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": 16.672173708006987, + "learning_rate": 1.6614745586708205e-07, + "loss": 0.6605, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3098912835121155, + "step": 5, + "valid_targets_mean": 7389.0, + "valid_targets_min": 5627 + }, + { + "epoch": 0.007275372862859221, + "grad_norm": 16.314322697317795, + "learning_rate": 3.7383177570093457e-07, + "loss": 0.7009, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23654916882514954, + "step": 10, + "valid_targets_mean": 1294.8, + "valid_targets_min": 422 + }, + { + "epoch": 0.010913059294288832, + "grad_norm": 15.154305669825192, + "learning_rate": 5.815160955347872e-07, + "loss": 0.6806, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.32596924901008606, + "step": 15, + "valid_targets_mean": 8539.9, + "valid_targets_min": 6761 + }, + { + "epoch": 0.014550745725718442, + "grad_norm": 13.472823488828476, + "learning_rate": 7.892004153686398e-07, + "loss": 0.6152, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2906034290790558, + "step": 20, + "valid_targets_mean": 7579.0, + "valid_targets_min": 6174 + }, + { + "epoch": 0.018188432157148052, + "grad_norm": 11.937048801548682, + "learning_rate": 9.968847352024923e-07, + "loss": 0.6185, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29628682136535645, + "step": 25, + "valid_targets_mean": 7328.2, + "valid_targets_min": 4839 + }, + { + "epoch": 0.021826118588577664, + "grad_norm": 8.915400489990493, + "learning_rate": 1.2045690550363449e-06, + "loss": 0.6127, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.32500478625297546, + "step": 30, + "valid_targets_mean": 6838.2, + "valid_targets_min": 5034 + }, + { + "epoch": 0.025463805020007276, + "grad_norm": 6.220568443019347, + "learning_rate": 1.4122533748701975e-06, + "loss": 0.5947, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.296002596616745, + "step": 35, + "valid_targets_mean": 7248.2, + "valid_targets_min": 5567 + }, + { + "epoch": 0.029101491451436885, + "grad_norm": 5.494528934238525, + "learning_rate": 1.6199376947040499e-06, + "loss": 0.5516, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2632763385772705, + "step": 40, + "valid_targets_mean": 6234.0, + "valid_targets_min": 5603 + }, + { + "epoch": 0.032739177882866496, + "grad_norm": 6.739352265927698, + "learning_rate": 1.8276220145379025e-06, + "loss": 0.5489, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12719041109085083, + "step": 45, + "valid_targets_mean": 602.0, + "valid_targets_min": 152 + }, + { + "epoch": 0.036376864314296105, + "grad_norm": 4.365508219314055, + "learning_rate": 2.0353063343717553e-06, + "loss": 0.4772, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22969640791416168, + "step": 50, + "valid_targets_mean": 7299.9, + "valid_targets_min": 6158 + }, + { + "epoch": 0.04001455074572572, + "grad_norm": 2.4526025958396356, + "learning_rate": 2.2429906542056077e-06, + "loss": 0.4548, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2125677764415741, + "step": 55, + "valid_targets_mean": 6140.0, + "valid_targets_min": 4648 + }, + { + "epoch": 0.04365223717715533, + "grad_norm": 1.703958530580982, + "learning_rate": 2.4506749740394605e-06, + "loss": 0.4333, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21467891335487366, + "step": 60, + "valid_targets_mean": 6603.0, + "valid_targets_min": 4882 + }, + { + "epoch": 0.04728992360858494, + "grad_norm": 1.166580850118028, + "learning_rate": 2.658359293873313e-06, + "loss": 0.4093, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20930930972099304, + "step": 65, + "valid_targets_mean": 8015.0, + "valid_targets_min": 6344 + }, + { + "epoch": 0.05092761004001455, + "grad_norm": 1.0175497627258907, + "learning_rate": 2.8660436137071652e-06, + "loss": 0.3969, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19320563971996307, + "step": 70, + "valid_targets_mean": 6305.2, + "valid_targets_min": 4622 + }, + { + "epoch": 0.05456529647144416, + "grad_norm": 0.8554899976971551, + "learning_rate": 3.073727933541018e-06, + "loss": 0.3991, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2037082016468048, + "step": 75, + "valid_targets_mean": 7890.8, + "valid_targets_min": 5485 + }, + { + "epoch": 0.05820298290287377, + "grad_norm": 0.7138087348113471, + "learning_rate": 3.2814122533748704e-06, + "loss": 0.3937, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17891252040863037, + "step": 80, + "valid_targets_mean": 7016.6, + "valid_targets_min": 4960 + }, + { + "epoch": 0.061840669334303384, + "grad_norm": 0.6522761461647447, + "learning_rate": 3.489096573208723e-06, + "loss": 0.3795, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17854025959968567, + "step": 85, + "valid_targets_mean": 6527.1, + "valid_targets_min": 5399 + }, + { + "epoch": 0.06547835576573299, + "grad_norm": 0.7421899918288901, + "learning_rate": 3.6967808930425756e-06, + "loss": 0.3773, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18866443634033203, + "step": 90, + "valid_targets_mean": 4995.0, + "valid_targets_min": 618 + }, + { + "epoch": 0.06911604219716261, + "grad_norm": 0.5883113743799047, + "learning_rate": 3.9044652128764284e-06, + "loss": 0.387, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.193133145570755, + "step": 95, + "valid_targets_mean": 7219.4, + "valid_targets_min": 5764 + }, + { + "epoch": 0.07275372862859221, + "grad_norm": 0.5319122996122065, + "learning_rate": 4.112149532710281e-06, + "loss": 0.3717, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1931883990764618, + "step": 100, + "valid_targets_mean": 7998.5, + "valid_targets_min": 5902 + }, + { + "epoch": 0.07639141506002183, + "grad_norm": 0.5072737311866509, + "learning_rate": 4.319833852544133e-06, + "loss": 0.3359, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15039585530757904, + "step": 105, + "valid_targets_mean": 7377.4, + "valid_targets_min": 5015 + }, + { + "epoch": 0.08002910149145144, + "grad_norm": 0.5011635487896844, + "learning_rate": 4.527518172377986e-06, + "loss": 0.3375, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1493467539548874, + "step": 110, + "valid_targets_mean": 6794.0, + "valid_targets_min": 5283 + }, + { + "epoch": 0.08366678792288104, + "grad_norm": 0.5646028134971479, + "learning_rate": 4.735202492211838e-06, + "loss": 0.3514, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19867245852947235, + "step": 115, + "valid_targets_mean": 6656.4, + "valid_targets_min": 4364 + }, + { + "epoch": 0.08730447435431066, + "grad_norm": 0.5010932133864977, + "learning_rate": 4.94288681204569e-06, + "loss": 0.3378, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16071778535842896, + "step": 120, + "valid_targets_mean": 7178.4, + "valid_targets_min": 4448 + }, + { + "epoch": 0.09094216078574027, + "grad_norm": 0.5003703965046953, + "learning_rate": 5.150571131879544e-06, + "loss": 0.3313, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18735583126544952, + "step": 125, + "valid_targets_mean": 7619.8, + "valid_targets_min": 5737 + }, + { + "epoch": 0.09457984721716987, + "grad_norm": 0.42943453345457483, + "learning_rate": 5.358255451713395e-06, + "loss": 0.2927, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14204931259155273, + "step": 130, + "valid_targets_mean": 8754.4, + "valid_targets_min": 5752 + }, + { + "epoch": 0.09821753364859949, + "grad_norm": 0.5252342399500064, + "learning_rate": 5.565939771547248e-06, + "loss": 0.2909, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.159476637840271, + "step": 135, + "valid_targets_mean": 6341.0, + "valid_targets_min": 4543 + }, + { + "epoch": 0.1018552200800291, + "grad_norm": 0.4792323545403113, + "learning_rate": 5.773624091381102e-06, + "loss": 0.2791, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15746693313121796, + "step": 140, + "valid_targets_mean": 7181.9, + "valid_targets_min": 4563 + }, + { + "epoch": 0.1054929065114587, + "grad_norm": 1.0594006048915074, + "learning_rate": 5.981308411214953e-06, + "loss": 0.3404, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12635990977287292, + "step": 145, + "valid_targets_mean": 1104.4, + "valid_targets_min": 161 + }, + { + "epoch": 0.10913059294288832, + "grad_norm": 0.42312512186263185, + "learning_rate": 6.188992731048806e-06, + "loss": 0.325, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14115752279758453, + "step": 150, + "valid_targets_mean": 7018.2, + "valid_targets_min": 5424 + }, + { + "epoch": 0.11276827937431794, + "grad_norm": 0.46597033040855856, + "learning_rate": 6.39667705088266e-06, + "loss": 0.3226, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16497910022735596, + "step": 155, + "valid_targets_mean": 7738.4, + "valid_targets_min": 5329 + }, + { + "epoch": 0.11640596580574754, + "grad_norm": 0.5067381251642002, + "learning_rate": 6.604361370716511e-06, + "loss": 0.3289, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16063794493675232, + "step": 160, + "valid_targets_mean": 6426.6, + "valid_targets_min": 5560 + }, + { + "epoch": 0.12004365223717715, + "grad_norm": 0.5724971269242243, + "learning_rate": 6.812045690550364e-06, + "loss": 0.3202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16376347839832306, + "step": 165, + "valid_targets_mean": 6169.9, + "valid_targets_min": 3446 + }, + { + "epoch": 0.12368133866860677, + "grad_norm": 0.4695931645462594, + "learning_rate": 7.019730010384217e-06, + "loss": 0.3046, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15643103420734406, + "step": 170, + "valid_targets_mean": 7204.6, + "valid_targets_min": 5381 + }, + { + "epoch": 0.12731902510003637, + "grad_norm": 0.46319899158409356, + "learning_rate": 7.227414330218069e-06, + "loss": 0.3088, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15746241807937622, + "step": 175, + "valid_targets_mean": 7600.4, + "valid_targets_min": 5241 + }, + { + "epoch": 0.13095671153146599, + "grad_norm": 0.49694383798712505, + "learning_rate": 7.435098650051922e-06, + "loss": 0.3002, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16453662514686584, + "step": 180, + "valid_targets_mean": 7098.5, + "valid_targets_min": 5607 + }, + { + "epoch": 0.1345943979628956, + "grad_norm": 0.48205189411525795, + "learning_rate": 7.642782969885774e-06, + "loss": 0.3053, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15910585224628448, + "step": 185, + "valid_targets_mean": 6955.1, + "valid_targets_min": 5530 + }, + { + "epoch": 0.13823208439432522, + "grad_norm": 0.47699884320387564, + "learning_rate": 7.850467289719627e-06, + "loss": 0.2985, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15197807550430298, + "step": 190, + "valid_targets_mean": 6748.1, + "valid_targets_min": 5364 + }, + { + "epoch": 0.14186977082575483, + "grad_norm": 0.5450233280486808, + "learning_rate": 8.058151609553479e-06, + "loss": 0.3022, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14742602407932281, + "step": 195, + "valid_targets_mean": 6273.9, + "valid_targets_min": 5438 + }, + { + "epoch": 0.14550745725718442, + "grad_norm": 0.48683047819035635, + "learning_rate": 8.265835929387332e-06, + "loss": 0.3023, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14779230952262878, + "step": 200, + "valid_targets_mean": 6419.5, + "valid_targets_min": 4879 + }, + { + "epoch": 0.14914514368861403, + "grad_norm": 0.487101785386386, + "learning_rate": 8.473520249221184e-06, + "loss": 0.2983, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14708268642425537, + "step": 205, + "valid_targets_mean": 6483.4, + "valid_targets_min": 4309 + }, + { + "epoch": 0.15278283012004365, + "grad_norm": 0.5718615016253239, + "learning_rate": 8.681204569055037e-06, + "loss": 0.2966, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16569852828979492, + "step": 210, + "valid_targets_mean": 6825.0, + "valid_targets_min": 5205 + }, + { + "epoch": 0.15642051655147327, + "grad_norm": 0.4711270483545283, + "learning_rate": 8.888888888888888e-06, + "loss": 0.2907, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13405171036720276, + "step": 215, + "valid_targets_mean": 6729.4, + "valid_targets_min": 5200 + }, + { + "epoch": 0.16005820298290288, + "grad_norm": 0.4814235431513613, + "learning_rate": 9.096573208722742e-06, + "loss": 0.2876, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14120830595493317, + "step": 220, + "valid_targets_mean": 6094.0, + "valid_targets_min": 5302 + }, + { + "epoch": 0.1636958894143325, + "grad_norm": 0.4955432478560671, + "learning_rate": 9.304257528556595e-06, + "loss": 0.2907, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15056359767913818, + "step": 225, + "valid_targets_mean": 7140.0, + "valid_targets_min": 5270 + }, + { + "epoch": 0.16733357584576208, + "grad_norm": 0.4900996174517632, + "learning_rate": 9.511941848390446e-06, + "loss": 0.2805, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13310673832893372, + "step": 230, + "valid_targets_mean": 7425.0, + "valid_targets_min": 5281 + }, + { + "epoch": 0.1709712622771917, + "grad_norm": 0.5127503525120441, + "learning_rate": 9.7196261682243e-06, + "loss": 0.2875, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17426514625549316, + "step": 235, + "valid_targets_mean": 7063.4, + "valid_targets_min": 5386 + }, + { + "epoch": 0.17460894870862131, + "grad_norm": 0.5481570941585517, + "learning_rate": 9.927310488058153e-06, + "loss": 0.2874, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15098632872104645, + "step": 240, + "valid_targets_mean": 6827.0, + "valid_targets_min": 4547 + }, + { + "epoch": 0.17824663514005093, + "grad_norm": 1.0509221900330894, + "learning_rate": 1.0134994807892004e-05, + "loss": 0.2662, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09195338189601898, + "step": 245, + "valid_targets_mean": 2047.0, + "valid_targets_min": 165 + }, + { + "epoch": 0.18188432157148055, + "grad_norm": 0.4773264136057681, + "learning_rate": 1.0342679127725856e-05, + "loss": 0.277, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14000239968299866, + "step": 250, + "valid_targets_mean": 6754.1, + "valid_targets_min": 5196 + }, + { + "epoch": 0.18552200800291016, + "grad_norm": 0.5005214637841252, + "learning_rate": 1.055036344755971e-05, + "loss": 0.2771, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1408223807811737, + "step": 255, + "valid_targets_mean": 5985.6, + "valid_targets_min": 5028 + }, + { + "epoch": 0.18915969443433975, + "grad_norm": 0.4885534828143598, + "learning_rate": 1.0758047767393562e-05, + "loss": 0.2822, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12849347293376923, + "step": 260, + "valid_targets_mean": 5963.9, + "valid_targets_min": 5237 + }, + { + "epoch": 0.19279738086576936, + "grad_norm": 0.5229435498574028, + "learning_rate": 1.0965732087227414e-05, + "loss": 0.2765, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14443373680114746, + "step": 265, + "valid_targets_mean": 6319.0, + "valid_targets_min": 5412 + }, + { + "epoch": 0.19643506729719898, + "grad_norm": 0.4662841243733873, + "learning_rate": 1.1173416407061269e-05, + "loss": 0.2619, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11246256530284882, + "step": 270, + "valid_targets_mean": 7125.8, + "valid_targets_min": 5080 + }, + { + "epoch": 0.2000727537286286, + "grad_norm": 0.5267094500724434, + "learning_rate": 1.138110072689512e-05, + "loss": 0.2769, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13702978193759918, + "step": 275, + "valid_targets_mean": 6123.5, + "valid_targets_min": 4521 + }, + { + "epoch": 0.2037104401600582, + "grad_norm": 1.2743698253312417, + "learning_rate": 1.1588785046728972e-05, + "loss": 0.3552, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23187868297100067, + "step": 280, + "valid_targets_mean": 6020.6, + "valid_targets_min": 2493 + }, + { + "epoch": 0.20734812659148782, + "grad_norm": 0.7878574872383237, + "learning_rate": 1.1796469366562827e-05, + "loss": 0.4341, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17532524466514587, + "step": 285, + "valid_targets_mean": 3953.5, + "valid_targets_min": 1177 + }, + { + "epoch": 0.2109858130229174, + "grad_norm": 1.0492182817934315, + "learning_rate": 1.2004153686396678e-05, + "loss": 0.4303, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2508031725883484, + "step": 290, + "valid_targets_mean": 4950.8, + "valid_targets_min": 2799 + }, + { + "epoch": 0.21462349945434703, + "grad_norm": 0.7403958282580348, + "learning_rate": 1.221183800623053e-05, + "loss": 0.4072, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22774533927440643, + "step": 295, + "valid_targets_mean": 5312.0, + "valid_targets_min": 2973 + }, + { + "epoch": 0.21826118588577664, + "grad_norm": 0.6781999166327676, + "learning_rate": 1.2419522326064383e-05, + "loss": 0.4003, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.187328040599823, + "step": 300, + "valid_targets_mean": 3775.5, + "valid_targets_min": 980 + }, + { + "epoch": 0.22189887231720626, + "grad_norm": 0.654960054506166, + "learning_rate": 1.2627206645898236e-05, + "loss": 0.4184, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21314597129821777, + "step": 305, + "valid_targets_mean": 5055.4, + "valid_targets_min": 1512 + }, + { + "epoch": 0.22553655874863587, + "grad_norm": 0.6063043539641743, + "learning_rate": 1.2834890965732088e-05, + "loss": 0.3843, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14725379645824432, + "step": 310, + "valid_targets_mean": 3600.8, + "valid_targets_min": 1848 + }, + { + "epoch": 0.2291742451800655, + "grad_norm": 0.5630804909745558, + "learning_rate": 1.3042575285565941e-05, + "loss": 0.3872, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14138510823249817, + "step": 315, + "valid_targets_mean": 3369.8, + "valid_targets_min": 1381 + }, + { + "epoch": 0.23281193161149508, + "grad_norm": 0.5341982095713987, + "learning_rate": 1.3250259605399793e-05, + "loss": 0.3916, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1985149383544922, + "step": 320, + "valid_targets_mean": 6283.2, + "valid_targets_min": 3600 + }, + { + "epoch": 0.2364496180429247, + "grad_norm": 0.6089381172795486, + "learning_rate": 1.3457943925233646e-05, + "loss": 0.4061, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1991836130619049, + "step": 325, + "valid_targets_mean": 5242.5, + "valid_targets_min": 2540 + }, + { + "epoch": 0.2400873044743543, + "grad_norm": 0.5550222277309973, + "learning_rate": 1.3665628245067499e-05, + "loss": 0.3867, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17959050834178925, + "step": 330, + "valid_targets_mean": 6246.5, + "valid_targets_min": 2157 + }, + { + "epoch": 0.24372499090578392, + "grad_norm": 0.6406829208335885, + "learning_rate": 1.387331256490135e-05, + "loss": 0.3871, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19637402892112732, + "step": 335, + "valid_targets_mean": 4556.5, + "valid_targets_min": 1720 + }, + { + "epoch": 0.24736267733721354, + "grad_norm": 0.6708609125185266, + "learning_rate": 1.4080996884735202e-05, + "loss": 0.3902, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21805864572525024, + "step": 340, + "valid_targets_mean": 5116.5, + "valid_targets_min": 1982 + }, + { + "epoch": 0.25100036376864315, + "grad_norm": 0.5829199071198174, + "learning_rate": 1.4288681204569057e-05, + "loss": 0.3819, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19399169087409973, + "step": 345, + "valid_targets_mean": 4464.5, + "valid_targets_min": 2718 + }, + { + "epoch": 0.25463805020007274, + "grad_norm": 0.6403435736125584, + "learning_rate": 1.4496365524402909e-05, + "loss": 0.3963, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18693894147872925, + "step": 350, + "valid_targets_mean": 3418.5, + "valid_targets_min": 1451 + }, + { + "epoch": 0.2582757366315024, + "grad_norm": 0.7590378865159076, + "learning_rate": 1.470404984423676e-05, + "loss": 0.3935, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20007114112377167, + "step": 355, + "valid_targets_mean": 3990.4, + "valid_targets_min": 1861 + }, + { + "epoch": 0.26191342306293197, + "grad_norm": 0.7066662647388957, + "learning_rate": 1.4911734164070615e-05, + "loss": 0.3854, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2105017900466919, + "step": 360, + "valid_targets_mean": 4065.4, + "valid_targets_min": 2100 + }, + { + "epoch": 0.26555110949436156, + "grad_norm": 0.6651258151688731, + "learning_rate": 1.5119418483904467e-05, + "loss": 0.3974, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21986253559589386, + "step": 365, + "valid_targets_mean": 4132.4, + "valid_targets_min": 3234 + }, + { + "epoch": 0.2691887959257912, + "grad_norm": 1.3065856314494502, + "learning_rate": 1.5327102803738318e-05, + "loss": 0.4045, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23988209664821625, + "step": 370, + "valid_targets_mean": 4246.2, + "valid_targets_min": 2650 + }, + { + "epoch": 0.2728264823572208, + "grad_norm": 0.7060256288142579, + "learning_rate": 1.5534787123572173e-05, + "loss": 0.3914, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16371214389801025, + "step": 375, + "valid_targets_mean": 4096.5, + "valid_targets_min": 1719 + }, + { + "epoch": 0.27646416878865043, + "grad_norm": 0.6651554253551132, + "learning_rate": 1.5742471443406025e-05, + "loss": 0.3955, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19506148993968964, + "step": 380, + "valid_targets_mean": 3933.1, + "valid_targets_min": 1020 + }, + { + "epoch": 0.28010185522008, + "grad_norm": 0.6736596405466624, + "learning_rate": 1.5950155763239876e-05, + "loss": 0.3725, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1945716142654419, + "step": 385, + "valid_targets_mean": 3551.1, + "valid_targets_min": 2130 + }, + { + "epoch": 0.28373954165150966, + "grad_norm": 0.6799095287634499, + "learning_rate": 1.615784008307373e-05, + "loss": 0.3652, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1561734676361084, + "step": 390, + "valid_targets_mean": 3535.0, + "valid_targets_min": 865 + }, + { + "epoch": 0.28737722808293925, + "grad_norm": 0.8447764843878364, + "learning_rate": 1.6365524402907583e-05, + "loss": 0.3762, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25565776228904724, + "step": 395, + "valid_targets_mean": 3716.5, + "valid_targets_min": 3144 + }, + { + "epoch": 0.29101491451436884, + "grad_norm": 0.6721247738585542, + "learning_rate": 1.6573208722741434e-05, + "loss": 0.3836, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.159881591796875, + "step": 400, + "valid_targets_mean": 3262.2, + "valid_targets_min": 2050 + }, + { + "epoch": 0.2946526009457985, + "grad_norm": 0.6597714204626142, + "learning_rate": 1.6780893042575286e-05, + "loss": 0.3814, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19321578741073608, + "step": 405, + "valid_targets_mean": 3296.5, + "valid_targets_min": 1570 + }, + { + "epoch": 0.29829028737722807, + "grad_norm": 0.6510309515839291, + "learning_rate": 1.698857736240914e-05, + "loss": 0.3622, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18300297856330872, + "step": 410, + "valid_targets_mean": 4806.4, + "valid_targets_min": 1795 + }, + { + "epoch": 0.3019279738086577, + "grad_norm": 0.5780547609708435, + "learning_rate": 1.7196261682242992e-05, + "loss": 0.3657, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.167499378323555, + "step": 415, + "valid_targets_mean": 4342.4, + "valid_targets_min": 1934 + }, + { + "epoch": 0.3055656602400873, + "grad_norm": 0.7193290278682505, + "learning_rate": 1.7403946002076844e-05, + "loss": 0.3766, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19165243208408356, + "step": 420, + "valid_targets_mean": 3675.4, + "valid_targets_min": 2557 + }, + { + "epoch": 0.3092033466715169, + "grad_norm": 0.6737247399866628, + "learning_rate": 1.76116303219107e-05, + "loss": 0.372, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20562180876731873, + "step": 425, + "valid_targets_mean": 4550.4, + "valid_targets_min": 2663 + }, + { + "epoch": 0.31284103310294653, + "grad_norm": 0.6369217154076988, + "learning_rate": 1.781931464174455e-05, + "loss": 0.3608, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18274179100990295, + "step": 430, + "valid_targets_mean": 4604.0, + "valid_targets_min": 2016 + }, + { + "epoch": 0.3164787195343761, + "grad_norm": 0.6975385744880191, + "learning_rate": 1.8026998961578402e-05, + "loss": 0.3693, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.163017138838768, + "step": 435, + "valid_targets_mean": 2976.8, + "valid_targets_min": 1367 + }, + { + "epoch": 0.32011640596580576, + "grad_norm": 0.6739676206922376, + "learning_rate": 1.8234683281412257e-05, + "loss": 0.3767, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17940731346607208, + "step": 440, + "valid_targets_mean": 3542.5, + "valid_targets_min": 2026 + }, + { + "epoch": 0.32375409239723535, + "grad_norm": 0.6950435910988002, + "learning_rate": 1.8442367601246108e-05, + "loss": 0.3565, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1703200340270996, + "step": 445, + "valid_targets_mean": 3247.2, + "valid_targets_min": 2033 + }, + { + "epoch": 0.327391778828665, + "grad_norm": 0.9268192353905462, + "learning_rate": 1.865005192107996e-05, + "loss": 0.379, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18782445788383484, + "step": 450, + "valid_targets_mean": 2201.6, + "valid_targets_min": 1283 + }, + { + "epoch": 0.3310294652600946, + "grad_norm": 0.8154368393433672, + "learning_rate": 1.8857736240913815e-05, + "loss": 0.3705, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19319942593574524, + "step": 455, + "valid_targets_mean": 2824.8, + "valid_targets_min": 1636 + }, + { + "epoch": 0.33466715169152417, + "grad_norm": 0.6928376035089021, + "learning_rate": 1.9065420560747666e-05, + "loss": 0.3681, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20118200778961182, + "step": 460, + "valid_targets_mean": 3844.4, + "valid_targets_min": 1629 + }, + { + "epoch": 0.3383048381229538, + "grad_norm": 0.7501101798643189, + "learning_rate": 1.9273104880581518e-05, + "loss": 0.358, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17122358083724976, + "step": 465, + "valid_targets_mean": 3048.6, + "valid_targets_min": 2080 + }, + { + "epoch": 0.3419425245543834, + "grad_norm": 0.7062953900028942, + "learning_rate": 1.948078920041537e-05, + "loss": 0.3631, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16683681309223175, + "step": 470, + "valid_targets_mean": 3111.8, + "valid_targets_min": 1153 + }, + { + "epoch": 0.34558021098581304, + "grad_norm": 0.7209923182327331, + "learning_rate": 1.9688473520249224e-05, + "loss": 0.3563, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1946922391653061, + "step": 475, + "valid_targets_mean": 4069.0, + "valid_targets_min": 1858 + }, + { + "epoch": 0.34921789741724263, + "grad_norm": 0.6733301482789495, + "learning_rate": 1.9896157840083076e-05, + "loss": 0.3723, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1562374234199524, + "step": 480, + "valid_targets_mean": 3415.2, + "valid_targets_min": 1639 + }, + { + "epoch": 0.35285558384867227, + "grad_norm": 0.7259666395918027, + "learning_rate": 2.0103842159916927e-05, + "loss": 0.3659, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14849652349948883, + "step": 485, + "valid_targets_mean": 3267.6, + "valid_targets_min": 1394 + }, + { + "epoch": 0.35649327028010186, + "grad_norm": 0.614834965547602, + "learning_rate": 2.031152647975078e-05, + "loss": 0.3567, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18810030817985535, + "step": 490, + "valid_targets_mean": 4775.9, + "valid_targets_min": 1844 + }, + { + "epoch": 0.36013095671153145, + "grad_norm": 0.6813151960410673, + "learning_rate": 2.0519210799584634e-05, + "loss": 0.3595, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16116054356098175, + "step": 495, + "valid_targets_mean": 3317.1, + "valid_targets_min": 1305 + }, + { + "epoch": 0.3637686431429611, + "grad_norm": 0.6721115117012006, + "learning_rate": 2.0726895119418485e-05, + "loss": 0.3546, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19223836064338684, + "step": 500, + "valid_targets_mean": 4290.8, + "valid_targets_min": 2185 + }, + { + "epoch": 0.3674063295743907, + "grad_norm": 0.6602628892876996, + "learning_rate": 2.0934579439252337e-05, + "loss": 0.3624, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21113836765289307, + "step": 505, + "valid_targets_mean": 4024.9, + "valid_targets_min": 2349 + }, + { + "epoch": 0.3710440160058203, + "grad_norm": 0.6269629006917696, + "learning_rate": 2.114226375908619e-05, + "loss": 0.3442, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1578657329082489, + "step": 510, + "valid_targets_mean": 3453.0, + "valid_targets_min": 1607 + }, + { + "epoch": 0.3746817024372499, + "grad_norm": 0.7557520335817234, + "learning_rate": 2.1349948078920043e-05, + "loss": 0.3707, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18425610661506653, + "step": 515, + "valid_targets_mean": 3389.8, + "valid_targets_min": 1507 + }, + { + "epoch": 0.3783193888686795, + "grad_norm": 0.6596154715541855, + "learning_rate": 2.1557632398753895e-05, + "loss": 0.3653, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16847243905067444, + "step": 520, + "valid_targets_mean": 3773.9, + "valid_targets_min": 1317 + }, + { + "epoch": 0.38195707530010914, + "grad_norm": 0.631828828107319, + "learning_rate": 2.1765316718587747e-05, + "loss": 0.3584, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16357772052288055, + "step": 525, + "valid_targets_mean": 3881.0, + "valid_targets_min": 3355 + }, + { + "epoch": 0.3855947617315387, + "grad_norm": 0.6151428888884226, + "learning_rate": 2.19730010384216e-05, + "loss": 0.3586, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17198525369167328, + "step": 530, + "valid_targets_mean": 3951.9, + "valid_targets_min": 2999 + }, + { + "epoch": 0.38923244816296837, + "grad_norm": 0.6294592439102643, + "learning_rate": 2.2180685358255453e-05, + "loss": 0.3579, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1596847027540207, + "step": 535, + "valid_targets_mean": 4086.5, + "valid_targets_min": 1739 + }, + { + "epoch": 0.39287013459439796, + "grad_norm": 0.6592842166012814, + "learning_rate": 2.2388369678089305e-05, + "loss": 0.3579, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15654650330543518, + "step": 540, + "valid_targets_mean": 3298.0, + "valid_targets_min": 1140 + }, + { + "epoch": 0.3965078210258276, + "grad_norm": 0.6726445761715082, + "learning_rate": 2.259605399792316e-05, + "loss": 0.3364, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1854420006275177, + "step": 545, + "valid_targets_mean": 4012.1, + "valid_targets_min": 2779 + }, + { + "epoch": 0.4001455074572572, + "grad_norm": 0.6864472466732854, + "learning_rate": 2.280373831775701e-05, + "loss": 0.3485, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17132800817489624, + "step": 550, + "valid_targets_mean": 3636.9, + "valid_targets_min": 1183 + }, + { + "epoch": 0.4037831938886868, + "grad_norm": 0.6749590129761743, + "learning_rate": 2.3011422637590863e-05, + "loss": 0.3342, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20462393760681152, + "step": 555, + "valid_targets_mean": 4269.0, + "valid_targets_min": 3041 + }, + { + "epoch": 0.4074208803201164, + "grad_norm": 0.6726871133042545, + "learning_rate": 2.3219106957424717e-05, + "loss": 0.3651, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18188972771167755, + "step": 560, + "valid_targets_mean": 3350.1, + "valid_targets_min": 1726 + }, + { + "epoch": 0.411058566751546, + "grad_norm": 0.5616952274598451, + "learning_rate": 2.3426791277258566e-05, + "loss": 0.3427, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15866096317768097, + "step": 565, + "valid_targets_mean": 4747.4, + "valid_targets_min": 1743 + }, + { + "epoch": 0.41469625318297565, + "grad_norm": 0.6505568616871292, + "learning_rate": 2.363447559709242e-05, + "loss": 0.3557, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17140597105026245, + "step": 570, + "valid_targets_mean": 3702.8, + "valid_targets_min": 1234 + }, + { + "epoch": 0.41833393961440524, + "grad_norm": 0.658503716659714, + "learning_rate": 2.3842159916926275e-05, + "loss": 0.3613, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18298622965812683, + "step": 575, + "valid_targets_mean": 4079.0, + "valid_targets_min": 2303 + }, + { + "epoch": 0.4219716260458348, + "grad_norm": 0.6506793067249006, + "learning_rate": 2.4049844236760124e-05, + "loss": 0.3611, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1794300377368927, + "step": 580, + "valid_targets_mean": 3830.8, + "valid_targets_min": 1651 + }, + { + "epoch": 0.42560931247726447, + "grad_norm": 0.7119657228568842, + "learning_rate": 2.425752855659398e-05, + "loss": 0.3519, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1880340278148651, + "step": 585, + "valid_targets_mean": 2876.2, + "valid_targets_min": 1374 + }, + { + "epoch": 0.42924699890869406, + "grad_norm": 0.9889690365229891, + "learning_rate": 2.4465212876427833e-05, + "loss": 0.3423, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17069831490516663, + "step": 590, + "valid_targets_mean": 3467.5, + "valid_targets_min": 1625 + }, + { + "epoch": 0.4328846853401237, + "grad_norm": 0.6701266335927107, + "learning_rate": 2.467289719626168e-05, + "loss": 0.3376, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1720227599143982, + "step": 595, + "valid_targets_mean": 4089.5, + "valid_targets_min": 997 + }, + { + "epoch": 0.4365223717715533, + "grad_norm": 0.5906633426443267, + "learning_rate": 2.4880581516095537e-05, + "loss": 0.3432, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14403122663497925, + "step": 600, + "valid_targets_mean": 3618.6, + "valid_targets_min": 1179 + }, + { + "epoch": 0.44016005820298293, + "grad_norm": 0.6604444389119205, + "learning_rate": 2.508826583592939e-05, + "loss": 0.3416, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16337966918945312, + "step": 605, + "valid_targets_mean": 3079.1, + "valid_targets_min": 1444 + }, + { + "epoch": 0.4437977446344125, + "grad_norm": 0.6286995452903261, + "learning_rate": 2.529595015576324e-05, + "loss": 0.3238, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14801639318466187, + "step": 610, + "valid_targets_mean": 3486.8, + "valid_targets_min": 1504 + }, + { + "epoch": 0.4474354310658421, + "grad_norm": 0.6710649461691846, + "learning_rate": 2.5503634475597095e-05, + "loss": 0.3368, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1698751598596573, + "step": 615, + "valid_targets_mean": 3998.4, + "valid_targets_min": 2302 + }, + { + "epoch": 0.45107311749727175, + "grad_norm": 0.7352765744124314, + "learning_rate": 2.571131879543095e-05, + "loss": 0.3465, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2148081660270691, + "step": 620, + "valid_targets_mean": 4282.5, + "valid_targets_min": 1818 + }, + { + "epoch": 0.45471080392870133, + "grad_norm": 0.818778736477047, + "learning_rate": 2.5919003115264798e-05, + "loss": 0.3688, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18084532022476196, + "step": 625, + "valid_targets_mean": 4221.9, + "valid_targets_min": 2633 + }, + { + "epoch": 0.458348490360131, + "grad_norm": 0.654379056249541, + "learning_rate": 2.6126687435098653e-05, + "loss": 0.3405, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17494402825832367, + "step": 630, + "valid_targets_mean": 3522.4, + "valid_targets_min": 2481 + }, + { + "epoch": 0.46198617679156057, + "grad_norm": 0.6486974768205178, + "learning_rate": 2.6334371754932508e-05, + "loss": 0.3537, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1898401826620102, + "step": 635, + "valid_targets_mean": 4058.8, + "valid_targets_min": 2013 + }, + { + "epoch": 0.46562386322299015, + "grad_norm": 0.7055990725541452, + "learning_rate": 2.6542056074766356e-05, + "loss": 0.3343, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18166470527648926, + "step": 640, + "valid_targets_mean": 3723.5, + "valid_targets_min": 1744 + }, + { + "epoch": 0.4692615496544198, + "grad_norm": 0.6326145471807061, + "learning_rate": 2.674974039460021e-05, + "loss": 0.3491, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18266236782073975, + "step": 645, + "valid_targets_mean": 4180.9, + "valid_targets_min": 1914 + }, + { + "epoch": 0.4728992360858494, + "grad_norm": 0.6774772263831336, + "learning_rate": 2.6957424714434066e-05, + "loss": 0.3337, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1472262442111969, + "step": 650, + "valid_targets_mean": 3668.6, + "valid_targets_min": 3065 + }, + { + "epoch": 0.476536922517279, + "grad_norm": 0.6250279861017435, + "learning_rate": 2.7165109034267914e-05, + "loss": 0.3241, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19464290142059326, + "step": 655, + "valid_targets_mean": 4473.2, + "valid_targets_min": 2894 + }, + { + "epoch": 0.4801746089487086, + "grad_norm": 0.6553873614963319, + "learning_rate": 2.737279335410177e-05, + "loss": 0.3506, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17545117437839508, + "step": 660, + "valid_targets_mean": 3395.6, + "valid_targets_min": 2359 + }, + { + "epoch": 0.48381229538013826, + "grad_norm": 0.7011305089713047, + "learning_rate": 2.758047767393562e-05, + "loss": 0.3368, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15108759701251984, + "step": 665, + "valid_targets_mean": 3114.5, + "valid_targets_min": 1463 + }, + { + "epoch": 0.48744998181156785, + "grad_norm": 0.6564714910343536, + "learning_rate": 2.778816199376947e-05, + "loss": 0.3496, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17418327927589417, + "step": 670, + "valid_targets_mean": 3588.1, + "valid_targets_min": 1426 + }, + { + "epoch": 0.49108766824299743, + "grad_norm": 0.6781327135126739, + "learning_rate": 2.7995846313603327e-05, + "loss": 0.3541, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17138399183750153, + "step": 675, + "valid_targets_mean": 3405.5, + "valid_targets_min": 1661 + }, + { + "epoch": 0.4947253546744271, + "grad_norm": 0.6027964594106526, + "learning_rate": 2.8203530633437178e-05, + "loss": 0.3439, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17667940258979797, + "step": 680, + "valid_targets_mean": 4457.8, + "valid_targets_min": 1529 + }, + { + "epoch": 0.49836304110585666, + "grad_norm": 0.8770150593730013, + "learning_rate": 2.841121495327103e-05, + "loss": 0.3772, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20397979021072388, + "step": 685, + "valid_targets_mean": 3272.9, + "valid_targets_min": 1450 + }, + { + "epoch": 0.5020007275372863, + "grad_norm": 0.6992872916258924, + "learning_rate": 2.8618899273104885e-05, + "loss": 0.3426, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15268027782440186, + "step": 690, + "valid_targets_mean": 3103.2, + "valid_targets_min": 1625 + }, + { + "epoch": 0.5056384139687159, + "grad_norm": 0.6217752656935459, + "learning_rate": 2.8826583592938736e-05, + "loss": 0.3501, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14535976946353912, + "step": 695, + "valid_targets_mean": 3952.2, + "valid_targets_min": 1174 + }, + { + "epoch": 0.5092761004001455, + "grad_norm": 0.6592984274992326, + "learning_rate": 2.9034267912772588e-05, + "loss": 0.3536, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18245969712734222, + "step": 700, + "valid_targets_mean": 4039.2, + "valid_targets_min": 2882 + }, + { + "epoch": 0.5129137868315751, + "grad_norm": 0.6073994513610984, + "learning_rate": 2.924195223260644e-05, + "loss": 0.356, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19549459218978882, + "step": 705, + "valid_targets_mean": 4594.0, + "valid_targets_min": 1946 + }, + { + "epoch": 0.5165514732630048, + "grad_norm": 0.6271238612973805, + "learning_rate": 2.9449636552440294e-05, + "loss": 0.346, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16739840805530548, + "step": 710, + "valid_targets_mean": 4082.6, + "valid_targets_min": 2000 + }, + { + "epoch": 0.5201891596944344, + "grad_norm": 0.6875695111008628, + "learning_rate": 2.9657320872274146e-05, + "loss": 0.3529, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18824011087417603, + "step": 715, + "valid_targets_mean": 4596.1, + "valid_targets_min": 2012 + }, + { + "epoch": 0.5238268461258639, + "grad_norm": 0.6542698134393353, + "learning_rate": 2.9865005192107997e-05, + "loss": 0.352, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1775708645582199, + "step": 720, + "valid_targets_mean": 3094.4, + "valid_targets_min": 719 + }, + { + "epoch": 0.5274645325572935, + "grad_norm": 0.6499498471243781, + "learning_rate": 3.0072689511941852e-05, + "loss": 0.3456, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1570858210325241, + "step": 725, + "valid_targets_mean": 4496.2, + "valid_targets_min": 3238 + }, + { + "epoch": 0.5311022189887231, + "grad_norm": 0.6446894679654944, + "learning_rate": 3.0280373831775704e-05, + "loss": 0.3307, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1490686535835266, + "step": 730, + "valid_targets_mean": 3816.8, + "valid_targets_min": 1372 + }, + { + "epoch": 0.5347399054201528, + "grad_norm": 0.7027666596169023, + "learning_rate": 3.0488058151609555e-05, + "loss": 0.3275, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1820855587720871, + "step": 735, + "valid_targets_mean": 3306.1, + "valid_targets_min": 2370 + }, + { + "epoch": 0.5383775918515824, + "grad_norm": 0.6118395941265976, + "learning_rate": 3.0695742471443414e-05, + "loss": 0.3393, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1468178778886795, + "step": 740, + "valid_targets_mean": 2958.1, + "valid_targets_min": 1576 + }, + { + "epoch": 0.542015278283012, + "grad_norm": 0.6599970439072205, + "learning_rate": 3.090342679127726e-05, + "loss": 0.3321, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1833878755569458, + "step": 745, + "valid_targets_mean": 3758.5, + "valid_targets_min": 1426 + }, + { + "epoch": 0.5456529647144416, + "grad_norm": 0.6618531214058636, + "learning_rate": 3.111111111111112e-05, + "loss": 0.3475, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15207602083683014, + "step": 750, + "valid_targets_mean": 3117.0, + "valid_targets_min": 1728 + }, + { + "epoch": 0.5492906511458713, + "grad_norm": 0.7703410978822947, + "learning_rate": 3.131879543094497e-05, + "loss": 0.3415, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18210090696811676, + "step": 755, + "valid_targets_mean": 3264.2, + "valid_targets_min": 1362 + }, + { + "epoch": 0.5529283375773009, + "grad_norm": 1.1808699073502564, + "learning_rate": 3.152647975077882e-05, + "loss": 0.3379, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.159324049949646, + "step": 760, + "valid_targets_mean": 3836.0, + "valid_targets_min": 852 + }, + { + "epoch": 0.5565660240087305, + "grad_norm": 0.7865572791780513, + "learning_rate": 3.173416407061267e-05, + "loss": 0.3478, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21431298553943634, + "step": 765, + "valid_targets_mean": 2764.1, + "valid_targets_min": 1195 + }, + { + "epoch": 0.56020371044016, + "grad_norm": 0.5888515535488492, + "learning_rate": 3.194184839044652e-05, + "loss": 0.3335, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17824387550354004, + "step": 770, + "valid_targets_mean": 4457.5, + "valid_targets_min": 1564 + }, + { + "epoch": 0.5638413968715896, + "grad_norm": 0.5460407955285632, + "learning_rate": 3.2149532710280374e-05, + "loss": 0.3277, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1395266354084015, + "step": 775, + "valid_targets_mean": 4371.2, + "valid_targets_min": 1975 + }, + { + "epoch": 0.5674790833030193, + "grad_norm": 0.6070915383944535, + "learning_rate": 3.235721703011423e-05, + "loss": 0.3201, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16940094530582428, + "step": 780, + "valid_targets_mean": 3965.8, + "valid_targets_min": 2426 + }, + { + "epoch": 0.5711167697344489, + "grad_norm": 0.5112623547139409, + "learning_rate": 3.2564901349948084e-05, + "loss": 0.3231, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16462098062038422, + "step": 785, + "valid_targets_mean": 6384.6, + "valid_targets_min": 2678 + }, + { + "epoch": 0.5747544561658785, + "grad_norm": 0.6046531976045818, + "learning_rate": 3.2772585669781936e-05, + "loss": 0.3346, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15996506810188293, + "step": 790, + "valid_targets_mean": 3543.2, + "valid_targets_min": 2826 + }, + { + "epoch": 0.5783921425973081, + "grad_norm": 0.6353173044613137, + "learning_rate": 3.298026998961579e-05, + "loss": 0.3205, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13481420278549194, + "step": 795, + "valid_targets_mean": 3155.8, + "valid_targets_min": 993 + }, + { + "epoch": 0.5820298290287377, + "grad_norm": 0.6985419850939057, + "learning_rate": 3.318795430944964e-05, + "loss": 0.3288, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1841023564338684, + "step": 800, + "valid_targets_mean": 4166.5, + "valid_targets_min": 2724 + }, + { + "epoch": 0.5856675154601674, + "grad_norm": 0.6731159567568743, + "learning_rate": 3.339563862928349e-05, + "loss": 0.329, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1551065742969513, + "step": 805, + "valid_targets_mean": 3079.5, + "valid_targets_min": 1360 + }, + { + "epoch": 0.589305201891597, + "grad_norm": 0.650556155380333, + "learning_rate": 3.360332294911734e-05, + "loss": 0.323, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15804484486579895, + "step": 810, + "valid_targets_mean": 4114.9, + "valid_targets_min": 2414 + }, + { + "epoch": 0.5929428883230266, + "grad_norm": 0.6183454826723159, + "learning_rate": 3.3811007268951193e-05, + "loss": 0.3488, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18963412940502167, + "step": 815, + "valid_targets_mean": 4524.8, + "valid_targets_min": 2425 + }, + { + "epoch": 0.5965805747544561, + "grad_norm": 0.699635594934599, + "learning_rate": 3.401869158878505e-05, + "loss": 0.3429, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17211580276489258, + "step": 820, + "valid_targets_mean": 3035.2, + "valid_targets_min": 1720 + }, + { + "epoch": 0.6002182611858857, + "grad_norm": 0.6849642086543869, + "learning_rate": 3.42263759086189e-05, + "loss": 0.3369, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16946591436862946, + "step": 825, + "valid_targets_mean": 3260.4, + "valid_targets_min": 1621 + }, + { + "epoch": 0.6038559476173154, + "grad_norm": 0.7831186209821384, + "learning_rate": 3.4434060228452755e-05, + "loss": 0.318, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17170177400112152, + "step": 830, + "valid_targets_mean": 3214.6, + "valid_targets_min": 1858 + }, + { + "epoch": 0.607493634048745, + "grad_norm": 0.6596492106534497, + "learning_rate": 3.4641744548286606e-05, + "loss": 0.3353, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15407197177410126, + "step": 835, + "valid_targets_mean": 3254.2, + "valid_targets_min": 1254 + }, + { + "epoch": 0.6111313204801746, + "grad_norm": 0.647909658024069, + "learning_rate": 3.484942886812046e-05, + "loss": 0.3241, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17729249596595764, + "step": 840, + "valid_targets_mean": 2994.4, + "valid_targets_min": 1039 + }, + { + "epoch": 0.6147690069116042, + "grad_norm": 0.6488204596481462, + "learning_rate": 3.505711318795431e-05, + "loss": 0.3429, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20890367031097412, + "step": 845, + "valid_targets_mean": 4625.4, + "valid_targets_min": 2526 + }, + { + "epoch": 0.6184066933430338, + "grad_norm": 0.6124217740362682, + "learning_rate": 3.526479750778816e-05, + "loss": 0.3301, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1785414069890976, + "step": 850, + "valid_targets_mean": 3571.8, + "valid_targets_min": 2409 + }, + { + "epoch": 0.6220443797744635, + "grad_norm": 0.6390456663286358, + "learning_rate": 3.547248182762202e-05, + "loss": 0.3334, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14413990080356598, + "step": 855, + "valid_targets_mean": 3691.5, + "valid_targets_min": 2918 + }, + { + "epoch": 0.6256820662058931, + "grad_norm": 0.6571469419852908, + "learning_rate": 3.568016614745587e-05, + "loss": 0.3376, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16710662841796875, + "step": 860, + "valid_targets_mean": 4063.1, + "valid_targets_min": 2179 + }, + { + "epoch": 0.6293197526373226, + "grad_norm": 0.8500621443652965, + "learning_rate": 3.588785046728972e-05, + "loss": 0.305, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10500035434961319, + "step": 865, + "valid_targets_mean": 2871.4, + "valid_targets_min": 955 + }, + { + "epoch": 0.6329574390687522, + "grad_norm": 0.6819170042753823, + "learning_rate": 3.6095534787123574e-05, + "loss": 0.2108, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07898104935884476, + "step": 870, + "valid_targets_mean": 1788.9, + "valid_targets_min": 669 + }, + { + "epoch": 0.6365951255001819, + "grad_norm": 0.5241319604378883, + "learning_rate": 3.6303219106957426e-05, + "loss": 0.1561, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06921590864658356, + "step": 875, + "valid_targets_mean": 3755.8, + "valid_targets_min": 2651 + }, + { + "epoch": 0.6402328119316115, + "grad_norm": 0.5455835808152676, + "learning_rate": 3.651090342679128e-05, + "loss": 0.1515, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08156849443912506, + "step": 880, + "valid_targets_mean": 3069.8, + "valid_targets_min": 710 + }, + { + "epoch": 0.6438704983630411, + "grad_norm": 0.5114349499247324, + "learning_rate": 3.6718587746625135e-05, + "loss": 0.1664, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.049888040870428085, + "step": 885, + "valid_targets_mean": 3618.5, + "valid_targets_min": 1365 + }, + { + "epoch": 0.6475081847944707, + "grad_norm": 0.5467282282638982, + "learning_rate": 3.692627206645899e-05, + "loss": 0.1476, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08316117525100708, + "step": 890, + "valid_targets_mean": 3279.9, + "valid_targets_min": 889 + }, + { + "epoch": 0.6511458712259003, + "grad_norm": 0.9895205763672333, + "learning_rate": 3.713395638629284e-05, + "loss": 0.1999, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11044557392597198, + "step": 895, + "valid_targets_mean": 1652.4, + "valid_targets_min": 559 + }, + { + "epoch": 0.65478355765733, + "grad_norm": 0.48965152370178944, + "learning_rate": 3.734164070612669e-05, + "loss": 0.1965, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06284298002719879, + "step": 900, + "valid_targets_mean": 3449.0, + "valid_targets_min": 1013 + }, + { + "epoch": 0.6584212440887596, + "grad_norm": 0.52733837479069, + "learning_rate": 3.754932502596054e-05, + "loss": 0.1625, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09367058426141739, + "step": 905, + "valid_targets_mean": 3726.1, + "valid_targets_min": 3222 + }, + { + "epoch": 0.6620589305201892, + "grad_norm": 0.7384899729113866, + "learning_rate": 3.775700934579439e-05, + "loss": 0.2046, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14074072241783142, + "step": 910, + "valid_targets_mean": 3205.8, + "valid_targets_min": 1949 + }, + { + "epoch": 0.6656966169516187, + "grad_norm": 0.3867350839543693, + "learning_rate": 3.796469366562825e-05, + "loss": 0.1228, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05376336723566055, + "step": 915, + "valid_targets_mean": 4059.8, + "valid_targets_min": 886 + }, + { + "epoch": 0.6693343033830483, + "grad_norm": 0.6644887518422271, + "learning_rate": 3.8172377985462096e-05, + "loss": 0.1468, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07888832688331604, + "step": 920, + "valid_targets_mean": 1858.2, + "valid_targets_min": 447 + }, + { + "epoch": 0.672971989814478, + "grad_norm": 0.49697054254571144, + "learning_rate": 3.8380062305295954e-05, + "loss": 0.1395, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05795615166425705, + "step": 925, + "valid_targets_mean": 3195.6, + "valid_targets_min": 758 + }, + { + "epoch": 0.6766096762459076, + "grad_norm": 0.6313662108831117, + "learning_rate": 3.8587746625129806e-05, + "loss": 0.1695, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09167777746915817, + "step": 930, + "valid_targets_mean": 1808.4, + "valid_targets_min": 604 + }, + { + "epoch": 0.6802473626773372, + "grad_norm": 0.5054261462340747, + "learning_rate": 3.879543094496366e-05, + "loss": 0.1636, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07724280655384064, + "step": 935, + "valid_targets_mean": 3314.6, + "valid_targets_min": 1743 + }, + { + "epoch": 0.6838850491087668, + "grad_norm": 0.5361016519204379, + "learning_rate": 3.900311526479751e-05, + "loss": 0.1421, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05760093405842781, + "step": 940, + "valid_targets_mean": 2440.4, + "valid_targets_min": 574 + }, + { + "epoch": 0.6875227355401964, + "grad_norm": 0.524120453457823, + "learning_rate": 3.921079958463137e-05, + "loss": 0.1381, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06556359678506851, + "step": 945, + "valid_targets_mean": 2535.6, + "valid_targets_min": 523 + }, + { + "epoch": 0.6911604219716261, + "grad_norm": 0.6425982922061564, + "learning_rate": 3.941848390446521e-05, + "loss": 0.1336, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10343382507562637, + "step": 950, + "valid_targets_mean": 3398.1, + "valid_targets_min": 2635 + }, + { + "epoch": 0.6947981084030557, + "grad_norm": 0.5437181054674852, + "learning_rate": 3.962616822429907e-05, + "loss": 0.1528, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06620605289936066, + "step": 955, + "valid_targets_mean": 3164.8, + "valid_targets_min": 812 + }, + { + "epoch": 0.6984357948344853, + "grad_norm": 0.5211338351014337, + "learning_rate": 3.983385254413292e-05, + "loss": 0.1545, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07680831104516983, + "step": 960, + "valid_targets_mean": 2463.9, + "valid_targets_min": 856 + }, + { + "epoch": 0.7020734812659148, + "grad_norm": 0.5770858311821477, + "learning_rate": 3.999999868458321e-05, + "loss": 0.1573, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09806957840919495, + "step": 965, + "valid_targets_mean": 2995.6, + "valid_targets_min": 718 + }, + { + "epoch": 0.7057111676973445, + "grad_norm": 0.707767060354504, + "learning_rate": 3.999995264501363e-05, + "loss": 0.1789, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18426083028316498, + "step": 970, + "valid_targets_mean": 2866.9, + "valid_targets_min": 1297 + }, + { + "epoch": 0.7093488541287741, + "grad_norm": 0.5997591195662267, + "learning_rate": 3.9999840834777436e-05, + "loss": 0.1698, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06399205327033997, + "step": 975, + "valid_targets_mean": 1661.1, + "valid_targets_min": 786 + }, + { + "epoch": 0.7129865405602037, + "grad_norm": 0.7541063063954828, + "learning_rate": 3.999966325424233e-05, + "loss": 0.1384, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09625499695539474, + "step": 980, + "valid_targets_mean": 1624.2, + "valid_targets_min": 550 + }, + { + "epoch": 0.7166242269916333, + "grad_norm": 0.4981481118994374, + "learning_rate": 3.999941990399228e-05, + "loss": 0.1861, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06664615124464035, + "step": 985, + "valid_targets_mean": 3254.6, + "valid_targets_min": 1304 + }, + { + "epoch": 0.7202619134230629, + "grad_norm": 0.5082209781737663, + "learning_rate": 3.999911078482757e-05, + "loss": 0.1688, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06212037429213524, + "step": 990, + "valid_targets_mean": 2833.0, + "valid_targets_min": 1247 + }, + { + "epoch": 0.7238995998544926, + "grad_norm": 0.6362456908707844, + "learning_rate": 3.9998735897764735e-05, + "loss": 0.143, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0882577896118164, + "step": 995, + "valid_targets_mean": 3056.5, + "valid_targets_min": 800 + }, + { + "epoch": 0.7275372862859222, + "grad_norm": 0.6768531713678688, + "learning_rate": 3.9998295244036615e-05, + "loss": 0.366, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15720057487487793, + "step": 1000, + "valid_targets_mean": 2580.8, + "valid_targets_min": 987 + }, + { + "epoch": 0.7311749727173518, + "grad_norm": 0.7578938555994541, + "learning_rate": 3.999778882509231e-05, + "loss": 0.1725, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14058971405029297, + "step": 1005, + "valid_targets_mean": 2245.6, + "valid_targets_min": 776 + }, + { + "epoch": 0.7348126591487814, + "grad_norm": 0.526799606713011, + "learning_rate": 3.999721664259721e-05, + "loss": 0.1459, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04046333581209183, + "step": 1010, + "valid_targets_mean": 1361.5, + "valid_targets_min": 658 + }, + { + "epoch": 0.7384503455802109, + "grad_norm": 0.561411684766126, + "learning_rate": 3.999657869843295e-05, + "loss": 0.261, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06742563843727112, + "step": 1015, + "valid_targets_mean": 1729.0, + "valid_targets_min": 666 + }, + { + "epoch": 0.7420880320116406, + "grad_norm": 0.38665629015674247, + "learning_rate": 3.999587499469746e-05, + "loss": 0.128, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05234804004430771, + "step": 1020, + "valid_targets_mean": 4975.0, + "valid_targets_min": 3529 + }, + { + "epoch": 0.7457257184430702, + "grad_norm": 0.4903464899873592, + "learning_rate": 3.999510553370487e-05, + "loss": 0.1423, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08424920588731766, + "step": 1025, + "valid_targets_mean": 3988.8, + "valid_targets_min": 1746 + }, + { + "epoch": 0.7493634048744998, + "grad_norm": 0.49242112343888367, + "learning_rate": 3.999427031798558e-05, + "loss": 0.1416, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09016582369804382, + "step": 1030, + "valid_targets_mean": 3222.4, + "valid_targets_min": 1226 + }, + { + "epoch": 0.7530010913059294, + "grad_norm": 0.4873912328145889, + "learning_rate": 3.9993369350286265e-05, + "loss": 0.132, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08642400801181793, + "step": 1035, + "valid_targets_mean": 4403.6, + "valid_targets_min": 3275 + }, + { + "epoch": 0.756638777737359, + "grad_norm": 0.42298209343103066, + "learning_rate": 3.999240263356977e-05, + "loss": 0.133, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0537312850356102, + "step": 1040, + "valid_targets_mean": 2546.2, + "valid_targets_min": 723 + }, + { + "epoch": 0.7602764641687887, + "grad_norm": 0.47351481128553036, + "learning_rate": 3.999137017101518e-05, + "loss": 0.1303, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07549460977315903, + "step": 1045, + "valid_targets_mean": 3348.6, + "valid_targets_min": 2234 + }, + { + "epoch": 0.7639141506002183, + "grad_norm": 0.3734975711878738, + "learning_rate": 3.999027196601781e-05, + "loss": 0.1242, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0408632829785347, + "step": 1050, + "valid_targets_mean": 3603.0, + "valid_targets_min": 1421 + }, + { + "epoch": 0.7675518370316479, + "grad_norm": 0.4851218541891473, + "learning_rate": 3.9989108022189134e-05, + "loss": 0.1291, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06940672546625137, + "step": 1055, + "valid_targets_mean": 2849.2, + "valid_targets_min": 1068 + }, + { + "epoch": 0.7711895234630775, + "grad_norm": 0.6064867710868221, + "learning_rate": 3.998787834335684e-05, + "loss": 0.1871, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05507325381040573, + "step": 1060, + "valid_targets_mean": 1587.9, + "valid_targets_min": 535 + }, + { + "epoch": 0.774827209894507, + "grad_norm": 0.45612441596919534, + "learning_rate": 3.998658293356477e-05, + "loss": 0.1137, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047272443771362305, + "step": 1065, + "valid_targets_mean": 1702.4, + "valid_targets_min": 585 + }, + { + "epoch": 0.7784648963259367, + "grad_norm": 0.4813491976947789, + "learning_rate": 3.9985221797072934e-05, + "loss": 0.127, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.061851028352975845, + "step": 1070, + "valid_targets_mean": 2588.6, + "valid_targets_min": 582 + }, + { + "epoch": 0.7821025827573663, + "grad_norm": 0.4636991824036082, + "learning_rate": 3.998379493835749e-05, + "loss": 0.1185, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06604774296283722, + "step": 1075, + "valid_targets_mean": 3144.0, + "valid_targets_min": 516 + }, + { + "epoch": 0.7857402691887959, + "grad_norm": 0.5319987561963951, + "learning_rate": 3.9982302362110716e-05, + "loss": 0.1881, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10227231681346893, + "step": 1080, + "valid_targets_mean": 3241.4, + "valid_targets_min": 999 + }, + { + "epoch": 0.7893779556202255, + "grad_norm": 0.5465936392812899, + "learning_rate": 3.9980744073241016e-05, + "loss": 0.1442, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.115080825984478, + "step": 1085, + "valid_targets_mean": 2021.1, + "valid_targets_min": 696 + }, + { + "epoch": 0.7930156420516552, + "grad_norm": 0.5442781696701228, + "learning_rate": 3.997912007687288e-05, + "loss": 0.168, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07837756723165512, + "step": 1090, + "valid_targets_mean": 2048.5, + "valid_targets_min": 635 + }, + { + "epoch": 0.7966533284830848, + "grad_norm": 0.4503261547255184, + "learning_rate": 3.997743037834689e-05, + "loss": 0.1451, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06273740530014038, + "step": 1095, + "valid_targets_mean": 2641.6, + "valid_targets_min": 816 + }, + { + "epoch": 0.8002910149145144, + "grad_norm": 0.6573834322340647, + "learning_rate": 3.99756749832197e-05, + "loss": 0.2058, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19539469480514526, + "step": 1100, + "valid_targets_mean": 2236.4, + "valid_targets_min": 949 + }, + { + "epoch": 0.803928701345944, + "grad_norm": 0.4925063324558213, + "learning_rate": 3.997385389726398e-05, + "loss": 0.1516, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07029861211776733, + "step": 1105, + "valid_targets_mean": 3270.9, + "valid_targets_min": 2201 + }, + { + "epoch": 0.8075663877773736, + "grad_norm": 0.9167476842224604, + "learning_rate": 3.997196712646845e-05, + "loss": 0.1707, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1203790158033371, + "step": 1110, + "valid_targets_mean": 1500.5, + "valid_targets_min": 687 + }, + { + "epoch": 0.8112040742088032, + "grad_norm": 0.5364903844495562, + "learning_rate": 3.997001467703784e-05, + "loss": 0.1347, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05906068533658981, + "step": 1115, + "valid_targets_mean": 2285.5, + "valid_targets_min": 449 + }, + { + "epoch": 0.8148417606402328, + "grad_norm": 0.4512273019277775, + "learning_rate": 3.9967996555392866e-05, + "loss": 0.1422, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.038344357162714005, + "step": 1120, + "valid_targets_mean": 2368.1, + "valid_targets_min": 658 + }, + { + "epoch": 0.8184794470716624, + "grad_norm": 0.731261953620943, + "learning_rate": 3.996591276817019e-05, + "loss": 0.1432, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05045781284570694, + "step": 1125, + "valid_targets_mean": 831.8, + "valid_targets_min": 516 + }, + { + "epoch": 0.822117133503092, + "grad_norm": 0.39035177265462845, + "learning_rate": 3.9963763322222443e-05, + "loss": 0.1156, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07647136598825455, + "step": 1130, + "valid_targets_mean": 4461.9, + "valid_targets_min": 2805 + }, + { + "epoch": 0.8257548199345216, + "grad_norm": 0.36863951255222555, + "learning_rate": 3.9961548224618153e-05, + "loss": 0.1112, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06329832971096039, + "step": 1135, + "valid_targets_mean": 3766.8, + "valid_targets_min": 2036 + }, + { + "epoch": 0.8293925063659513, + "grad_norm": 0.4263554525843365, + "learning_rate": 3.995926748264178e-05, + "loss": 0.1274, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.054403163492679596, + "step": 1140, + "valid_targets_mean": 2336.5, + "valid_targets_min": 681 + }, + { + "epoch": 0.8330301927973809, + "grad_norm": 0.37611880542730786, + "learning_rate": 3.995692110379362e-05, + "loss": 0.1495, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07528260350227356, + "step": 1145, + "valid_targets_mean": 3856.1, + "valid_targets_min": 3436 + }, + { + "epoch": 0.8366678792288105, + "grad_norm": 0.5447639148536129, + "learning_rate": 3.995450909578984e-05, + "loss": 0.1441, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09246468544006348, + "step": 1150, + "valid_targets_mean": 3165.5, + "valid_targets_min": 891 + }, + { + "epoch": 0.8403055656602401, + "grad_norm": 0.5423260475083603, + "learning_rate": 3.995203146656243e-05, + "loss": 0.1272, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0432354137301445, + "step": 1155, + "valid_targets_mean": 3550.6, + "valid_targets_min": 2573 + }, + { + "epoch": 0.8439432520916696, + "grad_norm": 0.37135457914180736, + "learning_rate": 3.994948822425918e-05, + "loss": 0.1216, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05804494023323059, + "step": 1160, + "valid_targets_mean": 2995.0, + "valid_targets_min": 928 + }, + { + "epoch": 0.8475809385230993, + "grad_norm": 0.48293333067300975, + "learning_rate": 3.9946879377243644e-05, + "loss": 0.1513, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08472256362438202, + "step": 1165, + "valid_targets_mean": 3839.6, + "valid_targets_min": 3392 + }, + { + "epoch": 0.8512186249545289, + "grad_norm": 0.32288572073710575, + "learning_rate": 3.9944204934095114e-05, + "loss": 0.1125, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06220956891775131, + "step": 1170, + "valid_targets_mean": 3855.0, + "valid_targets_min": 2848 + }, + { + "epoch": 0.8548563113859585, + "grad_norm": 0.5290631409005148, + "learning_rate": 3.9941464903608614e-05, + "loss": 0.1377, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07234556972980499, + "step": 1175, + "valid_targets_mean": 2586.0, + "valid_targets_min": 595 + }, + { + "epoch": 0.8584939978173881, + "grad_norm": 0.5046620975921794, + "learning_rate": 3.9938659294794854e-05, + "loss": 0.1428, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06970896571874619, + "step": 1180, + "valid_targets_mean": 2471.8, + "valid_targets_min": 1016 + }, + { + "epoch": 0.8621316842488177, + "grad_norm": 0.5804700197894945, + "learning_rate": 3.993578811688018e-05, + "loss": 0.1434, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10302190482616425, + "step": 1185, + "valid_targets_mean": 1567.8, + "valid_targets_min": 811 + }, + { + "epoch": 0.8657693706802474, + "grad_norm": 0.7420342831046403, + "learning_rate": 3.993285137930658e-05, + "loss": 0.1486, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0862768292427063, + "step": 1190, + "valid_targets_mean": 1569.1, + "valid_targets_min": 697 + }, + { + "epoch": 0.869407057111677, + "grad_norm": 0.7396894939916088, + "learning_rate": 3.992984909173165e-05, + "loss": 0.1344, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10438710451126099, + "step": 1195, + "valid_targets_mean": 1448.1, + "valid_targets_min": 736 + }, + { + "epoch": 0.8730447435431066, + "grad_norm": 0.44101170902357395, + "learning_rate": 3.9926781264028525e-05, + "loss": 0.1364, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04279227554798126, + "step": 1200, + "valid_targets_mean": 2415.5, + "valid_targets_min": 804 + }, + { + "epoch": 0.8766824299745362, + "grad_norm": 0.47102936237202425, + "learning_rate": 3.9923647906285884e-05, + "loss": 0.152, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.053733088076114655, + "step": 1205, + "valid_targets_mean": 3472.5, + "valid_targets_min": 1018 + }, + { + "epoch": 0.8803201164059659, + "grad_norm": 0.4027109428506898, + "learning_rate": 3.99204490288079e-05, + "loss": 0.1311, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06069178134202957, + "step": 1210, + "valid_targets_mean": 3623.8, + "valid_targets_min": 2369 + }, + { + "epoch": 0.8839578028373954, + "grad_norm": 0.4953922927393119, + "learning_rate": 3.991718464211421e-05, + "loss": 0.1191, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06853586435317993, + "step": 1215, + "valid_targets_mean": 3135.8, + "valid_targets_min": 1318 + }, + { + "epoch": 0.887595489268825, + "grad_norm": 0.4938275178628361, + "learning_rate": 3.9913854756939906e-05, + "loss": 0.1429, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09267303347587585, + "step": 1220, + "valid_targets_mean": 2258.2, + "valid_targets_min": 903 + }, + { + "epoch": 0.8912331757002546, + "grad_norm": 0.3939741184572268, + "learning_rate": 3.991045938423543e-05, + "loss": 0.1398, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06305704265832901, + "step": 1225, + "valid_targets_mean": 3460.4, + "valid_targets_min": 2346 + }, + { + "epoch": 0.8948708621316842, + "grad_norm": 0.7825001677893623, + "learning_rate": 3.990699853516661e-05, + "loss": 0.1203, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10577984154224396, + "step": 1230, + "valid_targets_mean": 2283.0, + "valid_targets_min": 737 + }, + { + "epoch": 0.8985085485631139, + "grad_norm": 0.37524690757320867, + "learning_rate": 3.9903472221114595e-05, + "loss": 0.1222, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.060406021773815155, + "step": 1235, + "valid_targets_mean": 3979.6, + "valid_targets_min": 2555 + }, + { + "epoch": 0.9021462349945435, + "grad_norm": 0.31561159437465247, + "learning_rate": 3.9899880453675806e-05, + "loss": 0.1186, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05468086525797844, + "step": 1240, + "valid_targets_mean": 4500.2, + "valid_targets_min": 2991 + }, + { + "epoch": 0.9057839214259731, + "grad_norm": 0.43526420576074204, + "learning_rate": 3.9896223244661925e-05, + "loss": 0.104, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0428628996014595, + "step": 1245, + "valid_targets_mean": 1992.4, + "valid_targets_min": 618 + }, + { + "epoch": 0.9094216078574027, + "grad_norm": 0.40880427679923703, + "learning_rate": 3.989250060609983e-05, + "loss": 0.1346, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.040370333939790726, + "step": 1250, + "valid_targets_mean": 3641.8, + "valid_targets_min": 2944 + }, + { + "epoch": 0.9130592942888323, + "grad_norm": 0.3742595184694629, + "learning_rate": 3.988871255023158e-05, + "loss": 0.1193, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07022063434123993, + "step": 1255, + "valid_targets_mean": 3686.1, + "valid_targets_min": 2702 + }, + { + "epoch": 0.916696980720262, + "grad_norm": 0.4127424664447629, + "learning_rate": 3.9884859089514336e-05, + "loss": 0.1537, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.034560155123472214, + "step": 1260, + "valid_targets_mean": 3341.6, + "valid_targets_min": 509 + }, + { + "epoch": 0.9203346671516915, + "grad_norm": 0.34132925372811895, + "learning_rate": 3.988094023662038e-05, + "loss": 0.1405, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.049870237708091736, + "step": 1265, + "valid_targets_mean": 2757.6, + "valid_targets_min": 637 + }, + { + "epoch": 0.9239723535831211, + "grad_norm": 0.45234133456232267, + "learning_rate": 3.987695600443701e-05, + "loss": 0.1571, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07380931824445724, + "step": 1270, + "valid_targets_mean": 2802.6, + "valid_targets_min": 737 + }, + { + "epoch": 0.9276100400145507, + "grad_norm": 0.3441424821112778, + "learning_rate": 3.987290640606653e-05, + "loss": 0.1144, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.053822845220565796, + "step": 1275, + "valid_targets_mean": 3312.9, + "valid_targets_min": 1292 + }, + { + "epoch": 0.9312477264459803, + "grad_norm": 0.6131523760422221, + "learning_rate": 3.986879145482623e-05, + "loss": 0.141, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06906868517398834, + "step": 1280, + "valid_targets_mean": 1049.6, + "valid_targets_min": 697 + }, + { + "epoch": 0.93488541287741, + "grad_norm": 0.3869239142656959, + "learning_rate": 3.986461116424829e-05, + "loss": 0.1227, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05666602402925491, + "step": 1285, + "valid_targets_mean": 3541.4, + "valid_targets_min": 923 + }, + { + "epoch": 0.9385230993088396, + "grad_norm": 0.8231050134009347, + "learning_rate": 3.9860365548079756e-05, + "loss": 0.1791, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10355181246995926, + "step": 1290, + "valid_targets_mean": 931.6, + "valid_targets_min": 531 + }, + { + "epoch": 0.9421607857402692, + "grad_norm": 0.5269674174888391, + "learning_rate": 3.985605462028252e-05, + "loss": 0.1636, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08680590242147446, + "step": 1295, + "valid_targets_mean": 1840.5, + "valid_targets_min": 844 + }, + { + "epoch": 0.9457984721716988, + "grad_norm": 0.5011502719725875, + "learning_rate": 3.985167839503324e-05, + "loss": 0.1285, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08993671834468842, + "step": 1300, + "valid_targets_mean": 3893.9, + "valid_targets_min": 2967 + }, + { + "epoch": 0.9494361586031284, + "grad_norm": 0.3975015798663951, + "learning_rate": 3.9847236886723324e-05, + "loss": 0.1256, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06715572625398636, + "step": 1305, + "valid_targets_mean": 3605.9, + "valid_targets_min": 2490 + }, + { + "epoch": 0.953073845034558, + "grad_norm": 0.4048857594294113, + "learning_rate": 3.984273010995884e-05, + "loss": 0.1264, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05451614409685135, + "step": 1310, + "valid_targets_mean": 2682.0, + "valid_targets_min": 970 + }, + { + "epoch": 0.9567115314659876, + "grad_norm": 0.4285568374791069, + "learning_rate": 3.983815807956054e-05, + "loss": 0.1151, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04970483109354973, + "step": 1315, + "valid_targets_mean": 2028.5, + "valid_targets_min": 729 + }, + { + "epoch": 0.9603492178974172, + "grad_norm": 0.9152369264982673, + "learning_rate": 3.9833520810563685e-05, + "loss": 0.1566, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11088594049215317, + "step": 1320, + "valid_targets_mean": 1028.9, + "valid_targets_min": 637 + }, + { + "epoch": 0.9639869043288468, + "grad_norm": 0.37939972837964403, + "learning_rate": 3.982881831821816e-05, + "loss": 0.1172, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04228641837835312, + "step": 1325, + "valid_targets_mean": 2879.2, + "valid_targets_min": 701 + }, + { + "epoch": 0.9676245907602765, + "grad_norm": 0.338661409567677, + "learning_rate": 3.982405061798829e-05, + "loss": 0.121, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05238931253552437, + "step": 1330, + "valid_targets_mean": 3671.1, + "valid_targets_min": 3131 + }, + { + "epoch": 0.9712622771917061, + "grad_norm": 0.4839085404721662, + "learning_rate": 3.9819217725552864e-05, + "loss": 0.1268, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07042280584573746, + "step": 1335, + "valid_targets_mean": 3241.5, + "valid_targets_min": 2018 + }, + { + "epoch": 0.9748999636231357, + "grad_norm": 0.3419252553517584, + "learning_rate": 3.981431965680504e-05, + "loss": 0.1306, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.054422955960035324, + "step": 1340, + "valid_targets_mean": 2491.6, + "valid_targets_min": 350 + }, + { + "epoch": 0.9785376500545653, + "grad_norm": 0.3113829919990609, + "learning_rate": 3.9809356427852325e-05, + "loss": 0.1085, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05991678684949875, + "step": 1345, + "valid_targets_mean": 4862.8, + "valid_targets_min": 3577 + }, + { + "epoch": 0.9821753364859949, + "grad_norm": 0.39519039084425117, + "learning_rate": 3.9804328055016495e-05, + "loss": 0.1229, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04635890573263168, + "step": 1350, + "valid_targets_mean": 1999.4, + "valid_targets_min": 507 + }, + { + "epoch": 0.9858130229174246, + "grad_norm": 0.5177156487942793, + "learning_rate": 3.979923455483356e-05, + "loss": 0.1372, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11809035390615463, + "step": 1355, + "valid_targets_mean": 3278.5, + "valid_targets_min": 2814 + }, + { + "epoch": 0.9894507093488542, + "grad_norm": 0.4346749334624095, + "learning_rate": 3.9794075944053704e-05, + "loss": 0.215, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06560631096363068, + "step": 1360, + "valid_targets_mean": 4006.2, + "valid_targets_min": 2347 + }, + { + "epoch": 0.9930883957802837, + "grad_norm": 0.5425095291119021, + "learning_rate": 3.9788852239641237e-05, + "loss": 0.1243, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07507210969924927, + "step": 1365, + "valid_targets_mean": 2447.4, + "valid_targets_min": 716 + }, + { + "epoch": 0.9967260822117133, + "grad_norm": 0.38282846016551636, + "learning_rate": 3.978356345877452e-05, + "loss": 0.1184, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0562497079372406, + "step": 1370, + "valid_targets_mean": 3209.6, + "valid_targets_min": 2245 + }, + { + "epoch": 1.0, + "grad_norm": 0.6283560188764783, + "learning_rate": 3.977820961884593e-05, + "loss": 0.1806, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25742000341415405, + "step": 1375, + "valid_targets_mean": 7922.2, + "valid_targets_min": 3513 + }, + { + "epoch": 1.0036376864314296, + "grad_norm": 0.45034151271268225, + "learning_rate": 3.977279073746179e-05, + "loss": 0.2781, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12958192825317383, + "step": 1380, + "valid_targets_mean": 7389.0, + "valid_targets_min": 5627 + }, + { + "epoch": 1.0072753728628592, + "grad_norm": 0.7957758274806169, + "learning_rate": 3.9767306832442314e-05, + "loss": 0.2775, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07887107878923416, + "step": 1385, + "valid_targets_mean": 1294.8, + "valid_targets_min": 422 + }, + { + "epoch": 1.0109130592942888, + "grad_norm": 0.3779983188699539, + "learning_rate": 3.9761757921821544e-05, + "loss": 0.2639, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13302335143089294, + "step": 1390, + "valid_targets_mean": 8539.9, + "valid_targets_min": 6761 + }, + { + "epoch": 1.0145507457257184, + "grad_norm": 0.3871754327940695, + "learning_rate": 3.975614402384731e-05, + "loss": 0.246, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11376132071018219, + "step": 1395, + "valid_targets_mean": 7579.0, + "valid_targets_min": 6174 + }, + { + "epoch": 1.0181884321571482, + "grad_norm": 0.43275193380180427, + "learning_rate": 3.975046515698114e-05, + "loss": 0.2491, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11483652889728546, + "step": 1400, + "valid_targets_mean": 7328.2, + "valid_targets_min": 4839 + }, + { + "epoch": 1.0218261185885777, + "grad_norm": 0.40855570133006935, + "learning_rate": 3.974472133989822e-05, + "loss": 0.2563, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13608838617801666, + "step": 1405, + "valid_targets_mean": 6838.2, + "valid_targets_min": 5034 + }, + { + "epoch": 1.0254638050200073, + "grad_norm": 0.40249638639450086, + "learning_rate": 3.973891259148733e-05, + "loss": 0.2659, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13695399463176727, + "step": 1410, + "valid_targets_mean": 7248.2, + "valid_targets_min": 5567 + }, + { + "epoch": 1.029101491451437, + "grad_norm": 0.40198400955280184, + "learning_rate": 3.973303893085078e-05, + "loss": 0.2488, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11573392897844315, + "step": 1415, + "valid_targets_mean": 6234.0, + "valid_targets_min": 5603 + }, + { + "epoch": 1.0327391778828665, + "grad_norm": 0.7378916027598909, + "learning_rate": 3.972710037730434e-05, + "loss": 0.251, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04065033793449402, + "step": 1420, + "valid_targets_mean": 602.0, + "valid_targets_min": 152 + }, + { + "epoch": 1.036376864314296, + "grad_norm": 0.3829057837844518, + "learning_rate": 3.97210969503772e-05, + "loss": 0.2367, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11710065603256226, + "step": 1425, + "valid_targets_mean": 7299.9, + "valid_targets_min": 6158 + }, + { + "epoch": 1.0400145507457257, + "grad_norm": 0.39430504343906464, + "learning_rate": 3.971502866981186e-05, + "loss": 0.2465, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11391253769397736, + "step": 1430, + "valid_targets_mean": 6140.0, + "valid_targets_min": 4648 + }, + { + "epoch": 1.0436522371771553, + "grad_norm": 0.40046750313528423, + "learning_rate": 3.9708895555564114e-05, + "loss": 0.2373, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12107068300247192, + "step": 1435, + "valid_targets_mean": 6603.0, + "valid_targets_min": 4882 + }, + { + "epoch": 1.0472899236085849, + "grad_norm": 0.39192280174124494, + "learning_rate": 3.970269762780297e-05, + "loss": 0.2363, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12057182937860489, + "step": 1440, + "valid_targets_mean": 8015.0, + "valid_targets_min": 6344 + }, + { + "epoch": 1.0509276100400144, + "grad_norm": 0.40139583660140166, + "learning_rate": 3.969643490691057e-05, + "loss": 0.2276, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10805411636829376, + "step": 1445, + "valid_targets_mean": 6305.2, + "valid_targets_min": 4622 + }, + { + "epoch": 1.0545652964714443, + "grad_norm": 0.3930289590922795, + "learning_rate": 3.969010741348211e-05, + "loss": 0.2362, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12318243086338043, + "step": 1450, + "valid_targets_mean": 7890.8, + "valid_targets_min": 5485 + }, + { + "epoch": 1.0582029829028738, + "grad_norm": 0.40066863780195006, + "learning_rate": 3.968371516832582e-05, + "loss": 0.2368, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10832932591438293, + "step": 1455, + "valid_targets_mean": 7016.6, + "valid_targets_min": 4960 + }, + { + "epoch": 1.0618406693343034, + "grad_norm": 0.4357597546434051, + "learning_rate": 3.9677258192462866e-05, + "loss": 0.2315, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11051399260759354, + "step": 1460, + "valid_targets_mean": 6527.1, + "valid_targets_min": 5399 + }, + { + "epoch": 1.065478355765733, + "grad_norm": 0.4781519571826574, + "learning_rate": 3.9670736507127284e-05, + "loss": 0.2412, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11723552644252777, + "step": 1465, + "valid_targets_mean": 4995.0, + "valid_targets_min": 618 + }, + { + "epoch": 1.0691160421971626, + "grad_norm": 0.4021383366665016, + "learning_rate": 3.96641501337659e-05, + "loss": 0.2525, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12497500330209732, + "step": 1470, + "valid_targets_mean": 7219.4, + "valid_targets_min": 5764 + }, + { + "epoch": 1.0727537286285922, + "grad_norm": 0.38340874278089554, + "learning_rate": 3.9657499094038264e-05, + "loss": 0.2337, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12226970493793488, + "step": 1475, + "valid_targets_mean": 7998.5, + "valid_targets_min": 5902 + }, + { + "epoch": 1.0763914150600218, + "grad_norm": 0.36669022973771304, + "learning_rate": 3.96507834098166e-05, + "loss": 0.2186, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09501726180315018, + "step": 1480, + "valid_targets_mean": 7377.4, + "valid_targets_min": 5015 + }, + { + "epoch": 1.0800291014914514, + "grad_norm": 0.37723037197971354, + "learning_rate": 3.964400310318571e-05, + "loss": 0.2288, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09689757227897644, + "step": 1485, + "valid_targets_mean": 6794.0, + "valid_targets_min": 5283 + }, + { + "epoch": 1.083666787922881, + "grad_norm": 0.4767510092347408, + "learning_rate": 3.9637158196442925e-05, + "loss": 0.2384, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13380497694015503, + "step": 1490, + "valid_targets_mean": 6656.4, + "valid_targets_min": 4364 + }, + { + "epoch": 1.0873044743543105, + "grad_norm": 0.40930945282349956, + "learning_rate": 3.963024871209798e-05, + "loss": 0.2341, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10803690552711487, + "step": 1495, + "valid_targets_mean": 7178.4, + "valid_targets_min": 4448 + }, + { + "epoch": 1.0909421607857404, + "grad_norm": 0.41388037829490687, + "learning_rate": 3.962327467287302e-05, + "loss": 0.233, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13676932454109192, + "step": 1500, + "valid_targets_mean": 7619.8, + "valid_targets_min": 5737 + }, + { + "epoch": 1.09457984721717, + "grad_norm": 0.35779796549155585, + "learning_rate": 3.961623610170244e-05, + "loss": 0.2056, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09769880771636963, + "step": 1505, + "valid_targets_mean": 8754.4, + "valid_targets_min": 5752 + }, + { + "epoch": 1.0982175336485995, + "grad_norm": 0.41957964077510124, + "learning_rate": 3.9609133021732884e-05, + "loss": 0.2062, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10785550624132156, + "step": 1510, + "valid_targets_mean": 6341.0, + "valid_targets_min": 4543 + }, + { + "epoch": 1.1018552200800291, + "grad_norm": 0.43525139861271794, + "learning_rate": 3.960196545632311e-05, + "loss": 0.2, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11581838876008987, + "step": 1515, + "valid_targets_mean": 7181.9, + "valid_targets_min": 4563 + }, + { + "epoch": 1.1054929065114587, + "grad_norm": 0.8236475805785468, + "learning_rate": 3.9594733429043966e-05, + "loss": 0.2456, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08820606768131256, + "step": 1520, + "valid_targets_mean": 1104.4, + "valid_targets_min": 161 + }, + { + "epoch": 1.1091305929428883, + "grad_norm": 0.381952260942012, + "learning_rate": 3.9587436963678266e-05, + "loss": 0.2327, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10200639814138412, + "step": 1525, + "valid_targets_mean": 7018.2, + "valid_targets_min": 5424 + }, + { + "epoch": 1.1127682793743179, + "grad_norm": 0.4255431561282304, + "learning_rate": 3.9580076084220735e-05, + "loss": 0.2398, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12220368534326553, + "step": 1530, + "valid_targets_mean": 7738.4, + "valid_targets_min": 5329 + }, + { + "epoch": 1.1164059658057475, + "grad_norm": 0.4178206776751286, + "learning_rate": 3.957265081487792e-05, + "loss": 0.2456, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11849737912416458, + "step": 1535, + "valid_targets_mean": 6426.6, + "valid_targets_min": 5560 + }, + { + "epoch": 1.120043652237177, + "grad_norm": 0.4221510986676782, + "learning_rate": 3.9565161180068144e-05, + "loss": 0.2385, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11909832060337067, + "step": 1540, + "valid_targets_mean": 6169.9, + "valid_targets_min": 3446 + }, + { + "epoch": 1.1236813386686069, + "grad_norm": 0.40631051370168514, + "learning_rate": 3.955760720442136e-05, + "loss": 0.2294, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1190079003572464, + "step": 1545, + "valid_targets_mean": 7204.6, + "valid_targets_min": 5381 + }, + { + "epoch": 1.1273190251000365, + "grad_norm": 0.39646876201617487, + "learning_rate": 3.9549988912779136e-05, + "loss": 0.2341, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12148076295852661, + "step": 1550, + "valid_targets_mean": 7600.4, + "valid_targets_min": 5241 + }, + { + "epoch": 1.130956711531466, + "grad_norm": 0.4133396726814436, + "learning_rate": 3.954230633019454e-05, + "loss": 0.2276, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12449182569980621, + "step": 1555, + "valid_targets_mean": 7098.5, + "valid_targets_min": 5607 + }, + { + "epoch": 1.1345943979628956, + "grad_norm": 0.43308350579728505, + "learning_rate": 3.9534559481932054e-05, + "loss": 0.2331, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12216755747795105, + "step": 1560, + "valid_targets_mean": 6955.1, + "valid_targets_min": 5530 + }, + { + "epoch": 1.1382320843943252, + "grad_norm": 0.42833953554284154, + "learning_rate": 3.952674839346751e-05, + "loss": 0.2313, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11783788353204727, + "step": 1565, + "valid_targets_mean": 6748.1, + "valid_targets_min": 5364 + }, + { + "epoch": 1.1418697708257548, + "grad_norm": 0.45861577873988296, + "learning_rate": 3.951887309048799e-05, + "loss": 0.2339, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1132352203130722, + "step": 1570, + "valid_targets_mean": 6273.9, + "valid_targets_min": 5438 + }, + { + "epoch": 1.1455074572571844, + "grad_norm": 0.438071370304024, + "learning_rate": 3.951093359889177e-05, + "loss": 0.2318, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11211521923542023, + "step": 1575, + "valid_targets_mean": 6419.5, + "valid_targets_min": 4879 + }, + { + "epoch": 1.149145143688614, + "grad_norm": 0.43323790429111475, + "learning_rate": 3.950292994478816e-05, + "loss": 0.2291, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11147815734148026, + "step": 1580, + "valid_targets_mean": 6483.4, + "valid_targets_min": 4309 + }, + { + "epoch": 1.1527828301200436, + "grad_norm": 0.496082033368415, + "learning_rate": 3.949486215449754e-05, + "loss": 0.225, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1322714388370514, + "step": 1585, + "valid_targets_mean": 6825.0, + "valid_targets_min": 5205 + }, + { + "epoch": 1.1564205165514734, + "grad_norm": 0.38239133833432287, + "learning_rate": 3.948673025455115e-05, + "loss": 0.2269, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10547439754009247, + "step": 1590, + "valid_targets_mean": 6729.4, + "valid_targets_min": 5200 + }, + { + "epoch": 1.160058202982903, + "grad_norm": 0.4403257108035881, + "learning_rate": 3.947853427169108e-05, + "loss": 0.2242, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10987573862075806, + "step": 1595, + "valid_targets_mean": 6094.0, + "valid_targets_min": 5302 + }, + { + "epoch": 1.1636958894143326, + "grad_norm": 0.4336448947116083, + "learning_rate": 3.947027423287017e-05, + "loss": 0.2285, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12097756564617157, + "step": 1600, + "valid_targets_mean": 7140.0, + "valid_targets_min": 5270 + }, + { + "epoch": 1.1673335758457621, + "grad_norm": 0.37844511526358493, + "learning_rate": 3.946195016525187e-05, + "loss": 0.2209, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10682767629623413, + "step": 1605, + "valid_targets_mean": 7425.0, + "valid_targets_min": 5281 + }, + { + "epoch": 1.1709712622771917, + "grad_norm": 0.4170297840844095, + "learning_rate": 3.945356209621025e-05, + "loss": 0.2287, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14239364862442017, + "step": 1610, + "valid_targets_mean": 7063.4, + "valid_targets_min": 5386 + }, + { + "epoch": 1.1746089487086213, + "grad_norm": 0.40705188666122216, + "learning_rate": 3.94451100533298e-05, + "loss": 0.2285, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11969711631536484, + "step": 1615, + "valid_targets_mean": 6827.0, + "valid_targets_min": 4547 + }, + { + "epoch": 1.178246635140051, + "grad_norm": 0.6813278460315642, + "learning_rate": 3.9436594064405415e-05, + "loss": 0.2116, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07327491790056229, + "step": 1620, + "valid_targets_mean": 2047.0, + "valid_targets_min": 165 + }, + { + "epoch": 1.1818843215714805, + "grad_norm": 0.3901082603473679, + "learning_rate": 3.942801415744228e-05, + "loss": 0.2196, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10928869992494583, + "step": 1625, + "valid_targets_mean": 6754.1, + "valid_targets_min": 5196 + }, + { + "epoch": 1.18552200800291, + "grad_norm": 0.45847320695657956, + "learning_rate": 3.941937036065576e-05, + "loss": 0.2207, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11200568079948425, + "step": 1630, + "valid_targets_mean": 5985.6, + "valid_targets_min": 5028 + }, + { + "epoch": 1.1891596944343397, + "grad_norm": 0.4328547578483813, + "learning_rate": 3.941066270247135e-05, + "loss": 0.2267, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.102421835064888, + "step": 1635, + "valid_targets_mean": 5963.9, + "valid_targets_min": 5237 + }, + { + "epoch": 1.1927973808657693, + "grad_norm": 0.41501646390795927, + "learning_rate": 3.940189121152453e-05, + "loss": 0.2218, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1153603196144104, + "step": 1640, + "valid_targets_mean": 6319.0, + "valid_targets_min": 5412 + }, + { + "epoch": 1.196435067297199, + "grad_norm": 0.37415097729015556, + "learning_rate": 3.939305591666072e-05, + "loss": 0.2108, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09063620120286942, + "step": 1645, + "valid_targets_mean": 7125.8, + "valid_targets_min": 5080 + }, + { + "epoch": 1.2000727537286286, + "grad_norm": 0.470591533668796, + "learning_rate": 3.938415684693514e-05, + "loss": 0.2237, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11060422658920288, + "step": 1650, + "valid_targets_mean": 6123.5, + "valid_targets_min": 4521 + }, + { + "epoch": 1.2037104401600582, + "grad_norm": 0.674801118662757, + "learning_rate": 3.937519403161275e-05, + "loss": 0.2781, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17820681631565094, + "step": 1655, + "valid_targets_mean": 6020.6, + "valid_targets_min": 2493 + }, + { + "epoch": 1.2073481265914878, + "grad_norm": 0.7940408610456344, + "learning_rate": 3.936616750016814e-05, + "loss": 0.3301, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12876705825328827, + "step": 1660, + "valid_targets_mean": 3953.5, + "valid_targets_min": 1177 + }, + { + "epoch": 1.2109858130229174, + "grad_norm": 0.602134930678445, + "learning_rate": 3.935707728228543e-05, + "loss": 0.3346, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19898059964179993, + "step": 1665, + "valid_targets_mean": 4950.8, + "valid_targets_min": 2799 + }, + { + "epoch": 1.214623499454347, + "grad_norm": 0.5625477548775495, + "learning_rate": 3.9347923407858175e-05, + "loss": 0.3192, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1800476312637329, + "step": 1670, + "valid_targets_mean": 5312.0, + "valid_targets_min": 2973 + }, + { + "epoch": 1.2182611858857766, + "grad_norm": 0.5709928374376457, + "learning_rate": 3.933870590698926e-05, + "loss": 0.3164, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14832507073879242, + "step": 1675, + "valid_targets_mean": 3775.5, + "valid_targets_min": 980 + }, + { + "epoch": 1.2218988723172062, + "grad_norm": 0.550521166636932, + "learning_rate": 3.932942480999083e-05, + "loss": 0.337, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1728513240814209, + "step": 1680, + "valid_targets_mean": 5055.4, + "valid_targets_min": 1512 + }, + { + "epoch": 1.2255365587486358, + "grad_norm": 0.5327619807065184, + "learning_rate": 3.932008014738414e-05, + "loss": 0.3092, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11707329005002975, + "step": 1685, + "valid_targets_mean": 3600.8, + "valid_targets_min": 1848 + }, + { + "epoch": 1.2291742451800656, + "grad_norm": 0.4867524866957371, + "learning_rate": 3.93106719498995e-05, + "loss": 0.3178, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1171652153134346, + "step": 1690, + "valid_targets_mean": 3369.8, + "valid_targets_min": 1381 + }, + { + "epoch": 1.2328119316114952, + "grad_norm": 0.47302539771216634, + "learning_rate": 3.930120024847616e-05, + "loss": 0.318, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16447412967681885, + "step": 1695, + "valid_targets_mean": 6283.2, + "valid_targets_min": 3600 + }, + { + "epoch": 1.2364496180429247, + "grad_norm": 0.5321951330976278, + "learning_rate": 3.929166507426219e-05, + "loss": 0.3302, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16255125403404236, + "step": 1700, + "valid_targets_mean": 5242.5, + "valid_targets_min": 2540 + }, + { + "epoch": 1.2400873044743543, + "grad_norm": 0.476435096062693, + "learning_rate": 3.928206645861442e-05, + "loss": 0.3161, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14480793476104736, + "step": 1705, + "valid_targets_mean": 6246.5, + "valid_targets_min": 2157 + }, + { + "epoch": 1.243724990905784, + "grad_norm": 0.5590299157912856, + "learning_rate": 3.927240443309827e-05, + "loss": 0.3197, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1628410518169403, + "step": 1710, + "valid_targets_mean": 4556.5, + "valid_targets_min": 1720 + }, + { + "epoch": 1.2473626773372135, + "grad_norm": 0.5321673140850889, + "learning_rate": 3.926267902948772e-05, + "loss": 0.323, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17951777577400208, + "step": 1715, + "valid_targets_mean": 5116.5, + "valid_targets_min": 1982 + }, + { + "epoch": 1.251000363768643, + "grad_norm": 0.5465550956835176, + "learning_rate": 3.925289027976517e-05, + "loss": 0.317, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16249269247055054, + "step": 1720, + "valid_targets_mean": 4464.5, + "valid_targets_min": 2718 + }, + { + "epoch": 1.2546380502000727, + "grad_norm": 0.6349671083202529, + "learning_rate": 3.9243038216121304e-05, + "loss": 0.3285, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1580926924943924, + "step": 1725, + "valid_targets_mean": 3418.5, + "valid_targets_min": 1451 + }, + { + "epoch": 1.2582757366315023, + "grad_norm": 0.5833567535374689, + "learning_rate": 3.923312287095505e-05, + "loss": 0.3268, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1676788181066513, + "step": 1730, + "valid_targets_mean": 3990.4, + "valid_targets_min": 1861 + }, + { + "epoch": 1.261913423062932, + "grad_norm": 0.5981375342516458, + "learning_rate": 3.922314427687343e-05, + "loss": 0.3185, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.176347553730011, + "step": 1735, + "valid_targets_mean": 4065.4, + "valid_targets_min": 2100 + }, + { + "epoch": 1.2655511094943614, + "grad_norm": 0.5992091862537661, + "learning_rate": 3.921310246669146e-05, + "loss": 0.3305, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18265917897224426, + "step": 1740, + "valid_targets_mean": 4132.4, + "valid_targets_min": 3234 + }, + { + "epoch": 1.2691887959257913, + "grad_norm": 0.6559607090299836, + "learning_rate": 3.920299747343204e-05, + "loss": 0.3372, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20549055933952332, + "step": 1745, + "valid_targets_mean": 4246.2, + "valid_targets_min": 2650 + }, + { + "epoch": 1.2728264823572208, + "grad_norm": 0.5765176065123253, + "learning_rate": 3.9192829330325856e-05, + "loss": 0.3271, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13869526982307434, + "step": 1750, + "valid_targets_mean": 4096.5, + "valid_targets_min": 1719 + }, + { + "epoch": 1.2764641687886504, + "grad_norm": 0.5880562792376881, + "learning_rate": 3.918259807081128e-05, + "loss": 0.3323, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1648177206516266, + "step": 1755, + "valid_targets_mean": 3933.1, + "valid_targets_min": 1020 + }, + { + "epoch": 1.28010185522008, + "grad_norm": 0.5698952900995367, + "learning_rate": 3.917230372853421e-05, + "loss": 0.3127, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16284897923469543, + "step": 1760, + "valid_targets_mean": 3551.1, + "valid_targets_min": 2130 + }, + { + "epoch": 1.2837395416515096, + "grad_norm": 0.5870026530148993, + "learning_rate": 3.916194633734803e-05, + "loss": 0.3081, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13259905576705933, + "step": 1765, + "valid_targets_mean": 3535.0, + "valid_targets_min": 865 + }, + { + "epoch": 1.2873772280829392, + "grad_norm": 0.686332186883629, + "learning_rate": 3.9151525931313425e-05, + "loss": 0.3165, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2147303819656372, + "step": 1770, + "valid_targets_mean": 3716.5, + "valid_targets_min": 3144 + }, + { + "epoch": 1.2910149145143688, + "grad_norm": 0.5573461443416352, + "learning_rate": 3.914104254469835e-05, + "loss": 0.3251, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.136275976896286, + "step": 1775, + "valid_targets_mean": 3262.2, + "valid_targets_min": 2050 + }, + { + "epoch": 1.2946526009457986, + "grad_norm": 0.5709337259289007, + "learning_rate": 3.9130496211977845e-05, + "loss": 0.3242, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16184990108013153, + "step": 1780, + "valid_targets_mean": 3296.5, + "valid_targets_min": 1570 + }, + { + "epoch": 1.298290287377228, + "grad_norm": 0.5481176472243965, + "learning_rate": 3.911988696783396e-05, + "loss": 0.3074, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15605130791664124, + "step": 1785, + "valid_targets_mean": 4806.4, + "valid_targets_min": 1795 + }, + { + "epoch": 1.3019279738086578, + "grad_norm": 0.5593576496892754, + "learning_rate": 3.910921484715563e-05, + "loss": 0.3108, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14285463094711304, + "step": 1790, + "valid_targets_mean": 4342.4, + "valid_targets_min": 1934 + }, + { + "epoch": 1.3055656602400874, + "grad_norm": 0.6102651941922826, + "learning_rate": 3.909847988503856e-05, + "loss": 0.3182, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16056188941001892, + "step": 1795, + "valid_targets_mean": 3675.4, + "valid_targets_min": 2557 + }, + { + "epoch": 1.309203346671517, + "grad_norm": 0.6054625607396906, + "learning_rate": 3.9087682116785114e-05, + "loss": 0.3159, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1763242930173874, + "step": 1800, + "valid_targets_mean": 4550.4, + "valid_targets_min": 2663 + }, + { + "epoch": 1.3128410331029465, + "grad_norm": 0.5485154213363367, + "learning_rate": 3.907682157790421e-05, + "loss": 0.308, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15230399370193481, + "step": 1805, + "valid_targets_mean": 4604.0, + "valid_targets_min": 2016 + }, + { + "epoch": 1.3164787195343761, + "grad_norm": 0.70230194893904, + "learning_rate": 3.906589830411116e-05, + "loss": 0.3144, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13870956003665924, + "step": 1810, + "valid_targets_mean": 2976.8, + "valid_targets_min": 1367 + }, + { + "epoch": 1.3201164059658057, + "grad_norm": 0.6627997708026013, + "learning_rate": 3.905491233132761e-05, + "loss": 0.3222, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1531262993812561, + "step": 1815, + "valid_targets_mean": 3542.5, + "valid_targets_min": 2026 + }, + { + "epoch": 1.3237540923972353, + "grad_norm": 0.6180043326700139, + "learning_rate": 3.904386369568137e-05, + "loss": 0.304, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14495117962360382, + "step": 1820, + "valid_targets_mean": 3247.2, + "valid_targets_min": 2033 + }, + { + "epoch": 1.327391778828665, + "grad_norm": 0.7511463314717345, + "learning_rate": 3.903275243350634e-05, + "loss": 0.3233, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16049274802207947, + "step": 1825, + "valid_targets_mean": 2201.6, + "valid_targets_min": 1283 + }, + { + "epoch": 1.3310294652600945, + "grad_norm": 0.6992033663180066, + "learning_rate": 3.9021578581342373e-05, + "loss": 0.3169, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16625064611434937, + "step": 1830, + "valid_targets_mean": 2824.8, + "valid_targets_min": 1636 + }, + { + "epoch": 1.3346671516915243, + "grad_norm": 0.6482413352541484, + "learning_rate": 3.901034217593512e-05, + "loss": 0.3154, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17135660350322723, + "step": 1835, + "valid_targets_mean": 3844.4, + "valid_targets_min": 1629 + }, + { + "epoch": 1.3383048381229539, + "grad_norm": 0.6342599615328701, + "learning_rate": 3.899904325423599e-05, + "loss": 0.3077, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14275401830673218, + "step": 1840, + "valid_targets_mean": 3048.6, + "valid_targets_min": 2080 + }, + { + "epoch": 1.3419425245543835, + "grad_norm": 0.7040285566089268, + "learning_rate": 3.8987681853401924e-05, + "loss": 0.3104, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14169757068157196, + "step": 1845, + "valid_targets_mean": 3111.8, + "valid_targets_min": 1153 + }, + { + "epoch": 1.345580210985813, + "grad_norm": 0.6007526159521345, + "learning_rate": 3.897625801079537e-05, + "loss": 0.3077, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16859573125839233, + "step": 1850, + "valid_targets_mean": 4069.0, + "valid_targets_min": 1858 + }, + { + "epoch": 1.3492178974172426, + "grad_norm": 0.558854775874792, + "learning_rate": 3.89647717639841e-05, + "loss": 0.3216, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13408693671226501, + "step": 1855, + "valid_targets_mean": 3415.2, + "valid_targets_min": 1639 + }, + { + "epoch": 1.3528555838486722, + "grad_norm": 0.6297937303424476, + "learning_rate": 3.8953223150741115e-05, + "loss": 0.3144, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1279762089252472, + "step": 1860, + "valid_targets_mean": 3267.6, + "valid_targets_min": 1394 + }, + { + "epoch": 1.3564932702801018, + "grad_norm": 0.5426577012589429, + "learning_rate": 3.894161220904449e-05, + "loss": 0.3074, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1627165526151657, + "step": 1865, + "valid_targets_mean": 4775.9, + "valid_targets_min": 1844 + }, + { + "epoch": 1.3601309567115314, + "grad_norm": 0.670785560862999, + "learning_rate": 3.8929938977077304e-05, + "loss": 0.3089, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13962198793888092, + "step": 1870, + "valid_targets_mean": 3317.1, + "valid_targets_min": 1305 + }, + { + "epoch": 1.363768643142961, + "grad_norm": 0.5844251197870406, + "learning_rate": 3.8918203493227445e-05, + "loss": 0.3071, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1659080982208252, + "step": 1875, + "valid_targets_mean": 4290.8, + "valid_targets_min": 2185 + }, + { + "epoch": 1.3674063295743908, + "grad_norm": 0.6105297958260628, + "learning_rate": 3.8906405796087545e-05, + "loss": 0.3124, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1823703497648239, + "step": 1880, + "valid_targets_mean": 4024.9, + "valid_targets_min": 2349 + }, + { + "epoch": 1.3710440160058204, + "grad_norm": 0.5652860601311969, + "learning_rate": 3.889454592445481e-05, + "loss": 0.2969, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13505664467811584, + "step": 1885, + "valid_targets_mean": 3453.0, + "valid_targets_min": 1607 + }, + { + "epoch": 1.37468170243725, + "grad_norm": 0.6622115017053676, + "learning_rate": 3.888262391733091e-05, + "loss": 0.3197, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15952789783477783, + "step": 1890, + "valid_targets_mean": 3389.8, + "valid_targets_min": 1507 + }, + { + "epoch": 1.3783193888686796, + "grad_norm": 0.6463449685928512, + "learning_rate": 3.887063981392187e-05, + "loss": 0.316, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14368987083435059, + "step": 1895, + "valid_targets_mean": 3773.9, + "valid_targets_min": 1317 + }, + { + "epoch": 1.3819570753001091, + "grad_norm": 0.6303291522424677, + "learning_rate": 3.885859365363789e-05, + "loss": 0.3095, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14195391535758972, + "step": 1900, + "valid_targets_mean": 3881.0, + "valid_targets_min": 3355 + }, + { + "epoch": 1.3855947617315387, + "grad_norm": 0.5758789251470144, + "learning_rate": 3.884648547609328e-05, + "loss": 0.3117, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1494072824716568, + "step": 1905, + "valid_targets_mean": 3951.9, + "valid_targets_min": 2999 + }, + { + "epoch": 1.3892324481629683, + "grad_norm": 0.5665132049058939, + "learning_rate": 3.883431532110627e-05, + "loss": 0.3091, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13890591263771057, + "step": 1910, + "valid_targets_mean": 4086.5, + "valid_targets_min": 1739 + }, + { + "epoch": 1.392870134594398, + "grad_norm": 0.5753576457919256, + "learning_rate": 3.882208322869891e-05, + "loss": 0.3107, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13475655019283295, + "step": 1915, + "valid_targets_mean": 3298.0, + "valid_targets_min": 1140 + }, + { + "epoch": 1.3965078210258275, + "grad_norm": 0.5736651484736572, + "learning_rate": 3.8809789239096956e-05, + "loss": 0.2923, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1610942780971527, + "step": 1920, + "valid_targets_mean": 4012.1, + "valid_targets_min": 2779 + }, + { + "epoch": 1.4001455074572573, + "grad_norm": 0.6014561037392774, + "learning_rate": 3.879743339272968e-05, + "loss": 0.3032, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1469648778438568, + "step": 1925, + "valid_targets_mean": 3636.9, + "valid_targets_min": 1183 + }, + { + "epoch": 1.4037831938886867, + "grad_norm": 0.6101294247276551, + "learning_rate": 3.87850157302298e-05, + "loss": 0.2896, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17685158550739288, + "step": 1930, + "valid_targets_mean": 4269.0, + "valid_targets_min": 3041 + }, + { + "epoch": 1.4074208803201165, + "grad_norm": 0.6208549688205313, + "learning_rate": 3.8772536292433314e-05, + "loss": 0.3176, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15794846415519714, + "step": 1935, + "valid_targets_mean": 3350.1, + "valid_targets_min": 1726 + }, + { + "epoch": 1.411058566751546, + "grad_norm": 0.512281998417535, + "learning_rate": 3.8759995120379355e-05, + "loss": 0.2979, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1391405165195465, + "step": 1940, + "valid_targets_mean": 4747.4, + "valid_targets_min": 1743 + }, + { + "epoch": 1.4146962531829756, + "grad_norm": 0.5404928074268446, + "learning_rate": 3.874739225531009e-05, + "loss": 0.3085, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1466849148273468, + "step": 1945, + "valid_targets_mean": 3702.8, + "valid_targets_min": 1234 + }, + { + "epoch": 1.4183339396144052, + "grad_norm": 0.6231618801959369, + "learning_rate": 3.873472773867056e-05, + "loss": 0.3158, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15846596658229828, + "step": 1950, + "valid_targets_mean": 4079.0, + "valid_targets_min": 2303 + }, + { + "epoch": 1.4219716260458348, + "grad_norm": 0.605872459211176, + "learning_rate": 3.8722001612108545e-05, + "loss": 0.3138, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1550457775592804, + "step": 1955, + "valid_targets_mean": 3830.8, + "valid_targets_min": 1651 + }, + { + "epoch": 1.4256093124772644, + "grad_norm": 0.6585396848979318, + "learning_rate": 3.870921391747443e-05, + "loss": 0.3057, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16306890547275543, + "step": 1960, + "valid_targets_mean": 2876.2, + "valid_targets_min": 1374 + }, + { + "epoch": 1.429246998908694, + "grad_norm": 0.7873666305191546, + "learning_rate": 3.869636469682109e-05, + "loss": 0.298, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15194176137447357, + "step": 1965, + "valid_targets_mean": 3467.5, + "valid_targets_min": 1625 + }, + { + "epoch": 1.4328846853401238, + "grad_norm": 0.587358505182511, + "learning_rate": 3.8683453992403704e-05, + "loss": 0.2928, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14921635389328003, + "step": 1970, + "valid_targets_mean": 4089.5, + "valid_targets_min": 997 + }, + { + "epoch": 1.4365223717715532, + "grad_norm": 0.5855458830530614, + "learning_rate": 3.8670481846679656e-05, + "loss": 0.2977, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1246376633644104, + "step": 1975, + "valid_targets_mean": 3618.6, + "valid_targets_min": 1179 + }, + { + "epoch": 1.440160058202983, + "grad_norm": 0.5942699003617847, + "learning_rate": 3.865744830230838e-05, + "loss": 0.2968, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14212846755981445, + "step": 1980, + "valid_targets_mean": 3079.1, + "valid_targets_min": 1444 + }, + { + "epoch": 1.4437977446344126, + "grad_norm": 0.815501717589764, + "learning_rate": 3.864435340215124e-05, + "loss": 0.2815, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12810811400413513, + "step": 1985, + "valid_targets_mean": 3486.8, + "valid_targets_min": 1504 + }, + { + "epoch": 1.4474354310658422, + "grad_norm": 0.5811123078765571, + "learning_rate": 3.863119718927132e-05, + "loss": 0.2914, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14832298457622528, + "step": 1990, + "valid_targets_mean": 3998.4, + "valid_targets_min": 2302 + }, + { + "epoch": 1.4510731174972717, + "grad_norm": 0.61648131738689, + "learning_rate": 3.8617979706933384e-05, + "loss": 0.3023, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18779967725276947, + "step": 1995, + "valid_targets_mean": 4282.5, + "valid_targets_min": 1818 + }, + { + "epoch": 1.4547108039287013, + "grad_norm": 0.5944691106467191, + "learning_rate": 3.860470099860368e-05, + "loss": 0.3232, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16084171831607819, + "step": 2000, + "valid_targets_mean": 4221.9, + "valid_targets_min": 2633 + }, + { + "epoch": 1.458348490360131, + "grad_norm": 0.592970359573761, + "learning_rate": 3.8591361107949766e-05, + "loss": 0.2954, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15279528498649597, + "step": 2005, + "valid_targets_mean": 3522.4, + "valid_targets_min": 2481 + }, + { + "epoch": 1.4619861767915605, + "grad_norm": 0.6017698381225525, + "learning_rate": 3.8577960078840424e-05, + "loss": 0.3056, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16447970271110535, + "step": 2010, + "valid_targets_mean": 4058.8, + "valid_targets_min": 2013 + }, + { + "epoch": 1.46562386322299, + "grad_norm": 0.6172396471201935, + "learning_rate": 3.85644979553455e-05, + "loss": 0.292, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15854698419570923, + "step": 2015, + "valid_targets_mean": 3723.5, + "valid_targets_min": 1744 + }, + { + "epoch": 1.4692615496544197, + "grad_norm": 0.6080808198965092, + "learning_rate": 3.855097478173574e-05, + "loss": 0.3052, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15987856686115265, + "step": 2020, + "valid_targets_mean": 4180.9, + "valid_targets_min": 1914 + }, + { + "epoch": 1.4728992360858495, + "grad_norm": 0.6229815983575846, + "learning_rate": 3.8537390602482644e-05, + "loss": 0.2921, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12944184243679047, + "step": 2025, + "valid_targets_mean": 3668.6, + "valid_targets_min": 3065 + }, + { + "epoch": 1.476536922517279, + "grad_norm": 0.5434681621837552, + "learning_rate": 3.8523745462258346e-05, + "loss": 0.2822, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17030569911003113, + "step": 2030, + "valid_targets_mean": 4473.2, + "valid_targets_min": 2894 + }, + { + "epoch": 1.4801746089487087, + "grad_norm": 0.6106139237787132, + "learning_rate": 3.851003940593546e-05, + "loss": 0.3046, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15291549265384674, + "step": 2035, + "valid_targets_mean": 3395.6, + "valid_targets_min": 2359 + }, + { + "epoch": 1.4838122953801383, + "grad_norm": 0.6476312887935898, + "learning_rate": 3.849627247858691e-05, + "loss": 0.2932, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13230586051940918, + "step": 2040, + "valid_targets_mean": 3114.5, + "valid_targets_min": 1463 + }, + { + "epoch": 1.4874499818115678, + "grad_norm": 0.6077477842721156, + "learning_rate": 3.848244472548581e-05, + "loss": 0.3035, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15149939060211182, + "step": 2045, + "valid_targets_mean": 3588.1, + "valid_targets_min": 1426 + }, + { + "epoch": 1.4910876682429974, + "grad_norm": 0.6034407177843439, + "learning_rate": 3.846855619210528e-05, + "loss": 0.3094, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14907614886760712, + "step": 2050, + "valid_targets_mean": 3405.5, + "valid_targets_min": 1661 + }, + { + "epoch": 1.494725354674427, + "grad_norm": 0.5773665913921754, + "learning_rate": 3.8454606924118345e-05, + "loss": 0.3012, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15548424422740936, + "step": 2055, + "valid_targets_mean": 4457.8, + "valid_targets_min": 1529 + }, + { + "epoch": 1.4983630411058566, + "grad_norm": 0.699843395164505, + "learning_rate": 3.844059696739774e-05, + "loss": 0.3302, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17992223799228668, + "step": 2060, + "valid_targets_mean": 3272.9, + "valid_targets_min": 1450 + }, + { + "epoch": 1.5020007275372862, + "grad_norm": 0.6186653594639459, + "learning_rate": 3.842652636801579e-05, + "loss": 0.2987, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1321069896221161, + "step": 2065, + "valid_targets_mean": 3103.2, + "valid_targets_min": 1625 + }, + { + "epoch": 1.505638413968716, + "grad_norm": 0.5537066897755315, + "learning_rate": 3.8412395172244236e-05, + "loss": 0.3071, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12781161069869995, + "step": 2070, + "valid_targets_mean": 3952.2, + "valid_targets_min": 1174 + }, + { + "epoch": 1.5092761004001454, + "grad_norm": 0.6355868275063258, + "learning_rate": 3.83982034265541e-05, + "loss": 0.3094, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15797901153564453, + "step": 2075, + "valid_targets_mean": 4039.2, + "valid_targets_min": 2882 + }, + { + "epoch": 1.5129137868315752, + "grad_norm": 0.5991457631943906, + "learning_rate": 3.838395117761551e-05, + "loss": 0.311, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1723627746105194, + "step": 2080, + "valid_targets_mean": 4594.0, + "valid_targets_min": 1946 + }, + { + "epoch": 1.5165514732630048, + "grad_norm": 0.550295639320478, + "learning_rate": 3.836963847229758e-05, + "loss": 0.3029, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1478092223405838, + "step": 2085, + "valid_targets_mean": 4082.6, + "valid_targets_min": 2000 + }, + { + "epoch": 1.5201891596944344, + "grad_norm": 0.5796027356897628, + "learning_rate": 3.8355265357668233e-05, + "loss": 0.3086, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16540232300758362, + "step": 2090, + "valid_targets_mean": 4596.1, + "valid_targets_min": 2012 + }, + { + "epoch": 1.523826846125864, + "grad_norm": 0.6098320346119771, + "learning_rate": 3.8340831880994044e-05, + "loss": 0.307, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15490400791168213, + "step": 2095, + "valid_targets_mean": 3094.4, + "valid_targets_min": 719 + }, + { + "epoch": 1.5274645325572935, + "grad_norm": 0.5260925928597147, + "learning_rate": 3.83263380897401e-05, + "loss": 0.3024, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13944070041179657, + "step": 2100, + "valid_targets_mean": 4496.2, + "valid_targets_min": 3238 + }, + { + "epoch": 1.5311022189887231, + "grad_norm": 0.5972126907174602, + "learning_rate": 3.831178403156982e-05, + "loss": 0.2894, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1293070763349533, + "step": 2105, + "valid_targets_mean": 3816.8, + "valid_targets_min": 1372 + }, + { + "epoch": 1.5347399054201527, + "grad_norm": 0.6844123323791906, + "learning_rate": 3.829716975434483e-05, + "loss": 0.286, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15981033444404602, + "step": 2110, + "valid_targets_mean": 3306.1, + "valid_targets_min": 2370 + }, + { + "epoch": 1.5383775918515825, + "grad_norm": 0.5830594421973586, + "learning_rate": 3.828249530612477e-05, + "loss": 0.297, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12970951199531555, + "step": 2115, + "valid_targets_mean": 2958.1, + "valid_targets_min": 1576 + }, + { + "epoch": 1.5420152782830119, + "grad_norm": 0.5923727857326865, + "learning_rate": 3.826776073516719e-05, + "loss": 0.2933, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1607396900653839, + "step": 2120, + "valid_targets_mean": 3758.5, + "valid_targets_min": 1426 + }, + { + "epoch": 1.5456529647144417, + "grad_norm": 0.6360026772248055, + "learning_rate": 3.825296608992731e-05, + "loss": 0.3035, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1332211196422577, + "step": 2125, + "valid_targets_mean": 3117.0, + "valid_targets_min": 1728 + }, + { + "epoch": 1.5492906511458713, + "grad_norm": 0.6803971860228673, + "learning_rate": 3.8238111419057935e-05, + "loss": 0.2984, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16000638902187347, + "step": 2130, + "valid_targets_mean": 3264.2, + "valid_targets_min": 1362 + }, + { + "epoch": 1.5529283375773009, + "grad_norm": 0.5608738240519487, + "learning_rate": 3.822319677140926e-05, + "loss": 0.2992, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1448010802268982, + "step": 2135, + "valid_targets_mean": 3836.0, + "valid_targets_min": 852 + }, + { + "epoch": 1.5565660240087305, + "grad_norm": 0.7249693833267111, + "learning_rate": 3.820822219602873e-05, + "loss": 0.3032, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18630772829055786, + "step": 2140, + "valid_targets_mean": 2764.1, + "valid_targets_min": 1195 + }, + { + "epoch": 1.56020371044016, + "grad_norm": 0.5737746084862969, + "learning_rate": 3.819318774216083e-05, + "loss": 0.2911, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15606160461902618, + "step": 2145, + "valid_targets_mean": 4457.5, + "valid_targets_min": 1564 + }, + { + "epoch": 1.5638413968715896, + "grad_norm": 0.5069181837081563, + "learning_rate": 3.8178093459247e-05, + "loss": 0.2872, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12255532294511795, + "step": 2150, + "valid_targets_mean": 4371.2, + "valid_targets_min": 1975 + }, + { + "epoch": 1.5674790833030192, + "grad_norm": 0.5470236415721637, + "learning_rate": 3.81629393969254e-05, + "loss": 0.2786, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14719901978969574, + "step": 2155, + "valid_targets_mean": 3965.8, + "valid_targets_min": 2426 + }, + { + "epoch": 1.571116769734449, + "grad_norm": 0.47058443091311936, + "learning_rate": 3.8147725605030785e-05, + "loss": 0.2828, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14284648001194, + "step": 2160, + "valid_targets_mean": 6384.6, + "valid_targets_min": 2678 + }, + { + "epoch": 1.5747544561658784, + "grad_norm": 0.5641020125005818, + "learning_rate": 3.8132452133594345e-05, + "loss": 0.2925, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13922816514968872, + "step": 2165, + "valid_targets_mean": 3543.2, + "valid_targets_min": 2826 + }, + { + "epoch": 1.5783921425973082, + "grad_norm": 0.7150638804936938, + "learning_rate": 3.8117119032843506e-05, + "loss": 0.2795, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1170087456703186, + "step": 2170, + "valid_targets_mean": 3155.8, + "valid_targets_min": 993 + }, + { + "epoch": 1.5820298290287376, + "grad_norm": 0.5925509028805128, + "learning_rate": 3.810172635320181e-05, + "loss": 0.2846, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15958797931671143, + "step": 2175, + "valid_targets_mean": 4166.5, + "valid_targets_min": 2724 + }, + { + "epoch": 1.5856675154601674, + "grad_norm": 0.6220705766770607, + "learning_rate": 3.808627414528871e-05, + "loss": 0.2864, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13429248332977295, + "step": 2180, + "valid_targets_mean": 3079.5, + "valid_targets_min": 1360 + }, + { + "epoch": 1.589305201891597, + "grad_norm": 0.5325761403657926, + "learning_rate": 3.807076245991943e-05, + "loss": 0.2823, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1388172209262848, + "step": 2185, + "valid_targets_mean": 4114.9, + "valid_targets_min": 2414 + }, + { + "epoch": 1.5929428883230266, + "grad_norm": 0.5965853844898403, + "learning_rate": 3.8055191348104795e-05, + "loss": 0.3047, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16585558652877808, + "step": 2190, + "valid_targets_mean": 4524.8, + "valid_targets_min": 2425 + }, + { + "epoch": 1.5965805747544561, + "grad_norm": 0.6322443967286924, + "learning_rate": 3.8039560861051036e-05, + "loss": 0.2973, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14818862080574036, + "step": 2195, + "valid_targets_mean": 3035.2, + "valid_targets_min": 1720 + }, + { + "epoch": 1.6002182611858857, + "grad_norm": 0.6582202164147167, + "learning_rate": 3.802387105015964e-05, + "loss": 0.2931, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14870315790176392, + "step": 2200, + "valid_targets_mean": 3260.4, + "valid_targets_min": 1621 + }, + { + "epoch": 1.6038559476173155, + "grad_norm": 0.6977052947496608, + "learning_rate": 3.800812196702722e-05, + "loss": 0.279, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15004289150238037, + "step": 2205, + "valid_targets_mean": 3214.6, + "valid_targets_min": 1858 + }, + { + "epoch": 1.607493634048745, + "grad_norm": 0.6228855571642352, + "learning_rate": 3.799231366344527e-05, + "loss": 0.2925, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13409893214702606, + "step": 2210, + "valid_targets_mean": 3254.2, + "valid_targets_min": 1254 + }, + { + "epoch": 1.6111313204801747, + "grad_norm": 0.5881210697588174, + "learning_rate": 3.797644619140005e-05, + "loss": 0.2833, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.155431866645813, + "step": 2215, + "valid_targets_mean": 2994.4, + "valid_targets_min": 1039 + }, + { + "epoch": 1.614769006911604, + "grad_norm": 0.5788886357868933, + "learning_rate": 3.796051960307239e-05, + "loss": 0.2988, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18311764299869537, + "step": 2220, + "valid_targets_mean": 4625.4, + "valid_targets_min": 2526 + }, + { + "epoch": 1.6184066933430339, + "grad_norm": 0.5767631182724388, + "learning_rate": 3.794453395083753e-05, + "loss": 0.2874, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15537132322788239, + "step": 2225, + "valid_targets_mean": 3571.8, + "valid_targets_min": 2409 + }, + { + "epoch": 1.6220443797744635, + "grad_norm": 0.5807359003801873, + "learning_rate": 3.792848928726496e-05, + "loss": 0.2914, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12560829520225525, + "step": 2230, + "valid_targets_mean": 3691.5, + "valid_targets_min": 2918 + }, + { + "epoch": 1.625682066205893, + "grad_norm": 0.6015921267895848, + "learning_rate": 3.7912385665118194e-05, + "loss": 0.295, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14455515146255493, + "step": 2235, + "valid_targets_mean": 4063.1, + "valid_targets_min": 2179 + }, + { + "epoch": 1.6293197526373226, + "grad_norm": 0.5201800095975557, + "learning_rate": 3.789622313735467e-05, + "loss": 0.253, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07017992436885834, + "step": 2240, + "valid_targets_mean": 2871.4, + "valid_targets_min": 955 + }, + { + "epoch": 1.6329574390687522, + "grad_norm": 0.4961683240395367, + "learning_rate": 3.7880001757125515e-05, + "loss": 0.1511, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.057317860424518585, + "step": 2245, + "valid_targets_mean": 1788.9, + "valid_targets_min": 669 + }, + { + "epoch": 1.636595125500182, + "grad_norm": 0.40795723729471994, + "learning_rate": 3.7863721577775416e-05, + "loss": 0.1129, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05178265646100044, + "step": 2250, + "valid_targets_mean": 3755.8, + "valid_targets_min": 2651 + }, + { + "epoch": 1.6402328119316114, + "grad_norm": 0.4352074661959599, + "learning_rate": 3.784738265284241e-05, + "loss": 0.1126, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06184077635407448, + "step": 2255, + "valid_targets_mean": 3069.8, + "valid_targets_min": 710 + }, + { + "epoch": 1.6438704983630412, + "grad_norm": 0.3698753747057345, + "learning_rate": 3.783098503605771e-05, + "loss": 0.1263, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03719334304332733, + "step": 2260, + "valid_targets_mean": 3618.5, + "valid_targets_min": 1365 + }, + { + "epoch": 1.6475081847944706, + "grad_norm": 0.45451165716858266, + "learning_rate": 3.781452878134557e-05, + "loss": 0.112, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05961167812347412, + "step": 2265, + "valid_targets_mean": 3279.9, + "valid_targets_min": 889 + }, + { + "epoch": 1.6511458712259004, + "grad_norm": 0.7614075332760967, + "learning_rate": 3.779801394282305e-05, + "loss": 0.1482, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08139972388744354, + "step": 2270, + "valid_targets_mean": 1652.4, + "valid_targets_min": 559 + }, + { + "epoch": 1.65478355765733, + "grad_norm": 0.35924429417999904, + "learning_rate": 3.778144057479988e-05, + "loss": 0.1502, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04681181162595749, + "step": 2275, + "valid_targets_mean": 3449.0, + "valid_targets_min": 1013 + }, + { + "epoch": 1.6584212440887596, + "grad_norm": 0.4053291343008899, + "learning_rate": 3.7764808731778255e-05, + "loss": 0.1255, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07239241153001785, + "step": 2280, + "valid_targets_mean": 3726.1, + "valid_targets_min": 3222 + }, + { + "epoch": 1.6620589305201892, + "grad_norm": 0.5491117974499979, + "learning_rate": 3.77481184684527e-05, + "loss": 0.1633, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11397881805896759, + "step": 2285, + "valid_targets_mean": 3205.8, + "valid_targets_min": 1949 + }, + { + "epoch": 1.6656966169516187, + "grad_norm": 0.30606967590085804, + "learning_rate": 3.773136983970979e-05, + "loss": 0.0958, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04392283037304878, + "step": 2290, + "valid_targets_mean": 4059.8, + "valid_targets_min": 886 + }, + { + "epoch": 1.6693343033830483, + "grad_norm": 0.5283488105853985, + "learning_rate": 3.771456290062811e-05, + "loss": 0.1151, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06183586269617081, + "step": 2295, + "valid_targets_mean": 1858.2, + "valid_targets_min": 447 + }, + { + "epoch": 1.672971989814478, + "grad_norm": 0.37217833855610777, + "learning_rate": 3.769769770647797e-05, + "loss": 0.1108, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04689063876867294, + "step": 2300, + "valid_targets_mean": 3195.6, + "valid_targets_min": 758 + }, + { + "epoch": 1.6766096762459077, + "grad_norm": 0.5104559907335485, + "learning_rate": 3.768077431272124e-05, + "loss": 0.1353, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07344531267881393, + "step": 2305, + "valid_targets_mean": 1808.4, + "valid_targets_min": 604 + }, + { + "epoch": 1.680247362677337, + "grad_norm": 0.374015412500195, + "learning_rate": 3.76637927750112e-05, + "loss": 0.1344, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.061076439917087555, + "step": 2310, + "valid_targets_mean": 3314.6, + "valid_targets_min": 1743 + }, + { + "epoch": 1.683885049108767, + "grad_norm": 0.38380798515951253, + "learning_rate": 3.764675314919234e-05, + "loss": 0.1131, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.046531111001968384, + "step": 2315, + "valid_targets_mean": 2440.4, + "valid_targets_min": 574 + }, + { + "epoch": 1.6875227355401963, + "grad_norm": 0.42245193458171654, + "learning_rate": 3.762965549130018e-05, + "loss": 0.1113, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05414202809333801, + "step": 2320, + "valid_targets_mean": 2535.6, + "valid_targets_min": 523 + }, + { + "epoch": 1.691160421971626, + "grad_norm": 0.5145047070291344, + "learning_rate": 3.761249985756104e-05, + "loss": 0.1096, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08456346392631531, + "step": 2325, + "valid_targets_mean": 3398.1, + "valid_targets_min": 2635 + }, + { + "epoch": 1.6947981084030557, + "grad_norm": 0.44320423662211966, + "learning_rate": 3.759528630439196e-05, + "loss": 0.1239, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.052373290061950684, + "step": 2330, + "valid_targets_mean": 3164.8, + "valid_targets_min": 812 + }, + { + "epoch": 1.6984357948344853, + "grad_norm": 0.401752385212484, + "learning_rate": 3.757801488840039e-05, + "loss": 0.128, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06265294551849365, + "step": 2335, + "valid_targets_mean": 2463.9, + "valid_targets_min": 856 + }, + { + "epoch": 1.7020734812659148, + "grad_norm": 0.45201763197413447, + "learning_rate": 3.756068566638412e-05, + "loss": 0.1301, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08166547864675522, + "step": 2340, + "valid_targets_mean": 2995.6, + "valid_targets_min": 718 + }, + { + "epoch": 1.7057111676973444, + "grad_norm": 0.5445025587054098, + "learning_rate": 3.754329869533099e-05, + "loss": 0.1506, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15904399752616882, + "step": 2345, + "valid_targets_mean": 2866.9, + "valid_targets_min": 1297 + }, + { + "epoch": 1.7093488541287742, + "grad_norm": 0.5039305103254105, + "learning_rate": 3.752585403241877e-05, + "loss": 0.1428, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.052191540598869324, + "step": 2350, + "valid_targets_mean": 1661.1, + "valid_targets_min": 786 + }, + { + "epoch": 1.7129865405602036, + "grad_norm": 0.6163343024650775, + "learning_rate": 3.7508351735014955e-05, + "loss": 0.1136, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07744116336107254, + "step": 2355, + "valid_targets_mean": 1624.2, + "valid_targets_min": 550 + }, + { + "epoch": 1.7166242269916334, + "grad_norm": 0.42661290707707217, + "learning_rate": 3.749079186067657e-05, + "loss": 0.1543, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05644921958446503, + "step": 2360, + "valid_targets_mean": 3254.6, + "valid_targets_min": 1304 + }, + { + "epoch": 1.7202619134230628, + "grad_norm": 0.4092778821800206, + "learning_rate": 3.747317446714998e-05, + "loss": 0.1422, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05108976364135742, + "step": 2365, + "valid_targets_mean": 2833.0, + "valid_targets_min": 1247 + }, + { + "epoch": 1.7238995998544926, + "grad_norm": 0.5392315976935373, + "learning_rate": 3.745549961237072e-05, + "loss": 0.1188, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0728115662932396, + "step": 2370, + "valid_targets_mean": 3056.5, + "valid_targets_min": 800 + }, + { + "epoch": 1.7275372862859222, + "grad_norm": 0.5703748098539458, + "learning_rate": 3.743776735446326e-05, + "loss": 0.3235, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1381148397922516, + "step": 2375, + "valid_targets_mean": 2580.8, + "valid_targets_min": 987 + }, + { + "epoch": 1.7311749727173518, + "grad_norm": 0.6367996549429168, + "learning_rate": 3.741997775174087e-05, + "loss": 0.1468, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11999819427728653, + "step": 2380, + "valid_targets_mean": 2245.6, + "valid_targets_min": 776 + }, + { + "epoch": 1.7348126591487814, + "grad_norm": 0.4115140896190236, + "learning_rate": 3.740213086270538e-05, + "loss": 0.1227, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.034233953803777695, + "step": 2385, + "valid_targets_mean": 1361.5, + "valid_targets_min": 658 + }, + { + "epoch": 1.738450345580211, + "grad_norm": 0.47797491827251565, + "learning_rate": 3.7384226746047035e-05, + "loss": 0.2306, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05631644278764725, + "step": 2390, + "valid_targets_mean": 1729.0, + "valid_targets_min": 666 + }, + { + "epoch": 1.7420880320116408, + "grad_norm": 0.2865778713532744, + "learning_rate": 3.7366265460644244e-05, + "loss": 0.1077, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04419219493865967, + "step": 2395, + "valid_targets_mean": 4975.0, + "valid_targets_min": 3529 + }, + { + "epoch": 1.7457257184430701, + "grad_norm": 0.4250694046462525, + "learning_rate": 3.734824706556344e-05, + "loss": 0.1207, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07206130027770996, + "step": 2400, + "valid_targets_mean": 3988.8, + "valid_targets_min": 1746 + }, + { + "epoch": 1.7493634048745, + "grad_norm": 0.4268170613590504, + "learning_rate": 3.7330171620058846e-05, + "loss": 0.1204, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07661005854606628, + "step": 2405, + "valid_targets_mean": 3222.4, + "valid_targets_min": 1226 + }, + { + "epoch": 1.7530010913059293, + "grad_norm": 0.38719903792268345, + "learning_rate": 3.7312039183572326e-05, + "loss": 0.1094, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07128511369228363, + "step": 2410, + "valid_targets_mean": 4403.6, + "valid_targets_min": 3275 + }, + { + "epoch": 1.756638777737359, + "grad_norm": 0.3576025725765885, + "learning_rate": 3.7293849815733134e-05, + "loss": 0.1101, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04545679688453674, + "step": 2415, + "valid_targets_mean": 2546.2, + "valid_targets_min": 723 + }, + { + "epoch": 1.7602764641687887, + "grad_norm": 0.39317118485916364, + "learning_rate": 3.727560357635774e-05, + "loss": 0.1077, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06300929188728333, + "step": 2420, + "valid_targets_mean": 3348.6, + "valid_targets_min": 2234 + }, + { + "epoch": 1.7639141506002183, + "grad_norm": 0.33903812267402533, + "learning_rate": 3.725730052544967e-05, + "loss": 0.104, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.034828782081604004, + "step": 2425, + "valid_targets_mean": 3603.0, + "valid_targets_min": 1421 + }, + { + "epoch": 1.7675518370316479, + "grad_norm": 0.41214064138967416, + "learning_rate": 3.723894072319926e-05, + "loss": 0.1081, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.058353863656520844, + "step": 2430, + "valid_targets_mean": 2849.2, + "valid_targets_min": 1068 + }, + { + "epoch": 1.7711895234630775, + "grad_norm": 0.4469108213424988, + "learning_rate": 3.7220524229983465e-05, + "loss": 0.1593, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.045006103813648224, + "step": 2435, + "valid_targets_mean": 1587.9, + "valid_targets_min": 535 + }, + { + "epoch": 1.774827209894507, + "grad_norm": 0.3691668838477204, + "learning_rate": 3.720205110636569e-05, + "loss": 0.0931, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03848590701818466, + "step": 2440, + "valid_targets_mean": 1702.4, + "valid_targets_min": 585 + }, + { + "epoch": 1.7784648963259366, + "grad_norm": 0.4474039652816741, + "learning_rate": 3.718352141309554e-05, + "loss": 0.1053, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05059252679347992, + "step": 2445, + "valid_targets_mean": 2588.6, + "valid_targets_min": 582 + }, + { + "epoch": 1.7821025827573664, + "grad_norm": 0.37074653415523934, + "learning_rate": 3.71649352111087e-05, + "loss": 0.0988, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0548858717083931, + "step": 2450, + "valid_targets_mean": 3144.0, + "valid_targets_min": 516 + }, + { + "epoch": 1.7857402691887958, + "grad_norm": 0.4708969328565874, + "learning_rate": 3.7146292561526654e-05, + "loss": 0.1598, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08324015140533447, + "step": 2455, + "valid_targets_mean": 3241.4, + "valid_targets_min": 999 + }, + { + "epoch": 1.7893779556202256, + "grad_norm": 0.5203065723245739, + "learning_rate": 3.7127593525656505e-05, + "loss": 0.1235, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10184566676616669, + "step": 2460, + "valid_targets_mean": 2021.1, + "valid_targets_min": 696 + }, + { + "epoch": 1.7930156420516552, + "grad_norm": 0.48427021539401394, + "learning_rate": 3.710883816499082e-05, + "loss": 0.1424, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06582730263471603, + "step": 2465, + "valid_targets_mean": 2048.5, + "valid_targets_min": 635 + }, + { + "epoch": 1.7966533284830848, + "grad_norm": 0.38040706052627077, + "learning_rate": 3.709002654120736e-05, + "loss": 0.1221, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05253373086452484, + "step": 2470, + "valid_targets_mean": 2641.6, + "valid_targets_min": 816 + }, + { + "epoch": 1.8002910149145144, + "grad_norm": 0.6104812931143585, + "learning_rate": 3.707115871616893e-05, + "loss": 0.1757, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17265313863754272, + "step": 2475, + "valid_targets_mean": 2236.4, + "valid_targets_min": 949 + }, + { + "epoch": 1.803928701345944, + "grad_norm": 0.39157121487771673, + "learning_rate": 3.705223475192315e-05, + "loss": 0.1299, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05986583232879639, + "step": 2480, + "valid_targets_mean": 3270.9, + "valid_targets_min": 2201 + }, + { + "epoch": 1.8075663877773736, + "grad_norm": 0.8432603543600838, + "learning_rate": 3.703325471070225e-05, + "loss": 0.1471, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10468660295009613, + "step": 2485, + "valid_targets_mean": 1500.5, + "valid_targets_min": 687 + }, + { + "epoch": 1.8112040742088031, + "grad_norm": 0.45666652647572725, + "learning_rate": 3.7014218654922884e-05, + "loss": 0.1092, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.049487486481666565, + "step": 2490, + "valid_targets_mean": 2285.5, + "valid_targets_min": 449 + }, + { + "epoch": 1.814841760640233, + "grad_norm": 0.35883839708871346, + "learning_rate": 3.699512664718589e-05, + "loss": 0.1206, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03356121852993965, + "step": 2495, + "valid_targets_mean": 2368.1, + "valid_targets_min": 658 + }, + { + "epoch": 1.8184794470716623, + "grad_norm": 0.5943469913543491, + "learning_rate": 3.6975978750276136e-05, + "loss": 0.1218, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.043844595551490784, + "step": 2500, + "valid_targets_mean": 831.8, + "valid_targets_min": 516 + }, + { + "epoch": 1.8221171335030921, + "grad_norm": 0.3423059505108621, + "learning_rate": 3.695677502716226e-05, + "loss": 0.0976, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0659906268119812, + "step": 2505, + "valid_targets_mean": 4461.9, + "valid_targets_min": 2805 + }, + { + "epoch": 1.8257548199345215, + "grad_norm": 0.32882649444701073, + "learning_rate": 3.69375155409965e-05, + "loss": 0.0942, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05426710844039917, + "step": 2510, + "valid_targets_mean": 3766.8, + "valid_targets_min": 2036 + }, + { + "epoch": 1.8293925063659513, + "grad_norm": 0.3778039978193167, + "learning_rate": 3.691820035511446e-05, + "loss": 0.1089, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04664384201169014, + "step": 2515, + "valid_targets_mean": 2336.5, + "valid_targets_min": 681 + }, + { + "epoch": 1.8330301927973809, + "grad_norm": 0.36138275468017916, + "learning_rate": 3.6898829533034926e-05, + "loss": 0.1279, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.061324190348386765, + "step": 2520, + "valid_targets_mean": 3856.1, + "valid_targets_min": 3436 + }, + { + "epoch": 1.8366678792288105, + "grad_norm": 0.5207513485096263, + "learning_rate": 3.687940313845964e-05, + "loss": 0.124, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08017132431268692, + "step": 2525, + "valid_targets_mean": 3165.5, + "valid_targets_min": 891 + }, + { + "epoch": 1.84030556566024, + "grad_norm": 0.2879470653468899, + "learning_rate": 3.685992123527311e-05, + "loss": 0.1098, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03742959350347519, + "step": 2530, + "valid_targets_mean": 3550.6, + "valid_targets_min": 2573 + }, + { + "epoch": 1.8439432520916696, + "grad_norm": 0.31340786054639913, + "learning_rate": 3.6840383887542366e-05, + "loss": 0.1037, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.049644142389297485, + "step": 2535, + "valid_targets_mean": 2995.0, + "valid_targets_min": 928 + }, + { + "epoch": 1.8475809385230995, + "grad_norm": 0.4343644280434679, + "learning_rate": 3.682079115951679e-05, + "loss": 0.1296, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07145219296216965, + "step": 2540, + "valid_targets_mean": 3839.6, + "valid_targets_min": 3392 + }, + { + "epoch": 1.8512186249545288, + "grad_norm": 0.2904139167716227, + "learning_rate": 3.680114311562785e-05, + "loss": 0.096, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.053274668753147125, + "step": 2545, + "valid_targets_mean": 3855.0, + "valid_targets_min": 2848 + }, + { + "epoch": 1.8548563113859586, + "grad_norm": 0.47048782870757444, + "learning_rate": 3.6781439820488974e-05, + "loss": 0.1171, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06100502610206604, + "step": 2550, + "valid_targets_mean": 2586.0, + "valid_targets_min": 595 + }, + { + "epoch": 1.858493997817388, + "grad_norm": 0.4381552971826222, + "learning_rate": 3.6761681338895245e-05, + "loss": 0.1191, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.061371996998786926, + "step": 2555, + "valid_targets_mean": 2471.8, + "valid_targets_min": 1016 + }, + { + "epoch": 1.8621316842488178, + "grad_norm": 0.5179846764410807, + "learning_rate": 3.6741867735823245e-05, + "loss": 0.1235, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0900430679321289, + "step": 2560, + "valid_targets_mean": 1567.8, + "valid_targets_min": 811 + }, + { + "epoch": 1.8657693706802474, + "grad_norm": 0.6316502161985346, + "learning_rate": 3.672199907643082e-05, + "loss": 0.1283, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07476092875003815, + "step": 2565, + "valid_targets_mean": 1569.1, + "valid_targets_min": 697 + }, + { + "epoch": 1.869407057111677, + "grad_norm": 0.641094857447289, + "learning_rate": 3.6702075426056875e-05, + "loss": 0.1143, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08838300406932831, + "step": 2570, + "valid_targets_mean": 1448.1, + "valid_targets_min": 736 + }, + { + "epoch": 1.8730447435431066, + "grad_norm": 0.3786507716759769, + "learning_rate": 3.6682096850221154e-05, + "loss": 0.1185, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03780680149793625, + "step": 2575, + "valid_targets_mean": 2415.5, + "valid_targets_min": 804 + }, + { + "epoch": 1.8766824299745362, + "grad_norm": 0.36316105867372167, + "learning_rate": 3.666206341462402e-05, + "loss": 0.1314, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04502163827419281, + "step": 2580, + "valid_targets_mean": 3472.5, + "valid_targets_min": 1018 + }, + { + "epoch": 1.880320116405966, + "grad_norm": 0.353256143035152, + "learning_rate": 3.664197518514626e-05, + "loss": 0.113, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05297588184475899, + "step": 2585, + "valid_targets_mean": 3623.8, + "valid_targets_min": 2369 + }, + { + "epoch": 1.8839578028373953, + "grad_norm": 0.4313801345155234, + "learning_rate": 3.662183222784883e-05, + "loss": 0.1021, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.059320539236068726, + "step": 2590, + "valid_targets_mean": 3135.8, + "valid_targets_min": 1318 + }, + { + "epoch": 1.8875954892688251, + "grad_norm": 0.49257719732815464, + "learning_rate": 3.660163460897268e-05, + "loss": 0.1241, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08133403956890106, + "step": 2595, + "valid_targets_mean": 2258.2, + "valid_targets_min": 903 + }, + { + "epoch": 1.8912331757002545, + "grad_norm": 0.3638370026779366, + "learning_rate": 3.658138239493851e-05, + "loss": 0.121, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.053712908178567886, + "step": 2600, + "valid_targets_mean": 3460.4, + "valid_targets_min": 2346 + }, + { + "epoch": 1.8948708621316843, + "grad_norm": 0.5925691584160387, + "learning_rate": 3.656107565234656e-05, + "loss": 0.1029, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09183934330940247, + "step": 2605, + "valid_targets_mean": 2283.0, + "valid_targets_min": 737 + }, + { + "epoch": 1.898508548563114, + "grad_norm": 0.3734072806702179, + "learning_rate": 3.6540714447976385e-05, + "loss": 0.1063, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.052989281713962555, + "step": 2610, + "valid_targets_mean": 3979.6, + "valid_targets_min": 2555 + }, + { + "epoch": 1.9021462349945435, + "grad_norm": 0.3066375839694084, + "learning_rate": 3.652029884878665e-05, + "loss": 0.1026, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047608330845832825, + "step": 2615, + "valid_targets_mean": 4500.2, + "valid_targets_min": 2991 + }, + { + "epoch": 1.905783921425973, + "grad_norm": 0.4009905277799659, + "learning_rate": 3.649982892191488e-05, + "loss": 0.0894, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03711732476949692, + "step": 2620, + "valid_targets_mean": 1992.4, + "valid_targets_min": 618 + }, + { + "epoch": 1.9094216078574027, + "grad_norm": 0.31797810687316186, + "learning_rate": 3.647930473467728e-05, + "loss": 0.1169, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03527629002928734, + "step": 2625, + "valid_targets_mean": 3641.8, + "valid_targets_min": 2944 + }, + { + "epoch": 1.9130592942888323, + "grad_norm": 0.3481385776243829, + "learning_rate": 3.6458726354568474e-05, + "loss": 0.1036, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0628269612789154, + "step": 2630, + "valid_targets_mean": 3686.1, + "valid_targets_min": 2702 + }, + { + "epoch": 1.9166969807202618, + "grad_norm": 0.3339853978403863, + "learning_rate": 3.643809384926133e-05, + "loss": 0.1352, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.028807982802391052, + "step": 2635, + "valid_targets_mean": 3341.6, + "valid_targets_min": 509 + }, + { + "epoch": 1.9203346671516917, + "grad_norm": 0.33336604734014724, + "learning_rate": 3.641740728660667e-05, + "loss": 0.1236, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.042243607342243195, + "step": 2640, + "valid_targets_mean": 2757.6, + "valid_targets_min": 637 + }, + { + "epoch": 1.923972353583121, + "grad_norm": 0.4356280875769784, + "learning_rate": 3.6396666734633115e-05, + "loss": 0.1375, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06324812024831772, + "step": 2645, + "valid_targets_mean": 2802.6, + "valid_targets_min": 737 + }, + { + "epoch": 1.9276100400145508, + "grad_norm": 0.31381580858155406, + "learning_rate": 3.637587226154683e-05, + "loss": 0.0982, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04663381725549698, + "step": 2650, + "valid_targets_mean": 3312.9, + "valid_targets_min": 1292 + }, + { + "epoch": 1.9312477264459802, + "grad_norm": 0.5358192387874385, + "learning_rate": 3.635502393573127e-05, + "loss": 0.1219, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.059162333607673645, + "step": 2655, + "valid_targets_mean": 1049.6, + "valid_targets_min": 697 + }, + { + "epoch": 1.93488541287741, + "grad_norm": 0.345385324336317, + "learning_rate": 3.633412182574704e-05, + "loss": 0.106, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04901306703686714, + "step": 2660, + "valid_targets_mean": 3541.4, + "valid_targets_min": 923 + }, + { + "epoch": 1.9385230993088396, + "grad_norm": 0.7848332560145095, + "learning_rate": 3.631316600033158e-05, + "loss": 0.1563, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08908266574144363, + "step": 2665, + "valid_targets_mean": 931.6, + "valid_targets_min": 531 + }, + { + "epoch": 1.9421607857402692, + "grad_norm": 0.5082442118515118, + "learning_rate": 3.629215652839898e-05, + "loss": 0.1456, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07691565901041031, + "step": 2670, + "valid_targets_mean": 1840.5, + "valid_targets_min": 844 + }, + { + "epoch": 1.9457984721716988, + "grad_norm": 0.44285333654170217, + "learning_rate": 3.627109347903974e-05, + "loss": 0.1118, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07734958082437515, + "step": 2675, + "valid_targets_mean": 3893.9, + "valid_targets_min": 2967 + }, + { + "epoch": 1.9494361586031284, + "grad_norm": 0.35530500816494354, + "learning_rate": 3.624997692152059e-05, + "loss": 0.1097, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05925382673740387, + "step": 2680, + "valid_targets_mean": 3605.9, + "valid_targets_min": 2490 + }, + { + "epoch": 1.9530738450345582, + "grad_norm": 0.35342985666445126, + "learning_rate": 3.622880692528417e-05, + "loss": 0.11, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04738998785614967, + "step": 2685, + "valid_targets_mean": 2682.0, + "valid_targets_min": 970 + }, + { + "epoch": 1.9567115314659875, + "grad_norm": 0.38756439271629023, + "learning_rate": 3.620758355994891e-05, + "loss": 0.0999, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.043250150978565216, + "step": 2690, + "valid_targets_mean": 2028.5, + "valid_targets_min": 729 + }, + { + "epoch": 1.9603492178974173, + "grad_norm": 0.8031943054422388, + "learning_rate": 3.61863068953087e-05, + "loss": 0.133, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09351123124361038, + "step": 2695, + "valid_targets_mean": 1028.9, + "valid_targets_min": 637 + }, + { + "epoch": 1.9639869043288467, + "grad_norm": 0.35202340139149235, + "learning_rate": 3.616497700133274e-05, + "loss": 0.1027, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03728501498699188, + "step": 2700, + "valid_targets_mean": 2879.2, + "valid_targets_min": 701 + }, + { + "epoch": 1.9676245907602765, + "grad_norm": 0.34285120977351874, + "learning_rate": 3.614359394816526e-05, + "loss": 0.1062, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0458427295088768, + "step": 2705, + "valid_targets_mean": 3671.1, + "valid_targets_min": 3131 + }, + { + "epoch": 1.971262277191706, + "grad_norm": 0.39451496493575217, + "learning_rate": 3.612215780612529e-05, + "loss": 0.1099, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06010882556438446, + "step": 2710, + "valid_targets_mean": 3241.5, + "valid_targets_min": 2018 + }, + { + "epoch": 1.9748999636231357, + "grad_norm": 0.3234935955934487, + "learning_rate": 3.610066864570649e-05, + "loss": 0.1132, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04774092137813568, + "step": 2715, + "valid_targets_mean": 2491.6, + "valid_targets_min": 350 + }, + { + "epoch": 1.9785376500545653, + "grad_norm": 0.27062408869137555, + "learning_rate": 3.6079126537576835e-05, + "loss": 0.095, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0524631142616272, + "step": 2720, + "valid_targets_mean": 4862.8, + "valid_targets_min": 3577 + }, + { + "epoch": 1.9821753364859949, + "grad_norm": 0.3495067308525934, + "learning_rate": 3.605753155257844e-05, + "loss": 0.1068, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0415499173104763, + "step": 2725, + "valid_targets_mean": 1999.4, + "valid_targets_min": 507 + }, + { + "epoch": 1.9858130229174247, + "grad_norm": 0.4487830616698078, + "learning_rate": 3.603588376172729e-05, + "loss": 0.12, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10398781299591064, + "step": 2730, + "valid_targets_mean": 3278.5, + "valid_targets_min": 2814 + }, + { + "epoch": 1.989450709348854, + "grad_norm": 0.3957327814518319, + "learning_rate": 3.601418323621305e-05, + "loss": 0.1927, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05769849941134453, + "step": 2735, + "valid_targets_mean": 4006.2, + "valid_targets_min": 2347 + }, + { + "epoch": 1.9930883957802839, + "grad_norm": 0.5195699661731401, + "learning_rate": 3.5992430047398785e-05, + "loss": 0.1092, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06545413285493851, + "step": 2740, + "valid_targets_mean": 2447.4, + "valid_targets_min": 716 + }, + { + "epoch": 1.9967260822117132, + "grad_norm": 0.3676449443794425, + "learning_rate": 3.597062426682074e-05, + "loss": 0.1034, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0491764098405838, + "step": 2745, + "valid_targets_mean": 3209.6, + "valid_targets_min": 2245 + }, + { + "epoch": 2.0, + "grad_norm": 0.5160642338367951, + "learning_rate": 3.5948765966188136e-05, + "loss": 0.1556, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19970330595970154, + "step": 2750, + "valid_targets_mean": 7922.2, + "valid_targets_min": 3513 + }, + { + "epoch": 2.00363768643143, + "grad_norm": 0.3751895263374549, + "learning_rate": 3.592685521738289e-05, + "loss": 0.2278, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10828530788421631, + "step": 2755, + "valid_targets_mean": 7389.0, + "valid_targets_min": 5627 + }, + { + "epoch": 2.007275372862859, + "grad_norm": 0.5826674070022475, + "learning_rate": 3.59048920924594e-05, + "loss": 0.2302, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06462503969669342, + "step": 2760, + "valid_targets_mean": 1294.8, + "valid_targets_min": 422 + }, + { + "epoch": 2.010913059294289, + "grad_norm": 0.3531067036091161, + "learning_rate": 3.588287666364429e-05, + "loss": 0.2202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1121969148516655, + "step": 2765, + "valid_targets_mean": 8539.9, + "valid_targets_min": 6761 + }, + { + "epoch": 2.0145507457257184, + "grad_norm": 0.38931784353922894, + "learning_rate": 3.586080900333623e-05, + "loss": 0.2098, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09733777493238449, + "step": 2770, + "valid_targets_mean": 7579.0, + "valid_targets_min": 6174 + }, + { + "epoch": 2.018188432157148, + "grad_norm": 0.38269804232575294, + "learning_rate": 3.583868918410561e-05, + "loss": 0.2129, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09734043478965759, + "step": 2775, + "valid_targets_mean": 7328.2, + "valid_targets_min": 4839 + }, + { + "epoch": 2.0218261185885775, + "grad_norm": 0.4188762564531471, + "learning_rate": 3.581651727869436e-05, + "loss": 0.2211, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11577893048524857, + "step": 2780, + "valid_targets_mean": 6838.2, + "valid_targets_min": 5034 + }, + { + "epoch": 2.0254638050200073, + "grad_norm": 0.3926179202590386, + "learning_rate": 3.5794293360015726e-05, + "loss": 0.228, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11768753826618195, + "step": 2785, + "valid_targets_mean": 7248.2, + "valid_targets_min": 5567 + }, + { + "epoch": 2.0291014914514367, + "grad_norm": 0.3711348606496126, + "learning_rate": 3.577201750115396e-05, + "loss": 0.2123, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09730631113052368, + "step": 2790, + "valid_targets_mean": 6234.0, + "valid_targets_min": 5603 + }, + { + "epoch": 2.0327391778828665, + "grad_norm": 0.7041685933498996, + "learning_rate": 3.574968977536414e-05, + "loss": 0.2111, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03259681165218353, + "step": 2795, + "valid_targets_mean": 602.0, + "valid_targets_min": 152 + }, + { + "epoch": 2.0363768643142963, + "grad_norm": 0.3734273743022818, + "learning_rate": 3.572731025607192e-05, + "loss": 0.2011, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10002113878726959, + "step": 2800, + "valid_targets_mean": 7299.9, + "valid_targets_min": 6158 + }, + { + "epoch": 2.0400145507457257, + "grad_norm": 0.4042814070606988, + "learning_rate": 3.5704879016873256e-05, + "loss": 0.2135, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09897083789110184, + "step": 2805, + "valid_targets_mean": 6140.0, + "valid_targets_min": 4648 + }, + { + "epoch": 2.0436522371771555, + "grad_norm": 0.4010099596620694, + "learning_rate": 3.568239613153421e-05, + "loss": 0.2058, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10558643937110901, + "step": 2810, + "valid_targets_mean": 6603.0, + "valid_targets_min": 4882 + }, + { + "epoch": 2.047289923608585, + "grad_norm": 0.3669301345715097, + "learning_rate": 3.5659861673990676e-05, + "loss": 0.2065, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10513758659362793, + "step": 2815, + "valid_targets_mean": 8015.0, + "valid_targets_min": 6344 + }, + { + "epoch": 2.0509276100400147, + "grad_norm": 0.41172325138509525, + "learning_rate": 3.563727571834814e-05, + "loss": 0.1977, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09315542876720428, + "step": 2820, + "valid_targets_mean": 6305.2, + "valid_targets_min": 4622 + }, + { + "epoch": 2.054565296471444, + "grad_norm": 0.37945208884912607, + "learning_rate": 3.561463833888145e-05, + "loss": 0.2063, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10794265568256378, + "step": 2825, + "valid_targets_mean": 7890.8, + "valid_targets_min": 5485 + }, + { + "epoch": 2.058202982902874, + "grad_norm": 0.38858190428217343, + "learning_rate": 3.559194961003455e-05, + "loss": 0.2074, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09541163593530655, + "step": 2830, + "valid_targets_mean": 7016.6, + "valid_targets_min": 4960 + }, + { + "epoch": 2.061840669334303, + "grad_norm": 0.39090149045439754, + "learning_rate": 3.556920960642027e-05, + "loss": 0.2032, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0977669432759285, + "step": 2835, + "valid_targets_mean": 6527.1, + "valid_targets_min": 5399 + }, + { + "epoch": 2.065478355765733, + "grad_norm": 0.4435294732856552, + "learning_rate": 3.554641840282004e-05, + "loss": 0.2106, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10202648490667343, + "step": 2840, + "valid_targets_mean": 4995.0, + "valid_targets_min": 618 + }, + { + "epoch": 2.0691160421971624, + "grad_norm": 0.3941591279483701, + "learning_rate": 3.5523576074183675e-05, + "loss": 0.2224, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10992235690355301, + "step": 2845, + "valid_targets_mean": 7219.4, + "valid_targets_min": 5764 + }, + { + "epoch": 2.072753728628592, + "grad_norm": 0.37156074890364443, + "learning_rate": 3.5500682695629096e-05, + "loss": 0.1983, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10518870502710342, + "step": 2850, + "valid_targets_mean": 7998.5, + "valid_targets_min": 5902 + }, + { + "epoch": 2.076391415060022, + "grad_norm": 0.3585474254069277, + "learning_rate": 3.547773834244214e-05, + "loss": 0.1917, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.083298459649086, + "step": 2855, + "valid_targets_mean": 7377.4, + "valid_targets_min": 5015 + }, + { + "epoch": 2.0800291014914514, + "grad_norm": 0.37173976620380944, + "learning_rate": 3.5454743090076244e-05, + "loss": 0.2018, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08512729406356812, + "step": 2860, + "valid_targets_mean": 6794.0, + "valid_targets_min": 5283 + }, + { + "epoch": 2.083666787922881, + "grad_norm": 0.39995580556698357, + "learning_rate": 3.543169701415225e-05, + "loss": 0.2104, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11859605461359024, + "step": 2865, + "valid_targets_mean": 6656.4, + "valid_targets_min": 4364 + }, + { + "epoch": 2.0873044743543105, + "grad_norm": 0.38771382659315395, + "learning_rate": 3.5408600190458115e-05, + "loss": 0.2067, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09514424949884415, + "step": 2870, + "valid_targets_mean": 7178.4, + "valid_targets_min": 4448 + }, + { + "epoch": 2.0909421607857404, + "grad_norm": 0.45418774734185147, + "learning_rate": 3.5385452694948694e-05, + "loss": 0.2059, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12124840915203094, + "step": 2875, + "valid_targets_mean": 7619.8, + "valid_targets_min": 5737 + }, + { + "epoch": 2.0945798472171697, + "grad_norm": 0.33912215416738634, + "learning_rate": 3.5362254603745494e-05, + "loss": 0.182, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08616577833890915, + "step": 2880, + "valid_targets_mean": 8754.4, + "valid_targets_min": 5752 + }, + { + "epoch": 2.0982175336485995, + "grad_norm": 0.40772882200969546, + "learning_rate": 3.5339005993136375e-05, + "loss": 0.1827, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09513247758150101, + "step": 2885, + "valid_targets_mean": 6341.0, + "valid_targets_min": 4543 + }, + { + "epoch": 2.101855220080029, + "grad_norm": 0.39675562240894374, + "learning_rate": 3.531570693957536e-05, + "loss": 0.1773, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10245552659034729, + "step": 2890, + "valid_targets_mean": 7181.9, + "valid_targets_min": 4563 + }, + { + "epoch": 2.1054929065114587, + "grad_norm": 0.7429960383973093, + "learning_rate": 3.529235751968233e-05, + "loss": 0.2155, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07499866932630539, + "step": 2895, + "valid_targets_mean": 1104.4, + "valid_targets_min": 161 + }, + { + "epoch": 2.1091305929428885, + "grad_norm": 0.37152049962987366, + "learning_rate": 3.526895781024282e-05, + "loss": 0.201, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09147954732179642, + "step": 2900, + "valid_targets_mean": 7018.2, + "valid_targets_min": 5424 + }, + { + "epoch": 2.112768279374318, + "grad_norm": 0.37694135196696443, + "learning_rate": 3.5245507888207746e-05, + "loss": 0.2132, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10887692868709564, + "step": 2905, + "valid_targets_mean": 7738.4, + "valid_targets_min": 5329 + }, + { + "epoch": 2.1164059658057477, + "grad_norm": 0.4747758160277102, + "learning_rate": 3.522200783069313e-05, + "loss": 0.2168, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10434703528881073, + "step": 2910, + "valid_targets_mean": 6426.6, + "valid_targets_min": 5560 + }, + { + "epoch": 2.120043652237177, + "grad_norm": 0.4034668124583497, + "learning_rate": 3.5198457714979875e-05, + "loss": 0.2111, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10518841445446014, + "step": 2915, + "valid_targets_mean": 6169.9, + "valid_targets_min": 3446 + }, + { + "epoch": 2.123681338668607, + "grad_norm": 0.5777227537638882, + "learning_rate": 3.517485761851352e-05, + "loss": 0.2042, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1060957983136177, + "step": 2920, + "valid_targets_mean": 7204.6, + "valid_targets_min": 5381 + }, + { + "epoch": 2.1273190251000362, + "grad_norm": 0.3751876840996291, + "learning_rate": 3.515120761890393e-05, + "loss": 0.2093, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10816653072834015, + "step": 2925, + "valid_targets_mean": 7600.4, + "valid_targets_min": 5241 + }, + { + "epoch": 2.130956711531466, + "grad_norm": 0.39897645916454916, + "learning_rate": 3.5127507793925124e-05, + "loss": 0.2031, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11103267222642899, + "step": 2930, + "valid_targets_mean": 7098.5, + "valid_targets_min": 5607 + }, + { + "epoch": 2.1345943979628954, + "grad_norm": 0.42058533989124924, + "learning_rate": 3.510375822151494e-05, + "loss": 0.209, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10965196788311005, + "step": 2935, + "valid_targets_mean": 6955.1, + "valid_targets_min": 5530 + }, + { + "epoch": 2.138232084394325, + "grad_norm": 0.4196500406075333, + "learning_rate": 3.5079958979774816e-05, + "loss": 0.2076, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.105963334441185, + "step": 2940, + "valid_targets_mean": 6748.1, + "valid_targets_min": 5364 + }, + { + "epoch": 2.141869770825755, + "grad_norm": 0.4546728164954694, + "learning_rate": 3.505611014696955e-05, + "loss": 0.2088, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10127680003643036, + "step": 2945, + "valid_targets_mean": 6273.9, + "valid_targets_min": 5438 + }, + { + "epoch": 2.1455074572571844, + "grad_norm": 0.39734574153677504, + "learning_rate": 3.503221180152701e-05, + "loss": 0.2059, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09925577044487, + "step": 2950, + "valid_targets_mean": 6419.5, + "valid_targets_min": 4879 + }, + { + "epoch": 2.149145143688614, + "grad_norm": 0.40197757388688693, + "learning_rate": 3.5008264022037876e-05, + "loss": 0.2037, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09937325119972229, + "step": 2955, + "valid_targets_mean": 6483.4, + "valid_targets_min": 4309 + }, + { + "epoch": 2.1527828301200436, + "grad_norm": 0.46700652136294263, + "learning_rate": 3.4984266887255415e-05, + "loss": 0.191, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11931593716144562, + "step": 2960, + "valid_targets_mean": 6825.0, + "valid_targets_min": 5205 + }, + { + "epoch": 2.1564205165514734, + "grad_norm": 0.4344835378043676, + "learning_rate": 3.496022047609518e-05, + "loss": 0.2029, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09423287957906723, + "step": 2965, + "valid_targets_mean": 6729.4, + "valid_targets_min": 5200 + }, + { + "epoch": 2.1600582029829027, + "grad_norm": 0.39825161449180846, + "learning_rate": 3.493612486763479e-05, + "loss": 0.2005, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09861066192388535, + "step": 2970, + "valid_targets_mean": 6094.0, + "valid_targets_min": 5302 + }, + { + "epoch": 2.1636958894143326, + "grad_norm": 0.41054246563502006, + "learning_rate": 3.4911980141113655e-05, + "loss": 0.2061, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10803300142288208, + "step": 2975, + "valid_targets_mean": 7140.0, + "valid_targets_min": 5270 + }, + { + "epoch": 2.167333575845762, + "grad_norm": 0.4152252234309137, + "learning_rate": 3.488778637593269e-05, + "loss": 0.1978, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09570389986038208, + "step": 2980, + "valid_targets_mean": 7425.0, + "valid_targets_min": 5281 + }, + { + "epoch": 2.1709712622771917, + "grad_norm": 0.44032841170031295, + "learning_rate": 3.486354365165408e-05, + "loss": 0.2067, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12877503037452698, + "step": 2985, + "valid_targets_mean": 7063.4, + "valid_targets_min": 5386 + }, + { + "epoch": 2.174608948708621, + "grad_norm": 0.40141557522398436, + "learning_rate": 3.4839252048001024e-05, + "loss": 0.2056, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10673506557941437, + "step": 2990, + "valid_targets_mean": 6827.0, + "valid_targets_min": 4547 + }, + { + "epoch": 2.178246635140051, + "grad_norm": 0.5953727865229269, + "learning_rate": 3.481491164485747e-05, + "loss": 0.1878, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.061618637293577194, + "step": 2995, + "valid_targets_mean": 2047.0, + "valid_targets_min": 165 + }, + { + "epoch": 2.1818843215714807, + "grad_norm": 0.3956980771227305, + "learning_rate": 3.479052252226783e-05, + "loss": 0.1947, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09779931604862213, + "step": 3000, + "valid_targets_mean": 6754.1, + "valid_targets_min": 5196 + }, + { + "epoch": 2.18552200800291, + "grad_norm": 0.42787706332713554, + "learning_rate": 3.476608476043674e-05, + "loss": 0.1986, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10137054324150085, + "step": 3005, + "valid_targets_mean": 5985.6, + "valid_targets_min": 5028 + }, + { + "epoch": 2.18915969443434, + "grad_norm": 0.4157218481917964, + "learning_rate": 3.474159843972877e-05, + "loss": 0.2056, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09290023148059845, + "step": 3010, + "valid_targets_mean": 5963.9, + "valid_targets_min": 5237 + }, + { + "epoch": 2.1927973808657693, + "grad_norm": 0.4518213760699321, + "learning_rate": 3.4717063640668195e-05, + "loss": 0.1998, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10422919690608978, + "step": 3015, + "valid_targets_mean": 6319.0, + "valid_targets_min": 5412 + }, + { + "epoch": 2.196435067297199, + "grad_norm": 0.3656262263402965, + "learning_rate": 3.469248044393873e-05, + "loss": 0.1904, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08172112703323364, + "step": 3020, + "valid_targets_mean": 7125.8, + "valid_targets_min": 5080 + }, + { + "epoch": 2.2000727537286284, + "grad_norm": 0.4281659573236867, + "learning_rate": 3.4667848930383203e-05, + "loss": 0.2032, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10026248544454575, + "step": 3025, + "valid_targets_mean": 6123.5, + "valid_targets_min": 4521 + }, + { + "epoch": 2.2037104401600582, + "grad_norm": 0.6081512845710995, + "learning_rate": 3.464316918100338e-05, + "loss": 0.2466, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15777519345283508, + "step": 3030, + "valid_targets_mean": 6020.6, + "valid_targets_min": 2493 + }, + { + "epoch": 2.207348126591488, + "grad_norm": 0.5691960225880356, + "learning_rate": 3.461844127695962e-05, + "loss": 0.2871, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11328042298555374, + "step": 3035, + "valid_targets_mean": 3953.5, + "valid_targets_min": 1177 + }, + { + "epoch": 2.2109858130229174, + "grad_norm": 0.5504067272920591, + "learning_rate": 3.4593665299570656e-05, + "loss": 0.2923, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17405429482460022, + "step": 3040, + "valid_targets_mean": 4950.8, + "valid_targets_min": 2799 + }, + { + "epoch": 2.214623499454347, + "grad_norm": 0.5467732354335214, + "learning_rate": 3.456884133031331e-05, + "loss": 0.2777, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15607950091362, + "step": 3045, + "valid_targets_mean": 5312.0, + "valid_targets_min": 2973 + }, + { + "epoch": 2.2182611858857766, + "grad_norm": 0.5439810016786308, + "learning_rate": 3.4543969450822216e-05, + "loss": 0.276, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1296633780002594, + "step": 3050, + "valid_targets_mean": 3775.5, + "valid_targets_min": 980 + }, + { + "epoch": 2.2218988723172064, + "grad_norm": 0.5062206186936642, + "learning_rate": 3.451904974288958e-05, + "loss": 0.2957, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15268778800964355, + "step": 3055, + "valid_targets_mean": 5055.4, + "valid_targets_min": 1512 + }, + { + "epoch": 2.2255365587486358, + "grad_norm": 0.48865423773198485, + "learning_rate": 3.4494082288464886e-05, + "loss": 0.2695, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10309544205665588, + "step": 3060, + "valid_targets_mean": 3600.8, + "valid_targets_min": 1848 + }, + { + "epoch": 2.2291742451800656, + "grad_norm": 0.48585353119085634, + "learning_rate": 3.4469067169654636e-05, + "loss": 0.2787, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10252232104539871, + "step": 3065, + "valid_targets_mean": 3369.8, + "valid_targets_min": 1381 + }, + { + "epoch": 2.232811931611495, + "grad_norm": 0.4761443975216412, + "learning_rate": 3.444400446872208e-05, + "loss": 0.2788, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14496013522148132, + "step": 3070, + "valid_targets_mean": 6283.2, + "valid_targets_min": 3600 + }, + { + "epoch": 2.2364496180429247, + "grad_norm": 0.5504267776883188, + "learning_rate": 3.4418894268086924e-05, + "loss": 0.2908, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14312861859798431, + "step": 3075, + "valid_targets_mean": 5242.5, + "valid_targets_min": 2540 + }, + { + "epoch": 2.240087304474354, + "grad_norm": 0.48164721609218153, + "learning_rate": 3.439373665032511e-05, + "loss": 0.2772, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12718579173088074, + "step": 3080, + "valid_targets_mean": 6246.5, + "valid_targets_min": 2157 + }, + { + "epoch": 2.243724990905784, + "grad_norm": 0.5570108109195289, + "learning_rate": 3.43685316981685e-05, + "loss": 0.2808, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1429150104522705, + "step": 3085, + "valid_targets_mean": 4556.5, + "valid_targets_min": 1720 + }, + { + "epoch": 2.2473626773372137, + "grad_norm": 0.5524332323600892, + "learning_rate": 3.434327949450461e-05, + "loss": 0.2834, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.156133770942688, + "step": 3090, + "valid_targets_mean": 5116.5, + "valid_targets_min": 1982 + }, + { + "epoch": 2.251000363768643, + "grad_norm": 0.5293165452648788, + "learning_rate": 3.431798012237634e-05, + "loss": 0.2803, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14376100897789001, + "step": 3095, + "valid_targets_mean": 4464.5, + "valid_targets_min": 2718 + }, + { + "epoch": 2.254638050200073, + "grad_norm": 0.5914086418526504, + "learning_rate": 3.429263366498173e-05, + "loss": 0.285, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1391942799091339, + "step": 3100, + "valid_targets_mean": 3418.5, + "valid_targets_min": 1451 + }, + { + "epoch": 2.2582757366315023, + "grad_norm": 0.5747026086074599, + "learning_rate": 3.426724020567365e-05, + "loss": 0.2837, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14594978094100952, + "step": 3105, + "valid_targets_mean": 3990.4, + "valid_targets_min": 1861 + }, + { + "epoch": 2.261913423062932, + "grad_norm": 0.6285910208281027, + "learning_rate": 3.424179982795952e-05, + "loss": 0.2773, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15464632213115692, + "step": 3110, + "valid_targets_mean": 4065.4, + "valid_targets_min": 2100 + }, + { + "epoch": 2.2655511094943614, + "grad_norm": 0.5867101629368412, + "learning_rate": 3.4216312615501085e-05, + "loss": 0.2888, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1599251925945282, + "step": 3115, + "valid_targets_mean": 4132.4, + "valid_targets_min": 3234 + }, + { + "epoch": 2.2691887959257913, + "grad_norm": 0.6575258721741737, + "learning_rate": 3.4190778652114076e-05, + "loss": 0.2948, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1787017285823822, + "step": 3120, + "valid_targets_mean": 4246.2, + "valid_targets_min": 2650 + }, + { + "epoch": 2.2728264823572206, + "grad_norm": 0.6176294055453587, + "learning_rate": 3.416519802176799e-05, + "loss": 0.2846, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12015119194984436, + "step": 3125, + "valid_targets_mean": 4096.5, + "valid_targets_min": 1719 + }, + { + "epoch": 2.2764641687886504, + "grad_norm": 0.5964385030953738, + "learning_rate": 3.4139570808585775e-05, + "loss": 0.2904, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14292794466018677, + "step": 3130, + "valid_targets_mean": 3933.1, + "valid_targets_min": 1020 + }, + { + "epoch": 2.28010185522008, + "grad_norm": 0.6664910118569715, + "learning_rate": 3.4113897096843585e-05, + "loss": 0.271, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14215406775474548, + "step": 3135, + "valid_targets_mean": 3551.1, + "valid_targets_min": 2130 + }, + { + "epoch": 2.2837395416515096, + "grad_norm": 0.5436575882438567, + "learning_rate": 3.408817697097047e-05, + "loss": 0.2686, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11503466963768005, + "step": 3140, + "valid_targets_mean": 3535.0, + "valid_targets_min": 865 + }, + { + "epoch": 2.2873772280829394, + "grad_norm": 0.6776919862710449, + "learning_rate": 3.406241051554813e-05, + "loss": 0.2751, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18645867705345154, + "step": 3145, + "valid_targets_mean": 3716.5, + "valid_targets_min": 3144 + }, + { + "epoch": 2.291014914514369, + "grad_norm": 0.591950325008901, + "learning_rate": 3.403659781531059e-05, + "loss": 0.2841, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12052954733371735, + "step": 3150, + "valid_targets_mean": 3262.2, + "valid_targets_min": 2050 + }, + { + "epoch": 2.2946526009457986, + "grad_norm": 0.5682429517500525, + "learning_rate": 3.401073895514401e-05, + "loss": 0.2829, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14154580235481262, + "step": 3155, + "valid_targets_mean": 3296.5, + "valid_targets_min": 1570 + }, + { + "epoch": 2.298290287377228, + "grad_norm": 0.5575148567073102, + "learning_rate": 3.398483402008629e-05, + "loss": 0.2685, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1370742917060852, + "step": 3160, + "valid_targets_mean": 4806.4, + "valid_targets_min": 1795 + }, + { + "epoch": 2.3019279738086578, + "grad_norm": 0.612009613868104, + "learning_rate": 3.395888309532687e-05, + "loss": 0.2725, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.126902773976326, + "step": 3165, + "valid_targets_mean": 4342.4, + "valid_targets_min": 1934 + }, + { + "epoch": 2.305565660240087, + "grad_norm": 0.7100250818864025, + "learning_rate": 3.393288626620645e-05, + "loss": 0.2771, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13965743780136108, + "step": 3170, + "valid_targets_mean": 3675.4, + "valid_targets_min": 2557 + }, + { + "epoch": 2.309203346671517, + "grad_norm": 0.6296930327693104, + "learning_rate": 3.3906843618216676e-05, + "loss": 0.277, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15611594915390015, + "step": 3175, + "valid_targets_mean": 4550.4, + "valid_targets_min": 2663 + }, + { + "epoch": 2.3128410331029468, + "grad_norm": 0.5796235228197228, + "learning_rate": 3.388075523699985e-05, + "loss": 0.2683, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13037574291229248, + "step": 3180, + "valid_targets_mean": 4604.0, + "valid_targets_min": 2016 + }, + { + "epoch": 2.316478719534376, + "grad_norm": 0.6334575251964438, + "learning_rate": 3.3854621208348696e-05, + "loss": 0.2759, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12338130921125412, + "step": 3185, + "valid_targets_mean": 2976.8, + "valid_targets_min": 1367 + }, + { + "epoch": 2.320116405965806, + "grad_norm": 0.6508436961281823, + "learning_rate": 3.3828441618206036e-05, + "loss": 0.2811, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13523150980472565, + "step": 3190, + "valid_targets_mean": 3542.5, + "valid_targets_min": 2026 + }, + { + "epoch": 2.3237540923972353, + "grad_norm": 0.6011106928993551, + "learning_rate": 3.3802216552664526e-05, + "loss": 0.2656, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1254008412361145, + "step": 3195, + "valid_targets_mean": 3247.2, + "valid_targets_min": 2033 + }, + { + "epoch": 2.327391778828665, + "grad_norm": 0.7350673287237015, + "learning_rate": 3.377594609796638e-05, + "loss": 0.2817, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13893890380859375, + "step": 3200, + "valid_targets_mean": 2201.6, + "valid_targets_min": 1283 + }, + { + "epoch": 2.3310294652600945, + "grad_norm": 0.6735113196246326, + "learning_rate": 3.374963034050307e-05, + "loss": 0.2774, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14697414636611938, + "step": 3205, + "valid_targets_mean": 2824.8, + "valid_targets_min": 1636 + }, + { + "epoch": 2.3346671516915243, + "grad_norm": 0.6247647986422843, + "learning_rate": 3.372326936681504e-05, + "loss": 0.2768, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.149931401014328, + "step": 3210, + "valid_targets_mean": 3844.4, + "valid_targets_min": 1629 + }, + { + "epoch": 2.3383048381229536, + "grad_norm": 0.6078420538838527, + "learning_rate": 3.369686326359144e-05, + "loss": 0.271, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12539994716644287, + "step": 3215, + "valid_targets_mean": 3048.6, + "valid_targets_min": 2080 + }, + { + "epoch": 2.3419425245543835, + "grad_norm": 0.6504939312075837, + "learning_rate": 3.3670412117669825e-05, + "loss": 0.2724, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1262730360031128, + "step": 3220, + "valid_targets_mean": 3111.8, + "valid_targets_min": 1153 + }, + { + "epoch": 2.345580210985813, + "grad_norm": 0.5859602206325825, + "learning_rate": 3.3643916016035874e-05, + "loss": 0.2727, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14854149520397186, + "step": 3225, + "valid_targets_mean": 4069.0, + "valid_targets_min": 1858 + }, + { + "epoch": 2.3492178974172426, + "grad_norm": 0.5676332468042427, + "learning_rate": 3.36173750458231e-05, + "loss": 0.2819, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11702656745910645, + "step": 3230, + "valid_targets_mean": 3415.2, + "valid_targets_min": 1639 + }, + { + "epoch": 2.3528555838486724, + "grad_norm": 0.6621362989981414, + "learning_rate": 3.359078929431259e-05, + "loss": 0.2742, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11234873533248901, + "step": 3235, + "valid_targets_mean": 3267.6, + "valid_targets_min": 1394 + }, + { + "epoch": 2.356493270280102, + "grad_norm": 0.5435662423921767, + "learning_rate": 3.356415884893268e-05, + "loss": 0.2703, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1432414948940277, + "step": 3240, + "valid_targets_mean": 4775.9, + "valid_targets_min": 1844 + }, + { + "epoch": 2.3601309567115316, + "grad_norm": 0.6069379143260903, + "learning_rate": 3.3537483797258664e-05, + "loss": 0.2741, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12383782863616943, + "step": 3245, + "valid_targets_mean": 3317.1, + "valid_targets_min": 1305 + }, + { + "epoch": 2.363768643142961, + "grad_norm": 0.7068484005160693, + "learning_rate": 3.3510764227012574e-05, + "loss": 0.2702, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14631207287311554, + "step": 3250, + "valid_targets_mean": 4290.8, + "valid_targets_min": 2185 + }, + { + "epoch": 2.367406329574391, + "grad_norm": 0.6241749823406223, + "learning_rate": 3.3484000226062794e-05, + "loss": 0.2731, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16094648838043213, + "step": 3255, + "valid_targets_mean": 4024.9, + "valid_targets_min": 2349 + }, + { + "epoch": 2.37104401600582, + "grad_norm": 0.5728439051387307, + "learning_rate": 3.345719188242385e-05, + "loss": 0.2613, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11801877617835999, + "step": 3260, + "valid_targets_mean": 3453.0, + "valid_targets_min": 1607 + }, + { + "epoch": 2.37468170243725, + "grad_norm": 0.635940114331161, + "learning_rate": 3.343033928425608e-05, + "loss": 0.2823, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14202754199504852, + "step": 3265, + "valid_targets_mean": 3389.8, + "valid_targets_min": 1507 + }, + { + "epoch": 2.3783193888686793, + "grad_norm": 0.6016808677547031, + "learning_rate": 3.340344251986537e-05, + "loss": 0.2771, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12453095614910126, + "step": 3270, + "valid_targets_mean": 3773.9, + "valid_targets_min": 1317 + }, + { + "epoch": 2.381957075300109, + "grad_norm": 0.6493207668192426, + "learning_rate": 3.337650167770281e-05, + "loss": 0.2733, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1271539032459259, + "step": 3275, + "valid_targets_mean": 3881.0, + "valid_targets_min": 3355 + }, + { + "epoch": 2.3855947617315385, + "grad_norm": 0.5616500847895824, + "learning_rate": 3.3349516846364496e-05, + "loss": 0.2756, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13402357697486877, + "step": 3280, + "valid_targets_mean": 3951.9, + "valid_targets_min": 2999 + }, + { + "epoch": 2.3892324481629683, + "grad_norm": 0.5743856544072798, + "learning_rate": 3.332248811459113e-05, + "loss": 0.2727, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1223636120557785, + "step": 3285, + "valid_targets_mean": 4086.5, + "valid_targets_min": 1739 + }, + { + "epoch": 2.392870134594398, + "grad_norm": 0.5607278813683847, + "learning_rate": 3.329541557126782e-05, + "loss": 0.2724, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11887595802545547, + "step": 3290, + "valid_targets_mean": 3298.0, + "valid_targets_min": 1140 + }, + { + "epoch": 2.3965078210258275, + "grad_norm": 0.6031537790719816, + "learning_rate": 3.326829930542374e-05, + "loss": 0.2566, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13889369368553162, + "step": 3295, + "valid_targets_mean": 4012.1, + "valid_targets_min": 2779 + }, + { + "epoch": 2.4001455074572573, + "grad_norm": 0.6073062311416269, + "learning_rate": 3.324113940623185e-05, + "loss": 0.2673, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12804676592350006, + "step": 3300, + "valid_targets_mean": 3636.9, + "valid_targets_min": 1183 + }, + { + "epoch": 2.4037831938886867, + "grad_norm": 0.598199890192643, + "learning_rate": 3.3213935963008576e-05, + "loss": 0.2555, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.156552255153656, + "step": 3305, + "valid_targets_mean": 4269.0, + "valid_targets_min": 3041 + }, + { + "epoch": 2.4074208803201165, + "grad_norm": 0.6006996187755297, + "learning_rate": 3.318668906521358e-05, + "loss": 0.2797, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1380569487810135, + "step": 3310, + "valid_targets_mean": 3350.1, + "valid_targets_min": 1726 + }, + { + "epoch": 2.411058566751546, + "grad_norm": 0.5100814351875782, + "learning_rate": 3.315939880244939e-05, + "loss": 0.2631, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12272945046424866, + "step": 3315, + "valid_targets_mean": 4747.4, + "valid_targets_min": 1743 + }, + { + "epoch": 2.4146962531829756, + "grad_norm": 0.5547879511912129, + "learning_rate": 3.3132065264461154e-05, + "loss": 0.2719, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1275254338979721, + "step": 3320, + "valid_targets_mean": 3702.8, + "valid_targets_min": 1234 + }, + { + "epoch": 2.4183339396144055, + "grad_norm": 0.5822517444321409, + "learning_rate": 3.310468854113636e-05, + "loss": 0.2797, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1412469893693924, + "step": 3325, + "valid_targets_mean": 4079.0, + "valid_targets_min": 2303 + }, + { + "epoch": 2.421971626045835, + "grad_norm": 0.5686084290047212, + "learning_rate": 3.307726872250447e-05, + "loss": 0.28, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13604910671710968, + "step": 3330, + "valid_targets_mean": 3830.8, + "valid_targets_min": 1651 + }, + { + "epoch": 2.4256093124772646, + "grad_norm": 0.6337671636328434, + "learning_rate": 3.304980589873669e-05, + "loss": 0.2692, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14259275794029236, + "step": 3335, + "valid_targets_mean": 2876.2, + "valid_targets_min": 1374 + }, + { + "epoch": 2.429246998908694, + "grad_norm": 0.6285636113999659, + "learning_rate": 3.302230016014564e-05, + "loss": 0.2631, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1368395984172821, + "step": 3340, + "valid_targets_mean": 3467.5, + "valid_targets_min": 1625 + }, + { + "epoch": 2.432884685340124, + "grad_norm": 0.5646127778561839, + "learning_rate": 3.299475159718508e-05, + "loss": 0.2569, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13010087609291077, + "step": 3345, + "valid_targets_mean": 4089.5, + "valid_targets_min": 997 + }, + { + "epoch": 2.436522371771553, + "grad_norm": 0.5892633485429684, + "learning_rate": 3.296716030044958e-05, + "loss": 0.2603, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11017262935638428, + "step": 3350, + "valid_targets_mean": 3618.6, + "valid_targets_min": 1179 + }, + { + "epoch": 2.440160058202983, + "grad_norm": 0.5796682519509309, + "learning_rate": 3.293952636067426e-05, + "loss": 0.2608, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12382735311985016, + "step": 3355, + "valid_targets_mean": 3079.1, + "valid_targets_min": 1444 + }, + { + "epoch": 2.4437977446344123, + "grad_norm": 0.5789140751694888, + "learning_rate": 3.2911849868734487e-05, + "loss": 0.2494, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11399437487125397, + "step": 3360, + "valid_targets_mean": 3486.8, + "valid_targets_min": 1504 + }, + { + "epoch": 2.447435431065842, + "grad_norm": 0.5776742364347436, + "learning_rate": 3.288413091564551e-05, + "loss": 0.2574, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12921425700187683, + "step": 3365, + "valid_targets_mean": 3998.4, + "valid_targets_min": 2302 + }, + { + "epoch": 2.4510731174972715, + "grad_norm": 0.6488104160079988, + "learning_rate": 3.285636959256227e-05, + "loss": 0.2663, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1654646396636963, + "step": 3370, + "valid_targets_mean": 4282.5, + "valid_targets_min": 1818 + }, + { + "epoch": 2.4547108039287013, + "grad_norm": 0.5912163397441839, + "learning_rate": 3.2828565990779e-05, + "loss": 0.2868, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14267028868198395, + "step": 3375, + "valid_targets_mean": 4221.9, + "valid_targets_min": 2633 + }, + { + "epoch": 2.458348490360131, + "grad_norm": 0.5859599968927008, + "learning_rate": 3.280072020172901e-05, + "loss": 0.2587, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13453692197799683, + "step": 3380, + "valid_targets_mean": 3522.4, + "valid_targets_min": 2481 + }, + { + "epoch": 2.4619861767915605, + "grad_norm": 0.6088880711927916, + "learning_rate": 3.277283231698429e-05, + "loss": 0.2683, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14552810788154602, + "step": 3385, + "valid_targets_mean": 4058.8, + "valid_targets_min": 2013 + }, + { + "epoch": 2.4656238632229903, + "grad_norm": 0.6124169745075544, + "learning_rate": 3.2744902428255336e-05, + "loss": 0.2577, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13899223506450653, + "step": 3390, + "valid_targets_mean": 3723.5, + "valid_targets_min": 1744 + }, + { + "epoch": 2.4692615496544197, + "grad_norm": 0.5722379000377743, + "learning_rate": 3.2716930627390704e-05, + "loss": 0.2693, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1410658359527588, + "step": 3395, + "valid_targets_mean": 4180.9, + "valid_targets_min": 1914 + }, + { + "epoch": 2.4728992360858495, + "grad_norm": 0.6096448168348366, + "learning_rate": 3.268891700637683e-05, + "loss": 0.2579, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11490217596292496, + "step": 3400, + "valid_targets_mean": 3668.6, + "valid_targets_min": 3065 + }, + { + "epoch": 2.476536922517279, + "grad_norm": 0.5825257918922361, + "learning_rate": 3.2660861657337646e-05, + "loss": 0.2488, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14985854923725128, + "step": 3405, + "valid_targets_mean": 4473.2, + "valid_targets_min": 2894 + }, + { + "epoch": 2.4801746089487087, + "grad_norm": 0.6220222886326096, + "learning_rate": 3.263276467253432e-05, + "loss": 0.2688, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13400928676128387, + "step": 3410, + "valid_targets_mean": 3395.6, + "valid_targets_min": 2359 + }, + { + "epoch": 2.4838122953801385, + "grad_norm": 0.6302895520978945, + "learning_rate": 3.2604626144364955e-05, + "loss": 0.2575, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11689244210720062, + "step": 3415, + "valid_targets_mean": 3114.5, + "valid_targets_min": 1463 + }, + { + "epoch": 2.487449981811568, + "grad_norm": 0.6145980685475172, + "learning_rate": 3.257644616536425e-05, + "loss": 0.2678, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13427552580833435, + "step": 3420, + "valid_targets_mean": 3588.1, + "valid_targets_min": 1426 + }, + { + "epoch": 2.491087668242997, + "grad_norm": 0.6010255592737749, + "learning_rate": 3.254822482820322e-05, + "loss": 0.2733, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13123303651809692, + "step": 3425, + "valid_targets_mean": 3405.5, + "valid_targets_min": 1661 + }, + { + "epoch": 2.494725354674427, + "grad_norm": 0.5477199729661505, + "learning_rate": 3.251996222568891e-05, + "loss": 0.2677, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1389879286289215, + "step": 3430, + "valid_targets_mean": 4457.8, + "valid_targets_min": 1529 + }, + { + "epoch": 2.498363041105857, + "grad_norm": 0.8481773438501398, + "learning_rate": 3.2491658450764024e-05, + "loss": 0.2896, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15845340490341187, + "step": 3435, + "valid_targets_mean": 3272.9, + "valid_targets_min": 1450 + }, + { + "epoch": 2.502000727537286, + "grad_norm": 0.6718391109766616, + "learning_rate": 3.246331359650671e-05, + "loss": 0.2628, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11781521141529083, + "step": 3440, + "valid_targets_mean": 3103.2, + "valid_targets_min": 1625 + }, + { + "epoch": 2.505638413968716, + "grad_norm": 0.5429138073428161, + "learning_rate": 3.243492775613018e-05, + "loss": 0.2727, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11309918016195297, + "step": 3445, + "valid_targets_mean": 3952.2, + "valid_targets_min": 1174 + }, + { + "epoch": 2.5092761004001454, + "grad_norm": 0.5952313438013163, + "learning_rate": 3.240650102298244e-05, + "loss": 0.2744, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1393856257200241, + "step": 3450, + "valid_targets_mean": 4039.2, + "valid_targets_min": 2882 + }, + { + "epoch": 2.512913786831575, + "grad_norm": 0.6088125679032266, + "learning_rate": 3.237803349054596e-05, + "loss": 0.2741, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15354037284851074, + "step": 3455, + "valid_targets_mean": 4594.0, + "valid_targets_min": 1946 + }, + { + "epoch": 2.5165514732630045, + "grad_norm": 0.5609887466760576, + "learning_rate": 3.234952525243741e-05, + "loss": 0.268, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12978559732437134, + "step": 3460, + "valid_targets_mean": 4082.6, + "valid_targets_min": 2000 + }, + { + "epoch": 2.5201891596944344, + "grad_norm": 0.705297934775914, + "learning_rate": 3.2320976402407286e-05, + "loss": 0.2721, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1452639102935791, + "step": 3465, + "valid_targets_mean": 4596.1, + "valid_targets_min": 2012 + }, + { + "epoch": 2.523826846125864, + "grad_norm": 0.6316288340963611, + "learning_rate": 3.229238703433966e-05, + "loss": 0.2716, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13632328808307648, + "step": 3470, + "valid_targets_mean": 3094.4, + "valid_targets_min": 719 + }, + { + "epoch": 2.5274645325572935, + "grad_norm": 0.6536853873111823, + "learning_rate": 3.2263757242251846e-05, + "loss": 0.2679, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12379848957061768, + "step": 3475, + "valid_targets_mean": 4496.2, + "valid_targets_min": 3238 + }, + { + "epoch": 2.531102218988723, + "grad_norm": 0.5897915697246868, + "learning_rate": 3.2235087120294105e-05, + "loss": 0.2549, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11371707171201706, + "step": 3480, + "valid_targets_mean": 3816.8, + "valid_targets_min": 1372 + }, + { + "epoch": 2.5347399054201527, + "grad_norm": 0.6835827233984606, + "learning_rate": 3.2206376762749284e-05, + "loss": 0.2532, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14215350151062012, + "step": 3485, + "valid_targets_mean": 3306.1, + "valid_targets_min": 2370 + }, + { + "epoch": 2.5383775918515825, + "grad_norm": 0.5681224140157343, + "learning_rate": 3.21776262640326e-05, + "loss": 0.2622, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1160813570022583, + "step": 3490, + "valid_targets_mean": 2958.1, + "valid_targets_min": 1576 + }, + { + "epoch": 2.542015278283012, + "grad_norm": 0.6296757370775974, + "learning_rate": 3.2148835718691235e-05, + "loss": 0.2592, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14146940410137177, + "step": 3495, + "valid_targets_mean": 3758.5, + "valid_targets_min": 1426 + }, + { + "epoch": 2.5456529647144417, + "grad_norm": 0.6395445235700964, + "learning_rate": 3.212000522140408e-05, + "loss": 0.2691, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1187351644039154, + "step": 3500, + "valid_targets_mean": 3117.0, + "valid_targets_min": 1728 + }, + { + "epoch": 2.5492906511458715, + "grad_norm": 0.7167796864119903, + "learning_rate": 3.209113486698143e-05, + "loss": 0.263, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14096613228321075, + "step": 3505, + "valid_targets_mean": 3264.2, + "valid_targets_min": 1362 + }, + { + "epoch": 2.552928337577301, + "grad_norm": 0.5582228148656896, + "learning_rate": 3.206222475036461e-05, + "loss": 0.2651, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12977878749370575, + "step": 3510, + "valid_targets_mean": 3836.0, + "valid_targets_min": 852 + }, + { + "epoch": 2.5565660240087302, + "grad_norm": 0.7199855416557892, + "learning_rate": 3.203327496662574e-05, + "loss": 0.2671, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16426774859428406, + "step": 3515, + "valid_targets_mean": 2764.1, + "valid_targets_min": 1195 + }, + { + "epoch": 2.56020371044016, + "grad_norm": 0.5650672663266725, + "learning_rate": 3.200428561096737e-05, + "loss": 0.2587, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13942572474479675, + "step": 3520, + "valid_targets_mean": 4457.5, + "valid_targets_min": 1564 + }, + { + "epoch": 2.56384139687159, + "grad_norm": 0.5552919625404192, + "learning_rate": 3.197525677872219e-05, + "loss": 0.2567, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10992539674043655, + "step": 3525, + "valid_targets_mean": 4371.2, + "valid_targets_min": 1975 + }, + { + "epoch": 2.567479083303019, + "grad_norm": 0.5626171916862984, + "learning_rate": 3.19461885653527e-05, + "loss": 0.2446, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1297529637813568, + "step": 3530, + "valid_targets_mean": 3965.8, + "valid_targets_min": 2426 + }, + { + "epoch": 2.571116769734449, + "grad_norm": 0.5004396062987074, + "learning_rate": 3.191708106645092e-05, + "loss": 0.2506, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12606431543827057, + "step": 3535, + "valid_targets_mean": 6384.6, + "valid_targets_min": 2678 + }, + { + "epoch": 2.5747544561658784, + "grad_norm": 0.5626832584509855, + "learning_rate": 3.188793437773807e-05, + "loss": 0.2594, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1233622282743454, + "step": 3540, + "valid_targets_mean": 3543.2, + "valid_targets_min": 2826 + }, + { + "epoch": 2.578392142597308, + "grad_norm": 0.5974069582978984, + "learning_rate": 3.1858748595064225e-05, + "loss": 0.2471, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1052197590470314, + "step": 3545, + "valid_targets_mean": 3155.8, + "valid_targets_min": 993 + }, + { + "epoch": 2.5820298290287376, + "grad_norm": 0.5732016940395837, + "learning_rate": 3.1829523814408024e-05, + "loss": 0.2494, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1401035189628601, + "step": 3550, + "valid_targets_mean": 4166.5, + "valid_targets_min": 2724 + }, + { + "epoch": 2.5856675154601674, + "grad_norm": 0.6054066982567247, + "learning_rate": 3.180026013187636e-05, + "loss": 0.2518, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11948444694280624, + "step": 3555, + "valid_targets_mean": 3079.5, + "valid_targets_min": 1360 + }, + { + "epoch": 2.589305201891597, + "grad_norm": 0.6281397381533048, + "learning_rate": 3.177095764370407e-05, + "loss": 0.249, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12264097481966019, + "step": 3560, + "valid_targets_mean": 4114.9, + "valid_targets_min": 2414 + }, + { + "epoch": 2.5929428883230266, + "grad_norm": 0.5756946400988691, + "learning_rate": 3.1741616446253574e-05, + "loss": 0.2705, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14798961579799652, + "step": 3565, + "valid_targets_mean": 4524.8, + "valid_targets_min": 2425 + }, + { + "epoch": 2.596580574754456, + "grad_norm": 0.6354632602830393, + "learning_rate": 3.171223663601463e-05, + "loss": 0.2609, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12842215597629547, + "step": 3570, + "valid_targets_mean": 3035.2, + "valid_targets_min": 1720 + }, + { + "epoch": 2.6002182611858857, + "grad_norm": 0.6598874407289866, + "learning_rate": 3.1682818309603915e-05, + "loss": 0.2574, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1312503218650818, + "step": 3575, + "valid_targets_mean": 3260.4, + "valid_targets_min": 1621 + }, + { + "epoch": 2.6038559476173155, + "grad_norm": 0.6491230191080161, + "learning_rate": 3.1653361563764846e-05, + "loss": 0.2456, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13304734230041504, + "step": 3580, + "valid_targets_mean": 3214.6, + "valid_targets_min": 1858 + }, + { + "epoch": 2.607493634048745, + "grad_norm": 0.6501890542567019, + "learning_rate": 3.1623866495367115e-05, + "loss": 0.2589, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11940549314022064, + "step": 3585, + "valid_targets_mean": 3254.2, + "valid_targets_min": 1254 + }, + { + "epoch": 2.6111313204801747, + "grad_norm": 0.6192279600811121, + "learning_rate": 3.159433320140648e-05, + "loss": 0.2497, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13544028997421265, + "step": 3590, + "valid_targets_mean": 2994.4, + "valid_targets_min": 1039 + }, + { + "epoch": 2.614769006911604, + "grad_norm": 0.5991286996472234, + "learning_rate": 3.156476177900438e-05, + "loss": 0.2644, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16194221377372742, + "step": 3595, + "valid_targets_mean": 4625.4, + "valid_targets_min": 2526 + }, + { + "epoch": 2.618406693343034, + "grad_norm": 0.5675548866290573, + "learning_rate": 3.153515232540767e-05, + "loss": 0.2542, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13777749240398407, + "step": 3600, + "valid_targets_mean": 3571.8, + "valid_targets_min": 2409 + }, + { + "epoch": 2.6220443797744633, + "grad_norm": 0.5990248373218381, + "learning_rate": 3.150550493798824e-05, + "loss": 0.2574, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11114709079265594, + "step": 3605, + "valid_targets_mean": 3691.5, + "valid_targets_min": 2918 + }, + { + "epoch": 2.625682066205893, + "grad_norm": 0.6014143450003461, + "learning_rate": 3.147581971424276e-05, + "loss": 0.2606, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12825147807598114, + "step": 3610, + "valid_targets_mean": 4063.1, + "valid_targets_min": 2179 + }, + { + "epoch": 2.629319752637323, + "grad_norm": 0.4993890478141338, + "learning_rate": 3.144609675179229e-05, + "loss": 0.2214, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06053350865840912, + "step": 3615, + "valid_targets_mean": 2871.4, + "valid_targets_min": 955 + }, + { + "epoch": 2.6329574390687522, + "grad_norm": 0.46289486666435914, + "learning_rate": 3.141633614838203e-05, + "loss": 0.1315, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04971493035554886, + "step": 3620, + "valid_targets_mean": 1788.9, + "valid_targets_min": 669 + }, + { + "epoch": 2.636595125500182, + "grad_norm": 0.366795955115012, + "learning_rate": 3.1386538001880934e-05, + "loss": 0.0977, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04621807858347893, + "step": 3625, + "valid_targets_mean": 3755.8, + "valid_targets_min": 2651 + }, + { + "epoch": 2.6402328119316114, + "grad_norm": 0.45937919984256786, + "learning_rate": 3.135670241028145e-05, + "loss": 0.0984, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.053791243582963943, + "step": 3630, + "valid_targets_mean": 3069.8, + "valid_targets_min": 710 + }, + { + "epoch": 2.643870498363041, + "grad_norm": 0.38735684547027494, + "learning_rate": 3.132682947169913e-05, + "loss": 0.1118, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03356508910655975, + "step": 3635, + "valid_targets_mean": 3618.5, + "valid_targets_min": 1365 + }, + { + "epoch": 2.6475081847944706, + "grad_norm": 0.44530524542581407, + "learning_rate": 3.1296919284372366e-05, + "loss": 0.0984, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.051966167986392975, + "step": 3640, + "valid_targets_mean": 3279.9, + "valid_targets_min": 889 + }, + { + "epoch": 2.6511458712259004, + "grad_norm": 0.699630416343122, + "learning_rate": 3.1266971946662045e-05, + "loss": 0.1315, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07253696024417877, + "step": 3645, + "valid_targets_mean": 1652.4, + "valid_targets_min": 559 + }, + { + "epoch": 2.65478355765733, + "grad_norm": 0.3686807985217438, + "learning_rate": 3.123698755705121e-05, + "loss": 0.1328, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.041426390409469604, + "step": 3650, + "valid_targets_mean": 3449.0, + "valid_targets_min": 1013 + }, + { + "epoch": 2.6584212440887596, + "grad_norm": 0.38369515266385995, + "learning_rate": 3.1206966214144764e-05, + "loss": 0.1101, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06235937401652336, + "step": 3655, + "valid_targets_mean": 3726.1, + "valid_targets_min": 3222 + }, + { + "epoch": 2.662058930520189, + "grad_norm": 0.5204553519315888, + "learning_rate": 3.1176908016669125e-05, + "loss": 0.1452, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1020481288433075, + "step": 3660, + "valid_targets_mean": 3205.8, + "valid_targets_min": 1949 + }, + { + "epoch": 2.6656966169516187, + "grad_norm": 0.2994707568521483, + "learning_rate": 3.114681306347191e-05, + "loss": 0.0843, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.039092160761356354, + "step": 3665, + "valid_targets_mean": 4059.8, + "valid_targets_min": 886 + }, + { + "epoch": 2.6693343033830486, + "grad_norm": 0.47764869590263553, + "learning_rate": 3.111668145352163e-05, + "loss": 0.1011, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05485702306032181, + "step": 3670, + "valid_targets_mean": 1858.2, + "valid_targets_min": 447 + }, + { + "epoch": 2.672971989814478, + "grad_norm": 0.3935953966157133, + "learning_rate": 3.108651328590728e-05, + "loss": 0.0977, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.041664764285087585, + "step": 3675, + "valid_targets_mean": 3195.6, + "valid_targets_min": 758 + }, + { + "epoch": 2.6766096762459077, + "grad_norm": 0.49312544116496915, + "learning_rate": 3.105630865983816e-05, + "loss": 0.1204, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06547340750694275, + "step": 3680, + "valid_targets_mean": 1808.4, + "valid_targets_min": 604 + }, + { + "epoch": 2.680247362677337, + "grad_norm": 0.3804041754715141, + "learning_rate": 3.102606767464341e-05, + "loss": 0.1199, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05331578850746155, + "step": 3685, + "valid_targets_mean": 3314.6, + "valid_targets_min": 1743 + }, + { + "epoch": 2.683885049108767, + "grad_norm": 0.3627093601947754, + "learning_rate": 3.0995790429771756e-05, + "loss": 0.0989, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.039569973945617676, + "step": 3690, + "valid_targets_mean": 2440.4, + "valid_targets_min": 574 + }, + { + "epoch": 2.6875227355401963, + "grad_norm": 0.4043148341294831, + "learning_rate": 3.0965477024791154e-05, + "loss": 0.0972, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04732849448919296, + "step": 3695, + "valid_targets_mean": 2535.6, + "valid_targets_min": 523 + }, + { + "epoch": 2.691160421971626, + "grad_norm": 0.49126442603256526, + "learning_rate": 3.093512755938849e-05, + "loss": 0.0969, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07462019473314285, + "step": 3700, + "valid_targets_mean": 3398.1, + "valid_targets_min": 2635 + }, + { + "epoch": 2.694798108403056, + "grad_norm": 0.3876936243101128, + "learning_rate": 3.090474213336923e-05, + "loss": 0.1089, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04585566371679306, + "step": 3705, + "valid_targets_mean": 3164.8, + "valid_targets_min": 812 + }, + { + "epoch": 2.6984357948344853, + "grad_norm": 0.37100467657932307, + "learning_rate": 3.087432084665709e-05, + "loss": 0.1129, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.054744258522987366, + "step": 3710, + "valid_targets_mean": 2463.9, + "valid_targets_min": 856 + }, + { + "epoch": 2.7020734812659146, + "grad_norm": 0.4240467964023532, + "learning_rate": 3.084386379929373e-05, + "loss": 0.114, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07003875821828842, + "step": 3715, + "valid_targets_mean": 2995.6, + "valid_targets_min": 718 + }, + { + "epoch": 2.7057111676973444, + "grad_norm": 0.5394869040261644, + "learning_rate": 3.0813371091438406e-05, + "loss": 0.1337, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1421079784631729, + "step": 3720, + "valid_targets_mean": 2866.9, + "valid_targets_min": 1297 + }, + { + "epoch": 2.7093488541287742, + "grad_norm": 0.490079988061846, + "learning_rate": 3.0782842823367634e-05, + "loss": 0.1276, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.046553391963243484, + "step": 3725, + "valid_targets_mean": 1661.1, + "valid_targets_min": 786 + }, + { + "epoch": 2.7129865405602036, + "grad_norm": 0.6082030807992259, + "learning_rate": 3.0752279095474867e-05, + "loss": 0.1, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06776092946529388, + "step": 3730, + "valid_targets_mean": 1624.2, + "valid_targets_min": 550 + }, + { + "epoch": 2.7166242269916334, + "grad_norm": 0.42009882332725174, + "learning_rate": 3.0721680008270206e-05, + "loss": 0.1368, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05056443437933922, + "step": 3735, + "valid_targets_mean": 3254.6, + "valid_targets_min": 1304 + }, + { + "epoch": 2.720261913423063, + "grad_norm": 0.39480233822206007, + "learning_rate": 3.0691045662379986e-05, + "loss": 0.1254, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04486966133117676, + "step": 3740, + "valid_targets_mean": 2833.0, + "valid_targets_min": 1247 + }, + { + "epoch": 2.7238995998544926, + "grad_norm": 0.5036788460446842, + "learning_rate": 3.0660376158546526e-05, + "loss": 0.1043, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06461863219738007, + "step": 3745, + "valid_targets_mean": 3056.5, + "valid_targets_min": 800 + }, + { + "epoch": 2.727537286285922, + "grad_norm": 0.5645470380338171, + "learning_rate": 3.0629671597627745e-05, + "loss": 0.2992, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12839019298553467, + "step": 3750, + "valid_targets_mean": 2580.8, + "valid_targets_min": 987 + }, + { + "epoch": 2.7311749727173518, + "grad_norm": 0.5496953203039707, + "learning_rate": 3.059893208059686e-05, + "loss": 0.1309, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10780438780784607, + "step": 3755, + "valid_targets_mean": 2245.6, + "valid_targets_min": 776 + }, + { + "epoch": 2.7348126591487816, + "grad_norm": 0.354142816253058, + "learning_rate": 3.056815770854203e-05, + "loss": 0.1084, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03022707998752594, + "step": 3760, + "valid_targets_mean": 1361.5, + "valid_targets_min": 658 + }, + { + "epoch": 2.738450345580211, + "grad_norm": 0.5067356442467635, + "learning_rate": 3.0537348582666044e-05, + "loss": 0.2105, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04984079673886299, + "step": 3765, + "valid_targets_mean": 1729.0, + "valid_targets_min": 666 + }, + { + "epoch": 2.7420880320116408, + "grad_norm": 0.303310023364222, + "learning_rate": 3.0506504804285977e-05, + "loss": 0.0958, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.040304262191057205, + "step": 3770, + "valid_targets_mean": 4975.0, + "valid_targets_min": 3529 + }, + { + "epoch": 2.74572571844307, + "grad_norm": 0.44352967559423523, + "learning_rate": 3.0475626474832868e-05, + "loss": 0.1079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06442651152610779, + "step": 3775, + "valid_targets_mean": 3988.8, + "valid_targets_min": 1746 + }, + { + "epoch": 2.7493634048745, + "grad_norm": 0.4469855195111906, + "learning_rate": 3.0444713695851375e-05, + "loss": 0.1065, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06729622185230255, + "step": 3780, + "valid_targets_mean": 3222.4, + "valid_targets_min": 1226 + }, + { + "epoch": 2.7530010913059293, + "grad_norm": 0.38190672403274933, + "learning_rate": 3.041376656899943e-05, + "loss": 0.0961, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06273630261421204, + "step": 3785, + "valid_targets_mean": 4403.6, + "valid_targets_min": 3275 + }, + { + "epoch": 2.756638777737359, + "grad_norm": 0.3618606062928999, + "learning_rate": 3.038278519604795e-05, + "loss": 0.0975, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.040049877017736435, + "step": 3790, + "valid_targets_mean": 2546.2, + "valid_targets_min": 723 + }, + { + "epoch": 2.760276464168789, + "grad_norm": 0.3951608319219096, + "learning_rate": 3.035176967888044e-05, + "loss": 0.0948, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.055870477110147476, + "step": 3795, + "valid_targets_mean": 3348.6, + "valid_targets_min": 2234 + }, + { + "epoch": 2.7639141506002183, + "grad_norm": 0.3302663556191806, + "learning_rate": 3.0320720119492707e-05, + "loss": 0.0913, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030992014333605766, + "step": 3800, + "valid_targets_mean": 3603.0, + "valid_targets_min": 1421 + }, + { + "epoch": 2.7675518370316476, + "grad_norm": 0.4098525200551233, + "learning_rate": 3.0289636619992505e-05, + "loss": 0.0956, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05131656676530838, + "step": 3805, + "valid_targets_mean": 2849.2, + "valid_targets_min": 1068 + }, + { + "epoch": 2.7711895234630775, + "grad_norm": 0.470537680444737, + "learning_rate": 3.02585192825992e-05, + "loss": 0.1426, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.040368691086769104, + "step": 3810, + "valid_targets_mean": 1587.9, + "valid_targets_min": 535 + }, + { + "epoch": 2.7748272098945073, + "grad_norm": 0.3574707934340061, + "learning_rate": 3.0227368209643437e-05, + "loss": 0.0814, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03336938098073006, + "step": 3815, + "valid_targets_mean": 1702.4, + "valid_targets_min": 585 + }, + { + "epoch": 2.7784648963259366, + "grad_norm": 0.42445662696175496, + "learning_rate": 3.0196183503566794e-05, + "loss": 0.0932, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04549098759889603, + "step": 3820, + "valid_targets_mean": 2588.6, + "valid_targets_min": 582 + }, + { + "epoch": 2.7821025827573664, + "grad_norm": 0.38092124281401657, + "learning_rate": 3.0164965266921462e-05, + "loss": 0.087, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04793877899646759, + "step": 3825, + "valid_targets_mean": 3144.0, + "valid_targets_min": 516 + }, + { + "epoch": 2.785740269188796, + "grad_norm": 0.4761361612284449, + "learning_rate": 3.01337136023699e-05, + "loss": 0.1411, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07162503898143768, + "step": 3830, + "valid_targets_mean": 3241.4, + "valid_targets_min": 999 + }, + { + "epoch": 2.7893779556202256, + "grad_norm": 0.4786818862183569, + "learning_rate": 3.0102428612684486e-05, + "loss": 0.1108, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09321946650743484, + "step": 3835, + "valid_targets_mean": 2021.1, + "valid_targets_min": 696 + }, + { + "epoch": 2.793015642051655, + "grad_norm": 0.4874811306402491, + "learning_rate": 3.0071110400747206e-05, + "loss": 0.126, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05804925784468651, + "step": 3840, + "valid_targets_mean": 2048.5, + "valid_targets_min": 635 + }, + { + "epoch": 2.796653328483085, + "grad_norm": 0.4087839869308148, + "learning_rate": 3.003975906954928e-05, + "loss": 0.1076, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.046706605702638626, + "step": 3845, + "valid_targets_mean": 2641.6, + "valid_targets_min": 816 + }, + { + "epoch": 2.8002910149145146, + "grad_norm": 0.5831439620642058, + "learning_rate": 3.0008374722190843e-05, + "loss": 0.1564, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15746557712554932, + "step": 3850, + "valid_targets_mean": 2236.4, + "valid_targets_min": 949 + }, + { + "epoch": 2.803928701345944, + "grad_norm": 0.4058684376460655, + "learning_rate": 2.9976957461880615e-05, + "loss": 0.1167, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.053821317851543427, + "step": 3855, + "valid_targets_mean": 3270.9, + "valid_targets_min": 2201 + }, + { + "epoch": 2.8075663877773733, + "grad_norm": 0.7204832404845884, + "learning_rate": 2.9945507391935557e-05, + "loss": 0.1317, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09420809149742126, + "step": 3860, + "valid_targets_mean": 1500.5, + "valid_targets_min": 687 + }, + { + "epoch": 2.811204074208803, + "grad_norm": 0.432213984830873, + "learning_rate": 2.9914024615780516e-05, + "loss": 0.0954, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04278099536895752, + "step": 3865, + "valid_targets_mean": 2285.5, + "valid_targets_min": 449 + }, + { + "epoch": 2.814841760640233, + "grad_norm": 0.384167806738454, + "learning_rate": 2.9882509236947887e-05, + "loss": 0.1055, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030389923602342606, + "step": 3870, + "valid_targets_mean": 2368.1, + "valid_targets_min": 658 + }, + { + "epoch": 2.8184794470716623, + "grad_norm": 0.6210936460075546, + "learning_rate": 2.9850961359077293e-05, + "loss": 0.1096, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04089265316724777, + "step": 3875, + "valid_targets_mean": 831.8, + "valid_targets_min": 516 + }, + { + "epoch": 2.822117133503092, + "grad_norm": 0.36654549601543507, + "learning_rate": 2.9819381085915225e-05, + "loss": 0.0869, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05952584743499756, + "step": 3880, + "valid_targets_mean": 4461.9, + "valid_targets_min": 2805 + }, + { + "epoch": 2.8257548199345215, + "grad_norm": 0.36937015359199293, + "learning_rate": 2.9787768521314717e-05, + "loss": 0.0832, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0489317923784256, + "step": 3885, + "valid_targets_mean": 3766.8, + "valid_targets_min": 2036 + }, + { + "epoch": 2.8293925063659513, + "grad_norm": 0.3636660595514668, + "learning_rate": 2.9756123769234978e-05, + "loss": 0.0969, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.041392117738723755, + "step": 3890, + "valid_targets_mean": 2336.5, + "valid_targets_min": 681 + }, + { + "epoch": 2.8330301927973807, + "grad_norm": 0.35518058439575534, + "learning_rate": 2.972444693374109e-05, + "loss": 0.1136, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05278894305229187, + "step": 3895, + "valid_targets_mean": 3856.1, + "valid_targets_min": 3436 + }, + { + "epoch": 2.8366678792288105, + "grad_norm": 0.47956316338214205, + "learning_rate": 2.9692738119003618e-05, + "loss": 0.1111, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07087892293930054, + "step": 3900, + "valid_targets_mean": 3165.5, + "valid_targets_min": 891 + }, + { + "epoch": 2.8403055656602403, + "grad_norm": 0.2743520149692237, + "learning_rate": 2.9660997429298305e-05, + "loss": 0.0977, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03344181180000305, + "step": 3905, + "valid_targets_mean": 3550.6, + "valid_targets_min": 2573 + }, + { + "epoch": 2.8439432520916696, + "grad_norm": 0.31241922383010473, + "learning_rate": 2.962922496900572e-05, + "loss": 0.0922, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04341200739145279, + "step": 3910, + "valid_targets_mean": 2995.0, + "valid_targets_min": 928 + }, + { + "epoch": 2.8475809385230995, + "grad_norm": 0.4112053902952101, + "learning_rate": 2.959742084261089e-05, + "loss": 0.1153, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06260673701763153, + "step": 3915, + "valid_targets_mean": 3839.6, + "valid_targets_min": 3392 + }, + { + "epoch": 2.851218624954529, + "grad_norm": 0.30437429880241884, + "learning_rate": 2.9565585154703008e-05, + "loss": 0.0848, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04707106202840805, + "step": 3920, + "valid_targets_mean": 3855.0, + "valid_targets_min": 2848 + }, + { + "epoch": 2.8548563113859586, + "grad_norm": 0.4599111592180491, + "learning_rate": 2.9533718009975033e-05, + "loss": 0.1042, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05379137024283409, + "step": 3925, + "valid_targets_mean": 2586.0, + "valid_targets_min": 595 + }, + { + "epoch": 2.858493997817388, + "grad_norm": 0.43927617020621973, + "learning_rate": 2.950181951322339e-05, + "loss": 0.1039, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0550314299762249, + "step": 3930, + "valid_targets_mean": 2471.8, + "valid_targets_min": 1016 + }, + { + "epoch": 2.862131684248818, + "grad_norm": 0.5080393066734379, + "learning_rate": 2.9469889769347592e-05, + "loss": 0.1098, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08142200112342834, + "step": 3935, + "valid_targets_mean": 1567.8, + "valid_targets_min": 811 + }, + { + "epoch": 2.8657693706802476, + "grad_norm": 0.6231271020584213, + "learning_rate": 2.943792888334991e-05, + "loss": 0.1137, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06609875708818436, + "step": 3940, + "valid_targets_mean": 1569.1, + "valid_targets_min": 697 + }, + { + "epoch": 2.869407057111677, + "grad_norm": 0.6281113713133436, + "learning_rate": 2.9405936960335043e-05, + "loss": 0.1, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07510620355606079, + "step": 3945, + "valid_targets_mean": 1448.1, + "valid_targets_min": 736 + }, + { + "epoch": 2.8730447435431063, + "grad_norm": 0.3867822761015305, + "learning_rate": 2.937391410550973e-05, + "loss": 0.1053, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.033857543021440506, + "step": 3950, + "valid_targets_mean": 2415.5, + "valid_targets_min": 804 + }, + { + "epoch": 2.876682429974536, + "grad_norm": 0.36785927006626923, + "learning_rate": 2.9341860424182457e-05, + "loss": 0.1173, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04019784927368164, + "step": 3955, + "valid_targets_mean": 3472.5, + "valid_targets_min": 1018 + }, + { + "epoch": 2.880320116405966, + "grad_norm": 0.362932871791688, + "learning_rate": 2.930977602176307e-05, + "loss": 0.1013, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0480780154466629, + "step": 3960, + "valid_targets_mean": 3623.8, + "valid_targets_min": 2369 + }, + { + "epoch": 2.8839578028373953, + "grad_norm": 0.4206385889860581, + "learning_rate": 2.927766100376245e-05, + "loss": 0.09, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.052347078919410706, + "step": 3965, + "valid_targets_mean": 3135.8, + "valid_targets_min": 1318 + }, + { + "epoch": 2.887595489268825, + "grad_norm": 0.5546409167582762, + "learning_rate": 2.9245515475792143e-05, + "loss": 0.1118, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07402493059635162, + "step": 3970, + "valid_targets_mean": 2258.2, + "valid_targets_min": 903 + }, + { + "epoch": 2.8912331757002545, + "grad_norm": 0.41448358451377926, + "learning_rate": 2.9213339543564055e-05, + "loss": 0.1085, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04812890663743019, + "step": 3975, + "valid_targets_mean": 3460.4, + "valid_targets_min": 2346 + }, + { + "epoch": 2.8948708621316843, + "grad_norm": 0.5828888856268895, + "learning_rate": 2.918113331289005e-05, + "loss": 0.0906, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08222638070583344, + "step": 3980, + "valid_targets_mean": 2283.0, + "valid_targets_min": 737 + }, + { + "epoch": 2.8985085485631137, + "grad_norm": 0.3636539431036967, + "learning_rate": 2.9148896889681643e-05, + "loss": 0.0951, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04695935547351837, + "step": 3985, + "valid_targets_mean": 3979.6, + "valid_targets_min": 2555 + }, + { + "epoch": 2.9021462349945435, + "grad_norm": 0.28299410838811495, + "learning_rate": 2.9116630379949643e-05, + "loss": 0.0922, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04223409295082092, + "step": 3990, + "valid_targets_mean": 4500.2, + "valid_targets_min": 2991 + }, + { + "epoch": 2.9057839214259733, + "grad_norm": 0.36925144449816544, + "learning_rate": 2.9084333889803783e-05, + "loss": 0.0794, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03296161815524101, + "step": 3995, + "valid_targets_mean": 1992.4, + "valid_targets_min": 618 + }, + { + "epoch": 2.9094216078574027, + "grad_norm": 0.3119294117172151, + "learning_rate": 2.905200752545241e-05, + "loss": 0.1049, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03177472949028015, + "step": 4000, + "valid_targets_mean": 3641.8, + "valid_targets_min": 2944 + }, + { + "epoch": 2.913059294288832, + "grad_norm": 0.3136746798389872, + "learning_rate": 2.9019651393202105e-05, + "loss": 0.0921, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.056320615112781525, + "step": 4005, + "valid_targets_mean": 3686.1, + "valid_targets_min": 2702 + }, + { + "epoch": 2.916696980720262, + "grad_norm": 0.37757781060977863, + "learning_rate": 2.8987265599457326e-05, + "loss": 0.1219, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.026903368532657623, + "step": 4010, + "valid_targets_mean": 3341.6, + "valid_targets_min": 509 + }, + { + "epoch": 2.9203346671516917, + "grad_norm": 0.31270051362999934, + "learning_rate": 2.8954850250720095e-05, + "loss": 0.1119, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03711850568652153, + "step": 4015, + "valid_targets_mean": 2757.6, + "valid_targets_min": 637 + }, + { + "epoch": 2.923972353583121, + "grad_norm": 0.4349686823320762, + "learning_rate": 2.892240545358962e-05, + "loss": 0.1232, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05590439215302467, + "step": 4020, + "valid_targets_mean": 2802.6, + "valid_targets_min": 737 + }, + { + "epoch": 2.927610040014551, + "grad_norm": 0.31293044086238875, + "learning_rate": 2.8889931314761947e-05, + "loss": 0.0869, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04133059084415436, + "step": 4025, + "valid_targets_mean": 3312.9, + "valid_targets_min": 1292 + }, + { + "epoch": 2.93124772644598, + "grad_norm": 0.6456558309782208, + "learning_rate": 2.885742794102962e-05, + "loss": 0.1082, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.052728064358234406, + "step": 4030, + "valid_targets_mean": 1049.6, + "valid_targets_min": 697 + }, + { + "epoch": 2.93488541287741, + "grad_norm": 0.375250867858344, + "learning_rate": 2.8824895439281317e-05, + "loss": 0.0938, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.043897178024053574, + "step": 4035, + "valid_targets_mean": 3541.4, + "valid_targets_min": 923 + }, + { + "epoch": 2.9385230993088394, + "grad_norm": 0.8166740742256334, + "learning_rate": 2.879233391650151e-05, + "loss": 0.1387, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0781305655837059, + "step": 4040, + "valid_targets_mean": 931.6, + "valid_targets_min": 531 + }, + { + "epoch": 2.942160785740269, + "grad_norm": 0.5044264160080727, + "learning_rate": 2.8759743479770103e-05, + "loss": 0.1304, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06985421478748322, + "step": 4045, + "valid_targets_mean": 1840.5, + "valid_targets_min": 844 + }, + { + "epoch": 2.945798472171699, + "grad_norm": 0.4716543131426424, + "learning_rate": 2.8727124236262094e-05, + "loss": 0.0994, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06737782061100006, + "step": 4050, + "valid_targets_mean": 3893.9, + "valid_targets_min": 2967 + }, + { + "epoch": 2.9494361586031284, + "grad_norm": 0.36185955384133334, + "learning_rate": 2.8694476293247196e-05, + "loss": 0.097, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05286838859319687, + "step": 4055, + "valid_targets_mean": 3605.9, + "valid_targets_min": 2490 + }, + { + "epoch": 2.953073845034558, + "grad_norm": 0.3799870624828868, + "learning_rate": 2.866179975808952e-05, + "loss": 0.0977, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.042296018451452255, + "step": 4060, + "valid_targets_mean": 2682.0, + "valid_targets_min": 970 + }, + { + "epoch": 2.9567115314659875, + "grad_norm": 0.39149978496619364, + "learning_rate": 2.8629094738247188e-05, + "loss": 0.088, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03831206634640694, + "step": 4065, + "valid_targets_mean": 2028.5, + "valid_targets_min": 729 + }, + { + "epoch": 2.9603492178974173, + "grad_norm": 0.7890767262143384, + "learning_rate": 2.8596361341272013e-05, + "loss": 0.1171, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08278489112854004, + "step": 4070, + "valid_targets_mean": 1028.9, + "valid_targets_min": 637 + }, + { + "epoch": 2.9639869043288467, + "grad_norm": 0.3424202886018702, + "learning_rate": 2.8563599674809105e-05, + "loss": 0.0911, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03332003206014633, + "step": 4075, + "valid_targets_mean": 2879.2, + "valid_targets_min": 701 + }, + { + "epoch": 2.9676245907602765, + "grad_norm": 0.32697452159036244, + "learning_rate": 2.8530809846596565e-05, + "loss": 0.0953, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04063595458865166, + "step": 4080, + "valid_targets_mean": 3671.1, + "valid_targets_min": 3131 + }, + { + "epoch": 2.9712622771917063, + "grad_norm": 0.3970752577404366, + "learning_rate": 2.8497991964465076e-05, + "loss": 0.0968, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05217598006129265, + "step": 4085, + "valid_targets_mean": 3241.5, + "valid_targets_min": 2018 + }, + { + "epoch": 2.9748999636231357, + "grad_norm": 0.33723258400378203, + "learning_rate": 2.846514613633761e-05, + "loss": 0.1016, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04364777356386185, + "step": 4090, + "valid_targets_mean": 2491.6, + "valid_targets_min": 350 + }, + { + "epoch": 2.978537650054565, + "grad_norm": 0.28991520686134536, + "learning_rate": 2.8432272470229008e-05, + "loss": 0.0852, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04675818234682083, + "step": 4095, + "valid_targets_mean": 4862.8, + "valid_targets_min": 3577 + }, + { + "epoch": 2.982175336485995, + "grad_norm": 0.36647286676943636, + "learning_rate": 2.8399371074245672e-05, + "loss": 0.0944, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.037208572030067444, + "step": 4100, + "valid_targets_mean": 1999.4, + "valid_targets_min": 507 + }, + { + "epoch": 2.9858130229174247, + "grad_norm": 0.4685771199035281, + "learning_rate": 2.8366442056585208e-05, + "loss": 0.1079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09521611034870148, + "step": 4105, + "valid_targets_mean": 3278.5, + "valid_targets_min": 2814 + }, + { + "epoch": 2.989450709348854, + "grad_norm": 0.3984111552426283, + "learning_rate": 2.833348552553604e-05, + "loss": 0.1765, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.052019376307725906, + "step": 4110, + "valid_targets_mean": 4006.2, + "valid_targets_min": 2347 + }, + { + "epoch": 2.993088395780284, + "grad_norm": 0.4833864467579129, + "learning_rate": 2.8300501589477073e-05, + "loss": 0.0976, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0584319606423378, + "step": 4115, + "valid_targets_mean": 2447.4, + "valid_targets_min": 716 + }, + { + "epoch": 2.996726082211713, + "grad_norm": 0.3590579225471038, + "learning_rate": 2.8267490356877328e-05, + "loss": 0.0918, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04403100162744522, + "step": 4120, + "valid_targets_mean": 3209.6, + "valid_targets_min": 2245 + }, + { + "epoch": 3.0, + "grad_norm": 0.5293382784631392, + "learning_rate": 2.8234451936295625e-05, + "loss": 0.1386, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16387367248535156, + "step": 4125, + "valid_targets_mean": 7922.2, + "valid_targets_min": 3513 + }, + { + "epoch": 3.00363768643143, + "grad_norm": 0.4157191083709455, + "learning_rate": 2.820138643638015e-05, + "loss": 0.2045, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09862853586673737, + "step": 4130, + "valid_targets_mean": 7389.0, + "valid_targets_min": 5627 + }, + { + "epoch": 3.007275372862859, + "grad_norm": 0.609951744635308, + "learning_rate": 2.8168293965868153e-05, + "loss": 0.2077, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05807187408208847, + "step": 4135, + "valid_targets_mean": 1294.8, + "valid_targets_min": 422 + }, + { + "epoch": 3.010913059294289, + "grad_norm": 0.36728880436624556, + "learning_rate": 2.81351746335856e-05, + "loss": 0.1979, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1018591821193695, + "step": 4140, + "valid_targets_mean": 8539.9, + "valid_targets_min": 6761 + }, + { + "epoch": 3.0145507457257184, + "grad_norm": 0.3759124796465622, + "learning_rate": 2.8102028548446772e-05, + "loss": 0.1911, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0892845168709755, + "step": 4145, + "valid_targets_mean": 7579.0, + "valid_targets_min": 6174 + }, + { + "epoch": 3.018188432157148, + "grad_norm": 0.4187138393583101, + "learning_rate": 2.8068855819453936e-05, + "loss": 0.1937, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08817440271377563, + "step": 4150, + "valid_targets_mean": 7328.2, + "valid_targets_min": 4839 + }, + { + "epoch": 3.0218261185885775, + "grad_norm": 0.45727414482655276, + "learning_rate": 2.8035656555696964e-05, + "loss": 0.2021, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10495422035455704, + "step": 4155, + "valid_targets_mean": 6838.2, + "valid_targets_min": 5034 + }, + { + "epoch": 3.0254638050200073, + "grad_norm": 0.40080315523657833, + "learning_rate": 2.800243086635301e-05, + "loss": 0.2071, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10678562521934509, + "step": 4160, + "valid_targets_mean": 7248.2, + "valid_targets_min": 5567 + }, + { + "epoch": 3.0291014914514367, + "grad_norm": 0.3985406501941626, + "learning_rate": 2.7969178860686112e-05, + "loss": 0.1923, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08802725374698639, + "step": 4165, + "valid_targets_mean": 6234.0, + "valid_targets_min": 5603 + }, + { + "epoch": 3.0327391778828665, + "grad_norm": 0.630813294587274, + "learning_rate": 2.793590064804687e-05, + "loss": 0.189, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029058294370770454, + "step": 4170, + "valid_targets_mean": 602.0, + "valid_targets_min": 152 + }, + { + "epoch": 3.0363768643142963, + "grad_norm": 0.40322375970933194, + "learning_rate": 2.7902596337872042e-05, + "loss": 0.1808, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09033234417438507, + "step": 4175, + "valid_targets_mean": 7299.9, + "valid_targets_min": 6158 + }, + { + "epoch": 3.0400145507457257, + "grad_norm": 0.379790412369749, + "learning_rate": 2.7869266039684223e-05, + "loss": 0.1946, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09078914672136307, + "step": 4180, + "valid_targets_mean": 6140.0, + "valid_targets_min": 4648 + }, + { + "epoch": 3.0436522371771555, + "grad_norm": 0.39679208359579876, + "learning_rate": 2.7835909863091483e-05, + "loss": 0.1874, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09636875241994858, + "step": 4185, + "valid_targets_mean": 6603.0, + "valid_targets_min": 4882 + }, + { + "epoch": 3.047289923608585, + "grad_norm": 0.3569518121219989, + "learning_rate": 2.7802527917786955e-05, + "loss": 0.1883, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09586949646472931, + "step": 4190, + "valid_targets_mean": 8015.0, + "valid_targets_min": 6344 + }, + { + "epoch": 3.0509276100400147, + "grad_norm": 0.3759460603727716, + "learning_rate": 2.776912031354856e-05, + "loss": 0.1797, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08524218946695328, + "step": 4195, + "valid_targets_mean": 6305.2, + "valid_targets_min": 4622 + }, + { + "epoch": 3.054565296471444, + "grad_norm": 0.3812646081866477, + "learning_rate": 2.7735687160238576e-05, + "loss": 0.1882, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09892039000988007, + "step": 4200, + "valid_targets_mean": 7890.8, + "valid_targets_min": 5485 + }, + { + "epoch": 3.058202982902874, + "grad_norm": 0.7446977861894493, + "learning_rate": 2.7702228567803295e-05, + "loss": 0.1894, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08702052384614944, + "step": 4205, + "valid_targets_mean": 7016.6, + "valid_targets_min": 4960 + }, + { + "epoch": 3.061840669334303, + "grad_norm": 0.40477439759405764, + "learning_rate": 2.7668744646272673e-05, + "loss": 0.1856, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0894547551870346, + "step": 4210, + "valid_targets_mean": 6527.1, + "valid_targets_min": 5399 + }, + { + "epoch": 3.065478355765733, + "grad_norm": 0.43481589165661744, + "learning_rate": 2.7635235505759966e-05, + "loss": 0.1915, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09212228655815125, + "step": 4215, + "valid_targets_mean": 4995.0, + "valid_targets_min": 618 + }, + { + "epoch": 3.0691160421971624, + "grad_norm": 0.3878246078460517, + "learning_rate": 2.760170125646137e-05, + "loss": 0.204, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10145001113414764, + "step": 4220, + "valid_targets_mean": 7219.4, + "valid_targets_min": 5764 + }, + { + "epoch": 3.072753728628592, + "grad_norm": 0.3845034374752069, + "learning_rate": 2.756814200865562e-05, + "loss": 0.1759, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09536157548427582, + "step": 4225, + "valid_targets_mean": 7998.5, + "valid_targets_min": 5902 + }, + { + "epoch": 3.076391415060022, + "grad_norm": 0.37506597624282223, + "learning_rate": 2.7534557872703705e-05, + "loss": 0.1758, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07570267468690872, + "step": 4230, + "valid_targets_mean": 7377.4, + "valid_targets_min": 5015 + }, + { + "epoch": 3.0800291014914514, + "grad_norm": 0.3536142536170405, + "learning_rate": 2.7500948959048424e-05, + "loss": 0.1849, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07815571129322052, + "step": 4235, + "valid_targets_mean": 6794.0, + "valid_targets_min": 5283 + }, + { + "epoch": 3.083666787922881, + "grad_norm": 0.40381278087991784, + "learning_rate": 2.7467315378214077e-05, + "loss": 0.1923, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10751962661743164, + "step": 4240, + "valid_targets_mean": 6656.4, + "valid_targets_min": 4364 + }, + { + "epoch": 3.0873044743543105, + "grad_norm": 0.3876362365228525, + "learning_rate": 2.743365724080607e-05, + "loss": 0.1891, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08710598200559616, + "step": 4245, + "valid_targets_mean": 7178.4, + "valid_targets_min": 4448 + }, + { + "epoch": 3.0909421607857404, + "grad_norm": 0.40462905221399453, + "learning_rate": 2.7399974657510583e-05, + "loss": 0.1884, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11193549633026123, + "step": 4250, + "valid_targets_mean": 7619.8, + "valid_targets_min": 5737 + }, + { + "epoch": 3.0945798472171697, + "grad_norm": 0.38954879069338355, + "learning_rate": 2.736626773909417e-05, + "loss": 0.1663, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07876700907945633, + "step": 4255, + "valid_targets_mean": 8754.4, + "valid_targets_min": 5752 + }, + { + "epoch": 3.0982175336485995, + "grad_norm": 0.42229383582807084, + "learning_rate": 2.733253659640342e-05, + "loss": 0.1673, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08713915944099426, + "step": 4260, + "valid_targets_mean": 6341.0, + "valid_targets_min": 4543 + }, + { + "epoch": 3.101855220080029, + "grad_norm": 0.38484864674450003, + "learning_rate": 2.729878134036458e-05, + "loss": 0.1621, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09316037595272064, + "step": 4265, + "valid_targets_mean": 7181.9, + "valid_targets_min": 4563 + }, + { + "epoch": 3.1054929065114587, + "grad_norm": 0.7229300678268955, + "learning_rate": 2.726500208198319e-05, + "loss": 0.1951, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06466048955917358, + "step": 4270, + "valid_targets_mean": 1104.4, + "valid_targets_min": 161 + }, + { + "epoch": 3.1091305929428885, + "grad_norm": 0.40527799391391683, + "learning_rate": 2.7231198932343747e-05, + "loss": 0.1787, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08429668843746185, + "step": 4275, + "valid_targets_mean": 7018.2, + "valid_targets_min": 5424 + }, + { + "epoch": 3.112768279374318, + "grad_norm": 0.3806048830990148, + "learning_rate": 2.7197372002609283e-05, + "loss": 0.1957, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09963764995336533, + "step": 4280, + "valid_targets_mean": 7738.4, + "valid_targets_min": 5329 + }, + { + "epoch": 3.1164059658057477, + "grad_norm": 0.39183382973558967, + "learning_rate": 2.7163521404021064e-05, + "loss": 0.1992, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09629352390766144, + "step": 4285, + "valid_targets_mean": 6426.6, + "valid_targets_min": 5560 + }, + { + "epoch": 3.120043652237177, + "grad_norm": 0.40720795441029006, + "learning_rate": 2.7129647247898162e-05, + "loss": 0.1935, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09571010619401932, + "step": 4290, + "valid_targets_mean": 6169.9, + "valid_targets_min": 3446 + }, + { + "epoch": 3.123681338668607, + "grad_norm": 0.4359459552144956, + "learning_rate": 2.7095749645637135e-05, + "loss": 0.1883, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09977784752845764, + "step": 4295, + "valid_targets_mean": 7204.6, + "valid_targets_min": 5381 + }, + { + "epoch": 3.1273190251000362, + "grad_norm": 0.3652236931678867, + "learning_rate": 2.706182870871165e-05, + "loss": 0.1928, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09928131103515625, + "step": 4300, + "valid_targets_mean": 7600.4, + "valid_targets_min": 5241 + }, + { + "epoch": 3.130956711531466, + "grad_norm": 0.46640746735798405, + "learning_rate": 2.7027884548672082e-05, + "loss": 0.1871, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10208265483379364, + "step": 4305, + "valid_targets_mean": 7098.5, + "valid_targets_min": 5607 + }, + { + "epoch": 3.1345943979628954, + "grad_norm": 0.4194398163533373, + "learning_rate": 2.6993917277145222e-05, + "loss": 0.1918, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1010596752166748, + "step": 4310, + "valid_targets_mean": 6955.1, + "valid_targets_min": 5530 + }, + { + "epoch": 3.138232084394325, + "grad_norm": 0.42554295407767356, + "learning_rate": 2.6959927005833817e-05, + "loss": 0.1915, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09818993508815765, + "step": 4315, + "valid_targets_mean": 6748.1, + "valid_targets_min": 5364 + }, + { + "epoch": 3.141869770825755, + "grad_norm": 0.4072207599104509, + "learning_rate": 2.6925913846516292e-05, + "loss": 0.1941, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09272346645593643, + "step": 4320, + "valid_targets_mean": 6273.9, + "valid_targets_min": 5438 + }, + { + "epoch": 3.1455074572571844, + "grad_norm": 0.4040697728903848, + "learning_rate": 2.6891877911046303e-05, + "loss": 0.1883, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09046670794487, + "step": 4325, + "valid_targets_mean": 6419.5, + "valid_targets_min": 4879 + }, + { + "epoch": 3.149145143688614, + "grad_norm": 0.39769528334321097, + "learning_rate": 2.6857819311352433e-05, + "loss": 0.187, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09106683731079102, + "step": 4330, + "valid_targets_mean": 6483.4, + "valid_targets_min": 4309 + }, + { + "epoch": 3.1527828301200436, + "grad_norm": 0.4561461979908356, + "learning_rate": 2.6823738159437787e-05, + "loss": 0.167, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10984981060028076, + "step": 4335, + "valid_targets_mean": 6825.0, + "valid_targets_min": 5205 + }, + { + "epoch": 3.1564205165514734, + "grad_norm": 0.40172286940426016, + "learning_rate": 2.6789634567379627e-05, + "loss": 0.1867, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08709568530321121, + "step": 4340, + "valid_targets_mean": 6729.4, + "valid_targets_min": 5200 + }, + { + "epoch": 3.1600582029829027, + "grad_norm": 0.411155767865407, + "learning_rate": 2.6755508647329033e-05, + "loss": 0.1844, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08994196355342865, + "step": 4345, + "valid_targets_mean": 6094.0, + "valid_targets_min": 5302 + }, + { + "epoch": 3.1636958894143326, + "grad_norm": 0.42418459743757575, + "learning_rate": 2.6721360511510476e-05, + "loss": 0.1893, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09878411889076233, + "step": 4350, + "valid_targets_mean": 7140.0, + "valid_targets_min": 5270 + }, + { + "epoch": 3.167333575845762, + "grad_norm": 0.3781009942450753, + "learning_rate": 2.6687190272221522e-05, + "loss": 0.1824, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08825981616973877, + "step": 4355, + "valid_targets_mean": 7425.0, + "valid_targets_min": 5281 + }, + { + "epoch": 3.1709712622771917, + "grad_norm": 0.4093829436563135, + "learning_rate": 2.6652998041832394e-05, + "loss": 0.1909, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11961458623409271, + "step": 4360, + "valid_targets_mean": 7063.4, + "valid_targets_min": 5386 + }, + { + "epoch": 3.174608948708621, + "grad_norm": 0.38447587310115894, + "learning_rate": 2.6618783932785662e-05, + "loss": 0.189, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09797022491693497, + "step": 4365, + "valid_targets_mean": 6827.0, + "valid_targets_min": 4547 + }, + { + "epoch": 3.178246635140051, + "grad_norm": 0.5266241232704508, + "learning_rate": 2.658454805759582e-05, + "loss": 0.169, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.051892612129449844, + "step": 4370, + "valid_targets_mean": 2047.0, + "valid_targets_min": 165 + }, + { + "epoch": 3.1818843215714807, + "grad_norm": 0.41673798389041194, + "learning_rate": 2.655029052884895e-05, + "loss": 0.1756, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08972398936748505, + "step": 4375, + "valid_targets_mean": 6754.1, + "valid_targets_min": 5196 + }, + { + "epoch": 3.18552200800291, + "grad_norm": 0.42652273073404817, + "learning_rate": 2.6516011459202356e-05, + "loss": 0.184, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09283125400543213, + "step": 4380, + "valid_targets_mean": 5985.6, + "valid_targets_min": 5028 + }, + { + "epoch": 3.18915969443434, + "grad_norm": 0.3982660978686799, + "learning_rate": 2.6481710961384152e-05, + "loss": 0.1895, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08609230816364288, + "step": 4385, + "valid_targets_mean": 5963.9, + "valid_targets_min": 5237 + }, + { + "epoch": 3.1927973808657693, + "grad_norm": 0.41736312059142305, + "learning_rate": 2.6447389148192944e-05, + "loss": 0.1842, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09648395329713821, + "step": 4390, + "valid_targets_mean": 6319.0, + "valid_targets_min": 5412 + }, + { + "epoch": 3.196435067297199, + "grad_norm": 0.4286158361486627, + "learning_rate": 2.6413046132497425e-05, + "loss": 0.1752, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0759158506989479, + "step": 4395, + "valid_targets_mean": 7125.8, + "valid_targets_min": 5080 + }, + { + "epoch": 3.2000727537286284, + "grad_norm": 0.41887926461193087, + "learning_rate": 2.6378682027236022e-05, + "loss": 0.1881, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09263383597135544, + "step": 4400, + "valid_targets_mean": 6123.5, + "valid_targets_min": 4521 + }, + { + "epoch": 3.2037104401600582, + "grad_norm": 0.5966288024147036, + "learning_rate": 2.6344296945416495e-05, + "loss": 0.2253, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14416128396987915, + "step": 4405, + "valid_targets_mean": 6020.6, + "valid_targets_min": 2493 + }, + { + "epoch": 3.207348126591488, + "grad_norm": 0.5584477271217165, + "learning_rate": 2.6309891000115605e-05, + "loss": 0.2573, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10154866427183151, + "step": 4410, + "valid_targets_mean": 3953.5, + "valid_targets_min": 1177 + }, + { + "epoch": 3.2109858130229174, + "grad_norm": 0.5394821159629227, + "learning_rate": 2.627546430447872e-05, + "loss": 0.2641, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15555283427238464, + "step": 4415, + "valid_targets_mean": 4950.8, + "valid_targets_min": 2799 + }, + { + "epoch": 3.214623499454347, + "grad_norm": 0.5314424007817011, + "learning_rate": 2.6241016971719435e-05, + "loss": 0.2497, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1387774795293808, + "step": 4420, + "valid_targets_mean": 5312.0, + "valid_targets_min": 2973 + }, + { + "epoch": 3.2182611858857766, + "grad_norm": 0.5696551365247292, + "learning_rate": 2.6206549115119236e-05, + "loss": 0.2476, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11646552383899689, + "step": 4425, + "valid_targets_mean": 3775.5, + "valid_targets_min": 980 + }, + { + "epoch": 3.2218988723172064, + "grad_norm": 0.51535173820986, + "learning_rate": 2.6172060848027066e-05, + "loss": 0.2665, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13705241680145264, + "step": 4430, + "valid_targets_mean": 5055.4, + "valid_targets_min": 1512 + }, + { + "epoch": 3.2255365587486358, + "grad_norm": 0.48955455365063616, + "learning_rate": 2.613755228385903e-05, + "loss": 0.2422, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09314122796058655, + "step": 4435, + "valid_targets_mean": 3600.8, + "valid_targets_min": 1848 + }, + { + "epoch": 3.2291742451800656, + "grad_norm": 0.4797345544470494, + "learning_rate": 2.610302353609795e-05, + "loss": 0.2505, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09227770566940308, + "step": 4440, + "valid_targets_mean": 3369.8, + "valid_targets_min": 1381 + }, + { + "epoch": 3.232811931611495, + "grad_norm": 0.723822350922482, + "learning_rate": 2.6068474718293035e-05, + "loss": 0.2493, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13072994351387024, + "step": 4445, + "valid_targets_mean": 6283.2, + "valid_targets_min": 3600 + }, + { + "epoch": 3.2364496180429247, + "grad_norm": 0.5018503189954827, + "learning_rate": 2.603390594405949e-05, + "loss": 0.2629, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13003017008304596, + "step": 4450, + "valid_targets_mean": 5242.5, + "valid_targets_min": 2540 + }, + { + "epoch": 3.240087304474354, + "grad_norm": 0.47377234803779916, + "learning_rate": 2.5999317327078147e-05, + "loss": 0.2484, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11421793699264526, + "step": 4455, + "valid_targets_mean": 6246.5, + "valid_targets_min": 2157 + }, + { + "epoch": 3.243724990905784, + "grad_norm": 0.5354823099277923, + "learning_rate": 2.5964708981095106e-05, + "loss": 0.2521, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12844307720661163, + "step": 4460, + "valid_targets_mean": 4556.5, + "valid_targets_min": 1720 + }, + { + "epoch": 3.2473626773372137, + "grad_norm": 0.5300456426255018, + "learning_rate": 2.5930081019921326e-05, + "loss": 0.2541, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13994041085243225, + "step": 4465, + "valid_targets_mean": 5116.5, + "valid_targets_min": 1982 + }, + { + "epoch": 3.251000363768643, + "grad_norm": 0.555529782409327, + "learning_rate": 2.589543355743229e-05, + "loss": 0.2512, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1292266845703125, + "step": 4470, + "valid_targets_mean": 4464.5, + "valid_targets_min": 2718 + }, + { + "epoch": 3.254638050200073, + "grad_norm": 0.6186196883063052, + "learning_rate": 2.5860766707567595e-05, + "loss": 0.2525, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12426377087831497, + "step": 4475, + "valid_targets_mean": 3418.5, + "valid_targets_min": 1451 + }, + { + "epoch": 3.2582757366315023, + "grad_norm": 0.5705011622777583, + "learning_rate": 2.5826080584330607e-05, + "loss": 0.2514, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12919899821281433, + "step": 4480, + "valid_targets_mean": 3990.4, + "valid_targets_min": 1861 + }, + { + "epoch": 3.261913423062932, + "grad_norm": 0.6408474732732089, + "learning_rate": 2.5791375301788068e-05, + "loss": 0.2461, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13842612504959106, + "step": 4485, + "valid_targets_mean": 4065.4, + "valid_targets_min": 2100 + }, + { + "epoch": 3.2655511094943614, + "grad_norm": 0.5942224020772513, + "learning_rate": 2.575665097406972e-05, + "loss": 0.2549, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1419658064842224, + "step": 4490, + "valid_targets_mean": 4132.4, + "valid_targets_min": 3234 + }, + { + "epoch": 3.2691887959257913, + "grad_norm": 0.6110064121597131, + "learning_rate": 2.572190771536795e-05, + "loss": 0.2624, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15811282396316528, + "step": 4495, + "valid_targets_mean": 4246.2, + "valid_targets_min": 2650 + }, + { + "epoch": 3.2728264823572206, + "grad_norm": 0.5850679102329469, + "learning_rate": 2.5687145639937393e-05, + "loss": 0.2547, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10699371248483658, + "step": 4500, + "valid_targets_mean": 4096.5, + "valid_targets_min": 1719 + }, + { + "epoch": 3.2764641687886504, + "grad_norm": 0.5823704692763212, + "learning_rate": 2.5652364862094562e-05, + "loss": 0.26, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12624040246009827, + "step": 4505, + "valid_targets_mean": 3933.1, + "valid_targets_min": 1020 + }, + { + "epoch": 3.28010185522008, + "grad_norm": 0.6095269436983027, + "learning_rate": 2.5617565496217465e-05, + "loss": 0.2393, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12802812457084656, + "step": 4510, + "valid_targets_mean": 3551.1, + "valid_targets_min": 2130 + }, + { + "epoch": 3.2837395416515096, + "grad_norm": 0.561837887154538, + "learning_rate": 2.5582747656745264e-05, + "loss": 0.2367, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1000700294971466, + "step": 4515, + "valid_targets_mean": 3535.0, + "valid_targets_min": 865 + }, + { + "epoch": 3.2873772280829394, + "grad_norm": 0.6998816118874007, + "learning_rate": 2.554791145817785e-05, + "loss": 0.2431, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16453173756599426, + "step": 4520, + "valid_targets_mean": 3716.5, + "valid_targets_min": 3144 + }, + { + "epoch": 3.291014914514369, + "grad_norm": 0.5686742522606334, + "learning_rate": 2.5513057015075483e-05, + "loss": 0.2523, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1081462949514389, + "step": 4525, + "valid_targets_mean": 3262.2, + "valid_targets_min": 2050 + }, + { + "epoch": 3.2946526009457986, + "grad_norm": 0.5769016304387705, + "learning_rate": 2.547818444205846e-05, + "loss": 0.2513, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1250273585319519, + "step": 4530, + "valid_targets_mean": 3296.5, + "valid_targets_min": 1570 + }, + { + "epoch": 3.298290287377228, + "grad_norm": 0.5582993632863197, + "learning_rate": 2.5443293853806634e-05, + "loss": 0.238, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12309233099222183, + "step": 4535, + "valid_targets_mean": 4806.4, + "valid_targets_min": 1795 + }, + { + "epoch": 3.3019279738086578, + "grad_norm": 0.5380679956990834, + "learning_rate": 2.5408385365059167e-05, + "loss": 0.2423, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11501941084861755, + "step": 4540, + "valid_targets_mean": 4342.4, + "valid_targets_min": 1934 + }, + { + "epoch": 3.305565660240087, + "grad_norm": 0.6190869717964917, + "learning_rate": 2.5373459090614048e-05, + "loss": 0.2464, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1255384087562561, + "step": 4545, + "valid_targets_mean": 3675.4, + "valid_targets_min": 2557 + }, + { + "epoch": 3.309203346671517, + "grad_norm": 0.5980207444486617, + "learning_rate": 2.533851514532777e-05, + "loss": 0.2473, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13984902203083038, + "step": 4550, + "valid_targets_mean": 4550.4, + "valid_targets_min": 2663 + }, + { + "epoch": 3.3128410331029468, + "grad_norm": 0.5657486038291921, + "learning_rate": 2.5303553644114933e-05, + "loss": 0.2368, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11461549997329712, + "step": 4555, + "valid_targets_mean": 4604.0, + "valid_targets_min": 2016 + }, + { + "epoch": 3.316478719534376, + "grad_norm": 0.6391067539268364, + "learning_rate": 2.5268574701947875e-05, + "loss": 0.2444, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10942800343036652, + "step": 4560, + "valid_targets_mean": 2976.8, + "valid_targets_min": 1367 + }, + { + "epoch": 3.320116405965806, + "grad_norm": 0.6324630848619399, + "learning_rate": 2.523357843385628e-05, + "loss": 0.249, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12059381604194641, + "step": 4565, + "valid_targets_mean": 3542.5, + "valid_targets_min": 2026 + }, + { + "epoch": 3.3237540923972353, + "grad_norm": 0.5621070778291817, + "learning_rate": 2.5198564954926813e-05, + "loss": 0.2358, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1088690459728241, + "step": 4570, + "valid_targets_mean": 3247.2, + "valid_targets_min": 2033 + }, + { + "epoch": 3.327391778828665, + "grad_norm": 0.7140276800198336, + "learning_rate": 2.5163534380302747e-05, + "loss": 0.2473, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1216677725315094, + "step": 4575, + "valid_targets_mean": 2201.6, + "valid_targets_min": 1283 + }, + { + "epoch": 3.3310294652600945, + "grad_norm": 0.6831628459126925, + "learning_rate": 2.5128486825183556e-05, + "loss": 0.2463, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13061967492103577, + "step": 4580, + "valid_targets_mean": 2824.8, + "valid_targets_min": 1636 + }, + { + "epoch": 3.3346671516915243, + "grad_norm": 0.6110237889992081, + "learning_rate": 2.5093422404824574e-05, + "loss": 0.2479, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.132030189037323, + "step": 4585, + "valid_targets_mean": 3844.4, + "valid_targets_min": 1629 + }, + { + "epoch": 3.3383048381229536, + "grad_norm": 0.6169430947146991, + "learning_rate": 2.5058341234536595e-05, + "loss": 0.241, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11137215793132782, + "step": 4590, + "valid_targets_mean": 3048.6, + "valid_targets_min": 2080 + }, + { + "epoch": 3.3419425245543835, + "grad_norm": 0.6585005075664031, + "learning_rate": 2.5023243429685478e-05, + "loss": 0.2414, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11240460723638535, + "step": 4595, + "valid_targets_mean": 3111.8, + "valid_targets_min": 1153 + }, + { + "epoch": 3.345580210985813, + "grad_norm": 0.6124595903654756, + "learning_rate": 2.4988129105691797e-05, + "loss": 0.2431, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1315760314464569, + "step": 4600, + "valid_targets_mean": 4069.0, + "valid_targets_min": 1858 + }, + { + "epoch": 3.3492178974172426, + "grad_norm": 0.5653436666747438, + "learning_rate": 2.4952998378030463e-05, + "loss": 0.249, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10359971225261688, + "step": 4605, + "valid_targets_mean": 3415.2, + "valid_targets_min": 1639 + }, + { + "epoch": 3.3528555838486724, + "grad_norm": 0.6129612442079944, + "learning_rate": 2.491785136223031e-05, + "loss": 0.2427, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0996951311826706, + "step": 4610, + "valid_targets_mean": 3267.6, + "valid_targets_min": 1394 + }, + { + "epoch": 3.356493270280102, + "grad_norm": 0.5479975124748936, + "learning_rate": 2.488268817387374e-05, + "loss": 0.2417, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12831532955169678, + "step": 4615, + "valid_targets_mean": 4775.9, + "valid_targets_min": 1844 + }, + { + "epoch": 3.3601309567115316, + "grad_norm": 0.6187310549283155, + "learning_rate": 2.484750892859636e-05, + "loss": 0.2423, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10849634557962418, + "step": 4620, + "valid_targets_mean": 3317.1, + "valid_targets_min": 1305 + }, + { + "epoch": 3.363768643142961, + "grad_norm": 0.5984259670631525, + "learning_rate": 2.4812313742086543e-05, + "loss": 0.2418, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13298459351062775, + "step": 4625, + "valid_targets_mean": 4290.8, + "valid_targets_min": 2185 + }, + { + "epoch": 3.367406329574391, + "grad_norm": 0.8227318179204328, + "learning_rate": 2.477710273008512e-05, + "loss": 0.2419, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1442309319972992, + "step": 4630, + "valid_targets_mean": 4024.9, + "valid_targets_min": 2349 + }, + { + "epoch": 3.37104401600582, + "grad_norm": 0.5549022418185744, + "learning_rate": 2.4741876008384944e-05, + "loss": 0.235, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10434092581272125, + "step": 4635, + "valid_targets_mean": 3453.0, + "valid_targets_min": 1607 + }, + { + "epoch": 3.37468170243725, + "grad_norm": 0.7066155047541945, + "learning_rate": 2.470663369283055e-05, + "loss": 0.2504, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1278536319732666, + "step": 4640, + "valid_targets_mean": 3389.8, + "valid_targets_min": 1507 + }, + { + "epoch": 3.3783193888686793, + "grad_norm": 0.5784765069221337, + "learning_rate": 2.4671375899317737e-05, + "loss": 0.2452, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10821263492107391, + "step": 4645, + "valid_targets_mean": 3773.9, + "valid_targets_min": 1317 + }, + { + "epoch": 3.381957075300109, + "grad_norm": 0.5696639676647047, + "learning_rate": 2.46361027437932e-05, + "loss": 0.2451, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11597596108913422, + "step": 4650, + "valid_targets_mean": 3881.0, + "valid_targets_min": 3355 + }, + { + "epoch": 3.3855947617315385, + "grad_norm": 0.5433639440466803, + "learning_rate": 2.4600814342254174e-05, + "loss": 0.2455, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12038057297468185, + "step": 4655, + "valid_targets_mean": 3951.9, + "valid_targets_min": 2999 + }, + { + "epoch": 3.3892324481629683, + "grad_norm": 0.5615830113048655, + "learning_rate": 2.4565510810748005e-05, + "loss": 0.2417, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10884769260883331, + "step": 4660, + "valid_targets_mean": 4086.5, + "valid_targets_min": 1739 + }, + { + "epoch": 3.392870134594398, + "grad_norm": 0.5768727574288592, + "learning_rate": 2.4530192265371817e-05, + "loss": 0.2408, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1046416312456131, + "step": 4665, + "valid_targets_mean": 3298.0, + "valid_targets_min": 1140 + }, + { + "epoch": 3.3965078210258275, + "grad_norm": 0.5871558430756996, + "learning_rate": 2.4494858822272094e-05, + "loss": 0.2272, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12178435921669006, + "step": 4670, + "valid_targets_mean": 4012.1, + "valid_targets_min": 2779 + }, + { + "epoch": 3.4001455074572573, + "grad_norm": 0.6115781845159424, + "learning_rate": 2.4459510597644322e-05, + "loss": 0.2385, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11243709921836853, + "step": 4675, + "valid_targets_mean": 3636.9, + "valid_targets_min": 1183 + }, + { + "epoch": 3.4037831938886867, + "grad_norm": 0.5917434187089635, + "learning_rate": 2.442414770773259e-05, + "loss": 0.2265, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13976138830184937, + "step": 4680, + "valid_targets_mean": 4269.0, + "valid_targets_min": 3041 + }, + { + "epoch": 3.4074208803201165, + "grad_norm": 0.5996350642917458, + "learning_rate": 2.4388770268829217e-05, + "loss": 0.249, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12027683109045029, + "step": 4685, + "valid_targets_mean": 3350.1, + "valid_targets_min": 1726 + }, + { + "epoch": 3.411058566751546, + "grad_norm": 0.5102262043038736, + "learning_rate": 2.435337839727436e-05, + "loss": 0.2344, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10933533310890198, + "step": 4690, + "valid_targets_mean": 4747.4, + "valid_targets_min": 1743 + }, + { + "epoch": 3.4146962531829756, + "grad_norm": 0.5760416761560045, + "learning_rate": 2.431797220945564e-05, + "loss": 0.2417, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11261209845542908, + "step": 4695, + "valid_targets_mean": 3702.8, + "valid_targets_min": 1234 + }, + { + "epoch": 3.4183339396144055, + "grad_norm": 0.6207616430002428, + "learning_rate": 2.428255182180777e-05, + "loss": 0.2481, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12631835043430328, + "step": 4700, + "valid_targets_mean": 4079.0, + "valid_targets_min": 2303 + }, + { + "epoch": 3.421971626045835, + "grad_norm": 0.602364427851666, + "learning_rate": 2.4247117350812143e-05, + "loss": 0.2491, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12005666643381119, + "step": 4705, + "valid_targets_mean": 3830.8, + "valid_targets_min": 1651 + }, + { + "epoch": 3.4256093124772646, + "grad_norm": 0.6320036194744546, + "learning_rate": 2.421166891299649e-05, + "loss": 0.2386, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12540699541568756, + "step": 4710, + "valid_targets_mean": 2876.2, + "valid_targets_min": 1374 + }, + { + "epoch": 3.429246998908694, + "grad_norm": 0.6520739375231831, + "learning_rate": 2.4176206624934438e-05, + "loss": 0.2331, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12044782191514969, + "step": 4715, + "valid_targets_mean": 3467.5, + "valid_targets_min": 1625 + }, + { + "epoch": 3.432884685340124, + "grad_norm": 0.5912498180162667, + "learning_rate": 2.414073060324519e-05, + "loss": 0.2281, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1140735000371933, + "step": 4720, + "valid_targets_mean": 4089.5, + "valid_targets_min": 997 + }, + { + "epoch": 3.436522371771553, + "grad_norm": 0.5645138860499421, + "learning_rate": 2.41052409645931e-05, + "loss": 0.2289, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09771278500556946, + "step": 4725, + "valid_targets_mean": 3618.6, + "valid_targets_min": 1179 + }, + { + "epoch": 3.440160058202983, + "grad_norm": 0.6003279319809036, + "learning_rate": 2.4069737825687302e-05, + "loss": 0.2291, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10813167691230774, + "step": 4730, + "valid_targets_mean": 3079.1, + "valid_targets_min": 1444 + }, + { + "epoch": 3.4437977446344123, + "grad_norm": 0.6328047959810414, + "learning_rate": 2.4034221303281328e-05, + "loss": 0.2195, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0983259379863739, + "step": 4735, + "valid_targets_mean": 3486.8, + "valid_targets_min": 1504 + }, + { + "epoch": 3.447435431065842, + "grad_norm": 0.5820360982349623, + "learning_rate": 2.3998691514172726e-05, + "loss": 0.2263, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11347955465316772, + "step": 4740, + "valid_targets_mean": 3998.4, + "valid_targets_min": 2302 + }, + { + "epoch": 3.4510731174972715, + "grad_norm": 0.65303598294505, + "learning_rate": 2.3963148575202665e-05, + "loss": 0.2357, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14733964204788208, + "step": 4745, + "valid_targets_mean": 4282.5, + "valid_targets_min": 1818 + }, + { + "epoch": 3.4547108039287013, + "grad_norm": 0.6149704465018314, + "learning_rate": 2.392759260325556e-05, + "loss": 0.2554, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12628173828125, + "step": 4750, + "valid_targets_mean": 4221.9, + "valid_targets_min": 2633 + }, + { + "epoch": 3.458348490360131, + "grad_norm": 0.6487411196205217, + "learning_rate": 2.389202371525869e-05, + "loss": 0.2284, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11891745030879974, + "step": 4755, + "valid_targets_mean": 3522.4, + "valid_targets_min": 2481 + }, + { + "epoch": 3.4619861767915605, + "grad_norm": 0.6390906801368823, + "learning_rate": 2.38564420281818e-05, + "loss": 0.2377, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12932664155960083, + "step": 4760, + "valid_targets_mean": 4058.8, + "valid_targets_min": 2013 + }, + { + "epoch": 3.4656238632229903, + "grad_norm": 0.6701161188325943, + "learning_rate": 2.382084765903674e-05, + "loss": 0.2286, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12234342843294144, + "step": 4765, + "valid_targets_mean": 3723.5, + "valid_targets_min": 1744 + }, + { + "epoch": 3.4692615496544197, + "grad_norm": 0.5987488498870587, + "learning_rate": 2.3785240724877054e-05, + "loss": 0.2381, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12440264225006104, + "step": 4770, + "valid_targets_mean": 4180.9, + "valid_targets_min": 1914 + }, + { + "epoch": 3.4728992360858495, + "grad_norm": 0.6030303001173395, + "learning_rate": 2.37496213427976e-05, + "loss": 0.2278, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10155526548624039, + "step": 4775, + "valid_targets_mean": 3668.6, + "valid_targets_min": 3065 + }, + { + "epoch": 3.476536922517279, + "grad_norm": 0.5663060451705713, + "learning_rate": 2.371398962993419e-05, + "loss": 0.2212, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13348785042762756, + "step": 4780, + "valid_targets_mean": 4473.2, + "valid_targets_min": 2894 + }, + { + "epoch": 3.4801746089487087, + "grad_norm": 0.6257290592880157, + "learning_rate": 2.3678345703463173e-05, + "loss": 0.2384, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11792676895856857, + "step": 4785, + "valid_targets_mean": 3395.6, + "valid_targets_min": 2359 + }, + { + "epoch": 3.4838122953801385, + "grad_norm": 0.6315762656485211, + "learning_rate": 2.364268968060107e-05, + "loss": 0.2273, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10366246849298477, + "step": 4790, + "valid_targets_mean": 3114.5, + "valid_targets_min": 1463 + }, + { + "epoch": 3.487449981811568, + "grad_norm": 0.6994636094739689, + "learning_rate": 2.3607021678604173e-05, + "loss": 0.2369, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11958909034729004, + "step": 4795, + "valid_targets_mean": 3588.1, + "valid_targets_min": 1426 + }, + { + "epoch": 3.491087668242997, + "grad_norm": 0.644016529360144, + "learning_rate": 2.357134181476818e-05, + "loss": 0.2437, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11642561107873917, + "step": 4800, + "valid_targets_mean": 3405.5, + "valid_targets_min": 1661 + }, + { + "epoch": 3.494725354674427, + "grad_norm": 0.5578170143343268, + "learning_rate": 2.3535650206427786e-05, + "loss": 0.2401, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12392961233854294, + "step": 4805, + "valid_targets_mean": 4457.8, + "valid_targets_min": 1529 + }, + { + "epoch": 3.498363041105857, + "grad_norm": 0.7024853064320605, + "learning_rate": 2.3499946970956308e-05, + "loss": 0.2567, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14314286410808563, + "step": 4810, + "valid_targets_mean": 3272.9, + "valid_targets_min": 1450 + }, + { + "epoch": 3.502000727537286, + "grad_norm": 0.6764325873286284, + "learning_rate": 2.3464232225765315e-05, + "loss": 0.2333, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10606211423873901, + "step": 4815, + "valid_targets_mean": 3103.2, + "valid_targets_min": 1625 + }, + { + "epoch": 3.505638413968716, + "grad_norm": 0.5626295185388045, + "learning_rate": 2.34285060883042e-05, + "loss": 0.244, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10113455355167389, + "step": 4820, + "valid_targets_mean": 3952.2, + "valid_targets_min": 1174 + }, + { + "epoch": 3.5092761004001454, + "grad_norm": 0.5774822425933687, + "learning_rate": 2.3392768676059848e-05, + "loss": 0.2435, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.122687928378582, + "step": 4825, + "valid_targets_mean": 4039.2, + "valid_targets_min": 2882 + }, + { + "epoch": 3.512913786831575, + "grad_norm": 0.5898536126089041, + "learning_rate": 2.335702010655621e-05, + "loss": 0.2439, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1370169073343277, + "step": 4830, + "valid_targets_mean": 4594.0, + "valid_targets_min": 1946 + }, + { + "epoch": 3.5165514732630045, + "grad_norm": 0.5836809821353771, + "learning_rate": 2.3321260497353918e-05, + "loss": 0.2408, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11454842984676361, + "step": 4835, + "valid_targets_mean": 4082.6, + "valid_targets_min": 2000 + }, + { + "epoch": 3.5201891596944344, + "grad_norm": 0.6146047187486914, + "learning_rate": 2.3285489966049922e-05, + "loss": 0.2416, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12982092797756195, + "step": 4840, + "valid_targets_mean": 4596.1, + "valid_targets_min": 2012 + }, + { + "epoch": 3.523826846125864, + "grad_norm": 0.6791293157154603, + "learning_rate": 2.3249708630277078e-05, + "loss": 0.2422, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12032167613506317, + "step": 4845, + "valid_targets_mean": 3094.4, + "valid_targets_min": 719 + }, + { + "epoch": 3.5274645325572935, + "grad_norm": 0.5749726832894869, + "learning_rate": 2.3213916607703787e-05, + "loss": 0.2403, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11222009360790253, + "step": 4850, + "valid_targets_mean": 4496.2, + "valid_targets_min": 3238 + }, + { + "epoch": 3.531102218988723, + "grad_norm": 0.5957305069749437, + "learning_rate": 2.3178114016033587e-05, + "loss": 0.2286, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10017383843660355, + "step": 4855, + "valid_targets_mean": 3816.8, + "valid_targets_min": 1372 + }, + { + "epoch": 3.5347399054201527, + "grad_norm": 0.6434776345954005, + "learning_rate": 2.3142300973004775e-05, + "loss": 0.2248, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12635159492492676, + "step": 4860, + "valid_targets_mean": 3306.1, + "valid_targets_min": 2370 + }, + { + "epoch": 3.5383775918515825, + "grad_norm": 0.5627040250088954, + "learning_rate": 2.3106477596390003e-05, + "loss": 0.2324, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10443396866321564, + "step": 4865, + "valid_targets_mean": 2958.1, + "valid_targets_min": 1576 + }, + { + "epoch": 3.542015278283012, + "grad_norm": 0.6162488499053235, + "learning_rate": 2.307064400399593e-05, + "loss": 0.2305, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12590399384498596, + "step": 4870, + "valid_targets_mean": 3758.5, + "valid_targets_min": 1426 + }, + { + "epoch": 3.5456529647144417, + "grad_norm": 0.7769577516941126, + "learning_rate": 2.3034800313662794e-05, + "loss": 0.2416, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10565987229347229, + "step": 4875, + "valid_targets_mean": 3117.0, + "valid_targets_min": 1728 + }, + { + "epoch": 3.5492906511458715, + "grad_norm": 0.8295639051559466, + "learning_rate": 2.2998946643264044e-05, + "loss": 0.2337, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12615841627120972, + "step": 4880, + "valid_targets_mean": 3264.2, + "valid_targets_min": 1362 + }, + { + "epoch": 3.552928337577301, + "grad_norm": 0.5816147754745505, + "learning_rate": 2.2963083110705953e-05, + "loss": 0.2377, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11687877774238586, + "step": 4885, + "valid_targets_mean": 3836.0, + "valid_targets_min": 852 + }, + { + "epoch": 3.5565660240087302, + "grad_norm": 0.7340327846949457, + "learning_rate": 2.2927209833927217e-05, + "loss": 0.2365, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1458575427532196, + "step": 4890, + "valid_targets_mean": 2764.1, + "valid_targets_min": 1195 + }, + { + "epoch": 3.56020371044016, + "grad_norm": 0.6058817140274564, + "learning_rate": 2.289132693089859e-05, + "loss": 0.2302, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12383746355772018, + "step": 4895, + "valid_targets_mean": 4457.5, + "valid_targets_min": 1564 + }, + { + "epoch": 3.56384139687159, + "grad_norm": 0.5551851650869939, + "learning_rate": 2.2855434519622456e-05, + "loss": 0.2297, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10075017064809799, + "step": 4900, + "valid_targets_mean": 4371.2, + "valid_targets_min": 1975 + }, + { + "epoch": 3.567479083303019, + "grad_norm": 0.5663142318750257, + "learning_rate": 2.2819532718132508e-05, + "loss": 0.218, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11594033241271973, + "step": 4905, + "valid_targets_mean": 3965.8, + "valid_targets_min": 2426 + }, + { + "epoch": 3.571116769734449, + "grad_norm": 0.5144626905581949, + "learning_rate": 2.2783621644493286e-05, + "loss": 0.2243, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11281321942806244, + "step": 4910, + "valid_targets_mean": 6384.6, + "valid_targets_min": 2678 + }, + { + "epoch": 3.5747544561658784, + "grad_norm": 0.7000020929918617, + "learning_rate": 2.2747701416799823e-05, + "loss": 0.2297, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10937657952308655, + "step": 4915, + "valid_targets_mean": 3543.2, + "valid_targets_min": 2826 + }, + { + "epoch": 3.578392142597308, + "grad_norm": 0.5904294168263171, + "learning_rate": 2.2711772153177284e-05, + "loss": 0.2191, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09370313584804535, + "step": 4920, + "valid_targets_mean": 3155.8, + "valid_targets_min": 993 + }, + { + "epoch": 3.5820298290287376, + "grad_norm": 0.5790807760011016, + "learning_rate": 2.267583397178051e-05, + "loss": 0.2192, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12285852432250977, + "step": 4925, + "valid_targets_mean": 4166.5, + "valid_targets_min": 2724 + }, + { + "epoch": 3.5856675154601674, + "grad_norm": 0.6174582670324816, + "learning_rate": 2.2639886990793705e-05, + "loss": 0.2217, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10573258996009827, + "step": 4930, + "valid_targets_mean": 3079.5, + "valid_targets_min": 1360 + }, + { + "epoch": 3.589305201891597, + "grad_norm": 0.5946542008003131, + "learning_rate": 2.2603931328429982e-05, + "loss": 0.2219, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11140136420726776, + "step": 4935, + "valid_targets_mean": 4114.9, + "valid_targets_min": 2414 + }, + { + "epoch": 3.5929428883230266, + "grad_norm": 0.6130045856646092, + "learning_rate": 2.2567967102931025e-05, + "loss": 0.2404, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13183709979057312, + "step": 4940, + "valid_targets_mean": 4524.8, + "valid_targets_min": 2425 + }, + { + "epoch": 3.596580574754456, + "grad_norm": 0.6489841303242022, + "learning_rate": 2.253199443256666e-05, + "loss": 0.2297, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11220933496952057, + "step": 4945, + "valid_targets_mean": 3035.2, + "valid_targets_min": 1720 + }, + { + "epoch": 3.6002182611858857, + "grad_norm": 0.6673242860013848, + "learning_rate": 2.249601343563449e-05, + "loss": 0.2274, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11637795716524124, + "step": 4950, + "valid_targets_mean": 3260.4, + "valid_targets_min": 1621 + }, + { + "epoch": 3.6038559476173155, + "grad_norm": 0.8341978076482921, + "learning_rate": 2.2460024230459522e-05, + "loss": 0.2173, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11685831099748611, + "step": 4955, + "valid_targets_mean": 3214.6, + "valid_targets_min": 1858 + }, + { + "epoch": 3.607493634048745, + "grad_norm": 0.6519440918168503, + "learning_rate": 2.242402693539371e-05, + "loss": 0.2295, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10721758008003235, + "step": 4960, + "valid_targets_mean": 3254.2, + "valid_targets_min": 1254 + }, + { + "epoch": 3.6111313204801747, + "grad_norm": 0.6511399261753961, + "learning_rate": 2.2388021668815657e-05, + "loss": 0.2241, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11937984079122543, + "step": 4965, + "valid_targets_mean": 2994.4, + "valid_targets_min": 1039 + }, + { + "epoch": 3.614769006911604, + "grad_norm": 0.6029276201017738, + "learning_rate": 2.2352008549130156e-05, + "loss": 0.2346, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14448794722557068, + "step": 4970, + "valid_targets_mean": 4625.4, + "valid_targets_min": 2526 + }, + { + "epoch": 3.618406693343034, + "grad_norm": 0.5916942506196884, + "learning_rate": 2.2315987694767837e-05, + "loss": 0.2271, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12231980264186859, + "step": 4975, + "valid_targets_mean": 3571.8, + "valid_targets_min": 2409 + }, + { + "epoch": 3.6220443797744633, + "grad_norm": 0.5901633974957422, + "learning_rate": 2.2279959224184754e-05, + "loss": 0.228, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09833040833473206, + "step": 4980, + "valid_targets_mean": 3691.5, + "valid_targets_min": 2918 + }, + { + "epoch": 3.625682066205893, + "grad_norm": 0.6036317214082011, + "learning_rate": 2.2243923255862025e-05, + "loss": 0.2314, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11447164416313171, + "step": 4985, + "valid_targets_mean": 4063.1, + "valid_targets_min": 2179 + }, + { + "epoch": 3.629319752637323, + "grad_norm": 0.4923362088079625, + "learning_rate": 2.22078799083054e-05, + "loss": 0.1957, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.053352825343608856, + "step": 4990, + "valid_targets_mean": 2871.4, + "valid_targets_min": 955 + }, + { + "epoch": 3.6329574390687522, + "grad_norm": 0.4414601428000912, + "learning_rate": 2.217182930004491e-05, + "loss": 0.1178, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04405904933810234, + "step": 4995, + "valid_targets_mean": 1788.9, + "valid_targets_min": 669 + }, + { + "epoch": 3.636595125500182, + "grad_norm": 0.3623385580015003, + "learning_rate": 2.213577154963448e-05, + "loss": 0.0872, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04181670397520065, + "step": 5000, + "valid_targets_mean": 3755.8, + "valid_targets_min": 2651 + }, + { + "epoch": 3.6402328119316114, + "grad_norm": 0.5882824323633015, + "learning_rate": 2.209970677565148e-05, + "loss": 0.0879, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04783633351325989, + "step": 5005, + "valid_targets_mean": 3069.8, + "valid_targets_min": 710 + }, + { + "epoch": 3.643870498363041, + "grad_norm": 0.3425382352154787, + "learning_rate": 2.2063635096696426e-05, + "loss": 0.1012, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030300896614789963, + "step": 5010, + "valid_targets_mean": 3618.5, + "valid_targets_min": 1365 + }, + { + "epoch": 3.6475081847944706, + "grad_norm": 0.43597490122492083, + "learning_rate": 2.2027556631392496e-05, + "loss": 0.0884, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0474068745970726, + "step": 5015, + "valid_targets_mean": 3279.9, + "valid_targets_min": 889 + }, + { + "epoch": 3.6511458712259004, + "grad_norm": 0.7739119227738408, + "learning_rate": 2.1991471498385226e-05, + "loss": 0.1192, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06611474603414536, + "step": 5020, + "valid_targets_mean": 1652.4, + "valid_targets_min": 559 + }, + { + "epoch": 3.65478355765733, + "grad_norm": 0.3368336340241663, + "learning_rate": 2.1955379816342046e-05, + "loss": 0.1197, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.037110213190317154, + "step": 5025, + "valid_targets_mean": 3449.0, + "valid_targets_min": 1013 + }, + { + "epoch": 3.6584212440887596, + "grad_norm": 0.3895180495728156, + "learning_rate": 2.191928170395194e-05, + "loss": 0.0987, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05450309067964554, + "step": 5030, + "valid_targets_mean": 3726.1, + "valid_targets_min": 3222 + }, + { + "epoch": 3.662058930520189, + "grad_norm": 0.4692469353013934, + "learning_rate": 2.1883177279925034e-05, + "loss": 0.1313, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09273035824298859, + "step": 5035, + "valid_targets_mean": 3205.8, + "valid_targets_min": 1949 + }, + { + "epoch": 3.6656966169516187, + "grad_norm": 0.30731264298330335, + "learning_rate": 2.184706666299222e-05, + "loss": 0.0753, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03514016419649124, + "step": 5040, + "valid_targets_mean": 4059.8, + "valid_targets_min": 886 + }, + { + "epoch": 3.6693343033830486, + "grad_norm": 0.5265068904772265, + "learning_rate": 2.1810949971904732e-05, + "loss": 0.0902, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04909496754407883, + "step": 5045, + "valid_targets_mean": 1858.2, + "valid_targets_min": 447 + }, + { + "epoch": 3.672971989814478, + "grad_norm": 0.37608004741068324, + "learning_rate": 2.17748273254338e-05, + "loss": 0.0874, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.037385012954473495, + "step": 5050, + "valid_targets_mean": 3195.6, + "valid_targets_min": 758 + }, + { + "epoch": 3.6766096762459077, + "grad_norm": 0.5040799820137605, + "learning_rate": 2.1738698842370234e-05, + "loss": 0.108, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05880262702703476, + "step": 5055, + "valid_targets_mean": 1808.4, + "valid_targets_min": 604 + }, + { + "epoch": 3.680247362677337, + "grad_norm": 0.390054132387286, + "learning_rate": 2.170256464152404e-05, + "loss": 0.1081, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04707545414566994, + "step": 5060, + "valid_targets_mean": 3314.6, + "valid_targets_min": 1743 + }, + { + "epoch": 3.683885049108767, + "grad_norm": 0.35522811086864176, + "learning_rate": 2.166642484172401e-05, + "loss": 0.0873, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03446710482239723, + "step": 5065, + "valid_targets_mean": 2440.4, + "valid_targets_min": 574 + }, + { + "epoch": 3.6875227355401963, + "grad_norm": 0.40865240556856547, + "learning_rate": 2.1630279561817375e-05, + "loss": 0.086, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04230947047472, + "step": 5070, + "valid_targets_mean": 2535.6, + "valid_targets_min": 523 + }, + { + "epoch": 3.691160421971626, + "grad_norm": 0.4830931001993365, + "learning_rate": 2.159412892066937e-05, + "loss": 0.0865, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06634676456451416, + "step": 5075, + "valid_targets_mean": 3398.1, + "valid_targets_min": 2635 + }, + { + "epoch": 3.694798108403056, + "grad_norm": 0.3865465906936746, + "learning_rate": 2.155797303716286e-05, + "loss": 0.0962, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04053504765033722, + "step": 5080, + "valid_targets_mean": 3164.8, + "valid_targets_min": 812 + }, + { + "epoch": 3.6984357948344853, + "grad_norm": 0.3959650977386073, + "learning_rate": 2.152181203019796e-05, + "loss": 0.1003, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04800565913319588, + "step": 5085, + "valid_targets_mean": 2463.9, + "valid_targets_min": 856 + }, + { + "epoch": 3.7020734812659146, + "grad_norm": 0.4555938541601627, + "learning_rate": 2.1485646018691627e-05, + "loss": 0.1004, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06002945452928543, + "step": 5090, + "valid_targets_mean": 2995.6, + "valid_targets_min": 718 + }, + { + "epoch": 3.7057111676973444, + "grad_norm": 0.520485937690199, + "learning_rate": 2.1449475121577268e-05, + "loss": 0.1201, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1282571256160736, + "step": 5095, + "valid_targets_mean": 2866.9, + "valid_targets_min": 1297 + }, + { + "epoch": 3.7093488541287742, + "grad_norm": 0.5057579563644704, + "learning_rate": 2.141329945780439e-05, + "loss": 0.1144, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04132267087697983, + "step": 5100, + "valid_targets_mean": 1661.1, + "valid_targets_min": 786 + }, + { + "epoch": 3.7129865405602036, + "grad_norm": 0.6248771302393089, + "learning_rate": 2.137711914633812e-05, + "loss": 0.0891, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06078476831316948, + "step": 5105, + "valid_targets_mean": 1624.2, + "valid_targets_min": 550 + }, + { + "epoch": 3.7166242269916334, + "grad_norm": 0.4362637890362872, + "learning_rate": 2.134093430615892e-05, + "loss": 0.1224, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04599619656801224, + "step": 5110, + "valid_targets_mean": 3254.6, + "valid_targets_min": 1304 + }, + { + "epoch": 3.720261913423063, + "grad_norm": 0.41207500508806333, + "learning_rate": 2.1304745056262123e-05, + "loss": 0.1122, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04035510867834091, + "step": 5115, + "valid_targets_mean": 2833.0, + "valid_targets_min": 1247 + }, + { + "epoch": 3.7238995998544926, + "grad_norm": 0.5194410357174574, + "learning_rate": 2.1268551515657552e-05, + "loss": 0.0926, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.057324860244989395, + "step": 5120, + "valid_targets_mean": 3056.5, + "valid_targets_min": 800 + }, + { + "epoch": 3.727537286285922, + "grad_norm": 0.5643513921730904, + "learning_rate": 2.123235380336917e-05, + "loss": 0.2811, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12033820897340775, + "step": 5125, + "valid_targets_mean": 2580.8, + "valid_targets_min": 987 + }, + { + "epoch": 3.7311749727173518, + "grad_norm": 0.5740959275971806, + "learning_rate": 2.1196152038434632e-05, + "loss": 0.1182, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0973605364561081, + "step": 5130, + "valid_targets_mean": 2245.6, + "valid_targets_min": 776 + }, + { + "epoch": 3.7348126591487816, + "grad_norm": 0.3912315671033847, + "learning_rate": 2.1159946339904936e-05, + "loss": 0.0973, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.027969777584075928, + "step": 5135, + "valid_targets_mean": 1361.5, + "valid_targets_min": 658 + }, + { + "epoch": 3.738450345580211, + "grad_norm": 0.504089971088435, + "learning_rate": 2.1123736826844003e-05, + "loss": 0.1947, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04456985741853714, + "step": 5140, + "valid_targets_mean": 1729.0, + "valid_targets_min": 666 + }, + { + "epoch": 3.7420880320116408, + "grad_norm": 0.3065164717284258, + "learning_rate": 2.1087523618328313e-05, + "loss": 0.0861, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03576495870947838, + "step": 5145, + "valid_targets_mean": 4975.0, + "valid_targets_min": 3529 + }, + { + "epoch": 3.74572571844307, + "grad_norm": 0.4744028599670395, + "learning_rate": 2.105130683344649e-05, + "loss": 0.0974, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05837929621338844, + "step": 5150, + "valid_targets_mean": 3988.8, + "valid_targets_min": 1746 + }, + { + "epoch": 3.7493634048745, + "grad_norm": 0.44398787454141025, + "learning_rate": 2.1015086591298918e-05, + "loss": 0.0951, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05947793275117874, + "step": 5155, + "valid_targets_mean": 3222.4, + "valid_targets_min": 1226 + }, + { + "epoch": 3.7530010913059293, + "grad_norm": 0.41856506802288646, + "learning_rate": 2.0978863010997356e-05, + "loss": 0.0855, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.056279443204402924, + "step": 5160, + "valid_targets_mean": 4403.6, + "valid_targets_min": 3275 + }, + { + "epoch": 3.756638777737359, + "grad_norm": 0.35157058242617284, + "learning_rate": 2.0942636211664528e-05, + "loss": 0.0871, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03564505651593208, + "step": 5165, + "valid_targets_mean": 2546.2, + "valid_targets_min": 723 + }, + { + "epoch": 3.760276464168789, + "grad_norm": 0.42419352841404584, + "learning_rate": 2.0906406312433764e-05, + "loss": 0.085, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.049980275332927704, + "step": 5170, + "valid_targets_mean": 3348.6, + "valid_targets_min": 2234 + }, + { + "epoch": 3.7639141506002183, + "grad_norm": 0.3177101789352681, + "learning_rate": 2.0870173432448566e-05, + "loss": 0.0822, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02766517549753189, + "step": 5175, + "valid_targets_mean": 3603.0, + "valid_targets_min": 1421 + }, + { + "epoch": 3.7675518370316476, + "grad_norm": 0.4284633926238813, + "learning_rate": 2.0833937690862266e-05, + "loss": 0.0847, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.045555345714092255, + "step": 5180, + "valid_targets_mean": 2849.2, + "valid_targets_min": 1068 + }, + { + "epoch": 3.7711895234630775, + "grad_norm": 0.4753180893811327, + "learning_rate": 2.0797699206837575e-05, + "loss": 0.1298, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03503882512450218, + "step": 5185, + "valid_targets_mean": 1587.9, + "valid_targets_min": 535 + }, + { + "epoch": 3.7748272098945073, + "grad_norm": 0.35524968550342634, + "learning_rate": 2.076145809954625e-05, + "loss": 0.0713, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02866739220917225, + "step": 5190, + "valid_targets_mean": 1702.4, + "valid_targets_min": 585 + }, + { + "epoch": 3.7784648963259366, + "grad_norm": 0.4200158367144281, + "learning_rate": 2.0725214488168646e-05, + "loss": 0.0829, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03983025997877121, + "step": 5195, + "valid_targets_mean": 2588.6, + "valid_targets_min": 582 + }, + { + "epoch": 3.7821025827573664, + "grad_norm": 0.4100584157517041, + "learning_rate": 2.068896849189338e-05, + "loss": 0.0772, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04206583648920059, + "step": 5200, + "valid_targets_mean": 3144.0, + "valid_targets_min": 516 + }, + { + "epoch": 3.785740269188796, + "grad_norm": 0.4870495427512827, + "learning_rate": 2.0652720229916906e-05, + "loss": 0.1256, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06180163100361824, + "step": 5205, + "valid_targets_mean": 3241.4, + "valid_targets_min": 999 + }, + { + "epoch": 3.7893779556202256, + "grad_norm": 0.49808272411489357, + "learning_rate": 2.0616469821443116e-05, + "loss": 0.1001, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0861666277050972, + "step": 5210, + "valid_targets_mean": 2021.1, + "valid_targets_min": 696 + }, + { + "epoch": 3.793015642051655, + "grad_norm": 0.49110982468094566, + "learning_rate": 2.058021738568297e-05, + "loss": 0.1118, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05045432597398758, + "step": 5215, + "valid_targets_mean": 2048.5, + "valid_targets_min": 635 + }, + { + "epoch": 3.796653328483085, + "grad_norm": 0.4117035197742263, + "learning_rate": 2.05439630418541e-05, + "loss": 0.0942, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04098989814519882, + "step": 5220, + "valid_targets_mean": 2641.6, + "valid_targets_min": 816 + }, + { + "epoch": 3.8002910149145146, + "grad_norm": 0.8985340486060293, + "learning_rate": 2.0507706909180395e-05, + "loss": 0.1402, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14424096047878265, + "step": 5225, + "valid_targets_mean": 2236.4, + "valid_targets_min": 949 + }, + { + "epoch": 3.803928701345944, + "grad_norm": 0.4079281492334926, + "learning_rate": 2.0471449106891644e-05, + "loss": 0.1052, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.048303425312042236, + "step": 5230, + "valid_targets_mean": 3270.9, + "valid_targets_min": 2201 + }, + { + "epoch": 3.8075663877773733, + "grad_norm": 0.7128532551186715, + "learning_rate": 2.043518975422312e-05, + "loss": 0.1181, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08494395017623901, + "step": 5235, + "valid_targets_mean": 1500.5, + "valid_targets_min": 687 + }, + { + "epoch": 3.811204074208803, + "grad_norm": 0.4651292875581144, + "learning_rate": 2.03989289704152e-05, + "loss": 0.0845, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0380450114607811, + "step": 5240, + "valid_targets_mean": 2285.5, + "valid_targets_min": 449 + }, + { + "epoch": 3.814841760640233, + "grad_norm": 0.382040676550468, + "learning_rate": 2.036266687471295e-05, + "loss": 0.0937, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.027340151369571686, + "step": 5245, + "valid_targets_mean": 2368.1, + "valid_targets_min": 658 + }, + { + "epoch": 3.8184794470716623, + "grad_norm": 0.6213954607139069, + "learning_rate": 2.0326403586365782e-05, + "loss": 0.0985, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.037498973309993744, + "step": 5250, + "valid_targets_mean": 831.8, + "valid_targets_min": 516 + }, + { + "epoch": 3.822117133503092, + "grad_norm": 0.35863223806599664, + "learning_rate": 2.029013922462699e-05, + "loss": 0.0779, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.054034292697906494, + "step": 5255, + "valid_targets_mean": 4461.9, + "valid_targets_min": 2805 + }, + { + "epoch": 3.8257548199345215, + "grad_norm": 0.36368818073774195, + "learning_rate": 2.025387390875343e-05, + "loss": 0.0743, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.044307418167591095, + "step": 5260, + "valid_targets_mean": 3766.8, + "valid_targets_min": 2036 + }, + { + "epoch": 3.8293925063659513, + "grad_norm": 0.3896870960034916, + "learning_rate": 2.021760775800509e-05, + "loss": 0.0868, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03668825328350067, + "step": 5265, + "valid_targets_mean": 2336.5, + "valid_targets_min": 681 + }, + { + "epoch": 3.8330301927973807, + "grad_norm": 0.3791507897162654, + "learning_rate": 2.0181340891644677e-05, + "loss": 0.1014, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04615131765604019, + "step": 5270, + "valid_targets_mean": 3856.1, + "valid_targets_min": 3436 + }, + { + "epoch": 3.8366678792288105, + "grad_norm": 0.5109407850844945, + "learning_rate": 2.0145073428937296e-05, + "loss": 0.1008, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06341229379177094, + "step": 5275, + "valid_targets_mean": 3165.5, + "valid_targets_min": 891 + }, + { + "epoch": 3.8403055656602403, + "grad_norm": 0.2932998162364287, + "learning_rate": 2.0108805489149977e-05, + "loss": 0.088, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030035603791475296, + "step": 5280, + "valid_targets_mean": 3550.6, + "valid_targets_min": 2573 + }, + { + "epoch": 3.8439432520916696, + "grad_norm": 0.33873487911549766, + "learning_rate": 2.0072537191551333e-05, + "loss": 0.0826, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03803408890962601, + "step": 5285, + "valid_targets_mean": 2995.0, + "valid_targets_min": 928 + }, + { + "epoch": 3.8475809385230995, + "grad_norm": 0.4532593357268297, + "learning_rate": 2.0036268655411146e-05, + "loss": 0.1034, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05615653097629547, + "step": 5290, + "valid_targets_mean": 3839.6, + "valid_targets_min": 3392 + }, + { + "epoch": 3.851218624954529, + "grad_norm": 0.30344600944678146, + "learning_rate": 2e-05, + "loss": 0.0749, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.041475169360637665, + "step": 5295, + "valid_targets_mean": 3855.0, + "valid_targets_min": 2848 + }, + { + "epoch": 3.8548563113859586, + "grad_norm": 0.4864739576085265, + "learning_rate": 1.9963731344588857e-05, + "loss": 0.0924, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04720839112997055, + "step": 5300, + "valid_targets_mean": 2586.0, + "valid_targets_min": 595 + }, + { + "epoch": 3.858493997817388, + "grad_norm": 0.4795572888167819, + "learning_rate": 1.9927462808448673e-05, + "loss": 0.0919, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04986552149057388, + "step": 5305, + "valid_targets_mean": 2471.8, + "valid_targets_min": 1016 + }, + { + "epoch": 3.862131684248818, + "grad_norm": 0.5366563825001275, + "learning_rate": 1.9891194510850033e-05, + "loss": 0.0975, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07327143847942352, + "step": 5310, + "valid_targets_mean": 1567.8, + "valid_targets_min": 811 + }, + { + "epoch": 3.8657693706802476, + "grad_norm": 0.6658148484340577, + "learning_rate": 1.9854926571062704e-05, + "loss": 0.101, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05824115127325058, + "step": 5315, + "valid_targets_mean": 1569.1, + "valid_targets_min": 697 + }, + { + "epoch": 3.869407057111677, + "grad_norm": 0.6695107834552159, + "learning_rate": 1.9818659108355326e-05, + "loss": 0.0881, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06518153101205826, + "step": 5320, + "valid_targets_mean": 1448.1, + "valid_targets_min": 736 + }, + { + "epoch": 3.8730447435431063, + "grad_norm": 0.4403884360408177, + "learning_rate": 1.9782392241994918e-05, + "loss": 0.0932, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030922863632440567, + "step": 5325, + "valid_targets_mean": 2415.5, + "valid_targets_min": 804 + }, + { + "epoch": 3.876682429974536, + "grad_norm": 0.5838578153862497, + "learning_rate": 1.9746126091246575e-05, + "loss": 0.1078, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03866460919380188, + "step": 5330, + "valid_targets_mean": 3472.5, + "valid_targets_min": 1018 + }, + { + "epoch": 3.880320116405966, + "grad_norm": 0.393542707472147, + "learning_rate": 1.9709860775373014e-05, + "loss": 0.0909, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.043517060577869415, + "step": 5335, + "valid_targets_mean": 3623.8, + "valid_targets_min": 2369 + }, + { + "epoch": 3.8839578028373953, + "grad_norm": 0.44512175296239614, + "learning_rate": 1.9673596413634225e-05, + "loss": 0.0795, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04587426781654358, + "step": 5340, + "valid_targets_mean": 3135.8, + "valid_targets_min": 1318 + }, + { + "epoch": 3.887595489268825, + "grad_norm": 0.4892417657571094, + "learning_rate": 1.9637333125287055e-05, + "loss": 0.1012, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06809043139219284, + "step": 5345, + "valid_targets_mean": 2258.2, + "valid_targets_min": 903 + }, + { + "epoch": 3.8912331757002545, + "grad_norm": 0.41680047895438094, + "learning_rate": 1.9601071029584803e-05, + "loss": 0.0967, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.043480031192302704, + "step": 5350, + "valid_targets_mean": 3460.4, + "valid_targets_min": 2346 + }, + { + "epoch": 3.8948708621316843, + "grad_norm": 0.6141448324773887, + "learning_rate": 1.9564810245776883e-05, + "loss": 0.08, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.073341503739357, + "step": 5355, + "valid_targets_mean": 2283.0, + "valid_targets_min": 737 + }, + { + "epoch": 3.8985085485631137, + "grad_norm": 0.37853518374315553, + "learning_rate": 1.9528550893108363e-05, + "loss": 0.0853, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0424114428460598, + "step": 5360, + "valid_targets_mean": 3979.6, + "valid_targets_min": 2555 + }, + { + "epoch": 3.9021462349945435, + "grad_norm": 0.28996964404267167, + "learning_rate": 1.949229309081961e-05, + "loss": 0.0821, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03688735142350197, + "step": 5365, + "valid_targets_mean": 4500.2, + "valid_targets_min": 2991 + }, + { + "epoch": 3.9057839214259733, + "grad_norm": 0.37966889622679556, + "learning_rate": 1.9456036958145906e-05, + "loss": 0.071, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029796846210956573, + "step": 5370, + "valid_targets_mean": 1992.4, + "valid_targets_min": 618 + }, + { + "epoch": 3.9094216078574027, + "grad_norm": 0.318339072489611, + "learning_rate": 1.941978261431703e-05, + "loss": 0.0944, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029032401740550995, + "step": 5375, + "valid_targets_mean": 3641.8, + "valid_targets_min": 2944 + }, + { + "epoch": 3.913059294288832, + "grad_norm": 0.33628611298952976, + "learning_rate": 1.9383530178556887e-05, + "loss": 0.0822, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0507139191031456, + "step": 5380, + "valid_targets_mean": 3686.1, + "valid_targets_min": 2702 + }, + { + "epoch": 3.916696980720262, + "grad_norm": 0.3447086057270459, + "learning_rate": 1.934727977008309e-05, + "loss": 0.1105, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.024055249989032745, + "step": 5385, + "valid_targets_mean": 3341.6, + "valid_targets_min": 509 + }, + { + "epoch": 3.9203346671516917, + "grad_norm": 0.3230815823542592, + "learning_rate": 1.9311031508106622e-05, + "loss": 0.1011, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03168732300400734, + "step": 5390, + "valid_targets_mean": 2757.6, + "valid_targets_min": 637 + }, + { + "epoch": 3.923972353583121, + "grad_norm": 0.4505791982994199, + "learning_rate": 1.927478551183136e-05, + "loss": 0.1101, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.049235112965106964, + "step": 5395, + "valid_targets_mean": 2802.6, + "valid_targets_min": 737 + }, + { + "epoch": 3.927610040014551, + "grad_norm": 0.32252083200005094, + "learning_rate": 1.9238541900453758e-05, + "loss": 0.0774, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03623335063457489, + "step": 5400, + "valid_targets_mean": 3312.9, + "valid_targets_min": 1292 + }, + { + "epoch": 3.93124772644598, + "grad_norm": 0.578337397921177, + "learning_rate": 1.920230079316243e-05, + "loss": 0.0962, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0475928857922554, + "step": 5405, + "valid_targets_mean": 1049.6, + "valid_targets_min": 697 + }, + { + "epoch": 3.93488541287741, + "grad_norm": 0.3893945041587812, + "learning_rate": 1.9166062309137737e-05, + "loss": 0.0825, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03989843651652336, + "step": 5410, + "valid_targets_mean": 3541.4, + "valid_targets_min": 923 + }, + { + "epoch": 3.9385230993088394, + "grad_norm": 0.7888509513237429, + "learning_rate": 1.912982656755144e-05, + "loss": 0.123, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06769990175962448, + "step": 5415, + "valid_targets_mean": 931.6, + "valid_targets_min": 531 + }, + { + "epoch": 3.942160785740269, + "grad_norm": 0.5279774365210688, + "learning_rate": 1.9093593687566246e-05, + "loss": 0.1173, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0630648285150528, + "step": 5420, + "valid_targets_mean": 1840.5, + "valid_targets_min": 844 + }, + { + "epoch": 3.945798472171699, + "grad_norm": 0.5042407723743552, + "learning_rate": 1.905736378833548e-05, + "loss": 0.0891, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.059978216886520386, + "step": 5425, + "valid_targets_mean": 3893.9, + "valid_targets_min": 2967 + }, + { + "epoch": 3.9494361586031284, + "grad_norm": 0.3699094562016425, + "learning_rate": 1.9021136989002654e-05, + "loss": 0.0866, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047110844403505325, + "step": 5430, + "valid_targets_mean": 3605.9, + "valid_targets_min": 2490 + }, + { + "epoch": 3.953073845034558, + "grad_norm": 0.37425835956892467, + "learning_rate": 1.8984913408701085e-05, + "loss": 0.0867, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.037735193967819214, + "step": 5435, + "valid_targets_mean": 2682.0, + "valid_targets_min": 970 + }, + { + "epoch": 3.9567115314659875, + "grad_norm": 0.46352905772471403, + "learning_rate": 1.8948693166553516e-05, + "loss": 0.0772, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.034436680376529694, + "step": 5440, + "valid_targets_mean": 2028.5, + "valid_targets_min": 729 + }, + { + "epoch": 3.9603492178974173, + "grad_norm": 0.8378724739189977, + "learning_rate": 1.8912476381671687e-05, + "loss": 0.1032, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07180199772119522, + "step": 5445, + "valid_targets_mean": 1028.9, + "valid_targets_min": 637 + }, + { + "epoch": 3.9639869043288467, + "grad_norm": 0.35045887589590013, + "learning_rate": 1.8876263173156004e-05, + "loss": 0.0806, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029849736019968987, + "step": 5450, + "valid_targets_mean": 2879.2, + "valid_targets_min": 701 + }, + { + "epoch": 3.9676245907602765, + "grad_norm": 0.3513641980086048, + "learning_rate": 1.8840053660095074e-05, + "loss": 0.0851, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03564560413360596, + "step": 5455, + "valid_targets_mean": 3671.1, + "valid_targets_min": 3131 + }, + { + "epoch": 3.9712622771917063, + "grad_norm": 0.43383595188079344, + "learning_rate": 1.8803847961565374e-05, + "loss": 0.0846, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.044742412865161896, + "step": 5460, + "valid_targets_mean": 3241.5, + "valid_targets_min": 2018 + }, + { + "epoch": 3.9748999636231357, + "grad_norm": 0.33161179689377634, + "learning_rate": 1.8767646196630837e-05, + "loss": 0.0939, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03894129395484924, + "step": 5465, + "valid_targets_mean": 2491.6, + "valid_targets_min": 350 + }, + { + "epoch": 3.978537650054565, + "grad_norm": 0.29488139759114407, + "learning_rate": 1.8731448484342448e-05, + "loss": 0.076, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.041373360902071, + "step": 5470, + "valid_targets_mean": 4862.8, + "valid_targets_min": 3577 + }, + { + "epoch": 3.982175336485995, + "grad_norm": 0.35295834518042046, + "learning_rate": 1.8695254943737887e-05, + "loss": 0.0835, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.033555276691913605, + "step": 5475, + "valid_targets_mean": 1999.4, + "valid_targets_min": 507 + }, + { + "epoch": 3.9858130229174247, + "grad_norm": 0.45099214247115943, + "learning_rate": 1.865906569384109e-05, + "loss": 0.0975, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08814767748117447, + "step": 5480, + "valid_targets_mean": 3278.5, + "valid_targets_min": 2814 + }, + { + "epoch": 3.989450709348854, + "grad_norm": 0.39342214054228625, + "learning_rate": 1.8622880853661888e-05, + "loss": 0.1638, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04658334702253342, + "step": 5485, + "valid_targets_mean": 4006.2, + "valid_targets_min": 2347 + }, + { + "epoch": 3.993088395780284, + "grad_norm": 0.4970052217985949, + "learning_rate": 1.858670054219562e-05, + "loss": 0.0867, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05141102522611618, + "step": 5490, + "valid_targets_mean": 2447.4, + "valid_targets_min": 716 + }, + { + "epoch": 3.996726082211713, + "grad_norm": 0.376136954974494, + "learning_rate": 1.855052487842273e-05, + "loss": 0.0812, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03884429484605789, + "step": 5495, + "valid_targets_mean": 3209.6, + "valid_targets_min": 2245 + }, + { + "epoch": 4.0, + "grad_norm": 0.5906815603466118, + "learning_rate": 1.8514353981308383e-05, + "loss": 0.1238, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1374020278453827, + "step": 5500, + "valid_targets_mean": 7922.2, + "valid_targets_min": 3513 + }, + { + "epoch": 4.003637686431429, + "grad_norm": 0.49445508212527367, + "learning_rate": 1.8478187969802045e-05, + "loss": 0.1904, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09235258400440216, + "step": 5505, + "valid_targets_mean": 7389.0, + "valid_targets_min": 5627 + }, + { + "epoch": 4.00727537286286, + "grad_norm": 0.6346689305428214, + "learning_rate": 1.8442026962837145e-05, + "loss": 0.1918, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05265560746192932, + "step": 5510, + "valid_targets_mean": 1294.8, + "valid_targets_min": 422 + }, + { + "epoch": 4.010913059294289, + "grad_norm": 0.37257902130835463, + "learning_rate": 1.8405871079330643e-05, + "loss": 0.1831, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09492306411266327, + "step": 5515, + "valid_targets_mean": 8539.9, + "valid_targets_min": 6761 + }, + { + "epoch": 4.014550745725718, + "grad_norm": 0.4008139116296186, + "learning_rate": 1.836972043818263e-05, + "loss": 0.1776, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08305317163467407, + "step": 5520, + "valid_targets_mean": 7579.0, + "valid_targets_min": 6174 + }, + { + "epoch": 4.018188432157148, + "grad_norm": 0.3899896588004054, + "learning_rate": 1.8333575158275996e-05, + "loss": 0.1803, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08173725008964539, + "step": 5525, + "valid_targets_mean": 7328.2, + "valid_targets_min": 4839 + }, + { + "epoch": 4.021826118588578, + "grad_norm": 0.3988174897481221, + "learning_rate": 1.8297435358475966e-05, + "loss": 0.1885, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0979163721203804, + "step": 5530, + "valid_targets_mean": 6838.2, + "valid_targets_min": 5034 + }, + { + "epoch": 4.025463805020007, + "grad_norm": 0.4173346530177005, + "learning_rate": 1.826130115762977e-05, + "loss": 0.1922, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09883152693510056, + "step": 5535, + "valid_targets_mean": 7248.2, + "valid_targets_min": 5567 + }, + { + "epoch": 4.029101491451437, + "grad_norm": 0.3866058982306686, + "learning_rate": 1.8225172674566204e-05, + "loss": 0.1778, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08141618221998215, + "step": 5540, + "valid_targets_mean": 6234.0, + "valid_targets_min": 5603 + }, + { + "epoch": 4.032739177882866, + "grad_norm": 0.6505615587003735, + "learning_rate": 1.818905002809527e-05, + "loss": 0.1733, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02584587223827839, + "step": 5545, + "valid_targets_mean": 602.0, + "valid_targets_min": 152 + }, + { + "epoch": 4.036376864314296, + "grad_norm": 0.415312540775909, + "learning_rate": 1.8152933337007793e-05, + "loss": 0.1659, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0843053087592125, + "step": 5550, + "valid_targets_mean": 7299.9, + "valid_targets_min": 6158 + }, + { + "epoch": 4.040014550745726, + "grad_norm": 0.3932583555163608, + "learning_rate": 1.8116822720074966e-05, + "loss": 0.1806, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08412586152553558, + "step": 5555, + "valid_targets_mean": 6140.0, + "valid_targets_min": 4648 + }, + { + "epoch": 4.043652237177155, + "grad_norm": 0.39760805174567215, + "learning_rate": 1.8080718296048068e-05, + "loss": 0.1738, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0895063728094101, + "step": 5560, + "valid_targets_mean": 6603.0, + "valid_targets_min": 4882 + }, + { + "epoch": 4.047289923608585, + "grad_norm": 0.3672113878857198, + "learning_rate": 1.804462018365796e-05, + "loss": 0.1746, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08881582319736481, + "step": 5565, + "valid_targets_mean": 8015.0, + "valid_targets_min": 6344 + }, + { + "epoch": 4.050927610040015, + "grad_norm": 0.3824107431622493, + "learning_rate": 1.800852850161478e-05, + "loss": 0.1669, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07899197936058044, + "step": 5570, + "valid_targets_mean": 6305.2, + "valid_targets_min": 4622 + }, + { + "epoch": 4.054565296471444, + "grad_norm": 0.3784084365275729, + "learning_rate": 1.7972443368607507e-05, + "loss": 0.1748, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09189162403345108, + "step": 5575, + "valid_targets_mean": 7890.8, + "valid_targets_min": 5485 + }, + { + "epoch": 4.058202982902873, + "grad_norm": 0.40100201120680257, + "learning_rate": 1.7936364903303577e-05, + "loss": 0.1774, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08325114846229553, + "step": 5580, + "valid_targets_mean": 7016.6, + "valid_targets_min": 4960 + }, + { + "epoch": 4.061840669334304, + "grad_norm": 0.4102033399807827, + "learning_rate": 1.7900293224348524e-05, + "loss": 0.1725, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08339834213256836, + "step": 5585, + "valid_targets_mean": 6527.1, + "valid_targets_min": 5399 + }, + { + "epoch": 4.065478355765733, + "grad_norm": 0.44965510365102324, + "learning_rate": 1.7864228450365525e-05, + "loss": 0.1769, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08455341309309006, + "step": 5590, + "valid_targets_mean": 4995.0, + "valid_targets_min": 618 + }, + { + "epoch": 4.069116042197162, + "grad_norm": 0.40476005581035607, + "learning_rate": 1.7828170699955093e-05, + "loss": 0.1896, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09496694803237915, + "step": 5595, + "valid_targets_mean": 7219.4, + "valid_targets_min": 5764 + }, + { + "epoch": 4.072753728628593, + "grad_norm": 0.3895714326206731, + "learning_rate": 1.779212009169461e-05, + "loss": 0.1581, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08789315819740295, + "step": 5600, + "valid_targets_mean": 7998.5, + "valid_targets_min": 5902 + }, + { + "epoch": 4.076391415060022, + "grad_norm": 0.38281650094453396, + "learning_rate": 1.7756076744137982e-05, + "loss": 0.1631, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06998445838689804, + "step": 5605, + "valid_targets_mean": 7377.4, + "valid_targets_min": 5015 + }, + { + "epoch": 4.080029101491451, + "grad_norm": 0.37482944099405163, + "learning_rate": 1.772004077581525e-05, + "loss": 0.1718, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07221399247646332, + "step": 5610, + "valid_targets_mean": 6794.0, + "valid_targets_min": 5283 + }, + { + "epoch": 4.083666787922881, + "grad_norm": 0.4134767011267427, + "learning_rate": 1.7684012305232166e-05, + "loss": 0.1782, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09920880198478699, + "step": 5615, + "valid_targets_mean": 6656.4, + "valid_targets_min": 4364 + }, + { + "epoch": 4.087304474354311, + "grad_norm": 0.385429289255785, + "learning_rate": 1.764799145086985e-05, + "loss": 0.1755, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08068601787090302, + "step": 5620, + "valid_targets_mean": 7178.4, + "valid_targets_min": 4448 + }, + { + "epoch": 4.09094216078574, + "grad_norm": 0.4281249684775914, + "learning_rate": 1.7611978331184346e-05, + "loss": 0.1745, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10421295464038849, + "step": 5625, + "valid_targets_mean": 7619.8, + "valid_targets_min": 5737 + }, + { + "epoch": 4.09457984721717, + "grad_norm": 0.35313053421316165, + "learning_rate": 1.7575973064606298e-05, + "loss": 0.1543, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07369157671928406, + "step": 5630, + "valid_targets_mean": 8754.4, + "valid_targets_min": 5752 + }, + { + "epoch": 4.098217533648599, + "grad_norm": 0.4642895060076819, + "learning_rate": 1.753997576954049e-05, + "loss": 0.1551, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08086036890745163, + "step": 5635, + "valid_targets_mean": 6341.0, + "valid_targets_min": 4543 + }, + { + "epoch": 4.101855220080029, + "grad_norm": 0.37633921091773515, + "learning_rate": 1.7503986564365512e-05, + "loss": 0.1505, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08632586896419525, + "step": 5640, + "valid_targets_mean": 7181.9, + "valid_targets_min": 4563 + }, + { + "epoch": 4.105492906511459, + "grad_norm": 0.7061647683285396, + "learning_rate": 1.7468005567433348e-05, + "loss": 0.1795, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05730462819337845, + "step": 5645, + "valid_targets_mean": 1104.4, + "valid_targets_min": 161 + }, + { + "epoch": 4.109130592942888, + "grad_norm": 0.4030203146804881, + "learning_rate": 1.743203289706898e-05, + "loss": 0.1632, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07856880128383636, + "step": 5650, + "valid_targets_mean": 7018.2, + "valid_targets_min": 5424 + }, + { + "epoch": 4.112768279374318, + "grad_norm": 0.40985215129212493, + "learning_rate": 1.7396068671570025e-05, + "loss": 0.1816, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09254579246044159, + "step": 5655, + "valid_targets_mean": 7738.4, + "valid_targets_min": 5329 + }, + { + "epoch": 4.116405965805748, + "grad_norm": 0.4302121542909593, + "learning_rate": 1.7360113009206305e-05, + "loss": 0.1852, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08936718106269836, + "step": 5660, + "valid_targets_mean": 6426.6, + "valid_targets_min": 5560 + }, + { + "epoch": 4.120043652237177, + "grad_norm": 0.42704379424379296, + "learning_rate": 1.7324166028219495e-05, + "loss": 0.1797, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08862088620662689, + "step": 5665, + "valid_targets_mean": 6169.9, + "valid_targets_min": 3446 + }, + { + "epoch": 4.123681338668606, + "grad_norm": 0.406656361225356, + "learning_rate": 1.7288227846822726e-05, + "loss": 0.1753, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09380245953798294, + "step": 5670, + "valid_targets_mean": 7204.6, + "valid_targets_min": 5381 + }, + { + "epoch": 4.127319025100037, + "grad_norm": 0.3756523864111663, + "learning_rate": 1.725229858320018e-05, + "loss": 0.1796, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0919438824057579, + "step": 5675, + "valid_targets_mean": 7600.4, + "valid_targets_min": 5241 + }, + { + "epoch": 4.130956711531466, + "grad_norm": 0.3964678723608219, + "learning_rate": 1.721637835550672e-05, + "loss": 0.1748, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09620088338851929, + "step": 5680, + "valid_targets_mean": 7098.5, + "valid_targets_min": 5607 + }, + { + "epoch": 4.134594397962895, + "grad_norm": 0.4066413913978783, + "learning_rate": 1.7180467281867495e-05, + "loss": 0.1783, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.094047412276268, + "step": 5685, + "valid_targets_mean": 6955.1, + "valid_targets_min": 5530 + }, + { + "epoch": 4.138232084394325, + "grad_norm": 0.4251132134801971, + "learning_rate": 1.7144565480377547e-05, + "loss": 0.1789, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09156718850135803, + "step": 5690, + "valid_targets_mean": 6748.1, + "valid_targets_min": 5364 + }, + { + "epoch": 4.141869770825755, + "grad_norm": 0.4050421516614434, + "learning_rate": 1.7108673069101423e-05, + "loss": 0.1812, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0859956294298172, + "step": 5695, + "valid_targets_mean": 6273.9, + "valid_targets_min": 5438 + }, + { + "epoch": 4.145507457257184, + "grad_norm": 0.3896128059619484, + "learning_rate": 1.7072790166072786e-05, + "loss": 0.1744, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08335354179143906, + "step": 5700, + "valid_targets_mean": 6419.5, + "valid_targets_min": 4879 + }, + { + "epoch": 4.149145143688614, + "grad_norm": 0.4045756043392928, + "learning_rate": 1.7036916889294053e-05, + "loss": 0.1732, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08373061567544937, + "step": 5705, + "valid_targets_mean": 6483.4, + "valid_targets_min": 4309 + }, + { + "epoch": 4.152782830120044, + "grad_norm": 0.47055325948575544, + "learning_rate": 1.7001053356735956e-05, + "loss": 0.1488, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1029350608587265, + "step": 5710, + "valid_targets_mean": 6825.0, + "valid_targets_min": 5205 + }, + { + "epoch": 4.156420516551473, + "grad_norm": 0.40858188285786645, + "learning_rate": 1.696519968633721e-05, + "loss": 0.1738, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0811498761177063, + "step": 5715, + "valid_targets_mean": 6729.4, + "valid_targets_min": 5200 + }, + { + "epoch": 4.160058202982903, + "grad_norm": 0.39546473750426536, + "learning_rate": 1.692935599600408e-05, + "loss": 0.1717, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08335229754447937, + "step": 5720, + "valid_targets_mean": 6094.0, + "valid_targets_min": 5302 + }, + { + "epoch": 4.163695889414332, + "grad_norm": 0.42129593230542384, + "learning_rate": 1.6893522403610004e-05, + "loss": 0.176, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0920780822634697, + "step": 5725, + "valid_targets_mean": 7140.0, + "valid_targets_min": 5270 + }, + { + "epoch": 4.167333575845762, + "grad_norm": 0.37434430037813055, + "learning_rate": 1.6857699026995235e-05, + "loss": 0.1697, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0819597989320755, + "step": 5730, + "valid_targets_mean": 7425.0, + "valid_targets_min": 5281 + }, + { + "epoch": 4.170971262277192, + "grad_norm": 0.4150130134654448, + "learning_rate": 1.6821885983966416e-05, + "loss": 0.1774, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11129307001829147, + "step": 5735, + "valid_targets_mean": 7063.4, + "valid_targets_min": 5386 + }, + { + "epoch": 4.174608948708621, + "grad_norm": 0.4006364099903819, + "learning_rate": 1.6786083392296217e-05, + "loss": 0.1756, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09026254713535309, + "step": 5740, + "valid_targets_mean": 6827.0, + "valid_targets_min": 4547 + }, + { + "epoch": 4.178246635140051, + "grad_norm": 0.5373430249461408, + "learning_rate": 1.6750291369722922e-05, + "loss": 0.154, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04350530356168747, + "step": 5745, + "valid_targets_mean": 2047.0, + "valid_targets_min": 165 + }, + { + "epoch": 4.181884321571481, + "grad_norm": 0.4412460017159865, + "learning_rate": 1.671451003395008e-05, + "loss": 0.1597, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08336615562438965, + "step": 5750, + "valid_targets_mean": 6754.1, + "valid_targets_min": 5196 + }, + { + "epoch": 4.18552200800291, + "grad_norm": 0.43589755225510995, + "learning_rate": 1.6678739502646092e-05, + "loss": 0.1709, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08603298664093018, + "step": 5755, + "valid_targets_mean": 5985.6, + "valid_targets_min": 5028 + }, + { + "epoch": 4.189159694434339, + "grad_norm": 0.41447422870259265, + "learning_rate": 1.6642979893443796e-05, + "loss": 0.1761, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0802798643708229, + "step": 5760, + "valid_targets_mean": 5963.9, + "valid_targets_min": 5237 + }, + { + "epoch": 4.19279738086577, + "grad_norm": 0.4577053389348319, + "learning_rate": 1.6607231323940155e-05, + "loss": 0.171, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09022487699985504, + "step": 5765, + "valid_targets_mean": 6319.0, + "valid_targets_min": 5412 + }, + { + "epoch": 4.196435067297199, + "grad_norm": 0.39935381001434445, + "learning_rate": 1.65714939116958e-05, + "loss": 0.1625, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07132437825202942, + "step": 5770, + "valid_targets_mean": 7125.8, + "valid_targets_min": 5080 + }, + { + "epoch": 4.200072753728628, + "grad_norm": 0.42207391878493233, + "learning_rate": 1.6535767774234695e-05, + "loss": 0.175, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08608950674533844, + "step": 5775, + "valid_targets_mean": 6123.5, + "valid_targets_min": 4521 + }, + { + "epoch": 4.203710440160058, + "grad_norm": 0.6090058658839225, + "learning_rate": 1.6500053029043702e-05, + "loss": 0.2076, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1338992416858673, + "step": 5780, + "valid_targets_mean": 6020.6, + "valid_targets_min": 2493 + }, + { + "epoch": 4.207348126591488, + "grad_norm": 0.5766909121939671, + "learning_rate": 1.646434979357222e-05, + "loss": 0.2338, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09264469146728516, + "step": 5785, + "valid_targets_mean": 3953.5, + "valid_targets_min": 1177 + }, + { + "epoch": 4.210985813022917, + "grad_norm": 0.5691692504462874, + "learning_rate": 1.642865818523183e-05, + "loss": 0.2413, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14126187562942505, + "step": 5790, + "valid_targets_mean": 4950.8, + "valid_targets_min": 2799 + }, + { + "epoch": 4.214623499454347, + "grad_norm": 0.5699665019723842, + "learning_rate": 1.639297832139583e-05, + "loss": 0.2274, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12575829029083252, + "step": 5795, + "valid_targets_mean": 5312.0, + "valid_targets_min": 2973 + }, + { + "epoch": 4.218261185885777, + "grad_norm": 0.575539999297184, + "learning_rate": 1.6357310319398937e-05, + "loss": 0.2266, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10618853569030762, + "step": 5800, + "valid_targets_mean": 3775.5, + "valid_targets_min": 980 + }, + { + "epoch": 4.221898872317206, + "grad_norm": 0.5435178801012488, + "learning_rate": 1.632165429653683e-05, + "loss": 0.245, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1255960613489151, + "step": 5805, + "valid_targets_mean": 5055.4, + "valid_targets_min": 1512 + }, + { + "epoch": 4.225536558748636, + "grad_norm": 0.5108065371346245, + "learning_rate": 1.628601037006582e-05, + "loss": 0.2198, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08539222925901413, + "step": 5810, + "valid_targets_mean": 3600.8, + "valid_targets_min": 1848 + }, + { + "epoch": 4.229174245180065, + "grad_norm": 0.4920916922801779, + "learning_rate": 1.6250378657202414e-05, + "loss": 0.2292, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08347669243812561, + "step": 5815, + "valid_targets_mean": 3369.8, + "valid_targets_min": 1381 + }, + { + "epoch": 4.232811931611495, + "grad_norm": 0.492087004118344, + "learning_rate": 1.6214759275122953e-05, + "loss": 0.2279, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1235499307513237, + "step": 5820, + "valid_targets_mean": 6283.2, + "valid_targets_min": 3600 + }, + { + "epoch": 4.236449618042925, + "grad_norm": 0.5374884997504846, + "learning_rate": 1.6179152340963268e-05, + "loss": 0.2404, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11848868429660797, + "step": 5825, + "valid_targets_mean": 5242.5, + "valid_targets_min": 2540 + }, + { + "epoch": 4.240087304474354, + "grad_norm": 0.4860785994578068, + "learning_rate": 1.6143557971818205e-05, + "loss": 0.2256, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10380423814058304, + "step": 5830, + "valid_targets_mean": 6246.5, + "valid_targets_min": 2157 + }, + { + "epoch": 4.243724990905784, + "grad_norm": 0.5565032599882889, + "learning_rate": 1.610797628474132e-05, + "loss": 0.2286, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11624015867710114, + "step": 5835, + "valid_targets_mean": 4556.5, + "valid_targets_min": 1720 + }, + { + "epoch": 4.247362677337214, + "grad_norm": 0.5841852777840628, + "learning_rate": 1.6072407396744447e-05, + "loss": 0.2303, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12602585554122925, + "step": 5840, + "valid_targets_mean": 5116.5, + "valid_targets_min": 1982 + }, + { + "epoch": 4.251000363768643, + "grad_norm": 0.5311784241072374, + "learning_rate": 1.6036851424797338e-05, + "loss": 0.2285, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11816704273223877, + "step": 5845, + "valid_targets_mean": 4464.5, + "valid_targets_min": 2718 + }, + { + "epoch": 4.2546380502000725, + "grad_norm": 0.6186200221224596, + "learning_rate": 1.6001308485827284e-05, + "loss": 0.2268, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11258639395236969, + "step": 5850, + "valid_targets_mean": 3418.5, + "valid_targets_min": 1451 + }, + { + "epoch": 4.258275736631503, + "grad_norm": 0.612227534066663, + "learning_rate": 1.5965778696718672e-05, + "loss": 0.2251, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11483729630708694, + "step": 5855, + "valid_targets_mean": 3990.4, + "valid_targets_min": 1861 + }, + { + "epoch": 4.261913423062932, + "grad_norm": 0.6270514291301228, + "learning_rate": 1.5930262174312704e-05, + "loss": 0.2213, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12583619356155396, + "step": 5860, + "valid_targets_mean": 4065.4, + "valid_targets_min": 2100 + }, + { + "epoch": 4.2655511094943614, + "grad_norm": 0.5935707512209718, + "learning_rate": 1.5894759035406902e-05, + "loss": 0.228, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12704241275787354, + "step": 5865, + "valid_targets_mean": 4132.4, + "valid_targets_min": 3234 + }, + { + "epoch": 4.269188795925791, + "grad_norm": 0.6376018289019716, + "learning_rate": 1.5859269396754814e-05, + "loss": 0.2353, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14031721651554108, + "step": 5870, + "valid_targets_mean": 4246.2, + "valid_targets_min": 2650 + }, + { + "epoch": 4.272826482357221, + "grad_norm": 0.597970500956161, + "learning_rate": 1.5823793375065566e-05, + "loss": 0.2276, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0952034741640091, + "step": 5875, + "valid_targets_mean": 4096.5, + "valid_targets_min": 1719 + }, + { + "epoch": 4.27646416878865, + "grad_norm": 0.5868384763077729, + "learning_rate": 1.5788331087003515e-05, + "loss": 0.234, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11258688569068909, + "step": 5880, + "valid_targets_mean": 3933.1, + "valid_targets_min": 1020 + }, + { + "epoch": 4.28010185522008, + "grad_norm": 0.614877562128478, + "learning_rate": 1.575288264918786e-05, + "loss": 0.213, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11548049747943878, + "step": 5885, + "valid_targets_mean": 3551.1, + "valid_targets_min": 2130 + }, + { + "epoch": 4.28373954165151, + "grad_norm": 0.6038260533261407, + "learning_rate": 1.5717448178192233e-05, + "loss": 0.2128, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08843228965997696, + "step": 5890, + "valid_targets_mean": 3535.0, + "valid_targets_min": 865 + }, + { + "epoch": 4.287377228082939, + "grad_norm": 0.7119301391974057, + "learning_rate": 1.568202779054437e-05, + "loss": 0.2166, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14763164520263672, + "step": 5895, + "valid_targets_mean": 3716.5, + "valid_targets_min": 3144 + }, + { + "epoch": 4.291014914514369, + "grad_norm": 0.6078189760255714, + "learning_rate": 1.5646621602725654e-05, + "loss": 0.226, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09747655689716339, + "step": 5900, + "valid_targets_mean": 3262.2, + "valid_targets_min": 2050 + }, + { + "epoch": 4.294652600945798, + "grad_norm": 0.6216836541448849, + "learning_rate": 1.5611229731170793e-05, + "loss": 0.2249, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11246366053819656, + "step": 5905, + "valid_targets_mean": 3296.5, + "valid_targets_min": 1570 + }, + { + "epoch": 4.298290287377228, + "grad_norm": 0.5633668098193757, + "learning_rate": 1.5575852292267414e-05, + "loss": 0.2143, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11098483204841614, + "step": 5910, + "valid_targets_mean": 4806.4, + "valid_targets_min": 1795 + }, + { + "epoch": 4.301927973808658, + "grad_norm": 0.5420468918659436, + "learning_rate": 1.5540489402355678e-05, + "loss": 0.2168, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10379049181938171, + "step": 5915, + "valid_targets_mean": 4342.4, + "valid_targets_min": 1934 + }, + { + "epoch": 4.305565660240087, + "grad_norm": 0.6358157935045365, + "learning_rate": 1.5505141177727913e-05, + "loss": 0.2202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1120682805776596, + "step": 5920, + "valid_targets_mean": 3675.4, + "valid_targets_min": 2557 + }, + { + "epoch": 4.3092033466715165, + "grad_norm": 0.6076788140107805, + "learning_rate": 1.5469807734628183e-05, + "loss": 0.2212, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1261780560016632, + "step": 5925, + "valid_targets_mean": 4550.4, + "valid_targets_min": 2663 + }, + { + "epoch": 4.312841033102947, + "grad_norm": 0.569080501902127, + "learning_rate": 1.5434489189252e-05, + "loss": 0.211, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10146931558847427, + "step": 5930, + "valid_targets_mean": 4604.0, + "valid_targets_min": 2016 + }, + { + "epoch": 4.316478719534376, + "grad_norm": 0.6502119917381123, + "learning_rate": 1.539918565774584e-05, + "loss": 0.2196, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09943878650665283, + "step": 5935, + "valid_targets_mean": 2976.8, + "valid_targets_min": 1367 + }, + { + "epoch": 4.3201164059658055, + "grad_norm": 0.6489520855798472, + "learning_rate": 1.5363897256206805e-05, + "loss": 0.2219, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10847680270671844, + "step": 5940, + "valid_targets_mean": 3542.5, + "valid_targets_min": 2026 + }, + { + "epoch": 4.323754092397236, + "grad_norm": 0.5863792223069811, + "learning_rate": 1.532862410068227e-05, + "loss": 0.2097, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.095295250415802, + "step": 5945, + "valid_targets_mean": 3247.2, + "valid_targets_min": 2033 + }, + { + "epoch": 4.327391778828665, + "grad_norm": 0.7405683313955196, + "learning_rate": 1.529336630716945e-05, + "loss": 0.2189, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10768456012010574, + "step": 5950, + "valid_targets_mean": 2201.6, + "valid_targets_min": 1283 + }, + { + "epoch": 4.3310294652600945, + "grad_norm": 0.7272899817560686, + "learning_rate": 1.525812399161506e-05, + "loss": 0.2193, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11754097044467926, + "step": 5955, + "valid_targets_mean": 2824.8, + "valid_targets_min": 1636 + }, + { + "epoch": 4.334667151691524, + "grad_norm": 0.6361706582121054, + "learning_rate": 1.5222897269914892e-05, + "loss": 0.2217, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11755922436714172, + "step": 5960, + "valid_targets_mean": 3844.4, + "valid_targets_min": 1629 + }, + { + "epoch": 4.338304838122954, + "grad_norm": 0.6482576567522217, + "learning_rate": 1.5187686257913466e-05, + "loss": 0.216, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1004336029291153, + "step": 5965, + "valid_targets_mean": 3048.6, + "valid_targets_min": 2080 + }, + { + "epoch": 4.3419425245543835, + "grad_norm": 0.6312057666402687, + "learning_rate": 1.515249107140365e-05, + "loss": 0.2164, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10184167325496674, + "step": 5970, + "valid_targets_mean": 3111.8, + "valid_targets_min": 1153 + }, + { + "epoch": 4.345580210985813, + "grad_norm": 0.6319547020100363, + "learning_rate": 1.511731182612626e-05, + "loss": 0.2183, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11787837743759155, + "step": 5975, + "valid_targets_mean": 4069.0, + "valid_targets_min": 1858 + }, + { + "epoch": 4.349217897417242, + "grad_norm": 0.5886133524093161, + "learning_rate": 1.5082148637769694e-05, + "loss": 0.2216, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09257040172815323, + "step": 5980, + "valid_targets_mean": 3415.2, + "valid_targets_min": 1639 + }, + { + "epoch": 4.352855583848672, + "grad_norm": 0.6353916996313892, + "learning_rate": 1.5047001621969537e-05, + "loss": 0.2159, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0877324715256691, + "step": 5985, + "valid_targets_mean": 3267.6, + "valid_targets_min": 1394 + }, + { + "epoch": 4.356493270280102, + "grad_norm": 0.5918219990476844, + "learning_rate": 1.5011870894308205e-05, + "loss": 0.2171, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11601825058460236, + "step": 5990, + "valid_targets_mean": 4775.9, + "valid_targets_min": 1844 + }, + { + "epoch": 4.360130956711531, + "grad_norm": 0.6016213790714964, + "learning_rate": 1.497675657031453e-05, + "loss": 0.2162, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09589147567749023, + "step": 5995, + "valid_targets_mean": 3317.1, + "valid_targets_min": 1305 + }, + { + "epoch": 4.363768643142961, + "grad_norm": 0.6292808923874497, + "learning_rate": 1.4941658765463412e-05, + "loss": 0.2167, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11942924559116364, + "step": 6000, + "valid_targets_mean": 4290.8, + "valid_targets_min": 2185 + }, + { + "epoch": 4.367406329574391, + "grad_norm": 0.6603853634562894, + "learning_rate": 1.4906577595175428e-05, + "loss": 0.2172, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13397620618343353, + "step": 6005, + "valid_targets_mean": 4024.9, + "valid_targets_min": 2349 + }, + { + "epoch": 4.37104401600582, + "grad_norm": 0.588431940819527, + "learning_rate": 1.4871513174816445e-05, + "loss": 0.2109, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09284106642007828, + "step": 6010, + "valid_targets_mean": 3453.0, + "valid_targets_min": 1607 + }, + { + "epoch": 4.3746817024372495, + "grad_norm": 0.6471888805402066, + "learning_rate": 1.483646561969726e-05, + "loss": 0.2253, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1170738935470581, + "step": 6015, + "valid_targets_mean": 3389.8, + "valid_targets_min": 1507 + }, + { + "epoch": 4.37831938886868, + "grad_norm": 0.6080374990960062, + "learning_rate": 1.4801435045073197e-05, + "loss": 0.2182, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09496614336967468, + "step": 6020, + "valid_targets_mean": 3773.9, + "valid_targets_min": 1317 + }, + { + "epoch": 4.381957075300109, + "grad_norm": 0.6141652867072032, + "learning_rate": 1.4766421566143726e-05, + "loss": 0.22, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10472684353590012, + "step": 6025, + "valid_targets_mean": 3881.0, + "valid_targets_min": 3355 + }, + { + "epoch": 4.3855947617315385, + "grad_norm": 0.6252643831269105, + "learning_rate": 1.4731425298052133e-05, + "loss": 0.2206, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10938552021980286, + "step": 6030, + "valid_targets_mean": 3951.9, + "valid_targets_min": 2999 + }, + { + "epoch": 4.389232448162969, + "grad_norm": 0.6082951671479347, + "learning_rate": 1.469644635588507e-05, + "loss": 0.2159, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0976148322224617, + "step": 6035, + "valid_targets_mean": 4086.5, + "valid_targets_min": 1739 + }, + { + "epoch": 4.392870134594398, + "grad_norm": 0.5889008263032748, + "learning_rate": 1.4661484854672234e-05, + "loss": 0.2155, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09392209351062775, + "step": 6040, + "valid_targets_mean": 3298.0, + "valid_targets_min": 1140 + }, + { + "epoch": 4.3965078210258275, + "grad_norm": 0.5965373766394706, + "learning_rate": 1.4626540909385954e-05, + "loss": 0.2033, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10743594169616699, + "step": 6045, + "valid_targets_mean": 4012.1, + "valid_targets_min": 2779 + }, + { + "epoch": 4.400145507457257, + "grad_norm": 0.645848620151031, + "learning_rate": 1.4591614634940836e-05, + "loss": 0.2142, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10016131401062012, + "step": 6050, + "valid_targets_mean": 3636.9, + "valid_targets_min": 1183 + }, + { + "epoch": 4.403783193888687, + "grad_norm": 0.614860226144204, + "learning_rate": 1.4556706146193374e-05, + "loss": 0.2021, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12577836215496063, + "step": 6055, + "valid_targets_mean": 4269.0, + "valid_targets_min": 3041 + }, + { + "epoch": 4.4074208803201165, + "grad_norm": 0.6199471507652365, + "learning_rate": 1.4521815557941547e-05, + "loss": 0.2236, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10665874928236008, + "step": 6060, + "valid_targets_mean": 3350.1, + "valid_targets_min": 1726 + }, + { + "epoch": 4.411058566751546, + "grad_norm": 0.5339748227298076, + "learning_rate": 1.4486942984924519e-05, + "loss": 0.2103, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09804654866456985, + "step": 6065, + "valid_targets_mean": 4747.4, + "valid_targets_min": 1743 + }, + { + "epoch": 4.414696253182976, + "grad_norm": 0.59680802457853, + "learning_rate": 1.4452088541822156e-05, + "loss": 0.2174, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10068153589963913, + "step": 6070, + "valid_targets_mean": 3702.8, + "valid_targets_min": 1234 + }, + { + "epoch": 4.4183339396144055, + "grad_norm": 0.6354155551649933, + "learning_rate": 1.4417252343254739e-05, + "loss": 0.2228, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11466504633426666, + "step": 6075, + "valid_targets_mean": 4079.0, + "valid_targets_min": 2303 + }, + { + "epoch": 4.421971626045835, + "grad_norm": 0.6181873876345484, + "learning_rate": 1.4382434503782538e-05, + "loss": 0.224, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10687104612588882, + "step": 6080, + "valid_targets_mean": 3830.8, + "valid_targets_min": 1651 + }, + { + "epoch": 4.425609312477264, + "grad_norm": 0.6575890379136939, + "learning_rate": 1.4347635137905445e-05, + "loss": 0.2139, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11098949611186981, + "step": 6085, + "valid_targets_mean": 2876.2, + "valid_targets_min": 1374 + }, + { + "epoch": 4.429246998908694, + "grad_norm": 0.6383721411006934, + "learning_rate": 1.4312854360062615e-05, + "loss": 0.2089, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10737615823745728, + "step": 6090, + "valid_targets_mean": 3467.5, + "valid_targets_min": 1625 + }, + { + "epoch": 4.432884685340124, + "grad_norm": 0.5892484967384606, + "learning_rate": 1.427809228463205e-05, + "loss": 0.2035, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10160204023122787, + "step": 6095, + "valid_targets_mean": 4089.5, + "valid_targets_min": 997 + }, + { + "epoch": 4.436522371771553, + "grad_norm": 0.624487768391819, + "learning_rate": 1.4243349025930286e-05, + "loss": 0.2032, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08693476021289825, + "step": 6100, + "valid_targets_mean": 3618.6, + "valid_targets_min": 1179 + }, + { + "epoch": 4.4401600582029825, + "grad_norm": 0.6170166568705118, + "learning_rate": 1.4208624698211937e-05, + "loss": 0.2037, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09555166959762573, + "step": 6105, + "valid_targets_mean": 3079.1, + "valid_targets_min": 1444 + }, + { + "epoch": 4.443797744634413, + "grad_norm": 0.6326745434595876, + "learning_rate": 1.41739194156694e-05, + "loss": 0.1959, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08613644540309906, + "step": 6110, + "valid_targets_mean": 3486.8, + "valid_targets_min": 1504 + }, + { + "epoch": 4.447435431065842, + "grad_norm": 0.6159934760230874, + "learning_rate": 1.4139233292432411e-05, + "loss": 0.2009, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10092556476593018, + "step": 6115, + "valid_targets_mean": 3998.4, + "valid_targets_min": 2302 + }, + { + "epoch": 4.4510731174972715, + "grad_norm": 0.6654815481699756, + "learning_rate": 1.4104566442567712e-05, + "loss": 0.2107, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1317807137966156, + "step": 6120, + "valid_targets_mean": 4282.5, + "valid_targets_min": 1818 + }, + { + "epoch": 4.454710803928702, + "grad_norm": 0.6677523962689633, + "learning_rate": 1.406991898007868e-05, + "loss": 0.2295, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11291779577732086, + "step": 6125, + "valid_targets_mean": 4221.9, + "valid_targets_min": 2633 + }, + { + "epoch": 4.458348490360131, + "grad_norm": 0.6289812391308073, + "learning_rate": 1.4035291018904896e-05, + "loss": 0.2043, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10719780623912811, + "step": 6130, + "valid_targets_mean": 3522.4, + "valid_targets_min": 2481 + }, + { + "epoch": 4.4619861767915605, + "grad_norm": 0.6543404234127959, + "learning_rate": 1.400068267292186e-05, + "loss": 0.2126, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11679226160049438, + "step": 6135, + "valid_targets_mean": 4058.8, + "valid_targets_min": 2013 + }, + { + "epoch": 4.46562386322299, + "grad_norm": 0.6970545520838164, + "learning_rate": 1.396609405594052e-05, + "loss": 0.2053, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10996805131435394, + "step": 6140, + "valid_targets_mean": 3723.5, + "valid_targets_min": 1744 + }, + { + "epoch": 4.46926154965442, + "grad_norm": 0.9190472260513889, + "learning_rate": 1.3931525281706972e-05, + "loss": 0.2127, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11103987693786621, + "step": 6145, + "valid_targets_mean": 4180.9, + "valid_targets_min": 1914 + }, + { + "epoch": 4.4728992360858495, + "grad_norm": 0.7719043730232972, + "learning_rate": 1.3896976463902054e-05, + "loss": 0.2052, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09050468355417252, + "step": 6150, + "valid_targets_mean": 3668.6, + "valid_targets_min": 3065 + }, + { + "epoch": 4.476536922517279, + "grad_norm": 0.5807046676143821, + "learning_rate": 1.3862447716140973e-05, + "loss": 0.1982, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11906305700540543, + "step": 6155, + "valid_targets_mean": 4473.2, + "valid_targets_min": 2894 + }, + { + "epoch": 4.480174608948708, + "grad_norm": 0.8258928133796488, + "learning_rate": 1.3827939151972937e-05, + "loss": 0.2135, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.104790098965168, + "step": 6160, + "valid_targets_mean": 3395.6, + "valid_targets_min": 2359 + }, + { + "epoch": 4.4838122953801385, + "grad_norm": 0.6443667271568224, + "learning_rate": 1.3793450884880767e-05, + "loss": 0.2024, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09202464669942856, + "step": 6165, + "valid_targets_mean": 3114.5, + "valid_targets_min": 1463 + }, + { + "epoch": 4.487449981811568, + "grad_norm": 0.722339980827066, + "learning_rate": 1.3758983028280568e-05, + "loss": 0.213, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10954804718494415, + "step": 6170, + "valid_targets_mean": 3588.1, + "valid_targets_min": 1426 + }, + { + "epoch": 4.491087668242997, + "grad_norm": 0.6620876798619384, + "learning_rate": 1.3724535695521289e-05, + "loss": 0.2197, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10517523437738419, + "step": 6175, + "valid_targets_mean": 3405.5, + "valid_targets_min": 1661 + }, + { + "epoch": 4.4947253546744275, + "grad_norm": 0.5775473663509162, + "learning_rate": 1.36901089998844e-05, + "loss": 0.2171, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1123870313167572, + "step": 6180, + "valid_targets_mean": 4457.8, + "valid_targets_min": 1529 + }, + { + "epoch": 4.498363041105857, + "grad_norm": 0.7175619237943752, + "learning_rate": 1.3655703054583512e-05, + "loss": 0.229, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12766525149345398, + "step": 6185, + "valid_targets_mean": 3272.9, + "valid_targets_min": 1450 + }, + { + "epoch": 4.502000727537286, + "grad_norm": 0.6985370757606914, + "learning_rate": 1.3621317972763981e-05, + "loss": 0.2095, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09593132138252258, + "step": 6190, + "valid_targets_mean": 3103.2, + "valid_targets_min": 1625 + }, + { + "epoch": 4.505638413968716, + "grad_norm": 0.5967350407163039, + "learning_rate": 1.3586953867502576e-05, + "loss": 0.2207, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09156215190887451, + "step": 6195, + "valid_targets_mean": 3952.2, + "valid_targets_min": 1174 + }, + { + "epoch": 4.509276100400146, + "grad_norm": 0.6227520681533228, + "learning_rate": 1.3552610851807062e-05, + "loss": 0.2176, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10852313041687012, + "step": 6200, + "valid_targets_mean": 4039.2, + "valid_targets_min": 2882 + }, + { + "epoch": 4.512913786831575, + "grad_norm": 0.6315949670260481, + "learning_rate": 1.3518289038615853e-05, + "loss": 0.2189, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12301957607269287, + "step": 6205, + "valid_targets_mean": 4594.0, + "valid_targets_min": 1946 + }, + { + "epoch": 4.5165514732630045, + "grad_norm": 0.6788120961905276, + "learning_rate": 1.3483988540797652e-05, + "loss": 0.2182, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10313686728477478, + "step": 6210, + "valid_targets_mean": 4082.6, + "valid_targets_min": 2000 + }, + { + "epoch": 4.520189159694434, + "grad_norm": 0.6582039260840834, + "learning_rate": 1.3449709471151051e-05, + "loss": 0.2164, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11530055105686188, + "step": 6215, + "valid_targets_mean": 4596.1, + "valid_targets_min": 2012 + }, + { + "epoch": 4.523826846125864, + "grad_norm": 0.6853702627921637, + "learning_rate": 1.3415451942404185e-05, + "loss": 0.2176, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10875741392374039, + "step": 6220, + "valid_targets_mean": 3094.4, + "valid_targets_min": 719 + }, + { + "epoch": 4.5274645325572935, + "grad_norm": 0.6127993915454828, + "learning_rate": 1.338121606721434e-05, + "loss": 0.2173, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10118193924427032, + "step": 6225, + "valid_targets_mean": 4496.2, + "valid_targets_min": 3238 + }, + { + "epoch": 4.531102218988723, + "grad_norm": 0.6098469424017015, + "learning_rate": 1.334700195816761e-05, + "loss": 0.2041, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08917822688817978, + "step": 6230, + "valid_targets_mean": 3816.8, + "valid_targets_min": 1372 + }, + { + "epoch": 4.534739905420153, + "grad_norm": 0.6648472764704159, + "learning_rate": 1.3312809727778486e-05, + "loss": 0.2017, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11358654499053955, + "step": 6235, + "valid_targets_mean": 3306.1, + "valid_targets_min": 2370 + }, + { + "epoch": 4.5383775918515825, + "grad_norm": 0.5850287276129595, + "learning_rate": 1.3278639488489527e-05, + "loss": 0.2082, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09406204521656036, + "step": 6240, + "valid_targets_mean": 2958.1, + "valid_targets_min": 1576 + }, + { + "epoch": 4.542015278283012, + "grad_norm": 0.6601277678630916, + "learning_rate": 1.3244491352670975e-05, + "loss": 0.2066, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11224832385778427, + "step": 6245, + "valid_targets_mean": 3758.5, + "valid_targets_min": 1426 + }, + { + "epoch": 4.545652964714441, + "grad_norm": 0.6695322246296539, + "learning_rate": 1.3210365432620371e-05, + "loss": 0.218, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09769761562347412, + "step": 6250, + "valid_targets_mean": 3117.0, + "valid_targets_min": 1728 + }, + { + "epoch": 4.5492906511458715, + "grad_norm": 0.7899394878449251, + "learning_rate": 1.3176261840562216e-05, + "loss": 0.2103, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1149214506149292, + "step": 6255, + "valid_targets_mean": 3264.2, + "valid_targets_min": 1362 + }, + { + "epoch": 4.552928337577301, + "grad_norm": 0.6001746207711385, + "learning_rate": 1.3142180688647574e-05, + "loss": 0.2143, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10592121630907059, + "step": 6260, + "valid_targets_mean": 3836.0, + "valid_targets_min": 852 + }, + { + "epoch": 4.55656602400873, + "grad_norm": 0.7249339371181663, + "learning_rate": 1.3108122088953704e-05, + "loss": 0.212, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1317460834980011, + "step": 6265, + "valid_targets_mean": 2764.1, + "valid_targets_min": 1195 + }, + { + "epoch": 4.56020371044016, + "grad_norm": 0.620510288658341, + "learning_rate": 1.3074086153483718e-05, + "loss": 0.2081, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1117672547698021, + "step": 6270, + "valid_targets_mean": 4457.5, + "valid_targets_min": 1564 + }, + { + "epoch": 4.56384139687159, + "grad_norm": 0.5977216165458243, + "learning_rate": 1.3040072994166186e-05, + "loss": 0.2077, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0921715795993805, + "step": 6275, + "valid_targets_mean": 4371.2, + "valid_targets_min": 1975 + }, + { + "epoch": 4.567479083303019, + "grad_norm": 0.5991617319916325, + "learning_rate": 1.3006082722854786e-05, + "loss": 0.1956, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10387928038835526, + "step": 6280, + "valid_targets_mean": 3965.8, + "valid_targets_min": 2426 + }, + { + "epoch": 4.571116769734449, + "grad_norm": 0.5256116701713895, + "learning_rate": 1.2972115451327918e-05, + "loss": 0.202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10245691239833832, + "step": 6285, + "valid_targets_mean": 6384.6, + "valid_targets_min": 2678 + }, + { + "epoch": 4.574754456165879, + "grad_norm": 0.6135646023126828, + "learning_rate": 1.2938171291288359e-05, + "loss": 0.2065, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1004362553358078, + "step": 6290, + "valid_targets_mean": 3543.2, + "valid_targets_min": 2826 + }, + { + "epoch": 4.578392142597308, + "grad_norm": 0.6569735829938845, + "learning_rate": 1.2904250354362874e-05, + "loss": 0.1972, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08405405282974243, + "step": 6295, + "valid_targets_mean": 3155.8, + "valid_targets_min": 993 + }, + { + "epoch": 4.582029829028738, + "grad_norm": 0.6322636273991732, + "learning_rate": 1.2870352752101841e-05, + "loss": 0.195, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10934832692146301, + "step": 6300, + "valid_targets_mean": 4166.5, + "valid_targets_min": 2724 + }, + { + "epoch": 4.585667515460168, + "grad_norm": 0.7004297374039018, + "learning_rate": 1.283647859597894e-05, + "loss": 0.1973, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09491347521543503, + "step": 6305, + "valid_targets_mean": 3079.5, + "valid_targets_min": 1360 + }, + { + "epoch": 4.589305201891597, + "grad_norm": 0.6287352966875873, + "learning_rate": 1.2802627997390715e-05, + "loss": 0.1992, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10119875520467758, + "step": 6310, + "valid_targets_mean": 4114.9, + "valid_targets_min": 2414 + }, + { + "epoch": 4.5929428883230266, + "grad_norm": 0.6366975958816601, + "learning_rate": 1.2768801067656258e-05, + "loss": 0.2157, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11901592463254929, + "step": 6315, + "valid_targets_mean": 4524.8, + "valid_targets_min": 2425 + }, + { + "epoch": 4.596580574754456, + "grad_norm": 0.6652163249494156, + "learning_rate": 1.2734997918016818e-05, + "loss": 0.2042, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09903083741664886, + "step": 6320, + "valid_targets_mean": 3035.2, + "valid_targets_min": 1720 + }, + { + "epoch": 4.600218261185885, + "grad_norm": 0.6854849984037822, + "learning_rate": 1.2701218659635428e-05, + "loss": 0.2024, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1036415696144104, + "step": 6325, + "valid_targets_mean": 3260.4, + "valid_targets_min": 1621 + }, + { + "epoch": 4.6038559476173155, + "grad_norm": 0.6924140491028604, + "learning_rate": 1.2667463403596592e-05, + "loss": 0.1956, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1063183844089508, + "step": 6330, + "valid_targets_mean": 3214.6, + "valid_targets_min": 1858 + }, + { + "epoch": 4.607493634048745, + "grad_norm": 0.6735453190616384, + "learning_rate": 1.2633732260905836e-05, + "loss": 0.2055, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09676910936832428, + "step": 6335, + "valid_targets_mean": 3254.2, + "valid_targets_min": 1254 + }, + { + "epoch": 4.611131320480174, + "grad_norm": 0.6625503755882753, + "learning_rate": 1.2600025342489422e-05, + "loss": 0.2014, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1069498136639595, + "step": 6340, + "valid_targets_mean": 2994.4, + "valid_targets_min": 1039 + }, + { + "epoch": 4.6147690069116045, + "grad_norm": 0.6253422590455798, + "learning_rate": 1.256634275919393e-05, + "loss": 0.21, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12980233132839203, + "step": 6345, + "valid_targets_mean": 4625.4, + "valid_targets_min": 2526 + }, + { + "epoch": 4.618406693343034, + "grad_norm": 0.6123419156026239, + "learning_rate": 1.2532684621785928e-05, + "loss": 0.2036, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10972052812576294, + "step": 6350, + "valid_targets_mean": 3571.8, + "valid_targets_min": 2409 + }, + { + "epoch": 4.622044379774463, + "grad_norm": 0.6098014784653196, + "learning_rate": 1.2499051040951585e-05, + "loss": 0.2039, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08793787658214569, + "step": 6355, + "valid_targets_mean": 3691.5, + "valid_targets_min": 2918 + }, + { + "epoch": 4.6256820662058935, + "grad_norm": 0.6813766816050257, + "learning_rate": 1.2465442127296297e-05, + "loss": 0.2073, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10273462533950806, + "step": 6360, + "valid_targets_mean": 4063.1, + "valid_targets_min": 2179 + }, + { + "epoch": 4.629319752637323, + "grad_norm": 0.4715836658861294, + "learning_rate": 1.2431857991344384e-05, + "loss": 0.1748, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04725448042154312, + "step": 6365, + "valid_targets_mean": 2871.4, + "valid_targets_min": 955 + }, + { + "epoch": 4.632957439068752, + "grad_norm": 0.4557970367260976, + "learning_rate": 1.2398298743538639e-05, + "loss": 0.1064, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03985024243593216, + "step": 6370, + "valid_targets_mean": 1788.9, + "valid_targets_min": 669 + }, + { + "epoch": 4.636595125500182, + "grad_norm": 0.36736617587893894, + "learning_rate": 1.2364764494240036e-05, + "loss": 0.0784, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03866071254014969, + "step": 6375, + "valid_targets_mean": 3755.8, + "valid_targets_min": 2651 + }, + { + "epoch": 4.640232811931612, + "grad_norm": 0.434971210496454, + "learning_rate": 1.2331255353727332e-05, + "loss": 0.0793, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.043811287730932236, + "step": 6380, + "valid_targets_mean": 3069.8, + "valid_targets_min": 710 + }, + { + "epoch": 4.643870498363041, + "grad_norm": 0.3470148711090211, + "learning_rate": 1.229777143219671e-05, + "loss": 0.0923, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02765033207833767, + "step": 6385, + "valid_targets_mean": 3618.5, + "valid_targets_min": 1365 + }, + { + "epoch": 4.647508184794471, + "grad_norm": 0.44156710470428145, + "learning_rate": 1.2264312839761434e-05, + "loss": 0.0796, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04348709434270859, + "step": 6390, + "valid_targets_mean": 3279.9, + "valid_targets_min": 889 + }, + { + "epoch": 4.6511458712259, + "grad_norm": 0.7825479645751586, + "learning_rate": 1.2230879686451438e-05, + "loss": 0.1107, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.061785418540239334, + "step": 6395, + "valid_targets_mean": 1652.4, + "valid_targets_min": 559 + }, + { + "epoch": 4.65478355765733, + "grad_norm": 0.3919109793948965, + "learning_rate": 1.2197472082213048e-05, + "loss": 0.1093, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03413020446896553, + "step": 6400, + "valid_targets_mean": 3449.0, + "valid_targets_min": 1013 + }, + { + "epoch": 4.65842124408876, + "grad_norm": 0.4275827119357158, + "learning_rate": 1.2164090136908526e-05, + "loss": 0.0896, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04819536581635475, + "step": 6405, + "valid_targets_mean": 3726.1, + "valid_targets_min": 3222 + }, + { + "epoch": 4.662058930520189, + "grad_norm": 0.5436909400623704, + "learning_rate": 1.213073396031578e-05, + "loss": 0.1208, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08598495274782181, + "step": 6410, + "valid_targets_mean": 3205.8, + "valid_targets_min": 1949 + }, + { + "epoch": 4.665696616951619, + "grad_norm": 0.3172576957542065, + "learning_rate": 1.2097403662127966e-05, + "loss": 0.0679, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03210180997848511, + "step": 6415, + "valid_targets_mean": 4059.8, + "valid_targets_min": 886 + }, + { + "epoch": 4.669334303383049, + "grad_norm": 0.4738061707203333, + "learning_rate": 1.2064099351953136e-05, + "loss": 0.0816, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.044295962899923325, + "step": 6420, + "valid_targets_mean": 1858.2, + "valid_targets_min": 447 + }, + { + "epoch": 4.672971989814478, + "grad_norm": 0.3810085585504696, + "learning_rate": 1.2030821139313895e-05, + "loss": 0.079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03367253392934799, + "step": 6425, + "valid_targets_mean": 3195.6, + "valid_targets_min": 758 + }, + { + "epoch": 4.676609676245907, + "grad_norm": 0.5046338980931883, + "learning_rate": 1.1997569133646993e-05, + "loss": 0.098, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.053939059376716614, + "step": 6430, + "valid_targets_mean": 1808.4, + "valid_targets_min": 604 + }, + { + "epoch": 4.6802473626773375, + "grad_norm": 0.4243522183286933, + "learning_rate": 1.1964343444303044e-05, + "loss": 0.0979, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04156392067670822, + "step": 6435, + "valid_targets_mean": 3314.6, + "valid_targets_min": 1743 + }, + { + "epoch": 4.683885049108767, + "grad_norm": 0.3783121884121069, + "learning_rate": 1.1931144180546074e-05, + "loss": 0.0777, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03058028593659401, + "step": 6440, + "valid_targets_mean": 2440.4, + "valid_targets_min": 574 + }, + { + "epoch": 4.687522735540196, + "grad_norm": 0.42871451129689536, + "learning_rate": 1.189797145155323e-05, + "loss": 0.0766, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.037871867418289185, + "step": 6445, + "valid_targets_mean": 2535.6, + "valid_targets_min": 523 + }, + { + "epoch": 4.691160421971626, + "grad_norm": 0.5326533470580298, + "learning_rate": 1.1864825366414405e-05, + "loss": 0.0784, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.060371845960617065, + "step": 6450, + "valid_targets_mean": 3398.1, + "valid_targets_min": 2635 + }, + { + "epoch": 4.694798108403056, + "grad_norm": 0.4392952456877791, + "learning_rate": 1.1831706034131849e-05, + "loss": 0.0864, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03697090968489647, + "step": 6455, + "valid_targets_mean": 3164.8, + "valid_targets_min": 812 + }, + { + "epoch": 4.698435794834485, + "grad_norm": 0.4222957989213417, + "learning_rate": 1.1798613563619861e-05, + "loss": 0.0901, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04246261715888977, + "step": 6460, + "valid_targets_mean": 2463.9, + "valid_targets_min": 856 + }, + { + "epoch": 4.702073481265915, + "grad_norm": 0.5381923401321657, + "learning_rate": 1.176554806370438e-05, + "loss": 0.0894, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05300880968570709, + "step": 6465, + "valid_targets_mean": 2995.6, + "valid_targets_min": 718 + }, + { + "epoch": 4.705711167697345, + "grad_norm": 0.5462907843341885, + "learning_rate": 1.1732509643122673e-05, + "loss": 0.1093, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11761785298585892, + "step": 6470, + "valid_targets_mean": 2866.9, + "valid_targets_min": 1297 + }, + { + "epoch": 4.709348854128774, + "grad_norm": 0.5035274656990405, + "learning_rate": 1.169949841052294e-05, + "loss": 0.1033, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.037283893674612045, + "step": 6475, + "valid_targets_mean": 1661.1, + "valid_targets_min": 786 + }, + { + "epoch": 4.712986540560204, + "grad_norm": 0.6336765795678937, + "learning_rate": 1.1666514474463963e-05, + "loss": 0.0794, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05520862340927124, + "step": 6480, + "valid_targets_mean": 1624.2, + "valid_targets_min": 550 + }, + { + "epoch": 4.716624226991633, + "grad_norm": 0.44714431433485097, + "learning_rate": 1.1633557943414797e-05, + "loss": 0.1106, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.041426822543144226, + "step": 6485, + "valid_targets_mean": 3254.6, + "valid_targets_min": 1304 + }, + { + "epoch": 4.720261913423063, + "grad_norm": 0.44085914888516614, + "learning_rate": 1.1600628925754331e-05, + "loss": 0.1008, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.036699213087558746, + "step": 6490, + "valid_targets_mean": 2833.0, + "valid_targets_min": 1247 + }, + { + "epoch": 4.723899599854493, + "grad_norm": 0.589381606086901, + "learning_rate": 1.1567727529771004e-05, + "loss": 0.0831, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05179157853126526, + "step": 6495, + "valid_targets_mean": 3056.5, + "valid_targets_min": 800 + }, + { + "epoch": 4.727537286285922, + "grad_norm": 0.5883055990626056, + "learning_rate": 1.1534853863662398e-05, + "loss": 0.267, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11429660022258759, + "step": 6500, + "valid_targets_mean": 2580.8, + "valid_targets_min": 987 + }, + { + "epoch": 4.731174972717351, + "grad_norm": 0.6066574644580103, + "learning_rate": 1.1502008035534924e-05, + "loss": 0.1067, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08849574625492096, + "step": 6505, + "valid_targets_mean": 2245.6, + "valid_targets_min": 776 + }, + { + "epoch": 4.734812659148782, + "grad_norm": 0.39203352863179497, + "learning_rate": 1.1469190153403443e-05, + "loss": 0.0875, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.025488954037427902, + "step": 6510, + "valid_targets_mean": 1361.5, + "valid_targets_min": 658 + }, + { + "epoch": 4.738450345580211, + "grad_norm": 0.49350794242118623, + "learning_rate": 1.1436400325190898e-05, + "loss": 0.1819, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03915173560380936, + "step": 6515, + "valid_targets_mean": 1729.0, + "valid_targets_min": 666 + }, + { + "epoch": 4.74208803201164, + "grad_norm": 0.3304719783880874, + "learning_rate": 1.1403638658727995e-05, + "loss": 0.0772, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03298184275627136, + "step": 6520, + "valid_targets_mean": 4975.0, + "valid_targets_min": 3529 + }, + { + "epoch": 4.745725718443071, + "grad_norm": 0.4831866419141741, + "learning_rate": 1.1370905261752815e-05, + "loss": 0.0891, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05346904695034027, + "step": 6525, + "valid_targets_mean": 3988.8, + "valid_targets_min": 1746 + }, + { + "epoch": 4.7493634048745, + "grad_norm": 0.4789943678450568, + "learning_rate": 1.1338200241910485e-05, + "loss": 0.0855, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05317910760641098, + "step": 6530, + "valid_targets_mean": 3222.4, + "valid_targets_min": 1226 + }, + { + "epoch": 4.753001091305929, + "grad_norm": 0.41801667751372207, + "learning_rate": 1.1305523706752812e-05, + "loss": 0.0766, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05118561536073685, + "step": 6535, + "valid_targets_mean": 4403.6, + "valid_targets_min": 3275 + }, + { + "epoch": 4.756638777737359, + "grad_norm": 0.36396195243200663, + "learning_rate": 1.1272875763737914e-05, + "loss": 0.0784, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03229308873414993, + "step": 6540, + "valid_targets_mean": 2546.2, + "valid_targets_min": 723 + }, + { + "epoch": 4.760276464168789, + "grad_norm": 0.4171445780297137, + "learning_rate": 1.1240256520229904e-05, + "loss": 0.076, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.045592300593853, + "step": 6545, + "valid_targets_mean": 3348.6, + "valid_targets_min": 2234 + }, + { + "epoch": 4.763914150600218, + "grad_norm": 0.352427311816711, + "learning_rate": 1.120766608349849e-05, + "loss": 0.0738, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02514420449733734, + "step": 6550, + "valid_targets_mean": 3603.0, + "valid_targets_min": 1421 + }, + { + "epoch": 4.767551837031648, + "grad_norm": 0.43712932298578205, + "learning_rate": 1.1175104560718686e-05, + "loss": 0.0755, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.040313418954610825, + "step": 6555, + "valid_targets_mean": 2849.2, + "valid_targets_min": 1068 + }, + { + "epoch": 4.771189523463077, + "grad_norm": 0.4438785521386183, + "learning_rate": 1.1142572058970389e-05, + "loss": 0.1175, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.031142400577664375, + "step": 6560, + "valid_targets_mean": 1587.9, + "valid_targets_min": 535 + }, + { + "epoch": 4.774827209894507, + "grad_norm": 0.3818402131088769, + "learning_rate": 1.111006868523806e-05, + "loss": 0.0632, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.025606829673051834, + "step": 6565, + "valid_targets_mean": 1702.4, + "valid_targets_min": 585 + }, + { + "epoch": 4.778464896325937, + "grad_norm": 0.4234236788425016, + "learning_rate": 1.1077594546410384e-05, + "loss": 0.0736, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03472496569156647, + "step": 6570, + "valid_targets_mean": 2588.6, + "valid_targets_min": 582 + }, + { + "epoch": 4.782102582757366, + "grad_norm": 0.4395638885281108, + "learning_rate": 1.1045149749279905e-05, + "loss": 0.0691, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03750314563512802, + "step": 6575, + "valid_targets_mean": 3144.0, + "valid_targets_min": 516 + }, + { + "epoch": 4.785740269188796, + "grad_norm": 0.5320450794130852, + "learning_rate": 1.1012734400542682e-05, + "loss": 0.1133, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05367879569530487, + "step": 6580, + "valid_targets_mean": 3241.4, + "valid_targets_min": 999 + }, + { + "epoch": 4.789377955620226, + "grad_norm": 0.6107193565247216, + "learning_rate": 1.09803486067979e-05, + "loss": 0.0912, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0808153748512268, + "step": 6585, + "valid_targets_mean": 2021.1, + "valid_targets_min": 696 + }, + { + "epoch": 4.793015642051655, + "grad_norm": 0.5121404054995107, + "learning_rate": 1.0947992474547595e-05, + "loss": 0.1001, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04429614916443825, + "step": 6590, + "valid_targets_mean": 2048.5, + "valid_targets_min": 635 + }, + { + "epoch": 4.796653328483085, + "grad_norm": 0.4102273132169571, + "learning_rate": 1.091566611019622e-05, + "loss": 0.0831, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.036619868129491806, + "step": 6595, + "valid_targets_mean": 2641.6, + "valid_targets_min": 816 + }, + { + "epoch": 4.800291014914515, + "grad_norm": 0.6584860554868548, + "learning_rate": 1.088336962005036e-05, + "loss": 0.1273, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.136690154671669, + "step": 6600, + "valid_targets_mean": 2236.4, + "valid_targets_min": 949 + }, + { + "epoch": 4.803928701345944, + "grad_norm": 0.4389226719161278, + "learning_rate": 1.0851103110318366e-05, + "loss": 0.0952, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04370023310184479, + "step": 6605, + "valid_targets_mean": 3270.9, + "valid_targets_min": 2201 + }, + { + "epoch": 4.807566387777373, + "grad_norm": 0.7356186260274163, + "learning_rate": 1.0818866687109957e-05, + "loss": 0.1067, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07638288289308548, + "step": 6610, + "valid_targets_mean": 1500.5, + "valid_targets_min": 687 + }, + { + "epoch": 4.811204074208804, + "grad_norm": 0.4723612864475983, + "learning_rate": 1.0786660456435955e-05, + "loss": 0.0748, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03331270068883896, + "step": 6615, + "valid_targets_mean": 2285.5, + "valid_targets_min": 449 + }, + { + "epoch": 4.814841760640233, + "grad_norm": 0.4073972527228057, + "learning_rate": 1.075448452420786e-05, + "loss": 0.0836, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0253402441740036, + "step": 6620, + "valid_targets_mean": 2368.1, + "valid_targets_min": 658 + }, + { + "epoch": 4.818479447071662, + "grad_norm": 0.6307346910192695, + "learning_rate": 1.0722338996237553e-05, + "loss": 0.0887, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.033539287745952606, + "step": 6625, + "valid_targets_mean": 831.8, + "valid_targets_min": 516 + }, + { + "epoch": 4.822117133503092, + "grad_norm": 0.3599077325535755, + "learning_rate": 1.0690223978236936e-05, + "loss": 0.0704, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04977177083492279, + "step": 6630, + "valid_targets_mean": 4461.9, + "valid_targets_min": 2805 + }, + { + "epoch": 4.825754819934522, + "grad_norm": 0.386732245841343, + "learning_rate": 1.0658139575817548e-05, + "loss": 0.0661, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03994423896074295, + "step": 6635, + "valid_targets_mean": 3766.8, + "valid_targets_min": 2036 + }, + { + "epoch": 4.829392506365951, + "grad_norm": 0.39817191217403836, + "learning_rate": 1.0626085894490272e-05, + "loss": 0.0781, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03285779058933258, + "step": 6640, + "valid_targets_mean": 2336.5, + "valid_targets_min": 681 + }, + { + "epoch": 4.833030192797381, + "grad_norm": 0.4054335798986033, + "learning_rate": 1.0594063039664962e-05, + "loss": 0.0917, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04155734181404114, + "step": 6645, + "valid_targets_mean": 3856.1, + "valid_targets_min": 3436 + }, + { + "epoch": 4.836667879228811, + "grad_norm": 0.5379388444055749, + "learning_rate": 1.0562071116650095e-05, + "loss": 0.0921, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0569240041077137, + "step": 6650, + "valid_targets_mean": 3165.5, + "valid_targets_min": 891 + }, + { + "epoch": 4.84030556566024, + "grad_norm": 0.31940798624247957, + "learning_rate": 1.053011023065242e-05, + "loss": 0.0795, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.027199801057577133, + "step": 6655, + "valid_targets_mean": 3550.6, + "valid_targets_min": 2573 + }, + { + "epoch": 4.84394325209167, + "grad_norm": 0.3460938130943227, + "learning_rate": 1.0498180486776617e-05, + "loss": 0.0745, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03397299349308014, + "step": 6660, + "valid_targets_mean": 2995.0, + "valid_targets_min": 928 + }, + { + "epoch": 4.847580938523099, + "grad_norm": 0.5391350475550482, + "learning_rate": 1.046628199002497e-05, + "loss": 0.0928, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05108436197042465, + "step": 6665, + "valid_targets_mean": 3839.6, + "valid_targets_min": 3392 + }, + { + "epoch": 4.851218624954529, + "grad_norm": 0.323326000464196, + "learning_rate": 1.0434414845296992e-05, + "loss": 0.0663, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03651874512434006, + "step": 6670, + "valid_targets_mean": 3855.0, + "valid_targets_min": 2848 + }, + { + "epoch": 4.854856311385959, + "grad_norm": 0.4714005391270297, + "learning_rate": 1.0402579157389115e-05, + "loss": 0.0828, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0413191094994545, + "step": 6675, + "valid_targets_mean": 2586.0, + "valid_targets_min": 595 + }, + { + "epoch": 4.858493997817388, + "grad_norm": 0.48895371992350767, + "learning_rate": 1.0370775030994295e-05, + "loss": 0.0816, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04428498446941376, + "step": 6680, + "valid_targets_mean": 2471.8, + "valid_targets_min": 1016 + }, + { + "epoch": 4.862131684248817, + "grad_norm": 0.5614578446398638, + "learning_rate": 1.0339002570701702e-05, + "loss": 0.0872, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06654151529073715, + "step": 6685, + "valid_targets_mean": 1567.8, + "valid_targets_min": 811 + }, + { + "epoch": 4.865769370680248, + "grad_norm": 0.6864016935447281, + "learning_rate": 1.0307261880996387e-05, + "loss": 0.09, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05250755697488785, + "step": 6690, + "valid_targets_mean": 1569.1, + "valid_targets_min": 697 + }, + { + "epoch": 4.869407057111677, + "grad_norm": 0.723017144588585, + "learning_rate": 1.027555306625891e-05, + "loss": 0.0779, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.058025624603033066, + "step": 6695, + "valid_targets_mean": 1448.1, + "valid_targets_min": 736 + }, + { + "epoch": 4.873044743543106, + "grad_norm": 0.429454839175072, + "learning_rate": 1.0243876230765024e-05, + "loss": 0.0824, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02732866071164608, + "step": 6700, + "valid_targets_mean": 2415.5, + "valid_targets_min": 804 + }, + { + "epoch": 4.876682429974537, + "grad_norm": 0.5363918086364604, + "learning_rate": 1.0212231478685286e-05, + "loss": 0.0977, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03533122316002846, + "step": 6705, + "valid_targets_mean": 3472.5, + "valid_targets_min": 1018 + }, + { + "epoch": 4.880320116405966, + "grad_norm": 0.41344628666575356, + "learning_rate": 1.0180618914084776e-05, + "loss": 0.0824, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0391097366809845, + "step": 6710, + "valid_targets_mean": 3623.8, + "valid_targets_min": 2369 + }, + { + "epoch": 4.883957802837395, + "grad_norm": 0.444558619022788, + "learning_rate": 1.0149038640922715e-05, + "loss": 0.0703, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04006372392177582, + "step": 6715, + "valid_targets_mean": 3135.8, + "valid_targets_min": 1318 + }, + { + "epoch": 4.887595489268825, + "grad_norm": 0.5194658636108421, + "learning_rate": 1.011749076305212e-05, + "loss": 0.0912, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06377632170915604, + "step": 6720, + "valid_targets_mean": 2258.2, + "valid_targets_min": 903 + }, + { + "epoch": 4.891233175700255, + "grad_norm": 0.442488828549806, + "learning_rate": 1.0085975384219493e-05, + "loss": 0.0869, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03866777941584587, + "step": 6725, + "valid_targets_mean": 3460.4, + "valid_targets_min": 2346 + }, + { + "epoch": 4.894870862131684, + "grad_norm": 0.6726545794319424, + "learning_rate": 1.0054492608064446e-05, + "loss": 0.071, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06590952724218369, + "step": 6730, + "valid_targets_mean": 2283.0, + "valid_targets_min": 737 + }, + { + "epoch": 4.898508548563114, + "grad_norm": 0.41977247835445164, + "learning_rate": 1.0023042538119383e-05, + "loss": 0.0766, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03807227313518524, + "step": 6735, + "valid_targets_mean": 3979.6, + "valid_targets_min": 2555 + }, + { + "epoch": 4.902146234994543, + "grad_norm": 0.32982402053732973, + "learning_rate": 9.991625277809167e-06, + "loss": 0.0734, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03251798823475838, + "step": 6740, + "valid_targets_mean": 4500.2, + "valid_targets_min": 2991 + }, + { + "epoch": 4.905783921425973, + "grad_norm": 0.3998023331764577, + "learning_rate": 9.960240930450728e-06, + "loss": 0.0634, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.026965569704771042, + "step": 6745, + "valid_targets_mean": 1992.4, + "valid_targets_min": 618 + }, + { + "epoch": 4.909421607857403, + "grad_norm": 0.3707630634637823, + "learning_rate": 9.928889599252802e-06, + "loss": 0.0861, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.026907937601208687, + "step": 6750, + "valid_targets_mean": 3641.8, + "valid_targets_min": 2944 + }, + { + "epoch": 4.913059294288832, + "grad_norm": 0.3443677299679073, + "learning_rate": 9.897571387315507e-06, + "loss": 0.0739, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04578782618045807, + "step": 6755, + "valid_targets_mean": 3686.1, + "valid_targets_min": 2702 + }, + { + "epoch": 4.916696980720262, + "grad_norm": 0.3553747064120349, + "learning_rate": 9.866286397630102e-06, + "loss": 0.1009, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.021052774041891098, + "step": 6760, + "valid_targets_mean": 3341.6, + "valid_targets_min": 509 + }, + { + "epoch": 4.920334667151692, + "grad_norm": 0.36903847945197477, + "learning_rate": 9.835034733078538e-06, + "loss": 0.0921, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.027694225311279297, + "step": 6765, + "valid_targets_mean": 2757.6, + "valid_targets_min": 637 + }, + { + "epoch": 4.923972353583121, + "grad_norm": 0.45865482784475703, + "learning_rate": 9.803816496433213e-06, + "loss": 0.1009, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.043617941439151764, + "step": 6770, + "valid_targets_mean": 2802.6, + "valid_targets_min": 737 + }, + { + "epoch": 4.92761004001455, + "grad_norm": 0.3515987014379923, + "learning_rate": 9.77263179035657e-06, + "loss": 0.0685, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03174688667058945, + "step": 6775, + "valid_targets_mean": 3312.9, + "valid_targets_min": 1292 + }, + { + "epoch": 4.931247726445981, + "grad_norm": 0.6103623922279477, + "learning_rate": 9.741480717400803e-06, + "loss": 0.0851, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04114576429128647, + "step": 6780, + "valid_targets_mean": 1049.6, + "valid_targets_min": 697 + }, + { + "epoch": 4.93488541287741, + "grad_norm": 0.3981517415260633, + "learning_rate": 9.710363380007501e-06, + "loss": 0.073, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03561249375343323, + "step": 6785, + "valid_targets_mean": 3541.4, + "valid_targets_min": 923 + }, + { + "epoch": 4.938523099308839, + "grad_norm": 1.0201085270169914, + "learning_rate": 9.679279880507294e-06, + "loss": 0.1109, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.060093462467193604, + "step": 6790, + "valid_targets_mean": 931.6, + "valid_targets_min": 531 + }, + { + "epoch": 4.942160785740269, + "grad_norm": 0.5656427896574032, + "learning_rate": 9.64823032111957e-06, + "loss": 0.1064, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05713935196399689, + "step": 6795, + "valid_targets_mean": 1840.5, + "valid_targets_min": 844 + }, + { + "epoch": 4.945798472171699, + "grad_norm": 0.5244217111870512, + "learning_rate": 9.617214803952055e-06, + "loss": 0.0798, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05276511609554291, + "step": 6800, + "valid_targets_mean": 3893.9, + "valid_targets_min": 2967 + }, + { + "epoch": 4.949436158603128, + "grad_norm": 0.39731354605548896, + "learning_rate": 9.586233431000567e-06, + "loss": 0.0775, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04224361479282379, + "step": 6805, + "valid_targets_mean": 3605.9, + "valid_targets_min": 2490 + }, + { + "epoch": 4.953073845034558, + "grad_norm": 0.420917558737506, + "learning_rate": 9.555286304148632e-06, + "loss": 0.0775, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03390621393918991, + "step": 6810, + "valid_targets_mean": 2682.0, + "valid_targets_min": 970 + }, + { + "epoch": 4.956711531465988, + "grad_norm": 0.491602229413475, + "learning_rate": 9.524373525167132e-06, + "loss": 0.0691, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03134778141975403, + "step": 6815, + "valid_targets_mean": 2028.5, + "valid_targets_min": 729 + }, + { + "epoch": 4.960349217897417, + "grad_norm": 1.0293003266602099, + "learning_rate": 9.493495195714028e-06, + "loss": 0.0923, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06407930701971054, + "step": 6820, + "valid_targets_mean": 1028.9, + "valid_targets_min": 637 + }, + { + "epoch": 4.963986904328847, + "grad_norm": 0.37632499744315867, + "learning_rate": 9.462651417333954e-06, + "loss": 0.072, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.026910055428743362, + "step": 6825, + "valid_targets_mean": 2879.2, + "valid_targets_min": 701 + }, + { + "epoch": 4.967624590760277, + "grad_norm": 0.3892968039613796, + "learning_rate": 9.431842291457974e-06, + "loss": 0.0759, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03147163987159729, + "step": 6830, + "valid_targets_mean": 3671.1, + "valid_targets_min": 3131 + }, + { + "epoch": 4.971262277191706, + "grad_norm": 0.5069360657900297, + "learning_rate": 9.401067919403148e-06, + "loss": 0.0751, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03921269625425339, + "step": 6835, + "valid_targets_mean": 3241.5, + "valid_targets_min": 2018 + }, + { + "epoch": 4.974899963623136, + "grad_norm": 0.34978253972944173, + "learning_rate": 9.370328402372257e-06, + "loss": 0.0845, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03464730083942413, + "step": 6840, + "valid_targets_mean": 2491.6, + "valid_targets_min": 350 + }, + { + "epoch": 4.978537650054565, + "grad_norm": 0.3196653309323728, + "learning_rate": 9.339623841453476e-06, + "loss": 0.0686, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03708865866065025, + "step": 6845, + "valid_targets_mean": 4862.8, + "valid_targets_min": 3577 + }, + { + "epoch": 4.982175336485994, + "grad_norm": 0.3702169705992094, + "learning_rate": 9.308954337620014e-06, + "loss": 0.0742, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030349217355251312, + "step": 6850, + "valid_targets_mean": 1999.4, + "valid_targets_min": 507 + }, + { + "epoch": 4.985813022917425, + "grad_norm": 0.4656323337914371, + "learning_rate": 9.278319991729801e-06, + "loss": 0.089, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08197690546512604, + "step": 6855, + "valid_targets_mean": 3278.5, + "valid_targets_min": 2814 + }, + { + "epoch": 4.989450709348854, + "grad_norm": 0.4171556178359351, + "learning_rate": 9.24772090452514e-06, + "loss": 0.1529, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.042212750762701035, + "step": 6860, + "valid_targets_mean": 4006.2, + "valid_targets_min": 2347 + }, + { + "epoch": 4.993088395780283, + "grad_norm": 0.5543444935068429, + "learning_rate": 9.217157176632378e-06, + "loss": 0.0779, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04535501077771187, + "step": 6865, + "valid_targets_mean": 2447.4, + "valid_targets_min": 716 + }, + { + "epoch": 4.996726082211714, + "grad_norm": 0.4039151368594099, + "learning_rate": 9.1866289085616e-06, + "loss": 0.0727, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.034677017480134964, + "step": 6870, + "valid_targets_mean": 3209.6, + "valid_targets_min": 2245 + }, + { + "epoch": 5.000727537286286, + "grad_norm": 1.1416467821892364, + "learning_rate": 9.156136200706265e-06, + "loss": 0.1437, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09463672339916229, + "step": 6875, + "valid_targets_mean": 8215.2, + "valid_targets_min": 6163 + }, + { + "epoch": 5.004365223717715, + "grad_norm": 0.6847670465311072, + "learning_rate": 9.125679153342912e-06, + "loss": 0.1911, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09924490749835968, + "step": 6880, + "valid_targets_mean": 8594.4, + "valid_targets_min": 6006 + }, + { + "epoch": 5.008002910149145, + "grad_norm": 0.5602166341967226, + "learning_rate": 9.095257866630774e-06, + "loss": 0.1763, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07732293009757996, + "step": 6885, + "valid_targets_mean": 3578.4, + "valid_targets_min": 289 + }, + { + "epoch": 5.011640596580575, + "grad_norm": 0.43531001429079946, + "learning_rate": 9.064872440611516e-06, + "loss": 0.1763, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08923172205686569, + "step": 6890, + "valid_targets_mean": 7019.9, + "valid_targets_min": 4985 + }, + { + "epoch": 5.015278283012004, + "grad_norm": 0.4077819585140151, + "learning_rate": 9.034522975208851e-06, + "loss": 0.1665, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08682319521903992, + "step": 6895, + "valid_targets_mean": 6956.6, + "valid_targets_min": 1457 + }, + { + "epoch": 5.018915969443434, + "grad_norm": 0.40362596369891823, + "learning_rate": 9.004209570228249e-06, + "loss": 0.1718, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08146252483129501, + "step": 6900, + "valid_targets_mean": 7361.6, + "valid_targets_min": 4961 + }, + { + "epoch": 5.022553655874864, + "grad_norm": 0.43121770079228916, + "learning_rate": 8.973932325356595e-06, + "loss": 0.1845, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10747069120407104, + "step": 6905, + "valid_targets_mean": 7058.5, + "valid_targets_min": 5985 + }, + { + "epoch": 5.026191342306293, + "grad_norm": 0.4106297879706851, + "learning_rate": 8.943691340161843e-06, + "loss": 0.1781, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09615565091371536, + "step": 6910, + "valid_targets_mean": 7664.2, + "valid_targets_min": 5169 + }, + { + "epoch": 5.029829028737723, + "grad_norm": 0.43257823564540093, + "learning_rate": 8.913486714092719e-06, + "loss": 0.1682, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08523549139499664, + "step": 6915, + "valid_targets_mean": 7111.8, + "valid_targets_min": 5348 + }, + { + "epoch": 5.033466715169152, + "grad_norm": 0.4584870256288148, + "learning_rate": 8.883318546478386e-06, + "loss": 0.1556, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07172805070877075, + "step": 6920, + "valid_targets_mean": 4238.9, + "valid_targets_min": 152 + }, + { + "epoch": 5.037104401600582, + "grad_norm": 0.43001874268271906, + "learning_rate": 8.853186936528089e-06, + "loss": 0.1618, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08758758008480072, + "step": 6925, + "valid_targets_mean": 7199.4, + "valid_targets_min": 4258 + }, + { + "epoch": 5.040742088032012, + "grad_norm": 0.4113756013614196, + "learning_rate": 8.823091983330883e-06, + "loss": 0.1675, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08103170990943909, + "step": 6930, + "valid_targets_mean": 7243.4, + "valid_targets_min": 5831 + }, + { + "epoch": 5.044379774463441, + "grad_norm": 0.4074133529991087, + "learning_rate": 8.793033785855243e-06, + "loss": 0.166, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08320692181587219, + "step": 6935, + "valid_targets_mean": 6950.6, + "valid_targets_min": 3862 + }, + { + "epoch": 5.0480174608948705, + "grad_norm": 0.4101854102702615, + "learning_rate": 8.763012442948795e-06, + "loss": 0.1646, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08831341564655304, + "step": 6940, + "valid_targets_mean": 6991.9, + "valid_targets_min": 5696 + }, + { + "epoch": 5.051655147326301, + "grad_norm": 0.5241420838104095, + "learning_rate": 8.733028053337957e-06, + "loss": 0.1581, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08797430247068405, + "step": 6945, + "valid_targets_mean": 6963.5, + "valid_targets_min": 4947 + }, + { + "epoch": 5.05529283375773, + "grad_norm": 0.39896337730948334, + "learning_rate": 8.703080715627639e-06, + "loss": 0.163, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07728785276412964, + "step": 6950, + "valid_targets_mean": 6436.9, + "valid_targets_min": 4988 + }, + { + "epoch": 5.0589305201891595, + "grad_norm": 0.38055443554446733, + "learning_rate": 8.67317052830088e-06, + "loss": 0.1674, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08168643712997437, + "step": 6955, + "valid_targets_mean": 7518.0, + "valid_targets_min": 5476 + }, + { + "epoch": 5.06256820662059, + "grad_norm": 0.36616736483979856, + "learning_rate": 8.643297589718557e-06, + "loss": 0.1613, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0763334333896637, + "step": 6960, + "valid_targets_mean": 7422.6, + "valid_targets_min": 5226 + }, + { + "epoch": 5.066205893052019, + "grad_norm": 0.4243403545501543, + "learning_rate": 8.613461998119066e-06, + "loss": 0.1708, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10530637949705124, + "step": 6965, + "valid_targets_mean": 7080.0, + "valid_targets_min": 3753 + }, + { + "epoch": 5.069843579483448, + "grad_norm": 0.49468764806976084, + "learning_rate": 8.583663851617971e-06, + "loss": 0.1861, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09384547919034958, + "step": 6970, + "valid_targets_mean": 6064.6, + "valid_targets_min": 5161 + }, + { + "epoch": 5.073481265914878, + "grad_norm": 0.42309858133320416, + "learning_rate": 8.553903248207714e-06, + "loss": 0.1334, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0785026103258133, + "step": 6975, + "valid_targets_mean": 7388.1, + "valid_targets_min": 5795 + }, + { + "epoch": 5.077118952346308, + "grad_norm": 0.43725189687503463, + "learning_rate": 8.524180285757253e-06, + "loss": 0.1579, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07613535970449448, + "step": 6980, + "valid_targets_mean": 7696.0, + "valid_targets_min": 4658 + }, + { + "epoch": 5.080756638777737, + "grad_norm": 0.405693362992898, + "learning_rate": 8.49449506201176e-06, + "loss": 0.161, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08673180639743805, + "step": 6985, + "valid_targets_mean": 7799.1, + "valid_targets_min": 5508 + }, + { + "epoch": 5.084394325209167, + "grad_norm": 0.4400263073424136, + "learning_rate": 8.464847674592338e-06, + "loss": 0.1709, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08398684114217758, + "step": 6990, + "valid_targets_mean": 7341.6, + "valid_targets_min": 6454 + }, + { + "epoch": 5.088032011640596, + "grad_norm": 0.4288424841052647, + "learning_rate": 8.435238220995623e-06, + "loss": 0.1635, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0722421258687973, + "step": 6995, + "valid_targets_mean": 6818.0, + "valid_targets_min": 4392 + }, + { + "epoch": 5.091669698072026, + "grad_norm": 0.3785158692736306, + "learning_rate": 8.405666798593532e-06, + "loss": 0.163, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07172627747058868, + "step": 7000, + "valid_targets_mean": 7857.6, + "valid_targets_min": 5394 + }, + { + "epoch": 5.095307384503456, + "grad_norm": 0.36232941013378556, + "learning_rate": 8.376133504632893e-06, + "loss": 0.1432, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.067245252430439, + "step": 7005, + "valid_targets_mean": 8314.1, + "valid_targets_min": 4837 + }, + { + "epoch": 5.098945070934885, + "grad_norm": 0.3480026091517385, + "learning_rate": 8.34663843623516e-06, + "loss": 0.1442, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06151001900434494, + "step": 7010, + "valid_targets_mean": 7368.2, + "valid_targets_min": 6027 + }, + { + "epoch": 5.102582757366315, + "grad_norm": 0.4422758957407754, + "learning_rate": 8.31718169039609e-06, + "loss": 0.1509, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0813760757446289, + "step": 7015, + "valid_targets_mean": 6435.0, + "valid_targets_min": 4238 + }, + { + "epoch": 5.106220443797745, + "grad_norm": 1.2724604157682624, + "learning_rate": 8.287763363985385e-06, + "loss": 0.15, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04369153082370758, + "step": 7020, + "valid_targets_mean": 247.8, + "valid_targets_min": 137 + }, + { + "epoch": 5.109858130229174, + "grad_norm": 0.42373980505800757, + "learning_rate": 8.258383553746435e-06, + "loss": 0.1689, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07961566746234894, + "step": 7025, + "valid_targets_mean": 6807.0, + "valid_targets_min": 5165 + }, + { + "epoch": 5.1134958166606035, + "grad_norm": 0.48331921342544587, + "learning_rate": 8.229042356295939e-06, + "loss": 0.1757, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09070520102977753, + "step": 7030, + "valid_targets_mean": 6989.9, + "valid_targets_min": 5104 + }, + { + "epoch": 5.117133503092034, + "grad_norm": 0.42155765007989343, + "learning_rate": 8.199739868123645e-06, + "loss": 0.1725, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09284138679504395, + "step": 7035, + "valid_targets_mean": 6692.2, + "valid_targets_min": 5068 + }, + { + "epoch": 5.120771189523463, + "grad_norm": 0.4306241303731025, + "learning_rate": 8.170476185591988e-06, + "loss": 0.1685, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10115309059619904, + "step": 7040, + "valid_targets_mean": 7419.6, + "valid_targets_min": 5578 + }, + { + "epoch": 5.1244088759548925, + "grad_norm": 0.40873214684421966, + "learning_rate": 8.141251404935783e-06, + "loss": 0.1635, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.083727166056633, + "step": 7045, + "valid_targets_mean": 7536.5, + "valid_targets_min": 5075 + }, + { + "epoch": 5.128046562386323, + "grad_norm": 0.3996654519682595, + "learning_rate": 8.112065622261937e-06, + "loss": 0.1699, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07687846571207047, + "step": 7050, + "valid_targets_mean": 7221.5, + "valid_targets_min": 5841 + }, + { + "epoch": 5.131684248817752, + "grad_norm": 0.40078935587062636, + "learning_rate": 8.082918933549073e-06, + "loss": 0.1651, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07973235845565796, + "step": 7055, + "valid_targets_mean": 7287.5, + "valid_targets_min": 5151 + }, + { + "epoch": 5.1353219352491815, + "grad_norm": 0.45316499809059957, + "learning_rate": 8.053811434647306e-06, + "loss": 0.1741, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08045566082000732, + "step": 7060, + "valid_targets_mean": 6738.4, + "valid_targets_min": 4042 + }, + { + "epoch": 5.138959621680611, + "grad_norm": 0.44826459491658105, + "learning_rate": 8.024743221277817e-06, + "loss": 0.1677, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09291350841522217, + "step": 7065, + "valid_targets_mean": 6331.0, + "valid_targets_min": 5496 + }, + { + "epoch": 5.142597308112041, + "grad_norm": 0.4151930005926884, + "learning_rate": 7.995714389032638e-06, + "loss": 0.1705, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08111736178398132, + "step": 7070, + "valid_targets_mean": 6271.1, + "valid_targets_min": 4695 + }, + { + "epoch": 5.14623499454347, + "grad_norm": 0.8205601311075956, + "learning_rate": 7.966725033374265e-06, + "loss": 0.166, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0857987031340599, + "step": 7075, + "valid_targets_mean": 5790.8, + "valid_targets_min": 5021 + }, + { + "epoch": 5.1498726809749, + "grad_norm": 0.47349808054409975, + "learning_rate": 7.93777524963539e-06, + "loss": 0.1582, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06169222295284271, + "step": 7080, + "valid_targets_mean": 3386.8, + "valid_targets_min": 2279 + }, + { + "epoch": 5.153510367406329, + "grad_norm": 0.440673546783875, + "learning_rate": 7.90886513301858e-06, + "loss": 0.1392, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08098767697811127, + "step": 7085, + "valid_targets_mean": 6240.5, + "valid_targets_min": 5093 + }, + { + "epoch": 5.157148053837759, + "grad_norm": 0.45138889953484773, + "learning_rate": 7.879994778595918e-06, + "loss": 0.1632, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08371904492378235, + "step": 7090, + "valid_targets_mean": 6329.5, + "valid_targets_min": 4589 + }, + { + "epoch": 5.160785740269189, + "grad_norm": 0.4030510707395975, + "learning_rate": 7.851164281308775e-06, + "loss": 0.1639, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08790479600429535, + "step": 7095, + "valid_targets_mean": 7589.2, + "valid_targets_min": 4963 + }, + { + "epoch": 5.164423426700618, + "grad_norm": 0.4191632810721022, + "learning_rate": 7.822373735967408e-06, + "loss": 0.1637, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0704013854265213, + "step": 7100, + "valid_targets_mean": 5577.9, + "valid_targets_min": 4458 + }, + { + "epoch": 5.168061113132048, + "grad_norm": 0.4794869688207334, + "learning_rate": 7.793623237250716e-06, + "loss": 0.1636, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0795382559299469, + "step": 7105, + "valid_targets_mean": 5432.5, + "valid_targets_min": 4433 + }, + { + "epoch": 5.171698799563478, + "grad_norm": 0.44436358787617536, + "learning_rate": 7.764912879705905e-06, + "loss": 0.1682, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08800412714481354, + "step": 7110, + "valid_targets_mean": 6361.8, + "valid_targets_min": 5201 + }, + { + "epoch": 5.175336485994907, + "grad_norm": 0.46873480084984337, + "learning_rate": 7.736242757748152e-06, + "loss": 0.1663, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07858821004629135, + "step": 7115, + "valid_targets_mean": 5824.4, + "valid_targets_min": 5129 + }, + { + "epoch": 5.1789741724263365, + "grad_norm": 0.7569703362212336, + "learning_rate": 7.707612965660345e-06, + "loss": 0.1286, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07862572371959686, + "step": 7120, + "valid_targets_mean": 1637.6, + "valid_targets_min": 137 + }, + { + "epoch": 5.182611858857767, + "grad_norm": 0.5560513628497672, + "learning_rate": 7.679023597592713e-06, + "loss": 0.1599, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08066793531179428, + "step": 7125, + "valid_targets_mean": 6474.6, + "valid_targets_min": 4751 + }, + { + "epoch": 5.186249545289196, + "grad_norm": 0.4826701927963093, + "learning_rate": 7.650474747562595e-06, + "loss": 0.1616, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08919551968574524, + "step": 7130, + "valid_targets_mean": 7186.2, + "valid_targets_min": 5358 + }, + { + "epoch": 5.1898872317206255, + "grad_norm": 0.4368327487658346, + "learning_rate": 7.621966509454044e-06, + "loss": 0.1665, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0829877182841301, + "step": 7135, + "valid_targets_mean": 7126.8, + "valid_targets_min": 4078 + }, + { + "epoch": 5.193524918152056, + "grad_norm": 0.4259584823043044, + "learning_rate": 7.5934989770175685e-06, + "loss": 0.1583, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07801656424999237, + "step": 7140, + "valid_targets_mean": 7022.1, + "valid_targets_min": 4940 + }, + { + "epoch": 5.197162604583485, + "grad_norm": 0.4508191018843489, + "learning_rate": 7.565072243869822e-06, + "loss": 0.1566, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08522289991378784, + "step": 7145, + "valid_targets_mean": 5900.6, + "valid_targets_min": 4779 + }, + { + "epoch": 5.2008002910149145, + "grad_norm": 0.52382580557286, + "learning_rate": 7.536686403493289e-06, + "loss": 0.1645, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08387278765439987, + "step": 7150, + "valid_targets_mean": 6290.1, + "valid_targets_min": 4620 + }, + { + "epoch": 5.204437977446344, + "grad_norm": 0.6117776615790714, + "learning_rate": 7.5083415492359804e-06, + "loss": 0.2071, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11673711240291595, + "step": 7155, + "valid_targets_mean": 5614.5, + "valid_targets_min": 3446 + }, + { + "epoch": 5.208075663877774, + "grad_norm": 0.6100112758111218, + "learning_rate": 7.480037774311104e-06, + "loss": 0.2207, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11624610424041748, + "step": 7160, + "valid_targets_mean": 4630.0, + "valid_targets_min": 1005 + }, + { + "epoch": 5.2117133503092035, + "grad_norm": 0.5227468189582934, + "learning_rate": 7.451775171796782e-06, + "loss": 0.2192, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08250939846038818, + "step": 7165, + "valid_targets_mean": 4669.6, + "valid_targets_min": 2054 + }, + { + "epoch": 5.215351036740633, + "grad_norm": 0.6130061905577443, + "learning_rate": 7.423553834635753e-06, + "loss": 0.2146, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09231794625520706, + "step": 7170, + "valid_targets_mean": 3745.0, + "valid_targets_min": 1561 + }, + { + "epoch": 5.218988723172062, + "grad_norm": 0.5940056500480482, + "learning_rate": 7.395373855635044e-06, + "loss": 0.2163, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11066854745149612, + "step": 7175, + "valid_targets_mean": 4284.0, + "valid_targets_min": 1184 + }, + { + "epoch": 5.2226264096034924, + "grad_norm": 0.49751047010102334, + "learning_rate": 7.367235327465683e-06, + "loss": 0.2212, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1091056764125824, + "step": 7180, + "valid_targets_mean": 6332.8, + "valid_targets_min": 1651 + }, + { + "epoch": 5.226264096034922, + "grad_norm": 0.4965296245821765, + "learning_rate": 7.3391383426623596e-06, + "loss": 0.2111, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10682186484336853, + "step": 7185, + "valid_targets_mean": 6446.5, + "valid_targets_min": 2231 + }, + { + "epoch": 5.229901782466351, + "grad_norm": 0.561886706650693, + "learning_rate": 7.311082993623173e-06, + "loss": 0.2105, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07878410071134567, + "step": 7190, + "valid_targets_mean": 3367.2, + "valid_targets_min": 467 + }, + { + "epoch": 5.233539468897781, + "grad_norm": 0.5191770019967609, + "learning_rate": 7.2830693726093e-06, + "loss": 0.2176, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13817280530929565, + "step": 7195, + "valid_targets_mean": 7194.1, + "valid_targets_min": 1748 + }, + { + "epoch": 5.237177155329211, + "grad_norm": 0.5660134818566327, + "learning_rate": 7.255097571744669e-06, + "loss": 0.2197, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11122316867113113, + "step": 7200, + "valid_targets_mean": 4078.5, + "valid_targets_min": 1697 + }, + { + "epoch": 5.24081484176064, + "grad_norm": 0.6511256598678392, + "learning_rate": 7.22716768301571e-06, + "loss": 0.2132, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0715332180261612, + "step": 7205, + "valid_targets_mean": 2089.5, + "valid_targets_min": 1223 + }, + { + "epoch": 5.2444525281920695, + "grad_norm": 0.5141476544712069, + "learning_rate": 7.199279798271002e-06, + "loss": 0.2126, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09484773129224777, + "step": 7210, + "valid_targets_mean": 5404.1, + "valid_targets_min": 1829 + }, + { + "epoch": 5.2480902146235, + "grad_norm": 0.5445812610542259, + "learning_rate": 7.171434009221001e-06, + "loss": 0.2148, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10296923667192459, + "step": 7215, + "valid_targets_mean": 4583.1, + "valid_targets_min": 1596 + }, + { + "epoch": 5.251727901054929, + "grad_norm": 0.6331165667674571, + "learning_rate": 7.14363040743774e-06, + "loss": 0.209, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11402976512908936, + "step": 7220, + "valid_targets_mean": 4666.8, + "valid_targets_min": 2041 + }, + { + "epoch": 5.2553655874863585, + "grad_norm": 0.5891899714444249, + "learning_rate": 7.115869084354492e-06, + "loss": 0.2081, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11394612491130829, + "step": 7225, + "valid_targets_mean": 4272.1, + "valid_targets_min": 1538 + }, + { + "epoch": 5.259003273917788, + "grad_norm": 0.6882555082596504, + "learning_rate": 7.0881501312655234e-06, + "loss": 0.2065, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09355896711349487, + "step": 7230, + "valid_targets_mean": 2928.0, + "valid_targets_min": 1739 + }, + { + "epoch": 5.262640960349218, + "grad_norm": 0.634070426637795, + "learning_rate": 7.060473639325738e-06, + "loss": 0.2072, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09569108486175537, + "step": 7235, + "valid_targets_mean": 3254.5, + "valid_targets_min": 935 + }, + { + "epoch": 5.2662786467806475, + "grad_norm": 0.6728269055484786, + "learning_rate": 7.032839699550422e-06, + "loss": 0.213, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1058521494269371, + "step": 7240, + "valid_targets_mean": 3535.2, + "valid_targets_min": 2704 + }, + { + "epoch": 5.269916333212077, + "grad_norm": 0.6755268277629889, + "learning_rate": 7.005248402814924e-06, + "loss": 0.2103, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09862123429775238, + "step": 7245, + "valid_targets_mean": 2987.6, + "valid_targets_min": 1173 + }, + { + "epoch": 5.273554019643507, + "grad_norm": 0.6118891504614263, + "learning_rate": 6.977699839854366e-06, + "loss": 0.2146, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09953857958316803, + "step": 7250, + "valid_targets_mean": 3700.6, + "valid_targets_min": 1670 + }, + { + "epoch": 5.2771917060749365, + "grad_norm": 0.6674140580600891, + "learning_rate": 6.950194101263321e-06, + "loss": 0.2105, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09250658750534058, + "step": 7255, + "valid_targets_mean": 3481.6, + "valid_targets_min": 1329 + }, + { + "epoch": 5.280829392506366, + "grad_norm": 0.6396886639498675, + "learning_rate": 6.922731277495532e-06, + "loss": 0.1983, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10669510066509247, + "step": 7260, + "valid_targets_mean": 3589.8, + "valid_targets_min": 2279 + }, + { + "epoch": 5.284467078937795, + "grad_norm": 0.6173408515463669, + "learning_rate": 6.895311458863645e-06, + "loss": 0.1897, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09615373611450195, + "step": 7265, + "valid_targets_mean": 5060.4, + "valid_targets_min": 2222 + }, + { + "epoch": 5.2881047653692255, + "grad_norm": 0.6172970700049095, + "learning_rate": 6.867934735538844e-06, + "loss": 0.2015, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10154391825199127, + "step": 7270, + "valid_targets_mean": 3898.6, + "valid_targets_min": 1849 + }, + { + "epoch": 5.291742451800655, + "grad_norm": 0.5925709877060346, + "learning_rate": 6.8406011975506224e-06, + "loss": 0.2054, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0879693478345871, + "step": 7275, + "valid_targets_mean": 3698.9, + "valid_targets_min": 1951 + }, + { + "epoch": 5.295380138232084, + "grad_norm": 0.6816507573736271, + "learning_rate": 6.81331093478643e-06, + "loss": 0.2115, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11203019320964813, + "step": 7280, + "valid_targets_mean": 4506.8, + "valid_targets_min": 1765 + }, + { + "epoch": 5.299017824663514, + "grad_norm": 0.6608593155621189, + "learning_rate": 6.786064036991426e-06, + "loss": 0.1966, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09521438181400299, + "step": 7285, + "valid_targets_mean": 2991.6, + "valid_targets_min": 2087 + }, + { + "epoch": 5.302655511094944, + "grad_norm": 0.595320565466497, + "learning_rate": 6.758860593768159e-06, + "loss": 0.1984, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.092640221118927, + "step": 7290, + "valid_targets_mean": 3994.1, + "valid_targets_min": 1603 + }, + { + "epoch": 5.306293197526373, + "grad_norm": 0.6243356135565736, + "learning_rate": 6.73170069457626e-06, + "loss": 0.2043, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10775718092918396, + "step": 7295, + "valid_targets_mean": 4391.9, + "valid_targets_min": 2092 + }, + { + "epoch": 5.3099308839578025, + "grad_norm": 0.5890167414911136, + "learning_rate": 6.704584428732184e-06, + "loss": 0.2009, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10066942870616913, + "step": 7300, + "valid_targets_mean": 3808.9, + "valid_targets_min": 1410 + }, + { + "epoch": 5.313568570389233, + "grad_norm": 0.6157049681879007, + "learning_rate": 6.677511885408874e-06, + "loss": 0.1916, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10587199032306671, + "step": 7305, + "valid_targets_mean": 5078.5, + "valid_targets_min": 2110 + }, + { + "epoch": 5.317206256820662, + "grad_norm": 0.6525732678894479, + "learning_rate": 6.65048315363551e-06, + "loss": 0.205, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09593355655670166, + "step": 7310, + "valid_targets_mean": 3447.4, + "valid_targets_min": 2420 + }, + { + "epoch": 5.3208439432520915, + "grad_norm": 0.6093565499282292, + "learning_rate": 6.623498322297192e-06, + "loss": 0.1981, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08943301439285278, + "step": 7315, + "valid_targets_mean": 4371.1, + "valid_targets_min": 1876 + }, + { + "epoch": 5.324481629683521, + "grad_norm": 0.6499176847199017, + "learning_rate": 6.596557480134638e-06, + "loss": 0.1993, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11322768777608871, + "step": 7320, + "valid_targets_mean": 3858.2, + "valid_targets_min": 2605 + }, + { + "epoch": 5.328119316114951, + "grad_norm": 0.689634249818503, + "learning_rate": 6.569660715743924e-06, + "loss": 0.1985, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09122210741043091, + "step": 7325, + "valid_targets_mean": 3383.9, + "valid_targets_min": 1177 + }, + { + "epoch": 5.3317570025463805, + "grad_norm": 0.7070680568340506, + "learning_rate": 6.542808117576149e-06, + "loss": 0.2022, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10135532915592194, + "step": 7330, + "valid_targets_mean": 3455.5, + "valid_targets_min": 1714 + }, + { + "epoch": 5.33539468897781, + "grad_norm": 0.606322288235217, + "learning_rate": 6.515999773937211e-06, + "loss": 0.202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11189623177051544, + "step": 7335, + "valid_targets_mean": 5547.5, + "valid_targets_min": 2033 + }, + { + "epoch": 5.33903237540924, + "grad_norm": 0.6430189424467577, + "learning_rate": 6.4892357729874365e-06, + "loss": 0.1968, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10169069468975067, + "step": 7340, + "valid_targets_mean": 3485.4, + "valid_targets_min": 1924 + }, + { + "epoch": 5.3426700618406695, + "grad_norm": 0.7178960832986984, + "learning_rate": 6.46251620274134e-06, + "loss": 0.2034, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0954156294465065, + "step": 7345, + "valid_targets_mean": 3080.2, + "valid_targets_min": 2098 + }, + { + "epoch": 5.346307748272099, + "grad_norm": 0.7112308380495792, + "learning_rate": 6.435841151067328e-06, + "loss": 0.2048, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11390557885169983, + "step": 7350, + "valid_targets_mean": 3393.4, + "valid_targets_min": 1363 + }, + { + "epoch": 5.349945434703528, + "grad_norm": 0.679810184567914, + "learning_rate": 6.40921070568741e-06, + "loss": 0.197, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10935913771390915, + "step": 7355, + "valid_targets_mean": 4115.4, + "valid_targets_min": 1363 + }, + { + "epoch": 5.3535831211349585, + "grad_norm": 0.6373958915171256, + "learning_rate": 6.382624954176904e-06, + "loss": 0.2026, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1226830706000328, + "step": 7360, + "valid_targets_mean": 4313.8, + "valid_targets_min": 2620 + }, + { + "epoch": 5.357220807566388, + "grad_norm": 0.734139470757665, + "learning_rate": 6.3560839839641344e-06, + "loss": 0.2006, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10395762324333191, + "step": 7365, + "valid_targets_mean": 3425.1, + "valid_targets_min": 2316 + }, + { + "epoch": 5.360858493997817, + "grad_norm": 0.65007845025175, + "learning_rate": 6.3295878823301855e-06, + "loss": 0.1983, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09647583961486816, + "step": 7370, + "valid_targets_mean": 3744.0, + "valid_targets_min": 2022 + }, + { + "epoch": 5.364496180429247, + "grad_norm": 0.6537882708978889, + "learning_rate": 6.3031367364085665e-06, + "loss": 0.193, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06819622218608856, + "step": 7375, + "valid_targets_mean": 2445.6, + "valid_targets_min": 1294 + }, + { + "epoch": 5.368133866860677, + "grad_norm": 0.78442270582941, + "learning_rate": 6.27673063318496e-06, + "loss": 0.2029, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0969153344631195, + "step": 7380, + "valid_targets_mean": 3556.1, + "valid_targets_min": 1700 + }, + { + "epoch": 5.371771553292106, + "grad_norm": 0.6484691755690373, + "learning_rate": 6.250369659496935e-06, + "loss": 0.2011, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0846802294254303, + "step": 7385, + "valid_targets_mean": 2714.0, + "valid_targets_min": 1586 + }, + { + "epoch": 5.375409239723536, + "grad_norm": 0.7507922534040055, + "learning_rate": 6.22405390203362e-06, + "loss": 0.1994, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10590636730194092, + "step": 7390, + "valid_targets_mean": 3266.4, + "valid_targets_min": 1358 + }, + { + "epoch": 5.379046926154966, + "grad_norm": 0.6397880562687964, + "learning_rate": 6.197783447335479e-06, + "loss": 0.2019, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09731410443782806, + "step": 7395, + "valid_targets_mean": 3782.9, + "valid_targets_min": 2290 + }, + { + "epoch": 5.382684612586395, + "grad_norm": 0.6987727770369699, + "learning_rate": 6.1715583817939714e-06, + "loss": 0.2092, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13707704842090607, + "step": 7400, + "valid_targets_mean": 4904.8, + "valid_targets_min": 2189 + }, + { + "epoch": 5.3863222990178246, + "grad_norm": 0.6626204394177894, + "learning_rate": 6.1453787916513085e-06, + "loss": 0.1989, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12689638137817383, + "step": 7405, + "valid_targets_mean": 3959.5, + "valid_targets_min": 2121 + }, + { + "epoch": 5.389959985449254, + "grad_norm": 0.6460976646944651, + "learning_rate": 6.119244763000156e-06, + "loss": 0.2, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10010427236557007, + "step": 7410, + "valid_targets_mean": 3689.4, + "valid_targets_min": 2011 + }, + { + "epoch": 5.393597671880684, + "grad_norm": 0.6083798156522772, + "learning_rate": 6.093156381783327e-06, + "loss": 0.194, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1083628386259079, + "step": 7415, + "valid_targets_mean": 5082.1, + "valid_targets_min": 2674 + }, + { + "epoch": 5.3972353583121135, + "grad_norm": 0.6968052497822441, + "learning_rate": 6.067113733793544e-06, + "loss": 0.1886, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1099071204662323, + "step": 7420, + "valid_targets_mean": 3660.9, + "valid_targets_min": 2266 + }, + { + "epoch": 5.400873044743543, + "grad_norm": 0.6329565659566032, + "learning_rate": 6.041116904673125e-06, + "loss": 0.1945, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08984982967376709, + "step": 7425, + "valid_targets_mean": 4285.5, + "valid_targets_min": 904 + }, + { + "epoch": 5.404510731174973, + "grad_norm": 0.6486772351007111, + "learning_rate": 6.01516597991372e-06, + "loss": 0.1954, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1152547299861908, + "step": 7430, + "valid_targets_mean": 3999.4, + "valid_targets_min": 1194 + }, + { + "epoch": 5.4081484176064025, + "grad_norm": 0.6235843792890207, + "learning_rate": 5.989261044856003e-06, + "loss": 0.1998, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09004076570272446, + "step": 7435, + "valid_targets_mean": 4011.8, + "valid_targets_min": 1658 + }, + { + "epoch": 5.411786104037832, + "grad_norm": 0.7061562185631421, + "learning_rate": 5.963402184689411e-06, + "loss": 0.1979, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0813688188791275, + "step": 7440, + "valid_targets_mean": 2728.8, + "valid_targets_min": 1423 + }, + { + "epoch": 5.415423790469261, + "grad_norm": 0.6330919668944716, + "learning_rate": 5.93758948445188e-06, + "loss": 0.1996, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10701604187488556, + "step": 7445, + "valid_targets_mean": 4504.4, + "valid_targets_min": 2320 + }, + { + "epoch": 5.4190614769006915, + "grad_norm": 0.6538573345607033, + "learning_rate": 5.911823029029531e-06, + "loss": 0.2024, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07564770430326462, + "step": 7450, + "valid_targets_mean": 3272.5, + "valid_targets_min": 1811 + }, + { + "epoch": 5.422699163332121, + "grad_norm": 0.5900701169655054, + "learning_rate": 5.8861029031564185e-06, + "loss": 0.2086, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09610539674758911, + "step": 7455, + "valid_targets_mean": 4425.9, + "valid_targets_min": 3026 + }, + { + "epoch": 5.42633684976355, + "grad_norm": 0.6645444717726084, + "learning_rate": 5.860429191414232e-06, + "loss": 0.1975, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08783581852912903, + "step": 7460, + "valid_targets_mean": 2719.1, + "valid_targets_min": 644 + }, + { + "epoch": 5.42997453619498, + "grad_norm": 0.6095630651937676, + "learning_rate": 5.834801978232019e-06, + "loss": 0.1937, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09069765359163284, + "step": 7465, + "valid_targets_mean": 3369.0, + "valid_targets_min": 1397 + }, + { + "epoch": 5.43361222262641, + "grad_norm": 0.6100443614528803, + "learning_rate": 5.809221347885929e-06, + "loss": 0.1839, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09447897970676422, + "step": 7470, + "valid_targets_mean": 4122.6, + "valid_targets_min": 1415 + }, + { + "epoch": 5.437249909057839, + "grad_norm": 0.6480830284239277, + "learning_rate": 5.7836873844989195e-06, + "loss": 0.1859, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10131757706403732, + "step": 7475, + "valid_targets_mean": 4136.6, + "valid_targets_min": 621 + }, + { + "epoch": 5.440887595489269, + "grad_norm": 0.6680036129861192, + "learning_rate": 5.758200172040484e-06, + "loss": 0.1864, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07572194933891296, + "step": 7480, + "valid_targets_mean": 3065.8, + "valid_targets_min": 1290 + }, + { + "epoch": 5.444525281920699, + "grad_norm": 0.7414386210594317, + "learning_rate": 5.732759794326355e-06, + "loss": 0.1814, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08234970271587372, + "step": 7485, + "valid_targets_mean": 2426.9, + "valid_targets_min": 857 + }, + { + "epoch": 5.448162968352128, + "grad_norm": 0.7120787454501439, + "learning_rate": 5.707366335018267e-06, + "loss": 0.1833, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09746937453746796, + "step": 7490, + "valid_targets_mean": 2949.1, + "valid_targets_min": 720 + }, + { + "epoch": 5.451800654783558, + "grad_norm": 0.7007943086801927, + "learning_rate": 5.682019877623666e-06, + "loss": 0.1936, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1129353791475296, + "step": 7495, + "valid_targets_mean": 3579.1, + "valid_targets_min": 1801 + }, + { + "epoch": 5.455438341214987, + "grad_norm": 0.6819647725604047, + "learning_rate": 5.656720505495399e-06, + "loss": 0.2118, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09540560841560364, + "step": 7500, + "valid_targets_mean": 3532.6, + "valid_targets_min": 1510 + }, + { + "epoch": 5.459076027646417, + "grad_norm": 0.7064826109821033, + "learning_rate": 5.631468301831509e-06, + "loss": 0.185, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08841721713542938, + "step": 7505, + "valid_targets_mean": 2807.9, + "valid_targets_min": 987 + }, + { + "epoch": 5.462713714077847, + "grad_norm": 0.6340856936119367, + "learning_rate": 5.606263349674894e-06, + "loss": 0.1972, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08617214858531952, + "step": 7510, + "valid_targets_mean": 3937.4, + "valid_targets_min": 996 + }, + { + "epoch": 5.466351400509276, + "grad_norm": 0.6365287831615708, + "learning_rate": 5.581105731913079e-06, + "loss": 0.1893, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08234523981809616, + "step": 7515, + "valid_targets_mean": 3954.5, + "valid_targets_min": 2001 + }, + { + "epoch": 5.469989086940705, + "grad_norm": 0.5496923059037093, + "learning_rate": 5.555995531277931e-06, + "loss": 0.1907, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07375232875347137, + "step": 7520, + "valid_targets_mean": 5059.6, + "valid_targets_min": 1767 + }, + { + "epoch": 5.4736267733721355, + "grad_norm": 0.5585585825651032, + "learning_rate": 5.530932830345368e-06, + "loss": 0.1928, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08105695247650146, + "step": 7525, + "valid_targets_mean": 4714.0, + "valid_targets_min": 1511 + }, + { + "epoch": 5.477264459803565, + "grad_norm": 0.7962125486388508, + "learning_rate": 5.50591771153512e-06, + "loss": 0.1881, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09318206459283829, + "step": 7530, + "valid_targets_mean": 2861.0, + "valid_targets_min": 974 + }, + { + "epoch": 5.480902146234994, + "grad_norm": 0.6281053409358629, + "learning_rate": 5.480950257110424e-06, + "loss": 0.1928, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08312389254570007, + "step": 7535, + "valid_targets_mean": 3792.6, + "valid_targets_min": 2339 + }, + { + "epoch": 5.4845398326664245, + "grad_norm": 0.6753220569079377, + "learning_rate": 5.4560305491777885e-06, + "loss": 0.1878, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08557170629501343, + "step": 7540, + "valid_targets_mean": 3333.1, + "valid_targets_min": 1249 + }, + { + "epoch": 5.488177519097854, + "grad_norm": 0.652086098519019, + "learning_rate": 5.431158669686696e-06, + "loss": 0.1969, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09584107249975204, + "step": 7545, + "valid_targets_mean": 3493.1, + "valid_targets_min": 1223 + }, + { + "epoch": 5.491815205529283, + "grad_norm": 0.7222541083308428, + "learning_rate": 5.4063347004293495e-06, + "loss": 0.2052, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10472215712070465, + "step": 7550, + "valid_targets_mean": 3285.8, + "valid_targets_min": 1318 + }, + { + "epoch": 5.495452891960713, + "grad_norm": 0.7255666782856913, + "learning_rate": 5.381558723040388e-06, + "loss": 0.2003, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09062521159648895, + "step": 7555, + "valid_targets_mean": 2070.5, + "valid_targets_min": 1002 + }, + { + "epoch": 5.499090578392143, + "grad_norm": 0.6617649536515426, + "learning_rate": 5.356830818996621e-06, + "loss": 0.2129, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09208090603351593, + "step": 7560, + "valid_targets_mean": 3328.1, + "valid_targets_min": 845 + }, + { + "epoch": 5.502728264823572, + "grad_norm": 0.6005022656194463, + "learning_rate": 5.332151069616797e-06, + "loss": 0.19, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10707007348537445, + "step": 7565, + "valid_targets_mean": 4657.4, + "valid_targets_min": 2747 + }, + { + "epoch": 5.506365951255002, + "grad_norm": 0.6945622159260968, + "learning_rate": 5.307519556061276e-06, + "loss": 0.2089, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10668567568063736, + "step": 7570, + "valid_targets_mean": 4166.4, + "valid_targets_min": 1241 + }, + { + "epoch": 5.510003637686431, + "grad_norm": 0.7346479896038832, + "learning_rate": 5.2829363593318075e-06, + "loss": 0.2014, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11057966202497482, + "step": 7575, + "valid_targets_mean": 3567.1, + "valid_targets_min": 992 + }, + { + "epoch": 5.513641324117861, + "grad_norm": 0.673027758635986, + "learning_rate": 5.258401560271238e-06, + "loss": 0.1982, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09393076598644257, + "step": 7580, + "valid_targets_mean": 3343.5, + "valid_targets_min": 1211 + }, + { + "epoch": 5.517279010549291, + "grad_norm": 0.7659727384606678, + "learning_rate": 5.2339152395632675e-06, + "loss": 0.2045, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11002011597156525, + "step": 7585, + "valid_targets_mean": 3081.6, + "valid_targets_min": 1889 + }, + { + "epoch": 5.52091669698072, + "grad_norm": 0.8406753345749028, + "learning_rate": 5.209477477732172e-06, + "loss": 0.1963, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08643898367881775, + "step": 7590, + "valid_targets_mean": 3233.9, + "valid_targets_min": 1624 + }, + { + "epoch": 5.52455438341215, + "grad_norm": 0.6960850392592949, + "learning_rate": 5.185088355142529e-06, + "loss": 0.2016, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1028476357460022, + "step": 7595, + "valid_targets_mean": 3817.4, + "valid_targets_min": 2763 + }, + { + "epoch": 5.52819206984358, + "grad_norm": 0.6932385181524168, + "learning_rate": 5.160747951998979e-06, + "loss": 0.2006, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09567970037460327, + "step": 7600, + "valid_targets_mean": 3037.0, + "valid_targets_min": 1928 + }, + { + "epoch": 5.531829756275009, + "grad_norm": 0.6511038215100928, + "learning_rate": 5.136456348345926e-06, + "loss": 0.1857, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09618302434682846, + "step": 7605, + "valid_targets_mean": 4733.2, + "valid_targets_min": 3147 + }, + { + "epoch": 5.535467442706439, + "grad_norm": 0.6290286342692134, + "learning_rate": 5.112213624067317e-06, + "loss": 0.1897, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08744131028652191, + "step": 7610, + "valid_targets_mean": 3348.4, + "valid_targets_min": 1844 + }, + { + "epoch": 5.539105129137869, + "grad_norm": 0.6863918883236108, + "learning_rate": 5.088019858886353e-06, + "loss": 0.1936, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10419254750013351, + "step": 7615, + "valid_targets_mean": 4360.2, + "valid_targets_min": 2330 + }, + { + "epoch": 5.542742815569298, + "grad_norm": 0.74853123177981, + "learning_rate": 5.06387513236521e-06, + "loss": 0.1982, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11319929361343384, + "step": 7620, + "valid_targets_mean": 3574.1, + "valid_targets_min": 900 + }, + { + "epoch": 5.546380502000727, + "grad_norm": 0.5915471181731582, + "learning_rate": 5.039779523904824e-06, + "loss": 0.1907, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09971155226230621, + "step": 7625, + "valid_targets_mean": 5092.8, + "valid_targets_min": 1713 + }, + { + "epoch": 5.550018188432157, + "grad_norm": 0.6763007591682828, + "learning_rate": 5.015733112744592e-06, + "loss": 0.2, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11057909578084946, + "step": 7630, + "valid_targets_mean": 4057.1, + "valid_targets_min": 981 + }, + { + "epoch": 5.553655874863587, + "grad_norm": 0.6464023589807169, + "learning_rate": 4.991735977962131e-06, + "loss": 0.1976, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11272943019866943, + "step": 7635, + "valid_targets_mean": 4127.9, + "valid_targets_min": 2098 + }, + { + "epoch": 5.557293561295016, + "grad_norm": 0.6676038113335158, + "learning_rate": 4.967788198473e-06, + "loss": 0.1917, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08872783929109573, + "step": 7640, + "valid_targets_mean": 3101.1, + "valid_targets_min": 1107 + }, + { + "epoch": 5.560931247726446, + "grad_norm": 0.6303830841186225, + "learning_rate": 4.943889853030452e-06, + "loss": 0.1944, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09278421103954315, + "step": 7645, + "valid_targets_mean": 3718.0, + "valid_targets_min": 2273 + }, + { + "epoch": 5.564568934157876, + "grad_norm": 0.7195279017920398, + "learning_rate": 4.920041020225186e-06, + "loss": 0.192, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08658171445131302, + "step": 7650, + "valid_targets_mean": 3316.0, + "valid_targets_min": 1920 + }, + { + "epoch": 5.568206620589305, + "grad_norm": 0.649190022452458, + "learning_rate": 4.896241778485065e-06, + "loss": 0.1786, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09506304562091827, + "step": 7655, + "valid_targets_mean": 4710.8, + "valid_targets_min": 1687 + }, + { + "epoch": 5.571844307020735, + "grad_norm": 0.6715441463581702, + "learning_rate": 4.872492206074881e-06, + "loss": 0.1871, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0819762796163559, + "step": 7660, + "valid_targets_mean": 3075.8, + "valid_targets_min": 1375 + }, + { + "epoch": 5.575481993452165, + "grad_norm": 0.6760220524523347, + "learning_rate": 4.84879238109607e-06, + "loss": 0.1896, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08513595163822174, + "step": 7665, + "valid_targets_mean": 3776.4, + "valid_targets_min": 906 + }, + { + "epoch": 5.579119679883594, + "grad_norm": 0.6759828318245477, + "learning_rate": 4.825142381486492e-06, + "loss": 0.1843, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09290358424186707, + "step": 7670, + "valid_targets_mean": 3393.9, + "valid_targets_min": 1874 + }, + { + "epoch": 5.582757366315024, + "grad_norm": 0.6248740470757713, + "learning_rate": 4.8015422850201285e-06, + "loss": 0.1775, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09320896863937378, + "step": 7675, + "valid_targets_mean": 3503.4, + "valid_targets_min": 1877 + }, + { + "epoch": 5.586395052746453, + "grad_norm": 0.6465137803416858, + "learning_rate": 4.7779921693068755e-06, + "loss": 0.1819, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08860233426094055, + "step": 7680, + "valid_targets_mean": 3744.4, + "valid_targets_min": 1697 + }, + { + "epoch": 5.590032739177883, + "grad_norm": 0.7064618738887267, + "learning_rate": 4.754492111792262e-06, + "loss": 0.1867, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09706763923168182, + "step": 7685, + "valid_targets_mean": 3015.1, + "valid_targets_min": 1679 + }, + { + "epoch": 5.593670425609313, + "grad_norm": 0.7248734116207157, + "learning_rate": 4.731042189757182e-06, + "loss": 0.1988, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09963656216859818, + "step": 7690, + "valid_targets_mean": 3292.0, + "valid_targets_min": 1695 + }, + { + "epoch": 5.597308112040742, + "grad_norm": 0.6612968724104933, + "learning_rate": 4.707642480317676e-06, + "loss": 0.1854, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08131450414657593, + "step": 7695, + "valid_targets_mean": 3148.8, + "valid_targets_min": 2114 + }, + { + "epoch": 5.600945798472171, + "grad_norm": 0.6784825958822017, + "learning_rate": 4.684293060424654e-06, + "loss": 0.1839, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0748187005519867, + "step": 7700, + "valid_targets_mean": 2852.0, + "valid_targets_min": 993 + }, + { + "epoch": 5.604583484903602, + "grad_norm": 0.6815836238398627, + "learning_rate": 4.660994006863631e-06, + "loss": 0.1831, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0888112336397171, + "step": 7705, + "valid_targets_mean": 3059.8, + "valid_targets_min": 1275 + }, + { + "epoch": 5.608221171335031, + "grad_norm": 0.6213740785845884, + "learning_rate": 4.637745396254516e-06, + "loss": 0.186, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09944784641265869, + "step": 7710, + "valid_targets_mean": 4479.0, + "valid_targets_min": 2250 + }, + { + "epoch": 5.61185885776646, + "grad_norm": 0.6640169508190984, + "learning_rate": 4.614547305051307e-06, + "loss": 0.1905, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11617793887853622, + "step": 7715, + "valid_targets_mean": 4952.2, + "valid_targets_min": 2892 + }, + { + "epoch": 5.615496544197891, + "grad_norm": 0.6706258314679535, + "learning_rate": 4.59139980954189e-06, + "loss": 0.1934, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.092521071434021, + "step": 7720, + "valid_targets_mean": 4026.4, + "valid_targets_min": 2935 + }, + { + "epoch": 5.61913423062932, + "grad_norm": 0.6755903012163939, + "learning_rate": 4.568302985847755e-06, + "loss": 0.1881, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08792385458946228, + "step": 7725, + "valid_targets_mean": 4085.2, + "valid_targets_min": 1817 + }, + { + "epoch": 5.622771917060749, + "grad_norm": 0.6570166002332302, + "learning_rate": 4.545256909923761e-06, + "loss": 0.1871, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09962716698646545, + "step": 7730, + "valid_targets_mean": 4050.4, + "valid_targets_min": 1727 + }, + { + "epoch": 5.626409603492179, + "grad_norm": 0.5875639897493753, + "learning_rate": 4.5222616575578695e-06, + "loss": 0.1853, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07115355134010315, + "step": 7735, + "valid_targets_mean": 3682.9, + "valid_targets_min": 1622 + }, + { + "epoch": 5.630047289923609, + "grad_norm": 0.5032722516750366, + "learning_rate": 4.499317304370909e-06, + "loss": 0.1508, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06456760317087173, + "step": 7740, + "valid_targets_mean": 2920.6, + "valid_targets_min": 1521 + }, + { + "epoch": 5.633684976355038, + "grad_norm": 0.41496694067666334, + "learning_rate": 4.476423925816333e-06, + "loss": 0.0916, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03742452338337898, + "step": 7745, + "valid_targets_mean": 3724.9, + "valid_targets_min": 2590 + }, + { + "epoch": 5.637322662786468, + "grad_norm": 0.43004716031142837, + "learning_rate": 4.4535815971799615e-06, + "loss": 0.0758, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03913837671279907, + "step": 7750, + "valid_targets_mean": 3167.9, + "valid_targets_min": 982 + }, + { + "epoch": 5.640960349217897, + "grad_norm": 0.4741494465111177, + "learning_rate": 4.430790393579733e-06, + "loss": 0.0823, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08193787932395935, + "step": 7755, + "valid_targets_mean": 2996.4, + "valid_targets_min": 927 + }, + { + "epoch": 5.644598035649327, + "grad_norm": 0.4208086870506883, + "learning_rate": 4.408050389965455e-06, + "loss": 0.0738, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.027150560170412064, + "step": 7760, + "valid_targets_mean": 1756.1, + "valid_targets_min": 641 + }, + { + "epoch": 5.648235722080757, + "grad_norm": 0.4792393651195527, + "learning_rate": 4.3853616611185526e-06, + "loss": 0.0772, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04014158621430397, + "step": 7765, + "valid_targets_mean": 2745.2, + "valid_targets_min": 859 + }, + { + "epoch": 5.651873408512186, + "grad_norm": 0.8639240715461735, + "learning_rate": 4.362724281651862e-06, + "loss": 0.1169, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05837690830230713, + "step": 7770, + "valid_targets_mean": 1268.5, + "valid_targets_min": 539 + }, + { + "epoch": 5.655511094943616, + "grad_norm": 0.43528227023563365, + "learning_rate": 4.340138326009326e-06, + "loss": 0.0918, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05404946953058243, + "step": 7775, + "valid_targets_mean": 2812.6, + "valid_targets_min": 920 + }, + { + "epoch": 5.659148781375046, + "grad_norm": 0.47217697124499947, + "learning_rate": 4.317603868465794e-06, + "loss": 0.0845, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04584541544318199, + "step": 7780, + "valid_targets_mean": 3495.0, + "valid_targets_min": 2295 + }, + { + "epoch": 5.662786467806475, + "grad_norm": 0.3839726101224589, + "learning_rate": 4.29512098312675e-06, + "loss": 0.112, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030235670506954193, + "step": 7785, + "valid_targets_mean": 3228.4, + "valid_targets_min": 738 + }, + { + "epoch": 5.666424154237904, + "grad_norm": 0.4181088979108228, + "learning_rate": 4.272689743928087e-06, + "loss": 0.0625, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04656219854950905, + "step": 7790, + "valid_targets_mean": 4290.0, + "valid_targets_min": 2706 + }, + { + "epoch": 5.670061840669335, + "grad_norm": 0.4233758801986377, + "learning_rate": 4.250310224635867e-06, + "loss": 0.0745, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0462779700756073, + "step": 7795, + "valid_targets_mean": 4529.8, + "valid_targets_min": 3259 + }, + { + "epoch": 5.673699527100764, + "grad_norm": 0.38828530542019923, + "learning_rate": 4.2279824988460485e-06, + "loss": 0.0731, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.033743955194950104, + "step": 7800, + "valid_targets_mean": 4198.2, + "valid_targets_min": 518 + }, + { + "epoch": 5.677337213532193, + "grad_norm": 0.6149112307699157, + "learning_rate": 4.205706639984284e-06, + "loss": 0.1149, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.043449513614177704, + "step": 7805, + "valid_targets_mean": 1873.5, + "valid_targets_min": 664 + }, + { + "epoch": 5.680974899963623, + "grad_norm": 0.4001824223536499, + "learning_rate": 4.183482721305641e-06, + "loss": 0.0673, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03490995988249779, + "step": 7810, + "valid_targets_mean": 3589.2, + "valid_targets_min": 2724 + }, + { + "epoch": 5.684612586395053, + "grad_norm": 0.49368264793662886, + "learning_rate": 4.161310815894397e-06, + "loss": 0.075, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.032060377299785614, + "step": 7815, + "valid_targets_mean": 1644.0, + "valid_targets_min": 518 + }, + { + "epoch": 5.688250272826482, + "grad_norm": 0.3986701661357499, + "learning_rate": 4.13919099666378e-06, + "loss": 0.0657, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03217446804046631, + "step": 7820, + "valid_targets_mean": 3376.5, + "valid_targets_min": 844 + }, + { + "epoch": 5.691887959257912, + "grad_norm": 0.4999272154372873, + "learning_rate": 4.1171233363557105e-06, + "loss": 0.0753, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03235863521695137, + "step": 7825, + "valid_targets_mean": 2660.2, + "valid_targets_min": 577 + }, + { + "epoch": 5.695525645689342, + "grad_norm": 0.4131563068524426, + "learning_rate": 4.095107907540612e-06, + "loss": 0.0783, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.036255016922950745, + "step": 7830, + "valid_targets_mean": 2450.9, + "valid_targets_min": 1081 + }, + { + "epoch": 5.699163332120771, + "grad_norm": 0.476513283187298, + "learning_rate": 4.073144782617111e-06, + "loss": 0.0849, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.035156235098838806, + "step": 7835, + "valid_targets_mean": 2564.8, + "valid_targets_min": 721 + }, + { + "epoch": 5.702801018552201, + "grad_norm": 0.4281603798655787, + "learning_rate": 4.051234033811864e-06, + "loss": 0.0819, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03701521456241608, + "step": 7840, + "valid_targets_mean": 2766.4, + "valid_targets_min": 615 + }, + { + "epoch": 5.70643870498363, + "grad_norm": 0.5926663410094402, + "learning_rate": 4.029375733179259e-06, + "loss": 0.124, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11648400127887726, + "step": 7845, + "valid_targets_mean": 3693.1, + "valid_targets_min": 2609 + }, + { + "epoch": 5.71007639141506, + "grad_norm": 0.3285021600364466, + "learning_rate": 4.007569952601222e-06, + "loss": 0.069, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.025256458669900894, + "step": 7850, + "valid_targets_mean": 4102.4, + "valid_targets_min": 3732 + }, + { + "epoch": 5.71371407784649, + "grad_norm": 0.466632980274218, + "learning_rate": 3.985816763786952e-06, + "loss": 0.0761, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029616517946124077, + "step": 7855, + "valid_targets_mean": 1369.8, + "valid_targets_min": 731 + }, + { + "epoch": 5.717351764277919, + "grad_norm": 0.4576231949860897, + "learning_rate": 3.964116238272708e-06, + "loss": 0.1047, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03944217413663864, + "step": 7860, + "valid_targets_mean": 3517.8, + "valid_targets_min": 1812 + }, + { + "epoch": 5.720989450709348, + "grad_norm": 0.404769244759034, + "learning_rate": 3.942468447421568e-06, + "loss": 0.0896, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.035283416509628296, + "step": 7865, + "valid_targets_mean": 4099.4, + "valid_targets_min": 3344 + }, + { + "epoch": 5.724627137140779, + "grad_norm": 0.7659647156219037, + "learning_rate": 3.920873462423167e-06, + "loss": 0.0991, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12065369635820389, + "step": 7870, + "valid_targets_mean": 2014.0, + "valid_targets_min": 568 + }, + { + "epoch": 5.728264823572208, + "grad_norm": 0.5521355621408831, + "learning_rate": 3.899331354293518e-06, + "loss": 0.2541, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04868510365486145, + "step": 7875, + "valid_targets_mean": 3413.2, + "valid_targets_min": 1248 + }, + { + "epoch": 5.731902510003637, + "grad_norm": 0.4713736324406221, + "learning_rate": 3.877842193874714e-06, + "loss": 0.0845, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.051058780401945114, + "step": 7880, + "valid_targets_mean": 3592.5, + "valid_targets_min": 2829 + }, + { + "epoch": 5.735540196435068, + "grad_norm": 0.538651586155841, + "learning_rate": 3.856406051834751e-06, + "loss": 0.0821, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05149976909160614, + "step": 7885, + "valid_targets_mean": 2591.2, + "valid_targets_min": 798 + }, + { + "epoch": 5.739177882866497, + "grad_norm": 0.44248020376656416, + "learning_rate": 3.835022998667266e-06, + "loss": 0.172, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03819961100816727, + "step": 7890, + "valid_targets_mean": 3274.5, + "valid_targets_min": 2594 + }, + { + "epoch": 5.742815569297926, + "grad_norm": 0.26717981233794824, + "learning_rate": 3.8136931046912983e-06, + "loss": 0.0642, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.023230068385601044, + "step": 7895, + "valid_targets_mean": 5338.6, + "valid_targets_min": 3239 + }, + { + "epoch": 5.746453255729357, + "grad_norm": 0.3668670833582055, + "learning_rate": 3.7924164400510945e-06, + "loss": 0.0869, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.024739600718021393, + "step": 7900, + "valid_targets_mean": 3009.1, + "valid_targets_min": 569 + }, + { + "epoch": 5.750090942160786, + "grad_norm": 0.4166403612979877, + "learning_rate": 3.7711930747158242e-06, + "loss": 0.0818, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.032855406403541565, + "step": 7905, + "valid_targets_mean": 3551.1, + "valid_targets_min": 846 + }, + { + "epoch": 5.753728628592215, + "grad_norm": 0.47423554432542825, + "learning_rate": 3.7500230784794145e-06, + "loss": 0.0718, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04294516146183014, + "step": 7910, + "valid_targets_mean": 3529.1, + "valid_targets_min": 1352 + }, + { + "epoch": 5.757366315023645, + "grad_norm": 0.46274294324865534, + "learning_rate": 3.7289065209602625e-06, + "loss": 0.0691, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03576917201280594, + "step": 7915, + "valid_targets_mean": 2796.4, + "valid_targets_min": 536 + }, + { + "epoch": 5.761004001455074, + "grad_norm": 0.3974097702853129, + "learning_rate": 3.7078434716010293e-06, + "loss": 0.0695, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.033707186579704285, + "step": 7920, + "valid_targets_mean": 3629.5, + "valid_targets_min": 2264 + }, + { + "epoch": 5.764641687886504, + "grad_norm": 0.4041831840955815, + "learning_rate": 3.6868339996684244e-06, + "loss": 0.0676, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03163086622953415, + "step": 7925, + "valid_targets_mean": 3847.5, + "valid_targets_min": 1865 + }, + { + "epoch": 5.768279374317934, + "grad_norm": 0.7463553551164961, + "learning_rate": 3.665878174252957e-06, + "loss": 0.1052, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18211999535560608, + "step": 7930, + "valid_targets_mean": 2020.4, + "valid_targets_min": 732 + }, + { + "epoch": 5.771917060749363, + "grad_norm": 0.4352069849163835, + "learning_rate": 3.6449760642687303e-06, + "loss": 0.0736, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03807666152715683, + "step": 7935, + "valid_targets_mean": 3760.4, + "valid_targets_min": 2719 + }, + { + "epoch": 5.775554747180793, + "grad_norm": 0.5457658355989019, + "learning_rate": 3.624127738453185e-06, + "loss": 0.0604, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04009000211954117, + "step": 7940, + "valid_targets_mean": 1354.6, + "valid_targets_min": 722 + }, + { + "epoch": 5.779192433612223, + "grad_norm": 0.3764802895171885, + "learning_rate": 3.6033332653668908e-06, + "loss": 0.0638, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02440740540623665, + "step": 7945, + "valid_targets_mean": 3962.9, + "valid_targets_min": 3167 + }, + { + "epoch": 5.782830120043652, + "grad_norm": 0.40638979351343907, + "learning_rate": 3.5825927133933358e-06, + "loss": 0.0634, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02310185134410858, + "step": 7950, + "valid_targets_mean": 2448.9, + "valid_targets_min": 802 + }, + { + "epoch": 5.786467806475082, + "grad_norm": 0.4695607082258077, + "learning_rate": 3.5619061507386765e-06, + "loss": 0.1087, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03632335364818573, + "step": 7955, + "valid_targets_mean": 3174.1, + "valid_targets_min": 908 + }, + { + "epoch": 5.790105492906512, + "grad_norm": 0.4181431509309617, + "learning_rate": 3.541273645431529e-06, + "loss": 0.0879, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05490988865494728, + "step": 7960, + "valid_targets_mean": 3718.0, + "valid_targets_min": 3104 + }, + { + "epoch": 5.793743179337941, + "grad_norm": 0.326640813935313, + "learning_rate": 3.520695265322727e-06, + "loss": 0.0843, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.020301470533013344, + "step": 7965, + "valid_targets_mean": 2740.0, + "valid_targets_min": 598 + }, + { + "epoch": 5.79738086576937, + "grad_norm": 1.033357776660184, + "learning_rate": 3.5001710780851284e-06, + "loss": 0.0915, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05413468927145004, + "step": 7970, + "valid_targets_mean": 819.4, + "valid_targets_min": 578 + }, + { + "epoch": 5.801018552200801, + "grad_norm": 0.6508212440941225, + "learning_rate": 3.4797011512133595e-06, + "loss": 0.1218, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04479039087891579, + "step": 7975, + "valid_targets_mean": 1079.6, + "valid_targets_min": 695 + }, + { + "epoch": 5.80465623863223, + "grad_norm": 0.48497919757016694, + "learning_rate": 3.4592855520236167e-06, + "loss": 0.0788, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.042157914489507675, + "step": 7980, + "valid_targets_mean": 2324.0, + "valid_targets_min": 845 + }, + { + "epoch": 5.808293925063659, + "grad_norm": 0.43238328661482794, + "learning_rate": 3.438924347653445e-06, + "loss": 0.0929, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.032174251973629, + "step": 7985, + "valid_targets_mean": 3022.2, + "valid_targets_min": 728 + }, + { + "epoch": 5.811931611495089, + "grad_norm": 0.653636370102615, + "learning_rate": 3.4186176050614937e-06, + "loss": 0.0744, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04990832507610321, + "step": 7990, + "valid_targets_mean": 1847.6, + "valid_targets_min": 820 + }, + { + "epoch": 5.815569297926519, + "grad_norm": 0.3600598356823773, + "learning_rate": 3.398365391027323e-06, + "loss": 0.0719, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03300601989030838, + "step": 7995, + "valid_targets_mean": 4479.2, + "valid_targets_min": 4044 + }, + { + "epoch": 5.819206984357948, + "grad_norm": 0.4245166503019682, + "learning_rate": 3.3781677721511773e-06, + "loss": 0.0817, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02435436099767685, + "step": 8000, + "valid_targets_mean": 4048.2, + "valid_targets_min": 580 + }, + { + "epoch": 5.822844670789378, + "grad_norm": 0.44659780707167784, + "learning_rate": 3.3580248148537485e-06, + "loss": 0.0673, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.033053670078516006, + "step": 8005, + "valid_targets_mean": 2770.8, + "valid_targets_min": 459 + }, + { + "epoch": 5.826482357220808, + "grad_norm": 0.42773027596696245, + "learning_rate": 3.3379365853759872e-06, + "loss": 0.0644, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04387069493532181, + "step": 8010, + "valid_targets_mean": 4135.4, + "valid_targets_min": 2558 + }, + { + "epoch": 5.830120043652237, + "grad_norm": 0.48052064227278946, + "learning_rate": 3.317903149778854e-06, + "loss": 0.0746, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05994393303990364, + "step": 8015, + "valid_targets_mean": 2926.4, + "valid_targets_min": 791 + }, + { + "epoch": 5.833757730083667, + "grad_norm": 0.4884794868651795, + "learning_rate": 3.2979245739431275e-06, + "loss": 0.0841, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04918165132403374, + "step": 8020, + "valid_targets_mean": 3447.4, + "valid_targets_min": 1165 + }, + { + "epoch": 5.837395416515096, + "grad_norm": 0.4204770950528825, + "learning_rate": 3.27800092356918e-06, + "loss": 0.0844, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04150998964905739, + "step": 8025, + "valid_targets_mean": 3704.6, + "valid_targets_min": 2901 + }, + { + "epoch": 5.841033102946526, + "grad_norm": 0.38726550573286683, + "learning_rate": 3.2581322641767588e-06, + "loss": 0.0714, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.026574641466140747, + "step": 8030, + "valid_targets_mean": 3872.1, + "valid_targets_min": 2674 + }, + { + "epoch": 5.844670789377956, + "grad_norm": 0.5320905912916885, + "learning_rate": 3.23831866110476e-06, + "loss": 0.0734, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04328707605600357, + "step": 8035, + "valid_targets_mean": 2819.6, + "valid_targets_min": 502 + }, + { + "epoch": 5.848308475809385, + "grad_norm": 0.42503092739070647, + "learning_rate": 3.21856017951103e-06, + "loss": 0.0803, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.023754674941301346, + "step": 8040, + "valid_targets_mean": 2963.4, + "valid_targets_min": 724 + }, + { + "epoch": 5.851946162240814, + "grad_norm": 0.3536869106215904, + "learning_rate": 3.19885688437215e-06, + "loss": 0.0599, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.027006492018699646, + "step": 8045, + "valid_targets_mean": 3650.0, + "valid_targets_min": 2769 + }, + { + "epoch": 5.855583848672245, + "grad_norm": 0.5506342555153401, + "learning_rate": 3.179208840483219e-06, + "loss": 0.0777, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03743523359298706, + "step": 8050, + "valid_targets_mean": 3052.1, + "valid_targets_min": 591 + }, + { + "epoch": 5.859221535103674, + "grad_norm": 0.4483039983639775, + "learning_rate": 3.15961611245764e-06, + "loss": 0.0757, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03647010028362274, + "step": 8055, + "valid_targets_mean": 2622.5, + "valid_targets_min": 387 + }, + { + "epoch": 5.862859221535103, + "grad_norm": 0.506369534585483, + "learning_rate": 3.1400787647268947e-06, + "loss": 0.0828, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04700222611427307, + "step": 8060, + "valid_targets_mean": 3589.4, + "valid_targets_min": 1648 + }, + { + "epoch": 5.866496907966534, + "grad_norm": 0.4653494538763903, + "learning_rate": 3.1205968615403615e-06, + "loss": 0.0784, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.031636111438274384, + "step": 8065, + "valid_targets_mean": 1455.0, + "valid_targets_min": 457 + }, + { + "epoch": 5.870134594397963, + "grad_norm": 0.8312464741477619, + "learning_rate": 3.1011704669650842e-06, + "loss": 0.0817, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0483405664563179, + "step": 8070, + "valid_targets_mean": 878.2, + "valid_targets_min": 482 + }, + { + "epoch": 5.873772280829392, + "grad_norm": 0.5811644324124312, + "learning_rate": 3.0817996448855482e-06, + "loss": 0.0752, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06588949263095856, + "step": 8075, + "valid_targets_mean": 2136.0, + "valid_targets_min": 808 + }, + { + "epoch": 5.877409967260822, + "grad_norm": 0.4673173485334166, + "learning_rate": 3.0624844590035096e-06, + "loss": 0.0825, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03072255663573742, + "step": 8080, + "valid_targets_mean": 2712.4, + "valid_targets_min": 1010 + }, + { + "epoch": 5.881047653692252, + "grad_norm": 0.4934640833339208, + "learning_rate": 3.043224972837744e-06, + "loss": 0.0772, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.042441628873348236, + "step": 8085, + "valid_targets_mean": 2711.9, + "valid_targets_min": 902 + }, + { + "epoch": 5.884685340123681, + "grad_norm": 0.5734924462134942, + "learning_rate": 3.024021249723865e-06, + "loss": 0.0665, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.042159706354141235, + "step": 8090, + "valid_targets_mean": 1882.1, + "valid_targets_min": 526 + }, + { + "epoch": 5.888323026555111, + "grad_norm": 0.6087842514327518, + "learning_rate": 3.0048733528141106e-06, + "loss": 0.0851, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.035335127264261246, + "step": 8095, + "valid_targets_mean": 1127.9, + "valid_targets_min": 637 + }, + { + "epoch": 5.89196071298654, + "grad_norm": 0.40316362335221434, + "learning_rate": 2.9857813450771213e-06, + "loss": 0.0757, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02658000960946083, + "step": 8100, + "valid_targets_mean": 3718.1, + "valid_targets_min": 1440 + }, + { + "epoch": 5.89559839941797, + "grad_norm": 0.4064640829840919, + "learning_rate": 2.9667452892977543e-06, + "loss": 0.0675, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0477389432489872, + "step": 8105, + "valid_targets_mean": 4313.0, + "valid_targets_min": 2341 + }, + { + "epoch": 5.8992360858494, + "grad_norm": 0.3694707582180142, + "learning_rate": 2.9477652480768506e-06, + "loss": 0.0709, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0340762622654438, + "step": 8110, + "valid_targets_mean": 3635.4, + "valid_targets_min": 762 + }, + { + "epoch": 5.902873772280829, + "grad_norm": 0.42984045598268095, + "learning_rate": 2.928841283831072e-06, + "loss": 0.0654, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.022145047783851624, + "step": 8115, + "valid_targets_mean": 1935.6, + "valid_targets_min": 667 + }, + { + "epoch": 5.906511458712259, + "grad_norm": 0.41741621094769904, + "learning_rate": 2.9099734587926477e-06, + "loss": 0.0606, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04102378338575363, + "step": 8120, + "valid_targets_mean": 4719.8, + "valid_targets_min": 4077 + }, + { + "epoch": 5.910149145143689, + "grad_norm": 0.37129574188842035, + "learning_rate": 2.8911618350091887e-06, + "loss": 0.0821, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.035389505326747894, + "step": 8125, + "valid_targets_mean": 3550.5, + "valid_targets_min": 784 + }, + { + "epoch": 5.913786831575118, + "grad_norm": 0.5462550058863899, + "learning_rate": 2.8724064743434988e-06, + "loss": 0.0926, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06464455276727676, + "step": 8130, + "valid_targets_mean": 3344.2, + "valid_targets_min": 1090 + }, + { + "epoch": 5.9174245180065475, + "grad_norm": 0.4540763934523156, + "learning_rate": 2.853707438473352e-06, + "loss": 0.0711, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03962421417236328, + "step": 8135, + "valid_targets_mean": 3255.9, + "valid_targets_min": 1189 + }, + { + "epoch": 5.921062204437978, + "grad_norm": 0.34332738809501495, + "learning_rate": 2.8350647888913018e-06, + "loss": 0.0804, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.022147882729768753, + "step": 8140, + "valid_targets_mean": 2780.1, + "valid_targets_min": 929 + }, + { + "epoch": 5.924699890869407, + "grad_norm": 0.36396584108726715, + "learning_rate": 2.8164785869044585e-06, + "loss": 0.0972, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.025651387870311737, + "step": 8145, + "valid_targets_mean": 3147.2, + "valid_targets_min": 987 + }, + { + "epoch": 5.928337577300836, + "grad_norm": 0.4756439787398049, + "learning_rate": 2.7979488936343215e-06, + "loss": 0.0684, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04675167053937912, + "step": 8150, + "valid_targets_mean": 3708.8, + "valid_targets_min": 3019 + }, + { + "epoch": 5.931975263732266, + "grad_norm": 0.43456686556899643, + "learning_rate": 2.77947577001654e-06, + "loss": 0.0743, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.031170830130577087, + "step": 8155, + "valid_targets_mean": 3040.9, + "valid_targets_min": 1045 + }, + { + "epoch": 5.935612950163696, + "grad_norm": 0.4059296901967331, + "learning_rate": 2.7610592768007437e-06, + "loss": 0.0698, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04947167634963989, + "step": 8160, + "valid_targets_mean": 2372.4, + "valid_targets_min": 647 + }, + { + "epoch": 5.939250636595125, + "grad_norm": 0.5907359738979803, + "learning_rate": 2.742699474550332e-06, + "loss": 0.1129, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07571597397327423, + "step": 8165, + "valid_targets_mean": 1277.0, + "valid_targets_min": 570 + }, + { + "epoch": 5.942888323026555, + "grad_norm": 0.5353294559196025, + "learning_rate": 2.7243964236422614e-06, + "loss": 0.0887, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04264019429683685, + "step": 8170, + "valid_targets_mean": 3175.0, + "valid_targets_min": 722 + }, + { + "epoch": 5.946526009457985, + "grad_norm": 0.4675051643091551, + "learning_rate": 2.7061501842668782e-06, + "loss": 0.0711, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.031131118535995483, + "step": 8175, + "valid_targets_mean": 2600.4, + "valid_targets_min": 635 + }, + { + "epoch": 5.950163695889414, + "grad_norm": 0.39780970906042795, + "learning_rate": 2.6879608164276793e-06, + "loss": 0.0709, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0339793786406517, + "step": 8180, + "valid_targets_mean": 3322.4, + "valid_targets_min": 2174 + }, + { + "epoch": 5.953801382320844, + "grad_norm": 0.42405350548291243, + "learning_rate": 2.6698283799411527e-06, + "loss": 0.0694, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03296683728694916, + "step": 8185, + "valid_targets_mean": 3355.9, + "valid_targets_min": 854 + }, + { + "epoch": 5.957439068752274, + "grad_norm": 0.4511936973061408, + "learning_rate": 2.6517529344365687e-06, + "loss": 0.0622, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.028288856148719788, + "step": 8190, + "valid_targets_mean": 2574.4, + "valid_targets_min": 785 + }, + { + "epoch": 5.961076755183703, + "grad_norm": 0.6805264397045886, + "learning_rate": 2.6337345393557614e-06, + "loss": 0.09, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0374947264790535, + "step": 8195, + "valid_targets_mean": 1653.1, + "valid_targets_min": 766 + }, + { + "epoch": 5.964714441615133, + "grad_norm": 0.4945460078847651, + "learning_rate": 2.615773253952969e-06, + "loss": 0.0637, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03786884620785713, + "step": 8200, + "valid_targets_mean": 3234.5, + "valid_targets_min": 797 + }, + { + "epoch": 5.968352128046562, + "grad_norm": 0.49717252851288735, + "learning_rate": 2.5978691372946174e-06, + "loss": 0.0695, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03298147767782211, + "step": 8205, + "valid_targets_mean": 2719.8, + "valid_targets_min": 786 + }, + { + "epoch": 5.971989814477992, + "grad_norm": 0.5199026651317178, + "learning_rate": 2.5800222482591375e-06, + "loss": 0.0702, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04534943401813507, + "step": 8210, + "valid_targets_mean": 2937.1, + "valid_targets_min": 1825 + }, + { + "epoch": 5.975627500909422, + "grad_norm": 0.4107638234514434, + "learning_rate": 2.5622326455367486e-06, + "loss": 0.0767, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04032134264707565, + "step": 8215, + "valid_targets_mean": 4866.9, + "valid_targets_min": 2274 + }, + { + "epoch": 5.979265187340851, + "grad_norm": 0.4541360107318601, + "learning_rate": 2.5445003876292872e-06, + "loss": 0.0637, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03362446650862694, + "step": 8220, + "valid_targets_mean": 2550.5, + "valid_targets_min": 678 + }, + { + "epoch": 5.9829028737722805, + "grad_norm": 0.5317383283850395, + "learning_rate": 2.526825532850019e-06, + "loss": 0.0708, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.054304249584674835, + "step": 8225, + "valid_targets_mean": 3518.4, + "valid_targets_min": 1017 + }, + { + "epoch": 5.986540560203711, + "grad_norm": 0.6445622716844768, + "learning_rate": 2.5092081393234315e-06, + "loss": 0.0972, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06996852159500122, + "step": 8230, + "valid_targets_mean": 2348.9, + "valid_targets_min": 1535 + }, + { + "epoch": 5.99017824663514, + "grad_norm": 0.34494633285648274, + "learning_rate": 2.4916482649850495e-06, + "loss": 0.1271, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.028302479535341263, + "step": 8235, + "valid_targets_mean": 3666.5, + "valid_targets_min": 1097 + }, + { + "epoch": 5.9938159330665695, + "grad_norm": 0.375049786055388, + "learning_rate": 2.4741459675812384e-06, + "loss": 0.0724, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.026802975684404373, + "step": 8240, + "valid_targets_mean": 3338.4, + "valid_targets_min": 2453 + }, + { + "epoch": 5.997453619498, + "grad_norm": 0.3541844444564547, + "learning_rate": 2.4567013046690156e-06, + "loss": 0.0665, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.027132321149110794, + "step": 8245, + "valid_targets_mean": 3709.8, + "valid_targets_min": 2555 + }, + { + "epoch": 6.000727537286286, + "grad_norm": 0.7022155308760905, + "learning_rate": 2.439314333615883e-06, + "loss": 0.1253, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0940237045288086, + "step": 8250, + "valid_targets_mean": 8215.2, + "valid_targets_min": 6163 + }, + { + "epoch": 6.004365223717715, + "grad_norm": 0.8879924311173548, + "learning_rate": 2.421985111599605e-06, + "loss": 0.1888, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09915944188833237, + "step": 8255, + "valid_targets_mean": 8594.4, + "valid_targets_min": 6006 + }, + { + "epoch": 6.008002910149145, + "grad_norm": 0.7632084569205972, + "learning_rate": 2.4047136956080454e-06, + "loss": 0.1736, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07527078688144684, + "step": 8260, + "valid_targets_mean": 3578.4, + "valid_targets_min": 289 + }, + { + "epoch": 6.011640596580575, + "grad_norm": 0.632056300786101, + "learning_rate": 2.3875001424389586e-06, + "loss": 0.174, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08850613236427307, + "step": 8265, + "valid_targets_mean": 7019.9, + "valid_targets_min": 4985 + }, + { + "epoch": 6.015278283012004, + "grad_norm": 0.5246206729375817, + "learning_rate": 2.3703445086998267e-06, + "loss": 0.1646, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08553560823202133, + "step": 8270, + "valid_targets_mean": 6956.6, + "valid_targets_min": 1457 + }, + { + "epoch": 6.018915969443434, + "grad_norm": 0.44950187993768104, + "learning_rate": 2.35324685080766e-06, + "loss": 0.1688, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07981151342391968, + "step": 8275, + "valid_targets_mean": 7361.6, + "valid_targets_min": 4961 + }, + { + "epoch": 6.022553655874864, + "grad_norm": 0.5059146444736967, + "learning_rate": 2.3362072249888e-06, + "loss": 0.1811, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10505921393632889, + "step": 8280, + "valid_targets_mean": 7058.5, + "valid_targets_min": 5985 + }, + { + "epoch": 6.026191342306293, + "grad_norm": 0.4601142969907781, + "learning_rate": 2.3192256872787677e-06, + "loss": 0.1742, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09413161128759384, + "step": 8285, + "valid_targets_mean": 7664.2, + "valid_targets_min": 5169 + }, + { + "epoch": 6.029829028737723, + "grad_norm": 0.45162947995089925, + "learning_rate": 2.3023022935220363e-06, + "loss": 0.164, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08361732959747314, + "step": 8290, + "valid_targets_mean": 7111.8, + "valid_targets_min": 5348 + }, + { + "epoch": 6.033466715169152, + "grad_norm": 0.4933846627357058, + "learning_rate": 2.28543709937189e-06, + "loss": 0.1509, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0694289281964302, + "step": 8295, + "valid_targets_mean": 4238.9, + "valid_targets_min": 152 + }, + { + "epoch": 6.037104401600582, + "grad_norm": 0.4604148658602071, + "learning_rate": 2.268630160290213e-06, + "loss": 0.1576, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08578993380069733, + "step": 8300, + "valid_targets_mean": 7199.4, + "valid_targets_min": 4258 + }, + { + "epoch": 6.040742088032012, + "grad_norm": 0.4318595956486393, + "learning_rate": 2.2518815315473108e-06, + "loss": 0.1637, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07919541001319885, + "step": 8305, + "valid_targets_mean": 7243.4, + "valid_targets_min": 5831 + }, + { + "epoch": 6.044379774463441, + "grad_norm": 0.42308892684370425, + "learning_rate": 2.235191268221746e-06, + "loss": 0.1621, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08090521395206451, + "step": 8310, + "valid_targets_mean": 6950.6, + "valid_targets_min": 3862 + }, + { + "epoch": 6.0480174608948705, + "grad_norm": 0.4090323519115128, + "learning_rate": 2.2185594252001263e-06, + "loss": 0.1603, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0860210657119751, + "step": 8315, + "valid_targets_mean": 6991.9, + "valid_targets_min": 5696 + }, + { + "epoch": 6.051655147326301, + "grad_norm": 0.4574078605262672, + "learning_rate": 2.2019860571769545e-06, + "loss": 0.154, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08625351637601852, + "step": 8320, + "valid_targets_mean": 6963.5, + "valid_targets_min": 4947 + }, + { + "epoch": 6.05529283375773, + "grad_norm": 0.4570917675850537, + "learning_rate": 2.1854712186544334e-06, + "loss": 0.1583, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07491280883550644, + "step": 8325, + "valid_targets_mean": 6436.9, + "valid_targets_min": 4988 + }, + { + "epoch": 6.0589305201891595, + "grad_norm": 0.4135559851501268, + "learning_rate": 2.1690149639422954e-06, + "loss": 0.1627, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07911977916955948, + "step": 8330, + "valid_targets_mean": 7518.0, + "valid_targets_min": 5476 + }, + { + "epoch": 6.06256820662059, + "grad_norm": 0.3931056472879538, + "learning_rate": 2.1526173471576018e-06, + "loss": 0.1569, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07426949590444565, + "step": 8335, + "valid_targets_mean": 7422.6, + "valid_targets_min": 5226 + }, + { + "epoch": 6.066205893052019, + "grad_norm": 0.45398004428685523, + "learning_rate": 2.1362784222245845e-06, + "loss": 0.1659, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10211780667304993, + "step": 8340, + "valid_targets_mean": 7080.0, + "valid_targets_min": 3753 + }, + { + "epoch": 6.069843579483448, + "grad_norm": 0.5051174453803092, + "learning_rate": 2.119998242874488e-06, + "loss": 0.1815, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09130923449993134, + "step": 8345, + "valid_targets_mean": 6064.6, + "valid_targets_min": 5161 + }, + { + "epoch": 6.073481265914878, + "grad_norm": 0.4053433669441014, + "learning_rate": 2.1037768626453347e-06, + "loss": 0.1273, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07640621811151505, + "step": 8350, + "valid_targets_mean": 7388.1, + "valid_targets_min": 5795 + }, + { + "epoch": 6.077118952346308, + "grad_norm": 0.4080240454567278, + "learning_rate": 2.087614334881811e-06, + "loss": 0.1534, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07413780689239502, + "step": 8355, + "valid_targets_mean": 7696.0, + "valid_targets_min": 4658 + }, + { + "epoch": 6.080756638777737, + "grad_norm": 0.4364525963249273, + "learning_rate": 2.0715107127350453e-06, + "loss": 0.1565, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08404910564422607, + "step": 8360, + "valid_targets_mean": 7799.1, + "valid_targets_min": 5508 + }, + { + "epoch": 6.084394325209167, + "grad_norm": 0.4728864963344184, + "learning_rate": 2.055466049162467e-06, + "loss": 0.166, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08176246285438538, + "step": 8365, + "valid_targets_mean": 7341.6, + "valid_targets_min": 6454 + }, + { + "epoch": 6.088032011640596, + "grad_norm": 0.4392515102583269, + "learning_rate": 2.0394803969276156e-06, + "loss": 0.1591, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07033687829971313, + "step": 8370, + "valid_targets_mean": 6818.0, + "valid_targets_min": 4392 + }, + { + "epoch": 6.091669698072026, + "grad_norm": 0.39410661776615913, + "learning_rate": 2.023553808599954e-06, + "loss": 0.1585, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0695308968424797, + "step": 8375, + "valid_targets_mean": 7857.6, + "valid_targets_min": 5394 + }, + { + "epoch": 6.095307384503456, + "grad_norm": 0.45032805739161996, + "learning_rate": 2.007686336554735e-06, + "loss": 0.1392, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06551136076450348, + "step": 8380, + "valid_targets_mean": 8314.1, + "valid_targets_min": 4837 + }, + { + "epoch": 6.098945070934885, + "grad_norm": 0.36248906832318767, + "learning_rate": 1.991878032972785e-06, + "loss": 0.1399, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.059478651732206345, + "step": 8385, + "valid_targets_mean": 7368.2, + "valid_targets_min": 6027 + }, + { + "epoch": 6.102582757366315, + "grad_norm": 0.42117654203764476, + "learning_rate": 1.976128949840361e-06, + "loss": 0.1465, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07895243912935257, + "step": 8390, + "valid_targets_mean": 6435.0, + "valid_targets_min": 4238 + }, + { + "epoch": 6.106220443797745, + "grad_norm": 1.2239733568343933, + "learning_rate": 1.9604391389489753e-06, + "loss": 0.1438, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04017515480518341, + "step": 8395, + "valid_targets_mean": 247.8, + "valid_targets_min": 137 + }, + { + "epoch": 6.109858130229174, + "grad_norm": 0.4923657557918222, + "learning_rate": 1.9448086518952115e-06, + "loss": 0.1635, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07729047536849976, + "step": 8400, + "valid_targets_mean": 6807.0, + "valid_targets_min": 5165 + }, + { + "epoch": 6.1134958166606035, + "grad_norm": 0.5188793377677856, + "learning_rate": 1.929237540080573e-06, + "loss": 0.1711, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08811140805482864, + "step": 8405, + "valid_targets_mean": 6989.9, + "valid_targets_min": 5104 + }, + { + "epoch": 6.117133503092034, + "grad_norm": 0.45631267425951877, + "learning_rate": 1.9137258547112904e-06, + "loss": 0.1672, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09005340933799744, + "step": 8410, + "valid_targets_mean": 6692.2, + "valid_targets_min": 5068 + }, + { + "epoch": 6.120771189523463, + "grad_norm": 0.4585286967623021, + "learning_rate": 1.8982736467981943e-06, + "loss": 0.1639, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.098460353910923, + "step": 8415, + "valid_targets_mean": 7419.6, + "valid_targets_min": 5578 + }, + { + "epoch": 6.1244088759548925, + "grad_norm": 0.40469832406253226, + "learning_rate": 1.8828809671564996e-06, + "loss": 0.1597, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08193356543779373, + "step": 8420, + "valid_targets_mean": 7536.5, + "valid_targets_min": 5075 + }, + { + "epoch": 6.128046562386323, + "grad_norm": 0.3887939231808057, + "learning_rate": 1.8675478664056635e-06, + "loss": 0.1657, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0748477652668953, + "step": 8425, + "valid_targets_mean": 7221.5, + "valid_targets_min": 5841 + }, + { + "epoch": 6.131684248817752, + "grad_norm": 0.4075237578529197, + "learning_rate": 1.852274394969218e-06, + "loss": 0.1605, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07729834318161011, + "step": 8430, + "valid_targets_mean": 7287.5, + "valid_targets_min": 5151 + }, + { + "epoch": 6.1353219352491815, + "grad_norm": 0.4637490231619464, + "learning_rate": 1.837060603074603e-06, + "loss": 0.1694, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07897293567657471, + "step": 8435, + "valid_targets_mean": 6738.4, + "valid_targets_min": 4042 + }, + { + "epoch": 6.138959621680611, + "grad_norm": 0.4668925238048148, + "learning_rate": 1.8219065407530045e-06, + "loss": 0.163, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08996886014938354, + "step": 8440, + "valid_targets_mean": 6331.0, + "valid_targets_min": 5496 + }, + { + "epoch": 6.142597308112041, + "grad_norm": 0.42929832632611525, + "learning_rate": 1.80681225783917e-06, + "loss": 0.1657, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07836571335792542, + "step": 8445, + "valid_targets_mean": 6271.1, + "valid_targets_min": 4695 + }, + { + "epoch": 6.14623499454347, + "grad_norm": 0.452694798539144, + "learning_rate": 1.7917778039712798e-06, + "loss": 0.1616, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0848751962184906, + "step": 8450, + "valid_targets_mean": 5790.8, + "valid_targets_min": 5021 + }, + { + "epoch": 6.1498726809749, + "grad_norm": 0.4884286421689847, + "learning_rate": 1.776803228590742e-06, + "loss": 0.1532, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05950256437063217, + "step": 8455, + "valid_targets_mean": 3386.8, + "valid_targets_min": 2279 + }, + { + "epoch": 6.153510367406329, + "grad_norm": 0.4244728419395087, + "learning_rate": 1.7618885809420706e-06, + "loss": 0.1332, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07865472882986069, + "step": 8460, + "valid_targets_mean": 6240.5, + "valid_targets_min": 5093 + }, + { + "epoch": 6.157148053837759, + "grad_norm": 0.43959460075064916, + "learning_rate": 1.7470339100726997e-06, + "loss": 0.1583, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08081591129302979, + "step": 8465, + "valid_targets_mean": 6329.5, + "valid_targets_min": 4589 + }, + { + "epoch": 6.160785740269189, + "grad_norm": 0.43198191131739777, + "learning_rate": 1.7322392648328179e-06, + "loss": 0.1586, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08507980406284332, + "step": 8470, + "valid_targets_mean": 7589.2, + "valid_targets_min": 4963 + }, + { + "epoch": 6.164423426700618, + "grad_norm": 0.4234151513664653, + "learning_rate": 1.7175046938752316e-06, + "loss": 0.1593, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06843066960573196, + "step": 8475, + "valid_targets_mean": 5577.9, + "valid_targets_min": 4458 + }, + { + "epoch": 6.168061113132048, + "grad_norm": 0.4537888356980943, + "learning_rate": 1.7028302456551782e-06, + "loss": 0.1588, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07723157107830048, + "step": 8480, + "valid_targets_mean": 5432.5, + "valid_targets_min": 4433 + }, + { + "epoch": 6.171698799563478, + "grad_norm": 0.6742745916149937, + "learning_rate": 1.6882159684301846e-06, + "loss": 0.1634, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08562292903661728, + "step": 8485, + "valid_targets_mean": 6361.8, + "valid_targets_min": 5201 + }, + { + "epoch": 6.175336485994907, + "grad_norm": 0.5811211168197454, + "learning_rate": 1.6736619102599073e-06, + "loss": 0.1612, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07633039355278015, + "step": 8490, + "valid_targets_mean": 5824.4, + "valid_targets_min": 5129 + }, + { + "epoch": 6.1789741724263365, + "grad_norm": 0.7462574472460597, + "learning_rate": 1.6591681190059582e-06, + "loss": 0.1228, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07242149114608765, + "step": 8495, + "valid_targets_mean": 1637.6, + "valid_targets_min": 137 + }, + { + "epoch": 6.182611858857767, + "grad_norm": 0.4224580878781636, + "learning_rate": 1.644734642331769e-06, + "loss": 0.1549, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07867047935724258, + "step": 8500, + "valid_targets_mean": 6474.6, + "valid_targets_min": 4751 + }, + { + "epoch": 6.186249545289196, + "grad_norm": 0.42300394755573734, + "learning_rate": 1.6303615277024222e-06, + "loss": 0.1562, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08644704520702362, + "step": 8505, + "valid_targets_mean": 7186.2, + "valid_targets_min": 5358 + }, + { + "epoch": 6.1898872317206255, + "grad_norm": 0.41470386026852957, + "learning_rate": 1.6160488223844972e-06, + "loss": 0.1617, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08024127781391144, + "step": 8510, + "valid_targets_mean": 7126.8, + "valid_targets_min": 4078 + }, + { + "epoch": 6.193524918152056, + "grad_norm": 0.4582752232249878, + "learning_rate": 1.6017965734459129e-06, + "loss": 0.1531, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07523344457149506, + "step": 8515, + "valid_targets_mean": 7022.1, + "valid_targets_min": 4940 + }, + { + "epoch": 6.197162604583485, + "grad_norm": 0.45232831146197966, + "learning_rate": 1.5876048277557688e-06, + "loss": 0.1513, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0824085921049118, + "step": 8520, + "valid_targets_mean": 5900.6, + "valid_targets_min": 4779 + }, + { + "epoch": 6.2008002910149145, + "grad_norm": 0.44950577176730705, + "learning_rate": 1.5734736319842125e-06, + "loss": 0.1592, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08161944895982742, + "step": 8525, + "valid_targets_mean": 6290.1, + "valid_targets_min": 4620 + }, + { + "epoch": 6.204437977446344, + "grad_norm": 0.6478265055015933, + "learning_rate": 1.5594030326022602e-06, + "loss": 0.2004, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11330418288707733, + "step": 8530, + "valid_targets_mean": 5614.5, + "valid_targets_min": 3446 + }, + { + "epoch": 6.208075663877774, + "grad_norm": 0.7018931052098141, + "learning_rate": 1.5453930758816605e-06, + "loss": 0.2132, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11321181058883667, + "step": 8535, + "valid_targets_mean": 4630.0, + "valid_targets_min": 1005 + }, + { + "epoch": 6.2117133503092035, + "grad_norm": 0.5751731369460551, + "learning_rate": 1.5314438078947258e-06, + "loss": 0.2128, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0803496390581131, + "step": 8540, + "valid_targets_mean": 4669.6, + "valid_targets_min": 2054 + }, + { + "epoch": 6.215351036740633, + "grad_norm": 0.6045007758852456, + "learning_rate": 1.5175552745141952e-06, + "loss": 0.2074, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08893614262342453, + "step": 8545, + "valid_targets_mean": 3745.0, + "valid_targets_min": 1561 + }, + { + "epoch": 6.218988723172062, + "grad_norm": 0.6217684903369846, + "learning_rate": 1.503727521413092e-06, + "loss": 0.2097, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10704738646745682, + "step": 8550, + "valid_targets_mean": 4284.0, + "valid_targets_min": 1184 + }, + { + "epoch": 6.2226264096034924, + "grad_norm": 0.5352609236226661, + "learning_rate": 1.4899605940645413e-06, + "loss": 0.2141, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10525930672883987, + "step": 8555, + "valid_targets_mean": 6332.8, + "valid_targets_min": 1651 + }, + { + "epoch": 6.226264096034922, + "grad_norm": 0.5471394063410301, + "learning_rate": 1.476254537741657e-06, + "loss": 0.2051, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10387635231018066, + "step": 8560, + "valid_targets_mean": 6446.5, + "valid_targets_min": 2231 + }, + { + "epoch": 6.229901782466351, + "grad_norm": 0.5973037168365218, + "learning_rate": 1.4626093975173627e-06, + "loss": 0.2037, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07628235220909119, + "step": 8565, + "valid_targets_mean": 3367.2, + "valid_targets_min": 467 + }, + { + "epoch": 6.233539468897781, + "grad_norm": 0.5366531277360588, + "learning_rate": 1.449025218264266e-06, + "loss": 0.2111, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1344883143901825, + "step": 8570, + "valid_targets_mean": 7194.1, + "valid_targets_min": 1748 + }, + { + "epoch": 6.237177155329211, + "grad_norm": 0.6130659138952005, + "learning_rate": 1.435502044654502e-06, + "loss": 0.2124, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10665645450353622, + "step": 8575, + "valid_targets_mean": 4078.5, + "valid_targets_min": 1697 + }, + { + "epoch": 6.24081484176064, + "grad_norm": 0.6834121301313114, + "learning_rate": 1.4220399211595748e-06, + "loss": 0.2062, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06865745782852173, + "step": 8580, + "valid_targets_mean": 2089.5, + "valid_targets_min": 1223 + }, + { + "epoch": 6.2444525281920695, + "grad_norm": 0.6090586156389747, + "learning_rate": 1.4086388920502403e-06, + "loss": 0.2056, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0915830135345459, + "step": 8585, + "valid_targets_mean": 5404.1, + "valid_targets_min": 1829 + }, + { + "epoch": 6.2480902146235, + "grad_norm": 0.5563223513726037, + "learning_rate": 1.3952990013963264e-06, + "loss": 0.2078, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10034674406051636, + "step": 8590, + "valid_targets_mean": 4583.1, + "valid_targets_min": 1596 + }, + { + "epoch": 6.251727901054929, + "grad_norm": 0.5874612570402994, + "learning_rate": 1.3820202930666148e-06, + "loss": 0.202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10953615605831146, + "step": 8595, + "valid_targets_mean": 4666.8, + "valid_targets_min": 2041 + }, + { + "epoch": 6.2553655874863585, + "grad_norm": 0.6129950888953899, + "learning_rate": 1.3688028107286888e-06, + "loss": 0.2007, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10942976921796799, + "step": 8600, + "valid_targets_mean": 4272.1, + "valid_targets_min": 1538 + }, + { + "epoch": 6.259003273917788, + "grad_norm": 0.6754450811759718, + "learning_rate": 1.355646597848772e-06, + "loss": 0.1984, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08985291421413422, + "step": 8605, + "valid_targets_mean": 2928.0, + "valid_targets_min": 1739 + }, + { + "epoch": 6.262640960349218, + "grad_norm": 0.621868556591415, + "learning_rate": 1.342551697691623e-06, + "loss": 0.1993, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09238231182098389, + "step": 8610, + "valid_targets_mean": 3254.5, + "valid_targets_min": 935 + }, + { + "epoch": 6.2662786467806475, + "grad_norm": 0.6486392944173593, + "learning_rate": 1.329518153320346e-06, + "loss": 0.2048, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1022038459777832, + "step": 8615, + "valid_targets_mean": 3535.2, + "valid_targets_min": 2704 + }, + { + "epoch": 6.269916333212077, + "grad_norm": 0.686276716540914, + "learning_rate": 1.3165460075963022e-06, + "loss": 0.2023, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09486167132854462, + "step": 8620, + "valid_targets_mean": 2987.6, + "valid_targets_min": 1173 + }, + { + "epoch": 6.273554019643507, + "grad_norm": 0.6105710268652234, + "learning_rate": 1.303635303178914e-06, + "loss": 0.2081, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09641323983669281, + "step": 8625, + "valid_targets_mean": 3700.6, + "valid_targets_min": 1670 + }, + { + "epoch": 6.2771917060749365, + "grad_norm": 0.650334912521023, + "learning_rate": 1.2907860825255747e-06, + "loss": 0.2026, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08880806714296341, + "step": 8630, + "valid_targets_mean": 3481.6, + "valid_targets_min": 1329 + }, + { + "epoch": 6.280829392506366, + "grad_norm": 0.640717919514835, + "learning_rate": 1.2779983878914637e-06, + "loss": 0.1905, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10280246287584305, + "step": 8635, + "valid_targets_mean": 3589.8, + "valid_targets_min": 2279 + }, + { + "epoch": 6.284467078937795, + "grad_norm": 0.6130000351960836, + "learning_rate": 1.2652722613294465e-06, + "loss": 0.1828, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09308261424303055, + "step": 8640, + "valid_targets_mean": 5060.4, + "valid_targets_min": 2222 + }, + { + "epoch": 6.2881047653692255, + "grad_norm": 0.649412117274213, + "learning_rate": 1.2526077446899177e-06, + "loss": 0.194, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09832559525966644, + "step": 8645, + "valid_targets_mean": 3898.6, + "valid_targets_min": 1849 + }, + { + "epoch": 6.291742451800655, + "grad_norm": 0.5969514319285444, + "learning_rate": 1.240004879620651e-06, + "loss": 0.1978, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08473603427410126, + "step": 8650, + "valid_targets_mean": 3698.9, + "valid_targets_min": 1951 + }, + { + "epoch": 6.295380138232084, + "grad_norm": 0.6306754483157647, + "learning_rate": 1.2274637075666962e-06, + "loss": 0.2036, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10783109813928604, + "step": 8655, + "valid_targets_mean": 4506.8, + "valid_targets_min": 1765 + }, + { + "epoch": 6.299017824663514, + "grad_norm": 0.690046909354114, + "learning_rate": 1.214984269770203e-06, + "loss": 0.1891, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09177146852016449, + "step": 8660, + "valid_targets_mean": 2991.6, + "valid_targets_min": 2087 + }, + { + "epoch": 6.302655511094944, + "grad_norm": 0.6160739967620503, + "learning_rate": 1.2025666072703234e-06, + "loss": 0.1909, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08940046280622482, + "step": 8665, + "valid_targets_mean": 3994.1, + "valid_targets_min": 1603 + }, + { + "epoch": 6.306293197526373, + "grad_norm": 0.8235995694486655, + "learning_rate": 1.1902107609030522e-06, + "loss": 0.1966, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10371425002813339, + "step": 8670, + "valid_targets_mean": 4391.9, + "valid_targets_min": 2092 + }, + { + "epoch": 6.3099308839578025, + "grad_norm": 0.6152723846357705, + "learning_rate": 1.177916771301091e-06, + "loss": 0.1929, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09667868167161942, + "step": 8675, + "valid_targets_mean": 3808.9, + "valid_targets_min": 1410 + }, + { + "epoch": 6.313568570389233, + "grad_norm": 0.5857366445948536, + "learning_rate": 1.1656846788937392e-06, + "loss": 0.1839, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1016744077205658, + "step": 8680, + "valid_targets_mean": 5078.5, + "valid_targets_min": 2110 + }, + { + "epoch": 6.317206256820662, + "grad_norm": 0.630436474370034, + "learning_rate": 1.1535145239067224e-06, + "loss": 0.1972, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09176885336637497, + "step": 8685, + "valid_targets_mean": 3447.4, + "valid_targets_min": 2420 + }, + { + "epoch": 6.3208439432520915, + "grad_norm": 0.6432925667639736, + "learning_rate": 1.1414063463621105e-06, + "loss": 0.1905, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08560145646333694, + "step": 8690, + "valid_targets_mean": 4371.1, + "valid_targets_min": 1876 + }, + { + "epoch": 6.324481629683521, + "grad_norm": 0.6821199653394837, + "learning_rate": 1.1293601860781366e-06, + "loss": 0.1926, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10832156240940094, + "step": 8695, + "valid_targets_mean": 3858.2, + "valid_targets_min": 2605 + }, + { + "epoch": 6.328119316114951, + "grad_norm": 0.678250608147653, + "learning_rate": 1.1173760826690927e-06, + "loss": 0.1902, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0882798433303833, + "step": 8700, + "valid_targets_mean": 3383.9, + "valid_targets_min": 1177 + }, + { + "epoch": 6.3317570025463805, + "grad_norm": 0.6863159512309185, + "learning_rate": 1.1054540755451936e-06, + "loss": 0.1942, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09751853346824646, + "step": 8705, + "valid_targets_mean": 3455.5, + "valid_targets_min": 1714 + }, + { + "epoch": 6.33539468897781, + "grad_norm": 0.5997030344153224, + "learning_rate": 1.093594203912456e-06, + "loss": 0.1954, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10792891681194305, + "step": 8710, + "valid_targets_mean": 5547.5, + "valid_targets_min": 2033 + }, + { + "epoch": 6.33903237540924, + "grad_norm": 0.6703537746175533, + "learning_rate": 1.0817965067725544e-06, + "loss": 0.1893, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09713058173656464, + "step": 8715, + "valid_targets_mean": 3485.4, + "valid_targets_min": 1924 + }, + { + "epoch": 6.3426700618406695, + "grad_norm": 0.6710924556976414, + "learning_rate": 1.0700610229226992e-06, + "loss": 0.1961, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09242476522922516, + "step": 8720, + "valid_targets_mean": 3080.2, + "valid_targets_min": 2098 + }, + { + "epoch": 6.346307748272099, + "grad_norm": 0.7188821116801972, + "learning_rate": 1.0583877909555106e-06, + "loss": 0.1973, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10941218584775925, + "step": 8725, + "valid_targets_mean": 3393.4, + "valid_targets_min": 1363 + }, + { + "epoch": 6.349945434703528, + "grad_norm": 0.6920559241275892, + "learning_rate": 1.0467768492588904e-06, + "loss": 0.1894, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10541440546512604, + "step": 8730, + "valid_targets_mean": 4115.4, + "valid_targets_min": 1363 + }, + { + "epoch": 6.3535831211349585, + "grad_norm": 0.6232375175121931, + "learning_rate": 1.0352282360159016e-06, + "loss": 0.1953, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1185319721698761, + "step": 8735, + "valid_targets_mean": 4313.8, + "valid_targets_min": 2620 + }, + { + "epoch": 6.357220807566388, + "grad_norm": 0.6665820943763031, + "learning_rate": 1.0237419892046364e-06, + "loss": 0.1935, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10018257796764374, + "step": 8740, + "valid_targets_mean": 3425.1, + "valid_targets_min": 2316 + }, + { + "epoch": 6.360858493997817, + "grad_norm": 0.6635362560778351, + "learning_rate": 1.0123181465980813e-06, + "loss": 0.191, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09305242449045181, + "step": 8745, + "valid_targets_mean": 3744.0, + "valid_targets_min": 2022 + }, + { + "epoch": 6.364496180429247, + "grad_norm": 0.6700051041240409, + "learning_rate": 1.000956745764019e-06, + "loss": 0.186, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06542357057332993, + "step": 8750, + "valid_targets_mean": 2445.6, + "valid_targets_min": 1294 + }, + { + "epoch": 6.368133866860677, + "grad_norm": 0.6648363906322654, + "learning_rate": 9.89657824064878e-07, + "loss": 0.1955, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0936518982052803, + "step": 8755, + "valid_targets_mean": 3556.1, + "valid_targets_min": 1700 + }, + { + "epoch": 6.371771553292106, + "grad_norm": 0.6796459728842329, + "learning_rate": 9.784214186576313e-07, + "loss": 0.194, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08168792724609375, + "step": 8760, + "valid_targets_mean": 2714.0, + "valid_targets_min": 1586 + }, + { + "epoch": 6.375409239723536, + "grad_norm": 0.7195486049275939, + "learning_rate": 9.67247566493661e-07, + "loss": 0.1918, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10128866881132126, + "step": 8765, + "valid_targets_mean": 3266.4, + "valid_targets_min": 1358 + }, + { + "epoch": 6.379046926154966, + "grad_norm": 0.6568287697553483, + "learning_rate": 9.561363043186356e-07, + "loss": 0.1943, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09369741380214691, + "step": 8770, + "valid_targets_mean": 3782.9, + "valid_targets_min": 2290 + }, + { + "epoch": 6.382684612586395, + "grad_norm": 0.6938811919226335, + "learning_rate": 9.450876686723975e-07, + "loss": 0.2022, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13313284516334534, + "step": 8775, + "valid_targets_mean": 4904.8, + "valid_targets_min": 2189 + }, + { + "epoch": 6.3863222990178246, + "grad_norm": 0.6749689205517695, + "learning_rate": 9.341016958888449e-07, + "loss": 0.1916, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12198731303215027, + "step": 8780, + "valid_targets_mean": 3959.5, + "valid_targets_min": 2121 + }, + { + "epoch": 6.389959985449254, + "grad_norm": 0.6724229962016504, + "learning_rate": 9.23178422095794e-07, + "loss": 0.1935, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0965539887547493, + "step": 8785, + "valid_targets_mean": 3689.4, + "valid_targets_min": 2011 + }, + { + "epoch": 6.393597671880684, + "grad_norm": 0.6216950721405592, + "learning_rate": 9.12317883214886e-07, + "loss": 0.187, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10437732934951782, + "step": 8790, + "valid_targets_mean": 5082.1, + "valid_targets_min": 2674 + }, + { + "epoch": 6.3972353583121135, + "grad_norm": 0.6520637917780099, + "learning_rate": 9.015201149614427e-07, + "loss": 0.182, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10639089345932007, + "step": 8795, + "valid_targets_mean": 3660.9, + "valid_targets_min": 2266 + }, + { + "epoch": 6.400873044743543, + "grad_norm": 0.5770955601333918, + "learning_rate": 8.907851528443734e-07, + "loss": 0.188, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08673425018787384, + "step": 8800, + "valid_targets_mean": 4285.5, + "valid_targets_min": 904 + }, + { + "epoch": 6.404510731174973, + "grad_norm": 0.6515263405887404, + "learning_rate": 8.801130321660411e-07, + "loss": 0.1888, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1116877868771553, + "step": 8805, + "valid_targets_mean": 3999.4, + "valid_targets_min": 1194 + }, + { + "epoch": 6.4081484176064025, + "grad_norm": 0.6279902508598477, + "learning_rate": 8.695037880221569e-07, + "loss": 0.1929, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08770826458930969, + "step": 8810, + "valid_targets_mean": 4011.8, + "valid_targets_min": 1658 + }, + { + "epoch": 6.411786104037832, + "grad_norm": 0.7153757791909153, + "learning_rate": 8.589574553016545e-07, + "loss": 0.1912, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07813868671655655, + "step": 8815, + "valid_targets_mean": 2728.8, + "valid_targets_min": 1423 + }, + { + "epoch": 6.415423790469261, + "grad_norm": 0.6615702733383715, + "learning_rate": 8.484740686865778e-07, + "loss": 0.193, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10356613993644714, + "step": 8820, + "valid_targets_mean": 4504.4, + "valid_targets_min": 2320 + }, + { + "epoch": 6.4190614769006915, + "grad_norm": 0.6455147675277527, + "learning_rate": 8.380536626519809e-07, + "loss": 0.1956, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07337168604135513, + "step": 8825, + "valid_targets_mean": 3272.5, + "valid_targets_min": 1811 + }, + { + "epoch": 6.422699163332121, + "grad_norm": 0.6145557929558106, + "learning_rate": 8.276962714657922e-07, + "loss": 0.202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09295627474784851, + "step": 8830, + "valid_targets_mean": 4425.9, + "valid_targets_min": 3026 + }, + { + "epoch": 6.42633684976355, + "grad_norm": 0.6627913423568262, + "learning_rate": 8.174019291887258e-07, + "loss": 0.1907, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08517342060804367, + "step": 8835, + "valid_targets_mean": 2719.1, + "valid_targets_min": 644 + }, + { + "epoch": 6.42997453619498, + "grad_norm": 0.6528935310009382, + "learning_rate": 8.071706696741466e-07, + "loss": 0.187, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08744402229785919, + "step": 8840, + "valid_targets_mean": 3369.0, + "valid_targets_min": 1397 + }, + { + "epoch": 6.43361222262641, + "grad_norm": 0.6448768255390662, + "learning_rate": 7.970025265679648e-07, + "loss": 0.1772, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09138321131467819, + "step": 8845, + "valid_targets_mean": 4122.6, + "valid_targets_min": 1415 + }, + { + "epoch": 6.437249909057839, + "grad_norm": 0.6379886695771694, + "learning_rate": 7.868975333085483e-07, + "loss": 0.1791, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09732823818922043, + "step": 8850, + "valid_targets_mean": 4136.6, + "valid_targets_min": 621 + }, + { + "epoch": 6.440887595489269, + "grad_norm": 0.6419065746957366, + "learning_rate": 7.76855723126575e-07, + "loss": 0.1797, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0729910358786583, + "step": 8855, + "valid_targets_mean": 3065.8, + "valid_targets_min": 1290 + }, + { + "epoch": 6.444525281920699, + "grad_norm": 0.7370929286068046, + "learning_rate": 7.668771290449539e-07, + "loss": 0.175, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07963927090167999, + "step": 8860, + "valid_targets_mean": 2426.9, + "valid_targets_min": 857 + }, + { + "epoch": 6.448162968352128, + "grad_norm": 0.710902099602013, + "learning_rate": 7.569617838787024e-07, + "loss": 0.1764, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09321784973144531, + "step": 8865, + "valid_targets_mean": 2949.1, + "valid_targets_min": 720 + }, + { + "epoch": 6.451800654783558, + "grad_norm": 0.6924207215531653, + "learning_rate": 7.471097202348377e-07, + "loss": 0.1866, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10883540660142899, + "step": 8870, + "valid_targets_mean": 3579.1, + "valid_targets_min": 1801 + }, + { + "epoch": 6.455438341214987, + "grad_norm": 0.6481835155154693, + "learning_rate": 7.373209705122808e-07, + "loss": 0.2047, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09211036562919617, + "step": 8875, + "valid_targets_mean": 3532.6, + "valid_targets_min": 1510 + }, + { + "epoch": 6.459076027646417, + "grad_norm": 0.7297875797710194, + "learning_rate": 7.275955669017332e-07, + "loss": 0.1784, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08426085859537125, + "step": 8880, + "valid_targets_mean": 2807.9, + "valid_targets_min": 987 + }, + { + "epoch": 6.462713714077847, + "grad_norm": 0.6366009426609505, + "learning_rate": 7.179335413855893e-07, + "loss": 0.1906, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08356069028377533, + "step": 8885, + "valid_targets_mean": 3937.4, + "valid_targets_min": 996 + }, + { + "epoch": 6.466351400509276, + "grad_norm": 0.633647017134607, + "learning_rate": 7.083349257378125e-07, + "loss": 0.1831, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07954810559749603, + "step": 8890, + "valid_targets_mean": 3954.5, + "valid_targets_min": 2001 + }, + { + "epoch": 6.469989086940705, + "grad_norm": 0.5784213894353188, + "learning_rate": 6.987997515238443e-07, + "loss": 0.1844, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07162293791770935, + "step": 8895, + "valid_targets_mean": 5059.6, + "valid_targets_min": 1767 + }, + { + "epoch": 6.4736267733721355, + "grad_norm": 0.5744459017133227, + "learning_rate": 6.89328050100504e-07, + "loss": 0.1864, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07844725996255875, + "step": 8900, + "valid_targets_mean": 4714.0, + "valid_targets_min": 1511 + }, + { + "epoch": 6.477264459803565, + "grad_norm": 0.7165246499828036, + "learning_rate": 6.799198526158668e-07, + "loss": 0.1822, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09037096053361893, + "step": 8905, + "valid_targets_mean": 2861.0, + "valid_targets_min": 974 + }, + { + "epoch": 6.480902146234994, + "grad_norm": 0.7028551416495389, + "learning_rate": 6.705751900091773e-07, + "loss": 0.1861, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08042994886636734, + "step": 8910, + "valid_targets_mean": 3792.6, + "valid_targets_min": 2339 + }, + { + "epoch": 6.4845398326664245, + "grad_norm": 0.6722895333623209, + "learning_rate": 6.612940930107381e-07, + "loss": 0.1814, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08294190466403961, + "step": 8915, + "valid_targets_mean": 3333.1, + "valid_targets_min": 1249 + }, + { + "epoch": 6.488177519097854, + "grad_norm": 0.6746584450175835, + "learning_rate": 6.520765921418259e-07, + "loss": 0.1897, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09232622385025024, + "step": 8920, + "valid_targets_mean": 3493.1, + "valid_targets_min": 1223 + }, + { + "epoch": 6.491815205529283, + "grad_norm": 0.7156921459383032, + "learning_rate": 6.429227177145669e-07, + "loss": 0.1985, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10127221792936325, + "step": 8925, + "valid_targets_mean": 3285.8, + "valid_targets_min": 1318 + }, + { + "epoch": 6.495452891960713, + "grad_norm": 0.7341838669219943, + "learning_rate": 6.33832499831859e-07, + "loss": 0.1941, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08726362884044647, + "step": 8930, + "valid_targets_mean": 2070.5, + "valid_targets_min": 1002 + }, + { + "epoch": 6.499090578392143, + "grad_norm": 0.7105623501434117, + "learning_rate": 6.248059683872476e-07, + "loss": 0.2056, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08885078132152557, + "step": 8935, + "valid_targets_mean": 3328.1, + "valid_targets_min": 845 + }, + { + "epoch": 6.502728264823572, + "grad_norm": 0.6188301669948522, + "learning_rate": 6.158431530648612e-07, + "loss": 0.1833, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10391568392515182, + "step": 8940, + "valid_targets_mean": 4657.4, + "valid_targets_min": 2747 + }, + { + "epoch": 6.506365951255002, + "grad_norm": 0.685738411223008, + "learning_rate": 6.069440833392848e-07, + "loss": 0.2029, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10299421846866608, + "step": 8945, + "valid_targets_mean": 4166.4, + "valid_targets_min": 1241 + }, + { + "epoch": 6.510003637686431, + "grad_norm": 0.7206041486000097, + "learning_rate": 5.981087884754711e-07, + "loss": 0.1945, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10663892328739166, + "step": 8950, + "valid_targets_mean": 3567.1, + "valid_targets_min": 992 + }, + { + "epoch": 6.513641324117861, + "grad_norm": 0.6624610395730864, + "learning_rate": 5.89337297528656e-07, + "loss": 0.1918, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09143456816673279, + "step": 8955, + "valid_targets_mean": 3343.5, + "valid_targets_min": 1211 + }, + { + "epoch": 6.517279010549291, + "grad_norm": 0.7850439907063984, + "learning_rate": 5.806296393442412e-07, + "loss": 0.198, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10617111623287201, + "step": 8960, + "valid_targets_mean": 3081.6, + "valid_targets_min": 1889 + }, + { + "epoch": 6.52091669698072, + "grad_norm": 0.7099684355611328, + "learning_rate": 5.719858425577251e-07, + "loss": 0.1898, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08416569232940674, + "step": 8965, + "valid_targets_mean": 3233.9, + "valid_targets_min": 1624 + }, + { + "epoch": 6.52455438341215, + "grad_norm": 0.6625047157037457, + "learning_rate": 5.634059355945876e-07, + "loss": 0.1947, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09973809123039246, + "step": 8970, + "valid_targets_mean": 3817.4, + "valid_targets_min": 2763 + }, + { + "epoch": 6.52819206984358, + "grad_norm": 0.7269697704685366, + "learning_rate": 5.548899466702029e-07, + "loss": 0.1936, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09195046871900558, + "step": 8975, + "valid_targets_mean": 3037.0, + "valid_targets_min": 1928 + }, + { + "epoch": 6.531829756275009, + "grad_norm": 0.6466919896835718, + "learning_rate": 5.464379037897539e-07, + "loss": 0.1793, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09277378022670746, + "step": 8980, + "valid_targets_mean": 4733.2, + "valid_targets_min": 3147 + }, + { + "epoch": 6.535467442706439, + "grad_norm": 0.6339840903766208, + "learning_rate": 5.380498347481266e-07, + "loss": 0.1835, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08492060005664825, + "step": 8985, + "valid_targets_mean": 3348.4, + "valid_targets_min": 1844 + }, + { + "epoch": 6.539105129137869, + "grad_norm": 0.701348384376016, + "learning_rate": 5.297257671298384e-07, + "loss": 0.1873, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10056929290294647, + "step": 8990, + "valid_targets_mean": 4360.2, + "valid_targets_min": 2330 + }, + { + "epoch": 6.542742815569298, + "grad_norm": 0.7415411407126674, + "learning_rate": 5.214657283089208e-07, + "loss": 0.1914, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10902796685695648, + "step": 8995, + "valid_targets_mean": 3574.1, + "valid_targets_min": 900 + }, + { + "epoch": 6.546380502000727, + "grad_norm": 0.5984258237447774, + "learning_rate": 5.132697454488545e-07, + "loss": 0.1842, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09636232256889343, + "step": 9000, + "valid_targets_mean": 5092.8, + "valid_targets_min": 1713 + }, + { + "epoch": 6.550018188432157, + "grad_norm": 0.7155664560177928, + "learning_rate": 5.051378455024635e-07, + "loss": 0.1927, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10691633820533752, + "step": 9005, + "valid_targets_mean": 4057.1, + "valid_targets_min": 981 + }, + { + "epoch": 6.553655874863587, + "grad_norm": 0.6474594317363322, + "learning_rate": 4.970700552118368e-07, + "loss": 0.1912, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10909921675920486, + "step": 9010, + "valid_targets_mean": 4127.9, + "valid_targets_min": 2098 + }, + { + "epoch": 6.557293561295016, + "grad_norm": 0.6870690330834758, + "learning_rate": 4.89066401108238e-07, + "loss": 0.1846, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08491425216197968, + "step": 9015, + "valid_targets_mean": 3101.1, + "valid_targets_min": 1107 + }, + { + "epoch": 6.560931247726446, + "grad_norm": 0.6198131497873299, + "learning_rate": 4.811269095120086e-07, + "loss": 0.1883, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08999783545732498, + "step": 9020, + "valid_targets_mean": 3718.0, + "valid_targets_min": 2273 + }, + { + "epoch": 6.564568934157876, + "grad_norm": 0.684761855897908, + "learning_rate": 4.732516065324921e-07, + "loss": 0.1861, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08337712287902832, + "step": 9025, + "valid_targets_mean": 3316.0, + "valid_targets_min": 1920 + }, + { + "epoch": 6.568206620589305, + "grad_norm": 0.6396966654007165, + "learning_rate": 4.6544051806794776e-07, + "loss": 0.1727, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09264329075813293, + "step": 9030, + "valid_targets_mean": 4710.8, + "valid_targets_min": 1687 + }, + { + "epoch": 6.571844307020735, + "grad_norm": 0.7010089116114495, + "learning_rate": 4.5769366980546124e-07, + "loss": 0.1812, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07932674884796143, + "step": 9035, + "valid_targets_mean": 3075.8, + "valid_targets_min": 1375 + }, + { + "epoch": 6.575481993452165, + "grad_norm": 0.6774570659802891, + "learning_rate": 4.500110872208652e-07, + "loss": 0.1835, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08248923718929291, + "step": 9040, + "valid_targets_mean": 3776.4, + "valid_targets_min": 906 + }, + { + "epoch": 6.579119679883594, + "grad_norm": 0.6980566804481833, + "learning_rate": 4.4239279557864114e-07, + "loss": 0.1779, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0895472913980484, + "step": 9045, + "valid_targets_mean": 3393.9, + "valid_targets_min": 1874 + }, + { + "epoch": 6.582757366315024, + "grad_norm": 0.6094062704445419, + "learning_rate": 4.3483881993185985e-07, + "loss": 0.1719, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08976786583662033, + "step": 9050, + "valid_targets_mean": 3503.4, + "valid_targets_min": 1877 + }, + { + "epoch": 6.586395052746453, + "grad_norm": 0.634759977482158, + "learning_rate": 4.2734918512207903e-07, + "loss": 0.1759, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08539655804634094, + "step": 9055, + "valid_targets_mean": 3744.4, + "valid_targets_min": 1697 + }, + { + "epoch": 6.590032739177883, + "grad_norm": 0.6933618003572481, + "learning_rate": 4.199239157792701e-07, + "loss": 0.1803, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09375739097595215, + "step": 9060, + "valid_targets_mean": 3015.1, + "valid_targets_min": 1679 + }, + { + "epoch": 6.593670425609313, + "grad_norm": 0.7036670153334643, + "learning_rate": 4.125630363217381e-07, + "loss": 0.1918, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09603090584278107, + "step": 9065, + "valid_targets_mean": 3292.0, + "valid_targets_min": 1695 + }, + { + "epoch": 6.597308112040742, + "grad_norm": 0.6466250211339908, + "learning_rate": 4.052665709560377e-07, + "loss": 0.1785, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07848048210144043, + "step": 9070, + "valid_targets_mean": 3148.8, + "valid_targets_min": 2114 + }, + { + "epoch": 6.600945798472171, + "grad_norm": 0.6598794983930097, + "learning_rate": 3.9803454367689064e-07, + "loss": 0.1769, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07219933718442917, + "step": 9075, + "valid_targets_mean": 2852.0, + "valid_targets_min": 993 + }, + { + "epoch": 6.604583484903602, + "grad_norm": 0.6628576413218599, + "learning_rate": 3.908669782671237e-07, + "loss": 0.1768, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08581328392028809, + "step": 9080, + "valid_targets_mean": 3059.8, + "valid_targets_min": 1275 + }, + { + "epoch": 6.608221171335031, + "grad_norm": 0.5869051523912915, + "learning_rate": 3.8376389829756664e-07, + "loss": 0.1794, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09547068923711777, + "step": 9085, + "valid_targets_mean": 4479.0, + "valid_targets_min": 2250 + }, + { + "epoch": 6.61185885776646, + "grad_norm": 0.6660959699835518, + "learning_rate": 3.7672532712698993e-07, + "loss": 0.1836, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11186611652374268, + "step": 9090, + "valid_targets_mean": 4952.2, + "valid_targets_min": 2892 + }, + { + "epoch": 6.615496544197891, + "grad_norm": 0.6490573016814258, + "learning_rate": 3.6975128790202264e-07, + "loss": 0.1867, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08971784263849258, + "step": 9095, + "valid_targets_mean": 4026.4, + "valid_targets_min": 2935 + }, + { + "epoch": 6.61913423062932, + "grad_norm": 0.6566725374730383, + "learning_rate": 3.628418035570791e-07, + "loss": 0.1815, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08526936173439026, + "step": 9100, + "valid_targets_mean": 4085.2, + "valid_targets_min": 1817 + }, + { + "epoch": 6.622771917060749, + "grad_norm": 0.687262585856772, + "learning_rate": 3.5599689681428795e-07, + "loss": 0.1806, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09663105010986328, + "step": 9105, + "valid_targets_mean": 4050.4, + "valid_targets_min": 1727 + }, + { + "epoch": 6.626409603492179, + "grad_norm": 0.5847492835096706, + "learning_rate": 3.4921659018340324e-07, + "loss": 0.1791, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06836828589439392, + "step": 9110, + "valid_targets_mean": 3682.9, + "valid_targets_min": 1622 + }, + { + "epoch": 6.630047289923609, + "grad_norm": 0.4829075946287008, + "learning_rate": 3.4250090596174233e-07, + "loss": 0.1451, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06295397877693176, + "step": 9115, + "valid_targets_mean": 2920.6, + "valid_targets_min": 1521 + }, + { + "epoch": 6.633684976355038, + "grad_norm": 0.4188319274789394, + "learning_rate": 3.3584986623410585e-07, + "loss": 0.0885, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03607305884361267, + "step": 9120, + "valid_targets_mean": 3724.9, + "valid_targets_min": 2590 + }, + { + "epoch": 6.637322662786468, + "grad_norm": 0.45408731341075553, + "learning_rate": 3.292634928727179e-07, + "loss": 0.0743, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03840715438127518, + "step": 9125, + "valid_targets_mean": 3167.9, + "valid_targets_min": 982 + }, + { + "epoch": 6.640960349217897, + "grad_norm": 0.49218723637784195, + "learning_rate": 3.227418075371325e-07, + "loss": 0.0813, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08046534657478333, + "step": 9130, + "valid_targets_mean": 2996.4, + "valid_targets_min": 927 + }, + { + "epoch": 6.644598035649327, + "grad_norm": 0.4004646084693425, + "learning_rate": 3.162848316741829e-07, + "loss": 0.0731, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.026624709367752075, + "step": 9135, + "valid_targets_mean": 1756.1, + "valid_targets_min": 641 + }, + { + "epoch": 6.648235722080757, + "grad_norm": 0.48509485256157103, + "learning_rate": 3.09892586517897e-07, + "loss": 0.076, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.039298541843891144, + "step": 9140, + "valid_targets_mean": 2745.2, + "valid_targets_min": 859 + }, + { + "epoch": 6.651873408512186, + "grad_norm": 0.8226369284933834, + "learning_rate": 3.035650930894396e-07, + "loss": 0.1142, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05786203593015671, + "step": 9145, + "valid_targets_mean": 1268.5, + "valid_targets_min": 539 + }, + { + "epoch": 6.655511094943616, + "grad_norm": 0.4205283614904237, + "learning_rate": 2.9730237219703026e-07, + "loss": 0.0902, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.052812378853559494, + "step": 9150, + "valid_targets_mean": 2812.6, + "valid_targets_min": 920 + }, + { + "epoch": 6.659148781375046, + "grad_norm": 0.4609554731204089, + "learning_rate": 2.9110444443588573e-07, + "loss": 0.0827, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04518943279981613, + "step": 9155, + "valid_targets_mean": 3495.0, + "valid_targets_min": 2295 + }, + { + "epoch": 6.662786467806475, + "grad_norm": 0.4057459581550531, + "learning_rate": 2.8497133018814407e-07, + "loss": 0.1105, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029872875660657883, + "step": 9160, + "valid_targets_mean": 3228.4, + "valid_targets_min": 738 + }, + { + "epoch": 6.666424154237904, + "grad_norm": 0.42399592613624437, + "learning_rate": 2.789030496228051e-07, + "loss": 0.0616, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04588189721107483, + "step": 9165, + "valid_targets_mean": 4290.0, + "valid_targets_min": 2706 + }, + { + "epoch": 6.670061840669335, + "grad_norm": 0.42798093628332406, + "learning_rate": 2.728996226956593e-07, + "loss": 0.0735, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0461779460310936, + "step": 9170, + "valid_targets_mean": 4529.8, + "valid_targets_min": 3259 + }, + { + "epoch": 6.673699527100764, + "grad_norm": 0.37787736483656453, + "learning_rate": 2.6696106914922525e-07, + "loss": 0.0723, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03311358764767647, + "step": 9175, + "valid_targets_mean": 4198.2, + "valid_targets_min": 518 + }, + { + "epoch": 6.677337213532193, + "grad_norm": 0.653185314880701, + "learning_rate": 2.6108740851267467e-07, + "loss": 0.1134, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.042892590165138245, + "step": 9180, + "valid_targets_mean": 1873.5, + "valid_targets_min": 664 + }, + { + "epoch": 6.680974899963623, + "grad_norm": 0.4158302064596771, + "learning_rate": 2.5527866010178536e-07, + "loss": 0.0669, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.034662507474422455, + "step": 9185, + "valid_targets_mean": 3589.2, + "valid_targets_min": 2724 + }, + { + "epoch": 6.684612586395053, + "grad_norm": 0.7018412103888166, + "learning_rate": 2.49534843018866e-07, + "loss": 0.0747, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0314330980181694, + "step": 9190, + "valid_targets_mean": 1644.0, + "valid_targets_min": 518 + }, + { + "epoch": 6.688250272826482, + "grad_norm": 0.38952181755726645, + "learning_rate": 2.4385597615269376e-07, + "loss": 0.0651, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.032368022948503494, + "step": 9195, + "valid_targets_mean": 3376.5, + "valid_targets_min": 844 + }, + { + "epoch": 6.691887959257912, + "grad_norm": 0.47773355017499686, + "learning_rate": 2.382420781784589e-07, + "loss": 0.0746, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0320374071598053, + "step": 9200, + "valid_targets_mean": 2660.2, + "valid_targets_min": 577 + }, + { + "epoch": 6.695525645689342, + "grad_norm": 0.41659383675320244, + "learning_rate": 2.3269316755768933e-07, + "loss": 0.0778, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.035704582929611206, + "step": 9205, + "valid_targets_mean": 2450.9, + "valid_targets_min": 1081 + }, + { + "epoch": 6.699163332120771, + "grad_norm": 0.4573982425735664, + "learning_rate": 2.2720926253821273e-07, + "loss": 0.0844, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0348493792116642, + "step": 9210, + "valid_targets_mean": 2564.8, + "valid_targets_min": 721 + }, + { + "epoch": 6.702801018552201, + "grad_norm": 0.4308688957110196, + "learning_rate": 2.2179038115407003e-07, + "loss": 0.0816, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03617878630757332, + "step": 9215, + "valid_targets_mean": 2766.4, + "valid_targets_min": 615 + }, + { + "epoch": 6.70643870498363, + "grad_norm": 0.563014159585323, + "learning_rate": 2.1643654122548207e-07, + "loss": 0.1224, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11430550366640091, + "step": 9220, + "valid_targets_mean": 3693.1, + "valid_targets_min": 2609 + }, + { + "epoch": 6.71007639141506, + "grad_norm": 0.32087251642077796, + "learning_rate": 2.1114776035876748e-07, + "loss": 0.0685, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02483009174466133, + "step": 9225, + "valid_targets_mean": 4102.4, + "valid_targets_min": 3732 + }, + { + "epoch": 6.71371407784649, + "grad_norm": 0.45708757753216245, + "learning_rate": 2.059240559463005e-07, + "loss": 0.0755, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029797296971082687, + "step": 9230, + "valid_targets_mean": 1369.8, + "valid_targets_min": 731 + }, + { + "epoch": 6.717351764277919, + "grad_norm": 0.4493964457447723, + "learning_rate": 2.0076544516644646e-07, + "loss": 0.1046, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.039776761084795, + "step": 9235, + "valid_targets_mean": 3517.8, + "valid_targets_min": 1812 + }, + { + "epoch": 6.720989450709348, + "grad_norm": 0.39688880094998147, + "learning_rate": 1.9567194498351093e-07, + "loss": 0.0894, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03570055961608887, + "step": 9240, + "valid_targets_mean": 4099.4, + "valid_targets_min": 3344 + }, + { + "epoch": 6.724627137140779, + "grad_norm": 0.7493925622737605, + "learning_rate": 1.906435721476796e-07, + "loss": 0.098, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11900442093610764, + "step": 9245, + "valid_targets_mean": 2014.0, + "valid_targets_min": 568 + }, + { + "epoch": 6.728264823572208, + "grad_norm": 0.518641874459602, + "learning_rate": 1.8568034319495832e-07, + "loss": 0.2527, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04781832545995712, + "step": 9250, + "valid_targets_mean": 3413.2, + "valid_targets_min": 1248 + }, + { + "epoch": 6.731902510003637, + "grad_norm": 0.4531927456387117, + "learning_rate": 1.8078227444713548e-07, + "loss": 0.083, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05080346018075943, + "step": 9255, + "valid_targets_mean": 3592.5, + "valid_targets_min": 2829 + }, + { + "epoch": 6.735540196435068, + "grad_norm": 0.5351916510048517, + "learning_rate": 1.7594938201170863e-07, + "loss": 0.0813, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05140184238553047, + "step": 9260, + "valid_targets_mean": 2591.2, + "valid_targets_min": 798 + }, + { + "epoch": 6.739177882866497, + "grad_norm": 0.45235678251011147, + "learning_rate": 1.7118168178184236e-07, + "loss": 0.1718, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0378408320248127, + "step": 9265, + "valid_targets_mean": 3274.5, + "valid_targets_min": 2594 + }, + { + "epoch": 6.742815569297926, + "grad_norm": 0.2457391014758876, + "learning_rate": 1.6647918943631937e-07, + "loss": 0.0635, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.023171255365014076, + "step": 9270, + "valid_targets_mean": 5338.6, + "valid_targets_min": 3239 + }, + { + "epoch": 6.746453255729357, + "grad_norm": 0.37304170888374344, + "learning_rate": 1.6184192043947166e-07, + "loss": 0.086, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.024363970384001732, + "step": 9275, + "valid_targets_mean": 3009.1, + "valid_targets_min": 569 + }, + { + "epoch": 6.750090942160786, + "grad_norm": 0.4220294863705037, + "learning_rate": 1.5726989004115622e-07, + "loss": 0.0807, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.032408908009529114, + "step": 9280, + "valid_targets_mean": 3551.1, + "valid_targets_min": 846 + }, + { + "epoch": 6.753728628592215, + "grad_norm": 0.46928130739258733, + "learning_rate": 1.5276311327667936e-07, + "loss": 0.0707, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04280420020222664, + "step": 9285, + "valid_targets_mean": 3529.1, + "valid_targets_min": 1352 + }, + { + "epoch": 6.757366315023645, + "grad_norm": 0.45718177518624864, + "learning_rate": 1.4832160496676128e-07, + "loss": 0.0682, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03570609167218208, + "step": 9290, + "valid_targets_mean": 2796.4, + "valid_targets_min": 536 + }, + { + "epoch": 6.761004001455074, + "grad_norm": 0.4113565986350081, + "learning_rate": 1.4394537971748946e-07, + "loss": 0.0687, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03309350088238716, + "step": 9295, + "valid_targets_mean": 3629.5, + "valid_targets_min": 2264 + }, + { + "epoch": 6.764641687886504, + "grad_norm": 0.4054869645020173, + "learning_rate": 1.3963445192025417e-07, + "loss": 0.0666, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.031219732016324997, + "step": 9300, + "valid_targets_mean": 3847.5, + "valid_targets_min": 1865 + }, + { + "epoch": 6.768279374317934, + "grad_norm": 0.7157715751277037, + "learning_rate": 1.3538883575171967e-07, + "loss": 0.104, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18031588196754456, + "step": 9305, + "valid_targets_mean": 2020.4, + "valid_targets_min": 732 + }, + { + "epoch": 6.771917060749363, + "grad_norm": 0.41762773267627895, + "learning_rate": 1.3120854517377324e-07, + "loss": 0.0726, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03796899691224098, + "step": 9310, + "valid_targets_mean": 3760.4, + "valid_targets_min": 2719 + }, + { + "epoch": 6.775554747180793, + "grad_norm": 0.5699094068265208, + "learning_rate": 1.2709359393347165e-07, + "loss": 0.06, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04004668444395065, + "step": 9315, + "valid_targets_mean": 1354.6, + "valid_targets_min": 722 + }, + { + "epoch": 6.779192433612223, + "grad_norm": 0.36035656018653267, + "learning_rate": 1.2304399556299917e-07, + "loss": 0.0632, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.024145206436514854, + "step": 9320, + "valid_targets_mean": 3962.9, + "valid_targets_min": 3167 + }, + { + "epoch": 6.782830120043652, + "grad_norm": 0.38845018590094216, + "learning_rate": 1.1905976337962533e-07, + "loss": 0.0631, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02289532497525215, + "step": 9325, + "valid_targets_mean": 2448.9, + "valid_targets_min": 802 + }, + { + "epoch": 6.786467806475082, + "grad_norm": 0.41021506577652317, + "learning_rate": 1.1514091048566489e-07, + "loss": 0.1079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03551233932375908, + "step": 9330, + "valid_targets_mean": 3174.1, + "valid_targets_min": 908 + }, + { + "epoch": 6.790105492906512, + "grad_norm": 0.3813583839694162, + "learning_rate": 1.1128744976842465e-07, + "loss": 0.0872, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05414692685008049, + "step": 9335, + "valid_targets_mean": 3718.0, + "valid_targets_min": 3104 + }, + { + "epoch": 6.793743179337941, + "grad_norm": 0.3181141339876425, + "learning_rate": 1.0749939390017006e-07, + "loss": 0.0835, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.019986433908343315, + "step": 9340, + "valid_targets_mean": 2740.0, + "valid_targets_min": 598 + }, + { + "epoch": 6.79738086576937, + "grad_norm": 0.9877345623576118, + "learning_rate": 1.0377675533807641e-07, + "loss": 0.0913, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05370524153113365, + "step": 9345, + "valid_targets_mean": 819.4, + "valid_targets_min": 578 + }, + { + "epoch": 6.801018552200801, + "grad_norm": 0.6074460743159467, + "learning_rate": 1.0011954632419551e-07, + "loss": 0.1213, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04436647146940231, + "step": 9350, + "valid_targets_mean": 1079.6, + "valid_targets_min": 695 + }, + { + "epoch": 6.80465623863223, + "grad_norm": 0.4669299099588689, + "learning_rate": 9.652777888541131e-08, + "loss": 0.0782, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04199977219104767, + "step": 9355, + "valid_targets_mean": 2324.0, + "valid_targets_min": 845 + }, + { + "epoch": 6.808293925063659, + "grad_norm": 0.42983394018600557, + "learning_rate": 9.300146483339323e-08, + "loss": 0.0924, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.032017309218645096, + "step": 9360, + "valid_targets_mean": 3022.2, + "valid_targets_min": 728 + }, + { + "epoch": 6.811931611495089, + "grad_norm": 0.6786555953200223, + "learning_rate": 8.954061576457396e-08, + "loss": 0.0733, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04938740283250809, + "step": 9365, + "valid_targets_mean": 1847.6, + "valid_targets_min": 820 + }, + { + "epoch": 6.815569297926519, + "grad_norm": 0.3367145106275689, + "learning_rate": 8.614524306009842e-08, + "loss": 0.0714, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0326002761721611, + "step": 9370, + "valid_targets_mean": 4479.2, + "valid_targets_min": 4044 + }, + { + "epoch": 6.819206984357948, + "grad_norm": 0.3653207811504584, + "learning_rate": 8.28153578857882e-08, + "loss": 0.0809, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.023903831839561462, + "step": 9375, + "valid_targets_mean": 4048.2, + "valid_targets_min": 580 + }, + { + "epoch": 6.822844670789378, + "grad_norm": 0.4328550873187656, + "learning_rate": 7.955097119210609e-08, + "loss": 0.0668, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.032907962799072266, + "step": 9380, + "valid_targets_mean": 2770.8, + "valid_targets_min": 459 + }, + { + "epoch": 6.826482357220808, + "grad_norm": 0.45105345434999033, + "learning_rate": 7.635209371412489e-08, + "loss": 0.064, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04325108230113983, + "step": 9385, + "valid_targets_mean": 4135.4, + "valid_targets_min": 2558 + }, + { + "epoch": 6.830120043652237, + "grad_norm": 0.470202379731346, + "learning_rate": 7.32187359714831e-08, + "loss": 0.0743, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05957060307264328, + "step": 9390, + "valid_targets_mean": 2926.4, + "valid_targets_min": 791 + }, + { + "epoch": 6.833757730083667, + "grad_norm": 0.46315594805418897, + "learning_rate": 7.015090826835603e-08, + "loss": 0.0836, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04875386506319046, + "step": 9395, + "valid_targets_mean": 3447.4, + "valid_targets_min": 1165 + }, + { + "epoch": 6.837395416515096, + "grad_norm": 0.4256070080742633, + "learning_rate": 6.714862069342021e-08, + "loss": 0.0841, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04151131212711334, + "step": 9400, + "valid_targets_mean": 3704.6, + "valid_targets_min": 2901 + }, + { + "epoch": 6.841033102946526, + "grad_norm": 0.388875212869588, + "learning_rate": 6.421188311982684e-08, + "loss": 0.0711, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.026393532752990723, + "step": 9405, + "valid_targets_mean": 3872.1, + "valid_targets_min": 2674 + }, + { + "epoch": 6.844670789377956, + "grad_norm": 0.4877925473100469, + "learning_rate": 6.13407052051529e-08, + "loss": 0.0729, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04247027635574341, + "step": 9410, + "valid_targets_mean": 2819.6, + "valid_targets_min": 502 + }, + { + "epoch": 6.848308475809385, + "grad_norm": 0.40684272644281727, + "learning_rate": 5.853509639138777e-08, + "loss": 0.0799, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.022995442152023315, + "step": 9415, + "valid_targets_mean": 2963.4, + "valid_targets_min": 724 + }, + { + "epoch": 6.851946162240814, + "grad_norm": 0.37459418196960753, + "learning_rate": 5.5795065904888926e-08, + "loss": 0.0598, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.027180785313248634, + "step": 9420, + "valid_targets_mean": 3650.0, + "valid_targets_min": 2769 + }, + { + "epoch": 6.855583848672245, + "grad_norm": 0.5256105807005791, + "learning_rate": 5.312062275636187e-08, + "loss": 0.0777, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.037905678153038025, + "step": 9425, + "valid_targets_mean": 3052.1, + "valid_targets_min": 591 + }, + { + "epoch": 6.859221535103674, + "grad_norm": 0.4399879505968927, + "learning_rate": 5.051177574082244e-08, + "loss": 0.0756, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.036372147500514984, + "step": 9430, + "valid_targets_mean": 2622.5, + "valid_targets_min": 387 + }, + { + "epoch": 6.862859221535103, + "grad_norm": 0.4906322414227212, + "learning_rate": 4.796853343757013e-08, + "loss": 0.0828, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04660686105489731, + "step": 9435, + "valid_targets_mean": 3589.4, + "valid_targets_min": 1648 + }, + { + "epoch": 6.866496907966534, + "grad_norm": 0.43520351600878515, + "learning_rate": 4.5490904210161443e-08, + "loss": 0.0786, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0321415439248085, + "step": 9440, + "valid_targets_mean": 1455.0, + "valid_targets_min": 457 + }, + { + "epoch": 6.870134594397963, + "grad_norm": 0.8352379790499505, + "learning_rate": 4.30788962063855e-08, + "loss": 0.082, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04850626736879349, + "step": 9445, + "valid_targets_mean": 878.2, + "valid_targets_min": 482 + }, + { + "epoch": 6.873772280829392, + "grad_norm": 0.6413813064801166, + "learning_rate": 4.073251735822403e-08, + "loss": 0.0755, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06545744091272354, + "step": 9450, + "valid_targets_mean": 2136.0, + "valid_targets_min": 808 + }, + { + "epoch": 6.877409967260822, + "grad_norm": 0.4124383222897515, + "learning_rate": 3.845177538184697e-08, + "loss": 0.082, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03021131455898285, + "step": 9455, + "valid_targets_mean": 2712.4, + "valid_targets_min": 1010 + }, + { + "epoch": 6.881047653692252, + "grad_norm": 0.4680565351324999, + "learning_rate": 3.6236677777561345e-08, + "loss": 0.0764, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04208286106586456, + "step": 9460, + "valid_targets_mean": 2711.9, + "valid_targets_min": 902 + }, + { + "epoch": 6.884685340123681, + "grad_norm": 0.5403717502789942, + "learning_rate": 3.408723182981355e-08, + "loss": 0.0661, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04137636721134186, + "step": 9465, + "valid_targets_mean": 1882.1, + "valid_targets_min": 526 + }, + { + "epoch": 6.888323026555111, + "grad_norm": 0.5881896219700764, + "learning_rate": 3.200344460713822e-08, + "loss": 0.0847, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.034964919090270996, + "step": 9470, + "valid_targets_mean": 1127.9, + "valid_targets_min": 637 + }, + { + "epoch": 6.89196071298654, + "grad_norm": 0.37572025271463305, + "learning_rate": 2.9985322962160504e-08, + "loss": 0.075, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.026309413835406303, + "step": 9475, + "valid_targets_mean": 3718.1, + "valid_targets_min": 1440 + }, + { + "epoch": 6.89559839941797, + "grad_norm": 0.40841486705109825, + "learning_rate": 2.8032873531553818e-08, + "loss": 0.067, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047584839165210724, + "step": 9480, + "valid_targets_mean": 4313.0, + "valid_targets_min": 2341 + }, + { + "epoch": 6.8992360858494, + "grad_norm": 0.3378152130728914, + "learning_rate": 2.61461027360288e-08, + "loss": 0.0704, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.033950164914131165, + "step": 9485, + "valid_targets_mean": 3635.4, + "valid_targets_min": 762 + }, + { + "epoch": 6.902873772280829, + "grad_norm": 0.4034880311024454, + "learning_rate": 2.4325016780308853e-08, + "loss": 0.065, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.022354822605848312, + "step": 9490, + "valid_targets_mean": 1935.6, + "valid_targets_min": 667 + }, + { + "epoch": 6.906511458712259, + "grad_norm": 0.4058324253899549, + "learning_rate": 2.256962165311016e-08, + "loss": 0.0602, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04055086895823479, + "step": 9495, + "valid_targets_mean": 4719.8, + "valid_targets_min": 4077 + }, + { + "epoch": 6.910149145143689, + "grad_norm": 0.3602890022835493, + "learning_rate": 2.0879923127123946e-08, + "loss": 0.0811, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.035397812724113464, + "step": 9500, + "valid_targets_mean": 3550.5, + "valid_targets_min": 784 + }, + { + "epoch": 6.913786831575118, + "grad_norm": 0.5486788646380928, + "learning_rate": 1.9255926758989795e-08, + "loss": 0.0919, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06386945396661758, + "step": 9505, + "valid_targets_mean": 3344.2, + "valid_targets_min": 1090 + }, + { + "epoch": 6.9174245180065475, + "grad_norm": 0.4401998297506189, + "learning_rate": 1.76976378892868e-08, + "loss": 0.0707, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03954967111349106, + "step": 9510, + "valid_targets_mean": 3255.9, + "valid_targets_min": 1189 + }, + { + "epoch": 6.921062204437978, + "grad_norm": 0.3289939582172149, + "learning_rate": 1.6205061642513565e-08, + "loss": 0.0798, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.021834682673215866, + "step": 9515, + "valid_targets_mean": 2780.1, + "valid_targets_min": 929 + }, + { + "epoch": 6.924699890869407, + "grad_norm": 0.36641212387206007, + "learning_rate": 1.4778202927068219e-08, + "loss": 0.0967, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.025437243282794952, + "step": 9520, + "valid_targets_mean": 3147.2, + "valid_targets_min": 987 + }, + { + "epoch": 6.928337577300836, + "grad_norm": 0.5174544251213739, + "learning_rate": 1.3417066435232885e-08, + "loss": 0.0682, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04704616218805313, + "step": 9525, + "valid_targets_mean": 3708.8, + "valid_targets_min": 3019 + }, + { + "epoch": 6.931975263732266, + "grad_norm": 0.4294588685047999, + "learning_rate": 1.2121656643164781e-08, + "loss": 0.0745, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03142224997282028, + "step": 9530, + "valid_targets_mean": 3040.9, + "valid_targets_min": 1045 + }, + { + "epoch": 6.935612950163696, + "grad_norm": 0.39783434547952773, + "learning_rate": 1.0891977810867372e-08, + "loss": 0.0698, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0497271828353405, + "step": 9535, + "valid_targets_mean": 2372.4, + "valid_targets_min": 647 + }, + { + "epoch": 6.939250636595125, + "grad_norm": 0.6018337177297728, + "learning_rate": 9.728033982192575e-09, + "loss": 0.1126, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07510608434677124, + "step": 9540, + "valid_targets_mean": 1277.0, + "valid_targets_min": 570 + }, + { + "epoch": 6.942888323026555, + "grad_norm": 0.48338398098958507, + "learning_rate": 8.629828984818568e-09, + "loss": 0.0887, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04306703060865402, + "step": 9545, + "valid_targets_mean": 3175.0, + "valid_targets_min": 722 + }, + { + "epoch": 6.946526009457985, + "grad_norm": 0.44000572465185683, + "learning_rate": 7.597366430234232e-09, + "loss": 0.0709, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030978510156273842, + "step": 9550, + "valid_targets_mean": 2600.4, + "valid_targets_min": 635 + }, + { + "epoch": 6.950163695889414, + "grad_norm": 0.3892064046693163, + "learning_rate": 6.630649713739168e-09, + "loss": 0.0705, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03366043418645859, + "step": 9555, + "valid_targets_mean": 3322.4, + "valid_targets_min": 2174 + }, + { + "epoch": 6.953801382320844, + "grad_norm": 0.3801568754305974, + "learning_rate": 5.729682014417037e-09, + "loss": 0.0695, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03285752609372139, + "step": 9560, + "valid_targets_mean": 3355.9, + "valid_targets_min": 854 + }, + { + "epoch": 6.957439068752274, + "grad_norm": 0.4389275595824972, + "learning_rate": 4.89446629514001e-09, + "loss": 0.0621, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.028004275634884834, + "step": 9565, + "valid_targets_mean": 2574.4, + "valid_targets_min": 785 + }, + { + "epoch": 6.961076755183703, + "grad_norm": 0.6430477158351365, + "learning_rate": 4.125005302548779e-09, + "loss": 0.089, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03752051293849945, + "step": 9570, + "valid_targets_mean": 1653.1, + "valid_targets_min": 766 + }, + { + "epoch": 6.964714441615133, + "grad_norm": 0.438519965583921, + "learning_rate": 3.421301567045898e-09, + "loss": 0.063, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03767339512705803, + "step": 9575, + "valid_targets_mean": 3234.5, + "valid_targets_min": 797 + }, + { + "epoch": 6.968352128046562, + "grad_norm": 0.4494038020911933, + "learning_rate": 2.783357402793563e-09, + "loss": 0.0689, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0329766646027565, + "step": 9580, + "valid_targets_mean": 2719.8, + "valid_targets_min": 786 + }, + { + "epoch": 6.971989814477992, + "grad_norm": 0.4587911169577132, + "learning_rate": 2.211174907693625e-09, + "loss": 0.0697, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04496292397379875, + "step": 9585, + "valid_targets_mean": 2937.1, + "valid_targets_min": 1825 + }, + { + "epoch": 6.975627500909422, + "grad_norm": 0.3519513203868607, + "learning_rate": 1.7047559633920353e-09, + "loss": 0.0752, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03976215422153473, + "step": 9590, + "valid_targets_mean": 4866.9, + "valid_targets_min": 2274 + }, + { + "epoch": 6.979265187340851, + "grad_norm": 0.4603975067317736, + "learning_rate": 1.26410223527218e-09, + "loss": 0.0629, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03335491567850113, + "step": 9595, + "valid_targets_mean": 2550.5, + "valid_targets_min": 678 + }, + { + "epoch": 6.9829028737722805, + "grad_norm": 0.49065513549844203, + "learning_rate": 8.89215172437119e-10, + "loss": 0.0703, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05341671407222748, + "step": 9600, + "valid_targets_mean": 3518.4, + "valid_targets_min": 1017 + }, + { + "epoch": 6.986540560203711, + "grad_norm": 0.5669924323887776, + "learning_rate": 5.800960077206874e-10, + "loss": 0.0963, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0694001093506813, + "step": 9605, + "valid_targets_mean": 2348.9, + "valid_targets_min": 1535 + }, + { + "epoch": 6.99017824663514, + "grad_norm": 0.37709660181478566, + "learning_rate": 3.3674575767417283e-10, + "loss": 0.1259, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.028170611709356308, + "step": 9610, + "valid_targets_mean": 3666.5, + "valid_targets_min": 1097 + }, + { + "epoch": 6.9938159330665695, + "grad_norm": 0.35959927921085916, + "learning_rate": 1.591652225663154e-10, + "loss": 0.0713, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.026346255093812943, + "step": 9615, + "valid_targets_mean": 3338.4, + "valid_targets_min": 2453 + }, + { + "epoch": 6.997453619498, + "grad_norm": 0.35229591190575504, + "learning_rate": 4.735498637442604e-11, + "loss": 0.0655, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.026771588250994682, + "step": 9620, + "valid_targets_mean": 3709.8, + "valid_targets_min": 2555 + }, + { + "epoch": 7.0, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11451305449008942, + "step": 9624, + "total_flos": 3.536242678171697e+18, + "train_loss": 0.04252522938566612, + "train_runtime": 67682.3505, + "train_samples_per_second": 2.274, + "train_steps_per_second": 0.142, + "valid_targets_mean": 7922.2, + "valid_targets_min": 3513 + } + ], + "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": 3.536242678171697e+18, + "train_batch_size": 1, + "trial_name": null, + "trial_params": null +}