, + "valid_targets_mean": 2995.5, + "valid_targets_min": 2033 + }, + { + "epoch": 2.8475809385230995, + "grad_norm": 0.5366620550580645, + "learning_rate": 2.959742084261089e-05, + "loss": 0.2391, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11457711458206177, + "step": 3915, + "valid_targets_mean": 3690.5, + "valid_targets_min": 1616 + }, + { + "epoch": 2.851218624954529, + "grad_norm": 0.6137011160848977, + "learning_rate": 2.9565585154703008e-05, + "loss": 0.2554, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11043548583984375, + "step": 3920, + "valid_targets_mean": 3303.9, + "valid_targets_min": 2000 + }, + { + "epoch": 2.8548563113859586, + "grad_norm": 0.5914198358682721, + "learning_rate": 2.9533718009975033e-05, + "loss": 0.2462, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1386929154396057, + "step": 3925, + "valid_targets_mean": 3788.9, + "valid_targets_min": 1750 + }, + { + "epoch": 2.858493997817388, + "grad_norm": 0.6231222438569067, + "learning_rate": 2.950181951322339e-05, + "loss": 0.2558, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14340519905090332, + "step": 3930, + "valid_targets_mean": 3898.2, + "valid_targets_min": 2126 + }, + { + "epoch": 2.862131684248818, + "grad_norm": 0.5850868215292919, + "learning_rate": 2.9469889769347592e-05, + "loss": 0.2649, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13771797716617584, + "step": 3935, + "valid_targets_mean": 3923.1, + "valid_targets_min": 3049 + }, + { + "epoch": 2.8657693706802476, + "grad_norm": 0.5788024037284814, + "learning_rate": 2.943792888334991e-05, + "loss": 0.2543, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14565597474575043, + "step": 3940, + "valid_targets_mean": 4337.1, + "valid_targets_min": 1571 + }, + { + "epoch": 2.869407057111677, + "grad_norm": 0.6375437938990752, + "learning_rate": 2.9405936960335043e-05, + "loss": 0.269, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12174739688634872, + "step": 3945, + "valid_targets_mean": 2998.0, + "valid_targets_min": 1487 + }, + { + "epoch": 2.8730447435431063, + "grad_norm": 0.5596278284016547, + "learning_rate": 2.937391410550973e-05, + "loss": 0.2583, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10632587969303131, + "step": 3950, + "valid_targets_mean": 3627.1, + "valid_targets_min": 1970 + }, + { + "epoch": 2.876682429974536, + "grad_norm": 0.5776167910003552, + "learning_rate": 2.9341860424182457e-05, + "loss": 0.2671, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11557529866695404, + "step": 3955, + "valid_targets_mean": 4038.0, + "valid_targets_min": 2492 + }, + { + "epoch": 2.880320116405966, + "grad_norm": 0.5749414394787049, + "learning_rate": 2.930977602176307e-05, + "loss": 0.2668, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11208498477935791, + "step": 3960, + "valid_targets_mean": 3211.2, + "valid_targets_min": 1066 + }, + { + "epoch": 2.8839578028373953, + "grad_norm": 0.5819761807851471, + "learning_rate": 2.927766100376245e-05, + "loss": 0.2643, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11607399582862854, + "step": 3965, + "valid_targets_mean": 3401.9, + "valid_targets_min": 2281 + }, + { + "epoch": 2.887595489268825, + "grad_norm": 0.6135255256013362, + "learning_rate": 2.9245515475792143e-05, + "loss": 0.2612, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12979558110237122, + "step": 3970, + "valid_targets_mean": 3662.2, + "valid_targets_min": 1969 + }, + { + "epoch": 2.8912331757002545, + "grad_norm": 0.6062616655160732, + "learning_rate": 2.9213339543564055e-05, + "loss": 0.2591, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11503790318965912, + "step": 3975, + "valid_targets_mean": 3189.6, + "valid_targets_min": 1005 + }, + { + "epoch": 2.8948708621316843, + "grad_norm": 0.628714597038046, + "learning_rate": 2.918113331289005e-05, + "loss": 0.2597, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11669173836708069, + "step": 3980, + "valid_targets_mean": 3546.5, + "valid_targets_min": 1980 + }, + { + "epoch": 2.8985085485631137, + "grad_norm": 0.581106063324451, + "learning_rate": 2.9148896889681643e-05, + "loss": 0.2611, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1484648734331131, + "step": 3985, + "valid_targets_mean": 4803.1, + "valid_targets_min": 1257 + }, + { + "epoch": 2.9021462349945435, + "grad_norm": 0.5879847890883845, + "learning_rate": 2.9116630379949643e-05, + "loss": 0.2522, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1323256492614746, + "step": 3990, + "valid_targets_mean": 3118.1, + "valid_targets_min": 1357 + }, + { + "epoch": 2.9057839214259733, + "grad_norm": 0.5465516617030629, + "learning_rate": 2.9084333889803783e-05, + "loss": 0.2393, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10648973286151886, + "step": 3995, + "valid_targets_mean": 4463.8, + "valid_targets_min": 1410 + }, + { + "epoch": 2.9094216078574027, + "grad_norm": 0.5705467415578057, + "learning_rate": 2.905200752545241e-05, + "loss": 0.2592, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1166662871837616, + "step": 4000, + "valid_targets_mean": 3429.8, + "valid_targets_min": 1390 + }, + { + "epoch": 2.913059294288832, + "grad_norm": 0.6163435943752998, + "learning_rate": 2.9019651393202105e-05, + "loss": 0.2452, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11069066822528839, + "step": 4005, + "valid_targets_mean": 3173.0, + "valid_targets_min": 790 + }, + { + "epoch": 2.916696980720262, + "grad_norm": 0.6036718326074603, + "learning_rate": 2.8987265599457326e-05, + "loss": 0.2681, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1336950957775116, + "step": 4010, + "valid_targets_mean": 3788.8, + "valid_targets_min": 1901 + }, + { + "epoch": 2.9203346671516917, + "grad_norm": 0.6318972524052862, + "learning_rate": 2.8954850250720095e-05, + "loss": 0.2538, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.116339311003685, + "step": 4015, + "valid_targets_mean": 2697.1, + "valid_targets_min": 1460 + }, + { + "epoch": 2.923972353583121, + "grad_norm": 0.6572597489480076, + "learning_rate": 2.892240545358962e-05, + "loss": 0.2534, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14784716069698334, + "step": 4020, + "valid_targets_mean": 5088.8, + "valid_targets_min": 952 + }, + { + "epoch": 2.927610040014551, + "grad_norm": 0.6620665029474391, + "learning_rate": 2.8889931314761947e-05, + "loss": 0.2546, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1337994784116745, + "step": 4025, + "valid_targets_mean": 3070.5, + "valid_targets_min": 1768 + }, + { + "epoch": 2.93124772644598, + "grad_norm": 0.5846147968076518, + "learning_rate": 2.885742794102962e-05, + "loss": 0.2565, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10425126552581787, + "step": 4030, + "valid_targets_mean": 3585.0, + "valid_targets_min": 1931 + }, + { + "epoch": 2.93488541287741, + "grad_norm": 0.4934189560873735, + "learning_rate": 2.8824895439281317e-05, + "loss": 0.2478, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11770500987768173, + "step": 4035, + "valid_targets_mean": 5915.4, + "valid_targets_min": 1233 + }, + { + "epoch": 2.9385230993088394, + "grad_norm": 0.5123245246908394, + "learning_rate": 2.879233391650151e-05, + "loss": 0.2432, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12215140461921692, + "step": 4040, + "valid_targets_mean": 5310.6, + "valid_targets_min": 1741 + }, + { + "epoch": 2.942160785740269, + "grad_norm": 0.5490257375036108, + "learning_rate": 2.8759743479770103e-05, + "loss": 0.245, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.116271011531353, + "step": 4045, + "valid_targets_mean": 3884.8, + "valid_targets_min": 856 + }, + { + "epoch": 2.945798472171699, + "grad_norm": 0.5738142047483945, + "learning_rate": 2.8727124236262094e-05, + "loss": 0.2473, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13274259865283966, + "step": 4050, + "valid_targets_mean": 3649.5, + "valid_targets_min": 1062 + }, + { + "epoch": 2.9494361586031284, + "grad_norm": 0.5833694873405221, + "learning_rate": 2.8694476293247196e-05, + "loss": 0.239, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11949186772108078, + "step": 4055, + "valid_targets_mean": 4412.6, + "valid_targets_min": 974 + }, + { + "epoch": 2.953073845034558, + "grad_norm": 0.6050746654209781, + "learning_rate": 2.866179975808952e-05, + "loss": 0.2435, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1292571872472763, + "step": 4060, + "valid_targets_mean": 3077.5, + "valid_targets_min": 1447 + }, + { + "epoch": 2.9567115314659875, + "grad_norm": 0.5836263545638236, + "learning_rate": 2.8629094738247188e-05, + "loss": 0.2398, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14119437336921692, + "step": 4065, + "valid_targets_mean": 4663.0, + "valid_targets_min": 2144 + }, + { + "epoch": 2.9603492178974173, + "grad_norm": 0.5619390007674765, + "learning_rate": 2.8596361341272013e-05, + "loss": 0.2406, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1377207636833191, + "step": 4070, + "valid_targets_mean": 4543.1, + "valid_targets_min": 1515 + }, + { + "epoch": 2.9639869043288467, + "grad_norm": 0.6871856263018866, + "learning_rate": 2.8563599674809105e-05, + "loss": 0.2583, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12471659481525421, + "step": 4075, + "valid_targets_mean": 3034.9, + "valid_targets_min": 1349 + }, + { + "epoch": 2.9676245907602765, + "grad_norm": 0.6312141006061917, + "learning_rate": 2.8530809846596565e-05, + "loss": 0.2585, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12723305821418762, + "step": 4080, + "valid_targets_mean": 3150.4, + "valid_targets_min": 1906 + }, + { + "epoch": 2.9712622771917063, + "grad_norm": 0.6356009438634392, + "learning_rate": 2.8497991964465076e-05, + "loss": 0.2455, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12955406308174133, + "step": 4085, + "valid_targets_mean": 3655.1, + "valid_targets_min": 2076 + }, + { + "epoch": 2.9748999636231357, + "grad_norm": 0.5667342074462254, + "learning_rate": 2.846514613633761e-05, + "loss": 0.242, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1029905378818512, + "step": 4090, + "valid_targets_mean": 2994.0, + "valid_targets_min": 1021 + }, + { + "epoch": 2.978537650054565, + "grad_norm": 0.6470389402961754, + "learning_rate": 2.8432272470229008e-05, + "loss": 0.2517, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1255868524312973, + "step": 4095, + "valid_targets_mean": 3226.9, + "valid_targets_min": 654 + }, + { + "epoch": 2.982175336485995, + "grad_norm": 0.5910396113389718, + "learning_rate": 2.8399371074245672e-05, + "loss": 0.2421, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14116665720939636, + "step": 4100, + "valid_targets_mean": 4790.8, + "valid_targets_min": 3074 + }, + { + "epoch": 2.9858130229174247, + "grad_norm": 0.6830546390606256, + "learning_rate": 2.8366442056585208e-05, + "loss": 0.2616, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12833862006664276, + "step": 4105, + "valid_targets_mean": 2604.0, + "valid_targets_min": 815 + }, + { + "epoch": 2.989450709348854, + "grad_norm": 0.5869793062466598, + "learning_rate": 2.833348552553604e-05, + "loss": 0.2497, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11382627487182617, + "step": 4110, + "valid_targets_mean": 3560.0, + "valid_targets_min": 2205 + }, + { + "epoch": 2.993088395780284, + "grad_norm": 0.5756849252647395, + "learning_rate": 2.8300501589477073e-05, + "loss": 0.2568, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13684064149856567, + "step": 4115, + "valid_targets_mean": 4216.0, + "valid_targets_min": 1793 + }, + { + "epoch": 2.996726082211713, + "grad_norm": 0.5942596113375226, + "learning_rate": 2.8267490356877328e-05, + "loss": 0.256, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11879312992095947, + "step": 4120, + "valid_targets_mean": 3134.5, + "valid_targets_min": 1217 + }, + { + "epoch": 3.0, + "grad_norm": 0.7543768011036392, + "learning_rate": 2.8234451936295625e-05, + "loss": 0.2309, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12745636701583862, + "step": 4125, + "valid_targets_mean": 1910.0, + "valid_targets_min": 393 + }, + { + "epoch": 3.00363768643143, + "grad_norm": 0.4287215417774143, + "learning_rate": 2.820138643638015e-05, + "loss": 0.1301, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05381636321544647, + "step": 4130, + "valid_targets_mean": 2722.6, + "valid_targets_min": 458 + }, + { + "epoch": 3.007275372862859, + "grad_norm": 0.4270633131127547, + "learning_rate": 2.8168293965868153e-05, + "loss": 0.1015, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05152353271842003, + "step": 4135, + "valid_targets_mean": 3113.1, + "valid_targets_min": 2461 + }, + { + "epoch": 3.010913059294289, + "grad_norm": 0.36945825376647345, + "learning_rate": 2.81351746335856e-05, + "loss": 0.0925, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05497678369283676, + "step": 4140, + "valid_targets_mean": 3648.1, + "valid_targets_min": 2068 + }, + { + "epoch": 3.0145507457257184, + "grad_norm": 0.41184015691909215, + "learning_rate": 2.8102028548446772e-05, + "loss": 0.1155, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.053780507296323776, + "step": 4145, + "valid_targets_mean": 3423.6, + "valid_targets_min": 2425 + }, + { + "epoch": 3.018188432157148, + "grad_norm": 0.44031586285122204, + "learning_rate": 2.8068855819453936e-05, + "loss": 0.0927, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04670848697423935, + "step": 4150, + "valid_targets_mean": 2724.1, + "valid_targets_min": 1198 + }, + { + "epoch": 3.0218261185885775, + "grad_norm": 0.7631268618655576, + "learning_rate": 2.8035656555696964e-05, + "loss": 0.1097, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07663515210151672, + "step": 4155, + "valid_targets_mean": 1232.5, + "valid_targets_min": 807 + }, + { + "epoch": 3.0254638050200073, + "grad_norm": 0.42881656200017365, + "learning_rate": 2.800243086635301e-05, + "loss": 0.1499, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06026969105005264, + "step": 4160, + "valid_targets_mean": 3487.4, + "valid_targets_min": 2351 + }, + { + "epoch": 3.0291014914514367, + "grad_norm": 0.45715399723043953, + "learning_rate": 2.7969178860686112e-05, + "loss": 0.1029, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07939073443412781, + "step": 4165, + "valid_targets_mean": 3126.4, + "valid_targets_min": 781 + }, + { + "epoch": 3.0327391778828665, + "grad_norm": 0.3404134964941753, + "learning_rate": 2.793590064804687e-05, + "loss": 0.1135, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.051661379635334015, + "step": 4170, + "valid_targets_mean": 3474.8, + "valid_targets_min": 799 + }, + { + "epoch": 3.0363768643142963, + "grad_norm": 0.2990684956969066, + "learning_rate": 2.7902596337872042e-05, + "loss": 0.1177, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04572618380188942, + "step": 4175, + "valid_targets_mean": 5026.1, + "valid_targets_min": 2802 + }, + { + "epoch": 3.0400145507457257, + "grad_norm": 0.3420876999337161, + "learning_rate": 2.7869266039684223e-05, + "loss": 0.0912, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03580169752240181, + "step": 4180, + "valid_targets_mean": 2611.0, + "valid_targets_min": 574 + }, + { + "epoch": 3.0436522371771555, + "grad_norm": 0.36181744680744127, + "learning_rate": 2.7835909863091483e-05, + "loss": 0.0992, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04692269116640091, + "step": 4185, + "valid_targets_mean": 4036.5, + "valid_targets_min": 2642 + }, + { + "epoch": 3.047289923608585, + "grad_norm": 0.5506085523489712, + "learning_rate": 2.7802527917786955e-05, + "loss": 0.1028, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06711885333061218, + "step": 4190, + "valid_targets_mean": 1960.5, + "valid_targets_min": 593 + }, + { + "epoch": 3.0509276100400147, + "grad_norm": 0.3826583761426045, + "learning_rate": 2.776912031354856e-05, + "loss": 0.1293, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04094811901450157, + "step": 4195, + "valid_targets_mean": 3354.2, + "valid_targets_min": 1353 + }, + { + "epoch": 3.054565296471444, + "grad_norm": 0.4645761776564635, + "learning_rate": 2.7735687160238576e-05, + "loss": 0.095, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05478697270154953, + "step": 4200, + "valid_targets_mean": 2300.9, + "valid_targets_min": 818 + }, + { + "epoch": 3.058202982902874, + "grad_norm": 0.33711812040945555, + "learning_rate": 2.7702228567803295e-05, + "loss": 0.0948, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03377291560173035, + "step": 4205, + "valid_targets_mean": 2295.8, + "valid_targets_min": 934 + }, + { + "epoch": 3.061840669334303, + "grad_norm": 0.376099509318961, + "learning_rate": 2.7668744646272673e-05, + "loss": 0.0867, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04774352163076401, + "step": 4210, + "valid_targets_mean": 3524.4, + "valid_targets_min": 2798 + }, + { + "epoch": 3.065478355765733, + "grad_norm": 0.42584230541745577, + "learning_rate": 2.7635235505759966e-05, + "loss": 0.1097, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04377376288175583, + "step": 4215, + "valid_targets_mean": 2810.1, + "valid_targets_min": 471 + }, + { + "epoch": 3.0691160421971624, + "grad_norm": 0.4161932690331968, + "learning_rate": 2.760170125646137e-05, + "loss": 0.1103, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06973842531442642, + "step": 4220, + "valid_targets_mean": 3042.8, + "valid_targets_min": 903 + }, + { + "epoch": 3.072753728628592, + "grad_norm": 0.40055539232535053, + "learning_rate": 2.756814200865562e-05, + "loss": 0.1094, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04281292110681534, + "step": 4225, + "valid_targets_mean": 3081.6, + "valid_targets_min": 474 + }, + { + "epoch": 3.076391415060022, + "grad_norm": 0.40070639381288975, + "learning_rate": 2.7534557872703705e-05, + "loss": 0.111, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05405103415250778, + "step": 4230, + "valid_targets_mean": 3468.0, + "valid_targets_min": 640 + }, + { + "epoch": 3.0800291014914514, + "grad_norm": 0.3321494120233214, + "learning_rate": 2.7500948959048424e-05, + "loss": 0.1443, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03560910001397133, + "step": 4235, + "valid_targets_mean": 3438.4, + "valid_targets_min": 902 + }, + { + "epoch": 3.083666787922881, + "grad_norm": 0.39989722528073435, + "learning_rate": 2.7467315378214077e-05, + "loss": 0.0939, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.046598583459854126, + "step": 4240, + "valid_targets_mean": 3212.8, + "valid_targets_min": 1837 + }, + { + "epoch": 3.0873044743543105, + "grad_norm": 0.7387876156197377, + "learning_rate": 2.743365724080607e-05, + "loss": 0.1262, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09828364849090576, + "step": 4245, + "valid_targets_mean": 1358.8, + "valid_targets_min": 756 + }, + { + "epoch": 3.0909421607857404, + "grad_norm": 0.5294843584946793, + "learning_rate": 2.7399974657510583e-05, + "loss": 0.1276, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09029585123062134, + "step": 4250, + "valid_targets_mean": 2896.2, + "valid_targets_min": 1072 + }, + { + "epoch": 3.0945798472171697, + "grad_norm": 0.3575879537147911, + "learning_rate": 2.736626773909417e-05, + "loss": 0.0994, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.032179757952690125, + "step": 4255, + "valid_targets_mean": 2908.1, + "valid_targets_min": 895 + }, + { + "epoch": 3.0982175336485995, + "grad_norm": 0.6038146054922319, + "learning_rate": 2.733253659640342e-05, + "loss": 0.2216, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19517835974693298, + "step": 4260, + "valid_targets_mean": 2562.2, + "valid_targets_min": 878 + }, + { + "epoch": 3.101855220080029, + "grad_norm": 0.42588371258804786, + "learning_rate": 2.729878134036458e-05, + "loss": 0.2035, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.046075351536273956, + "step": 4265, + "valid_targets_mean": 2087.6, + "valid_targets_min": 756 + }, + { + "epoch": 3.1054929065114587, + "grad_norm": 0.32700514534949815, + "learning_rate": 2.726500208198319e-05, + "loss": 0.112, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03978538513183594, + "step": 4270, + "valid_targets_mean": 3421.8, + "valid_targets_min": 2963 + }, + { + "epoch": 3.1091305929428885, + "grad_norm": 0.7434576365293648, + "learning_rate": 2.7231198932343747e-05, + "loss": 0.183, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23474599421024323, + "step": 4275, + "valid_targets_mean": 2375.2, + "valid_targets_min": 1153 + }, + { + "epoch": 3.112768279374318, + "grad_norm": 0.3074090905511868, + "learning_rate": 2.7197372002609283e-05, + "loss": 0.1206, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03939659893512726, + "step": 4280, + "valid_targets_mean": 4886.2, + "valid_targets_min": 1700 + }, + { + "epoch": 3.1164059658057477, + "grad_norm": 0.3192597504200334, + "learning_rate": 2.7163521404021064e-05, + "loss": 0.0978, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.032613445073366165, + "step": 4285, + "valid_targets_mean": 3965.1, + "valid_targets_min": 789 + }, + { + "epoch": 3.120043652237177, + "grad_norm": 0.354985431450355, + "learning_rate": 2.7129647247898162e-05, + "loss": 0.1065, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05446261912584305, + "step": 4290, + "valid_targets_mean": 3771.5, + "valid_targets_min": 995 + }, + { + "epoch": 3.123681338668607, + "grad_norm": 0.39320797787978967, + "learning_rate": 2.7095749645637135e-05, + "loss": 0.0962, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.032891303300857544, + "step": 4295, + "valid_targets_mean": 2220.8, + "valid_targets_min": 488 + }, + { + "epoch": 3.1273190251000362, + "grad_norm": 0.42517655062719134, + "learning_rate": 2.706182870871165e-05, + "loss": 0.0971, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05703073740005493, + "step": 4300, + "valid_targets_mean": 3635.4, + "valid_targets_min": 2437 + }, + { + "epoch": 3.130956711531466, + "grad_norm": 0.516003445513525, + "learning_rate": 2.7027884548672082e-05, + "loss": 0.0927, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04004015401005745, + "step": 4305, + "valid_targets_mean": 1752.1, + "valid_targets_min": 712 + }, + { + "epoch": 3.1345943979628954, + "grad_norm": 0.3467676115218773, + "learning_rate": 2.6993917277145222e-05, + "loss": 0.0926, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.046710897237062454, + "step": 4310, + "valid_targets_mean": 3713.4, + "valid_targets_min": 2422 + }, + { + "epoch": 3.138232084394325, + "grad_norm": 0.3900285448246925, + "learning_rate": 2.6959927005833817e-05, + "loss": 0.0837, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.045447416603565216, + "step": 4315, + "valid_targets_mean": 3998.5, + "valid_targets_min": 3643 + }, + { + "epoch": 3.141869770825755, + "grad_norm": 0.3189884151548082, + "learning_rate": 2.6925913846516292e-05, + "loss": 0.1449, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03279590234160423, + "step": 4320, + "valid_targets_mean": 3319.0, + "valid_targets_min": 772 + }, + { + "epoch": 3.1455074572571844, + "grad_norm": 0.370758215303547, + "learning_rate": 2.6891877911046303e-05, + "loss": 0.0773, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.032603129744529724, + "step": 4325, + "valid_targets_mean": 2934.8, + "valid_targets_min": 722 + }, + { + "epoch": 3.149145143688614, + "grad_norm": 0.33463574571641586, + "learning_rate": 2.6857819311352433e-05, + "loss": 0.0865, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03405681997537613, + "step": 4330, + "valid_targets_mean": 3656.4, + "valid_targets_min": 2788 + }, + { + "epoch": 3.1527828301200436, + "grad_norm": 0.3706708672864502, + "learning_rate": 2.6823738159437787e-05, + "loss": 0.0876, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04544834792613983, + "step": 4335, + "valid_targets_mean": 3202.5, + "valid_targets_min": 948 + }, + { + "epoch": 3.1564205165514734, + "grad_norm": 0.5664165969451758, + "learning_rate": 2.6789634567379627e-05, + "loss": 0.1078, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.052561357617378235, + "step": 4340, + "valid_targets_mean": 1002.0, + "valid_targets_min": 490 + }, + { + "epoch": 3.1600582029829027, + "grad_norm": 0.3543154322238887, + "learning_rate": 2.6755508647329033e-05, + "loss": 0.121, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0520610511302948, + "step": 4345, + "valid_targets_mean": 3751.9, + "valid_targets_min": 2800 + }, + { + "epoch": 3.1636958894143326, + "grad_norm": 0.49260281424999225, + "learning_rate": 2.6721360511510476e-05, + "loss": 0.129, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08149917423725128, + "step": 4350, + "valid_targets_mean": 2641.8, + "valid_targets_min": 896 + }, + { + "epoch": 3.167333575845762, + "grad_norm": 0.4607714192364014, + "learning_rate": 2.6687190272221522e-05, + "loss": 0.1054, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05892542749643326, + "step": 4355, + "valid_targets_mean": 3199.0, + "valid_targets_min": 1333 + }, + { + "epoch": 3.1709712622771917, + "grad_norm": 0.4603327770257462, + "learning_rate": 2.6652998041832394e-05, + "loss": 0.1202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05408715829253197, + "step": 4360, + "valid_targets_mean": 2548.1, + "valid_targets_min": 693 + }, + { + "epoch": 3.174608948708621, + "grad_norm": 0.38625066426727866, + "learning_rate": 2.6618783932785662e-05, + "loss": 0.1347, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04113394394516945, + "step": 4365, + "valid_targets_mean": 2482.9, + "valid_targets_min": 803 + }, + { + "epoch": 3.178246635140051, + "grad_norm": 0.3862165485632013, + "learning_rate": 2.658454805759582e-05, + "loss": 0.1007, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05237935110926628, + "step": 4370, + "valid_targets_mean": 3239.2, + "valid_targets_min": 2525 + }, + { + "epoch": 3.1818843215714807, + "grad_norm": 0.5225360274236779, + "learning_rate": 2.655029052884895e-05, + "loss": 0.1242, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04548477381467819, + "step": 4375, + "valid_targets_mean": 3806.1, + "valid_targets_min": 2393 + }, + { + "epoch": 3.18552200800291, + "grad_norm": 0.41076875067176494, + "learning_rate": 2.6516011459202356e-05, + "loss": 0.1064, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047077976167201996, + "step": 4380, + "valid_targets_mean": 2990.9, + "valid_targets_min": 804 + }, + { + "epoch": 3.18915969443434, + "grad_norm": 0.3619084100967564, + "learning_rate": 2.6481710961384152e-05, + "loss": 0.0968, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047198586165905, + "step": 4385, + "valid_targets_mean": 4885.8, + "valid_targets_min": 3695 + }, + { + "epoch": 3.1927973808657693, + "grad_norm": 0.49910741903659817, + "learning_rate": 2.6447389148192944e-05, + "loss": 0.0881, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.053359534591436386, + "step": 4390, + "valid_targets_mean": 3128.8, + "valid_targets_min": 596 + }, + { + "epoch": 3.196435067297199, + "grad_norm": 0.28983307775216033, + "learning_rate": 2.6413046132497425e-05, + "loss": 0.0857, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03270778805017471, + "step": 4395, + "valid_targets_mean": 4567.8, + "valid_targets_min": 3804 + }, + { + "epoch": 3.2000727537286284, + "grad_norm": 0.3317313617205048, + "learning_rate": 2.6378682027236022e-05, + "loss": 0.0925, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03664100170135498, + "step": 4400, + "valid_targets_mean": 3853.1, + "valid_targets_min": 2862 + }, + { + "epoch": 3.2037104401600582, + "grad_norm": 0.5552639836605017, + "learning_rate": 2.6344296945416495e-05, + "loss": 0.1149, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06586731970310211, + "step": 4405, + "valid_targets_mean": 1072.6, + "valid_targets_min": 497 + }, + { + "epoch": 3.207348126591488, + "grad_norm": 0.3598978214169791, + "learning_rate": 2.6309891000115605e-05, + "loss": 0.0962, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05427849292755127, + "step": 4410, + "valid_targets_mean": 4191.0, + "valid_targets_min": 3918 + }, + { + "epoch": 3.2109858130229174, + "grad_norm": 0.5421500914542011, + "learning_rate": 2.627546430447872e-05, + "loss": 0.1091, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04944329708814621, + "step": 4415, + "valid_targets_mean": 2121.9, + "valid_targets_min": 785 + }, + { + "epoch": 3.214623499454347, + "grad_norm": 0.3308496590316084, + "learning_rate": 2.6241016971719435e-05, + "loss": 0.079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.040540844202041626, + "step": 4420, + "valid_targets_mean": 3451.6, + "valid_targets_min": 905 + }, + { + "epoch": 3.2182611858857766, + "grad_norm": 0.40366014102244474, + "learning_rate": 2.6206549115119236e-05, + "loss": 0.1039, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07467225939035416, + "step": 4425, + "valid_targets_mean": 2929.2, + "valid_targets_min": 657 + }, + { + "epoch": 3.2218988723172064, + "grad_norm": 0.36238149024010136, + "learning_rate": 2.6172060848027066e-05, + "loss": 0.0955, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04922741651535034, + "step": 4430, + "valid_targets_mean": 3798.9, + "valid_targets_min": 3184 + }, + { + "epoch": 3.2255365587486358, + "grad_norm": 0.47390553042949995, + "learning_rate": 2.613755228385903e-05, + "loss": 0.086, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.052453458309173584, + "step": 4435, + "valid_targets_mean": 1463.9, + "valid_targets_min": 627 + }, + { + "epoch": 3.2291742451800656, + "grad_norm": 0.46804442440229715, + "learning_rate": 2.610302353609795e-05, + "loss": 0.0971, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04647493362426758, + "step": 4440, + "valid_targets_mean": 1701.5, + "valid_targets_min": 827 + }, + { + "epoch": 3.232811931611495, + "grad_norm": 0.46268472209082717, + "learning_rate": 2.6068474718293035e-05, + "loss": 0.1002, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04509687423706055, + "step": 4445, + "valid_targets_mean": 1789.9, + "valid_targets_min": 940 + }, + { + "epoch": 3.2364496180429247, + "grad_norm": 0.393682947804984, + "learning_rate": 2.603390594405949e-05, + "loss": 0.1163, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04263824224472046, + "step": 4450, + "valid_targets_mean": 2943.8, + "valid_targets_min": 958 + }, + { + "epoch": 3.240087304474354, + "grad_norm": 0.4394941122455063, + "learning_rate": 2.5999317327078147e-05, + "loss": 0.0958, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0396265909075737, + "step": 4455, + "valid_targets_mean": 1859.2, + "valid_targets_min": 719 + }, + { + "epoch": 3.243724990905784, + "grad_norm": 0.5090788205316978, + "learning_rate": 2.5964708981095106e-05, + "loss": 0.1177, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.036966800689697266, + "step": 4460, + "valid_targets_mean": 1703.9, + "valid_targets_min": 759 + }, + { + "epoch": 3.2473626773372137, + "grad_norm": 0.8410057678920028, + "learning_rate": 2.5930081019921326e-05, + "loss": 0.1192, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1211915910243988, + "step": 4465, + "valid_targets_mean": 1556.6, + "valid_targets_min": 521 + }, + { + "epoch": 3.251000363768643, + "grad_norm": 0.40735096506741686, + "learning_rate": 2.589543355743229e-05, + "loss": 0.0975, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05928069353103638, + "step": 4470, + "valid_targets_mean": 3267.6, + "valid_targets_min": 1032 + }, + { + "epoch": 3.254638050200073, + "grad_norm": 0.2929553744149724, + "learning_rate": 2.5860766707567595e-05, + "loss": 0.0825, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03371032699942589, + "step": 4475, + "valid_targets_mean": 3609.0, + "valid_targets_min": 2741 + }, + { + "epoch": 3.2582757366315023, + "grad_norm": 0.35905281624750407, + "learning_rate": 2.5826080584330607e-05, + "loss": 0.095, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.042490120977163315, + "step": 4480, + "valid_targets_mean": 2735.8, + "valid_targets_min": 1692 + }, + { + "epoch": 3.261913423062932, + "grad_norm": 0.40827007806623855, + "learning_rate": 2.5791375301788068e-05, + "loss": 0.1256, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.053532056510448456, + "step": 4485, + "valid_targets_mean": 3693.4, + "valid_targets_min": 2215 + }, + { + "epoch": 3.2655511094943614, + "grad_norm": 0.3415324052216411, + "learning_rate": 2.575665097406972e-05, + "loss": 0.0804, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0357334278523922, + "step": 4490, + "valid_targets_mean": 3109.9, + "valid_targets_min": 723 + }, + { + "epoch": 3.2691887959257913, + "grad_norm": 0.3267890142760794, + "learning_rate": 2.572190771536795e-05, + "loss": 0.0957, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04056313633918762, + "step": 4495, + "valid_targets_mean": 4984.2, + "valid_targets_min": 3282 + }, + { + "epoch": 3.2728264823572206, + "grad_norm": 0.2965626878361444, + "learning_rate": 2.5687145639937393e-05, + "loss": 0.0914, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03863619267940521, + "step": 4500, + "valid_targets_mean": 3863.4, + "valid_targets_min": 995 + }, + { + "epoch": 3.2764641687886504, + "grad_norm": 0.2830157158146061, + "learning_rate": 2.5652364862094562e-05, + "loss": 0.0777, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.036777086555957794, + "step": 4505, + "valid_targets_mean": 4201.0, + "valid_targets_min": 949 + }, + { + "epoch": 3.28010185522008, + "grad_norm": 0.3993128740872454, + "learning_rate": 2.5617565496217465e-05, + "loss": 0.1006, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06263408064842224, + "step": 4510, + "valid_targets_mean": 4321.6, + "valid_targets_min": 2724 + }, + { + "epoch": 3.2837395416515096, + "grad_norm": 0.3156643422749747, + "learning_rate": 2.5582747656745264e-05, + "loss": 0.0843, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02916128747165203, + "step": 4515, + "valid_targets_mean": 3384.5, + "valid_targets_min": 424 + }, + { + "epoch": 3.2873772280829394, + "grad_norm": 0.37241515630056654, + "learning_rate": 2.554791145817785e-05, + "loss": 0.1201, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03795400634407997, + "step": 4520, + "valid_targets_mean": 2987.2, + "valid_targets_min": 531 + }, + { + "epoch": 3.291014914514369, + "grad_norm": 0.35153484717645267, + "learning_rate": 2.5513057015075483e-05, + "loss": 0.114, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03742213547229767, + "step": 4525, + "valid_targets_mean": 3469.5, + "valid_targets_min": 1195 + }, + { + "epoch": 3.2946526009457986, + "grad_norm": 0.36408396186761965, + "learning_rate": 2.547818444205846e-05, + "loss": 0.0808, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07128562033176422, + "step": 4530, + "valid_targets_mean": 2682.8, + "valid_targets_min": 722 + }, + { + "epoch": 3.298290287377228, + "grad_norm": 0.34153093131834356, + "learning_rate": 2.5443293853806634e-05, + "loss": 0.1121, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04040738195180893, + "step": 4535, + "valid_targets_mean": 3496.9, + "valid_targets_min": 1251 + }, + { + "epoch": 3.3019279738086578, + "grad_norm": 0.403442501763785, + "learning_rate": 2.5408385365059167e-05, + "loss": 0.0908, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.050956886261701584, + "step": 4540, + "valid_targets_mean": 2912.5, + "valid_targets_min": 970 + }, + { + "epoch": 3.305565660240087, + "grad_norm": 0.5856314907598824, + "learning_rate": 2.5373459090614048e-05, + "loss": 0.1049, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05067780986428261, + "step": 4545, + "valid_targets_mean": 1423.6, + "valid_targets_min": 634 + }, + { + "epoch": 3.309203346671517, + "grad_norm": 0.5738301219799696, + "learning_rate": 2.533851514532777e-05, + "loss": 0.1176, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07840394228696823, + "step": 4550, + "valid_targets_mean": 1423.9, + "valid_targets_min": 685 + }, + { + "epoch": 3.3128410331029468, + "grad_norm": 0.3663283713030322, + "learning_rate": 2.5303553644114933e-05, + "loss": 0.1279, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.036331646144390106, + "step": 4555, + "valid_targets_mean": 2932.5, + "valid_targets_min": 593 + }, + { + "epoch": 3.316478719534376, + "grad_norm": 0.30431717816559906, + "learning_rate": 2.5268574701947875e-05, + "loss": 0.0985, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.037071093916893005, + "step": 4560, + "valid_targets_mean": 3945.8, + "valid_targets_min": 3129 + }, + { + "epoch": 3.320116405965806, + "grad_norm": 0.4164052858153671, + "learning_rate": 2.523357843385628e-05, + "loss": 0.1021, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05604838952422142, + "step": 4565, + "valid_targets_mean": 3621.1, + "valid_targets_min": 859 + }, + { + "epoch": 3.3237540923972353, + "grad_norm": 0.443911771831657, + "learning_rate": 2.5198564954926813e-05, + "loss": 0.0918, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04908209294080734, + "step": 4570, + "valid_targets_mean": 2553.0, + "valid_targets_min": 616 + }, + { + "epoch": 3.327391778828665, + "grad_norm": 0.4003296041938948, + "learning_rate": 2.5163534380302747e-05, + "loss": 0.0903, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04099854454398155, + "step": 4575, + "valid_targets_mean": 2690.5, + "valid_targets_min": 541 + }, + { + "epoch": 3.3310294652600945, + "grad_norm": 0.6443270105154282, + "learning_rate": 2.5128486825183556e-05, + "loss": 0.0874, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.055260926485061646, + "step": 4580, + "valid_targets_mean": 906.4, + "valid_targets_min": 598 + }, + { + "epoch": 3.3346671516915243, + "grad_norm": 0.404060662851538, + "learning_rate": 2.5093422404824574e-05, + "loss": 0.1046, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04591308534145355, + "step": 4585, + "valid_targets_mean": 3147.2, + "valid_targets_min": 1101 + }, + { + "epoch": 3.3383048381229536, + "grad_norm": 0.4800816577584834, + "learning_rate": 2.5058341234536595e-05, + "loss": 0.0859, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03849026560783386, + "step": 4590, + "valid_targets_mean": 2331.8, + "valid_targets_min": 680 + }, + { + "epoch": 3.3419425245543835, + "grad_norm": 0.3734660876379632, + "learning_rate": 2.5023243429685478e-05, + "loss": 0.0882, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04736591875553131, + "step": 4595, + "valid_targets_mean": 2577.0, + "valid_targets_min": 774 + }, + { + "epoch": 3.345580210985813, + "grad_norm": 0.4681275617693876, + "learning_rate": 2.4988129105691797e-05, + "loss": 0.1044, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.056144267320632935, + "step": 4600, + "valid_targets_mean": 4140.1, + "valid_targets_min": 1474 + }, + { + "epoch": 3.3492178974172426, + "grad_norm": 0.2261006833349661, + "learning_rate": 2.4952998378030463e-05, + "loss": 0.0806, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.028667740523815155, + "step": 4605, + "valid_targets_mean": 5498.9, + "valid_targets_min": 3677 + }, + { + "epoch": 3.3528555838486724, + "grad_norm": 0.36798225168060944, + "learning_rate": 2.491785136223031e-05, + "loss": 0.0884, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.053660616278648376, + "step": 4610, + "valid_targets_mean": 3352.5, + "valid_targets_min": 675 + }, + { + "epoch": 3.356493270280102, + "grad_norm": 0.3765255466672178, + "learning_rate": 2.488268817387374e-05, + "loss": 0.0852, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03555317968130112, + "step": 4615, + "valid_targets_mean": 3902.5, + "valid_targets_min": 795 + }, + { + "epoch": 3.3601309567115316, + "grad_norm": 0.5098592906744859, + "learning_rate": 2.484750892859636e-05, + "loss": 0.1918, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08414119482040405, + "step": 4620, + "valid_targets_mean": 2495.0, + "valid_targets_min": 605 + }, + { + "epoch": 3.363768643142961, + "grad_norm": 0.3215095012170358, + "learning_rate": 2.4812313742086543e-05, + "loss": 0.0846, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0438154898583889, + "step": 4625, + "valid_targets_mean": 3298.6, + "valid_targets_min": 985 + }, + { + "epoch": 3.367406329574391, + "grad_norm": 0.37008556103292095, + "learning_rate": 2.477710273008512e-05, + "loss": 0.0972, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04828403890132904, + "step": 4630, + "valid_targets_mean": 3505.2, + "valid_targets_min": 878 + }, + { + "epoch": 3.37104401600582, + "grad_norm": 0.5872954461335614, + "learning_rate": 2.4741876008384944e-05, + "loss": 0.1189, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12595906853675842, + "step": 4635, + "valid_targets_mean": 2289.5, + "valid_targets_min": 1214 + }, + { + "epoch": 3.37468170243725, + "grad_norm": 0.45121110787178426, + "learning_rate": 2.470663369283055e-05, + "loss": 0.187, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12383411079645157, + "step": 4640, + "valid_targets_mean": 8496.5, + "valid_targets_min": 6565 + }, + { + "epoch": 3.3783193888686793, + "grad_norm": 0.4210373359709796, + "learning_rate": 2.4671375899317737e-05, + "loss": 0.2007, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08909362554550171, + "step": 4645, + "valid_targets_mean": 5663.5, + "valid_targets_min": 3833 + }, + { + "epoch": 3.381957075300109, + "grad_norm": 0.3810637404478252, + "learning_rate": 2.46361027437932e-05, + "loss": 0.1918, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09538260847330093, + "step": 4650, + "valid_targets_mean": 6654.1, + "valid_targets_min": 5324 + }, + { + "epoch": 3.3855947617315385, + "grad_norm": 0.4543810782907732, + "learning_rate": 2.4600814342254174e-05, + "loss": 0.1836, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09072799980640411, + "step": 4655, + "valid_targets_mean": 7635.1, + "valid_targets_min": 5672 + }, + { + "epoch": 3.3892324481629683, + "grad_norm": 0.3306758902649885, + "learning_rate": 2.4565510810748005e-05, + "loss": 0.1893, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0854153260588646, + "step": 4660, + "valid_targets_mean": 8517.0, + "valid_targets_min": 5045 + }, + { + "epoch": 3.392870134594398, + "grad_norm": 0.3841165090451712, + "learning_rate": 2.4530192265371817e-05, + "loss": 0.1935, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09400700032711029, + "step": 4665, + "valid_targets_mean": 6759.1, + "valid_targets_min": 3152 + }, + { + "epoch": 3.3965078210258275, + "grad_norm": 0.35095694794798266, + "learning_rate": 2.4494858822272094e-05, + "loss": 0.1991, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08769888430833817, + "step": 4670, + "valid_targets_mean": 7822.8, + "valid_targets_min": 4820 + }, + { + "epoch": 3.4001455074572573, + "grad_norm": 0.3678179870262964, + "learning_rate": 2.4459510597644322e-05, + "loss": 0.1884, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0757410079240799, + "step": 4675, + "valid_targets_mean": 6852.2, + "valid_targets_min": 4135 + }, + { + "epoch": 3.4037831938886867, + "grad_norm": 0.4049544129584414, + "learning_rate": 2.442414770773259e-05, + "loss": 0.1853, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08173209428787231, + "step": 4680, + "valid_targets_mean": 4794.2, + "valid_targets_min": 3230 + }, + { + "epoch": 3.4074208803201165, + "grad_norm": 0.3698709674598867, + "learning_rate": 2.4388770268829217e-05, + "loss": 0.1681, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09619638323783875, + "step": 4685, + "valid_targets_mean": 7691.0, + "valid_targets_min": 5900 + }, + { + "epoch": 3.411058566751546, + "grad_norm": 0.39477711238644175, + "learning_rate": 2.435337839727436e-05, + "loss": 0.1878, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.088360995054245, + "step": 4690, + "valid_targets_mean": 6510.9, + "valid_targets_min": 5639 + }, + { + "epoch": 3.4146962531829756, + "grad_norm": 0.4011818301983857, + "learning_rate": 2.431797220945564e-05, + "loss": 0.177, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0920364186167717, + "step": 4695, + "valid_targets_mean": 6096.8, + "valid_targets_min": 4669 + }, + { + "epoch": 3.4183339396144055, + "grad_norm": 0.36303149991580924, + "learning_rate": 2.428255182180777e-05, + "loss": 0.184, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09486152231693268, + "step": 4700, + "valid_targets_mean": 6812.2, + "valid_targets_min": 5160 + }, + { + "epoch": 3.421971626045835, + "grad_norm": 0.35622468316421296, + "learning_rate": 2.4247117350812143e-05, + "loss": 0.1762, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08160389214754105, + "step": 4705, + "valid_targets_mean": 6980.4, + "valid_targets_min": 5151 + }, + { + "epoch": 3.4256093124772646, + "grad_norm": 0.36533413740639853, + "learning_rate": 2.421166891299649e-05, + "loss": 0.1791, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09681347757577896, + "step": 4710, + "valid_targets_mean": 7417.1, + "valid_targets_min": 5932 + }, + { + "epoch": 3.429246998908694, + "grad_norm": 0.3845419277495396, + "learning_rate": 2.4176206624934438e-05, + "loss": 0.1873, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08905616402626038, + "step": 4715, + "valid_targets_mean": 6258.5, + "valid_targets_min": 5569 + }, + { + "epoch": 3.432884685340124, + "grad_norm": 0.3811194547688403, + "learning_rate": 2.414073060324519e-05, + "loss": 0.1779, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1009448915719986, + "step": 4720, + "valid_targets_mean": 7689.1, + "valid_targets_min": 6223 + }, + { + "epoch": 3.436522371771553, + "grad_norm": 0.4926531867045376, + "learning_rate": 2.41052409645931e-05, + "loss": 0.1876, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10826841741800308, + "step": 4725, + "valid_targets_mean": 4371.9, + "valid_targets_min": 1172 + }, + { + "epoch": 3.440160058202983, + "grad_norm": 0.3946697544549626, + "learning_rate": 2.4069737825687302e-05, + "loss": 0.1973, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10552500188350677, + "step": 4730, + "valid_targets_mean": 6936.2, + "valid_targets_min": 5619 + }, + { + "epoch": 3.4437977446344123, + "grad_norm": 0.6756979625666991, + "learning_rate": 2.4034221303281328e-05, + "loss": 0.1732, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11715242266654968, + "step": 4735, + "valid_targets_mean": 3273.2, + "valid_targets_min": 442 + }, + { + "epoch": 3.447435431065842, + "grad_norm": 0.359427188376704, + "learning_rate": 2.3998691514172726e-05, + "loss": 0.1693, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08864973485469818, + "step": 4740, + "valid_targets_mean": 8141.1, + "valid_targets_min": 5061 + }, + { + "epoch": 3.4510731174972715, + "grad_norm": 0.3843708927392185, + "learning_rate": 2.3963148575202665e-05, + "loss": 0.1788, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0900273472070694, + "step": 4745, + "valid_targets_mean": 7371.1, + "valid_targets_min": 5112 + }, + { + "epoch": 3.4547108039287013, + "grad_norm": 0.38790244382415523, + "learning_rate": 2.392759260325556e-05, + "loss": 0.178, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07887575030326843, + "step": 4750, + "valid_targets_mean": 5659.4, + "valid_targets_min": 4626 + }, + { + "epoch": 3.458348490360131, + "grad_norm": 0.36817952472018384, + "learning_rate": 2.389202371525869e-05, + "loss": 0.1847, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08105608820915222, + "step": 4755, + "valid_targets_mean": 7801.8, + "valid_targets_min": 6172 + }, + { + "epoch": 3.4619861767915605, + "grad_norm": 0.36355277167794897, + "learning_rate": 2.38564420281818e-05, + "loss": 0.1768, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08919825404882431, + "step": 4760, + "valid_targets_mean": 7547.4, + "valid_targets_min": 5804 + }, + { + "epoch": 3.4656238632229903, + "grad_norm": 0.38797775685880864, + "learning_rate": 2.382084765903674e-05, + "loss": 0.1682, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08065015077590942, + "step": 4765, + "valid_targets_mean": 6474.2, + "valid_targets_min": 5050 + }, + { + "epoch": 3.4692615496544197, + "grad_norm": 0.39168951158642135, + "learning_rate": 2.3785240724877054e-05, + "loss": 0.1579, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08171379566192627, + "step": 4770, + "valid_targets_mean": 7225.4, + "valid_targets_min": 5298 + }, + { + "epoch": 3.4728992360858495, + "grad_norm": 0.36624766941167125, + "learning_rate": 2.37496213427976e-05, + "loss": 0.1567, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07413630932569504, + "step": 4775, + "valid_targets_mean": 6442.9, + "valid_targets_min": 4679 + }, + { + "epoch": 3.476536922517279, + "grad_norm": 0.46309819156822746, + "learning_rate": 2.371398962993419e-05, + "loss": 0.1856, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07415753602981567, + "step": 4780, + "valid_targets_mean": 3651.9, + "valid_targets_min": 1801 + }, + { + "epoch": 3.4801746089487087, + "grad_norm": 0.4611966905534764, + "learning_rate": 2.3678345703463173e-05, + "loss": 0.177, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10450230538845062, + "step": 4785, + "valid_targets_mean": 7512.8, + "valid_targets_min": 5331 + }, + { + "epoch": 3.4838122953801385, + "grad_norm": 0.41071678724529975, + "learning_rate": 2.364268968060107e-05, + "loss": 0.1863, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09761945903301239, + "step": 4790, + "valid_targets_mean": 7108.8, + "valid_targets_min": 4822 + }, + { + "epoch": 3.487449981811568, + "grad_norm": 0.3808582970746324, + "learning_rate": 2.3607021678604173e-05, + "loss": 0.1913, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0842888206243515, + "step": 4795, + "valid_targets_mean": 7125.8, + "valid_targets_min": 4623 + }, + { + "epoch": 3.491087668242997, + "grad_norm": 0.47121975880533973, + "learning_rate": 2.357134181476818e-05, + "loss": 0.1887, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08456876873970032, + "step": 4800, + "valid_targets_mean": 6269.2, + "valid_targets_min": 5458 + }, + { + "epoch": 3.494725354674427, + "grad_norm": 0.3964586137169603, + "learning_rate": 2.3535650206427786e-05, + "loss": 0.1787, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0912339836359024, + "step": 4805, + "valid_targets_mean": 7455.1, + "valid_targets_min": 6008 + }, + { + "epoch": 3.498363041105857, + "grad_norm": 0.4089936207322316, + "learning_rate": 2.3499946970956308e-05, + "loss": 0.1888, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0928640365600586, + "step": 4810, + "valid_targets_mean": 7104.4, + "valid_targets_min": 5386 + }, + { + "epoch": 3.502000727537286, + "grad_norm": 0.3782522942606185, + "learning_rate": 2.3464232225765315e-05, + "loss": 0.1792, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09134906530380249, + "step": 4815, + "valid_targets_mean": 6532.6, + "valid_targets_min": 4855 + }, + { + "epoch": 3.505638413968716, + "grad_norm": 0.37433630113282457, + "learning_rate": 2.34285060883042e-05, + "loss": 0.1846, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09020563215017319, + "step": 4820, + "valid_targets_mean": 6743.9, + "valid_targets_min": 5613 + }, + { + "epoch": 3.5092761004001454, + "grad_norm": 0.3890842924390144, + "learning_rate": 2.3392768676059848e-05, + "loss": 0.1839, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09263330698013306, + "step": 4825, + "valid_targets_mean": 6373.6, + "valid_targets_min": 4961 + }, + { + "epoch": 3.512913786831575, + "grad_norm": 0.4042639316665441, + "learning_rate": 2.335702010655621e-05, + "loss": 0.1872, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09637884050607681, + "step": 4830, + "valid_targets_mean": 6461.2, + "valid_targets_min": 5286 + }, + { + "epoch": 3.5165514732630045, + "grad_norm": 0.3796096515237625, + "learning_rate": 2.3321260497353918e-05, + "loss": 0.1828, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09597183763980865, + "step": 4835, + "valid_targets_mean": 6780.5, + "valid_targets_min": 5201 + }, + { + "epoch": 3.5201891596944344, + "grad_norm": 0.3977257817245041, + "learning_rate": 2.3285489966049922e-05, + "loss": 0.1816, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09119761735200882, + "step": 4840, + "valid_targets_mean": 5710.8, + "valid_targets_min": 4838 + }, + { + "epoch": 3.523826846125864, + "grad_norm": 1.3437144391381461, + "learning_rate": 2.3249708630277078e-05, + "loss": 0.1514, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13297277688980103, + "step": 4845, + "valid_targets_mean": 3235.6, + "valid_targets_min": 176 + }, + { + "epoch": 3.5274645325572935, + "grad_norm": 0.4097847298160484, + "learning_rate": 2.3213916607703787e-05, + "loss": 0.1839, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0841015875339508, + "step": 4850, + "valid_targets_mean": 6674.8, + "valid_targets_min": 4582 + }, + { + "epoch": 3.531102218988723, + "grad_norm": 0.3904456791045362, + "learning_rate": 2.3178114016033587e-05, + "loss": 0.1785, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08899042010307312, + "step": 4855, + "valid_targets_mean": 6157.8, + "valid_targets_min": 5385 + }, + { + "epoch": 3.5347399054201527, + "grad_norm": 0.40074076801870795, + "learning_rate": 2.3142300973004775e-05, + "loss": 0.1813, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08565524220466614, + "step": 4860, + "valid_targets_mean": 6006.2, + "valid_targets_min": 4478 + }, + { + "epoch": 3.5383775918515825, + "grad_norm": 0.410279050721928, + "learning_rate": 2.3106477596390003e-05, + "loss": 0.1777, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08927357196807861, + "step": 4865, + "valid_targets_mean": 5893.4, + "valid_targets_min": 3911 + }, + { + "epoch": 3.542015278283012, + "grad_norm": 0.40822052653192054, + "learning_rate": 2.307064400399593e-05, + "loss": 0.1774, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0881558358669281, + "step": 4870, + "valid_targets_mean": 6134.2, + "valid_targets_min": 5318 + }, + { + "epoch": 3.5456529647144417, + "grad_norm": 0.4150001913163252, + "learning_rate": 2.3034800313662794e-05, + "loss": 0.1858, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09278759360313416, + "step": 4875, + "valid_targets_mean": 5474.4, + "valid_targets_min": 4959 + }, + { + "epoch": 3.5492906511458715, + "grad_norm": 0.54802474045807, + "learning_rate": 2.2998946643264044e-05, + "loss": 0.177, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05490536615252495, + "step": 4880, + "valid_targets_mean": 1804.1, + "valid_targets_min": 842 + }, + { + "epoch": 3.552928337577301, + "grad_norm": 0.49559910633324156, + "learning_rate": 2.2963083110705953e-05, + "loss": 0.1518, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08161012828350067, + "step": 4885, + "valid_targets_mean": 6118.8, + "valid_targets_min": 4899 + }, + { + "epoch": 3.5565660240087302, + "grad_norm": 0.4000308821431107, + "learning_rate": 2.2927209833927217e-05, + "loss": 0.1747, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08208855241537094, + "step": 4890, + "valid_targets_mean": 6773.8, + "valid_targets_min": 5046 + }, + { + "epoch": 3.56020371044016, + "grad_norm": 0.4172416706504058, + "learning_rate": 2.289132693089859e-05, + "loss": 0.1854, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0945064127445221, + "step": 4895, + "valid_targets_mean": 6633.6, + "valid_targets_min": 5284 + }, + { + "epoch": 3.56384139687159, + "grad_norm": 0.3952928605756737, + "learning_rate": 2.2855434519622456e-05, + "loss": 0.177, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09049314260482788, + "step": 4900, + "valid_targets_mean": 6055.2, + "valid_targets_min": 4928 + }, + { + "epoch": 3.567479083303019, + "grad_norm": 0.3919483571715627, + "learning_rate": 2.2819532718132508e-05, + "loss": 0.173, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09322609752416611, + "step": 4905, + "valid_targets_mean": 7259.4, + "valid_targets_min": 5758 + }, + { + "epoch": 3.571116769734449, + "grad_norm": 0.45974398369608954, + "learning_rate": 2.2783621644493286e-05, + "loss": 0.1772, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09171155840158463, + "step": 4910, + "valid_targets_mean": 5731.8, + "valid_targets_min": 4629 + }, + { + "epoch": 3.5747544561658784, + "grad_norm": 0.710690506362282, + "learning_rate": 2.2747701416799823e-05, + "loss": 0.2021, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18977713584899902, + "step": 4915, + "valid_targets_mean": 5360.9, + "valid_targets_min": 2317 + }, + { + "epoch": 3.578392142597308, + "grad_norm": 0.5133326835100587, + "learning_rate": 2.2711772153177284e-05, + "loss": 0.2512, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13477376103401184, + "step": 4920, + "valid_targets_mean": 5303.1, + "valid_targets_min": 533 + }, + { + "epoch": 3.5820298290287376, + "grad_norm": 0.5652893069154128, + "learning_rate": 2.267583397178051e-05, + "loss": 0.2464, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1277260035276413, + "step": 4925, + "valid_targets_mean": 4545.8, + "valid_targets_min": 835 + }, + { + "epoch": 3.5856675154601674, + "grad_norm": 0.6429924695882341, + "learning_rate": 2.2639886990793705e-05, + "loss": 0.2419, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13330325484275818, + "step": 4930, + "valid_targets_mean": 3618.9, + "valid_targets_min": 1201 + }, + { + "epoch": 3.589305201891597, + "grad_norm": 0.5626044555601487, + "learning_rate": 2.2603931328429982e-05, + "loss": 0.2307, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10827826708555222, + "step": 4935, + "valid_targets_mean": 4253.5, + "valid_targets_min": 1880 + }, + { + "epoch": 3.5929428883230266, + "grad_norm": 0.5268410293992923, + "learning_rate": 2.2567967102931025e-05, + "loss": 0.2551, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17985832691192627, + "step": 4940, + "valid_targets_mean": 6610.1, + "valid_targets_min": 2171 + }, + { + "epoch": 3.596580574754456, + "grad_norm": 0.4615436587790331, + "learning_rate": 2.253199443256666e-05, + "loss": 0.2332, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13218903541564941, + "step": 4945, + "valid_targets_mean": 7171.4, + "valid_targets_min": 4500 + }, + { + "epoch": 3.6002182611858857, + "grad_norm": 0.5112047182579162, + "learning_rate": 2.249601343563449e-05, + "loss": 0.2398, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12573397159576416, + "step": 4950, + "valid_targets_mean": 6207.4, + "valid_targets_min": 1855 + }, + { + "epoch": 3.6038559476173155, + "grad_norm": 0.552808766051254, + "learning_rate": 2.2460024230459522e-05, + "loss": 0.2382, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12190352380275726, + "step": 4955, + "valid_targets_mean": 4075.9, + "valid_targets_min": 1024 + }, + { + "epoch": 3.607493634048745, + "grad_norm": 0.5097494997680447, + "learning_rate": 2.242402693539371e-05, + "loss": 0.249, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10620151460170746, + "step": 4960, + "valid_targets_mean": 4767.1, + "valid_targets_min": 2107 + }, + { + "epoch": 3.6111313204801747, + "grad_norm": 0.5677885659284854, + "learning_rate": 2.2388021668815657e-05, + "loss": 0.2444, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13467761874198914, + "step": 4965, + "valid_targets_mean": 4211.4, + "valid_targets_min": 802 + }, + { + "epoch": 3.614769006911604, + "grad_norm": 0.44513655251968914, + "learning_rate": 2.2352008549130156e-05, + "loss": 0.2343, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12193802744150162, + "step": 4970, + "valid_targets_mean": 5345.1, + "valid_targets_min": 1597 + }, + { + "epoch": 3.618406693343034, + "grad_norm": 0.6069340528692686, + "learning_rate": 2.2315987694767837e-05, + "loss": 0.2474, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12675432860851288, + "step": 4975, + "valid_targets_mean": 3671.4, + "valid_targets_min": 1650 + }, + { + "epoch": 3.6220443797744633, + "grad_norm": 0.5193269537131178, + "learning_rate": 2.2279959224184754e-05, + "loss": 0.2373, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12448069453239441, + "step": 4980, + "valid_targets_mean": 4799.1, + "valid_targets_min": 1385 + }, + { + "epoch": 3.625682066205893, + "grad_norm": 0.5831449000886, + "learning_rate": 2.2243923255862025e-05, + "loss": 0.2348, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12927505373954773, + "step": 4985, + "valid_targets_mean": 4023.5, + "valid_targets_min": 2908 + }, + { + "epoch": 3.629319752637323, + "grad_norm": 0.5123832213066304, + "learning_rate": 2.22078799083054e-05, + "loss": 0.2417, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10974849760532379, + "step": 4990, + "valid_targets_mean": 5064.6, + "valid_targets_min": 2785 + }, + { + "epoch": 3.6329574390687522, + "grad_norm": 0.6099028740108954, + "learning_rate": 2.217182930004491e-05, + "loss": 0.2331, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11671056598424911, + "step": 4995, + "valid_targets_mean": 3886.4, + "valid_targets_min": 1210 + }, + { + "epoch": 3.636595125500182, + "grad_norm": 0.591591406488122, + "learning_rate": 2.213577154963448e-05, + "loss": 0.2403, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1341969072818756, + "step": 5000, + "valid_targets_mean": 3708.8, + "valid_targets_min": 1064 + }, + { + "epoch": 3.6402328119316114, + "grad_norm": 0.6649186258830714, + "learning_rate": 2.209970677565148e-05, + "loss": 0.2511, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12801742553710938, + "step": 5005, + "valid_targets_mean": 3276.2, + "valid_targets_min": 1678 + }, + { + "epoch": 3.643870498363041, + "grad_norm": 0.6021068479256197, + "learning_rate": 2.2063635096696426e-05, + "loss": 0.2433, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12707066535949707, + "step": 5010, + "valid_targets_mean": 3520.1, + "valid_targets_min": 671 + }, + { + "epoch": 3.6475081847944706, + "grad_norm": 0.6232625537000929, + "learning_rate": 2.2027556631392496e-05, + "loss": 0.2398, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11610487848520279, + "step": 5015, + "valid_targets_mean": 3488.5, + "valid_targets_min": 2000 + }, + { + "epoch": 3.6511458712259004, + "grad_norm": 0.6011088020931473, + "learning_rate": 2.1991471498385226e-05, + "loss": 0.224, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13889220356941223, + "step": 5020, + "valid_targets_mean": 4402.6, + "valid_targets_min": 1532 + }, + { + "epoch": 3.65478355765733, + "grad_norm": 0.5765964476295736, + "learning_rate": 2.1955379816342046e-05, + "loss": 0.2314, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08994786441326141, + "step": 5025, + "valid_targets_mean": 3749.4, + "valid_targets_min": 1051 + }, + { + "epoch": 3.6584212440887596, + "grad_norm": 0.6235936283198645, + "learning_rate": 2.191928170395194e-05, + "loss": 0.2238, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11185075342655182, + "step": 5030, + "valid_targets_mean": 3368.1, + "valid_targets_min": 1345 + }, + { + "epoch": 3.662058930520189, + "grad_norm": 0.6307460865444124, + "learning_rate": 2.1883177279925034e-05, + "loss": 0.2405, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11603544652462006, + "step": 5035, + "valid_targets_mean": 4128.9, + "valid_targets_min": 2231 + }, + { + "epoch": 3.6656966169516187, + "grad_norm": 0.6531826589060592, + "learning_rate": 2.184706666299222e-05, + "loss": 0.2421, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10511721670627594, + "step": 5040, + "valid_targets_mean": 3672.6, + "valid_targets_min": 2127 + }, + { + "epoch": 3.6693343033830486, + "grad_norm": 0.6783809103696131, + "learning_rate": 2.1810949971904732e-05, + "loss": 0.2294, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10793808102607727, + "step": 5045, + "valid_targets_mean": 3006.8, + "valid_targets_min": 1265 + }, + { + "epoch": 3.672971989814478, + "grad_norm": 0.6175902377019231, + "learning_rate": 2.17748273254338e-05, + "loss": 0.2299, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09916140139102936, + "step": 5050, + "valid_targets_mean": 2981.9, + "valid_targets_min": 908 + }, + { + "epoch": 3.6766096762459077, + "grad_norm": 0.6350392809125993, + "learning_rate": 2.1738698842370234e-05, + "loss": 0.2305, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1317286193370819, + "step": 5055, + "valid_targets_mean": 3147.5, + "valid_targets_min": 1264 + }, + { + "epoch": 3.680247362677337, + "grad_norm": 0.6425101503514785, + "learning_rate": 2.170256464152404e-05, + "loss": 0.2292, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10621222853660583, + "step": 5060, + "valid_targets_mean": 2517.9, + "valid_targets_min": 926 + }, + { + "epoch": 3.683885049108767, + "grad_norm": 0.6472187558968999, + "learning_rate": 2.166642484172401e-05, + "loss": 0.2326, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.114153191447258, + "step": 5065, + "valid_targets_mean": 3204.5, + "valid_targets_min": 1016 + }, + { + "epoch": 3.6875227355401963, + "grad_norm": 0.6173876662089426, + "learning_rate": 2.1630279561817375e-05, + "loss": 0.2265, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11225185543298721, + "step": 5070, + "valid_targets_mean": 3525.0, + "valid_targets_min": 2019 + }, + { + "epoch": 3.691160421971626, + "grad_norm": 0.644244791895765, + "learning_rate": 2.159412892066937e-05, + "loss": 0.2314, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1372176706790924, + "step": 5075, + "valid_targets_mean": 4745.4, + "valid_targets_min": 3023 + }, + { + "epoch": 3.694798108403056, + "grad_norm": 0.6022243512102662, + "learning_rate": 2.155797303716286e-05, + "loss": 0.2239, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10248969495296478, + "step": 5080, + "valid_targets_mean": 3503.6, + "valid_targets_min": 1110 + }, + { + "epoch": 3.6984357948344853, + "grad_norm": 0.6191575815101891, + "learning_rate": 2.152181203019796e-05, + "loss": 0.2277, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13408900797367096, + "step": 5085, + "valid_targets_mean": 4206.5, + "valid_targets_min": 1460 + }, + { + "epoch": 3.7020734812659146, + "grad_norm": 0.607144452145269, + "learning_rate": 2.1485646018691627e-05, + "loss": 0.2287, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1244189441204071, + "step": 5090, + "valid_targets_mean": 3938.6, + "valid_targets_min": 2558 + }, + { + "epoch": 3.7057111676973444, + "grad_norm": 0.5877802885650723, + "learning_rate": 2.1449475121577268e-05, + "loss": 0.2392, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10661394149065018, + "step": 5095, + "valid_targets_mean": 3421.1, + "valid_targets_min": 1673 + }, + { + "epoch": 3.7093488541287742, + "grad_norm": 0.6412977238944128, + "learning_rate": 2.141329945780439e-05, + "loss": 0.2279, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10688023269176483, + "step": 5100, + "valid_targets_mean": 3549.6, + "valid_targets_min": 1860 + }, + { + "epoch": 3.7129865405602036, + "grad_norm": 0.5359037796348759, + "learning_rate": 2.137711914633812e-05, + "loss": 0.2273, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1073889434337616, + "step": 5105, + "valid_targets_mean": 4414.0, + "valid_targets_min": 2495 + }, + { + "epoch": 3.7166242269916334, + "grad_norm": 0.6348911730747517, + "learning_rate": 2.134093430615892e-05, + "loss": 0.2277, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12980222702026367, + "step": 5110, + "valid_targets_mean": 3861.9, + "valid_targets_min": 1945 + }, + { + "epoch": 3.720261913423063, + "grad_norm": 0.5981008065006139, + "learning_rate": 2.1304745056262123e-05, + "loss": 0.2363, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11580340564250946, + "step": 5115, + "valid_targets_mean": 4488.5, + "valid_targets_min": 1908 + }, + { + "epoch": 3.7238995998544926, + "grad_norm": 0.5872512131500627, + "learning_rate": 2.1268551515657552e-05, + "loss": 0.228, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11840249598026276, + "step": 5120, + "valid_targets_mean": 3507.5, + "valid_targets_min": 1880 + }, + { + "epoch": 3.727537286285922, + "grad_norm": 0.58891431783559, + "learning_rate": 2.123235380336917e-05, + "loss": 0.2242, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11315446346998215, + "step": 5125, + "valid_targets_mean": 4491.5, + "valid_targets_min": 2492 + }, + { + "epoch": 3.7311749727173518, + "grad_norm": 0.6001308463224355, + "learning_rate": 2.1196152038434632e-05, + "loss": 0.2274, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11253811419010162, + "step": 5130, + "valid_targets_mean": 4228.2, + "valid_targets_min": 3004 + }, + { + "epoch": 3.7348126591487816, + "grad_norm": 0.6583542772471446, + "learning_rate": 2.1159946339904936e-05, + "loss": 0.2213, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11470526456832886, + "step": 5135, + "valid_targets_mean": 3899.9, + "valid_targets_min": 1833 + }, + { + "epoch": 3.738450345580211, + "grad_norm": 0.6563178530501849, + "learning_rate": 2.1123736826844003e-05, + "loss": 0.225, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11485277116298676, + "step": 5140, + "valid_targets_mean": 3114.2, + "valid_targets_min": 1182 + }, + { + "epoch": 3.7420880320116408, + "grad_norm": 0.5810101541745151, + "learning_rate": 2.1087523618328313e-05, + "loss": 0.2276, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11349155753850937, + "step": 5145, + "valid_targets_mean": 4926.9, + "valid_targets_min": 1869 + }, + { + "epoch": 3.74572571844307, + "grad_norm": 0.5749063681681545, + "learning_rate": 2.105130683344649e-05, + "loss": 0.2283, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1340569257736206, + "step": 5150, + "valid_targets_mean": 4382.4, + "valid_targets_min": 2498 + }, + { + "epoch": 3.7493634048745, + "grad_norm": 0.669961642593916, + "learning_rate": 2.1015086591298918e-05, + "loss": 0.2416, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11251956224441528, + "step": 5155, + "valid_targets_mean": 3098.8, + "valid_targets_min": 988 + }, + { + "epoch": 3.7530010913059293, + "grad_norm": 0.6070387288037116, + "learning_rate": 2.0978863010997356e-05, + "loss": 0.2238, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11906497180461884, + "step": 5160, + "valid_targets_mean": 3733.5, + "valid_targets_min": 1534 + }, + { + "epoch": 3.756638777737359, + "grad_norm": 0.5502500561359225, + "learning_rate": 2.0942636211664528e-05, + "loss": 0.23, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10084741562604904, + "step": 5165, + "valid_targets_mean": 4108.1, + "valid_targets_min": 2252 + }, + { + "epoch": 3.760276464168789, + "grad_norm": 0.7462263044089149, + "learning_rate": 2.0906406312433764e-05, + "loss": 0.2319, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11911963671445847, + "step": 5170, + "valid_targets_mean": 3446.6, + "valid_targets_min": 1436 + }, + { + "epoch": 3.7639141506002183, + "grad_norm": 0.6412735661285742, + "learning_rate": 2.0870173432448566e-05, + "loss": 0.2289, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10536488890647888, + "step": 5175, + "valid_targets_mean": 3819.2, + "valid_targets_min": 2333 + }, + { + "epoch": 3.7675518370316476, + "grad_norm": 0.631669976720977, + "learning_rate": 2.0833937690862266e-05, + "loss": 0.2209, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10178577154874802, + "step": 5180, + "valid_targets_mean": 3889.8, + "valid_targets_min": 1272 + }, + { + "epoch": 3.7711895234630775, + "grad_norm": 0.6101916370646752, + "learning_rate": 2.0797699206837575e-05, + "loss": 0.2206, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11349606513977051, + "step": 5185, + "valid_targets_mean": 3233.0, + "valid_targets_min": 1319 + }, + { + "epoch": 3.7748272098945073, + "grad_norm": 0.5615223756888936, + "learning_rate": 2.076145809954625e-05, + "loss": 0.2085, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09186767041683197, + "step": 5190, + "valid_targets_mean": 3186.6, + "valid_targets_min": 1194 + }, + { + "epoch": 3.7784648963259366, + "grad_norm": 0.650305290415346, + "learning_rate": 2.0725214488168646e-05, + "loss": 0.242, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10046033561229706, + "step": 5195, + "valid_targets_mean": 2710.6, + "valid_targets_min": 1453 + }, + { + "epoch": 3.7821025827573664, + "grad_norm": 0.5200111863606415, + "learning_rate": 2.068896849189338e-05, + "loss": 0.2233, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09310400485992432, + "step": 5200, + "valid_targets_mean": 5344.5, + "valid_targets_min": 3603 + }, + { + "epoch": 3.785740269188796, + "grad_norm": 0.5849221256333798, + "learning_rate": 2.0652720229916906e-05, + "loss": 0.235, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10216125845909119, + "step": 5205, + "valid_targets_mean": 3067.5, + "valid_targets_min": 1256 + }, + { + "epoch": 3.7893779556202256, + "grad_norm": 0.6535662092215815, + "learning_rate": 2.0616469821443116e-05, + "loss": 0.2296, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15201467275619507, + "step": 5210, + "valid_targets_mean": 4443.9, + "valid_targets_min": 1336 + }, + { + "epoch": 3.793015642051655, + "grad_norm": 0.6394944714524744, + "learning_rate": 2.058021738568297e-05, + "loss": 0.2376, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11878855526447296, + "step": 5215, + "valid_targets_mean": 3581.8, + "valid_targets_min": 1275 + }, + { + "epoch": 3.796653328483085, + "grad_norm": 0.6133697086011416, + "learning_rate": 2.05439630418541e-05, + "loss": 0.2277, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12785664200782776, + "step": 5220, + "valid_targets_mean": 3770.1, + "valid_targets_min": 2202 + }, + { + "epoch": 3.8002910149145146, + "grad_norm": 0.5881174719264395, + "learning_rate": 2.0507706909180395e-05, + "loss": 0.2206, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1308748722076416, + "step": 5225, + "valid_targets_mean": 4112.8, + "valid_targets_min": 3332 + }, + { + "epoch": 3.803928701345944, + "grad_norm": 0.6357644736215303, + "learning_rate": 2.0471449106891644e-05, + "loss": 0.2197, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11513872444629669, + "step": 5230, + "valid_targets_mean": 3260.0, + "valid_targets_min": 2195 + }, + { + "epoch": 3.8075663877773733, + "grad_norm": 0.6285686305488973, + "learning_rate": 2.043518975422312e-05, + "loss": 0.22, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10969813168048859, + "step": 5235, + "valid_targets_mean": 4039.6, + "valid_targets_min": 2343 + }, + { + "epoch": 3.811204074208803, + "grad_norm": 0.581098015300979, + "learning_rate": 2.03989289704152e-05, + "loss": 0.2136, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09730441868305206, + "step": 5240, + "valid_targets_mean": 2678.8, + "valid_targets_min": 1026 + }, + { + "epoch": 3.814841760640233, + "grad_norm": 0.6428189900891887, + "learning_rate": 2.036266687471295e-05, + "loss": 0.2168, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10830096900463104, + "step": 5245, + "valid_targets_mean": 3754.5, + "valid_targets_min": 1042 + }, + { + "epoch": 3.8184794470716623, + "grad_norm": 0.6326576667741963, + "learning_rate": 2.0326403586365782e-05, + "loss": 0.2046, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10399054735898972, + "step": 5250, + "valid_targets_mean": 3301.6, + "valid_targets_min": 1274 + }, + { + "epoch": 3.822117133503092, + "grad_norm": 0.6848779545669462, + "learning_rate": 2.029013922462699e-05, + "loss": 0.222, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11134923249483109, + "step": 5255, + "valid_targets_mean": 3632.2, + "valid_targets_min": 914 + }, + { + "epoch": 3.8257548199345215, + "grad_norm": 0.5641317007637904, + "learning_rate": 2.025387390875343e-05, + "loss": 0.2457, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10083137452602386, + "step": 5260, + "valid_targets_mean": 4342.5, + "valid_targets_min": 2093 + }, + { + "epoch": 3.8293925063659513, + "grad_norm": 0.7137370248936561, + "learning_rate": 2.021760775800509e-05, + "loss": 0.2246, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11637339740991592, + "step": 5265, + "valid_targets_mean": 3609.6, + "valid_targets_min": 1683 + }, + { + "epoch": 3.8330301927973807, + "grad_norm": 0.5525803549326115, + "learning_rate": 2.0181340891644677e-05, + "loss": 0.2344, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10778437554836273, + "step": 5270, + "valid_targets_mean": 4096.2, + "valid_targets_min": 2253 + }, + { + "epoch": 3.8366678792288105, + "grad_norm": 0.5648852778175543, + "learning_rate": 2.0145073428937296e-05, + "loss": 0.2242, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1065383180975914, + "step": 5275, + "valid_targets_mean": 3794.4, + "valid_targets_min": 1789 + }, + { + "epoch": 3.8403055656602403, + "grad_norm": 0.6024413851522801, + "learning_rate": 2.0108805489149977e-05, + "loss": 0.2275, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11417102068662643, + "step": 5280, + "valid_targets_mean": 3753.8, + "valid_targets_min": 1586 + }, + { + "epoch": 3.8439432520916696, + "grad_norm": 0.6445868752209662, + "learning_rate": 2.0072537191551333e-05, + "loss": 0.2235, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10088599473237991, + "step": 5285, + "valid_targets_mean": 2995.5, + "valid_targets_min": 2033 + }, + { + "epoch": 3.8475809385230995, + "grad_norm": 0.7162815669131108, + "learning_rate": 2.0036268655411146e-05, + "loss": 0.2121, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10161931812763214, + "step": 5290, + "valid_targets_mean": 3690.5, + "valid_targets_min": 1616 + }, + { + "epoch": 3.851218624954529, + "grad_norm": 0.6056969717857391, + "learning_rate": 2e-05, + "loss": 0.2272, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09984216094017029, + "step": 5295, + "valid_targets_mean": 3303.9, + "valid_targets_min": 2000 + }, + { + "epoch": 3.8548563113859586, + "grad_norm": 0.6113868829658972, + "learning_rate": 1.9963731344588857e-05, + "loss": 0.2186, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12194693088531494, + "step": 5300, + "valid_targets_mean": 3788.9, + "valid_targets_min": 1750 + }, + { + "epoch": 3.858493997817388, + "grad_norm": 0.65874088711381, + "learning_rate": 1.9927462808448673e-05, + "loss": 0.2259, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1281115710735321, + "step": 5305, + "valid_targets_mean": 3898.2, + "valid_targets_min": 2126 + }, + { + "epoch": 3.862131684248818, + "grad_norm": 0.6433800456506442, + "learning_rate": 1.9891194510850033e-05, + "loss": 0.2369, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12369626760482788, + "step": 5310, + "valid_targets_mean": 3923.1, + "valid_targets_min": 3049 + }, + { + "epoch": 3.8657693706802476, + "grad_norm": 0.5909855743873148, + "learning_rate": 1.9854926571062704e-05, + "loss": 0.226, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1302732229232788, + "step": 5315, + "valid_targets_mean": 4337.1, + "valid_targets_min": 1571 + }, + { + "epoch": 3.869407057111677, + "grad_norm": 0.6534087137908258, + "learning_rate": 1.9818659108355326e-05, + "loss": 0.2385, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10928314924240112, + "step": 5320, + "valid_targets_mean": 2998.0, + "valid_targets_min": 1487 + }, + { + "epoch": 3.8730447435431063, + "grad_norm": 0.5892970944694524, + "learning_rate": 1.9782392241994918e-05, + "loss": 0.2291, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09445537626743317, + "step": 5325, + "valid_targets_mean": 3627.1, + "valid_targets_min": 1970 + }, + { + "epoch": 3.876682429974536, + "grad_norm": 0.5941960008609036, + "learning_rate": 1.9746126091246575e-05, + "loss": 0.2401, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10478130728006363, + "step": 5330, + "valid_targets_mean": 4038.0, + "valid_targets_min": 2492 + }, + { + "epoch": 3.880320116405966, + "grad_norm": 0.6884819828024702, + "learning_rate": 1.9709860775373014e-05, + "loss": 0.2377, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10003891587257385, + "step": 5335, + "valid_targets_mean": 3211.2, + "valid_targets_min": 1066 + }, + { + "epoch": 3.8839578028373953, + "grad_norm": 0.5896001517324118, + "learning_rate": 1.9673596413634225e-05, + "loss": 0.2356, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10323205590248108, + "step": 5340, + "valid_targets_mean": 3401.9, + "valid_targets_min": 2281 + }, + { + "epoch": 3.887595489268825, + "grad_norm": 0.6504774562657795, + "learning_rate": 1.9637333125287055e-05, + "loss": 0.2346, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11640578508377075, + "step": 5345, + "valid_targets_mean": 3662.2, + "valid_targets_min": 1969 + }, + { + "epoch": 3.8912331757002545, + "grad_norm": 0.6850678010926895, + "learning_rate": 1.9601071029584803e-05, + "loss": 0.2293, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10384529829025269, + "step": 5350, + "valid_targets_mean": 3189.6, + "valid_targets_min": 1005 + }, + { + "epoch": 3.8948708621316843, + "grad_norm": 0.6408542876925832, + "learning_rate": 1.9564810245776883e-05, + "loss": 0.2302, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10389432311058044, + "step": 5355, + "valid_targets_mean": 3546.5, + "valid_targets_min": 1980 + }, + { + "epoch": 3.8985085485631137, + "grad_norm": 0.5882278099988311, + "learning_rate": 1.9528550893108363e-05, + "loss": 0.233, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13294747471809387, + "step": 5360, + "valid_targets_mean": 4803.1, + "valid_targets_min": 1257 + }, + { + "epoch": 3.9021462349945435, + "grad_norm": 0.6312848625699897, + "learning_rate": 1.949229309081961e-05, + "loss": 0.2227, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11631964892148972, + "step": 5365, + "valid_targets_mean": 3118.1, + "valid_targets_min": 1357 + }, + { + "epoch": 3.9057839214259733, + "grad_norm": 0.5407214889945463, + "learning_rate": 1.9456036958145906e-05, + "loss": 0.2123, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09496937692165375, + "step": 5370, + "valid_targets_mean": 4463.8, + "valid_targets_min": 1410 + }, + { + "epoch": 3.9094216078574027, + "grad_norm": 0.6375817087254536, + "learning_rate": 1.941978261431703e-05, + "loss": 0.2303, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10226564854383469, + "step": 5375, + "valid_targets_mean": 3429.8, + "valid_targets_min": 1390 + }, + { + "epoch": 3.913059294288832, + "grad_norm": 0.6625481483545319, + "learning_rate": 1.9383530178556887e-05, + "loss": 0.2177, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09840551018714905, + "step": 5380, + "valid_targets_mean": 3173.0, + "valid_targets_min": 790 + }, + { + "epoch": 3.916696980720262, + "grad_norm": 0.6034232205538901, + "learning_rate": 1.934727977008309e-05, + "loss": 0.2386, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12093979865312576, + "step": 5385, + "valid_targets_mean": 3788.8, + "valid_targets_min": 1901 + }, + { + "epoch": 3.9203346671516917, + "grad_norm": 0.6497470901165474, + "learning_rate": 1.9311031508106622e-05, + "loss": 0.2256, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10182322561740875, + "step": 5390, + "valid_targets_mean": 2697.1, + "valid_targets_min": 1460 + }, + { + "epoch": 3.923972353583121, + "grad_norm": 0.6344905744894002, + "learning_rate": 1.927478551183136e-05, + "loss": 0.227, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13488560914993286, + "step": 5395, + "valid_targets_mean": 5088.8, + "valid_targets_min": 952 + }, + { + "epoch": 3.927610040014551, + "grad_norm": 0.7141722186240833, + "learning_rate": 1.9238541900453758e-05, + "loss": 0.2265, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11794304847717285, + "step": 5400, + "valid_targets_mean": 3070.5, + "valid_targets_min": 1768 + }, + { + "epoch": 3.93124772644598, + "grad_norm": 0.5927828253997296, + "learning_rate": 1.920230079316243e-05, + "loss": 0.2289, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09280972927808762, + "step": 5405, + "valid_targets_mean": 3585.0, + "valid_targets_min": 1931 + }, + { + "epoch": 3.93488541287741, + "grad_norm": 0.5672265755880941, + "learning_rate": 1.9166062309137737e-05, + "loss": 0.2216, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1056184470653534, + "step": 5410, + "valid_targets_mean": 5915.4, + "valid_targets_min": 1233 + }, + { + "epoch": 3.9385230993088394, + "grad_norm": 0.5446257054494565, + "learning_rate": 1.912982656755144e-05, + "loss": 0.2165, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10914193838834763, + "step": 5415, + "valid_targets_mean": 5310.6, + "valid_targets_min": 1741 + }, + { + "epoch": 3.942160785740269, + "grad_norm": 0.5607713663661205, + "learning_rate": 1.9093593687566246e-05, + "loss": 0.2187, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1028391420841217, + "step": 5420, + "valid_targets_mean": 3884.8, + "valid_targets_min": 856 + }, + { + "epoch": 3.945798472171699, + "grad_norm": 0.5947764995372521, + "learning_rate": 1.905736378833548e-05, + "loss": 0.2203, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11742674559354782, + "step": 5425, + "valid_targets_mean": 3649.5, + "valid_targets_min": 1062 + }, + { + "epoch": 3.9494361586031284, + "grad_norm": 0.5518894829787054, + "learning_rate": 1.9021136989002654e-05, + "loss": 0.2129, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10768945515155792, + "step": 5430, + "valid_targets_mean": 4412.6, + "valid_targets_min": 974 + }, + { + "epoch": 3.953073845034558, + "grad_norm": 0.6401698617515541, + "learning_rate": 1.8984913408701085e-05, + "loss": 0.2156, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11455243825912476, + "step": 5435, + "valid_targets_mean": 3077.5, + "valid_targets_min": 1447 + }, + { + "epoch": 3.9567115314659875, + "grad_norm": 0.5925829578924068, + "learning_rate": 1.8948693166553516e-05, + "loss": 0.2138, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1244969442486763, + "step": 5440, + "valid_targets_mean": 4663.0, + "valid_targets_min": 2144 + }, + { + "epoch": 3.9603492178974173, + "grad_norm": 0.621680409575374, + "learning_rate": 1.8912476381671687e-05, + "loss": 0.2128, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12215929478406906, + "step": 5445, + "valid_targets_mean": 4543.1, + "valid_targets_min": 1515 + }, + { + "epoch": 3.9639869043288467, + "grad_norm": 0.6973064119070888, + "learning_rate": 1.8876263173156004e-05, + "loss": 0.2327, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11164948344230652, + "step": 5450, + "valid_targets_mean": 3034.9, + "valid_targets_min": 1349 + }, + { + "epoch": 3.9676245907602765, + "grad_norm": 0.6361292754938587, + "learning_rate": 1.8840053660095074e-05, + "loss": 0.2298, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11279813945293427, + "step": 5455, + "valid_targets_mean": 3150.4, + "valid_targets_min": 1906 + }, + { + "epoch": 3.9712622771917063, + "grad_norm": 0.6494360834232921, + "learning_rate": 1.8803847961565374e-05, + "loss": 0.2171, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11465411633253098, + "step": 5460, + "valid_targets_mean": 3655.1, + "valid_targets_min": 2076 + }, + { + "epoch": 3.9748999636231357, + "grad_norm": 0.5915214876039027, + "learning_rate": 1.8767646196630837e-05, + "loss": 0.2136, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0917111337184906, + "step": 5465, + "valid_targets_mean": 2994.0, + "valid_targets_min": 1021 + }, + { + "epoch": 3.978537650054565, + "grad_norm": 0.6891408596032598, + "learning_rate": 1.8731448484342448e-05, + "loss": 0.2235, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11071023344993591, + "step": 5470, + "valid_targets_mean": 3226.9, + "valid_targets_min": 654 + }, + { + "epoch": 3.982175336485995, + "grad_norm": 0.590136406372101, + "learning_rate": 1.8695254943737887e-05, + "loss": 0.2175, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1263342648744583, + "step": 5475, + "valid_targets_mean": 4790.8, + "valid_targets_min": 3074 + }, + { + "epoch": 3.9858130229174247, + "grad_norm": 0.7116339755943882, + "learning_rate": 1.865906569384109e-05, + "loss": 0.2321, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11262650787830353, + "step": 5480, + "valid_targets_mean": 2604.0, + "valid_targets_min": 815 + }, + { + "epoch": 3.989450709348854, + "grad_norm": 0.6088766245010181, + "learning_rate": 1.8622880853661888e-05, + "loss": 0.2226, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10210122168064117, + "step": 5485, + "valid_targets_mean": 3560.0, + "valid_targets_min": 2205 + }, + { + "epoch": 3.993088395780284, + "grad_norm": 0.6003039601430649, + "learning_rate": 1.858670054219562e-05, + "loss": 0.2294, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12317419052124023, + "step": 5490, + "valid_targets_mean": 4216.0, + "valid_targets_min": 1793 + }, + { + "epoch": 3.996726082211713, + "grad_norm": 0.6405604346878275, + "learning_rate": 1.855052487842273e-05, + "loss": 0.2283, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10642292350530624, + "step": 5495, + "valid_targets_mean": 3134.5, + "valid_targets_min": 1217 + }, + { + "epoch": 4.0, + "grad_norm": 0.7854243841604341, + "learning_rate": 1.8514353981308383e-05, + "loss": 0.2039, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10046437382698059, + "step": 5500, + "valid_targets_mean": 1910.0, + "valid_targets_min": 393 + }, + { + "epoch": 4.003637686431429, + "grad_norm": 0.47508137807009887, + "learning_rate": 1.8478187969802045e-05, + "loss": 0.1159, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04844546318054199, + "step": 5505, + "valid_targets_mean": 2722.6, + "valid_targets_min": 458 + }, + { + "epoch": 4.00727537286286, + "grad_norm": 0.43177404184069323, + "learning_rate": 1.8442026962837145e-05, + "loss": 0.0909, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.045549578964710236, + "step": 5510, + "valid_targets_mean": 3113.1, + "valid_targets_min": 2461 + }, + { + "epoch": 4.010913059294289, + "grad_norm": 0.3729076604823543, + "learning_rate": 1.8405871079330643e-05, + "loss": 0.0825, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.048076242208480835, + "step": 5515, + "valid_targets_mean": 3648.1, + "valid_targets_min": 2068 + }, + { + "epoch": 4.014550745725718, + "grad_norm": 0.4947203422071337, + "learning_rate": 1.836972043818263e-05, + "loss": 0.1053, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04965710639953613, + "step": 5520, + "valid_targets_mean": 3423.6, + "valid_targets_min": 2425 + }, + { + "epoch": 4.018188432157148, + "grad_norm": 0.49205612598004717, + "learning_rate": 1.8333575158275996e-05, + "loss": 0.0838, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0421437993645668, + "step": 5525, + "valid_targets_mean": 2724.1, + "valid_targets_min": 1198 + }, + { + "epoch": 4.021826118588578, + "grad_norm": 0.7489115318073554, + "learning_rate": 1.8297435358475966e-05, + "loss": 0.0994, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0709318295121193, + "step": 5530, + "valid_targets_mean": 1232.5, + "valid_targets_min": 807 + }, + { + "epoch": 4.025463805020007, + "grad_norm": 0.477427840425181, + "learning_rate": 1.826130115762977e-05, + "loss": 0.1369, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05451330542564392, + "step": 5535, + "valid_targets_mean": 3487.4, + "valid_targets_min": 2351 + }, + { + "epoch": 4.029101491451437, + "grad_norm": 0.5056025418040886, + "learning_rate": 1.8225172674566204e-05, + "loss": 0.0928, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0718500167131424, + "step": 5540, + "valid_targets_mean": 3126.4, + "valid_targets_min": 781 + }, + { + "epoch": 4.032739177882866, + "grad_norm": 0.34132221671688934, + "learning_rate": 1.818905002809527e-05, + "loss": 0.1023, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04493570327758789, + "step": 5545, + "valid_targets_mean": 3474.8, + "valid_targets_min": 799 + }, + { + "epoch": 4.036376864314296, + "grad_norm": 0.3137542574431029, + "learning_rate": 1.8152933337007793e-05, + "loss": 0.1068, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04022137075662613, + "step": 5550, + "valid_targets_mean": 5026.1, + "valid_targets_min": 2802 + }, + { + "epoch": 4.040014550745726, + "grad_norm": 0.33940747208685307, + "learning_rate": 1.8116822720074966e-05, + "loss": 0.0814, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.032157689332962036, + "step": 5555, + "valid_targets_mean": 2611.0, + "valid_targets_min": 574 + }, + { + "epoch": 4.043652237177155, + "grad_norm": 0.36087658219620455, + "learning_rate": 1.8080718296048068e-05, + "loss": 0.089, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04190497472882271, + "step": 5560, + "valid_targets_mean": 4036.5, + "valid_targets_min": 2642 + }, + { + "epoch": 4.047289923608585, + "grad_norm": 0.5433535834621037, + "learning_rate": 1.804462018365796e-05, + "loss": 0.0925, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.060748204588890076, + "step": 5565, + "valid_targets_mean": 1960.5, + "valid_targets_min": 593 + }, + { + "epoch": 4.050927610040015, + "grad_norm": 0.39911977503677387, + "learning_rate": 1.800852850161478e-05, + "loss": 0.1175, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.036281686276197433, + "step": 5570, + "valid_targets_mean": 3354.2, + "valid_targets_min": 1353 + }, + { + "epoch": 4.054565296471444, + "grad_norm": 0.49915032372572854, + "learning_rate": 1.7972443368607507e-05, + "loss": 0.0843, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.048760510981082916, + "step": 5575, + "valid_targets_mean": 2300.9, + "valid_targets_min": 818 + }, + { + "epoch": 4.058202982902873, + "grad_norm": 0.34027006781766916, + "learning_rate": 1.7936364903303577e-05, + "loss": 0.0844, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02959313616156578, + "step": 5580, + "valid_targets_mean": 2295.8, + "valid_targets_min": 934 + }, + { + "epoch": 4.061840669334304, + "grad_norm": 0.355062212090508, + "learning_rate": 1.7900293224348524e-05, + "loss": 0.0766, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.041722558438777924, + "step": 5585, + "valid_targets_mean": 3524.4, + "valid_targets_min": 2798 + }, + { + "epoch": 4.065478355765733, + "grad_norm": 0.4493108962650758, + "learning_rate": 1.7864228450365525e-05, + "loss": 0.0979, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.038800932466983795, + "step": 5590, + "valid_targets_mean": 2810.1, + "valid_targets_min": 471 + }, + { + "epoch": 4.069116042197162, + "grad_norm": 0.43774734511548413, + "learning_rate": 1.7828170699955093e-05, + "loss": 0.0985, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06264286488294601, + "step": 5595, + "valid_targets_mean": 3042.8, + "valid_targets_min": 903 + }, + { + "epoch": 4.072753728628593, + "grad_norm": 0.41962218388460787, + "learning_rate": 1.779212009169461e-05, + "loss": 0.0975, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.038495905697345734, + "step": 5600, + "valid_targets_mean": 3081.6, + "valid_targets_min": 474 + }, + { + "epoch": 4.076391415060022, + "grad_norm": 0.4161742118672979, + "learning_rate": 1.7756076744137982e-05, + "loss": 0.0993, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04713970422744751, + "step": 5605, + "valid_targets_mean": 3468.0, + "valid_targets_min": 640 + }, + { + "epoch": 4.080029101491451, + "grad_norm": 0.35223215764950067, + "learning_rate": 1.772004077581525e-05, + "loss": 0.1307, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03238498419523239, + "step": 5610, + "valid_targets_mean": 3438.4, + "valid_targets_min": 902 + }, + { + "epoch": 4.083666787922881, + "grad_norm": 0.40567821606230203, + "learning_rate": 1.7684012305232166e-05, + "loss": 0.0832, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04149336367845535, + "step": 5615, + "valid_targets_mean": 3212.8, + "valid_targets_min": 1837 + }, + { + "epoch": 4.087304474354311, + "grad_norm": 0.719599222670336, + "learning_rate": 1.764799145086985e-05, + "loss": 0.1131, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08572384715080261, + "step": 5620, + "valid_targets_mean": 1358.8, + "valid_targets_min": 756 + }, + { + "epoch": 4.09094216078574, + "grad_norm": 0.5595037153334608, + "learning_rate": 1.7611978331184346e-05, + "loss": 0.1139, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08240849524736404, + "step": 5625, + "valid_targets_mean": 2896.2, + "valid_targets_min": 1072 + }, + { + "epoch": 4.09457984721717, + "grad_norm": 0.3807945566013535, + "learning_rate": 1.7575973064606298e-05, + "loss": 0.0891, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.028254041448235512, + "step": 5630, + "valid_targets_mean": 2908.1, + "valid_targets_min": 895 + }, + { + "epoch": 4.098217533648599, + "grad_norm": 0.6119436315422381, + "learning_rate": 1.753997576954049e-05, + "loss": 0.2052, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18530578911304474, + "step": 5635, + "valid_targets_mean": 2562.2, + "valid_targets_min": 878 + }, + { + "epoch": 4.101855220080029, + "grad_norm": 0.47574368424901675, + "learning_rate": 1.7503986564365512e-05, + "loss": 0.1897, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.041331954300403595, + "step": 5640, + "valid_targets_mean": 2087.6, + "valid_targets_min": 756 + }, + { + "epoch": 4.105492906511459, + "grad_norm": 0.35537244695724873, + "learning_rate": 1.7468005567433348e-05, + "loss": 0.1005, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.036058492958545685, + "step": 5645, + "valid_targets_mean": 3421.8, + "valid_targets_min": 2963 + }, + { + "epoch": 4.109130592942888, + "grad_norm": 0.8378979214043236, + "learning_rate": 1.743203289706898e-05, + "loss": 0.1701, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2253960818052292, + "step": 5650, + "valid_targets_mean": 2375.2, + "valid_targets_min": 1153 + }, + { + "epoch": 4.112768279374318, + "grad_norm": 0.3028219566389386, + "learning_rate": 1.7396068671570025e-05, + "loss": 0.109, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.034852106124162674, + "step": 5655, + "valid_targets_mean": 4886.2, + "valid_targets_min": 1700 + }, + { + "epoch": 4.116405965805748, + "grad_norm": 0.34336237657872454, + "learning_rate": 1.7360113009206305e-05, + "loss": 0.0887, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02934516780078411, + "step": 5660, + "valid_targets_mean": 3965.1, + "valid_targets_min": 789 + }, + { + "epoch": 4.120043652237177, + "grad_norm": 0.39246697301687633, + "learning_rate": 1.7324166028219495e-05, + "loss": 0.095, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04932279512286186, + "step": 5665, + "valid_targets_mean": 3771.5, + "valid_targets_min": 995 + }, + { + "epoch": 4.123681338668606, + "grad_norm": 0.37046213444481435, + "learning_rate": 1.7288227846822726e-05, + "loss": 0.0858, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030095692723989487, + "step": 5670, + "valid_targets_mean": 2220.8, + "valid_targets_min": 488 + }, + { + "epoch": 4.127319025100037, + "grad_norm": 0.4583836450611509, + "learning_rate": 1.725229858320018e-05, + "loss": 0.0866, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.051320742815732956, + "step": 5675, + "valid_targets_mean": 3635.4, + "valid_targets_min": 2437 + }, + { + "epoch": 4.130956711531466, + "grad_norm": 0.4842586495070214, + "learning_rate": 1.721637835550672e-05, + "loss": 0.0819, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03604227304458618, + "step": 5680, + "valid_targets_mean": 1752.1, + "valid_targets_min": 712 + }, + { + "epoch": 4.134594397962895, + "grad_norm": 0.40379943081863245, + "learning_rate": 1.7180467281867495e-05, + "loss": 0.0836, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.041926104575395584, + "step": 5685, + "valid_targets_mean": 3713.4, + "valid_targets_min": 2422 + }, + { + "epoch": 4.138232084394325, + "grad_norm": 0.37709035510056593, + "learning_rate": 1.7144565480377547e-05, + "loss": 0.0746, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04119100421667099, + "step": 5690, + "valid_targets_mean": 3998.5, + "valid_targets_min": 3643 + }, + { + "epoch": 4.141869770825755, + "grad_norm": 0.39703786226340276, + "learning_rate": 1.7108673069101423e-05, + "loss": 0.1302, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02937202714383602, + "step": 5695, + "valid_targets_mean": 3319.0, + "valid_targets_min": 772 + }, + { + "epoch": 4.145507457257184, + "grad_norm": 0.38584777121216757, + "learning_rate": 1.7072790166072786e-05, + "loss": 0.0684, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.028986606746912003, + "step": 5700, + "valid_targets_mean": 2934.8, + "valid_targets_min": 722 + }, + { + "epoch": 4.149145143688614, + "grad_norm": 0.37523282564498867, + "learning_rate": 1.7036916889294053e-05, + "loss": 0.0766, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03080163151025772, + "step": 5705, + "valid_targets_mean": 3656.4, + "valid_targets_min": 2788 + }, + { + "epoch": 4.152782830120044, + "grad_norm": 0.3856910589457599, + "learning_rate": 1.7001053356735956e-05, + "loss": 0.0783, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.040719982236623764, + "step": 5710, + "valid_targets_mean": 3202.5, + "valid_targets_min": 948 + }, + { + "epoch": 4.156420516551473, + "grad_norm": 0.6099894263315112, + "learning_rate": 1.696519968633721e-05, + "loss": 0.0968, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.048052381724119186, + "step": 5715, + "valid_targets_mean": 1002.0, + "valid_targets_min": 490 + }, + { + "epoch": 4.160058202982903, + "grad_norm": 0.3643498775090436, + "learning_rate": 1.692935599600408e-05, + "loss": 0.1087, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0471794530749321, + "step": 5720, + "valid_targets_mean": 3751.9, + "valid_targets_min": 2800 + }, + { + "epoch": 4.163695889414332, + "grad_norm": 0.5011087882106846, + "learning_rate": 1.6893522403610004e-05, + "loss": 0.1158, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0698692798614502, + "step": 5725, + "valid_targets_mean": 2641.8, + "valid_targets_min": 896 + }, + { + "epoch": 4.167333575845762, + "grad_norm": 0.4630952049477068, + "learning_rate": 1.6857699026995235e-05, + "loss": 0.092, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05100418999791145, + "step": 5730, + "valid_targets_mean": 3199.0, + "valid_targets_min": 1333 + }, + { + "epoch": 4.170971262277192, + "grad_norm": 0.487533381505839, + "learning_rate": 1.6821885983966416e-05, + "loss": 0.1067, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.048301566392183304, + "step": 5735, + "valid_targets_mean": 2548.1, + "valid_targets_min": 693 + }, + { + "epoch": 4.174608948708621, + "grad_norm": 0.3981392873285051, + "learning_rate": 1.6786083392296217e-05, + "loss": 0.122, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03749905899167061, + "step": 5740, + "valid_targets_mean": 2482.9, + "valid_targets_min": 803 + }, + { + "epoch": 4.178246635140051, + "grad_norm": 0.4088028149245119, + "learning_rate": 1.6750291369722922e-05, + "loss": 0.0905, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.046542029827833176, + "step": 5745, + "valid_targets_mean": 3239.2, + "valid_targets_min": 2525 + }, + { + "epoch": 4.181884321571481, + "grad_norm": 0.49514018168197027, + "learning_rate": 1.671451003395008e-05, + "loss": 0.112, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.040806859731674194, + "step": 5750, + "valid_targets_mean": 3806.1, + "valid_targets_min": 2393 + }, + { + "epoch": 4.18552200800291, + "grad_norm": 0.409217942124554, + "learning_rate": 1.6678739502646092e-05, + "loss": 0.0946, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04138227924704552, + "step": 5755, + "valid_targets_mean": 2990.9, + "valid_targets_min": 804 + }, + { + "epoch": 4.189159694434339, + "grad_norm": 0.3983925297455034, + "learning_rate": 1.6642979893443796e-05, + "loss": 0.0869, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04352351278066635, + "step": 5760, + "valid_targets_mean": 4885.8, + "valid_targets_min": 3695 + }, + { + "epoch": 4.19279738086577, + "grad_norm": 0.4578502204910818, + "learning_rate": 1.6607231323940155e-05, + "loss": 0.079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0466909185051918, + "step": 5765, + "valid_targets_mean": 3128.8, + "valid_targets_min": 596 + }, + { + "epoch": 4.196435067297199, + "grad_norm": 0.3030454577119551, + "learning_rate": 1.65714939116958e-05, + "loss": 0.0767, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029646337032318115, + "step": 5770, + "valid_targets_mean": 4567.8, + "valid_targets_min": 3804 + }, + { + "epoch": 4.200072753728628, + "grad_norm": 0.36369466995691585, + "learning_rate": 1.6535767774234695e-05, + "loss": 0.0835, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.033713437616825104, + "step": 5775, + "valid_targets_mean": 3853.1, + "valid_targets_min": 2862 + }, + { + "epoch": 4.203710440160058, + "grad_norm": 0.5578676257808485, + "learning_rate": 1.6500053029043702e-05, + "loss": 0.1033, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05905359238386154, + "step": 5780, + "valid_targets_mean": 1072.6, + "valid_targets_min": 497 + }, + { + "epoch": 4.207348126591488, + "grad_norm": 0.41202443148752255, + "learning_rate": 1.646434979357222e-05, + "loss": 0.0866, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0491841696202755, + "step": 5785, + "valid_targets_mean": 4191.0, + "valid_targets_min": 3918 + }, + { + "epoch": 4.210985813022917, + "grad_norm": 0.5411500595654912, + "learning_rate": 1.642865818523183e-05, + "loss": 0.0985, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04362528771162033, + "step": 5790, + "valid_targets_mean": 2121.9, + "valid_targets_min": 785 + }, + { + "epoch": 4.214623499454347, + "grad_norm": 0.35423189995589294, + "learning_rate": 1.639297832139583e-05, + "loss": 0.0711, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03572291135787964, + "step": 5795, + "valid_targets_mean": 3451.6, + "valid_targets_min": 905 + }, + { + "epoch": 4.218261185885777, + "grad_norm": 0.43368395143632693, + "learning_rate": 1.6357310319398937e-05, + "loss": 0.0929, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06749680638313293, + "step": 5800, + "valid_targets_mean": 2929.2, + "valid_targets_min": 657 + }, + { + "epoch": 4.221898872317206, + "grad_norm": 0.3964611315790159, + "learning_rate": 1.632165429653683e-05, + "loss": 0.085, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.043853700160980225, + "step": 5805, + "valid_targets_mean": 3798.9, + "valid_targets_min": 3184 + }, + { + "epoch": 4.225536558748636, + "grad_norm": 0.5113975228095675, + "learning_rate": 1.628601037006582e-05, + "loss": 0.0769, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04714503139257431, + "step": 5810, + "valid_targets_mean": 1463.9, + "valid_targets_min": 627 + }, + { + "epoch": 4.229174245180065, + "grad_norm": 0.5037299887793051, + "learning_rate": 1.6250378657202414e-05, + "loss": 0.0857, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04180711507797241, + "step": 5815, + "valid_targets_mean": 1701.5, + "valid_targets_min": 827 + }, + { + "epoch": 4.232811931611495, + "grad_norm": 0.4879466648549181, + "learning_rate": 1.6214759275122953e-05, + "loss": 0.0891, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03998861089348793, + "step": 5820, + "valid_targets_mean": 1789.9, + "valid_targets_min": 940 + }, + { + "epoch": 4.236449618042925, + "grad_norm": 0.49333805932431574, + "learning_rate": 1.6179152340963268e-05, + "loss": 0.1036, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.039158690720796585, + "step": 5825, + "valid_targets_mean": 2943.8, + "valid_targets_min": 958 + }, + { + "epoch": 4.240087304474354, + "grad_norm": 0.4298355902791748, + "learning_rate": 1.6143557971818205e-05, + "loss": 0.085, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03558417782187462, + "step": 5830, + "valid_targets_mean": 1859.2, + "valid_targets_min": 719 + }, + { + "epoch": 4.243724990905784, + "grad_norm": 0.5006556805777524, + "learning_rate": 1.610797628474132e-05, + "loss": 0.1041, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03351663798093796, + "step": 5835, + "valid_targets_mean": 1703.9, + "valid_targets_min": 759 + }, + { + "epoch": 4.247362677337214, + "grad_norm": 0.7728847717879501, + "learning_rate": 1.6072407396744447e-05, + "loss": 0.1077, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11155924201011658, + "step": 5840, + "valid_targets_mean": 1556.6, + "valid_targets_min": 521 + }, + { + "epoch": 4.251000363768643, + "grad_norm": 0.4551440895186116, + "learning_rate": 1.6036851424797338e-05, + "loss": 0.0878, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05278254300355911, + "step": 5845, + "valid_targets_mean": 3267.6, + "valid_targets_min": 1032 + }, + { + "epoch": 4.2546380502000725, + "grad_norm": 0.30882277698518007, + "learning_rate": 1.6001308485827284e-05, + "loss": 0.0737, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029455725103616714, + "step": 5850, + "valid_targets_mean": 3609.0, + "valid_targets_min": 2741 + }, + { + "epoch": 4.258275736631503, + "grad_norm": 0.38013965396625365, + "learning_rate": 1.5965778696718672e-05, + "loss": 0.0848, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.037792496383190155, + "step": 5855, + "valid_targets_mean": 2735.8, + "valid_targets_min": 1692 + }, + { + "epoch": 4.261913423062932, + "grad_norm": 0.40017063042262324, + "learning_rate": 1.5930262174312704e-05, + "loss": 0.1129, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047459736466407776, + "step": 5860, + "valid_targets_mean": 3693.4, + "valid_targets_min": 2215 + }, + { + "epoch": 4.2655511094943614, + "grad_norm": 0.36610516070616483, + "learning_rate": 1.5894759035406902e-05, + "loss": 0.0717, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03134572505950928, + "step": 5865, + "valid_targets_mean": 3109.9, + "valid_targets_min": 723 + }, + { + "epoch": 4.269188795925791, + "grad_norm": 0.34014521568523604, + "learning_rate": 1.5859269396754814e-05, + "loss": 0.0858, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03693245351314545, + "step": 5870, + "valid_targets_mean": 4984.2, + "valid_targets_min": 3282 + }, + { + "epoch": 4.272826482357221, + "grad_norm": 0.3071546711400201, + "learning_rate": 1.5823793375065566e-05, + "loss": 0.0815, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.033801496028900146, + "step": 5875, + "valid_targets_mean": 3863.4, + "valid_targets_min": 995 + }, + { + "epoch": 4.27646416878865, + "grad_norm": 0.28377558173127576, + "learning_rate": 1.5788331087003515e-05, + "loss": 0.0696, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.033096663653850555, + "step": 5880, + "valid_targets_mean": 4201.0, + "valid_targets_min": 949 + }, + { + "epoch": 4.28010185522008, + "grad_norm": 0.3958593078730338, + "learning_rate": 1.575288264918786e-05, + "loss": 0.0907, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.057266540825366974, + "step": 5885, + "valid_targets_mean": 4321.6, + "valid_targets_min": 2724 + }, + { + "epoch": 4.28373954165151, + "grad_norm": 0.3471672579431372, + "learning_rate": 1.5717448178192233e-05, + "loss": 0.0762, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.025314774364233017, + "step": 5890, + "valid_targets_mean": 3384.5, + "valid_targets_min": 424 + }, + { + "epoch": 4.287377228082939, + "grad_norm": 0.3853918106618121, + "learning_rate": 1.568202779054437e-05, + "loss": 0.1089, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03343036025762558, + "step": 5895, + "valid_targets_mean": 2987.2, + "valid_targets_min": 531 + }, + { + "epoch": 4.291014914514369, + "grad_norm": 0.33175378636141195, + "learning_rate": 1.5646621602725654e-05, + "loss": 0.104, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.033136673271656036, + "step": 5900, + "valid_targets_mean": 3469.5, + "valid_targets_min": 1195 + }, + { + "epoch": 4.294652600945798, + "grad_norm": 0.3913326737612808, + "learning_rate": 1.5611229731170793e-05, + "loss": 0.0711, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06447508186101913, + "step": 5905, + "valid_targets_mean": 2682.8, + "valid_targets_min": 722 + }, + { + "epoch": 4.298290287377228, + "grad_norm": 0.3786384343329823, + "learning_rate": 1.5575852292267414e-05, + "loss": 0.1006, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03585771471261978, + "step": 5910, + "valid_targets_mean": 3496.9, + "valid_targets_min": 1251 + }, + { + "epoch": 4.301927973808658, + "grad_norm": 0.4350138030070068, + "learning_rate": 1.5540489402355678e-05, + "loss": 0.0806, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0447450652718544, + "step": 5915, + "valid_targets_mean": 2912.5, + "valid_targets_min": 970 + }, + { + "epoch": 4.305565660240087, + "grad_norm": 0.5515078427469016, + "learning_rate": 1.5505141177727913e-05, + "loss": 0.0922, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04508994147181511, + "step": 5920, + "valid_targets_mean": 1423.6, + "valid_targets_min": 634 + }, + { + "epoch": 4.3092033466715165, + "grad_norm": 0.5854140640505225, + "learning_rate": 1.5469807734628183e-05, + "loss": 0.1056, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07098647207021713, + "step": 5925, + "valid_targets_mean": 1423.9, + "valid_targets_min": 685 + }, + { + "epoch": 4.312841033102947, + "grad_norm": 0.40954228153431144, + "learning_rate": 1.5434489189252e-05, + "loss": 0.1144, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.031893499195575714, + "step": 5930, + "valid_targets_mean": 2932.5, + "valid_targets_min": 593 + }, + { + "epoch": 4.316478719534376, + "grad_norm": 0.3472579183275617, + "learning_rate": 1.539918565774584e-05, + "loss": 0.088, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.032529138028621674, + "step": 5935, + "valid_targets_mean": 3945.8, + "valid_targets_min": 3129 + }, + { + "epoch": 4.3201164059658055, + "grad_norm": 0.42955493134530875, + "learning_rate": 1.5363897256206805e-05, + "loss": 0.0905, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.049513645470142365, + "step": 5940, + "valid_targets_mean": 3621.1, + "valid_targets_min": 859 + }, + { + "epoch": 4.323754092397236, + "grad_norm": 0.4906594582364949, + "learning_rate": 1.532862410068227e-05, + "loss": 0.0815, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04312261566519737, + "step": 5945, + "valid_targets_mean": 2553.0, + "valid_targets_min": 616 + }, + { + "epoch": 4.327391778828665, + "grad_norm": 0.42159771168829957, + "learning_rate": 1.529336630716945e-05, + "loss": 0.0797, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.036384858191013336, + "step": 5950, + "valid_targets_mean": 2690.5, + "valid_targets_min": 541 + }, + { + "epoch": 4.3310294652600945, + "grad_norm": 0.7591040565432277, + "learning_rate": 1.525812399161506e-05, + "loss": 0.0764, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.049666374921798706, + "step": 5955, + "valid_targets_mean": 906.4, + "valid_targets_min": 598 + }, + { + "epoch": 4.334667151691524, + "grad_norm": 0.43748071390789345, + "learning_rate": 1.5222897269914892e-05, + "loss": 0.0927, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04071364179253578, + "step": 5960, + "valid_targets_mean": 3147.2, + "valid_targets_min": 1101 + }, + { + "epoch": 4.338304838122954, + "grad_norm": 0.5200972112503011, + "learning_rate": 1.5187686257913466e-05, + "loss": 0.077, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.035416826605796814, + "step": 5965, + "valid_targets_mean": 2331.8, + "valid_targets_min": 680 + }, + { + "epoch": 4.3419425245543835, + "grad_norm": 0.4191763678825635, + "learning_rate": 1.515249107140365e-05, + "loss": 0.0777, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.041577018797397614, + "step": 5970, + "valid_targets_mean": 2577.0, + "valid_targets_min": 774 + }, + { + "epoch": 4.345580210985813, + "grad_norm": 0.4141086261896435, + "learning_rate": 1.511731182612626e-05, + "loss": 0.0935, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.051594045013189316, + "step": 5975, + "valid_targets_mean": 4140.1, + "valid_targets_min": 1474 + }, + { + "epoch": 4.349217897417242, + "grad_norm": 0.24142057633907354, + "learning_rate": 1.5082148637769694e-05, + "loss": 0.072, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.025543535128235817, + "step": 5980, + "valid_targets_mean": 5498.9, + "valid_targets_min": 3677 + }, + { + "epoch": 4.352855583848672, + "grad_norm": 0.38191426788444527, + "learning_rate": 1.5047001621969537e-05, + "loss": 0.0784, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.046430543065071106, + "step": 5985, + "valid_targets_mean": 3352.5, + "valid_targets_min": 675 + }, + { + "epoch": 4.356493270280102, + "grad_norm": 0.34081368033316656, + "learning_rate": 1.5011870894308205e-05, + "loss": 0.0758, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03203882277011871, + "step": 5990, + "valid_targets_mean": 3902.5, + "valid_targets_min": 795 + }, + { + "epoch": 4.360130956711531, + "grad_norm": 0.5353288341967958, + "learning_rate": 1.497675657031453e-05, + "loss": 0.1788, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0781446248292923, + "step": 5995, + "valid_targets_mean": 2495.0, + "valid_targets_min": 605 + }, + { + "epoch": 4.363768643142961, + "grad_norm": 0.3374388882296259, + "learning_rate": 1.4941658765463412e-05, + "loss": 0.075, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03828607499599457, + "step": 6000, + "valid_targets_mean": 3298.6, + "valid_targets_min": 985 + }, + { + "epoch": 4.367406329574391, + "grad_norm": 0.4279547663418045, + "learning_rate": 1.4906577595175428e-05, + "loss": 0.0863, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04187878966331482, + "step": 6005, + "valid_targets_mean": 3505.2, + "valid_targets_min": 878 + }, + { + "epoch": 4.37104401600582, + "grad_norm": 0.599677297914672, + "learning_rate": 1.4871513174816445e-05, + "loss": 0.1078, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11834871768951416, + "step": 6010, + "valid_targets_mean": 2289.5, + "valid_targets_min": 1214 + }, + { + "epoch": 4.3746817024372495, + "grad_norm": 0.5696956840220913, + "learning_rate": 1.483646561969726e-05, + "loss": 0.1761, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11600963771343231, + "step": 6015, + "valid_targets_mean": 8496.5, + "valid_targets_min": 6565 + }, + { + "epoch": 4.37831938886868, + "grad_norm": 0.4425826147667315, + "learning_rate": 1.4801435045073197e-05, + "loss": 0.1887, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08283234387636185, + "step": 6020, + "valid_targets_mean": 5663.5, + "valid_targets_min": 3833 + }, + { + "epoch": 4.381957075300109, + "grad_norm": 0.4230720399288085, + "learning_rate": 1.4766421566143726e-05, + "loss": 0.1772, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08994975686073303, + "step": 6025, + "valid_targets_mean": 6654.1, + "valid_targets_min": 5324 + }, + { + "epoch": 4.3855947617315385, + "grad_norm": 0.3851159472004513, + "learning_rate": 1.4731425298052133e-05, + "loss": 0.1716, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0857534408569336, + "step": 6030, + "valid_targets_mean": 7635.1, + "valid_targets_min": 5672 + }, + { + "epoch": 4.389232448162969, + "grad_norm": 0.3564944061317814, + "learning_rate": 1.469644635588507e-05, + "loss": 0.1768, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07931319624185562, + "step": 6035, + "valid_targets_mean": 8517.0, + "valid_targets_min": 5045 + }, + { + "epoch": 4.392870134594398, + "grad_norm": 0.40749487760600017, + "learning_rate": 1.4661484854672234e-05, + "loss": 0.1819, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08788798004388809, + "step": 6040, + "valid_targets_mean": 6759.1, + "valid_targets_min": 3152 + }, + { + "epoch": 4.3965078210258275, + "grad_norm": 0.40052058571775107, + "learning_rate": 1.4626540909385954e-05, + "loss": 0.1854, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08185853064060211, + "step": 6045, + "valid_targets_mean": 7822.8, + "valid_targets_min": 4820 + }, + { + "epoch": 4.400145507457257, + "grad_norm": 0.3586299893343649, + "learning_rate": 1.4591614634940836e-05, + "loss": 0.1753, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07046092301607132, + "step": 6050, + "valid_targets_mean": 6852.2, + "valid_targets_min": 4135 + }, + { + "epoch": 4.403783193888687, + "grad_norm": 0.4348792988707946, + "learning_rate": 1.4556706146193374e-05, + "loss": 0.173, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07519887387752533, + "step": 6055, + "valid_targets_mean": 4794.2, + "valid_targets_min": 3230 + }, + { + "epoch": 4.4074208803201165, + "grad_norm": 0.3913283853301071, + "learning_rate": 1.4521815557941547e-05, + "loss": 0.1536, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0890219658613205, + "step": 6060, + "valid_targets_mean": 7691.0, + "valid_targets_min": 5900 + }, + { + "epoch": 4.411058566751546, + "grad_norm": 0.41190769419368467, + "learning_rate": 1.4486942984924519e-05, + "loss": 0.175, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08249206840991974, + "step": 6065, + "valid_targets_mean": 6510.9, + "valid_targets_min": 5639 + }, + { + "epoch": 4.414696253182976, + "grad_norm": 0.4204313778437715, + "learning_rate": 1.4452088541822156e-05, + "loss": 0.1647, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08555063605308533, + "step": 6070, + "valid_targets_mean": 6096.8, + "valid_targets_min": 4669 + }, + { + "epoch": 4.4183339396144055, + "grad_norm": 0.4013289509689533, + "learning_rate": 1.4417252343254739e-05, + "loss": 0.1717, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08796443045139313, + "step": 6075, + "valid_targets_mean": 6812.2, + "valid_targets_min": 5160 + }, + { + "epoch": 4.421971626045835, + "grad_norm": 0.37551437412008803, + "learning_rate": 1.4382434503782538e-05, + "loss": 0.1647, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07602357864379883, + "step": 6080, + "valid_targets_mean": 6980.4, + "valid_targets_min": 5151 + }, + { + "epoch": 4.425609312477264, + "grad_norm": 0.37139006338220987, + "learning_rate": 1.4347635137905445e-05, + "loss": 0.1668, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0899597704410553, + "step": 6085, + "valid_targets_mean": 7417.1, + "valid_targets_min": 5932 + }, + { + "epoch": 4.429246998908694, + "grad_norm": 0.38948026853331585, + "learning_rate": 1.4312854360062615e-05, + "loss": 0.1749, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08285940438508987, + "step": 6090, + "valid_targets_mean": 6258.5, + "valid_targets_min": 5569 + }, + { + "epoch": 4.432884685340124, + "grad_norm": 0.3927432977052218, + "learning_rate": 1.427809228463205e-05, + "loss": 0.167, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0940733253955841, + "step": 6095, + "valid_targets_mean": 7689.1, + "valid_targets_min": 6223 + }, + { + "epoch": 4.436522371771553, + "grad_norm": 0.508711011995544, + "learning_rate": 1.4243349025930286e-05, + "loss": 0.1746, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09931532293558121, + "step": 6100, + "valid_targets_mean": 4371.9, + "valid_targets_min": 1172 + }, + { + "epoch": 4.4401600582029825, + "grad_norm": 0.4201712142025694, + "learning_rate": 1.4208624698211937e-05, + "loss": 0.184, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09860621392726898, + "step": 6105, + "valid_targets_mean": 6936.2, + "valid_targets_min": 5619 + }, + { + "epoch": 4.443797744634413, + "grad_norm": 0.6901400680398599, + "learning_rate": 1.41739194156694e-05, + "loss": 0.1577, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10339194536209106, + "step": 6110, + "valid_targets_mean": 3273.2, + "valid_targets_min": 442 + }, + { + "epoch": 4.447435431065842, + "grad_norm": 0.38049309791747526, + "learning_rate": 1.4139233292432411e-05, + "loss": 0.1577, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08326970785856247, + "step": 6115, + "valid_targets_mean": 8141.1, + "valid_targets_min": 5061 + }, + { + "epoch": 4.4510731174972715, + "grad_norm": 0.41154835593495287, + "learning_rate": 1.4104566442567712e-05, + "loss": 0.1669, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08395469933748245, + "step": 6120, + "valid_targets_mean": 7371.1, + "valid_targets_min": 5112 + }, + { + "epoch": 4.454710803928702, + "grad_norm": 0.4140699591269555, + "learning_rate": 1.406991898007868e-05, + "loss": 0.1656, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07373280823230743, + "step": 6125, + "valid_targets_mean": 5659.4, + "valid_targets_min": 4626 + }, + { + "epoch": 4.458348490360131, + "grad_norm": 0.36428262379710313, + "learning_rate": 1.4035291018904896e-05, + "loss": 0.1722, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0759967714548111, + "step": 6130, + "valid_targets_mean": 7801.8, + "valid_targets_min": 6172 + }, + { + "epoch": 4.4619861767915605, + "grad_norm": 0.3915273961028772, + "learning_rate": 1.400068267292186e-05, + "loss": 0.1645, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08310458064079285, + "step": 6135, + "valid_targets_mean": 7547.4, + "valid_targets_min": 5804 + }, + { + "epoch": 4.46562386322299, + "grad_norm": 0.3765489675634409, + "learning_rate": 1.396609405594052e-05, + "loss": 0.1577, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07479551434516907, + "step": 6140, + "valid_targets_mean": 6474.2, + "valid_targets_min": 5050 + }, + { + "epoch": 4.46926154965442, + "grad_norm": 0.39490980705462364, + "learning_rate": 1.3931525281706972e-05, + "loss": 0.1471, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07651837170124054, + "step": 6145, + "valid_targets_mean": 7225.4, + "valid_targets_min": 5298 + }, + { + "epoch": 4.4728992360858495, + "grad_norm": 0.3643926026402985, + "learning_rate": 1.3896976463902054e-05, + "loss": 0.1458, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06921494007110596, + "step": 6150, + "valid_targets_mean": 6442.9, + "valid_targets_min": 4679 + }, + { + "epoch": 4.476536922517279, + "grad_norm": 0.47778446276702463, + "learning_rate": 1.3862447716140973e-05, + "loss": 0.1727, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06839483976364136, + "step": 6155, + "valid_targets_mean": 3651.9, + "valid_targets_min": 1801 + }, + { + "epoch": 4.480174608948708, + "grad_norm": 0.4217338941178669, + "learning_rate": 1.3827939151972937e-05, + "loss": 0.1598, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09741295874118805, + "step": 6160, + "valid_targets_mean": 7512.8, + "valid_targets_min": 5331 + }, + { + "epoch": 4.4838122953801385, + "grad_norm": 0.4435137915615606, + "learning_rate": 1.3793450884880767e-05, + "loss": 0.1737, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09181414544582367, + "step": 6165, + "valid_targets_mean": 7108.8, + "valid_targets_min": 4822 + }, + { + "epoch": 4.487449981811568, + "grad_norm": 0.49929566081096194, + "learning_rate": 1.3758983028280568e-05, + "loss": 0.1787, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07795166969299316, + "step": 6170, + "valid_targets_mean": 7125.8, + "valid_targets_min": 4623 + }, + { + "epoch": 4.491087668242997, + "grad_norm": 0.4407184749715861, + "learning_rate": 1.3724535695521289e-05, + "loss": 0.1766, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08003628998994827, + "step": 6175, + "valid_targets_mean": 6269.2, + "valid_targets_min": 5458 + }, + { + "epoch": 4.4947253546744275, + "grad_norm": 0.38328777090971217, + "learning_rate": 1.36901089998844e-05, + "loss": 0.1672, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08524587750434875, + "step": 6180, + "valid_targets_mean": 7455.1, + "valid_targets_min": 6008 + }, + { + "epoch": 4.498363041105857, + "grad_norm": 0.40070817587113283, + "learning_rate": 1.3655703054583512e-05, + "loss": 0.1763, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08703937381505966, + "step": 6185, + "valid_targets_mean": 7104.4, + "valid_targets_min": 5386 + }, + { + "epoch": 4.502000727537286, + "grad_norm": 0.44110776638705385, + "learning_rate": 1.3621317972763981e-05, + "loss": 0.1678, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08502298593521118, + "step": 6190, + "valid_targets_mean": 6532.6, + "valid_targets_min": 4855 + }, + { + "epoch": 4.505638413968716, + "grad_norm": 0.3902331472672831, + "learning_rate": 1.3586953867502576e-05, + "loss": 0.1718, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08421805500984192, + "step": 6195, + "valid_targets_mean": 6743.9, + "valid_targets_min": 5613 + }, + { + "epoch": 4.509276100400146, + "grad_norm": 0.4515743883553881, + "learning_rate": 1.3552610851807062e-05, + "loss": 0.172, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08649154007434845, + "step": 6200, + "valid_targets_mean": 6373.6, + "valid_targets_min": 4961 + }, + { + "epoch": 4.512913786831575, + "grad_norm": 0.4289752951577046, + "learning_rate": 1.3518289038615853e-05, + "loss": 0.1749, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09018933773040771, + "step": 6205, + "valid_targets_mean": 6461.2, + "valid_targets_min": 5286 + }, + { + "epoch": 4.5165514732630045, + "grad_norm": 0.38950771702504067, + "learning_rate": 1.3483988540797652e-05, + "loss": 0.1702, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08902765065431595, + "step": 6210, + "valid_targets_mean": 6780.5, + "valid_targets_min": 5201 + }, + { + "epoch": 4.520189159694434, + "grad_norm": 0.4226123114435249, + "learning_rate": 1.3449709471151051e-05, + "loss": 0.1691, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08445990830659866, + "step": 6215, + "valid_targets_mean": 5710.8, + "valid_targets_min": 4838 + }, + { + "epoch": 4.523826846125864, + "grad_norm": 0.692420905238489, + "learning_rate": 1.3415451942404185e-05, + "loss": 0.1379, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12581872940063477, + "step": 6220, + "valid_targets_mean": 3235.6, + "valid_targets_min": 176 + }, + { + "epoch": 4.5274645325572935, + "grad_norm": 0.445570891026041, + "learning_rate": 1.338121606721434e-05, + "loss": 0.1718, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07805676758289337, + "step": 6225, + "valid_targets_mean": 6674.8, + "valid_targets_min": 4582 + }, + { + "epoch": 4.531102218988723, + "grad_norm": 0.41354798862433134, + "learning_rate": 1.334700195816761e-05, + "loss": 0.1669, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08325718343257904, + "step": 6230, + "valid_targets_mean": 6157.8, + "valid_targets_min": 5385 + }, + { + "epoch": 4.534739905420153, + "grad_norm": 0.41705158804938935, + "learning_rate": 1.3312809727778486e-05, + "loss": 0.1697, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07977943122386932, + "step": 6235, + "valid_targets_mean": 6006.2, + "valid_targets_min": 4478 + }, + { + "epoch": 4.5383775918515825, + "grad_norm": 0.4155766089994334, + "learning_rate": 1.3278639488489527e-05, + "loss": 0.166, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08346584439277649, + "step": 6240, + "valid_targets_mean": 5893.4, + "valid_targets_min": 3911 + }, + { + "epoch": 4.542015278283012, + "grad_norm": 0.41958750551163554, + "learning_rate": 1.3244491352670975e-05, + "loss": 0.1653, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0822872743010521, + "step": 6245, + "valid_targets_mean": 6134.2, + "valid_targets_min": 5318 + }, + { + "epoch": 4.545652964714441, + "grad_norm": 0.4468226071675573, + "learning_rate": 1.3210365432620371e-05, + "loss": 0.1733, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08607763051986694, + "step": 6250, + "valid_targets_mean": 5474.4, + "valid_targets_min": 4959 + }, + { + "epoch": 4.5492906511458715, + "grad_norm": 0.5879336028377282, + "learning_rate": 1.3176261840562216e-05, + "loss": 0.1639, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0488097220659256, + "step": 6255, + "valid_targets_mean": 1804.1, + "valid_targets_min": 842 + }, + { + "epoch": 4.552928337577301, + "grad_norm": 0.4685985113963561, + "learning_rate": 1.3142180688647574e-05, + "loss": 0.138, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0762399211525917, + "step": 6260, + "valid_targets_mean": 6118.8, + "valid_targets_min": 4899 + }, + { + "epoch": 4.55656602400873, + "grad_norm": 0.4129472943721311, + "learning_rate": 1.3108122088953704e-05, + "loss": 0.1632, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0764441266655922, + "step": 6265, + "valid_targets_mean": 6773.8, + "valid_targets_min": 5046 + }, + { + "epoch": 4.56020371044016, + "grad_norm": 0.5986776975869527, + "learning_rate": 1.3074086153483718e-05, + "loss": 0.173, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08815701305866241, + "step": 6270, + "valid_targets_mean": 6633.6, + "valid_targets_min": 5284 + }, + { + "epoch": 4.56384139687159, + "grad_norm": 0.4113736031683725, + "learning_rate": 1.3040072994166186e-05, + "loss": 0.1652, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08418386429548264, + "step": 6275, + "valid_targets_mean": 6055.2, + "valid_targets_min": 4928 + }, + { + "epoch": 4.567479083303019, + "grad_norm": 0.4074593657653936, + "learning_rate": 1.3006082722854786e-05, + "loss": 0.162, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0868585854768753, + "step": 6280, + "valid_targets_mean": 7259.4, + "valid_targets_min": 5758 + }, + { + "epoch": 4.571116769734449, + "grad_norm": 0.44219751991633816, + "learning_rate": 1.2972115451327918e-05, + "loss": 0.1655, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08573442697525024, + "step": 6285, + "valid_targets_mean": 5731.8, + "valid_targets_min": 4629 + }, + { + "epoch": 4.574754456165879, + "grad_norm": 0.7175633107186672, + "learning_rate": 1.2938171291288359e-05, + "loss": 0.1871, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17345711588859558, + "step": 6290, + "valid_targets_mean": 5360.9, + "valid_targets_min": 2317 + }, + { + "epoch": 4.578392142597308, + "grad_norm": 0.5518101004158592, + "learning_rate": 1.2904250354362874e-05, + "loss": 0.2305, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12164200842380524, + "step": 6295, + "valid_targets_mean": 5303.1, + "valid_targets_min": 533 + }, + { + "epoch": 4.582029829028738, + "grad_norm": 0.5754697351066769, + "learning_rate": 1.2870352752101841e-05, + "loss": 0.2265, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11665134876966476, + "step": 6300, + "valid_targets_mean": 4545.8, + "valid_targets_min": 835 + }, + { + "epoch": 4.585667515460168, + "grad_norm": 0.645413918048563, + "learning_rate": 1.283647859597894e-05, + "loss": 0.2221, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12209512293338776, + "step": 6305, + "valid_targets_mean": 3618.9, + "valid_targets_min": 1201 + }, + { + "epoch": 4.589305201891597, + "grad_norm": 0.5501811054500719, + "learning_rate": 1.2802627997390715e-05, + "loss": 0.2129, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09908602386713028, + "step": 6310, + "valid_targets_mean": 4253.5, + "valid_targets_min": 1880 + }, + { + "epoch": 4.5929428883230266, + "grad_norm": 0.6109149981910231, + "learning_rate": 1.2768801067656258e-05, + "loss": 0.2339, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16492760181427002, + "step": 6315, + "valid_targets_mean": 6610.1, + "valid_targets_min": 2171 + }, + { + "epoch": 4.596580574754456, + "grad_norm": 0.5396632608257826, + "learning_rate": 1.2734997918016818e-05, + "loss": 0.2135, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12043086439371109, + "step": 6320, + "valid_targets_mean": 7171.4, + "valid_targets_min": 4500 + }, + { + "epoch": 4.600218261185885, + "grad_norm": 0.502430474373435, + "learning_rate": 1.2701218659635428e-05, + "loss": 0.2201, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11684626340866089, + "step": 6325, + "valid_targets_mean": 6207.4, + "valid_targets_min": 1855 + }, + { + "epoch": 4.6038559476173155, + "grad_norm": 0.6100555780996599, + "learning_rate": 1.2667463403596592e-05, + "loss": 0.2175, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10947791486978531, + "step": 6330, + "valid_targets_mean": 4075.9, + "valid_targets_min": 1024 + }, + { + "epoch": 4.607493634048745, + "grad_norm": 0.6017071910262811, + "learning_rate": 1.2633732260905836e-05, + "loss": 0.2295, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09756974875926971, + "step": 6335, + "valid_targets_mean": 4767.1, + "valid_targets_min": 2107 + }, + { + "epoch": 4.611131320480174, + "grad_norm": 0.5959168230776188, + "learning_rate": 1.2600025342489422e-05, + "loss": 0.2229, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12152259796857834, + "step": 6340, + "valid_targets_mean": 4211.4, + "valid_targets_min": 802 + }, + { + "epoch": 4.6147690069116045, + "grad_norm": 0.4734470729678315, + "learning_rate": 1.256634275919393e-05, + "loss": 0.2132, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11078476160764694, + "step": 6345, + "valid_targets_mean": 5345.1, + "valid_targets_min": 1597 + }, + { + "epoch": 4.618406693343034, + "grad_norm": 0.9624496589144286, + "learning_rate": 1.2532684621785928e-05, + "loss": 0.2264, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11552584916353226, + "step": 6350, + "valid_targets_mean": 3671.4, + "valid_targets_min": 1650 + }, + { + "epoch": 4.622044379774463, + "grad_norm": 0.5562286681000476, + "learning_rate": 1.2499051040951585e-05, + "loss": 0.2173, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1131507158279419, + "step": 6355, + "valid_targets_mean": 4799.1, + "valid_targets_min": 1385 + }, + { + "epoch": 4.6256820662058935, + "grad_norm": 0.578794053939905, + "learning_rate": 1.2465442127296297e-05, + "loss": 0.2117, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11656546592712402, + "step": 6360, + "valid_targets_mean": 4023.5, + "valid_targets_min": 2908 + }, + { + "epoch": 4.629319752637323, + "grad_norm": 0.5219023057458381, + "learning_rate": 1.2431857991344384e-05, + "loss": 0.2178, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0985490158200264, + "step": 6365, + "valid_targets_mean": 5064.6, + "valid_targets_min": 2785 + }, + { + "epoch": 4.632957439068752, + "grad_norm": 0.6764384236496572, + "learning_rate": 1.2398298743538639e-05, + "loss": 0.211, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10558735579252243, + "step": 6370, + "valid_targets_mean": 3886.4, + "valid_targets_min": 1210 + }, + { + "epoch": 4.636595125500182, + "grad_norm": 0.6264390446410325, + "learning_rate": 1.2364764494240036e-05, + "loss": 0.2163, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12082573026418686, + "step": 6375, + "valid_targets_mean": 3708.8, + "valid_targets_min": 1064 + }, + { + "epoch": 4.640232811931612, + "grad_norm": 0.6892106692347273, + "learning_rate": 1.2331255353727332e-05, + "loss": 0.2274, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1169440746307373, + "step": 6380, + "valid_targets_mean": 3276.2, + "valid_targets_min": 1678 + }, + { + "epoch": 4.643870498363041, + "grad_norm": 0.6267864299169623, + "learning_rate": 1.229777143219671e-05, + "loss": 0.2185, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11541140824556351, + "step": 6385, + "valid_targets_mean": 3520.1, + "valid_targets_min": 671 + }, + { + "epoch": 4.647508184794471, + "grad_norm": 0.6517213846543736, + "learning_rate": 1.2264312839761434e-05, + "loss": 0.2173, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1043749675154686, + "step": 6390, + "valid_targets_mean": 3488.5, + "valid_targets_min": 2000 + }, + { + "epoch": 4.6511458712259, + "grad_norm": 0.6311624059878068, + "learning_rate": 1.2230879686451438e-05, + "loss": 0.2006, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12485212087631226, + "step": 6395, + "valid_targets_mean": 4402.6, + "valid_targets_min": 1532 + }, + { + "epoch": 4.65478355765733, + "grad_norm": 0.6111287884921068, + "learning_rate": 1.2197472082213048e-05, + "loss": 0.2086, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08060477674007416, + "step": 6400, + "valid_targets_mean": 3749.4, + "valid_targets_min": 1051 + }, + { + "epoch": 4.65842124408876, + "grad_norm": 0.6444861518542919, + "learning_rate": 1.2164090136908526e-05, + "loss": 0.201, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10161945223808289, + "step": 6405, + "valid_targets_mean": 3368.1, + "valid_targets_min": 1345 + }, + { + "epoch": 4.662058930520189, + "grad_norm": 0.6374646587635184, + "learning_rate": 1.213073396031578e-05, + "loss": 0.2178, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10618400573730469, + "step": 6410, + "valid_targets_mean": 4128.9, + "valid_targets_min": 2231 + }, + { + "epoch": 4.665696616951619, + "grad_norm": 0.6522404890000051, + "learning_rate": 1.2097403662127966e-05, + "loss": 0.2189, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09713581204414368, + "step": 6415, + "valid_targets_mean": 3672.6, + "valid_targets_min": 2127 + }, + { + "epoch": 4.669334303383049, + "grad_norm": 0.6540593050100753, + "learning_rate": 1.2064099351953136e-05, + "loss": 0.2087, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09841892123222351, + "step": 6420, + "valid_targets_mean": 3006.8, + "valid_targets_min": 1265 + }, + { + "epoch": 4.672971989814478, + "grad_norm": 0.621066109429132, + "learning_rate": 1.2030821139313895e-05, + "loss": 0.2086, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08791917562484741, + "step": 6425, + "valid_targets_mean": 2981.9, + "valid_targets_min": 908 + }, + { + "epoch": 4.676609676245907, + "grad_norm": 0.6584147968263147, + "learning_rate": 1.1997569133646993e-05, + "loss": 0.2076, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11763627827167511, + "step": 6430, + "valid_targets_mean": 3147.5, + "valid_targets_min": 1264 + }, + { + "epoch": 4.6802473626773375, + "grad_norm": 0.6298694919237081, + "learning_rate": 1.1964343444303044e-05, + "loss": 0.2061, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09583032131195068, + "step": 6435, + "valid_targets_mean": 2517.9, + "valid_targets_min": 926 + }, + { + "epoch": 4.683885049108767, + "grad_norm": 0.6701567713362214, + "learning_rate": 1.1931144180546074e-05, + "loss": 0.2088, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10230707377195358, + "step": 6440, + "valid_targets_mean": 3204.5, + "valid_targets_min": 1016 + }, + { + "epoch": 4.687522735540196, + "grad_norm": 0.6208255717685753, + "learning_rate": 1.189797145155323e-05, + "loss": 0.2047, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10046559572219849, + "step": 6445, + "valid_targets_mean": 3525.0, + "valid_targets_min": 2019 + }, + { + "epoch": 4.691160421971626, + "grad_norm": 0.6272655345756317, + "learning_rate": 1.1864825366414405e-05, + "loss": 0.2079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12601499259471893, + "step": 6450, + "valid_targets_mean": 4745.4, + "valid_targets_min": 3023 + }, + { + "epoch": 4.694798108403056, + "grad_norm": 0.6146972984711753, + "learning_rate": 1.1831706034131849e-05, + "loss": 0.2021, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09143291413784027, + "step": 6455, + "valid_targets_mean": 3503.6, + "valid_targets_min": 1110 + }, + { + "epoch": 4.698435794834485, + "grad_norm": 0.7399279255182569, + "learning_rate": 1.1798613563619861e-05, + "loss": 0.2027, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11826576292514801, + "step": 6460, + "valid_targets_mean": 4206.5, + "valid_targets_min": 1460 + }, + { + "epoch": 4.702073481265915, + "grad_norm": 0.6397446069287527, + "learning_rate": 1.176554806370438e-05, + "loss": 0.2042, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11238368600606918, + "step": 6465, + "valid_targets_mean": 3938.6, + "valid_targets_min": 2558 + }, + { + "epoch": 4.705711167697345, + "grad_norm": 0.6451943669573549, + "learning_rate": 1.1732509643122673e-05, + "loss": 0.2152, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09567653387784958, + "step": 6470, + "valid_targets_mean": 3421.1, + "valid_targets_min": 1673 + }, + { + "epoch": 4.709348854128774, + "grad_norm": 0.6039117915339294, + "learning_rate": 1.169949841052294e-05, + "loss": 0.2057, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09724710881710052, + "step": 6475, + "valid_targets_mean": 3549.6, + "valid_targets_min": 1860 + }, + { + "epoch": 4.712986540560204, + "grad_norm": 0.5808388096549929, + "learning_rate": 1.1666514474463963e-05, + "loss": 0.2049, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0987212210893631, + "step": 6480, + "valid_targets_mean": 4414.0, + "valid_targets_min": 2495 + }, + { + "epoch": 4.716624226991633, + "grad_norm": 0.6312997787383451, + "learning_rate": 1.1633557943414797e-05, + "loss": 0.2058, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11610773950815201, + "step": 6485, + "valid_targets_mean": 3861.9, + "valid_targets_min": 1945 + }, + { + "epoch": 4.720261913423063, + "grad_norm": 0.6315792355497565, + "learning_rate": 1.1600628925754331e-05, + "loss": 0.2122, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10350020229816437, + "step": 6490, + "valid_targets_mean": 4488.5, + "valid_targets_min": 1908 + }, + { + "epoch": 4.723899599854493, + "grad_norm": 0.6223945096053899, + "learning_rate": 1.1567727529771004e-05, + "loss": 0.2047, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10728933662176132, + "step": 6495, + "valid_targets_mean": 3507.5, + "valid_targets_min": 1880 + }, + { + "epoch": 4.727537286285922, + "grad_norm": 0.6128033135692099, + "learning_rate": 1.1534853863662398e-05, + "loss": 0.202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10026122629642487, + "step": 6500, + "valid_targets_mean": 4491.5, + "valid_targets_min": 2492 + }, + { + "epoch": 4.731174972717351, + "grad_norm": 0.6149956899250494, + "learning_rate": 1.1502008035534924e-05, + "loss": 0.2058, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1003032699227333, + "step": 6505, + "valid_targets_mean": 4228.2, + "valid_targets_min": 3004 + }, + { + "epoch": 4.734812659148782, + "grad_norm": 0.6327099645966063, + "learning_rate": 1.1469190153403443e-05, + "loss": 0.1998, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10490190982818604, + "step": 6510, + "valid_targets_mean": 3899.9, + "valid_targets_min": 1833 + }, + { + "epoch": 4.738450345580211, + "grad_norm": 0.6761200181574752, + "learning_rate": 1.1436400325190898e-05, + "loss": 0.2015, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10284453630447388, + "step": 6515, + "valid_targets_mean": 3114.2, + "valid_targets_min": 1182 + }, + { + "epoch": 4.74208803201164, + "grad_norm": 0.5920982447417442, + "learning_rate": 1.1403638658727995e-05, + "loss": 0.2068, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10319872200489044, + "step": 6520, + "valid_targets_mean": 4926.9, + "valid_targets_min": 1869 + }, + { + "epoch": 4.745725718443071, + "grad_norm": 0.6076412512149691, + "learning_rate": 1.1370905261752815e-05, + "loss": 0.205, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11891406029462814, + "step": 6525, + "valid_targets_mean": 4382.4, + "valid_targets_min": 2498 + }, + { + "epoch": 4.7493634048745, + "grad_norm": 0.6957004528144002, + "learning_rate": 1.1338200241910485e-05, + "loss": 0.2169, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10094984620809555, + "step": 6530, + "valid_targets_mean": 3098.8, + "valid_targets_min": 988 + }, + { + "epoch": 4.753001091305929, + "grad_norm": 0.6416006694994362, + "learning_rate": 1.1305523706752812e-05, + "loss": 0.2014, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10813774913549423, + "step": 6535, + "valid_targets_mean": 3733.5, + "valid_targets_min": 1534 + }, + { + "epoch": 4.756638777737359, + "grad_norm": 0.7525045446436269, + "learning_rate": 1.1272875763737914e-05, + "loss": 0.2076, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08951827883720398, + "step": 6540, + "valid_targets_mean": 4108.1, + "valid_targets_min": 2252 + }, + { + "epoch": 4.760276464168789, + "grad_norm": 0.7705805691281489, + "learning_rate": 1.1240256520229904e-05, + "loss": 0.2092, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10944414138793945, + "step": 6545, + "valid_targets_mean": 3446.6, + "valid_targets_min": 1436 + }, + { + "epoch": 4.763914150600218, + "grad_norm": 0.6304784450106393, + "learning_rate": 1.120766608349849e-05, + "loss": 0.2058, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09495945274829865, + "step": 6550, + "valid_targets_mean": 3819.2, + "valid_targets_min": 2333 + }, + { + "epoch": 4.767551837031648, + "grad_norm": 0.6199576838366923, + "learning_rate": 1.1175104560718686e-05, + "loss": 0.1995, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09225665032863617, + "step": 6555, + "valid_targets_mean": 3889.8, + "valid_targets_min": 1272 + }, + { + "epoch": 4.771189523463077, + "grad_norm": 0.6299790923131132, + "learning_rate": 1.1142572058970389e-05, + "loss": 0.1991, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10326541215181351, + "step": 6560, + "valid_targets_mean": 3233.0, + "valid_targets_min": 1319 + }, + { + "epoch": 4.774827209894507, + "grad_norm": 0.5810992745952577, + "learning_rate": 1.111006868523806e-05, + "loss": 0.1869, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0835062563419342, + "step": 6565, + "valid_targets_mean": 3186.6, + "valid_targets_min": 1194 + }, + { + "epoch": 4.778464896325937, + "grad_norm": 0.6777203286911709, + "learning_rate": 1.1077594546410384e-05, + "loss": 0.2196, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0921110063791275, + "step": 6570, + "valid_targets_mean": 2710.6, + "valid_targets_min": 1453 + }, + { + "epoch": 4.782102582757366, + "grad_norm": 0.5535128285413637, + "learning_rate": 1.1045149749279905e-05, + "loss": 0.2023, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08525358140468597, + "step": 6575, + "valid_targets_mean": 5344.5, + "valid_targets_min": 3603 + }, + { + "epoch": 4.785740269188796, + "grad_norm": 0.7387148498524838, + "learning_rate": 1.1012734400542682e-05, + "loss": 0.214, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09214930236339569, + "step": 6580, + "valid_targets_mean": 3067.5, + "valid_targets_min": 1256 + }, + { + "epoch": 4.789377955620226, + "grad_norm": 0.6819167945030455, + "learning_rate": 1.09803486067979e-05, + "loss": 0.2075, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13691744208335876, + "step": 6585, + "valid_targets_mean": 4443.9, + "valid_targets_min": 1336 + }, + { + "epoch": 4.793015642051655, + "grad_norm": 0.6469616523721617, + "learning_rate": 1.0947992474547595e-05, + "loss": 0.2155, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10695698112249374, + "step": 6590, + "valid_targets_mean": 3581.8, + "valid_targets_min": 1275 + }, + { + "epoch": 4.796653328483085, + "grad_norm": 0.6441397077689758, + "learning_rate": 1.091566611019622e-05, + "loss": 0.2063, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11472078412771225, + "step": 6595, + "valid_targets_mean": 3770.1, + "valid_targets_min": 2202 + }, + { + "epoch": 4.800291014914515, + "grad_norm": 0.6305643586105915, + "learning_rate": 1.088336962005036e-05, + "loss": 0.1991, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11849327385425568, + "step": 6600, + "valid_targets_mean": 4112.8, + "valid_targets_min": 3332 + }, + { + "epoch": 4.803928701345944, + "grad_norm": 0.664274263957975, + "learning_rate": 1.0851103110318366e-05, + "loss": 0.1974, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10195624828338623, + "step": 6605, + "valid_targets_mean": 3260.0, + "valid_targets_min": 2195 + }, + { + "epoch": 4.807566387777373, + "grad_norm": 0.6401408442787736, + "learning_rate": 1.0818866687109957e-05, + "loss": 0.1973, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09940888732671738, + "step": 6610, + "valid_targets_mean": 4039.6, + "valid_targets_min": 2343 + }, + { + "epoch": 4.811204074208804, + "grad_norm": 0.6056448048052465, + "learning_rate": 1.0786660456435955e-05, + "loss": 0.1911, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0866222232580185, + "step": 6615, + "valid_targets_mean": 2678.8, + "valid_targets_min": 1026 + }, + { + "epoch": 4.814841760640233, + "grad_norm": 0.7288487140474567, + "learning_rate": 1.075448452420786e-05, + "loss": 0.1955, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09735588729381561, + "step": 6620, + "valid_targets_mean": 3754.5, + "valid_targets_min": 1042 + }, + { + "epoch": 4.818479447071662, + "grad_norm": 0.6342287083671793, + "learning_rate": 1.0722338996237553e-05, + "loss": 0.183, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09269433468580246, + "step": 6625, + "valid_targets_mean": 3301.6, + "valid_targets_min": 1274 + }, + { + "epoch": 4.822117133503092, + "grad_norm": 0.6955357985992475, + "learning_rate": 1.0690223978236936e-05, + "loss": 0.2003, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1000731810927391, + "step": 6630, + "valid_targets_mean": 3632.2, + "valid_targets_min": 914 + }, + { + "epoch": 4.825754819934522, + "grad_norm": 0.6033119526742248, + "learning_rate": 1.0658139575817548e-05, + "loss": 0.223, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09067384153604507, + "step": 6635, + "valid_targets_mean": 4342.5, + "valid_targets_min": 2093 + }, + { + "epoch": 4.829392506365951, + "grad_norm": 0.6650384208274835, + "learning_rate": 1.0626085894490272e-05, + "loss": 0.2028, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10550171136856079, + "step": 6640, + "valid_targets_mean": 3609.6, + "valid_targets_min": 1683 + }, + { + "epoch": 4.833030192797381, + "grad_norm": 0.6114423313188353, + "learning_rate": 1.0594063039664962e-05, + "loss": 0.2115, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.098120778799057, + "step": 6645, + "valid_targets_mean": 4096.2, + "valid_targets_min": 2253 + }, + { + "epoch": 4.836667879228811, + "grad_norm": 0.57324292497613, + "learning_rate": 1.0562071116650095e-05, + "loss": 0.2038, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09750476479530334, + "step": 6650, + "valid_targets_mean": 3794.4, + "valid_targets_min": 1789 + }, + { + "epoch": 4.84030556566024, + "grad_norm": 0.6789599799956623, + "learning_rate": 1.053011023065242e-05, + "loss": 0.2052, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10277897119522095, + "step": 6655, + "valid_targets_mean": 3753.8, + "valid_targets_min": 1586 + }, + { + "epoch": 4.84394325209167, + "grad_norm": 0.7061371048719564, + "learning_rate": 1.0498180486776617e-05, + "loss": 0.2024, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09115615487098694, + "step": 6660, + "valid_targets_mean": 2995.5, + "valid_targets_min": 2033 + }, + { + "epoch": 4.847580938523099, + "grad_norm": 0.6086176582018035, + "learning_rate": 1.046628199002497e-05, + "loss": 0.1924, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09480108320713043, + "step": 6665, + "valid_targets_mean": 3690.5, + "valid_targets_min": 1616 + }, + { + "epoch": 4.851218624954529, + "grad_norm": 0.6220999994634886, + "learning_rate": 1.0434414845296992e-05, + "loss": 0.2049, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09085287153720856, + "step": 6670, + "valid_targets_mean": 3303.9, + "valid_targets_min": 2000 + }, + { + "epoch": 4.854856311385959, + "grad_norm": 0.6257335373278716, + "learning_rate": 1.0402579157389115e-05, + "loss": 0.1976, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1091274544596672, + "step": 6675, + "valid_targets_mean": 3788.9, + "valid_targets_min": 1750 + }, + { + "epoch": 4.858493997817388, + "grad_norm": 0.6996574680250727, + "learning_rate": 1.0370775030994295e-05, + "loss": 0.2031, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11670731753110886, + "step": 6680, + "valid_targets_mean": 3898.2, + "valid_targets_min": 2126 + }, + { + "epoch": 4.862131684248817, + "grad_norm": 0.670680506234839, + "learning_rate": 1.0339002570701702e-05, + "loss": 0.2146, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11267900466918945, + "step": 6685, + "valid_targets_mean": 3923.1, + "valid_targets_min": 3049 + }, + { + "epoch": 4.865769370680248, + "grad_norm": 0.6471070911418823, + "learning_rate": 1.0307261880996387e-05, + "loss": 0.2037, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11813917011022568, + "step": 6690, + "valid_targets_mean": 4337.1, + "valid_targets_min": 1571 + }, + { + "epoch": 4.869407057111677, + "grad_norm": 0.7161860924846117, + "learning_rate": 1.027555306625891e-05, + "loss": 0.2151, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09921123087406158, + "step": 6695, + "valid_targets_mean": 2998.0, + "valid_targets_min": 1487 + }, + { + "epoch": 4.873044743543106, + "grad_norm": 0.6094441727035125, + "learning_rate": 1.0243876230765024e-05, + "loss": 0.2067, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0853734165430069, + "step": 6700, + "valid_targets_mean": 3627.1, + "valid_targets_min": 1970 + }, + { + "epoch": 4.876682429974537, + "grad_norm": 0.642595027873774, + "learning_rate": 1.0212231478685286e-05, + "loss": 0.2193, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09581203758716583, + "step": 6705, + "valid_targets_mean": 4038.0, + "valid_targets_min": 2492 + }, + { + "epoch": 4.880320116405966, + "grad_norm": 0.6770529583245282, + "learning_rate": 1.0180618914084776e-05, + "loss": 0.2154, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.090870201587677, + "step": 6710, + "valid_targets_mean": 3211.2, + "valid_targets_min": 1066 + }, + { + "epoch": 4.883957802837395, + "grad_norm": 0.6391918809873867, + "learning_rate": 1.0149038640922715e-05, + "loss": 0.2132, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09326095879077911, + "step": 6715, + "valid_targets_mean": 3401.9, + "valid_targets_min": 2281 + }, + { + "epoch": 4.887595489268825, + "grad_norm": 0.669366365434005, + "learning_rate": 1.011749076305212e-05, + "loss": 0.2136, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10599018633365631, + "step": 6720, + "valid_targets_mean": 3662.2, + "valid_targets_min": 1969 + }, + { + "epoch": 4.891233175700255, + "grad_norm": 0.6513652631164422, + "learning_rate": 1.0085975384219493e-05, + "loss": 0.2068, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09536653757095337, + "step": 6725, + "valid_targets_mean": 3189.6, + "valid_targets_min": 1005 + }, + { + "epoch": 4.894870862131684, + "grad_norm": 0.6861482702276924, + "learning_rate": 1.0054492608064446e-05, + "loss": 0.2076, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0931592509150505, + "step": 6730, + "valid_targets_mean": 3546.5, + "valid_targets_min": 1980 + }, + { + "epoch": 4.898508548563114, + "grad_norm": 0.647561915271615, + "learning_rate": 1.0023042538119383e-05, + "loss": 0.2117, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12084192782640457, + "step": 6735, + "valid_targets_mean": 4803.1, + "valid_targets_min": 1257 + }, + { + "epoch": 4.902146234994543, + "grad_norm": 0.6597306331161269, + "learning_rate": 9.991625277809167e-06, + "loss": 0.2, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10438212752342224, + "step": 6740, + "valid_targets_mean": 3118.1, + "valid_targets_min": 1357 + }, + { + "epoch": 4.905783921425973, + "grad_norm": 0.6439657246544176, + "learning_rate": 9.960240930450728e-06, + "loss": 0.1918, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08579470962285995, + "step": 6745, + "valid_targets_mean": 4463.8, + "valid_targets_min": 1410 + }, + { + "epoch": 4.909421607857403, + "grad_norm": 0.6417397967600736, + "learning_rate": 9.928889599252802e-06, + "loss": 0.209, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09263204038143158, + "step": 6750, + "valid_targets_mean": 3429.8, + "valid_targets_min": 1390 + }, + { + "epoch": 4.913059294288832, + "grad_norm": 0.6818690025311759, + "learning_rate": 9.897571387315507e-06, + "loss": 0.1973, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08999520540237427, + "step": 6755, + "valid_targets_mean": 3173.0, + "valid_targets_min": 790 + }, + { + "epoch": 4.916696980720262, + "grad_norm": 0.6223426730897159, + "learning_rate": 9.866286397630102e-06, + "loss": 0.2167, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1105850487947464, + "step": 6760, + "valid_targets_mean": 3788.8, + "valid_targets_min": 1901 + }, + { + "epoch": 4.920334667151692, + "grad_norm": 0.6664160767132498, + "learning_rate": 9.835034733078538e-06, + "loss": 0.2039, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09159524738788605, + "step": 6765, + "valid_targets_mean": 2697.1, + "valid_targets_min": 1460 + }, + { + "epoch": 4.923972353583121, + "grad_norm": 0.642574783692351, + "learning_rate": 9.803816496433213e-06, + "loss": 0.2064, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12431442737579346, + "step": 6770, + "valid_targets_mean": 5088.8, + "valid_targets_min": 952 + }, + { + "epoch": 4.92761004001455, + "grad_norm": 0.7402594423835748, + "learning_rate": 9.77263179035657e-06, + "loss": 0.2061, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10670942068099976, + "step": 6775, + "valid_targets_mean": 3070.5, + "valid_targets_min": 1768 + }, + { + "epoch": 4.931247726445981, + "grad_norm": 0.8283170830822155, + "learning_rate": 9.741480717400803e-06, + "loss": 0.208, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08409285545349121, + "step": 6780, + "valid_targets_mean": 3585.0, + "valid_targets_min": 1931 + }, + { + "epoch": 4.93488541287741, + "grad_norm": 0.5615595814881356, + "learning_rate": 9.710363380007501e-06, + "loss": 0.2021, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09720393270254135, + "step": 6785, + "valid_targets_mean": 5915.4, + "valid_targets_min": 1233 + }, + { + "epoch": 4.938523099308839, + "grad_norm": 0.5763024366594448, + "learning_rate": 9.679279880507294e-06, + "loss": 0.1967, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09970632940530777, + "step": 6790, + "valid_targets_mean": 5310.6, + "valid_targets_min": 1741 + }, + { + "epoch": 4.942160785740269, + "grad_norm": 0.5885269178703451, + "learning_rate": 9.64823032111957e-06, + "loss": 0.199, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09312334656715393, + "step": 6795, + "valid_targets_mean": 3884.8, + "valid_targets_min": 856 + }, + { + "epoch": 4.945798472171699, + "grad_norm": 0.6250339486197213, + "learning_rate": 9.617214803952055e-06, + "loss": 0.1996, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10566447675228119, + "step": 6800, + "valid_targets_mean": 3649.5, + "valid_targets_min": 1062 + }, + { + "epoch": 4.949436158603128, + "grad_norm": 0.5840173497050609, + "learning_rate": 9.586233431000567e-06, + "loss": 0.1923, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0972767323255539, + "step": 6805, + "valid_targets_mean": 4412.6, + "valid_targets_min": 974 + }, + { + "epoch": 4.953073845034558, + "grad_norm": 0.656819854403405, + "learning_rate": 9.555286304148632e-06, + "loss": 0.1944, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10316130518913269, + "step": 6810, + "valid_targets_mean": 3077.5, + "valid_targets_min": 1447 + }, + { + "epoch": 4.956711531465988, + "grad_norm": 0.6168193027945126, + "learning_rate": 9.524373525167132e-06, + "loss": 0.1926, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11088547855615616, + "step": 6815, + "valid_targets_mean": 4663.0, + "valid_targets_min": 2144 + }, + { + "epoch": 4.960349217897417, + "grad_norm": 0.6643862278908077, + "learning_rate": 9.493495195714028e-06, + "loss": 0.1924, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11088418960571289, + "step": 6820, + "valid_targets_mean": 4543.1, + "valid_targets_min": 1515 + }, + { + "epoch": 4.963986904328847, + "grad_norm": 0.7361203603027506, + "learning_rate": 9.462651417333954e-06, + "loss": 0.2105, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1013547033071518, + "step": 6825, + "valid_targets_mean": 3034.9, + "valid_targets_min": 1349 + }, + { + "epoch": 4.967624590760277, + "grad_norm": 0.6387947801095519, + "learning_rate": 9.431842291457974e-06, + "loss": 0.2074, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.101687490940094, + "step": 6830, + "valid_targets_mean": 3150.4, + "valid_targets_min": 1906 + }, + { + "epoch": 4.971262277191706, + "grad_norm": 0.6827999755930357, + "learning_rate": 9.401067919403148e-06, + "loss": 0.1951, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10326459258794785, + "step": 6835, + "valid_targets_mean": 3655.1, + "valid_targets_min": 2076 + }, + { + "epoch": 4.974899963623136, + "grad_norm": 0.6061043907263892, + "learning_rate": 9.370328402372257e-06, + "loss": 0.1921, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08264756202697754, + "step": 6840, + "valid_targets_mean": 2994.0, + "valid_targets_min": 1021 + }, + { + "epoch": 4.978537650054565, + "grad_norm": 0.6839738798148667, + "learning_rate": 9.339623841453476e-06, + "loss": 0.203, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1000542864203453, + "step": 6845, + "valid_targets_mean": 3226.9, + "valid_targets_min": 654 + }, + { + "epoch": 4.982175336485994, + "grad_norm": 0.6383825240366025, + "learning_rate": 9.308954337620014e-06, + "loss": 0.1977, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11440642178058624, + "step": 6850, + "valid_targets_mean": 4790.8, + "valid_targets_min": 3074 + }, + { + "epoch": 4.985813022917425, + "grad_norm": 0.7500829972825747, + "learning_rate": 9.278319991729801e-06, + "loss": 0.2096, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10051862895488739, + "step": 6855, + "valid_targets_mean": 2604.0, + "valid_targets_min": 815 + }, + { + "epoch": 4.989450709348854, + "grad_norm": 0.6456201760932595, + "learning_rate": 9.24772090452514e-06, + "loss": 0.2021, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09308859705924988, + "step": 6860, + "valid_targets_mean": 3560.0, + "valid_targets_min": 2205 + }, + { + "epoch": 4.993088395780283, + "grad_norm": 0.628166341862447, + "learning_rate": 9.217157176632378e-06, + "loss": 0.209, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11224481463432312, + "step": 6865, + "valid_targets_mean": 4216.0, + "valid_targets_min": 1793 + }, + { + "epoch": 4.996726082211714, + "grad_norm": 0.6335218311355936, + "learning_rate": 9.1866289085616e-06, + "loss": 0.2075, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09714855253696442, + "step": 6870, + "valid_targets_mean": 3134.5, + "valid_targets_min": 1217 + }, + { + "epoch": 5.0, + "grad_norm": 0.7624139079173295, + "learning_rate": 9.156136200706265e-06, + "loss": 0.1839, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08307020366191864, + "step": 6875, + "valid_targets_mean": 1910.0, + "valid_targets_min": 393 + }, + { + "epoch": 5.003637686431429, + "grad_norm": 0.4834850969299207, + "learning_rate": 9.125679153342912e-06, + "loss": 0.1068, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04412412643432617, + "step": 6880, + "valid_targets_mean": 2722.6, + "valid_targets_min": 458 + }, + { + "epoch": 5.00727537286286, + "grad_norm": 0.4524256162485089, + "learning_rate": 9.095257866630774e-06, + "loss": 0.0827, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04133915156126022, + "step": 6885, + "valid_targets_mean": 3113.1, + "valid_targets_min": 2461 + }, + { + "epoch": 5.010913059294289, + "grad_norm": 0.37637083500833096, + "learning_rate": 9.064872440611516e-06, + "loss": 0.0745, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.042610540986061096, + "step": 6890, + "valid_targets_mean": 3648.1, + "valid_targets_min": 2068 + }, + { + "epoch": 5.014550745725718, + "grad_norm": 0.5544781662590217, + "learning_rate": 9.034522975208851e-06, + "loss": 0.0965, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.045214422047138214, + "step": 6895, + "valid_targets_mean": 3423.6, + "valid_targets_min": 2425 + }, + { + "epoch": 5.018188432157148, + "grad_norm": 0.4460210770090794, + "learning_rate": 9.004209570228249e-06, + "loss": 0.0763, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.038082271814346313, + "step": 6900, + "valid_targets_mean": 2724.1, + "valid_targets_min": 1198 + }, + { + "epoch": 5.021826118588578, + "grad_norm": 0.9147056737042617, + "learning_rate": 8.973932325356595e-06, + "loss": 0.0917, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06764447689056396, + "step": 6905, + "valid_targets_mean": 1232.5, + "valid_targets_min": 807 + }, + { + "epoch": 5.025463805020007, + "grad_norm": 0.4646045707834218, + "learning_rate": 8.943691340161843e-06, + "loss": 0.1281, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04966093972325325, + "step": 6910, + "valid_targets_mean": 3487.4, + "valid_targets_min": 2351 + }, + { + "epoch": 5.029101491451437, + "grad_norm": 0.5646278847378118, + "learning_rate": 8.913486714092719e-06, + "loss": 0.0857, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06584298610687256, + "step": 6915, + "valid_targets_mean": 3126.4, + "valid_targets_min": 781 + }, + { + "epoch": 5.032739177882866, + "grad_norm": 0.361297803418113, + "learning_rate": 8.883318546478386e-06, + "loss": 0.0934, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04038098827004433, + "step": 6920, + "valid_targets_mean": 3474.8, + "valid_targets_min": 799 + }, + { + "epoch": 5.036376864314296, + "grad_norm": 0.36360732483251107, + "learning_rate": 8.853186936528089e-06, + "loss": 0.0993, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03646401688456535, + "step": 6925, + "valid_targets_mean": 5026.1, + "valid_targets_min": 2802 + }, + { + "epoch": 5.040014550745726, + "grad_norm": 0.3844442731231517, + "learning_rate": 8.823091983330883e-06, + "loss": 0.074, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029461167752742767, + "step": 6930, + "valid_targets_mean": 2611.0, + "valid_targets_min": 574 + }, + { + "epoch": 5.043652237177155, + "grad_norm": 0.3782788384169436, + "learning_rate": 8.793033785855243e-06, + "loss": 0.0811, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03839336335659027, + "step": 6935, + "valid_targets_mean": 4036.5, + "valid_targets_min": 2642 + }, + { + "epoch": 5.047289923608585, + "grad_norm": 0.537141705177125, + "learning_rate": 8.763012442948795e-06, + "loss": 0.0838, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.054927848279476166, + "step": 6940, + "valid_targets_mean": 1960.5, + "valid_targets_min": 593 + }, + { + "epoch": 5.050927610040015, + "grad_norm": 0.4260798291634585, + "learning_rate": 8.733028053337957e-06, + "loss": 0.1083, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0319986566901207, + "step": 6945, + "valid_targets_mean": 3354.2, + "valid_targets_min": 1353 + }, + { + "epoch": 5.054565296471444, + "grad_norm": 0.5438896879617444, + "learning_rate": 8.703080715627639e-06, + "loss": 0.0761, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04345440864562988, + "step": 6950, + "valid_targets_mean": 2300.9, + "valid_targets_min": 818 + }, + { + "epoch": 5.058202982902873, + "grad_norm": 0.3480282170476537, + "learning_rate": 8.67317052830088e-06, + "loss": 0.0759, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.026791565120220184, + "step": 6955, + "valid_targets_mean": 2295.8, + "valid_targets_min": 934 + }, + { + "epoch": 5.061840669334304, + "grad_norm": 0.3752241000215591, + "learning_rate": 8.643297589718557e-06, + "loss": 0.0694, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.037308964878320694, + "step": 6960, + "valid_targets_mean": 3524.4, + "valid_targets_min": 2798 + }, + { + "epoch": 5.065478355765733, + "grad_norm": 0.5633050414702512, + "learning_rate": 8.613461998119066e-06, + "loss": 0.0891, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.034928228706121445, + "step": 6965, + "valid_targets_mean": 2810.1, + "valid_targets_min": 471 + }, + { + "epoch": 5.069116042197162, + "grad_norm": 0.47103848359724954, + "learning_rate": 8.583663851617971e-06, + "loss": 0.0894, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05751679092645645, + "step": 6970, + "valid_targets_mean": 3042.8, + "valid_targets_min": 903 + }, + { + "epoch": 5.072753728628593, + "grad_norm": 0.43996409955396976, + "learning_rate": 8.553903248207714e-06, + "loss": 0.0885, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.034712422639131546, + "step": 6975, + "valid_targets_mean": 3081.6, + "valid_targets_min": 474 + }, + { + "epoch": 5.076391415060022, + "grad_norm": 0.4337826202048859, + "learning_rate": 8.524180285757253e-06, + "loss": 0.09, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.042026765644550323, + "step": 6980, + "valid_targets_mean": 3468.0, + "valid_targets_min": 640 + }, + { + "epoch": 5.080029101491451, + "grad_norm": 0.3402385634337198, + "learning_rate": 8.49449506201176e-06, + "loss": 0.1205, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029103241860866547, + "step": 6985, + "valid_targets_mean": 3438.4, + "valid_targets_min": 902 + }, + { + "epoch": 5.083666787922881, + "grad_norm": 0.3998533895683887, + "learning_rate": 8.464847674592338e-06, + "loss": 0.0745, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0369945764541626, + "step": 6990, + "valid_targets_mean": 3212.8, + "valid_targets_min": 1837 + }, + { + "epoch": 5.087304474354311, + "grad_norm": 0.74767614698203, + "learning_rate": 8.435238220995623e-06, + "loss": 0.1024, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07584372907876968, + "step": 6995, + "valid_targets_mean": 1358.8, + "valid_targets_min": 756 + }, + { + "epoch": 5.09094216078574, + "grad_norm": 0.5607369340517905, + "learning_rate": 8.405666798593532e-06, + "loss": 0.1033, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07532108575105667, + "step": 7000, + "valid_targets_mean": 2896.2, + "valid_targets_min": 1072 + }, + { + "epoch": 5.09457984721717, + "grad_norm": 0.4152532316521971, + "learning_rate": 8.376133504632893e-06, + "loss": 0.0813, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.025922097265720367, + "step": 7005, + "valid_targets_mean": 2908.1, + "valid_targets_min": 895 + }, + { + "epoch": 5.098217533648599, + "grad_norm": 0.6659061911118538, + "learning_rate": 8.34663843623516e-06, + "loss": 0.1937, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17876693606376648, + "step": 7010, + "valid_targets_mean": 2562.2, + "valid_targets_min": 878 + }, + { + "epoch": 5.101855220080029, + "grad_norm": 0.5447428433003172, + "learning_rate": 8.31718169039609e-06, + "loss": 0.1789, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0368511937558651, + "step": 7015, + "valid_targets_mean": 2087.6, + "valid_targets_min": 756 + }, + { + "epoch": 5.105492906511459, + "grad_norm": 0.3522234598360031, + "learning_rate": 8.287763363985385e-06, + "loss": 0.0906, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03206147253513336, + "step": 7020, + "valid_targets_mean": 3421.8, + "valid_targets_min": 2963 + }, + { + "epoch": 5.109130592942888, + "grad_norm": 0.8072401532747078, + "learning_rate": 8.258383553746435e-06, + "loss": 0.1598, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21943265199661255, + "step": 7025, + "valid_targets_mean": 2375.2, + "valid_targets_min": 1153 + }, + { + "epoch": 5.112768279374318, + "grad_norm": 0.2924197162303623, + "learning_rate": 8.229042356295939e-06, + "loss": 0.1002, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.031125444918870926, + "step": 7030, + "valid_targets_mean": 4886.2, + "valid_targets_min": 1700 + }, + { + "epoch": 5.116405965805748, + "grad_norm": 0.35815181324557743, + "learning_rate": 8.199739868123645e-06, + "loss": 0.0809, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.026504548266530037, + "step": 7035, + "valid_targets_mean": 3965.1, + "valid_targets_min": 789 + }, + { + "epoch": 5.120043652237177, + "grad_norm": 0.4352974234085096, + "learning_rate": 8.170476185591988e-06, + "loss": 0.0865, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04590744525194168, + "step": 7040, + "valid_targets_mean": 3771.5, + "valid_targets_min": 995 + }, + { + "epoch": 5.123681338668606, + "grad_norm": 0.5244366945688731, + "learning_rate": 8.141251404935783e-06, + "loss": 0.0776, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02735268324613571, + "step": 7045, + "valid_targets_mean": 2220.8, + "valid_targets_min": 488 + }, + { + "epoch": 5.127319025100037, + "grad_norm": 0.4311791412392709, + "learning_rate": 8.112065622261937e-06, + "loss": 0.0787, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04630735516548157, + "step": 7050, + "valid_targets_mean": 3635.4, + "valid_targets_min": 2437 + }, + { + "epoch": 5.130956711531466, + "grad_norm": 0.4950691392608991, + "learning_rate": 8.082918933549073e-06, + "loss": 0.0739, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03328017145395279, + "step": 7055, + "valid_targets_mean": 1752.1, + "valid_targets_min": 712 + }, + { + "epoch": 5.134594397962895, + "grad_norm": 0.45890412692132476, + "learning_rate": 8.053811434647306e-06, + "loss": 0.0757, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.038077034056186676, + "step": 7060, + "valid_targets_mean": 3713.4, + "valid_targets_min": 2422 + }, + { + "epoch": 5.138232084394325, + "grad_norm": 0.41816426709132404, + "learning_rate": 8.024743221277817e-06, + "loss": 0.0675, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03780101612210274, + "step": 7065, + "valid_targets_mean": 3998.5, + "valid_targets_min": 3643 + }, + { + "epoch": 5.141869770825755, + "grad_norm": 0.3693113102650201, + "learning_rate": 7.995714389032638e-06, + "loss": 0.1197, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02686825394630432, + "step": 7070, + "valid_targets_mean": 3319.0, + "valid_targets_min": 772 + }, + { + "epoch": 5.145507457257184, + "grad_norm": 0.39183647924131665, + "learning_rate": 7.966725033374265e-06, + "loss": 0.0613, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.025884218513965607, + "step": 7075, + "valid_targets_mean": 2934.8, + "valid_targets_min": 722 + }, + { + "epoch": 5.149145143688614, + "grad_norm": 0.3507917857410961, + "learning_rate": 7.93777524963539e-06, + "loss": 0.0685, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02814413234591484, + "step": 7080, + "valid_targets_mean": 3656.4, + "valid_targets_min": 2788 + }, + { + "epoch": 5.152782830120044, + "grad_norm": 0.41656482242304665, + "learning_rate": 7.90886513301858e-06, + "loss": 0.0706, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03668323531746864, + "step": 7085, + "valid_targets_mean": 3202.5, + "valid_targets_min": 948 + }, + { + "epoch": 5.156420516551473, + "grad_norm": 0.6101615893638521, + "learning_rate": 7.879994778595918e-06, + "loss": 0.0872, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04316336661577225, + "step": 7090, + "valid_targets_mean": 1002.0, + "valid_targets_min": 490 + }, + { + "epoch": 5.160058202982903, + "grad_norm": 0.4006887390912012, + "learning_rate": 7.851164281308775e-06, + "loss": 0.0986, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04276715964078903, + "step": 7095, + "valid_targets_mean": 3751.9, + "valid_targets_min": 2800 + }, + { + "epoch": 5.163695889414332, + "grad_norm": 0.5613817618190012, + "learning_rate": 7.822373735967408e-06, + "loss": 0.1059, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06156430020928383, + "step": 7100, + "valid_targets_mean": 2641.8, + "valid_targets_min": 896 + }, + { + "epoch": 5.167333575845762, + "grad_norm": 0.4823042705322211, + "learning_rate": 7.793623237250716e-06, + "loss": 0.0819, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.044846922159194946, + "step": 7105, + "valid_targets_mean": 3199.0, + "valid_targets_min": 1333 + }, + { + "epoch": 5.170971262277192, + "grad_norm": 0.5036532441685991, + "learning_rate": 7.764912879705905e-06, + "loss": 0.0963, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04325459524989128, + "step": 7110, + "valid_targets_mean": 2548.1, + "valid_targets_min": 693 + }, + { + "epoch": 5.174608948708621, + "grad_norm": 0.44941066125734125, + "learning_rate": 7.736242757748152e-06, + "loss": 0.112, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03488607704639435, + "step": 7115, + "valid_targets_mean": 2482.9, + "valid_targets_min": 803 + }, + { + "epoch": 5.178246635140051, + "grad_norm": 0.4249349210498175, + "learning_rate": 7.707612965660345e-06, + "loss": 0.082, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04229549691081047, + "step": 7120, + "valid_targets_mean": 3239.2, + "valid_targets_min": 2525 + }, + { + "epoch": 5.181884321571481, + "grad_norm": 0.5114199302937515, + "learning_rate": 7.679023597592713e-06, + "loss": 0.1016, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03714393824338913, + "step": 7125, + "valid_targets_mean": 3806.1, + "valid_targets_min": 2393 + }, + { + "epoch": 5.18552200800291, + "grad_norm": 0.5054969964170948, + "learning_rate": 7.650474747562595e-06, + "loss": 0.0845, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03761124983429909, + "step": 7130, + "valid_targets_mean": 2990.9, + "valid_targets_min": 804 + }, + { + "epoch": 5.189159694434339, + "grad_norm": 0.4148108950481197, + "learning_rate": 7.621966509454044e-06, + "loss": 0.0793, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.040075432509183884, + "step": 7135, + "valid_targets_mean": 4885.8, + "valid_targets_min": 3695 + }, + { + "epoch": 5.19279738086577, + "grad_norm": 0.4943325242940144, + "learning_rate": 7.5934989770175685e-06, + "loss": 0.0714, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0408380925655365, + "step": 7140, + "valid_targets_mean": 3128.8, + "valid_targets_min": 596 + }, + { + "epoch": 5.196435067297199, + "grad_norm": 0.32523174785299647, + "learning_rate": 7.565072243869822e-06, + "loss": 0.0693, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0269252210855484, + "step": 7145, + "valid_targets_mean": 4567.8, + "valid_targets_min": 3804 + }, + { + "epoch": 5.200072753728628, + "grad_norm": 0.3703800843539961, + "learning_rate": 7.536686403493289e-06, + "loss": 0.0757, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030664190649986267, + "step": 7150, + "valid_targets_mean": 3853.1, + "valid_targets_min": 2862 + }, + { + "epoch": 5.203710440160058, + "grad_norm": 0.5926662665350232, + "learning_rate": 7.5083415492359804e-06, + "loss": 0.0946, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05477459728717804, + "step": 7155, + "valid_targets_mean": 1072.6, + "valid_targets_min": 497 + }, + { + "epoch": 5.207348126591488, + "grad_norm": 0.39502070380916837, + "learning_rate": 7.480037774311104e-06, + "loss": 0.0795, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04477190598845482, + "step": 7160, + "valid_targets_mean": 4191.0, + "valid_targets_min": 3918 + }, + { + "epoch": 5.210985813022917, + "grad_norm": 0.5558577392331604, + "learning_rate": 7.451775171796782e-06, + "loss": 0.0902, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03855915367603302, + "step": 7165, + "valid_targets_mean": 2121.9, + "valid_targets_min": 785 + }, + { + "epoch": 5.214623499454347, + "grad_norm": 0.3609861194116582, + "learning_rate": 7.423553834635753e-06, + "loss": 0.0645, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03197663277387619, + "step": 7170, + "valid_targets_mean": 3451.6, + "valid_targets_min": 905 + }, + { + "epoch": 5.218261185885777, + "grad_norm": 0.4531025416948121, + "learning_rate": 7.395373855635044e-06, + "loss": 0.0844, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.061146993190050125, + "step": 7175, + "valid_targets_mean": 2929.2, + "valid_targets_min": 657 + }, + { + "epoch": 5.221898872317206, + "grad_norm": 0.414385072397305, + "learning_rate": 7.367235327465683e-06, + "loss": 0.0769, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03954493999481201, + "step": 7180, + "valid_targets_mean": 3798.9, + "valid_targets_min": 3184 + }, + { + "epoch": 5.225536558748636, + "grad_norm": 0.5653485449546104, + "learning_rate": 7.3391383426623596e-06, + "loss": 0.0698, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04343792051076889, + "step": 7185, + "valid_targets_mean": 1463.9, + "valid_targets_min": 627 + }, + { + "epoch": 5.229174245180065, + "grad_norm": 0.5245644751143034, + "learning_rate": 7.311082993623173e-06, + "loss": 0.0773, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03775782138109207, + "step": 7190, + "valid_targets_mean": 1701.5, + "valid_targets_min": 827 + }, + { + "epoch": 5.232811931611495, + "grad_norm": 0.5144531735984624, + "learning_rate": 7.2830693726093e-06, + "loss": 0.0799, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03625495731830597, + "step": 7195, + "valid_targets_mean": 1789.9, + "valid_targets_min": 940 + }, + { + "epoch": 5.236449618042925, + "grad_norm": 0.44004366612786827, + "learning_rate": 7.255097571744669e-06, + "loss": 0.0935, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03633845970034599, + "step": 7200, + "valid_targets_mean": 2943.8, + "valid_targets_min": 958 + }, + { + "epoch": 5.240087304474354, + "grad_norm": 0.457179821949183, + "learning_rate": 7.22716768301571e-06, + "loss": 0.0762, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03227534890174866, + "step": 7205, + "valid_targets_mean": 1859.2, + "valid_targets_min": 719 + }, + { + "epoch": 5.243724990905784, + "grad_norm": 0.5098149973616596, + "learning_rate": 7.199279798271002e-06, + "loss": 0.0933, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03085312992334366, + "step": 7210, + "valid_targets_mean": 1703.9, + "valid_targets_min": 759 + }, + { + "epoch": 5.247362677337214, + "grad_norm": 0.7985610233132641, + "learning_rate": 7.171434009221001e-06, + "loss": 0.0973, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10171546787023544, + "step": 7215, + "valid_targets_mean": 1556.6, + "valid_targets_min": 521 + }, + { + "epoch": 5.251000363768643, + "grad_norm": 0.4889291012854042, + "learning_rate": 7.14363040743774e-06, + "loss": 0.0791, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0480334609746933, + "step": 7220, + "valid_targets_mean": 3267.6, + "valid_targets_min": 1032 + }, + { + "epoch": 5.2546380502000725, + "grad_norm": 0.3453403890211141, + "learning_rate": 7.115869084354492e-06, + "loss": 0.0662, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02615818940103054, + "step": 7225, + "valid_targets_mean": 3609.0, + "valid_targets_min": 2741 + }, + { + "epoch": 5.258275736631503, + "grad_norm": 0.3905508888577857, + "learning_rate": 7.0881501312655234e-06, + "loss": 0.0758, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03419473022222519, + "step": 7230, + "valid_targets_mean": 2735.8, + "valid_targets_min": 1692 + }, + { + "epoch": 5.261913423062932, + "grad_norm": 0.4197869160365788, + "learning_rate": 7.060473639325738e-06, + "loss": 0.1031, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.042926859110593796, + "step": 7235, + "valid_targets_mean": 3693.4, + "valid_targets_min": 2215 + }, + { + "epoch": 5.2655511094943614, + "grad_norm": 0.38437698160670863, + "learning_rate": 7.032839699550422e-06, + "loss": 0.0636, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.028215356171131134, + "step": 7240, + "valid_targets_mean": 3109.9, + "valid_targets_min": 723 + }, + { + "epoch": 5.269188795925791, + "grad_norm": 0.3577499927109396, + "learning_rate": 7.005248402814924e-06, + "loss": 0.078, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.033782362937927246, + "step": 7245, + "valid_targets_mean": 4984.2, + "valid_targets_min": 3282 + }, + { + "epoch": 5.272826482357221, + "grad_norm": 0.3712406317164906, + "learning_rate": 6.977699839854366e-06, + "loss": 0.0739, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.031025735661387444, + "step": 7250, + "valid_targets_mean": 3863.4, + "valid_targets_min": 995 + }, + { + "epoch": 5.27646416878865, + "grad_norm": 0.3126808338434537, + "learning_rate": 6.950194101263321e-06, + "loss": 0.0626, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02998102828860283, + "step": 7255, + "valid_targets_mean": 4201.0, + "valid_targets_min": 949 + }, + { + "epoch": 5.28010185522008, + "grad_norm": 0.4632624097519686, + "learning_rate": 6.922731277495532e-06, + "loss": 0.0829, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05324278026819229, + "step": 7260, + "valid_targets_mean": 4321.6, + "valid_targets_min": 2724 + }, + { + "epoch": 5.28373954165151, + "grad_norm": 0.3388166905076338, + "learning_rate": 6.895311458863645e-06, + "loss": 0.0694, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.022668879479169846, + "step": 7265, + "valid_targets_mean": 3384.5, + "valid_targets_min": 424 + }, + { + "epoch": 5.287377228082939, + "grad_norm": 0.408976501324214, + "learning_rate": 6.867934735538844e-06, + "loss": 0.1, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02927073836326599, + "step": 7270, + "valid_targets_mean": 2987.2, + "valid_targets_min": 531 + }, + { + "epoch": 5.291014914514369, + "grad_norm": 0.3439387013032728, + "learning_rate": 6.8406011975506224e-06, + "loss": 0.0957, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029715554788708687, + "step": 7275, + "valid_targets_mean": 3469.5, + "valid_targets_min": 1195 + }, + { + "epoch": 5.294652600945798, + "grad_norm": 0.4197391547382634, + "learning_rate": 6.81331093478643e-06, + "loss": 0.064, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05956321954727173, + "step": 7280, + "valid_targets_mean": 2682.8, + "valid_targets_min": 722 + }, + { + "epoch": 5.298290287377228, + "grad_norm": 0.39335163601540973, + "learning_rate": 6.786064036991426e-06, + "loss": 0.0916, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03253050893545151, + "step": 7285, + "valid_targets_mean": 3496.9, + "valid_targets_min": 1251 + }, + { + "epoch": 5.301927973808658, + "grad_norm": 0.42679330437063373, + "learning_rate": 6.758860593768159e-06, + "loss": 0.0728, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03972616791725159, + "step": 7290, + "valid_targets_mean": 2912.5, + "valid_targets_min": 970 + }, + { + "epoch": 5.305565660240087, + "grad_norm": 0.6127313493446648, + "learning_rate": 6.73170069457626e-06, + "loss": 0.0817, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.039961330592632294, + "step": 7295, + "valid_targets_mean": 1423.6, + "valid_targets_min": 634 + }, + { + "epoch": 5.3092033466715165, + "grad_norm": 1.1822510291790165, + "learning_rate": 6.704584428732184e-06, + "loss": 0.0959, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06498440355062485, + "step": 7300, + "valid_targets_mean": 1423.9, + "valid_targets_min": 685 + }, + { + "epoch": 5.312841033102947, + "grad_norm": 0.44059822907617, + "learning_rate": 6.677511885408874e-06, + "loss": 0.1041, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.028044860810041428, + "step": 7305, + "valid_targets_mean": 2932.5, + "valid_targets_min": 593 + }, + { + "epoch": 5.316478719534376, + "grad_norm": 0.37190396515625207, + "learning_rate": 6.65048315363551e-06, + "loss": 0.0792, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.028502220287919044, + "step": 7310, + "valid_targets_mean": 3945.8, + "valid_targets_min": 3129 + }, + { + "epoch": 5.3201164059658055, + "grad_norm": 0.49603316257058255, + "learning_rate": 6.623498322297192e-06, + "loss": 0.0816, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04402817785739899, + "step": 7315, + "valid_targets_mean": 3621.1, + "valid_targets_min": 859 + }, + { + "epoch": 5.323754092397236, + "grad_norm": 0.5247125214869726, + "learning_rate": 6.596557480134638e-06, + "loss": 0.0734, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03837627172470093, + "step": 7320, + "valid_targets_mean": 2553.0, + "valid_targets_min": 616 + }, + { + "epoch": 5.327391778828665, + "grad_norm": 0.4443032346650512, + "learning_rate": 6.569660715743924e-06, + "loss": 0.0714, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03224431723356247, + "step": 7325, + "valid_targets_mean": 2690.5, + "valid_targets_min": 541 + }, + { + "epoch": 5.3310294652600945, + "grad_norm": 0.7849505296062362, + "learning_rate": 6.542808117576149e-06, + "loss": 0.0684, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04495672881603241, + "step": 7330, + "valid_targets_mean": 906.4, + "valid_targets_min": 598 + }, + { + "epoch": 5.334667151691524, + "grad_norm": 0.4454796819220204, + "learning_rate": 6.515999773937211e-06, + "loss": 0.0834, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03678368031978607, + "step": 7335, + "valid_targets_mean": 3147.2, + "valid_targets_min": 1101 + }, + { + "epoch": 5.338304838122954, + "grad_norm": 0.5130910841858659, + "learning_rate": 6.4892357729874365e-06, + "loss": 0.0691, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03200775757431984, + "step": 7340, + "valid_targets_mean": 2331.8, + "valid_targets_min": 680 + }, + { + "epoch": 5.3419425245543835, + "grad_norm": 0.40070113312269956, + "learning_rate": 6.46251620274134e-06, + "loss": 0.0691, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.036414578557014465, + "step": 7345, + "valid_targets_mean": 2577.0, + "valid_targets_min": 774 + }, + { + "epoch": 5.345580210985813, + "grad_norm": 0.48288558734619136, + "learning_rate": 6.435841151067328e-06, + "loss": 0.0852, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04752586409449577, + "step": 7350, + "valid_targets_mean": 4140.1, + "valid_targets_min": 1474 + }, + { + "epoch": 5.349217897417242, + "grad_norm": 0.27002504869008, + "learning_rate": 6.40921070568741e-06, + "loss": 0.0651, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.023120813071727753, + "step": 7355, + "valid_targets_mean": 5498.9, + "valid_targets_min": 3677 + }, + { + "epoch": 5.352855583848672, + "grad_norm": 0.4249989280005605, + "learning_rate": 6.382624954176904e-06, + "loss": 0.0707, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04140477627515793, + "step": 7360, + "valid_targets_mean": 3352.5, + "valid_targets_min": 675 + }, + { + "epoch": 5.356493270280102, + "grad_norm": 0.3952269695870162, + "learning_rate": 6.3560839839641344e-06, + "loss": 0.0684, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029594851657748222, + "step": 7365, + "valid_targets_mean": 3902.5, + "valid_targets_min": 795 + }, + { + "epoch": 5.360130956711531, + "grad_norm": 0.5407431657687977, + "learning_rate": 6.3295878823301855e-06, + "loss": 0.1686, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07363144308328629, + "step": 7370, + "valid_targets_mean": 2495.0, + "valid_targets_min": 605 + }, + { + "epoch": 5.363768643142961, + "grad_norm": 0.36985750760100605, + "learning_rate": 6.3031367364085665e-06, + "loss": 0.0675, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0338042676448822, + "step": 7375, + "valid_targets_mean": 3298.6, + "valid_targets_min": 985 + }, + { + "epoch": 5.367406329574391, + "grad_norm": 0.4603686243622257, + "learning_rate": 6.27673063318496e-06, + "loss": 0.078, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03716839477419853, + "step": 7380, + "valid_targets_mean": 3505.2, + "valid_targets_min": 878 + }, + { + "epoch": 5.37104401600582, + "grad_norm": 0.6988699865952612, + "learning_rate": 6.250369659496935e-06, + "loss": 0.099, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11366989463567734, + "step": 7385, + "valid_targets_mean": 2289.5, + "valid_targets_min": 1214 + }, + { + "epoch": 5.3746817024372495, + "grad_norm": 0.7226723444227648, + "learning_rate": 6.22405390203362e-06, + "loss": 0.1708, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11264634877443314, + "step": 7390, + "valid_targets_mean": 8496.5, + "valid_targets_min": 6565 + }, + { + "epoch": 5.37831938886868, + "grad_norm": 0.5631734596033368, + "learning_rate": 6.197783447335479e-06, + "loss": 0.1835, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07904249429702759, + "step": 7395, + "valid_targets_mean": 5663.5, + "valid_targets_min": 3833 + }, + { + "epoch": 5.381957075300109, + "grad_norm": 0.4962181823891549, + "learning_rate": 6.1715583817939714e-06, + "loss": 0.1671, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08655880391597748, + "step": 7400, + "valid_targets_mean": 6654.1, + "valid_targets_min": 5324 + }, + { + "epoch": 5.3855947617315385, + "grad_norm": 0.4466542399955791, + "learning_rate": 6.1453787916513085e-06, + "loss": 0.1651, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08224038779735565, + "step": 7405, + "valid_targets_mean": 7635.1, + "valid_targets_min": 5672 + }, + { + "epoch": 5.389232448162969, + "grad_norm": 0.4716822694609194, + "learning_rate": 6.119244763000156e-06, + "loss": 0.1697, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07575672119855881, + "step": 7410, + "valid_targets_mean": 8517.0, + "valid_targets_min": 5045 + }, + { + "epoch": 5.392870134594398, + "grad_norm": 0.41323180146529975, + "learning_rate": 6.093156381783327e-06, + "loss": 0.1743, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0845370814204216, + "step": 7415, + "valid_targets_mean": 6759.1, + "valid_targets_min": 3152 + }, + { + "epoch": 5.3965078210258275, + "grad_norm": 0.3830320012336569, + "learning_rate": 6.067113733793544e-06, + "loss": 0.1774, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07888054847717285, + "step": 7420, + "valid_targets_mean": 7822.8, + "valid_targets_min": 4820 + }, + { + "epoch": 5.400145507457257, + "grad_norm": 0.3770076462550648, + "learning_rate": 6.041116904673125e-06, + "loss": 0.1673, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06678445637226105, + "step": 7425, + "valid_targets_mean": 6852.2, + "valid_targets_min": 4135 + }, + { + "epoch": 5.403783193888687, + "grad_norm": 0.4793426479734725, + "learning_rate": 6.01516597991372e-06, + "loss": 0.1645, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07082075625658035, + "step": 7430, + "valid_targets_mean": 4794.2, + "valid_targets_min": 3230 + }, + { + "epoch": 5.4074208803201165, + "grad_norm": 0.39208229182131793, + "learning_rate": 5.989261044856003e-06, + "loss": 0.1442, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0844431072473526, + "step": 7435, + "valid_targets_mean": 7691.0, + "valid_targets_min": 5900 + }, + { + "epoch": 5.411058566751546, + "grad_norm": 0.4196594766109365, + "learning_rate": 5.963402184689411e-06, + "loss": 0.1674, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07888361066579819, + "step": 7440, + "valid_targets_mean": 6510.9, + "valid_targets_min": 5639 + }, + { + "epoch": 5.414696253182976, + "grad_norm": 0.4388759353532579, + "learning_rate": 5.93758948445188e-06, + "loss": 0.1573, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08189737796783447, + "step": 7445, + "valid_targets_mean": 6096.8, + "valid_targets_min": 4669 + }, + { + "epoch": 5.4183339396144055, + "grad_norm": 0.41042450689497956, + "learning_rate": 5.911823029029531e-06, + "loss": 0.1639, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08374934643507004, + "step": 7450, + "valid_targets_mean": 6812.2, + "valid_targets_min": 5160 + }, + { + "epoch": 5.421971626045835, + "grad_norm": 0.38286517134780107, + "learning_rate": 5.8861029031564185e-06, + "loss": 0.157, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07253463566303253, + "step": 7455, + "valid_targets_mean": 6980.4, + "valid_targets_min": 5151 + }, + { + "epoch": 5.425609312477264, + "grad_norm": 0.3885344964294181, + "learning_rate": 5.860429191414232e-06, + "loss": 0.1584, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08540088683366776, + "step": 7460, + "valid_targets_mean": 7417.1, + "valid_targets_min": 5932 + }, + { + "epoch": 5.429246998908694, + "grad_norm": 0.3947423571918331, + "learning_rate": 5.834801978232019e-06, + "loss": 0.1663, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07878325134515762, + "step": 7465, + "valid_targets_mean": 6258.5, + "valid_targets_min": 5569 + }, + { + "epoch": 5.432884685340124, + "grad_norm": 0.40178902664448984, + "learning_rate": 5.809221347885929e-06, + "loss": 0.1589, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09002530574798584, + "step": 7470, + "valid_targets_mean": 7689.1, + "valid_targets_min": 6223 + }, + { + "epoch": 5.436522371771553, + "grad_norm": 0.5301398398897944, + "learning_rate": 5.7836873844989195e-06, + "loss": 0.1661, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09332171082496643, + "step": 7475, + "valid_targets_mean": 4371.9, + "valid_targets_min": 1172 + }, + { + "epoch": 5.4401600582029825, + "grad_norm": 0.44532335395901024, + "learning_rate": 5.758200172040484e-06, + "loss": 0.1754, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09414127469062805, + "step": 7480, + "valid_targets_mean": 6936.2, + "valid_targets_min": 5619 + }, + { + "epoch": 5.443797744634413, + "grad_norm": 0.6823466862164559, + "learning_rate": 5.732759794326355e-06, + "loss": 0.1473, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09428075700998306, + "step": 7485, + "valid_targets_mean": 3273.2, + "valid_targets_min": 442 + }, + { + "epoch": 5.447435431065842, + "grad_norm": 0.3763916133911704, + "learning_rate": 5.707366335018267e-06, + "loss": 0.1499, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0792102962732315, + "step": 7490, + "valid_targets_mean": 8141.1, + "valid_targets_min": 5061 + }, + { + "epoch": 5.4510731174972715, + "grad_norm": 0.4227743983781728, + "learning_rate": 5.682019877623666e-06, + "loss": 0.1587, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07964727282524109, + "step": 7495, + "valid_targets_mean": 7371.1, + "valid_targets_min": 5112 + }, + { + "epoch": 5.454710803928702, + "grad_norm": 0.4185294205454082, + "learning_rate": 5.656720505495399e-06, + "loss": 0.1574, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07004642486572266, + "step": 7500, + "valid_targets_mean": 5659.4, + "valid_targets_min": 4626 + }, + { + "epoch": 5.458348490360131, + "grad_norm": 0.39255911113358327, + "learning_rate": 5.631468301831509e-06, + "loss": 0.1639, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07255950570106506, + "step": 7505, + "valid_targets_mean": 7801.8, + "valid_targets_min": 6172 + }, + { + "epoch": 5.4619861767915605, + "grad_norm": 0.6664369217144795, + "learning_rate": 5.606263349674894e-06, + "loss": 0.1565, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07924732565879822, + "step": 7510, + "valid_targets_mean": 7547.4, + "valid_targets_min": 5804 + }, + { + "epoch": 5.46562386322299, + "grad_norm": 0.3945962420065962, + "learning_rate": 5.581105731913079e-06, + "loss": 0.1501, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07094433903694153, + "step": 7515, + "valid_targets_mean": 6474.2, + "valid_targets_min": 5050 + }, + { + "epoch": 5.46926154965442, + "grad_norm": 0.38893022004344274, + "learning_rate": 5.555995531277931e-06, + "loss": 0.14, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07292284071445465, + "step": 7520, + "valid_targets_mean": 7225.4, + "valid_targets_min": 5298 + }, + { + "epoch": 5.4728992360858495, + "grad_norm": 0.4211857811426324, + "learning_rate": 5.530932830345368e-06, + "loss": 0.1386, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06579746305942535, + "step": 7525, + "valid_targets_mean": 6442.9, + "valid_targets_min": 4679 + }, + { + "epoch": 5.476536922517279, + "grad_norm": 0.5210692938577276, + "learning_rate": 5.50591771153512e-06, + "loss": 0.1635, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06426037102937698, + "step": 7530, + "valid_targets_mean": 3651.9, + "valid_targets_min": 1801 + }, + { + "epoch": 5.480174608948708, + "grad_norm": 0.45082954658127883, + "learning_rate": 5.480950257110424e-06, + "loss": 0.1483, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09224918484687805, + "step": 7535, + "valid_targets_mean": 7512.8, + "valid_targets_min": 5331 + }, + { + "epoch": 5.4838122953801385, + "grad_norm": 0.4663344615443458, + "learning_rate": 5.4560305491777885e-06, + "loss": 0.1654, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08798010647296906, + "step": 7540, + "valid_targets_mean": 7108.8, + "valid_targets_min": 4822 + }, + { + "epoch": 5.487449981811568, + "grad_norm": 0.43754334151495755, + "learning_rate": 5.431158669686696e-06, + "loss": 0.1709, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07461196929216385, + "step": 7545, + "valid_targets_mean": 7125.8, + "valid_targets_min": 4623 + }, + { + "epoch": 5.491087668242997, + "grad_norm": 0.4088515194385727, + "learning_rate": 5.4063347004293495e-06, + "loss": 0.1685, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07658769190311432, + "step": 7550, + "valid_targets_mean": 6269.2, + "valid_targets_min": 5458 + }, + { + "epoch": 5.4947253546744275, + "grad_norm": 0.41253508428809654, + "learning_rate": 5.381558723040388e-06, + "loss": 0.1593, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08086318522691727, + "step": 7555, + "valid_targets_mean": 7455.1, + "valid_targets_min": 6008 + }, + { + "epoch": 5.498363041105857, + "grad_norm": 0.42166429378559966, + "learning_rate": 5.356830818996621e-06, + "loss": 0.1683, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08315180242061615, + "step": 7560, + "valid_targets_mean": 7104.4, + "valid_targets_min": 5386 + }, + { + "epoch": 5.502000727537286, + "grad_norm": 0.4176546262502344, + "learning_rate": 5.332151069616797e-06, + "loss": 0.1602, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0812249481678009, + "step": 7565, + "valid_targets_mean": 6532.6, + "valid_targets_min": 4855 + }, + { + "epoch": 5.505638413968716, + "grad_norm": 0.39777969370068905, + "learning_rate": 5.307519556061276e-06, + "loss": 0.164, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08043235540390015, + "step": 7570, + "valid_targets_mean": 6743.9, + "valid_targets_min": 5613 + }, + { + "epoch": 5.509276100400146, + "grad_norm": 0.4341566413591557, + "learning_rate": 5.2829363593318075e-06, + "loss": 0.1639, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08326408267021179, + "step": 7575, + "valid_targets_mean": 6373.6, + "valid_targets_min": 4961 + }, + { + "epoch": 5.512913786831575, + "grad_norm": 0.4452060644307391, + "learning_rate": 5.258401560271238e-06, + "loss": 0.1668, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08607415854930878, + "step": 7580, + "valid_targets_mean": 6461.2, + "valid_targets_min": 5286 + }, + { + "epoch": 5.5165514732630045, + "grad_norm": 0.43127354386240374, + "learning_rate": 5.2339152395632675e-06, + "loss": 0.162, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08456918597221375, + "step": 7585, + "valid_targets_mean": 6780.5, + "valid_targets_min": 5201 + }, + { + "epoch": 5.520189159694434, + "grad_norm": 0.48319159468571077, + "learning_rate": 5.209477477732172e-06, + "loss": 0.1609, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08016848564147949, + "step": 7590, + "valid_targets_mean": 5710.8, + "valid_targets_min": 4838 + }, + { + "epoch": 5.523826846125864, + "grad_norm": 0.6945525711970426, + "learning_rate": 5.185088355142529e-06, + "loss": 0.1276, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11619526147842407, + "step": 7595, + "valid_targets_mean": 3235.6, + "valid_targets_min": 176 + }, + { + "epoch": 5.5274645325572935, + "grad_norm": 0.42104441935961046, + "learning_rate": 5.160747951998979e-06, + "loss": 0.1633, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07417930662631989, + "step": 7600, + "valid_targets_mean": 6674.8, + "valid_targets_min": 4582 + }, + { + "epoch": 5.531102218988723, + "grad_norm": 0.430862440583841, + "learning_rate": 5.136456348345926e-06, + "loss": 0.1586, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0793251320719719, + "step": 7605, + "valid_targets_mean": 6157.8, + "valid_targets_min": 5385 + }, + { + "epoch": 5.534739905420153, + "grad_norm": 0.45638825100266367, + "learning_rate": 5.112213624067317e-06, + "loss": 0.1615, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07603725790977478, + "step": 7610, + "valid_targets_mean": 6006.2, + "valid_targets_min": 4478 + }, + { + "epoch": 5.5383775918515825, + "grad_norm": 0.4338304621505074, + "learning_rate": 5.088019858886353e-06, + "loss": 0.1579, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07923339307308197, + "step": 7615, + "valid_targets_mean": 5893.4, + "valid_targets_min": 3911 + }, + { + "epoch": 5.542015278283012, + "grad_norm": 0.4211385791137756, + "learning_rate": 5.06387513236521e-06, + "loss": 0.1571, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07844312489032745, + "step": 7620, + "valid_targets_mean": 6134.2, + "valid_targets_min": 5318 + }, + { + "epoch": 5.545652964714441, + "grad_norm": 0.4467076229639607, + "learning_rate": 5.039779523904824e-06, + "loss": 0.1648, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08192823082208633, + "step": 7625, + "valid_targets_mean": 5474.4, + "valid_targets_min": 4959 + }, + { + "epoch": 5.5492906511458715, + "grad_norm": 0.5944612779254935, + "learning_rate": 5.015733112744592e-06, + "loss": 0.1546, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04446254298090935, + "step": 7630, + "valid_targets_mean": 1804.1, + "valid_targets_min": 842 + }, + { + "epoch": 5.552928337577301, + "grad_norm": 0.45766188843610256, + "learning_rate": 4.991735977962131e-06, + "loss": 0.1289, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0722193643450737, + "step": 7635, + "valid_targets_mean": 6118.8, + "valid_targets_min": 4899 + }, + { + "epoch": 5.55656602400873, + "grad_norm": 0.41276875678140634, + "learning_rate": 4.967788198473e-06, + "loss": 0.1545, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0718635767698288, + "step": 7640, + "valid_targets_mean": 6773.8, + "valid_targets_min": 5046 + }, + { + "epoch": 5.56020371044016, + "grad_norm": 0.4791641885444281, + "learning_rate": 4.943889853030452e-06, + "loss": 0.1648, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08457650989294052, + "step": 7645, + "valid_targets_mean": 6633.6, + "valid_targets_min": 5284 + }, + { + "epoch": 5.56384139687159, + "grad_norm": 0.433982481251753, + "learning_rate": 4.920041020225186e-06, + "loss": 0.1567, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07967168837785721, + "step": 7650, + "valid_targets_mean": 6055.2, + "valid_targets_min": 4928 + }, + { + "epoch": 5.567479083303019, + "grad_norm": 0.4276445855415657, + "learning_rate": 4.896241778485065e-06, + "loss": 0.154, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08257752656936646, + "step": 7655, + "valid_targets_mean": 7259.4, + "valid_targets_min": 5758 + }, + { + "epoch": 5.571116769734449, + "grad_norm": 0.4425005567823872, + "learning_rate": 4.872492206074881e-06, + "loss": 0.1572, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08087639510631561, + "step": 7660, + "valid_targets_mean": 5731.8, + "valid_targets_min": 4629 + }, + { + "epoch": 5.574754456165879, + "grad_norm": 0.72206523431684, + "learning_rate": 4.84879238109607e-06, + "loss": 0.1779, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16300134360790253, + "step": 7665, + "valid_targets_mean": 5360.9, + "valid_targets_min": 2317 + }, + { + "epoch": 5.578392142597308, + "grad_norm": 0.6251541482722512, + "learning_rate": 4.825142381486492e-06, + "loss": 0.2177, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11427206546068192, + "step": 7670, + "valid_targets_mean": 5303.1, + "valid_targets_min": 533 + }, + { + "epoch": 5.582029829028738, + "grad_norm": 0.6132758337081321, + "learning_rate": 4.8015422850201285e-06, + "loss": 0.2142, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10906066000461578, + "step": 7675, + "valid_targets_mean": 4545.8, + "valid_targets_min": 835 + }, + { + "epoch": 5.585667515460168, + "grad_norm": 0.6548813988873464, + "learning_rate": 4.7779921693068755e-06, + "loss": 0.2099, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11494287103414536, + "step": 7680, + "valid_targets_mean": 3618.9, + "valid_targets_min": 1201 + }, + { + "epoch": 5.589305201891597, + "grad_norm": 0.5959109366699014, + "learning_rate": 4.754492111792262e-06, + "loss": 0.2009, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09357195347547531, + "step": 7685, + "valid_targets_mean": 4253.5, + "valid_targets_min": 1880 + }, + { + "epoch": 5.5929428883230266, + "grad_norm": 0.5654369879072534, + "learning_rate": 4.731042189757182e-06, + "loss": 0.2212, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15617477893829346, + "step": 7690, + "valid_targets_mean": 6610.1, + "valid_targets_min": 2171 + }, + { + "epoch": 5.596580574754456, + "grad_norm": 0.5005486719417936, + "learning_rate": 4.707642480317676e-06, + "loss": 0.2012, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11480434983968735, + "step": 7695, + "valid_targets_mean": 7171.4, + "valid_targets_min": 4500 + }, + { + "epoch": 5.600218261185885, + "grad_norm": 0.5155233694289717, + "learning_rate": 4.684293060424654e-06, + "loss": 0.2078, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11116056144237518, + "step": 7700, + "valid_targets_mean": 6207.4, + "valid_targets_min": 1855 + }, + { + "epoch": 5.6038559476173155, + "grad_norm": 0.6511227751992686, + "learning_rate": 4.660994006863631e-06, + "loss": 0.2052, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1022200882434845, + "step": 7705, + "valid_targets_mean": 4075.9, + "valid_targets_min": 1024 + }, + { + "epoch": 5.607493634048745, + "grad_norm": 0.5652163215386843, + "learning_rate": 4.637745396254516e-06, + "loss": 0.2175, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09313727915287018, + "step": 7710, + "valid_targets_mean": 4767.1, + "valid_targets_min": 2107 + }, + { + "epoch": 5.611131320480174, + "grad_norm": 0.6683621702520854, + "learning_rate": 4.614547305051307e-06, + "loss": 0.2098, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11404265463352203, + "step": 7715, + "valid_targets_mean": 4211.4, + "valid_targets_min": 802 + }, + { + "epoch": 5.6147690069116045, + "grad_norm": 0.4826386168825257, + "learning_rate": 4.59139980954189e-06, + "loss": 0.2005, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10423771291971207, + "step": 7720, + "valid_targets_mean": 5345.1, + "valid_targets_min": 1597 + }, + { + "epoch": 5.618406693343034, + "grad_norm": 0.6391862619711574, + "learning_rate": 4.568302985847755e-06, + "loss": 0.2144, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11117672920227051, + "step": 7725, + "valid_targets_mean": 3671.4, + "valid_targets_min": 1650 + }, + { + "epoch": 5.622044379774463, + "grad_norm": 0.6154642519794076, + "learning_rate": 4.545256909923761e-06, + "loss": 0.2047, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10608255863189697, + "step": 7730, + "valid_targets_mean": 4799.1, + "valid_targets_min": 1385 + }, + { + "epoch": 5.6256820662058935, + "grad_norm": 0.595529389778941, + "learning_rate": 4.5222616575578695e-06, + "loss": 0.1974, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10875016450881958, + "step": 7735, + "valid_targets_mean": 4023.5, + "valid_targets_min": 2908 + }, + { + "epoch": 5.629319752637323, + "grad_norm": 0.5609598929427366, + "learning_rate": 4.499317304370909e-06, + "loss": 0.203, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09192369878292084, + "step": 7740, + "valid_targets_mean": 5064.6, + "valid_targets_min": 2785 + }, + { + "epoch": 5.632957439068752, + "grad_norm": 0.7087611768209884, + "learning_rate": 4.476423925816333e-06, + "loss": 0.1971, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09869243204593658, + "step": 7745, + "valid_targets_mean": 3886.4, + "valid_targets_min": 1210 + }, + { + "epoch": 5.636595125500182, + "grad_norm": 0.6415787431633968, + "learning_rate": 4.4535815971799615e-06, + "loss": 0.2023, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11284425854682922, + "step": 7750, + "valid_targets_mean": 3708.8, + "valid_targets_min": 1064 + }, + { + "epoch": 5.640232811931612, + "grad_norm": 0.7083948936497451, + "learning_rate": 4.430790393579733e-06, + "loss": 0.2127, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10994810611009598, + "step": 7755, + "valid_targets_mean": 3276.2, + "valid_targets_min": 1678 + }, + { + "epoch": 5.643870498363041, + "grad_norm": 0.6377501540626633, + "learning_rate": 4.408050389965455e-06, + "loss": 0.2036, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10835543274879456, + "step": 7760, + "valid_targets_mean": 3520.1, + "valid_targets_min": 671 + }, + { + "epoch": 5.647508184794471, + "grad_norm": 0.6810719418522665, + "learning_rate": 4.3853616611185526e-06, + "loss": 0.2039, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09780985116958618, + "step": 7765, + "valid_targets_mean": 3488.5, + "valid_targets_min": 2000 + }, + { + "epoch": 5.6511458712259, + "grad_norm": 0.646361341447219, + "learning_rate": 4.362724281651862e-06, + "loss": 0.1867, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11614830791950226, + "step": 7770, + "valid_targets_mean": 4402.6, + "valid_targets_min": 1532 + }, + { + "epoch": 5.65478355765733, + "grad_norm": 0.6242984896204661, + "learning_rate": 4.340138326009326e-06, + "loss": 0.1948, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07546952366828918, + "step": 7775, + "valid_targets_mean": 3749.4, + "valid_targets_min": 1051 + }, + { + "epoch": 5.65842124408876, + "grad_norm": 0.6465283110325533, + "learning_rate": 4.317603868465794e-06, + "loss": 0.1872, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09486444294452667, + "step": 7780, + "valid_targets_mean": 3368.1, + "valid_targets_min": 1345 + }, + { + "epoch": 5.662058930520189, + "grad_norm": 0.643363597384762, + "learning_rate": 4.29512098312675e-06, + "loss": 0.2033, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09919878095388412, + "step": 7785, + "valid_targets_mean": 4128.9, + "valid_targets_min": 2231 + }, + { + "epoch": 5.665696616951619, + "grad_norm": 0.6942914315635799, + "learning_rate": 4.272689743928087e-06, + "loss": 0.2042, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09136289358139038, + "step": 7790, + "valid_targets_mean": 3672.6, + "valid_targets_min": 2127 + }, + { + "epoch": 5.669334303383049, + "grad_norm": 0.8256373371983696, + "learning_rate": 4.250310224635867e-06, + "loss": 0.1954, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09276250749826431, + "step": 7795, + "valid_targets_mean": 3006.8, + "valid_targets_min": 1265 + }, + { + "epoch": 5.672971989814478, + "grad_norm": 0.7209275067216181, + "learning_rate": 4.2279824988460485e-06, + "loss": 0.1948, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08123257011175156, + "step": 7800, + "valid_targets_mean": 2981.9, + "valid_targets_min": 908 + }, + { + "epoch": 5.676609676245907, + "grad_norm": 0.7756143686441812, + "learning_rate": 4.205706639984284e-06, + "loss": 0.1938, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10961326211690903, + "step": 7805, + "valid_targets_mean": 3147.5, + "valid_targets_min": 1264 + }, + { + "epoch": 5.6802473626773375, + "grad_norm": 0.646323572262216, + "learning_rate": 4.183482721305641e-06, + "loss": 0.1923, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08962377160787582, + "step": 7810, + "valid_targets_mean": 2517.9, + "valid_targets_min": 926 + }, + { + "epoch": 5.683885049108767, + "grad_norm": 0.7490080810059698, + "learning_rate": 4.161310815894397e-06, + "loss": 0.1948, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09486709535121918, + "step": 7815, + "valid_targets_mean": 3204.5, + "valid_targets_min": 1016 + }, + { + "epoch": 5.687522735540196, + "grad_norm": 0.6274757217627094, + "learning_rate": 4.13919099666378e-06, + "loss": 0.1911, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09284915030002594, + "step": 7820, + "valid_targets_mean": 3525.0, + "valid_targets_min": 2019 + }, + { + "epoch": 5.691160421971626, + "grad_norm": 0.6337358757582208, + "learning_rate": 4.1171233363557105e-06, + "loss": 0.193, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11828829348087311, + "step": 7825, + "valid_targets_mean": 4745.4, + "valid_targets_min": 3023 + }, + { + "epoch": 5.694798108403056, + "grad_norm": 0.6518366478971095, + "learning_rate": 4.095107907540612e-06, + "loss": 0.1879, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08485589921474457, + "step": 7830, + "valid_targets_mean": 3503.6, + "valid_targets_min": 1110 + }, + { + "epoch": 5.698435794834485, + "grad_norm": 0.6769842223446785, + "learning_rate": 4.073144782617111e-06, + "loss": 0.1886, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10957793146371841, + "step": 7835, + "valid_targets_mean": 4206.5, + "valid_targets_min": 1460 + }, + { + "epoch": 5.702073481265915, + "grad_norm": 0.6678581997495683, + "learning_rate": 4.051234033811864e-06, + "loss": 0.1888, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10500785708427429, + "step": 7840, + "valid_targets_mean": 3938.6, + "valid_targets_min": 2558 + }, + { + "epoch": 5.705711167697345, + "grad_norm": 0.6488094762629959, + "learning_rate": 4.029375733179259e-06, + "loss": 0.2003, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08887283504009247, + "step": 7845, + "valid_targets_mean": 3421.1, + "valid_targets_min": 1673 + }, + { + "epoch": 5.709348854128774, + "grad_norm": 0.6398799388211451, + "learning_rate": 4.007569952601222e-06, + "loss": 0.192, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09037714451551437, + "step": 7850, + "valid_targets_mean": 3549.6, + "valid_targets_min": 1860 + }, + { + "epoch": 5.712986540560204, + "grad_norm": 0.6566982246595748, + "learning_rate": 3.985816763786952e-06, + "loss": 0.1911, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09378645569086075, + "step": 7855, + "valid_targets_mean": 4414.0, + "valid_targets_min": 2495 + }, + { + "epoch": 5.716624226991633, + "grad_norm": 0.6548719534572124, + "learning_rate": 3.964116238272708e-06, + "loss": 0.1922, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10823580622673035, + "step": 7860, + "valid_targets_mean": 3861.9, + "valid_targets_min": 1945 + }, + { + "epoch": 5.720261913423063, + "grad_norm": 0.6542010312170575, + "learning_rate": 3.942468447421568e-06, + "loss": 0.2, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09585648775100708, + "step": 7865, + "valid_targets_mean": 4488.5, + "valid_targets_min": 1908 + }, + { + "epoch": 5.723899599854493, + "grad_norm": 0.6516372800181811, + "learning_rate": 3.920873462423167e-06, + "loss": 0.1903, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1004343181848526, + "step": 7870, + "valid_targets_mean": 3507.5, + "valid_targets_min": 1880 + }, + { + "epoch": 5.727537286285922, + "grad_norm": 0.6364549434976475, + "learning_rate": 3.899331354293518e-06, + "loss": 0.1878, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09232114255428314, + "step": 7875, + "valid_targets_mean": 4491.5, + "valid_targets_min": 2492 + }, + { + "epoch": 5.731174972717351, + "grad_norm": 0.6319612890529416, + "learning_rate": 3.877842193874714e-06, + "loss": 0.1917, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09266478568315506, + "step": 7880, + "valid_targets_mean": 4228.2, + "valid_targets_min": 3004 + }, + { + "epoch": 5.734812659148782, + "grad_norm": 0.66875681256989, + "learning_rate": 3.856406051834751e-06, + "loss": 0.1863, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0974518358707428, + "step": 7885, + "valid_targets_mean": 3899.9, + "valid_targets_min": 1833 + }, + { + "epoch": 5.738450345580211, + "grad_norm": 0.7341758357314973, + "learning_rate": 3.835022998667266e-06, + "loss": 0.1872, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09497388452291489, + "step": 7890, + "valid_targets_mean": 3114.2, + "valid_targets_min": 1182 + }, + { + "epoch": 5.74208803201164, + "grad_norm": 0.6551007040566378, + "learning_rate": 3.8136931046912983e-06, + "loss": 0.1935, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09711603820323944, + "step": 7895, + "valid_targets_mean": 4926.9, + "valid_targets_min": 1869 + }, + { + "epoch": 5.745725718443071, + "grad_norm": 0.6581176474393102, + "learning_rate": 3.7924164400510945e-06, + "loss": 0.1911, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11074376106262207, + "step": 7900, + "valid_targets_mean": 4382.4, + "valid_targets_min": 2498 + }, + { + "epoch": 5.7493634048745, + "grad_norm": 0.6954083797896067, + "learning_rate": 3.7711930747158242e-06, + "loss": 0.2019, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09420367330312729, + "step": 7905, + "valid_targets_mean": 3098.8, + "valid_targets_min": 988 + }, + { + "epoch": 5.753001091305929, + "grad_norm": 0.6312916647123977, + "learning_rate": 3.7500230784794145e-06, + "loss": 0.1881, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10167976468801498, + "step": 7910, + "valid_targets_mean": 3733.5, + "valid_targets_min": 1534 + }, + { + "epoch": 5.756638777737359, + "grad_norm": 0.6002791424800878, + "learning_rate": 3.7289065209602625e-06, + "loss": 0.1956, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0846383273601532, + "step": 7915, + "valid_targets_mean": 4108.1, + "valid_targets_min": 2252 + }, + { + "epoch": 5.760276464168789, + "grad_norm": 0.7800345658186676, + "learning_rate": 3.7078434716010293e-06, + "loss": 0.1952, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10326851904392242, + "step": 7920, + "valid_targets_mean": 3446.6, + "valid_targets_min": 1436 + }, + { + "epoch": 5.763914150600218, + "grad_norm": 0.6693030964974499, + "learning_rate": 3.6868339996684244e-06, + "loss": 0.1918, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0884346216917038, + "step": 7925, + "valid_targets_mean": 3819.2, + "valid_targets_min": 2333 + }, + { + "epoch": 5.767551837031648, + "grad_norm": 0.6450391095029231, + "learning_rate": 3.665878174252957e-06, + "loss": 0.1865, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08642492443323135, + "step": 7930, + "valid_targets_mean": 3889.8, + "valid_targets_min": 1272 + }, + { + "epoch": 5.771189523463077, + "grad_norm": 0.6535314412107917, + "learning_rate": 3.6449760642687303e-06, + "loss": 0.1867, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09708479791879654, + "step": 7935, + "valid_targets_mean": 3233.0, + "valid_targets_min": 1319 + }, + { + "epoch": 5.774827209894507, + "grad_norm": 0.6340063608749432, + "learning_rate": 3.624127738453185e-06, + "loss": 0.1739, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07828659564256668, + "step": 7940, + "valid_targets_mean": 3186.6, + "valid_targets_min": 1194 + }, + { + "epoch": 5.778464896325937, + "grad_norm": 0.6872976082466106, + "learning_rate": 3.6033332653668908e-06, + "loss": 0.2056, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08664410561323166, + "step": 7945, + "valid_targets_mean": 2710.6, + "valid_targets_min": 1453 + }, + { + "epoch": 5.782102582757366, + "grad_norm": 0.6121257193263314, + "learning_rate": 3.5825927133933358e-06, + "loss": 0.1896, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08067971467971802, + "step": 7950, + "valid_targets_mean": 5344.5, + "valid_targets_min": 3603 + }, + { + "epoch": 5.785740269188796, + "grad_norm": 0.6592802875183024, + "learning_rate": 3.5619061507386765e-06, + "loss": 0.2015, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08765257894992828, + "step": 7955, + "valid_targets_mean": 3067.5, + "valid_targets_min": 1256 + }, + { + "epoch": 5.789377955620226, + "grad_norm": 0.72348000142629, + "learning_rate": 3.541273645431529e-06, + "loss": 0.1945, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12835977971553802, + "step": 7960, + "valid_targets_mean": 4443.9, + "valid_targets_min": 1336 + }, + { + "epoch": 5.793015642051655, + "grad_norm": 0.6760108669013773, + "learning_rate": 3.520695265322727e-06, + "loss": 0.202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09975667297840118, + "step": 7965, + "valid_targets_mean": 3581.8, + "valid_targets_min": 1275 + }, + { + "epoch": 5.796653328483085, + "grad_norm": 0.9351696036828149, + "learning_rate": 3.5001710780851284e-06, + "loss": 0.1931, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10712091624736786, + "step": 7970, + "valid_targets_mean": 3770.1, + "valid_targets_min": 2202 + }, + { + "epoch": 5.800291014914515, + "grad_norm": 0.6440772111126771, + "learning_rate": 3.4797011512133595e-06, + "loss": 0.1865, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11148664355278015, + "step": 7975, + "valid_targets_mean": 4112.8, + "valid_targets_min": 3332 + }, + { + "epoch": 5.803928701345944, + "grad_norm": 0.6684180569478338, + "learning_rate": 3.4592855520236167e-06, + "loss": 0.1842, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09458138793706894, + "step": 7980, + "valid_targets_mean": 3260.0, + "valid_targets_min": 2195 + }, + { + "epoch": 5.807566387777373, + "grad_norm": 0.6509093365763523, + "learning_rate": 3.438924347653445e-06, + "loss": 0.1839, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09279956668615341, + "step": 7985, + "valid_targets_mean": 4039.6, + "valid_targets_min": 2343 + }, + { + "epoch": 5.811204074208804, + "grad_norm": 0.6381516710606598, + "learning_rate": 3.4186176050614937e-06, + "loss": 0.1777, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08008937537670135, + "step": 7990, + "valid_targets_mean": 2678.8, + "valid_targets_min": 1026 + }, + { + "epoch": 5.814841760640233, + "grad_norm": 0.7002266921157347, + "learning_rate": 3.398365391027323e-06, + "loss": 0.1824, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09111490845680237, + "step": 7995, + "valid_targets_mean": 3754.5, + "valid_targets_min": 1042 + }, + { + "epoch": 5.818479447071662, + "grad_norm": 0.7029221667021586, + "learning_rate": 3.3781677721511773e-06, + "loss": 0.1692, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08540008962154388, + "step": 8000, + "valid_targets_mean": 3301.6, + "valid_targets_min": 1274 + }, + { + "epoch": 5.822117133503092, + "grad_norm": 0.7036861150046985, + "learning_rate": 3.3580248148537485e-06, + "loss": 0.1862, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09279727190732956, + "step": 8005, + "valid_targets_mean": 3632.2, + "valid_targets_min": 914 + }, + { + "epoch": 5.825754819934522, + "grad_norm": 0.622046547195297, + "learning_rate": 3.3379365853759872e-06, + "loss": 0.209, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08438567072153091, + "step": 8010, + "valid_targets_mean": 4342.5, + "valid_targets_min": 2093 + }, + { + "epoch": 5.829392506365951, + "grad_norm": 0.6946547275540209, + "learning_rate": 3.317903149778854e-06, + "loss": 0.1892, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09815973788499832, + "step": 8015, + "valid_targets_mean": 3609.6, + "valid_targets_min": 1683 + }, + { + "epoch": 5.833030192797381, + "grad_norm": 0.6285344940793238, + "learning_rate": 3.2979245739431275e-06, + "loss": 0.1979, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0921814814209938, + "step": 8020, + "valid_targets_mean": 4096.2, + "valid_targets_min": 2253 + }, + { + "epoch": 5.836667879228811, + "grad_norm": 0.6447372621592952, + "learning_rate": 3.27800092356918e-06, + "loss": 0.1917, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09213989973068237, + "step": 8025, + "valid_targets_mean": 3794.4, + "valid_targets_min": 1789 + }, + { + "epoch": 5.84030556566024, + "grad_norm": 0.7190024434872343, + "learning_rate": 3.2581322641767588e-06, + "loss": 0.1927, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09685018658638, + "step": 8030, + "valid_targets_mean": 3753.8, + "valid_targets_min": 1586 + }, + { + "epoch": 5.84394325209167, + "grad_norm": 0.7157321498856416, + "learning_rate": 3.23831866110476e-06, + "loss": 0.1902, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08564841747283936, + "step": 8035, + "valid_targets_mean": 2995.5, + "valid_targets_min": 2033 + }, + { + "epoch": 5.847580938523099, + "grad_norm": 0.6104453770692604, + "learning_rate": 3.21856017951103e-06, + "loss": 0.1807, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08898060023784637, + "step": 8040, + "valid_targets_mean": 3690.5, + "valid_targets_min": 1616 + }, + { + "epoch": 5.851218624954529, + "grad_norm": 0.6338444617504116, + "learning_rate": 3.19885688437215e-06, + "loss": 0.1917, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08519577980041504, + "step": 8045, + "valid_targets_mean": 3303.9, + "valid_targets_min": 2000 + }, + { + "epoch": 5.854856311385959, + "grad_norm": 0.6549674512955367, + "learning_rate": 3.179208840483219e-06, + "loss": 0.1842, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10119380056858063, + "step": 8050, + "valid_targets_mean": 3788.9, + "valid_targets_min": 1750 + }, + { + "epoch": 5.858493997817388, + "grad_norm": 0.7110076182121746, + "learning_rate": 3.15961611245764e-06, + "loss": 0.1895, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1099613681435585, + "step": 8055, + "valid_targets_mean": 3898.2, + "valid_targets_min": 2126 + }, + { + "epoch": 5.862131684248817, + "grad_norm": 0.6941932120536121, + "learning_rate": 3.1400787647268947e-06, + "loss": 0.2015, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10639913380146027, + "step": 8060, + "valid_targets_mean": 3923.1, + "valid_targets_min": 3049 + }, + { + "epoch": 5.865769370680248, + "grad_norm": 0.6790489393499264, + "learning_rate": 3.1205968615403615e-06, + "loss": 0.1911, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11137130111455917, + "step": 8065, + "valid_targets_mean": 4337.1, + "valid_targets_min": 1571 + }, + { + "epoch": 5.869407057111677, + "grad_norm": 0.7185676915970799, + "learning_rate": 3.1011704669650842e-06, + "loss": 0.2014, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09340154379606247, + "step": 8070, + "valid_targets_mean": 2998.0, + "valid_targets_min": 1487 + }, + { + "epoch": 5.873044743543106, + "grad_norm": 0.6231777259621331, + "learning_rate": 3.0817996448855482e-06, + "loss": 0.1932, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07955539971590042, + "step": 8075, + "valid_targets_mean": 3627.1, + "valid_targets_min": 1970 + }, + { + "epoch": 5.876682429974537, + "grad_norm": 0.643730039625055, + "learning_rate": 3.0624844590035096e-06, + "loss": 0.2069, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09050670266151428, + "step": 8080, + "valid_targets_mean": 4038.0, + "valid_targets_min": 2492 + }, + { + "epoch": 5.880320116405966, + "grad_norm": 0.6905204605045329, + "learning_rate": 3.043224972837744e-06, + "loss": 0.2018, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08497107028961182, + "step": 8085, + "valid_targets_mean": 3211.2, + "valid_targets_min": 1066 + }, + { + "epoch": 5.883957802837395, + "grad_norm": 0.6464025848297961, + "learning_rate": 3.024021249723865e-06, + "loss": 0.2002, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0867832601070404, + "step": 8090, + "valid_targets_mean": 3401.9, + "valid_targets_min": 2281 + }, + { + "epoch": 5.887595489268825, + "grad_norm": 0.6862153016626994, + "learning_rate": 3.0048733528141106e-06, + "loss": 0.2007, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09990978986024857, + "step": 8095, + "valid_targets_mean": 3662.2, + "valid_targets_min": 1969 + }, + { + "epoch": 5.891233175700255, + "grad_norm": 0.7081767795930606, + "learning_rate": 2.9857813450771213e-06, + "loss": 0.1934, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09026861190795898, + "step": 8100, + "valid_targets_mean": 3189.6, + "valid_targets_min": 1005 + }, + { + "epoch": 5.894870862131684, + "grad_norm": 0.7111852297794045, + "learning_rate": 2.9667452892977543e-06, + "loss": 0.1943, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08734816312789917, + "step": 8105, + "valid_targets_mean": 3546.5, + "valid_targets_min": 1980 + }, + { + "epoch": 5.898508548563114, + "grad_norm": 0.7272946311080127, + "learning_rate": 2.9477652480768506e-06, + "loss": 0.1987, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11377128958702087, + "step": 8110, + "valid_targets_mean": 4803.1, + "valid_targets_min": 1257 + }, + { + "epoch": 5.902146234994543, + "grad_norm": 0.6820293282548535, + "learning_rate": 2.928841283831072e-06, + "loss": 0.1867, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0970475971698761, + "step": 8115, + "valid_targets_mean": 3118.1, + "valid_targets_min": 1357 + }, + { + "epoch": 5.905783921425973, + "grad_norm": 0.6103748840218605, + "learning_rate": 2.9099734587926477e-06, + "loss": 0.1797, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08083633333444595, + "step": 8120, + "valid_targets_mean": 4463.8, + "valid_targets_min": 1410 + }, + { + "epoch": 5.909421607857403, + "grad_norm": 0.6965056981110872, + "learning_rate": 2.8911618350091887e-06, + "loss": 0.1962, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08663342893123627, + "step": 8125, + "valid_targets_mean": 3429.8, + "valid_targets_min": 1390 + }, + { + "epoch": 5.913059294288832, + "grad_norm": 0.6976968284870314, + "learning_rate": 2.8724064743434988e-06, + "loss": 0.1855, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08465567976236343, + "step": 8130, + "valid_targets_mean": 3173.0, + "valid_targets_min": 790 + }, + { + "epoch": 5.916696980720262, + "grad_norm": 0.7020416461503929, + "learning_rate": 2.853707438473352e-06, + "loss": 0.2042, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10430674999952316, + "step": 8135, + "valid_targets_mean": 3788.8, + "valid_targets_min": 1901 + }, + { + "epoch": 5.920334667151692, + "grad_norm": 0.7109780382400118, + "learning_rate": 2.8350647888913018e-06, + "loss": 0.1913, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08578915148973465, + "step": 8140, + "valid_targets_mean": 2697.1, + "valid_targets_min": 1460 + }, + { + "epoch": 5.923972353583121, + "grad_norm": 0.6522265888349529, + "learning_rate": 2.8164785869044585e-06, + "loss": 0.1943, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11750302463769913, + "step": 8145, + "valid_targets_mean": 5088.8, + "valid_targets_min": 952 + }, + { + "epoch": 5.92761004001455, + "grad_norm": 0.7807103589253218, + "learning_rate": 2.7979488936343215e-06, + "loss": 0.1933, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10032657533884048, + "step": 8150, + "valid_targets_mean": 3070.5, + "valid_targets_min": 1768 + }, + { + "epoch": 5.931247726445981, + "grad_norm": 0.690366073509498, + "learning_rate": 2.77947577001654e-06, + "loss": 0.1964, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08066092431545258, + "step": 8155, + "valid_targets_mean": 3585.0, + "valid_targets_min": 1931 + }, + { + "epoch": 5.93488541287741, + "grad_norm": 0.6459461752020543, + "learning_rate": 2.7610592768007437e-06, + "loss": 0.1909, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09195667505264282, + "step": 8160, + "valid_targets_mean": 5915.4, + "valid_targets_min": 1233 + }, + { + "epoch": 5.938523099308839, + "grad_norm": 0.6684510936157754, + "learning_rate": 2.742699474550332e-06, + "loss": 0.1854, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09438403695821762, + "step": 8165, + "valid_targets_mean": 5310.6, + "valid_targets_min": 1741 + }, + { + "epoch": 5.942160785740269, + "grad_norm": 0.704745692678583, + "learning_rate": 2.7243964236422614e-06, + "loss": 0.1876, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08752848953008652, + "step": 8170, + "valid_targets_mean": 3884.8, + "valid_targets_min": 856 + }, + { + "epoch": 5.945798472171699, + "grad_norm": 0.6654020203985466, + "learning_rate": 2.7061501842668782e-06, + "loss": 0.1885, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09932336956262589, + "step": 8175, + "valid_targets_mean": 3649.5, + "valid_targets_min": 1062 + }, + { + "epoch": 5.949436158603128, + "grad_norm": 0.6133815265030181, + "learning_rate": 2.6879608164276793e-06, + "loss": 0.1821, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09161056578159332, + "step": 8180, + "valid_targets_mean": 4412.6, + "valid_targets_min": 974 + }, + { + "epoch": 5.953073845034558, + "grad_norm": 0.7418591553467809, + "learning_rate": 2.6698283799411527e-06, + "loss": 0.1827, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09696438908576965, + "step": 8185, + "valid_targets_mean": 3077.5, + "valid_targets_min": 1447 + }, + { + "epoch": 5.956711531465988, + "grad_norm": 0.6671002943056219, + "learning_rate": 2.6517529344365687e-06, + "loss": 0.1804, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10416644811630249, + "step": 8190, + "valid_targets_mean": 4663.0, + "valid_targets_min": 2144 + }, + { + "epoch": 5.960349217897417, + "grad_norm": 0.6660010166834326, + "learning_rate": 2.6337345393557614e-06, + "loss": 0.1807, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10428169369697571, + "step": 8195, + "valid_targets_mean": 4543.1, + "valid_targets_min": 1515 + }, + { + "epoch": 5.963986904328847, + "grad_norm": 0.7464482128473694, + "learning_rate": 2.615773253952969e-06, + "loss": 0.1983, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09561319649219513, + "step": 8200, + "valid_targets_mean": 3034.9, + "valid_targets_min": 1349 + }, + { + "epoch": 5.967624590760277, + "grad_norm": 0.6984313991020112, + "learning_rate": 2.5978691372946174e-06, + "loss": 0.195, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09490586817264557, + "step": 8205, + "valid_targets_mean": 3150.4, + "valid_targets_min": 1906 + }, + { + "epoch": 5.971262277191706, + "grad_norm": 0.7003712763879207, + "learning_rate": 2.5800222482591375e-06, + "loss": 0.1831, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09672945737838745, + "step": 8210, + "valid_targets_mean": 3655.1, + "valid_targets_min": 2076 + }, + { + "epoch": 5.974899963623136, + "grad_norm": 0.6199355093974543, + "learning_rate": 2.5622326455367486e-06, + "loss": 0.1797, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07759780436754227, + "step": 8215, + "valid_targets_mean": 2994.0, + "valid_targets_min": 1021 + }, + { + "epoch": 5.978537650054565, + "grad_norm": 0.7201319297824156, + "learning_rate": 2.5445003876292872e-06, + "loss": 0.191, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09335920214653015, + "step": 8220, + "valid_targets_mean": 3226.9, + "valid_targets_min": 654 + }, + { + "epoch": 5.982175336485994, + "grad_norm": 0.6460765031325312, + "learning_rate": 2.526825532850019e-06, + "loss": 0.1866, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1075495257973671, + "step": 8225, + "valid_targets_mean": 4790.8, + "valid_targets_min": 3074 + }, + { + "epoch": 5.985813022917425, + "grad_norm": 0.7877334906943314, + "learning_rate": 2.5092081393234315e-06, + "loss": 0.1977, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09398342669010162, + "step": 8230, + "valid_targets_mean": 2604.0, + "valid_targets_min": 815 + }, + { + "epoch": 5.989450709348854, + "grad_norm": 0.6497680641977175, + "learning_rate": 2.4916482649850495e-06, + "loss": 0.1903, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08751532435417175, + "step": 8235, + "valid_targets_mean": 3560.0, + "valid_targets_min": 2205 + }, + { + "epoch": 5.993088395780283, + "grad_norm": 0.6558967079985312, + "learning_rate": 2.4741459675812384e-06, + "loss": 0.1968, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10613460838794708, + "step": 8240, + "valid_targets_mean": 4216.0, + "valid_targets_min": 1793 + }, + { + "epoch": 5.996726082211714, + "grad_norm": 0.7091318142663681, + "learning_rate": 2.4567013046690156e-06, + "loss": 0.1951, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09108114242553711, + "step": 8245, + "valid_targets_mean": 3134.5, + "valid_targets_min": 1217 + }, + { + "epoch": 6.000727537286286, + "grad_norm": 1.1744063194030963, + "learning_rate": 2.439314333615883e-06, + "loss": 0.1916, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05465805530548096, + "step": 8250, + "valid_targets_mean": 1525.2, + "valid_targets_min": 534 + }, + { + "epoch": 6.004365223717715, + "grad_norm": 0.44734030700425187, + "learning_rate": 2.421985111599605e-06, + "loss": 0.0952, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03248581662774086, + "step": 8255, + "valid_targets_mean": 2735.4, + "valid_targets_min": 834 + }, + { + "epoch": 6.008002910149145, + "grad_norm": 0.4171943362696413, + "learning_rate": 2.4047136956080454e-06, + "loss": 0.0787, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03711620718240738, + "step": 8260, + "valid_targets_mean": 3253.9, + "valid_targets_min": 2790 + }, + { + "epoch": 6.011640596580575, + "grad_norm": 0.4032956810324009, + "learning_rate": 2.3875001424389586e-06, + "loss": 0.0725, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03649858385324478, + "step": 8265, + "valid_targets_mean": 3383.1, + "valid_targets_min": 2133 + }, + { + "epoch": 6.015278283012004, + "grad_norm": 0.4327203499455479, + "learning_rate": 2.3703445086998267e-06, + "loss": 0.0921, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.040198374539613724, + "step": 8270, + "valid_targets_mean": 2954.8, + "valid_targets_min": 2338 + }, + { + "epoch": 6.018915969443434, + "grad_norm": 0.43374785706003616, + "learning_rate": 2.35324685080766e-06, + "loss": 0.0713, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03631129860877991, + "step": 8275, + "valid_targets_mean": 2680.4, + "valid_targets_min": 662 + }, + { + "epoch": 6.022553655874864, + "grad_norm": 1.2230723586031433, + "learning_rate": 2.3362072249888e-06, + "loss": 0.0991, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0746355727314949, + "step": 8280, + "valid_targets_mean": 1101.0, + "valid_targets_min": 559 + }, + { + "epoch": 6.026191342306293, + "grad_norm": 0.4137786857050217, + "learning_rate": 2.3192256872787677e-06, + "loss": 0.1093, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03225291520357132, + "step": 8285, + "valid_targets_mean": 3490.6, + "valid_targets_min": 1091 + }, + { + "epoch": 6.029829028737723, + "grad_norm": 0.43456560300376956, + "learning_rate": 2.3023022935220363e-06, + "loss": 0.0822, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03752252832055092, + "step": 8290, + "valid_targets_mean": 3061.8, + "valid_targets_min": 947 + }, + { + "epoch": 6.033466715169152, + "grad_norm": 0.5595465283507853, + "learning_rate": 2.28543709937189e-06, + "loss": 0.1059, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09113454818725586, + "step": 8295, + "valid_targets_mean": 2125.4, + "valid_targets_min": 859 + }, + { + "epoch": 6.037104401600582, + "grad_norm": 0.3271702796784013, + "learning_rate": 2.268630160290213e-06, + "loss": 0.0744, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03985653817653656, + "step": 8300, + "valid_targets_mean": 5837.8, + "valid_targets_min": 636 + }, + { + "epoch": 6.040742088032012, + "grad_norm": 0.45662610843364504, + "learning_rate": 2.2518815315473108e-06, + "loss": 0.0746, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.041903093457221985, + "step": 8305, + "valid_targets_mean": 2185.0, + "valid_targets_min": 826 + }, + { + "epoch": 6.044379774463441, + "grad_norm": 0.34255531291609803, + "learning_rate": 2.235191268221746e-06, + "loss": 0.0735, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03324306011199951, + "step": 8310, + "valid_targets_mean": 3446.5, + "valid_targets_min": 1096 + }, + { + "epoch": 6.0480174608948705, + "grad_norm": 0.4982728523190991, + "learning_rate": 2.2185594252001263e-06, + "loss": 0.0906, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03918091207742691, + "step": 8315, + "valid_targets_mean": 2343.1, + "valid_targets_min": 688 + }, + { + "epoch": 6.051655147326301, + "grad_norm": 0.4191910164677894, + "learning_rate": 2.2019860571769545e-06, + "loss": 0.0927, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02716805413365364, + "step": 8320, + "valid_targets_mean": 2938.5, + "valid_targets_min": 553 + }, + { + "epoch": 6.05529283375773, + "grad_norm": 0.4853632673626972, + "learning_rate": 2.1854712186544334e-06, + "loss": 0.074, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.039732713252305984, + "step": 8325, + "valid_targets_mean": 2988.9, + "valid_targets_min": 566 + }, + { + "epoch": 6.0589305201891595, + "grad_norm": 0.3428520514706945, + "learning_rate": 2.1690149639422954e-06, + "loss": 0.0685, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.027942605316638947, + "step": 8330, + "valid_targets_mean": 3230.2, + "valid_targets_min": 1111 + }, + { + "epoch": 6.06256820662059, + "grad_norm": 0.4591129105948867, + "learning_rate": 2.1526173471576018e-06, + "loss": 0.073, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05215325206518173, + "step": 8335, + "valid_targets_mean": 1232.1, + "valid_targets_min": 622 + }, + { + "epoch": 6.066205893052019, + "grad_norm": 0.5845468423849308, + "learning_rate": 2.1362784222245845e-06, + "loss": 0.0848, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03981347009539604, + "step": 8340, + "valid_targets_mean": 2132.5, + "valid_targets_min": 696 + }, + { + "epoch": 6.069843579483448, + "grad_norm": 0.49853747536383713, + "learning_rate": 2.119998242874488e-06, + "loss": 0.0822, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.041917718946933746, + "step": 8345, + "valid_targets_mean": 3225.2, + "valid_targets_min": 1922 + }, + { + "epoch": 6.073481265914878, + "grad_norm": 0.500682055398399, + "learning_rate": 2.1037768626453347e-06, + "loss": 0.0826, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03136047348380089, + "step": 8350, + "valid_targets_mean": 2008.5, + "valid_targets_min": 802 + }, + { + "epoch": 6.077118952346308, + "grad_norm": 0.5681289754139265, + "learning_rate": 2.087614334881811e-06, + "loss": 0.092, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06482289731502533, + "step": 8355, + "valid_targets_mean": 1754.1, + "valid_targets_min": 957 + }, + { + "epoch": 6.080756638777737, + "grad_norm": 0.5058819972416035, + "learning_rate": 2.0715107127350453e-06, + "loss": 0.1083, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04073405638337135, + "step": 8360, + "valid_targets_mean": 2498.5, + "valid_targets_min": 641 + }, + { + "epoch": 6.084394325209167, + "grad_norm": 0.4610279161789922, + "learning_rate": 2.055466049162467e-06, + "loss": 0.0702, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04540270194411278, + "step": 8365, + "valid_targets_mean": 2364.6, + "valid_targets_min": 924 + }, + { + "epoch": 6.088032011640596, + "grad_norm": 0.5655294308767235, + "learning_rate": 2.0394803969276156e-06, + "loss": 0.1064, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07651930302381516, + "step": 8370, + "valid_targets_mean": 2385.5, + "valid_targets_min": 700 + }, + { + "epoch": 6.091669698072026, + "grad_norm": 0.5010966541748249, + "learning_rate": 2.023553808599954e-06, + "loss": 0.0922, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07019326090812683, + "step": 8375, + "valid_targets_mean": 3363.2, + "valid_targets_min": 2731 + }, + { + "epoch": 6.095307384503456, + "grad_norm": 0.5463292274849768, + "learning_rate": 2.007686336554735e-06, + "loss": 0.0791, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06699395179748535, + "step": 8380, + "valid_targets_mean": 2615.0, + "valid_targets_min": 893 + }, + { + "epoch": 6.098945070934885, + "grad_norm": 0.8512828515449898, + "learning_rate": 1.991878032972785e-06, + "loss": 0.2422, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26900383830070496, + "step": 8385, + "valid_targets_mean": 2615.9, + "valid_targets_min": 922 + }, + { + "epoch": 6.102582757366315, + "grad_norm": 0.47390210165474755, + "learning_rate": 1.976128949840361e-06, + "loss": 0.112, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02851167879998684, + "step": 8390, + "valid_targets_mean": 1357.4, + "valid_targets_min": 507 + }, + { + "epoch": 6.106220443797745, + "grad_norm": 0.44197439042470044, + "learning_rate": 1.9604391389489753e-06, + "loss": 0.0902, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04483911395072937, + "step": 8395, + "valid_targets_mean": 3236.9, + "valid_targets_min": 615 + }, + { + "epoch": 6.109858130229174, + "grad_norm": 0.5316409660774585, + "learning_rate": 1.9448086518952115e-06, + "loss": 0.1715, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.061240632086992264, + "step": 8400, + "valid_targets_mean": 3023.1, + "valid_targets_min": 826 + }, + { + "epoch": 6.1134958166606035, + "grad_norm": 0.47382055591361816, + "learning_rate": 1.929237540080573e-06, + "loss": 0.0763, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03914313018321991, + "step": 8405, + "valid_targets_mean": 2924.4, + "valid_targets_min": 851 + }, + { + "epoch": 6.117133503092034, + "grad_norm": 0.42652502702772677, + "learning_rate": 1.9137258547112904e-06, + "loss": 0.0776, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.049810171127319336, + "step": 8410, + "valid_targets_mean": 4528.4, + "valid_targets_min": 2665 + }, + { + "epoch": 6.120771189523463, + "grad_norm": 0.38300341533129556, + "learning_rate": 1.8982736467981943e-06, + "loss": 0.0789, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.026908889412879944, + "step": 8415, + "valid_targets_mean": 2820.6, + "valid_targets_min": 595 + }, + { + "epoch": 6.1244088759548925, + "grad_norm": 0.48253703525354563, + "learning_rate": 1.8828809671564996e-06, + "loss": 0.073, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04525480046868324, + "step": 8420, + "valid_targets_mean": 2583.1, + "valid_targets_min": 825 + }, + { + "epoch": 6.128046562386323, + "grad_norm": 0.45106256662260047, + "learning_rate": 1.8675478664056635e-06, + "loss": 0.0731, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.033943358808755875, + "step": 8425, + "valid_targets_mean": 3771.5, + "valid_targets_min": 672 + }, + { + "epoch": 6.131684248817752, + "grad_norm": 0.35719347127036916, + "learning_rate": 1.852274394969218e-06, + "loss": 0.067, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.027508588507771492, + "step": 8430, + "valid_targets_mean": 3165.0, + "valid_targets_min": 1956 + }, + { + "epoch": 6.1353219352491815, + "grad_norm": 0.39110376491429577, + "learning_rate": 1.837060603074603e-06, + "loss": 0.074, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03375666216015816, + "step": 8435, + "valid_targets_mean": 3719.8, + "valid_targets_min": 2540 + }, + { + "epoch": 6.138959621680611, + "grad_norm": 0.44902619395337096, + "learning_rate": 1.8219065407530045e-06, + "loss": 0.0654, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.046320926398038864, + "step": 8440, + "valid_targets_mean": 3447.5, + "valid_targets_min": 1240 + }, + { + "epoch": 6.142597308112041, + "grad_norm": 0.31743317891668027, + "learning_rate": 1.80681225783917e-06, + "loss": 0.1065, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02056090161204338, + "step": 8445, + "valid_targets_mean": 2975.2, + "valid_targets_min": 623 + }, + { + "epoch": 6.14623499454347, + "grad_norm": 0.3863955040375285, + "learning_rate": 1.7917778039712798e-06, + "loss": 0.0614, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.031008237972855568, + "step": 8450, + "valid_targets_mean": 2960.1, + "valid_targets_min": 583 + }, + { + "epoch": 6.1498726809749, + "grad_norm": 0.39162124269131526, + "learning_rate": 1.776803228590742e-06, + "loss": 0.0664, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03139938414096832, + "step": 8455, + "valid_targets_mean": 3723.5, + "valid_targets_min": 3230 + }, + { + "epoch": 6.153510367406329, + "grad_norm": 0.40645630064506777, + "learning_rate": 1.7618885809420706e-06, + "loss": 0.0639, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.025481581687927246, + "step": 8460, + "valid_targets_mean": 2520.9, + "valid_targets_min": 647 + }, + { + "epoch": 6.157148053837759, + "grad_norm": 0.7803717741586691, + "learning_rate": 1.7470339100726997e-06, + "loss": 0.1094, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08562277257442474, + "step": 8465, + "valid_targets_mean": 1094.5, + "valid_targets_min": 195 + }, + { + "epoch": 6.160785740269189, + "grad_norm": 0.38555218960275867, + "learning_rate": 1.7322392648328179e-06, + "loss": 0.0728, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05408021807670593, + "step": 8470, + "valid_targets_mean": 3635.9, + "valid_targets_min": 3050 + }, + { + "epoch": 6.164423426700618, + "grad_norm": 0.8629197563715347, + "learning_rate": 1.7175046938752316e-06, + "loss": 0.1028, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06241467595100403, + "step": 8475, + "valid_targets_mean": 2382.9, + "valid_targets_min": 1087 + }, + { + "epoch": 6.168061113132048, + "grad_norm": 0.529109541946274, + "learning_rate": 1.7028302456551782e-06, + "loss": 0.0722, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.032474685460329056, + "step": 8480, + "valid_targets_mean": 3025.5, + "valid_targets_min": 847 + }, + { + "epoch": 6.171698799563478, + "grad_norm": 0.5832944519638508, + "learning_rate": 1.6882159684301846e-06, + "loss": 0.0965, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05389542132616043, + "step": 8485, + "valid_targets_mean": 2263.1, + "valid_targets_min": 916 + }, + { + "epoch": 6.175336485994907, + "grad_norm": 0.47528795018398684, + "learning_rate": 1.6736619102599073e-06, + "loss": 0.1015, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03645186126232147, + "step": 8490, + "valid_targets_mean": 2567.0, + "valid_targets_min": 1062 + }, + { + "epoch": 6.1789741724263365, + "grad_norm": 0.6889755905855955, + "learning_rate": 1.6591681190059582e-06, + "loss": 0.1001, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06236148253083229, + "step": 8495, + "valid_targets_mean": 1443.4, + "valid_targets_min": 771 + }, + { + "epoch": 6.182611858857767, + "grad_norm": 0.47451239711457016, + "learning_rate": 1.644734642331769e-06, + "loss": 0.0724, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03476029261946678, + "step": 8500, + "valid_targets_mean": 2753.2, + "valid_targets_min": 707 + }, + { + "epoch": 6.186249545289196, + "grad_norm": 0.39281076668903153, + "learning_rate": 1.6303615277024222e-06, + "loss": 0.0791, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.035612087696790695, + "step": 8505, + "valid_targets_mean": 3665.8, + "valid_targets_min": 2850 + }, + { + "epoch": 6.1898872317206255, + "grad_norm": 0.5030256469959341, + "learning_rate": 1.6160488223844972e-06, + "loss": 0.0784, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029855545610189438, + "step": 8510, + "valid_targets_mean": 2238.2, + "valid_targets_min": 749 + }, + { + "epoch": 6.193524918152056, + "grad_norm": 0.4008859340429753, + "learning_rate": 1.6017965734459129e-06, + "loss": 0.0665, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.035867076367139816, + "step": 8515, + "valid_targets_mean": 4488.2, + "valid_targets_min": 1052 + }, + { + "epoch": 6.197162604583485, + "grad_norm": 0.36770483057133274, + "learning_rate": 1.5876048277557688e-06, + "loss": 0.0629, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03395641967654228, + "step": 8520, + "valid_targets_mean": 4258.2, + "valid_targets_min": 2540 + }, + { + "epoch": 6.2008002910149145, + "grad_norm": 0.36397940859527594, + "learning_rate": 1.5734736319842125e-06, + "loss": 0.0735, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.027728509157896042, + "step": 8525, + "valid_targets_mean": 3782.5, + "valid_targets_min": 1859 + }, + { + "epoch": 6.204437977446344, + "grad_norm": 0.3725114408413663, + "learning_rate": 1.5594030326022602e-06, + "loss": 0.0868, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.027011023834347725, + "step": 8530, + "valid_targets_mean": 2879.5, + "valid_targets_min": 893 + }, + { + "epoch": 6.208075663877774, + "grad_norm": 0.5035313877209104, + "learning_rate": 1.5453930758816605e-06, + "loss": 0.0833, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03949464485049248, + "step": 8535, + "valid_targets_mean": 2072.6, + "valid_targets_min": 753 + }, + { + "epoch": 6.2117133503092035, + "grad_norm": 0.32678514518840024, + "learning_rate": 1.5314438078947258e-06, + "loss": 0.0774, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02284541353583336, + "step": 8540, + "valid_targets_mean": 4205.6, + "valid_targets_min": 3934 + }, + { + "epoch": 6.215351036740633, + "grad_norm": 0.5524499412975836, + "learning_rate": 1.5175552745141952e-06, + "loss": 0.0681, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.039183832705020905, + "step": 8545, + "valid_targets_mean": 3174.9, + "valid_targets_min": 1576 + }, + { + "epoch": 6.218988723172062, + "grad_norm": 0.539655293717086, + "learning_rate": 1.503727521413092e-06, + "loss": 0.0787, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04134761542081833, + "step": 8550, + "valid_targets_mean": 3241.8, + "valid_targets_min": 1006 + }, + { + "epoch": 6.2226264096034924, + "grad_norm": 0.3451414327051248, + "learning_rate": 1.4899605940645413e-06, + "loss": 0.0651, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.023175060749053955, + "step": 8555, + "valid_targets_mean": 3445.5, + "valid_targets_min": 980 + }, + { + "epoch": 6.226264096034922, + "grad_norm": 0.6069131075458013, + "learning_rate": 1.476254537741657e-06, + "loss": 0.0742, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04750014841556549, + "step": 8560, + "valid_targets_mean": 2083.9, + "valid_targets_min": 745 + }, + { + "epoch": 6.229901782466351, + "grad_norm": 0.5413340373404614, + "learning_rate": 1.4626093975173627e-06, + "loss": 0.0723, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.045820388942956924, + "step": 8565, + "valid_targets_mean": 2768.8, + "valid_targets_min": 1020 + }, + { + "epoch": 6.233539468897781, + "grad_norm": 0.6300476810600338, + "learning_rate": 1.449025218264266e-06, + "loss": 0.0742, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.049257226288318634, + "step": 8570, + "valid_targets_mean": 2314.8, + "valid_targets_min": 967 + }, + { + "epoch": 6.237177155329211, + "grad_norm": 0.44907274383302226, + "learning_rate": 1.435502044654502e-06, + "loss": 0.0897, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03163832798600197, + "step": 8575, + "valid_targets_mean": 2462.0, + "valid_targets_min": 1015 + }, + { + "epoch": 6.24081484176064, + "grad_norm": 0.640398386186174, + "learning_rate": 1.4220399211595748e-06, + "loss": 0.071, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04077361524105072, + "step": 8580, + "valid_targets_mean": 1214.9, + "valid_targets_min": 705 + }, + { + "epoch": 6.2444525281920695, + "grad_norm": 0.379929261522601, + "learning_rate": 1.4086388920502403e-06, + "loss": 0.0816, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02503812126815319, + "step": 8585, + "valid_targets_mean": 2364.8, + "valid_targets_min": 622 + }, + { + "epoch": 6.2480902146235, + "grad_norm": 0.4149120552174689, + "learning_rate": 1.3952990013963264e-06, + "loss": 0.0965, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03039976954460144, + "step": 8590, + "valid_targets_mean": 3702.2, + "valid_targets_min": 1322 + }, + { + "epoch": 6.251727901054929, + "grad_norm": 0.433385106559487, + "learning_rate": 1.3820202930666148e-06, + "loss": 0.0738, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.036639075726270676, + "step": 8595, + "valid_targets_mean": 2863.1, + "valid_targets_min": 1011 + }, + { + "epoch": 6.2553655874863585, + "grad_norm": 0.4939850176200701, + "learning_rate": 1.3688028107286888e-06, + "loss": 0.0642, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04004495218396187, + "step": 8600, + "valid_targets_mean": 3106.1, + "valid_targets_min": 874 + }, + { + "epoch": 6.259003273917788, + "grad_norm": 0.5189747215704622, + "learning_rate": 1.355646597848772e-06, + "loss": 0.0784, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.053952958434820175, + "step": 8605, + "valid_targets_mean": 3161.4, + "valid_targets_min": 1420 + }, + { + "epoch": 6.262640960349218, + "grad_norm": 0.4450470257075342, + "learning_rate": 1.342551697691623e-06, + "loss": 0.088, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.031179944053292274, + "step": 8610, + "valid_targets_mean": 2796.0, + "valid_targets_min": 1109 + }, + { + "epoch": 6.2662786467806475, + "grad_norm": 0.41658350659909704, + "learning_rate": 1.329518153320346e-06, + "loss": 0.0576, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02920685149729252, + "step": 8615, + "valid_targets_mean": 1822.9, + "valid_targets_min": 684 + }, + { + "epoch": 6.269916333212077, + "grad_norm": 0.48210258798128836, + "learning_rate": 1.3165460075963022e-06, + "loss": 0.0766, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03600240498781204, + "step": 8620, + "valid_targets_mean": 2389.8, + "valid_targets_min": 671 + }, + { + "epoch": 6.273554019643507, + "grad_norm": 0.3999859004305225, + "learning_rate": 1.303635303178914e-06, + "loss": 0.0673, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03169986978173256, + "step": 8625, + "valid_targets_mean": 4513.8, + "valid_targets_min": 1057 + }, + { + "epoch": 6.2771917060749365, + "grad_norm": 0.4087930775465746, + "learning_rate": 1.2907860825255747e-06, + "loss": 0.0591, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02867048606276512, + "step": 8630, + "valid_targets_mean": 3533.1, + "valid_targets_min": 1243 + }, + { + "epoch": 6.280829392506366, + "grad_norm": 0.3918142296224141, + "learning_rate": 1.2779983878914637e-06, + "loss": 0.0803, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03566394001245499, + "step": 8635, + "valid_targets_mean": 3911.2, + "valid_targets_min": 3322 + }, + { + "epoch": 6.284467078937795, + "grad_norm": 0.5388532929621077, + "learning_rate": 1.2652722613294465e-06, + "loss": 0.0678, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04117462784051895, + "step": 8640, + "valid_targets_mean": 1820.0, + "valid_targets_min": 680 + }, + { + "epoch": 6.2881047653692255, + "grad_norm": 0.4557402309532139, + "learning_rate": 1.2526077446899177e-06, + "loss": 0.0958, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03897159919142723, + "step": 8645, + "valid_targets_mean": 3371.9, + "valid_targets_min": 640 + }, + { + "epoch": 6.291742451800655, + "grad_norm": 0.308957317036644, + "learning_rate": 1.240004879620651e-06, + "loss": 0.0825, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.01961485855281353, + "step": 8650, + "valid_targets_mean": 3327.0, + "valid_targets_min": 929 + }, + { + "epoch": 6.295380138232084, + "grad_norm": 0.5868809442393074, + "learning_rate": 1.2274637075666962e-06, + "loss": 0.0864, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.052756547927856445, + "step": 8655, + "valid_targets_mean": 3318.0, + "valid_targets_min": 2380 + }, + { + "epoch": 6.299017824663514, + "grad_norm": 0.3960989181859142, + "learning_rate": 1.214984269770203e-06, + "loss": 0.0658, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.034377750009298325, + "step": 8660, + "valid_targets_mean": 3709.4, + "valid_targets_min": 2619 + }, + { + "epoch": 6.302655511094944, + "grad_norm": 0.6632789712716916, + "learning_rate": 1.2025666072703234e-06, + "loss": 0.075, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.044711560010910034, + "step": 8665, + "valid_targets_mean": 1574.6, + "valid_targets_min": 693 + }, + { + "epoch": 6.306293197526373, + "grad_norm": 0.412634534945064, + "learning_rate": 1.1902107609030522e-06, + "loss": 0.0711, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03933592885732651, + "step": 8670, + "valid_targets_mean": 3885.1, + "valid_targets_min": 3129 + }, + { + "epoch": 6.3099308839578025, + "grad_norm": 0.7022342167406869, + "learning_rate": 1.177916771301091e-06, + "loss": 0.0953, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04452413320541382, + "step": 8675, + "valid_targets_mean": 1611.0, + "valid_targets_min": 543 + }, + { + "epoch": 6.313568570389233, + "grad_norm": 0.45253399573350717, + "learning_rate": 1.1656846788937392e-06, + "loss": 0.0969, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.044975146651268005, + "step": 8680, + "valid_targets_mean": 2319.6, + "valid_targets_min": 503 + }, + { + "epoch": 6.317206256820662, + "grad_norm": 0.5426356269473899, + "learning_rate": 1.1535145239067224e-06, + "loss": 0.0761, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05086173117160797, + "step": 8685, + "valid_targets_mean": 1987.9, + "valid_targets_min": 625 + }, + { + "epoch": 6.3208439432520915, + "grad_norm": 0.44470054115523855, + "learning_rate": 1.1414063463621105e-06, + "loss": 0.0722, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03267371281981468, + "step": 8690, + "valid_targets_mean": 3822.8, + "valid_targets_min": 2822 + }, + { + "epoch": 6.324481629683521, + "grad_norm": 0.4324155344759206, + "learning_rate": 1.1293601860781366e-06, + "loss": 0.0703, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04304149001836777, + "step": 8695, + "valid_targets_mean": 3724.5, + "valid_targets_min": 2987 + }, + { + "epoch": 6.328119316114951, + "grad_norm": 0.4035580561020715, + "learning_rate": 1.1173760826690927e-06, + "loss": 0.062, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.025457989424467087, + "step": 8700, + "valid_targets_mean": 3022.0, + "valid_targets_min": 856 + }, + { + "epoch": 6.3317570025463805, + "grad_norm": 0.9864455351271603, + "learning_rate": 1.1054540755451936e-06, + "loss": 0.0716, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.039650216698646545, + "step": 8705, + "valid_targets_mean": 764.2, + "valid_targets_min": 563 + }, + { + "epoch": 6.33539468897781, + "grad_norm": 0.4126969456682851, + "learning_rate": 1.093594203912456e-06, + "loss": 0.0717, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029915517196059227, + "step": 8710, + "valid_targets_mean": 3117.1, + "valid_targets_min": 751 + }, + { + "epoch": 6.33903237540924, + "grad_norm": 0.5106625702727047, + "learning_rate": 1.0817965067725544e-06, + "loss": 0.0654, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04161466285586357, + "step": 8715, + "valid_targets_mean": 3566.1, + "valid_targets_min": 980 + }, + { + "epoch": 6.3426700618406695, + "grad_norm": 0.5024904259156272, + "learning_rate": 1.0700610229226992e-06, + "loss": 0.0655, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03574306145310402, + "step": 8720, + "valid_targets_mean": 3589.8, + "valid_targets_min": 2538 + }, + { + "epoch": 6.346307748272099, + "grad_norm": 0.43492721780198385, + "learning_rate": 1.0583877909555106e-06, + "loss": 0.075, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02578561007976532, + "step": 8725, + "valid_targets_mean": 4182.8, + "valid_targets_min": 485 + }, + { + "epoch": 6.349945434703528, + "grad_norm": 0.33710581185240773, + "learning_rate": 1.0467768492588904e-06, + "loss": 0.0635, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.031049154698848724, + "step": 8730, + "valid_targets_mean": 4832.8, + "valid_targets_min": 2551 + }, + { + "epoch": 6.3535831211349585, + "grad_norm": 0.38037735308996584, + "learning_rate": 1.0352282360159016e-06, + "loss": 0.0671, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.034391772001981735, + "step": 8735, + "valid_targets_mean": 4733.6, + "valid_targets_min": 3365 + }, + { + "epoch": 6.357220807566388, + "grad_norm": 0.49023648242735307, + "learning_rate": 1.0237419892046364e-06, + "loss": 0.0721, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05556822568178177, + "step": 8740, + "valid_targets_mean": 2954.4, + "valid_targets_min": 872 + }, + { + "epoch": 6.360858493997817, + "grad_norm": 0.40664562363759993, + "learning_rate": 1.0123181465980813e-06, + "loss": 0.1557, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03481421247124672, + "step": 8745, + "valid_targets_mean": 2590.9, + "valid_targets_min": 1015 + }, + { + "epoch": 6.364496180429247, + "grad_norm": 0.45490537195777986, + "learning_rate": 1.000956745764019e-06, + "loss": 0.0672, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05418315529823303, + "step": 8750, + "valid_targets_mean": 2342.4, + "valid_targets_min": 967 + }, + { + "epoch": 6.368133866860677, + "grad_norm": 0.3936661475660268, + "learning_rate": 9.89657824064878e-07, + "loss": 0.0668, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.027985725551843643, + "step": 8755, + "valid_targets_mean": 3335.1, + "valid_targets_min": 886 + }, + { + "epoch": 6.371771553292106, + "grad_norm": 0.5663798000185188, + "learning_rate": 9.784214186576313e-07, + "loss": 0.1042, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07290384918451309, + "step": 8760, + "valid_targets_mean": 6534.1, + "valid_targets_min": 3513 + }, + { + "epoch": 6.375409239723536, + "grad_norm": 0.8805934966469552, + "learning_rate": 9.67247566493661e-07, + "loss": 0.1847, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09770999848842621, + "step": 8765, + "valid_targets_mean": 8040.8, + "valid_targets_min": 5627 + }, + { + "epoch": 6.379046926154966, + "grad_norm": 0.947610462091583, + "learning_rate": 9.561363043186356e-07, + "loss": 0.1725, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04272324591875076, + "step": 8770, + "valid_targets_mean": 1206.4, + "valid_targets_min": 154 + }, + { + "epoch": 6.382684612586395, + "grad_norm": 0.6503191086134215, + "learning_rate": 9.450876686723975e-07, + "loss": 0.1719, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08603615313768387, + "step": 8775, + "valid_targets_mean": 8257.1, + "valid_targets_min": 4391 + }, + { + "epoch": 6.3863222990178246, + "grad_norm": 0.6695306141089974, + "learning_rate": 9.341016958888449e-07, + "loss": 0.1674, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07560458779335022, + "step": 8780, + "valid_targets_mean": 7369.2, + "valid_targets_min": 5609 + }, + { + "epoch": 6.389959985449254, + "grad_norm": 0.61790365660674, + "learning_rate": 9.23178422095794e-07, + "loss": 0.1687, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08042816817760468, + "step": 8785, + "valid_targets_mean": 8364.1, + "valid_targets_min": 4900 + }, + { + "epoch": 6.393597671880684, + "grad_norm": 0.5664432837944859, + "learning_rate": 9.12317883214886e-07, + "loss": 0.1791, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08908683806657791, + "step": 8790, + "valid_targets_mean": 6989.9, + "valid_targets_min": 5203 + }, + { + "epoch": 6.3972353583121135, + "grad_norm": 0.6042345553864555, + "learning_rate": 9.015201149614427e-07, + "loss": 0.1781, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09172607958316803, + "step": 8795, + "valid_targets_mean": 6957.9, + "valid_targets_min": 4939 + }, + { + "epoch": 6.400873044743543, + "grad_norm": 0.5800607222835366, + "learning_rate": 8.907851528443734e-07, + "loss": 0.1657, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08786244690418243, + "step": 8800, + "valid_targets_mean": 6731.4, + "valid_targets_min": 5603 + }, + { + "epoch": 6.404510731174973, + "grad_norm": 0.8249176604401638, + "learning_rate": 8.801130321660411e-07, + "loss": 0.1531, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05159953981637955, + "step": 8805, + "valid_targets_mean": 1784.4, + "valid_targets_min": 152 + }, + { + "epoch": 6.4081484176064025, + "grad_norm": 0.5248316714212427, + "learning_rate": 8.695037880221569e-07, + "loss": 0.1525, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08698199689388275, + "step": 8810, + "valid_targets_mean": 7188.8, + "valid_targets_min": 6165 + }, + { + "epoch": 6.411786104037832, + "grad_norm": 0.48809776782625375, + "learning_rate": 8.589574553016545e-07, + "loss": 0.1632, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07594908773899078, + "step": 8815, + "valid_targets_mean": 6071.9, + "valid_targets_min": 4131 + }, + { + "epoch": 6.415423790469261, + "grad_norm": 0.49782795770057936, + "learning_rate": 8.484740686865778e-07, + "loss": 0.1589, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07991297543048859, + "step": 8820, + "valid_targets_mean": 6849.2, + "valid_targets_min": 5210 + }, + { + "epoch": 6.4190614769006915, + "grad_norm": 0.4631219996517776, + "learning_rate": 8.380536626519809e-07, + "loss": 0.1631, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07452963292598724, + "step": 8825, + "valid_targets_mean": 7063.6, + "valid_targets_min": 5229 + }, + { + "epoch": 6.422699163332121, + "grad_norm": 0.4405464820451204, + "learning_rate": 8.276962714657922e-07, + "loss": 0.1524, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07358582317829132, + "step": 8830, + "valid_targets_mean": 6557.4, + "valid_targets_min": 4917 + }, + { + "epoch": 6.42633684976355, + "grad_norm": 0.4512886722529708, + "learning_rate": 8.174019291887258e-07, + "loss": 0.1603, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08150118589401245, + "step": 8835, + "valid_targets_mean": 7429.8, + "valid_targets_min": 5485 + }, + { + "epoch": 6.42997453619498, + "grad_norm": 0.422299668239269, + "learning_rate": 8.071706696741466e-07, + "loss": 0.1629, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07526423037052155, + "step": 8840, + "valid_targets_mean": 7177.1, + "valid_targets_min": 4936 + }, + { + "epoch": 6.43361222262641, + "grad_norm": 0.4310845590314784, + "learning_rate": 7.970025265679648e-07, + "loss": 0.1579, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07308314740657806, + "step": 8845, + "valid_targets_mean": 6854.6, + "valid_targets_min": 5399 + }, + { + "epoch": 6.437249909057839, + "grad_norm": 0.49642443383234497, + "learning_rate": 7.868975333085483e-07, + "loss": 0.1646, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06461305916309357, + "step": 8850, + "valid_targets_mean": 4393.0, + "valid_targets_min": 176 + }, + { + "epoch": 6.440887595489269, + "grad_norm": 0.4723809848097252, + "learning_rate": 7.76855723126575e-07, + "loss": 0.1785, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09491999447345734, + "step": 8855, + "valid_targets_mean": 7331.6, + "valid_targets_min": 5516 + }, + { + "epoch": 6.444525281920699, + "grad_norm": 0.407547894635402, + "learning_rate": 7.668771290449539e-07, + "loss": 0.1346, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06739763915538788, + "step": 8860, + "valid_targets_mean": 7282.9, + "valid_targets_min": 6023 + }, + { + "epoch": 6.448162968352128, + "grad_norm": 0.42227660640689757, + "learning_rate": 7.569617838787024e-07, + "loss": 0.1501, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07323046773672104, + "step": 8865, + "valid_targets_mean": 7764.1, + "valid_targets_min": 5015 + }, + { + "epoch": 6.451800654783558, + "grad_norm": 0.42924153772419693, + "learning_rate": 7.471097202348377e-07, + "loss": 0.1558, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06430743634700775, + "step": 8870, + "valid_targets_mean": 6513.1, + "valid_targets_min": 5498 + }, + { + "epoch": 6.455438341214987, + "grad_norm": 0.5019770452592123, + "learning_rate": 7.373209705122808e-07, + "loss": 0.1631, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09016607701778412, + "step": 8875, + "valid_targets_mean": 6089.0, + "valid_targets_min": 4364 + }, + { + "epoch": 6.459076027646417, + "grad_norm": 0.45600177526334507, + "learning_rate": 7.275955669017332e-07, + "loss": 0.1606, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07765991985797882, + "step": 8880, + "valid_targets_mean": 6736.5, + "valid_targets_min": 4448 + }, + { + "epoch": 6.462713714077847, + "grad_norm": 0.46969776488959125, + "learning_rate": 7.179335413855893e-07, + "loss": 0.1585, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09270019829273224, + "step": 8885, + "valid_targets_mean": 7164.1, + "valid_targets_min": 5361 + }, + { + "epoch": 6.466351400509276, + "grad_norm": 0.3633659114248918, + "learning_rate": 7.083349257378125e-07, + "loss": 0.1391, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06534329801797867, + "step": 8890, + "valid_targets_mean": 9142.8, + "valid_targets_min": 5332 + }, + { + "epoch": 6.469989086940705, + "grad_norm": 0.4823285379472773, + "learning_rate": 6.987997515238443e-07, + "loss": 0.1427, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07594664394855499, + "step": 8895, + "valid_targets_mean": 6898.2, + "valid_targets_min": 4543 + }, + { + "epoch": 6.4736267733721355, + "grad_norm": 0.5032376403951011, + "learning_rate": 6.89328050100504e-07, + "loss": 0.1375, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08504600077867508, + "step": 8900, + "valid_targets_mean": 7036.5, + "valid_targets_min": 4563 + }, + { + "epoch": 6.477264459803565, + "grad_norm": 0.9592858977656941, + "learning_rate": 6.799198526158668e-07, + "loss": 0.1554, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.01893843151628971, + "step": 8905, + "valid_targets_mean": 303.4, + "valid_targets_min": 137 + }, + { + "epoch": 6.480902146234994, + "grad_norm": 0.45994413100777826, + "learning_rate": 6.705751900091773e-07, + "loss": 0.1518, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07345838099718094, + "step": 8910, + "valid_targets_mean": 6170.5, + "valid_targets_min": 5103 + }, + { + "epoch": 6.4845398326664245, + "grad_norm": 0.4981841617933808, + "learning_rate": 6.612940930107381e-07, + "loss": 0.1646, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0805969387292862, + "step": 8915, + "valid_targets_mean": 6543.5, + "valid_targets_min": 4970 + }, + { + "epoch": 6.488177519097854, + "grad_norm": 0.4968066353019299, + "learning_rate": 6.520765921418259e-07, + "loss": 0.1665, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08517226576805115, + "step": 8920, + "valid_targets_mean": 6977.5, + "valid_targets_min": 5516 + }, + { + "epoch": 6.491815205529283, + "grad_norm": 0.44973008105973733, + "learning_rate": 6.429227177145669e-07, + "loss": 0.164, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07331424951553345, + "step": 8925, + "valid_targets_mean": 6267.0, + "valid_targets_min": 3446 + }, + { + "epoch": 6.495452891960713, + "grad_norm": 1.1150746961260314, + "learning_rate": 6.33832499831859e-07, + "loss": 0.1604, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08266744762659073, + "step": 8930, + "valid_targets_mean": 7138.0, + "valid_targets_min": 5457 + }, + { + "epoch": 6.499090578392143, + "grad_norm": 0.4293061915333132, + "learning_rate": 6.248059683872476e-07, + "loss": 0.1651, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08413954824209213, + "step": 8935, + "valid_targets_mean": 7686.0, + "valid_targets_min": 5241 + }, + { + "epoch": 6.502728264823572, + "grad_norm": 0.43302412744616287, + "learning_rate": 6.158431530648612e-07, + "loss": 0.1587, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07528438419103622, + "step": 8940, + "valid_targets_mean": 6100.4, + "valid_targets_min": 4826 + }, + { + "epoch": 6.506365951255002, + "grad_norm": 0.45926968567451015, + "learning_rate": 6.069440833392848e-07, + "loss": 0.1631, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0963195264339447, + "step": 8945, + "valid_targets_mean": 7517.6, + "valid_targets_min": 5757 + }, + { + "epoch": 6.510003637686431, + "grad_norm": 0.4503414477096914, + "learning_rate": 5.981087884754711e-07, + "loss": 0.1613, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0855419859290123, + "step": 8950, + "valid_targets_mean": 6934.8, + "valid_targets_min": 5789 + }, + { + "epoch": 6.513641324117861, + "grad_norm": 0.44575517103610435, + "learning_rate": 5.89337297528656e-07, + "loss": 0.1615, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07873673737049103, + "step": 8955, + "valid_targets_mean": 6668.5, + "valid_targets_min": 4984 + }, + { + "epoch": 6.517279010549291, + "grad_norm": 0.46453418295272575, + "learning_rate": 5.806296393442412e-07, + "loss": 0.1584, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08166180551052094, + "step": 8960, + "valid_targets_mean": 7219.6, + "valid_targets_min": 5532 + }, + { + "epoch": 6.52091669698072, + "grad_norm": 0.4340080333084046, + "learning_rate": 5.719858425577251e-07, + "loss": 0.1562, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06769324839115143, + "step": 8965, + "valid_targets_mean": 5596.8, + "valid_targets_min": 4003 + }, + { + "epoch": 6.52455438341215, + "grad_norm": 0.44525502425167673, + "learning_rate": 5.634059355945876e-07, + "loss": 0.1282, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08140580356121063, + "step": 8970, + "valid_targets_mean": 6938.2, + "valid_targets_min": 4381 + }, + { + "epoch": 6.52819206984358, + "grad_norm": 0.40210110987885667, + "learning_rate": 5.548899466702029e-07, + "loss": 0.1566, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07129770517349243, + "step": 8975, + "valid_targets_mean": 6382.5, + "valid_targets_min": 5479 + }, + { + "epoch": 6.531829756275009, + "grad_norm": 0.42776736298807305, + "learning_rate": 5.464379037897539e-07, + "loss": 0.1565, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07721640169620514, + "step": 8980, + "valid_targets_mean": 6278.9, + "valid_targets_min": 5413 + }, + { + "epoch": 6.535467442706439, + "grad_norm": 0.4198384857734038, + "learning_rate": 5.380498347481266e-07, + "loss": 0.1588, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07614041864871979, + "step": 8985, + "valid_targets_mean": 7032.0, + "valid_targets_min": 5270 + }, + { + "epoch": 6.539105129137869, + "grad_norm": 0.40236160348219974, + "learning_rate": 5.297257671298384e-07, + "loss": 0.1531, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07596487551927567, + "step": 8990, + "valid_targets_mean": 7515.1, + "valid_targets_min": 5210 + }, + { + "epoch": 6.542742815569298, + "grad_norm": 0.4324303333399569, + "learning_rate": 5.214657283089208e-07, + "loss": 0.1595, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09667275100946426, + "step": 8995, + "valid_targets_mean": 7108.4, + "valid_targets_min": 5235 + }, + { + "epoch": 6.546380502000727, + "grad_norm": 0.4366199462442665, + "learning_rate": 5.132697454488545e-07, + "loss": 0.1599, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07238341867923737, + "step": 9000, + "valid_targets_mean": 6046.8, + "valid_targets_min": 4547 + }, + { + "epoch": 6.550018188432157, + "grad_norm": 0.641070212278311, + "learning_rate": 5.051378455024635e-07, + "loss": 0.13, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.013506947085261345, + "step": 9005, + "valid_targets_mean": 321.0, + "valid_targets_min": 125 + }, + { + "epoch": 6.553655874863587, + "grad_norm": 0.41260015223885094, + "learning_rate": 4.970700552118368e-07, + "loss": 0.145, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07229682803153992, + "step": 9010, + "valid_targets_mean": 5921.0, + "valid_targets_min": 4812 + }, + { + "epoch": 6.557293561295016, + "grad_norm": 0.468757796502411, + "learning_rate": 4.89066401108238e-07, + "loss": 0.1522, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07735782861709595, + "step": 9015, + "valid_targets_mean": 6027.8, + "valid_targets_min": 5028 + }, + { + "epoch": 6.560931247726446, + "grad_norm": 0.42950766897619996, + "learning_rate": 4.811269095120086e-07, + "loss": 0.1614, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0800497904419899, + "step": 9020, + "valid_targets_mean": 6486.1, + "valid_targets_min": 5177 + }, + { + "epoch": 6.564568934157876, + "grad_norm": 0.4333749090785557, + "learning_rate": 4.732516065324921e-07, + "loss": 0.153, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07765417546033859, + "step": 9025, + "valid_targets_mean": 6566.8, + "valid_targets_min": 5575 + }, + { + "epoch": 6.568206620589305, + "grad_norm": 0.3974084640877036, + "learning_rate": 4.6544051806794776e-07, + "loss": 0.147, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06253786385059357, + "step": 9030, + "valid_targets_mean": 6367.0, + "valid_targets_min": 5080 + }, + { + "epoch": 6.571844307020735, + "grad_norm": 0.435342117042205, + "learning_rate": 4.5769366980546124e-07, + "loss": 0.1576, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07456903904676437, + "step": 9035, + "valid_targets_mean": 6229.6, + "valid_targets_min": 4779 + }, + { + "epoch": 6.575481993452165, + "grad_norm": 0.6232652043434258, + "learning_rate": 4.500110872208652e-07, + "loss": 0.1859, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1131189614534378, + "step": 9040, + "valid_targets_mean": 6415.2, + "valid_targets_min": 1971 + }, + { + "epoch": 6.579119679883594, + "grad_norm": 0.6809123163947903, + "learning_rate": 4.4239279557864114e-07, + "loss": 0.2101, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12304944545030594, + "step": 9045, + "valid_targets_mean": 5431.6, + "valid_targets_min": 1631 + }, + { + "epoch": 6.582757366315024, + "grad_norm": 0.6682500338015337, + "learning_rate": 4.3483881993185985e-07, + "loss": 0.2139, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12996098399162292, + "step": 9050, + "valid_targets_mean": 6106.4, + "valid_targets_min": 3528 + }, + { + "epoch": 6.586395052746453, + "grad_norm": 0.6016244949515843, + "learning_rate": 4.2734918512207903e-07, + "loss": 0.2004, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08334504067897797, + "step": 9055, + "valid_targets_mean": 4173.6, + "valid_targets_min": 2736 + }, + { + "epoch": 6.590032739177883, + "grad_norm": 0.6355934704330896, + "learning_rate": 4.199239157792701e-07, + "loss": 0.2049, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08299373835325241, + "step": 9060, + "valid_targets_mean": 3603.9, + "valid_targets_min": 1439 + }, + { + "epoch": 6.593670425609313, + "grad_norm": 0.61572426948566, + "learning_rate": 4.125630363217381e-07, + "loss": 0.2111, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09605669230222702, + "step": 9065, + "valid_targets_mean": 4739.2, + "valid_targets_min": 1838 + }, + { + "epoch": 6.597308112040742, + "grad_norm": 0.5665501745326615, + "learning_rate": 4.052665709560377e-07, + "loss": 0.1945, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11633896082639694, + "step": 9070, + "valid_targets_mean": 5608.2, + "valid_targets_min": 2405 + }, + { + "epoch": 6.600945798472171, + "grad_norm": 0.5937661238629611, + "learning_rate": 3.9803454367689064e-07, + "loss": 0.2052, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08373953402042389, + "step": 9075, + "valid_targets_mean": 3184.2, + "valid_targets_min": 1654 + }, + { + "epoch": 6.604583484903602, + "grad_norm": 0.5714778791275026, + "learning_rate": 3.908669782671237e-07, + "loss": 0.2017, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10007348656654358, + "step": 9080, + "valid_targets_mean": 5555.0, + "valid_targets_min": 3805 + }, + { + "epoch": 6.608221171335031, + "grad_norm": 0.6236960786904175, + "learning_rate": 3.8376389829756664e-07, + "loss": 0.215, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09843029081821442, + "step": 9085, + "valid_targets_mean": 4677.1, + "valid_targets_min": 1544 + }, + { + "epoch": 6.61185885776646, + "grad_norm": 0.5188989156257346, + "learning_rate": 3.7672532712698993e-07, + "loss": 0.2008, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09717538952827454, + "step": 9090, + "valid_targets_mean": 5710.6, + "valid_targets_min": 953 + }, + { + "epoch": 6.615496544197891, + "grad_norm": 0.6108105633137265, + "learning_rate": 3.6975128790202264e-07, + "loss": 0.2049, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11747584491968155, + "step": 9095, + "valid_targets_mean": 5297.4, + "valid_targets_min": 1265 + }, + { + "epoch": 6.61913423062932, + "grad_norm": 0.5634401313765528, + "learning_rate": 3.628418035570791e-07, + "loss": 0.2048, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09857988357543945, + "step": 9100, + "valid_targets_mean": 5766.1, + "valid_targets_min": 2566 + }, + { + "epoch": 6.622771917060749, + "grad_norm": 0.5902752294132522, + "learning_rate": 3.5599689681428795e-07, + "loss": 0.2035, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0878555029630661, + "step": 9105, + "valid_targets_mean": 3699.2, + "valid_targets_min": 1759 + }, + { + "epoch": 6.626409603492179, + "grad_norm": 0.6641620108564258, + "learning_rate": 3.4921659018340324e-07, + "loss": 0.1953, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09167471528053284, + "step": 9110, + "valid_targets_mean": 3070.5, + "valid_targets_min": 1451 + }, + { + "epoch": 6.630047289923609, + "grad_norm": 0.6221640797115767, + "learning_rate": 3.4250090596174233e-07, + "loss": 0.1957, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09388487040996552, + "step": 9115, + "valid_targets_mean": 3941.4, + "valid_targets_min": 1785 + }, + { + "epoch": 6.633684976355038, + "grad_norm": 0.6192056707107922, + "learning_rate": 3.3584986623410585e-07, + "loss": 0.1952, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10646381229162216, + "step": 9120, + "valid_targets_mean": 4304.8, + "valid_targets_min": 2345 + }, + { + "epoch": 6.637322662786468, + "grad_norm": 0.634724813053147, + "learning_rate": 3.292634928727179e-07, + "loss": 0.1989, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10682906210422516, + "step": 9125, + "valid_targets_mean": 3978.0, + "valid_targets_min": 2767 + }, + { + "epoch": 6.640960349217897, + "grad_norm": 0.6762443046118477, + "learning_rate": 3.227418075371325e-07, + "loss": 0.2066, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09616175293922424, + "step": 9130, + "valid_targets_mean": 3366.4, + "valid_targets_min": 1205 + }, + { + "epoch": 6.644598035649327, + "grad_norm": 0.5861443511926535, + "learning_rate": 3.162848316741829e-07, + "loss": 0.1982, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09885793179273605, + "step": 9135, + "valid_targets_mean": 4209.6, + "valid_targets_min": 2004 + }, + { + "epoch": 6.648235722080757, + "grad_norm": 0.6197888825045001, + "learning_rate": 3.09892586517897e-07, + "loss": 0.2028, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1089317724108696, + "step": 9140, + "valid_targets_mean": 4580.6, + "valid_targets_min": 1469 + }, + { + "epoch": 6.651873408512186, + "grad_norm": 0.6695081153423252, + "learning_rate": 3.035650930894396e-07, + "loss": 0.1846, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11764448881149292, + "step": 9145, + "valid_targets_mean": 4466.8, + "valid_targets_min": 2610 + }, + { + "epoch": 6.655511094943616, + "grad_norm": 0.6304885856264902, + "learning_rate": 2.9730237219703026e-07, + "loss": 0.1882, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09724260866641998, + "step": 9150, + "valid_targets_mean": 4223.9, + "valid_targets_min": 2046 + }, + { + "epoch": 6.659148781375046, + "grad_norm": 0.6735558696999867, + "learning_rate": 2.9110444443588573e-07, + "loss": 0.1875, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12535706162452698, + "step": 9155, + "valid_targets_mean": 4617.2, + "valid_targets_min": 3373 + }, + { + "epoch": 6.662786467806475, + "grad_norm": 0.655728205448021, + "learning_rate": 2.8497133018814407e-07, + "loss": 0.1983, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10089924186468124, + "step": 9160, + "valid_targets_mean": 3745.6, + "valid_targets_min": 1747 + }, + { + "epoch": 6.666424154237904, + "grad_norm": 0.673760700072197, + "learning_rate": 2.789030496228051e-07, + "loss": 0.201, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08270110189914703, + "step": 9165, + "valid_targets_mean": 2923.9, + "valid_targets_min": 1166 + }, + { + "epoch": 6.670061840669335, + "grad_norm": 0.608628995000805, + "learning_rate": 2.728996226956593e-07, + "loss": 0.1911, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09401462972164154, + "step": 9170, + "valid_targets_mean": 4002.2, + "valid_targets_min": 1537 + }, + { + "epoch": 6.673699527100764, + "grad_norm": 0.6110763165870186, + "learning_rate": 2.6696106914922525e-07, + "loss": 0.19, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09339277446269989, + "step": 9175, + "valid_targets_mean": 4572.6, + "valid_targets_min": 3119 + }, + { + "epoch": 6.677337213532193, + "grad_norm": 0.6483867947846704, + "learning_rate": 2.6108740851267467e-07, + "loss": 0.1917, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09184390306472778, + "step": 9180, + "valid_targets_mean": 3756.0, + "valid_targets_min": 2562 + }, + { + "epoch": 6.680974899963623, + "grad_norm": 0.6334543432696639, + "learning_rate": 2.5527866010178536e-07, + "loss": 0.1907, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0959712415933609, + "step": 9185, + "valid_targets_mean": 4084.4, + "valid_targets_min": 2659 + }, + { + "epoch": 6.684612586395053, + "grad_norm": 0.6231185254577087, + "learning_rate": 2.49534843018866e-07, + "loss": 0.1855, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08765615522861481, + "step": 9190, + "valid_targets_mean": 3971.2, + "valid_targets_min": 1349 + }, + { + "epoch": 6.688250272826482, + "grad_norm": 0.6480167663486762, + "learning_rate": 2.4385597615269376e-07, + "loss": 0.1898, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0789773017168045, + "step": 9195, + "valid_targets_mean": 3099.9, + "valid_targets_min": 1567 + }, + { + "epoch": 6.691887959257912, + "grad_norm": 0.7086197253279554, + "learning_rate": 2.382420781784589e-07, + "loss": 0.1902, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07857057452201843, + "step": 9200, + "valid_targets_mean": 2787.2, + "valid_targets_min": 1500 + }, + { + "epoch": 6.695525645689342, + "grad_norm": 0.6181476541997896, + "learning_rate": 2.3269316755768933e-07, + "loss": 0.1813, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08673426508903503, + "step": 9205, + "valid_targets_mean": 3431.1, + "valid_targets_min": 2033 + }, + { + "epoch": 6.699163332120771, + "grad_norm": 0.8650783341632495, + "learning_rate": 2.2720926253821273e-07, + "loss": 0.1924, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10629917681217194, + "step": 9210, + "valid_targets_mean": 2845.0, + "valid_targets_min": 1346 + }, + { + "epoch": 6.702801018552201, + "grad_norm": 0.7985695935727825, + "learning_rate": 2.2179038115407003e-07, + "loss": 0.1891, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09637464582920074, + "step": 9215, + "valid_targets_mean": 2877.2, + "valid_targets_min": 1821 + }, + { + "epoch": 6.70643870498363, + "grad_norm": 0.8058008286314676, + "learning_rate": 2.1643654122548207e-07, + "loss": 0.1898, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0932677686214447, + "step": 9220, + "valid_targets_mean": 3490.1, + "valid_targets_min": 2535 + }, + { + "epoch": 6.71007639141506, + "grad_norm": 0.6572807136555466, + "learning_rate": 2.1114776035876748e-07, + "loss": 0.1871, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08622194826602936, + "step": 9225, + "valid_targets_mean": 3276.2, + "valid_targets_min": 2352 + }, + { + "epoch": 6.71371407784649, + "grad_norm": 0.7014344864745535, + "learning_rate": 2.059240559463005e-07, + "loss": 0.1907, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09426698088645935, + "step": 9230, + "valid_targets_mean": 3097.9, + "valid_targets_min": 1153 + }, + { + "epoch": 6.717351764277919, + "grad_norm": 0.6208158090590651, + "learning_rate": 2.0076544516644646e-07, + "loss": 0.1891, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09258632361888885, + "step": 9235, + "valid_targets_mean": 3562.4, + "valid_targets_min": 2072 + }, + { + "epoch": 6.720989450709348, + "grad_norm": 0.6467074617140777, + "learning_rate": 1.9567194498351093e-07, + "loss": 0.1928, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08063232898712158, + "step": 9240, + "valid_targets_mean": 3239.8, + "valid_targets_min": 922 + }, + { + "epoch": 6.724627137140779, + "grad_norm": 0.6261921404446175, + "learning_rate": 1.906435721476796e-07, + "loss": 0.1861, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07426629960536957, + "step": 9245, + "valid_targets_mean": 3867.5, + "valid_targets_min": 1417 + }, + { + "epoch": 6.728264823572208, + "grad_norm": 0.6253617235972133, + "learning_rate": 1.8568034319495832e-07, + "loss": 0.1923, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08791183680295944, + "step": 9250, + "valid_targets_mean": 4348.8, + "valid_targets_min": 1844 + }, + { + "epoch": 6.731902510003637, + "grad_norm": 0.6641837096569552, + "learning_rate": 1.8078227444713548e-07, + "loss": 0.1838, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10204626619815826, + "step": 9255, + "valid_targets_mean": 3871.9, + "valid_targets_min": 1261 + }, + { + "epoch": 6.735540196435068, + "grad_norm": 0.6378276604679073, + "learning_rate": 1.7594938201170863e-07, + "loss": 0.1862, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10780606418848038, + "step": 9260, + "valid_targets_mean": 4526.2, + "valid_targets_min": 2077 + }, + { + "epoch": 6.739177882866497, + "grad_norm": 0.6284389577937012, + "learning_rate": 1.7118168178184236e-07, + "loss": 0.1847, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1028466746211052, + "step": 9265, + "valid_targets_mean": 4245.9, + "valid_targets_min": 2724 + }, + { + "epoch": 6.742815569297926, + "grad_norm": 0.6412656282784194, + "learning_rate": 1.6647918943631937e-07, + "loss": 0.1884, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11818064749240875, + "step": 9270, + "valid_targets_mean": 4940.1, + "valid_targets_min": 2292 + }, + { + "epoch": 6.746453255729357, + "grad_norm": 0.7149270362884091, + "learning_rate": 1.6184192043947166e-07, + "loss": 0.1913, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09232407063245773, + "step": 9275, + "valid_targets_mean": 3041.5, + "valid_targets_min": 1377 + }, + { + "epoch": 6.750090942160786, + "grad_norm": 0.603127287886402, + "learning_rate": 1.5726989004115622e-07, + "loss": 0.192, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09093765914440155, + "step": 9280, + "valid_targets_mean": 3381.0, + "valid_targets_min": 2395 + }, + { + "epoch": 6.753728628592215, + "grad_norm": 0.6598546718598943, + "learning_rate": 1.5276311327667936e-07, + "loss": 0.1895, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08933129161596298, + "step": 9285, + "valid_targets_mean": 3055.2, + "valid_targets_min": 654 + }, + { + "epoch": 6.757366315023645, + "grad_norm": 0.6082128454788234, + "learning_rate": 1.4832160496676128e-07, + "loss": 0.1926, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08989579975605011, + "step": 9290, + "valid_targets_mean": 3746.1, + "valid_targets_min": 1706 + }, + { + "epoch": 6.761004001455074, + "grad_norm": 0.628391494513897, + "learning_rate": 1.4394537971748946e-07, + "loss": 0.1907, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10210548341274261, + "step": 9295, + "valid_targets_mean": 4521.0, + "valid_targets_min": 1739 + }, + { + "epoch": 6.764641687886504, + "grad_norm": 0.638322560720791, + "learning_rate": 1.3963445192025417e-07, + "loss": 0.1898, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0781676322221756, + "step": 9300, + "valid_targets_mean": 2924.1, + "valid_targets_min": 1140 + }, + { + "epoch": 6.768279374317934, + "grad_norm": 0.6087907253569477, + "learning_rate": 1.3538883575171967e-07, + "loss": 0.1757, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.082227423787117, + "step": 9305, + "valid_targets_mean": 3621.6, + "valid_targets_min": 2501 + }, + { + "epoch": 6.771917060749363, + "grad_norm": 0.5912728072083482, + "learning_rate": 1.3120854517377324e-07, + "loss": 0.1875, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09463781118392944, + "step": 9310, + "valid_targets_mean": 4796.2, + "valid_targets_min": 3232 + }, + { + "epoch": 6.775554747180793, + "grad_norm": 1.9293488105601104, + "learning_rate": 1.2709359393347165e-07, + "loss": 0.1746, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09143277257680893, + "step": 9315, + "valid_targets_mean": 4139.2, + "valid_targets_min": 1101 + }, + { + "epoch": 6.779192433612223, + "grad_norm": 0.683428532268907, + "learning_rate": 1.2304399556299917e-07, + "loss": 0.2044, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1154882162809372, + "step": 9320, + "valid_targets_mean": 4282.5, + "valid_targets_min": 2079 + }, + { + "epoch": 6.782830120043652, + "grad_norm": 0.6438237721773814, + "learning_rate": 1.1905976337962533e-07, + "loss": 0.1901, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08787211775779724, + "step": 9325, + "valid_targets_mean": 3253.9, + "valid_targets_min": 1743 + }, + { + "epoch": 6.786467806475082, + "grad_norm": 0.6282659936434376, + "learning_rate": 1.1514091048566489e-07, + "loss": 0.193, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09667611122131348, + "step": 9330, + "valid_targets_mean": 3566.6, + "valid_targets_min": 1234 + }, + { + "epoch": 6.790105492906512, + "grad_norm": 0.6952836821558238, + "learning_rate": 1.1128744976842465e-07, + "loss": 0.195, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0931648313999176, + "step": 9335, + "valid_targets_mean": 4002.4, + "valid_targets_min": 1822 + }, + { + "epoch": 6.793743179337941, + "grad_norm": 0.62549042307452, + "learning_rate": 1.0749939390017006e-07, + "loss": 0.1974, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08386340737342834, + "step": 9340, + "valid_targets_mean": 3697.1, + "valid_targets_min": 1651 + }, + { + "epoch": 6.79738086576937, + "grad_norm": 0.6734547710859565, + "learning_rate": 1.0377675533807641e-07, + "loss": 0.1921, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10287788510322571, + "step": 9345, + "valid_targets_mean": 3215.8, + "valid_targets_min": 1374 + }, + { + "epoch": 6.801018552200801, + "grad_norm": 0.6300690657737684, + "learning_rate": 1.0011954632419551e-07, + "loss": 0.18, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09985345602035522, + "step": 9350, + "valid_targets_mean": 4574.8, + "valid_targets_min": 2725 + }, + { + "epoch": 6.80465623863223, + "grad_norm": 0.6150374971168788, + "learning_rate": 9.652777888541131e-08, + "loss": 0.1836, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10298489779233932, + "step": 9355, + "valid_targets_mean": 4331.6, + "valid_targets_min": 1966 + }, + { + "epoch": 6.808293925063659, + "grad_norm": 0.7339869140621635, + "learning_rate": 9.300146483339323e-08, + "loss": 0.1788, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09358154237270355, + "step": 9360, + "valid_targets_mean": 3265.5, + "valid_targets_min": 1407 + }, + { + "epoch": 6.811931611495089, + "grad_norm": 0.6516220633446448, + "learning_rate": 8.954061576457396e-08, + "loss": 0.18, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09653577208518982, + "step": 9365, + "valid_targets_mean": 3937.5, + "valid_targets_min": 1444 + }, + { + "epoch": 6.815569297926519, + "grad_norm": 0.6456997701029382, + "learning_rate": 8.614524306009842e-08, + "loss": 0.1742, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08693601191043854, + "step": 9370, + "valid_targets_mean": 3222.6, + "valid_targets_min": 1504 + }, + { + "epoch": 6.819206984357948, + "grad_norm": 0.64866087632252, + "learning_rate": 8.28153578857882e-08, + "loss": 0.1705, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.072783924639225, + "step": 9375, + "valid_targets_mean": 2694.6, + "valid_targets_min": 858 + }, + { + "epoch": 6.822844670789378, + "grad_norm": 0.6687062657876721, + "learning_rate": 7.955097119210609e-08, + "loss": 0.1865, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09193064272403717, + "step": 9380, + "valid_targets_mean": 3778.6, + "valid_targets_min": 1288 + }, + { + "epoch": 6.826482357220808, + "grad_norm": 1.1279129590979862, + "learning_rate": 7.635209371412489e-08, + "loss": 0.2029, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09083223342895508, + "step": 9385, + "valid_targets_mean": 4008.2, + "valid_targets_min": 1537 + }, + { + "epoch": 6.830120043652237, + "grad_norm": 0.6930355472867588, + "learning_rate": 7.32187359714831e-08, + "loss": 0.1898, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11253908276557922, + "step": 9390, + "valid_targets_mean": 4195.2, + "valid_targets_min": 2642 + }, + { + "epoch": 6.833757730083667, + "grad_norm": 0.6394543443263588, + "learning_rate": 7.015090826835603e-08, + "loss": 0.1931, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10663770139217377, + "step": 9395, + "valid_targets_mean": 4148.8, + "valid_targets_min": 2013 + }, + { + "epoch": 6.837395416515096, + "grad_norm": 0.6447410586550241, + "learning_rate": 6.714862069342021e-08, + "loss": 0.1869, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09736479073762894, + "step": 9400, + "valid_targets_mean": 4112.2, + "valid_targets_min": 1744 + }, + { + "epoch": 6.841033102946526, + "grad_norm": 0.6400297945692115, + "learning_rate": 6.421188311982684e-08, + "loss": 0.1922, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08869846165180206, + "step": 9405, + "valid_targets_mean": 4169.5, + "valid_targets_min": 1914 + }, + { + "epoch": 6.844670789377956, + "grad_norm": 0.6655364412810247, + "learning_rate": 6.13407052051529e-08, + "loss": 0.1855, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07831773161888123, + "step": 9410, + "valid_targets_mean": 3479.8, + "valid_targets_min": 1181 + }, + { + "epoch": 6.848308475809385, + "grad_norm": 0.6394097700342698, + "learning_rate": 5.853509639138777e-08, + "loss": 0.1773, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09875177592039108, + "step": 9415, + "valid_targets_mean": 4103.2, + "valid_targets_min": 1937 + }, + { + "epoch": 6.851946162240814, + "grad_norm": 0.7239916997499976, + "learning_rate": 5.5795065904888926e-08, + "loss": 0.1935, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10490909218788147, + "step": 9420, + "valid_targets_mean": 3203.0, + "valid_targets_min": 1649 + }, + { + "epoch": 6.855583848672245, + "grad_norm": 0.7315948785196692, + "learning_rate": 5.312062275636187e-08, + "loss": 0.1805, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08180540800094604, + "step": 9425, + "valid_targets_mean": 2847.4, + "valid_targets_min": 1197 + }, + { + "epoch": 6.859221535103674, + "grad_norm": 0.6530382425205488, + "learning_rate": 5.051177574082244e-08, + "loss": 0.1889, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09313544631004333, + "step": 9430, + "valid_targets_mean": 3441.4, + "valid_targets_min": 1824 + }, + { + "epoch": 6.862859221535103, + "grad_norm": 0.7179412900274088, + "learning_rate": 4.796853343757013e-08, + "loss": 0.1973, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08854475617408752, + "step": 9435, + "valid_targets_mean": 2920.4, + "valid_targets_min": 538 + }, + { + "epoch": 6.866496907966534, + "grad_norm": 0.6504178715409094, + "learning_rate": 4.5490904210161443e-08, + "loss": 0.1899, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08997946977615356, + "step": 9440, + "valid_targets_mean": 3919.2, + "valid_targets_min": 537 + }, + { + "epoch": 6.870134594397963, + "grad_norm": 0.7158671258387808, + "learning_rate": 4.30788962063855e-08, + "loss": 0.2055, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11576934158802032, + "step": 9445, + "valid_targets_mean": 3630.1, + "valid_targets_min": 2169 + }, + { + "epoch": 6.873772280829392, + "grad_norm": 0.6811870415461045, + "learning_rate": 4.073251735822403e-08, + "loss": 0.1872, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09216710925102234, + "step": 9450, + "valid_targets_mean": 3427.9, + "valid_targets_min": 1625 + }, + { + "epoch": 6.877409967260822, + "grad_norm": 0.6559423159808994, + "learning_rate": 3.845177538184697e-08, + "loss": 0.2017, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08505384624004364, + "step": 9455, + "valid_targets_mean": 3678.0, + "valid_targets_min": 1997 + }, + { + "epoch": 6.881047653692252, + "grad_norm": 0.8974156900821202, + "learning_rate": 3.6236677777561345e-08, + "loss": 0.2011, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11590538918972015, + "step": 9460, + "valid_targets_mean": 4401.4, + "valid_targets_min": 2935 + }, + { + "epoch": 6.884685340123681, + "grad_norm": 0.6006318582032555, + "learning_rate": 3.408723182981355e-08, + "loss": 0.1963, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10319983959197998, + "step": 9465, + "valid_targets_mean": 5173.2, + "valid_targets_min": 2652 + }, + { + "epoch": 6.888323026555111, + "grad_norm": 0.6921350750396422, + "learning_rate": 3.200344460713822e-08, + "loss": 0.2002, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11386243999004364, + "step": 9470, + "valid_targets_mean": 4345.4, + "valid_targets_min": 1381 + }, + { + "epoch": 6.89196071298654, + "grad_norm": 0.6501354529472395, + "learning_rate": 2.9985322962160504e-08, + "loss": 0.1913, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0788707286119461, + "step": 9475, + "valid_targets_mean": 3375.8, + "valid_targets_min": 637 + }, + { + "epoch": 6.89559839941797, + "grad_norm": 0.6793041493489081, + "learning_rate": 2.8032873531553818e-08, + "loss": 0.193, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09024740755558014, + "step": 9480, + "valid_targets_mean": 3335.4, + "valid_targets_min": 719 + }, + { + "epoch": 6.8992360858494, + "grad_norm": 0.6443854747301324, + "learning_rate": 2.61461027360288e-08, + "loss": 0.1947, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08147332817316055, + "step": 9485, + "valid_targets_mean": 4440.4, + "valid_targets_min": 2182 + }, + { + "epoch": 6.902873772280829, + "grad_norm": 0.6793674535040548, + "learning_rate": 2.4325016780308853e-08, + "loss": 0.1833, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10034150630235672, + "step": 9490, + "valid_targets_mean": 4045.1, + "valid_targets_min": 1372 + }, + { + "epoch": 6.906511458712259, + "grad_norm": 0.7417344187568157, + "learning_rate": 2.256962165311016e-08, + "loss": 0.1833, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11274001002311707, + "step": 9495, + "valid_targets_mean": 4512.6, + "valid_targets_min": 2768 + }, + { + "epoch": 6.910149145143689, + "grad_norm": 0.657465120456138, + "learning_rate": 2.0879923127123946e-08, + "loss": 0.1898, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07671374827623367, + "step": 9500, + "valid_targets_mean": 2918.2, + "valid_targets_min": 530 + }, + { + "epoch": 6.913786831575118, + "grad_norm": 0.7062001553894033, + "learning_rate": 1.9255926758989795e-08, + "loss": 0.1875, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10334539413452148, + "step": 9505, + "valid_targets_mean": 3749.5, + "valid_targets_min": 611 + }, + { + "epoch": 6.9174245180065475, + "grad_norm": 0.6311711525193738, + "learning_rate": 1.76976378892868e-08, + "loss": 0.1995, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08968045562505722, + "step": 9510, + "valid_targets_mean": 3685.0, + "valid_targets_min": 1728 + }, + { + "epoch": 6.921062204437978, + "grad_norm": 0.6869124429922997, + "learning_rate": 1.6205061642513565e-08, + "loss": 0.1906, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09583739936351776, + "step": 9515, + "valid_targets_mean": 3281.9, + "valid_targets_min": 1362 + }, + { + "epoch": 6.924699890869407, + "grad_norm": 0.6334133222834776, + "learning_rate": 1.4778202927068219e-08, + "loss": 0.1927, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08698935061693192, + "step": 9520, + "valid_targets_mean": 3194.5, + "valid_targets_min": 852 + }, + { + "epoch": 6.928337577300836, + "grad_norm": 0.6945563015773214, + "learning_rate": 1.3417066435232885e-08, + "loss": 0.1927, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07487818598747253, + "step": 9525, + "valid_targets_mean": 2485.8, + "valid_targets_min": 1195 + }, + { + "epoch": 6.931975263732266, + "grad_norm": 0.664077125810339, + "learning_rate": 1.2121656643164781e-08, + "loss": 0.1956, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10741528123617172, + "step": 9530, + "valid_targets_mean": 4512.6, + "valid_targets_min": 1397 + }, + { + "epoch": 6.935612950163696, + "grad_norm": 0.6758930104280826, + "learning_rate": 1.0891977810867372e-08, + "loss": 0.1874, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07957504689693451, + "step": 9535, + "valid_targets_mean": 3585.1, + "valid_targets_min": 1975 + }, + { + "epoch": 6.939250636595125, + "grad_norm": 0.632738948077458, + "learning_rate": 9.728033982192575e-09, + "loss": 0.1781, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09706446528434753, + "step": 9540, + "valid_targets_mean": 4138.1, + "valid_targets_min": 2564 + }, + { + "epoch": 6.942888323026555, + "grad_norm": 0.5925100439795884, + "learning_rate": 8.629828984818568e-09, + "loss": 0.1855, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08914048969745636, + "step": 9545, + "valid_targets_mean": 5947.9, + "valid_targets_min": 2595 + }, + { + "epoch": 6.946526009457985, + "grad_norm": 0.6126316609896713, + "learning_rate": 7.597366430234232e-09, + "loss": 0.1848, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10090913623571396, + "step": 9550, + "valid_targets_mean": 3978.2, + "valid_targets_min": 2845 + }, + { + "epoch": 6.950163695889414, + "grad_norm": 0.7658281140992341, + "learning_rate": 6.630649713739168e-09, + "loss": 0.1783, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07699738442897797, + "step": 9555, + "valid_targets_mean": 2705.5, + "valid_targets_min": 993 + }, + { + "epoch": 6.953801382320844, + "grad_norm": 0.6241237848563637, + "learning_rate": 5.729682014417037e-09, + "loss": 0.176, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0777236670255661, + "step": 9560, + "valid_targets_mean": 3516.2, + "valid_targets_min": 1575 + }, + { + "epoch": 6.957439068752274, + "grad_norm": 0.6813612384866146, + "learning_rate": 4.89446629514001e-09, + "loss": 0.1824, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10267175734043121, + "step": 9565, + "valid_targets_mean": 3603.6, + "valid_targets_min": 1360 + }, + { + "epoch": 6.961076755183703, + "grad_norm": 0.625299205728955, + "learning_rate": 4.125005302548779e-09, + "loss": 0.1783, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09203311800956726, + "step": 9570, + "valid_targets_mean": 3558.2, + "valid_targets_min": 2278 + }, + { + "epoch": 6.964714441615133, + "grad_norm": 0.6498218935371392, + "learning_rate": 3.421301567045898e-09, + "loss": 0.1977, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1062920093536377, + "step": 9575, + "valid_targets_mean": 4035.0, + "valid_targets_min": 1454 + }, + { + "epoch": 6.968352128046562, + "grad_norm": 0.6673951674221013, + "learning_rate": 2.783357402793563e-09, + "loss": 0.1852, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09096456319093704, + "step": 9580, + "valid_targets_mean": 3655.6, + "valid_targets_min": 1720 + }, + { + "epoch": 6.971989814477992, + "grad_norm": 0.7173461731356011, + "learning_rate": 2.211174907693625e-09, + "loss": 0.1858, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1027454286813736, + "step": 9585, + "valid_targets_mean": 3471.6, + "valid_targets_min": 1278 + }, + { + "epoch": 6.975627500909422, + "grad_norm": 0.7194942424499599, + "learning_rate": 1.7047559633920353e-09, + "loss": 0.1767, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10155707597732544, + "step": 9590, + "valid_targets_mean": 3191.1, + "valid_targets_min": 1858 + }, + { + "epoch": 6.979265187340851, + "grad_norm": 0.6755454732893658, + "learning_rate": 1.26410223527218e-09, + "loss": 0.187, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07620224356651306, + "step": 9595, + "valid_targets_mean": 2954.8, + "valid_targets_min": 628 + }, + { + "epoch": 6.9829028737722805, + "grad_norm": 0.7263477397213285, + "learning_rate": 8.89215172437119e-10, + "loss": 0.1851, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08595754206180573, + "step": 9600, + "valid_targets_mean": 2642.0, + "valid_targets_min": 1039 + }, + { + "epoch": 6.986540560203711, + "grad_norm": 0.6249723451349228, + "learning_rate": 5.800960077206874e-10, + "loss": 0.1931, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08751708269119263, + "step": 9605, + "valid_targets_mean": 4193.1, + "valid_targets_min": 1737 + }, + { + "epoch": 6.99017824663514, + "grad_norm": 0.6860592741110955, + "learning_rate": 3.3674575767417283e-10, + "loss": 0.1898, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11643706262111664, + "step": 9610, + "valid_targets_mean": 3987.8, + "valid_targets_min": 809 + }, + { + "epoch": 6.9938159330665695, + "grad_norm": 0.6245804612969832, + "learning_rate": 1.591652225663154e-10, + "loss": 0.1868, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09224496781826019, + "step": 9615, + "valid_targets_mean": 3707.1, + "valid_targets_min": 1780 + }, + { + "epoch": 6.997453619498, + "grad_norm": 0.6258120168403039, + "learning_rate": 4.735498637442604e-11, + "loss": 0.1935, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09841226041316986, + "step": 9620, + "valid_targets_mean": 4457.8, + "valid_targets_min": 2185 + }, + { + "epoch": 7.0, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0703798308968544, + "step": 9624, + "total_flos": 3.5396290752681083e+18, + "train_loss": 0.038640074608406225, + "train_runtime": 58770.888, + "train_samples_per_second": 2.619, + "train_steps_per_second": 0.164, + "valid_targets_mean": 1910.0, + "valid_targets_min": 393 + } + ], + "logging_steps": 5, + "max_steps": 9625, + "num_input_tokens_seen": 0, + "num_train_epochs": 7, + "save_steps": 200, + "stateful_callbacks": { + "TrainerControl": { + "args": { + "should_epoch_stop": false, + "should_evaluate": false, + "should_log": false, + "should_save": true, + "should_training_stop": false + }, + "attributes": {} + } + }, + "total_flos": 3.5396290752681083e+18, + "train_batch_size": 1, + "trial_name": null, + "trial_params": null +}