diff --git "a/trainer_state.json" "b/trainer_state.json" new file mode 100644--- /dev/null +++ "b/trainer_state.json" @@ -0,0 +1,20122 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 7.0, + "eval_steps": 500, + "global_step": 9127, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.0038358266206367474, + "grad_norm": 16.734935497428065, + "learning_rate": 1.752464403066813e-07, + "loss": 0.6602, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.30977684259414673, + "step": 5, + "valid_targets_mean": 7389.0, + "valid_targets_min": 5627 + }, + { + "epoch": 0.007671653241273495, + "grad_norm": 16.47501680957501, + "learning_rate": 3.943044906900329e-07, + "loss": 0.7005, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2366279661655426, + "step": 10, + "valid_targets_mean": 1294.8, + "valid_targets_min": 422 + }, + { + "epoch": 0.011507479861910242, + "grad_norm": 15.013850200341626, + "learning_rate": 6.133625410733845e-07, + "loss": 0.6797, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.32489901781082153, + "step": 15, + "valid_targets_mean": 8539.9, + "valid_targets_min": 6761 + }, + { + "epoch": 0.01534330648254699, + "grad_norm": 13.453022819682884, + "learning_rate": 8.324205914567362e-07, + "loss": 0.6144, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29006752371788025, + "step": 20, + "valid_targets_mean": 7579.0, + "valid_targets_min": 6174 + }, + { + "epoch": 0.019179133103183737, + "grad_norm": 11.55765841573356, + "learning_rate": 1.0514786418400876e-06, + "loss": 0.6149, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29434677958488464, + "step": 25, + "valid_targets_mean": 7328.2, + "valid_targets_min": 4839 + }, + { + "epoch": 0.023014959723820484, + "grad_norm": 8.288523937443522, + "learning_rate": 1.2705366922234393e-06, + "loss": 0.6088, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3221447169780731, + "step": 30, + "valid_targets_mean": 6838.2, + "valid_targets_min": 5034 + }, + { + "epoch": 0.02685078634445723, + "grad_norm": 5.984685658990618, + "learning_rate": 1.4895947426067908e-06, + "loss": 0.5899, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29333993792533875, + "step": 35, + "valid_targets_mean": 7248.2, + "valid_targets_min": 5567 + }, + { + "epoch": 0.03068661296509398, + "grad_norm": 5.377661797668344, + "learning_rate": 1.7086527929901425e-06, + "loss": 0.5468, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26035892963409424, + "step": 40, + "valid_targets_mean": 6234.0, + "valid_targets_min": 5603 + }, + { + "epoch": 0.03452243958573072, + "grad_norm": 6.674265700552884, + "learning_rate": 1.9277108433734943e-06, + "loss": 0.5424, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1257585883140564, + "step": 45, + "valid_targets_mean": 602.0, + "valid_targets_min": 152 + }, + { + "epoch": 0.038358266206367474, + "grad_norm": 4.009823370124004, + "learning_rate": 2.1467688937568456e-06, + "loss": 0.4697, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2261408418416977, + "step": 50, + "valid_targets_mean": 7299.9, + "valid_targets_min": 6158 + }, + { + "epoch": 0.04219409282700422, + "grad_norm": 2.2963089011322135, + "learning_rate": 2.3658269441401973e-06, + "loss": 0.4497, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21069017052650452, + "step": 55, + "valid_targets_mean": 6140.0, + "valid_targets_min": 4648 + }, + { + "epoch": 0.04602991944764097, + "grad_norm": 1.5374086974072751, + "learning_rate": 2.584884994523549e-06, + "loss": 0.4296, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21293629705905914, + "step": 60, + "valid_targets_mean": 6603.0, + "valid_targets_min": 4882 + }, + { + "epoch": 0.04986574606827771, + "grad_norm": 1.0882981389974995, + "learning_rate": 2.8039430449069006e-06, + "loss": 0.4061, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20768508315086365, + "step": 65, + "valid_targets_mean": 8015.0, + "valid_targets_min": 6344 + }, + { + "epoch": 0.05370157268891446, + "grad_norm": 0.9562516789313762, + "learning_rate": 3.0230010952902523e-06, + "loss": 0.3939, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19166187942028046, + "step": 70, + "valid_targets_mean": 6305.2, + "valid_targets_min": 4622 + }, + { + "epoch": 0.05753739930955121, + "grad_norm": 0.812861503298975, + "learning_rate": 3.242059145673604e-06, + "loss": 0.3959, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20208412408828735, + "step": 75, + "valid_targets_mean": 7890.8, + "valid_targets_min": 5485 + }, + { + "epoch": 0.06137322593018796, + "grad_norm": 0.6900432805927756, + "learning_rate": 3.4611171960569557e-06, + "loss": 0.3907, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1776198446750641, + "step": 80, + "valid_targets_mean": 7016.6, + "valid_targets_min": 4960 + }, + { + "epoch": 0.06520905255082471, + "grad_norm": 0.6600086507458631, + "learning_rate": 3.680175246440307e-06, + "loss": 0.3767, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17716944217681885, + "step": 85, + "valid_targets_mean": 6527.1, + "valid_targets_min": 5399 + }, + { + "epoch": 0.06904487917146145, + "grad_norm": 0.7244473697831765, + "learning_rate": 3.899233296823659e-06, + "loss": 0.3745, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18706785142421722, + "step": 90, + "valid_targets_mean": 4995.0, + "valid_targets_min": 618 + }, + { + "epoch": 0.0728807057920982, + "grad_norm": 0.5995688902821968, + "learning_rate": 4.11829134720701e-06, + "loss": 0.3839, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19158749282360077, + "step": 95, + "valid_targets_mean": 7219.4, + "valid_targets_min": 5764 + }, + { + "epoch": 0.07671653241273495, + "grad_norm": 0.5284136453933524, + "learning_rate": 4.337349397590362e-06, + "loss": 0.3681, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1916421353816986, + "step": 100, + "valid_targets_mean": 7998.5, + "valid_targets_min": 5902 + }, + { + "epoch": 0.0805523590333717, + "grad_norm": 0.529789660804858, + "learning_rate": 4.556407447973713e-06, + "loss": 0.3333, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14918303489685059, + "step": 105, + "valid_targets_mean": 7377.4, + "valid_targets_min": 5015 + }, + { + "epoch": 0.08438818565400844, + "grad_norm": 0.482906978545994, + "learning_rate": 4.7754654983570645e-06, + "loss": 0.3351, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14809924364089966, + "step": 110, + "valid_targets_mean": 6794.0, + "valid_targets_min": 5283 + }, + { + "epoch": 0.08822401227464519, + "grad_norm": 0.5405740652873254, + "learning_rate": 4.9945235487404166e-06, + "loss": 0.3488, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19710281491279602, + "step": 115, + "valid_targets_mean": 6656.4, + "valid_targets_min": 4364 + }, + { + "epoch": 0.09205983889528194, + "grad_norm": 0.4819331829885772, + "learning_rate": 5.213581599123768e-06, + "loss": 0.3353, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15957677364349365, + "step": 120, + "valid_targets_mean": 7178.4, + "valid_targets_min": 4448 + }, + { + "epoch": 0.09589566551591867, + "grad_norm": 0.5104868287021335, + "learning_rate": 5.43263964950712e-06, + "loss": 0.3291, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1861436814069748, + "step": 125, + "valid_targets_mean": 7619.8, + "valid_targets_min": 5737 + }, + { + "epoch": 0.09973149213655542, + "grad_norm": 0.4975508996084737, + "learning_rate": 5.651697699890471e-06, + "loss": 0.2907, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14112728834152222, + "step": 130, + "valid_targets_mean": 8754.4, + "valid_targets_min": 5752 + }, + { + "epoch": 0.10356731875719218, + "grad_norm": 0.501575305710743, + "learning_rate": 5.870755750273823e-06, + "loss": 0.2891, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1584150195121765, + "step": 135, + "valid_targets_mean": 6341.0, + "valid_targets_min": 4543 + }, + { + "epoch": 0.10740314537782893, + "grad_norm": 0.9149702547321956, + "learning_rate": 6.089813800657175e-06, + "loss": 0.2774, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15664801001548767, + "step": 140, + "valid_targets_mean": 7181.9, + "valid_targets_min": 4563 + }, + { + "epoch": 0.11123897199846566, + "grad_norm": 1.5878920658680282, + "learning_rate": 6.308871851040527e-06, + "loss": 0.3386, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12577520310878754, + "step": 145, + "valid_targets_mean": 1104.4, + "valid_targets_min": 161 + }, + { + "epoch": 0.11507479861910241, + "grad_norm": 0.42462381600999033, + "learning_rate": 6.527929901423879e-06, + "loss": 0.3234, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14038443565368652, + "step": 150, + "valid_targets_mean": 7018.2, + "valid_targets_min": 5424 + }, + { + "epoch": 0.11891062523973916, + "grad_norm": 0.46620730012167644, + "learning_rate": 6.746987951807229e-06, + "loss": 0.3214, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1642071008682251, + "step": 155, + "valid_targets_mean": 7738.4, + "valid_targets_min": 5329 + }, + { + "epoch": 0.12274645186037592, + "grad_norm": 0.6027829892168501, + "learning_rate": 6.96604600219058e-06, + "loss": 0.3276, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15978729724884033, + "step": 160, + "valid_targets_mean": 6426.6, + "valid_targets_min": 5560 + }, + { + "epoch": 0.12658227848101267, + "grad_norm": 0.5009827063334908, + "learning_rate": 7.1851040525739325e-06, + "loss": 0.3189, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16298848390579224, + "step": 165, + "valid_targets_mean": 6169.9, + "valid_targets_min": 3446 + }, + { + "epoch": 0.13041810510164942, + "grad_norm": 0.500938767243081, + "learning_rate": 7.404162102957284e-06, + "loss": 0.3033, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15571677684783936, + "step": 170, + "valid_targets_mean": 7204.6, + "valid_targets_min": 5381 + }, + { + "epoch": 0.13425393172228614, + "grad_norm": 0.48906062676040096, + "learning_rate": 7.623220153340636e-06, + "loss": 0.3075, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1569436490535736, + "step": 175, + "valid_targets_mean": 7600.4, + "valid_targets_min": 5241 + }, + { + "epoch": 0.1380897583429229, + "grad_norm": 0.4990705077464035, + "learning_rate": 7.842278203723988e-06, + "loss": 0.2989, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16396324336528778, + "step": 180, + "valid_targets_mean": 7098.5, + "valid_targets_min": 5607 + }, + { + "epoch": 0.14192558496355964, + "grad_norm": 0.5368314758300365, + "learning_rate": 8.06133625410734e-06, + "loss": 0.3039, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1584191620349884, + "step": 185, + "valid_targets_mean": 6955.1, + "valid_targets_min": 5530 + }, + { + "epoch": 0.1457614115841964, + "grad_norm": 0.5440848484709342, + "learning_rate": 8.28039430449069e-06, + "loss": 0.2973, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15149888396263123, + "step": 190, + "valid_targets_mean": 6748.1, + "valid_targets_min": 5364 + }, + { + "epoch": 0.14959723820483314, + "grad_norm": 0.5079119708240225, + "learning_rate": 8.499452354874043e-06, + "loss": 0.3009, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14683541655540466, + "step": 195, + "valid_targets_mean": 6273.9, + "valid_targets_min": 5438 + }, + { + "epoch": 0.1534330648254699, + "grad_norm": 0.4598919295175204, + "learning_rate": 8.718510405257395e-06, + "loss": 0.3009, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14699074625968933, + "step": 200, + "valid_targets_mean": 6419.5, + "valid_targets_min": 4879 + }, + { + "epoch": 0.15726889144610665, + "grad_norm": 0.48210922440009263, + "learning_rate": 8.937568455640747e-06, + "loss": 0.2968, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14629018306732178, + "step": 205, + "valid_targets_mean": 6483.4, + "valid_targets_min": 4309 + }, + { + "epoch": 0.1611047180667434, + "grad_norm": 0.500177216476559, + "learning_rate": 9.156626506024097e-06, + "loss": 0.2952, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16482025384902954, + "step": 210, + "valid_targets_mean": 6825.0, + "valid_targets_min": 5205 + }, + { + "epoch": 0.16494054468738012, + "grad_norm": 0.4718729419336097, + "learning_rate": 9.375684556407448e-06, + "loss": 0.2893, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13331450521945953, + "step": 215, + "valid_targets_mean": 6729.4, + "valid_targets_min": 5200 + }, + { + "epoch": 0.16877637130801687, + "grad_norm": 0.48812974131059683, + "learning_rate": 9.5947426067908e-06, + "loss": 0.2862, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14071591198444366, + "step": 220, + "valid_targets_mean": 6094.0, + "valid_targets_min": 5302 + }, + { + "epoch": 0.17261219792865362, + "grad_norm": 0.5141184588361866, + "learning_rate": 9.813800657174152e-06, + "loss": 0.2893, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14972206950187683, + "step": 225, + "valid_targets_mean": 7140.0, + "valid_targets_min": 5270 + }, + { + "epoch": 0.17644802454929037, + "grad_norm": 0.49735108309191883, + "learning_rate": 1.0032858707557504e-05, + "loss": 0.2791, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13243457674980164, + "step": 230, + "valid_targets_mean": 7425.0, + "valid_targets_min": 5281 + }, + { + "epoch": 0.18028385116992712, + "grad_norm": 0.5143180680992367, + "learning_rate": 1.0251916757940854e-05, + "loss": 0.2861, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17353805899620056, + "step": 235, + "valid_targets_mean": 7063.4, + "valid_targets_min": 5386 + }, + { + "epoch": 0.18411967779056387, + "grad_norm": 0.4935873548180119, + "learning_rate": 1.0470974808324206e-05, + "loss": 0.2862, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.150490403175354, + "step": 240, + "valid_targets_mean": 6827.0, + "valid_targets_min": 4547 + }, + { + "epoch": 0.18795550441120062, + "grad_norm": 0.8719200771797969, + "learning_rate": 1.0690032858707557e-05, + "loss": 0.2647, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09142240136861801, + "step": 245, + "valid_targets_mean": 2047.0, + "valid_targets_min": 165 + }, + { + "epoch": 0.19179133103183735, + "grad_norm": 0.4679253078958614, + "learning_rate": 1.0909090909090909e-05, + "loss": 0.2756, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1393997073173523, + "step": 250, + "valid_targets_mean": 6754.1, + "valid_targets_min": 5196 + }, + { + "epoch": 0.1956271576524741, + "grad_norm": 0.797382393697359, + "learning_rate": 1.1128148959474261e-05, + "loss": 0.2757, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13992202281951904, + "step": 255, + "valid_targets_mean": 5985.6, + "valid_targets_min": 5028 + }, + { + "epoch": 0.19946298427311085, + "grad_norm": 0.5173897429554096, + "learning_rate": 1.1347207009857613e-05, + "loss": 0.2808, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12789398431777954, + "step": 260, + "valid_targets_mean": 5963.9, + "valid_targets_min": 5237 + }, + { + "epoch": 0.2032988108937476, + "grad_norm": 0.48599651417524753, + "learning_rate": 1.1566265060240964e-05, + "loss": 0.2751, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14384804666042328, + "step": 265, + "valid_targets_mean": 6319.0, + "valid_targets_min": 5412 + }, + { + "epoch": 0.20713463751438435, + "grad_norm": 0.457291552531284, + "learning_rate": 1.1785323110624316e-05, + "loss": 0.2608, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1122150719165802, + "step": 270, + "valid_targets_mean": 7125.8, + "valid_targets_min": 5080 + }, + { + "epoch": 0.2109704641350211, + "grad_norm": 0.5277078931727495, + "learning_rate": 1.2004381161007668e-05, + "loss": 0.2757, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13621293008327484, + "step": 275, + "valid_targets_mean": 6123.5, + "valid_targets_min": 4521 + }, + { + "epoch": 0.21480629075565785, + "grad_norm": 2.8110309777301614, + "learning_rate": 1.222343921139102e-05, + "loss": 0.3343, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24442338943481445, + "step": 280, + "valid_targets_mean": 1886.0, + "valid_targets_min": 1111 + }, + { + "epoch": 0.2186421173762946, + "grad_norm": 1.5473742647756052, + "learning_rate": 1.2442497261774372e-05, + "loss": 0.4158, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.208450049161911, + "step": 285, + "valid_targets_mean": 1497.6, + "valid_targets_min": 893 + }, + { + "epoch": 0.22247794399693133, + "grad_norm": 1.1623872694058264, + "learning_rate": 1.2661555312157722e-05, + "loss": 0.3653, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12746557593345642, + "step": 290, + "valid_targets_mean": 1356.0, + "valid_targets_min": 914 + }, + { + "epoch": 0.22631377061756808, + "grad_norm": 1.2708853262987865, + "learning_rate": 1.2880613362541074e-05, + "loss": 0.3475, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1466427445411682, + "step": 295, + "valid_targets_mean": 1324.5, + "valid_targets_min": 680 + }, + { + "epoch": 0.23014959723820483, + "grad_norm": 1.0243911663720016, + "learning_rate": 1.3099671412924427e-05, + "loss": 0.337, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15540817379951477, + "step": 300, + "valid_targets_mean": 1931.8, + "valid_targets_min": 876 + }, + { + "epoch": 0.23398542385884158, + "grad_norm": 1.0619169151116759, + "learning_rate": 1.3318729463307779e-05, + "loss": 0.3226, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14255644381046295, + "step": 305, + "valid_targets_mean": 1292.2, + "valid_targets_min": 734 + }, + { + "epoch": 0.23782125047947833, + "grad_norm": 0.9642244234813864, + "learning_rate": 1.3537787513691129e-05, + "loss": 0.3273, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12808261811733246, + "step": 310, + "valid_targets_mean": 1469.6, + "valid_targets_min": 655 + }, + { + "epoch": 0.24165707710011508, + "grad_norm": 1.1669025321111033, + "learning_rate": 1.3756845564074481e-05, + "loss": 0.2963, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15637479722499847, + "step": 315, + "valid_targets_mean": 1548.4, + "valid_targets_min": 737 + }, + { + "epoch": 0.24549290372075183, + "grad_norm": 1.0913528100040142, + "learning_rate": 1.3975903614457833e-05, + "loss": 0.3139, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1346641480922699, + "step": 320, + "valid_targets_mean": 1316.5, + "valid_targets_min": 759 + }, + { + "epoch": 0.24932873034138858, + "grad_norm": 1.1508242462369775, + "learning_rate": 1.4194961664841185e-05, + "loss": 0.2929, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11255344748497009, + "step": 325, + "valid_targets_mean": 1249.6, + "valid_targets_min": 753 + }, + { + "epoch": 0.25316455696202533, + "grad_norm": 1.0216178090704038, + "learning_rate": 1.4414019715224536e-05, + "loss": 0.2917, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18646572530269623, + "step": 330, + "valid_targets_mean": 1833.0, + "valid_targets_min": 927 + }, + { + "epoch": 0.2570003835826621, + "grad_norm": 1.0544385666551388, + "learning_rate": 1.4633077765607888e-05, + "loss": 0.3017, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14336952567100525, + "step": 335, + "valid_targets_mean": 1338.4, + "valid_targets_min": 679 + }, + { + "epoch": 0.26083621020329883, + "grad_norm": 1.3409182233026566, + "learning_rate": 1.485213581599124e-05, + "loss": 0.2892, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12809902429580688, + "step": 340, + "valid_targets_mean": 1244.6, + "valid_targets_min": 771 + }, + { + "epoch": 0.2646720368239356, + "grad_norm": 1.029929305831354, + "learning_rate": 1.5071193866374592e-05, + "loss": 0.2705, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12857429683208466, + "step": 345, + "valid_targets_mean": 1732.5, + "valid_targets_min": 986 + }, + { + "epoch": 0.2685078634445723, + "grad_norm": 1.0372241760025775, + "learning_rate": 1.5290251916757942e-05, + "loss": 0.2785, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12367382645606995, + "step": 350, + "valid_targets_mean": 1389.9, + "valid_targets_min": 1103 + }, + { + "epoch": 0.27234369006520903, + "grad_norm": 1.0345105326920743, + "learning_rate": 1.5509309967141293e-05, + "loss": 0.2818, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15311776101589203, + "step": 355, + "valid_targets_mean": 1593.2, + "valid_targets_min": 675 + }, + { + "epoch": 0.2761795166858458, + "grad_norm": 1.0249265002293595, + "learning_rate": 1.5728368017524643e-05, + "loss": 0.279, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17026084661483765, + "step": 360, + "valid_targets_mean": 1926.6, + "valid_targets_min": 974 + }, + { + "epoch": 0.28001534330648253, + "grad_norm": 1.0757630662547273, + "learning_rate": 1.5947426067907997e-05, + "loss": 0.2851, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16669753193855286, + "step": 365, + "valid_targets_mean": 1953.1, + "valid_targets_min": 869 + }, + { + "epoch": 0.2838511699271193, + "grad_norm": 1.0658120443443597, + "learning_rate": 1.6166484118291347e-05, + "loss": 0.2818, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12514963746070862, + "step": 370, + "valid_targets_mean": 1327.2, + "valid_targets_min": 1023 + }, + { + "epoch": 0.28768699654775604, + "grad_norm": 0.963302521411966, + "learning_rate": 1.63855421686747e-05, + "loss": 0.2841, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15024003386497498, + "step": 375, + "valid_targets_mean": 1717.9, + "valid_targets_min": 709 + }, + { + "epoch": 0.2915228231683928, + "grad_norm": 1.256054332078691, + "learning_rate": 1.660460021905805e-05, + "loss": 0.2694, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1166786253452301, + "step": 380, + "valid_targets_mean": 1139.9, + "valid_targets_min": 658 + }, + { + "epoch": 0.29535864978902954, + "grad_norm": 1.0588427544350336, + "learning_rate": 1.6823658269441402e-05, + "loss": 0.2791, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13326531648635864, + "step": 385, + "valid_targets_mean": 1444.8, + "valid_targets_min": 662 + }, + { + "epoch": 0.2991944764096663, + "grad_norm": 1.1021955817900322, + "learning_rate": 1.7042716319824756e-05, + "loss": 0.2846, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09856607019901276, + "step": 390, + "valid_targets_mean": 1242.2, + "valid_targets_min": 859 + }, + { + "epoch": 0.30303030303030304, + "grad_norm": 1.0141903876246645, + "learning_rate": 1.7261774370208106e-05, + "loss": 0.2566, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15726488828659058, + "step": 395, + "valid_targets_mean": 1569.2, + "valid_targets_min": 950 + }, + { + "epoch": 0.3068661296509398, + "grad_norm": 1.0719904271579026, + "learning_rate": 1.7480832420591457e-05, + "loss": 0.2695, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12624512612819672, + "step": 400, + "valid_targets_mean": 1539.8, + "valid_targets_min": 686 + }, + { + "epoch": 0.31070195627157654, + "grad_norm": 0.9183456608084654, + "learning_rate": 1.769989047097481e-05, + "loss": 0.2628, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14598432183265686, + "step": 405, + "valid_targets_mean": 1610.6, + "valid_targets_min": 702 + }, + { + "epoch": 0.3145377828922133, + "grad_norm": 0.998835392306809, + "learning_rate": 1.791894852135816e-05, + "loss": 0.2574, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09772234410047531, + "step": 410, + "valid_targets_mean": 1190.4, + "valid_targets_min": 779 + }, + { + "epoch": 0.31837360951285004, + "grad_norm": 1.049527742447451, + "learning_rate": 1.8138006571741515e-05, + "loss": 0.275, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13988231122493744, + "step": 415, + "valid_targets_mean": 1467.1, + "valid_targets_min": 822 + }, + { + "epoch": 0.3222094361334868, + "grad_norm": 1.2055607306265592, + "learning_rate": 1.8357064622124865e-05, + "loss": 0.252, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11006900668144226, + "step": 420, + "valid_targets_mean": 1350.5, + "valid_targets_min": 868 + }, + { + "epoch": 0.3260452627541235, + "grad_norm": 1.0984063055270805, + "learning_rate": 1.8576122672508216e-05, + "loss": 0.2625, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1078415960073471, + "step": 425, + "valid_targets_mean": 1283.5, + "valid_targets_min": 781 + }, + { + "epoch": 0.32988108937476024, + "grad_norm": 0.9801484266671924, + "learning_rate": 1.879518072289157e-05, + "loss": 0.2636, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10969997942447662, + "step": 430, + "valid_targets_mean": 1303.2, + "valid_targets_min": 793 + }, + { + "epoch": 0.333716915995397, + "grad_norm": 1.050665285644886, + "learning_rate": 1.901423877327492e-05, + "loss": 0.2609, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11194390803575516, + "step": 435, + "valid_targets_mean": 1309.1, + "valid_targets_min": 806 + }, + { + "epoch": 0.33755274261603374, + "grad_norm": 0.9490761264213051, + "learning_rate": 1.923329682365827e-05, + "loss": 0.2587, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10698190331459045, + "step": 440, + "valid_targets_mean": 1397.6, + "valid_targets_min": 803 + }, + { + "epoch": 0.3413885692366705, + "grad_norm": 1.0145479533241633, + "learning_rate": 1.9452354874041624e-05, + "loss": 0.2552, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09630389511585236, + "step": 445, + "valid_targets_mean": 1187.8, + "valid_targets_min": 660 + }, + { + "epoch": 0.34522439585730724, + "grad_norm": 1.026120087997192, + "learning_rate": 1.9671412924424974e-05, + "loss": 0.2384, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10423795133829117, + "step": 450, + "valid_targets_mean": 1547.0, + "valid_targets_min": 853 + }, + { + "epoch": 0.349060222477944, + "grad_norm": 0.9717511359693604, + "learning_rate": 1.9890470974808328e-05, + "loss": 0.2577, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13358120620250702, + "step": 455, + "valid_targets_mean": 1704.2, + "valid_targets_min": 719 + }, + { + "epoch": 0.35289604909858074, + "grad_norm": 0.9522714065389901, + "learning_rate": 2.0109529025191675e-05, + "loss": 0.2594, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09310548007488251, + "step": 460, + "valid_targets_mean": 1255.0, + "valid_targets_min": 765 + }, + { + "epoch": 0.3567318757192175, + "grad_norm": 0.9407842169365302, + "learning_rate": 2.032858707557503e-05, + "loss": 0.2533, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12559077143669128, + "step": 465, + "valid_targets_mean": 1550.8, + "valid_targets_min": 875 + }, + { + "epoch": 0.36056770233985425, + "grad_norm": 1.0489012269303666, + "learning_rate": 2.054764512595838e-05, + "loss": 0.2516, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12418130040168762, + "step": 470, + "valid_targets_mean": 1362.0, + "valid_targets_min": 739 + }, + { + "epoch": 0.364403528960491, + "grad_norm": 1.00888735991545, + "learning_rate": 2.0766703176341733e-05, + "loss": 0.2608, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1448429822921753, + "step": 475, + "valid_targets_mean": 1666.8, + "valid_targets_min": 760 + }, + { + "epoch": 0.36823935558112775, + "grad_norm": 1.050858527274344, + "learning_rate": 2.0985761226725084e-05, + "loss": 0.2443, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10663163661956787, + "step": 480, + "valid_targets_mean": 1319.6, + "valid_targets_min": 1046 + }, + { + "epoch": 0.3720751822017645, + "grad_norm": 0.9262489910051214, + "learning_rate": 2.1204819277108437e-05, + "loss": 0.2634, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1415494829416275, + "step": 485, + "valid_targets_mean": 1727.9, + "valid_targets_min": 851 + }, + { + "epoch": 0.37591100882240125, + "grad_norm": 1.0225125315809178, + "learning_rate": 2.1423877327491788e-05, + "loss": 0.2621, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13438208401203156, + "step": 490, + "valid_targets_mean": 1643.2, + "valid_targets_min": 907 + }, + { + "epoch": 0.379746835443038, + "grad_norm": 1.0035335714307403, + "learning_rate": 2.164293537787514e-05, + "loss": 0.2672, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1140209287405014, + "step": 495, + "valid_targets_mean": 1478.5, + "valid_targets_min": 764 + }, + { + "epoch": 0.3835826620636747, + "grad_norm": 0.9064416863587725, + "learning_rate": 2.186199342825849e-05, + "loss": 0.2465, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1154487282037735, + "step": 500, + "valid_targets_mean": 1522.6, + "valid_targets_min": 822 + }, + { + "epoch": 0.38741848868431145, + "grad_norm": 0.9612163746054043, + "learning_rate": 2.2081051478641842e-05, + "loss": 0.246, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12227274477481842, + "step": 505, + "valid_targets_mean": 1589.0, + "valid_targets_min": 982 + }, + { + "epoch": 0.3912543153049482, + "grad_norm": 1.042342007817629, + "learning_rate": 2.2300109529025193e-05, + "loss": 0.2586, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12591767311096191, + "step": 510, + "valid_targets_mean": 1574.8, + "valid_targets_min": 816 + }, + { + "epoch": 0.39509014192558495, + "grad_norm": 1.024469729994657, + "learning_rate": 2.2519167579408543e-05, + "loss": 0.2476, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12528133392333984, + "step": 515, + "valid_targets_mean": 1422.0, + "valid_targets_min": 721 + }, + { + "epoch": 0.3989259685462217, + "grad_norm": 1.001616568292112, + "learning_rate": 2.2738225629791897e-05, + "loss": 0.2401, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09153550118207932, + "step": 520, + "valid_targets_mean": 1150.6, + "valid_targets_min": 741 + }, + { + "epoch": 0.40276179516685845, + "grad_norm": 1.215812688596626, + "learning_rate": 2.2957283680175247e-05, + "loss": 0.2491, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13487711548805237, + "step": 525, + "valid_targets_mean": 1684.2, + "valid_targets_min": 734 + }, + { + "epoch": 0.4065976217874952, + "grad_norm": 0.9572506952104678, + "learning_rate": 2.31763417305586e-05, + "loss": 0.2518, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1491519808769226, + "step": 530, + "valid_targets_mean": 1937.1, + "valid_targets_min": 744 + }, + { + "epoch": 0.41043344840813195, + "grad_norm": 0.9585098902258874, + "learning_rate": 2.3395399780941948e-05, + "loss": 0.249, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14101262390613556, + "step": 535, + "valid_targets_mean": 1842.2, + "valid_targets_min": 620 + }, + { + "epoch": 0.4142692750287687, + "grad_norm": 0.9666762178177184, + "learning_rate": 2.3614457831325302e-05, + "loss": 0.2422, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14275363087654114, + "step": 540, + "valid_targets_mean": 1637.6, + "valid_targets_min": 785 + }, + { + "epoch": 0.41810510164940545, + "grad_norm": 0.8139323772658648, + "learning_rate": 2.3833515881708652e-05, + "loss": 0.2472, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12145279347896576, + "step": 545, + "valid_targets_mean": 1969.9, + "valid_targets_min": 740 + }, + { + "epoch": 0.4219409282700422, + "grad_norm": 1.0510633532641227, + "learning_rate": 2.4052573932092006e-05, + "loss": 0.2429, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11080403625965118, + "step": 550, + "valid_targets_mean": 1432.1, + "valid_targets_min": 711 + }, + { + "epoch": 0.42577675489067895, + "grad_norm": 0.9455904180936224, + "learning_rate": 2.4271631982475357e-05, + "loss": 0.2436, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13420063257217407, + "step": 555, + "valid_targets_mean": 1695.6, + "valid_targets_min": 1122 + }, + { + "epoch": 0.4296125815113157, + "grad_norm": 1.1118682257404255, + "learning_rate": 2.449069003285871e-05, + "loss": 0.2412, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12985703349113464, + "step": 560, + "valid_targets_mean": 1524.5, + "valid_targets_min": 722 + }, + { + "epoch": 0.43344840813195246, + "grad_norm": 1.0262499120030573, + "learning_rate": 2.470974808324206e-05, + "loss": 0.2276, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10387195646762848, + "step": 565, + "valid_targets_mean": 1289.2, + "valid_targets_min": 661 + }, + { + "epoch": 0.4372842347525892, + "grad_norm": 1.0446972341050431, + "learning_rate": 2.4928806133625415e-05, + "loss": 0.2339, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11148473620414734, + "step": 570, + "valid_targets_mean": 1519.5, + "valid_targets_min": 1011 + }, + { + "epoch": 0.44112006137322596, + "grad_norm": 0.9766382395801256, + "learning_rate": 2.5147864184008765e-05, + "loss": 0.2449, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1290050595998764, + "step": 575, + "valid_targets_mean": 1731.4, + "valid_targets_min": 1277 + }, + { + "epoch": 0.44495588799386265, + "grad_norm": 0.9844663713651677, + "learning_rate": 2.536692223439212e-05, + "loss": 0.2319, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12323303520679474, + "step": 580, + "valid_targets_mean": 1486.1, + "valid_targets_min": 807 + }, + { + "epoch": 0.4487917146144994, + "grad_norm": 0.9800717091314614, + "learning_rate": 2.5585980284775466e-05, + "loss": 0.2451, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13714760541915894, + "step": 585, + "valid_targets_mean": 1549.9, + "valid_targets_min": 679 + }, + { + "epoch": 0.45262754123513615, + "grad_norm": 0.9646267942215914, + "learning_rate": 2.580503833515882e-05, + "loss": 0.2404, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11489401757717133, + "step": 590, + "valid_targets_mean": 1656.0, + "valid_targets_min": 1294 + }, + { + "epoch": 0.4564633678557729, + "grad_norm": 1.045999685951308, + "learning_rate": 2.602409638554217e-05, + "loss": 0.2412, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10923351347446442, + "step": 595, + "valid_targets_mean": 1368.8, + "valid_targets_min": 790 + }, + { + "epoch": 0.46029919447640966, + "grad_norm": 1.0016015033264485, + "learning_rate": 2.6243154435925524e-05, + "loss": 0.2366, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1331334263086319, + "step": 600, + "valid_targets_mean": 1765.8, + "valid_targets_min": 592 + }, + { + "epoch": 0.4641350210970464, + "grad_norm": 0.9578349364041602, + "learning_rate": 2.6462212486308874e-05, + "loss": 0.2419, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12723052501678467, + "step": 605, + "valid_targets_mean": 1586.8, + "valid_targets_min": 662 + }, + { + "epoch": 0.46797084771768316, + "grad_norm": 1.0216894556368463, + "learning_rate": 2.6681270536692228e-05, + "loss": 0.2459, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11931857466697693, + "step": 610, + "valid_targets_mean": 1336.9, + "valid_targets_min": 622 + }, + { + "epoch": 0.4718066743383199, + "grad_norm": 0.9861569322179687, + "learning_rate": 2.690032858707558e-05, + "loss": 0.2425, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07638566195964813, + "step": 615, + "valid_targets_mean": 1097.1, + "valid_targets_min": 625 + }, + { + "epoch": 0.47564250095895666, + "grad_norm": 0.9616924030236756, + "learning_rate": 2.7119386637458932e-05, + "loss": 0.2553, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09577345848083496, + "step": 620, + "valid_targets_mean": 1484.9, + "valid_targets_min": 888 + }, + { + "epoch": 0.4794783275795934, + "grad_norm": 0.9671648035590452, + "learning_rate": 2.733844468784228e-05, + "loss": 0.247, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1297050267457962, + "step": 625, + "valid_targets_mean": 1757.0, + "valid_targets_min": 635 + }, + { + "epoch": 0.48331415420023016, + "grad_norm": 1.0867329455358807, + "learning_rate": 2.755750273822563e-05, + "loss": 0.2496, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14081266522407532, + "step": 630, + "valid_targets_mean": 1476.1, + "valid_targets_min": 522 + }, + { + "epoch": 0.4871499808208669, + "grad_norm": 1.0631276524709687, + "learning_rate": 2.7776560788608983e-05, + "loss": 0.2355, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10674621164798737, + "step": 635, + "valid_targets_mean": 1230.9, + "valid_targets_min": 697 + }, + { + "epoch": 0.49098580744150366, + "grad_norm": 0.9104110395316168, + "learning_rate": 2.7995618838992334e-05, + "loss": 0.2332, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12736645340919495, + "step": 640, + "valid_targets_mean": 1671.0, + "valid_targets_min": 1010 + }, + { + "epoch": 0.4948216340621404, + "grad_norm": 1.002118535703065, + "learning_rate": 2.8214676889375688e-05, + "loss": 0.239, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11902131885290146, + "step": 645, + "valid_targets_mean": 1411.0, + "valid_targets_min": 769 + }, + { + "epoch": 0.49865746068277716, + "grad_norm": 0.8999569829256253, + "learning_rate": 2.8433734939759038e-05, + "loss": 0.2414, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10108102858066559, + "step": 650, + "valid_targets_mean": 1298.6, + "valid_targets_min": 745 + }, + { + "epoch": 0.5024932873034139, + "grad_norm": 1.0613703384439306, + "learning_rate": 2.8652792990142392e-05, + "loss": 0.2368, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10034029185771942, + "step": 655, + "valid_targets_mean": 1118.9, + "valid_targets_min": 680 + }, + { + "epoch": 0.5063291139240507, + "grad_norm": 0.8436411148121848, + "learning_rate": 2.887185104052574e-05, + "loss": 0.2297, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14048156142234802, + "step": 660, + "valid_targets_mean": 1905.9, + "valid_targets_min": 1347 + }, + { + "epoch": 0.5101649405446874, + "grad_norm": 0.9863620057734538, + "learning_rate": 2.9090909090909093e-05, + "loss": 0.2457, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17079246044158936, + "step": 665, + "valid_targets_mean": 1636.0, + "valid_targets_min": 851 + }, + { + "epoch": 0.5140007671653242, + "grad_norm": 1.0562726034341003, + "learning_rate": 2.9309967141292443e-05, + "loss": 0.231, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1036631390452385, + "step": 670, + "valid_targets_mean": 1457.2, + "valid_targets_min": 903 + }, + { + "epoch": 0.5178365937859609, + "grad_norm": 0.9314320767734602, + "learning_rate": 2.9529025191675797e-05, + "loss": 0.2285, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12763053178787231, + "step": 675, + "valid_targets_mean": 1674.4, + "valid_targets_min": 1240 + }, + { + "epoch": 0.5216724204065977, + "grad_norm": 0.7259273761211261, + "learning_rate": 2.9748083242059147e-05, + "loss": 0.2266, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09470324218273163, + "step": 680, + "valid_targets_mean": 1697.1, + "valid_targets_min": 917 + }, + { + "epoch": 0.5255082470272344, + "grad_norm": 0.9810786205637814, + "learning_rate": 2.99671412924425e-05, + "loss": 0.2133, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09834206104278564, + "step": 685, + "valid_targets_mean": 1580.9, + "valid_targets_min": 930 + }, + { + "epoch": 0.5293440736478712, + "grad_norm": 0.9738721624868321, + "learning_rate": 3.018619934282585e-05, + "loss": 0.2365, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12216811627149582, + "step": 690, + "valid_targets_mean": 1655.1, + "valid_targets_min": 1227 + }, + { + "epoch": 0.5331799002685078, + "grad_norm": 0.8408043642842551, + "learning_rate": 3.0405257393209205e-05, + "loss": 0.2311, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11649586260318756, + "step": 695, + "valid_targets_mean": 1565.8, + "valid_targets_min": 870 + }, + { + "epoch": 0.5370157268891446, + "grad_norm": 0.9547799956255943, + "learning_rate": 3.0624315443592556e-05, + "loss": 0.2347, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11278624832630157, + "step": 700, + "valid_targets_mean": 1564.6, + "valid_targets_min": 822 + }, + { + "epoch": 0.5408515535097813, + "grad_norm": 1.1171981247117677, + "learning_rate": 3.084337349397591e-05, + "loss": 0.2385, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10538198798894882, + "step": 705, + "valid_targets_mean": 1540.5, + "valid_targets_min": 1214 + }, + { + "epoch": 0.5446873801304181, + "grad_norm": 0.9881968540879705, + "learning_rate": 3.1062431544359257e-05, + "loss": 0.2274, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10644804686307907, + "step": 710, + "valid_targets_mean": 1268.6, + "valid_targets_min": 591 + }, + { + "epoch": 0.5485232067510548, + "grad_norm": 0.9469220607558081, + "learning_rate": 3.128148959474261e-05, + "loss": 0.2428, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08846846222877502, + "step": 715, + "valid_targets_mean": 1232.5, + "valid_targets_min": 646 + }, + { + "epoch": 0.5523590333716916, + "grad_norm": 0.8764748029513397, + "learning_rate": 3.150054764512596e-05, + "loss": 0.3383, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20138436555862427, + "step": 720, + "valid_targets_mean": 5097.2, + "valid_targets_min": 1971 + }, + { + "epoch": 0.5561948599923283, + "grad_norm": 0.6998875555834904, + "learning_rate": 3.171960569550932e-05, + "loss": 0.4058, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22296258807182312, + "step": 725, + "valid_targets_mean": 5217.0, + "valid_targets_min": 1631 + }, + { + "epoch": 0.5600306866129651, + "grad_norm": 0.6433036060831373, + "learning_rate": 3.1938663745892665e-05, + "loss": 0.3978, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.25297731161117554, + "step": 730, + "valid_targets_mean": 5966.0, + "valid_targets_min": 3294 + }, + { + "epoch": 0.5638665132336018, + "grad_norm": 0.5863796788815303, + "learning_rate": 3.215772179627602e-05, + "loss": 0.3769, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15809233486652374, + "step": 735, + "valid_targets_mean": 4246.1, + "valid_targets_min": 2736 + }, + { + "epoch": 0.5677023398542386, + "grad_norm": 0.6033511096991282, + "learning_rate": 3.2376779846659366e-05, + "loss": 0.3726, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.152872696518898, + "step": 740, + "valid_targets_mean": 3896.2, + "valid_targets_min": 1439 + }, + { + "epoch": 0.5715381664748753, + "grad_norm": 0.5863472612274271, + "learning_rate": 3.259583789704272e-05, + "loss": 0.3895, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14980095624923706, + "step": 745, + "valid_targets_mean": 4481.8, + "valid_targets_min": 1512 + }, + { + "epoch": 0.5753739930955121, + "grad_norm": 0.566112423601653, + "learning_rate": 3.281489594742607e-05, + "loss": 0.357, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1976863145828247, + "step": 750, + "valid_targets_mean": 5127.9, + "valid_targets_min": 2405 + }, + { + "epoch": 0.5792098197161488, + "grad_norm": 0.5515694985292244, + "learning_rate": 3.303395399780942e-05, + "loss": 0.3681, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14896513521671295, + "step": 755, + "valid_targets_mean": 3253.8, + "valid_targets_min": 1654 + }, + { + "epoch": 0.5830456463367856, + "grad_norm": 0.528369336444888, + "learning_rate": 3.3253012048192774e-05, + "loss": 0.3632, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.170309916138649, + "step": 760, + "valid_targets_mean": 5750.2, + "valid_targets_min": 4644 + }, + { + "epoch": 0.5868814729574223, + "grad_norm": 0.6391067952216819, + "learning_rate": 3.347207009857612e-05, + "loss": 0.3807, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18046733736991882, + "step": 765, + "valid_targets_mean": 4141.6, + "valid_targets_min": 1544 + }, + { + "epoch": 0.5907172995780591, + "grad_norm": 0.563497656727892, + "learning_rate": 3.3691128148959475e-05, + "loss": 0.3621, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17420697212219238, + "step": 770, + "valid_targets_mean": 5869.8, + "valid_targets_min": 953 + }, + { + "epoch": 0.5945531261986958, + "grad_norm": 0.5573566619484841, + "learning_rate": 3.391018619934283e-05, + "loss": 0.3629, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19688981771469116, + "step": 775, + "valid_targets_mean": 4973.6, + "valid_targets_min": 1265 + }, + { + "epoch": 0.5983889528193326, + "grad_norm": 0.5307044694458916, + "learning_rate": 3.412924424972618e-05, + "loss": 0.3591, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17081880569458008, + "step": 780, + "valid_targets_mean": 5359.9, + "valid_targets_min": 1982 + }, + { + "epoch": 0.6022247794399693, + "grad_norm": 0.5789909202143201, + "learning_rate": 3.434830230010953e-05, + "loss": 0.3647, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15739955008029938, + "step": 785, + "valid_targets_mean": 3892.5, + "valid_targets_min": 1759 + }, + { + "epoch": 0.6060606060606061, + "grad_norm": 0.7127170350394364, + "learning_rate": 3.456736035049288e-05, + "loss": 0.3719, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17377883195877075, + "step": 790, + "valid_targets_mean": 3185.8, + "valid_targets_min": 1451 + }, + { + "epoch": 0.6098964326812428, + "grad_norm": 0.6240766929929672, + "learning_rate": 3.478641840087624e-05, + "loss": 0.3673, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18613335490226746, + "step": 795, + "valid_targets_mean": 4283.2, + "valid_targets_min": 1861 + }, + { + "epoch": 0.6137322593018796, + "grad_norm": 0.6906312334295291, + "learning_rate": 3.500547645125959e-05, + "loss": 0.3637, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18791157007217407, + "step": 800, + "valid_targets_mean": 4340.8, + "valid_targets_min": 2345 + }, + { + "epoch": 0.6175680859225163, + "grad_norm": 0.663102172162758, + "learning_rate": 3.522453450164294e-05, + "loss": 0.3748, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19062283635139465, + "step": 805, + "valid_targets_mean": 3923.9, + "valid_targets_min": 2767 + }, + { + "epoch": 0.6214039125431531, + "grad_norm": 0.6936361684386131, + "learning_rate": 3.544359255202629e-05, + "loss": 0.3802, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20406955480575562, + "step": 810, + "valid_targets_mean": 3651.8, + "valid_targets_min": 1205 + }, + { + "epoch": 0.6252397391637898, + "grad_norm": 0.6405597426033891, + "learning_rate": 3.566265060240964e-05, + "loss": 0.3665, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15071913599967957, + "step": 815, + "valid_targets_mean": 3835.4, + "valid_targets_min": 2004 + }, + { + "epoch": 0.6290755657844266, + "grad_norm": 0.6069719042856164, + "learning_rate": 3.588170865279299e-05, + "loss": 0.3711, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17999251186847687, + "step": 820, + "valid_targets_mean": 3778.0, + "valid_targets_min": 1020 + }, + { + "epoch": 0.6329113924050633, + "grad_norm": 0.6519288883381018, + "learning_rate": 3.6100766703176346e-05, + "loss": 0.3504, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20555254817008972, + "step": 825, + "valid_targets_mean": 4307.8, + "valid_targets_min": 2308 + }, + { + "epoch": 0.6367472190257001, + "grad_norm": 0.5984637881073982, + "learning_rate": 3.63198247535597e-05, + "loss": 0.3466, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1959816813468933, + "step": 830, + "valid_targets_mean": 4553.9, + "valid_targets_min": 2046 + }, + { + "epoch": 0.6405830456463368, + "grad_norm": 0.6529203002456284, + "learning_rate": 3.653888280394305e-05, + "loss": 0.3495, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24104225635528564, + "step": 835, + "valid_targets_mean": 4660.2, + "valid_targets_min": 3373 + }, + { + "epoch": 0.6444188722669736, + "grad_norm": 0.6194718108989191, + "learning_rate": 3.67579408543264e-05, + "loss": 0.3616, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1746080368757248, + "step": 840, + "valid_targets_mean": 3727.4, + "valid_targets_min": 1747 + }, + { + "epoch": 0.6482546988876102, + "grad_norm": 0.6166549484939656, + "learning_rate": 3.697699890470975e-05, + "loss": 0.3628, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16639012098312378, + "step": 845, + "valid_targets_mean": 3326.1, + "valid_targets_min": 1570 + }, + { + "epoch": 0.652090525508247, + "grad_norm": 0.6249737365636636, + "learning_rate": 3.71960569550931e-05, + "loss": 0.3445, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16949501633644104, + "step": 850, + "valid_targets_mean": 4600.8, + "valid_targets_min": 1537 + }, + { + "epoch": 0.6559263521288837, + "grad_norm": 0.5740735753012594, + "learning_rate": 3.7415115005476456e-05, + "loss": 0.3455, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14278164505958557, + "step": 855, + "valid_targets_mean": 4328.9, + "valid_targets_min": 1934 + }, + { + "epoch": 0.6597621787495205, + "grad_norm": 0.6485404059104144, + "learning_rate": 3.76341730558598e-05, + "loss": 0.3547, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15070801973342896, + "step": 860, + "valid_targets_mean": 3494.0, + "valid_targets_min": 2562 + }, + { + "epoch": 0.6635980053701572, + "grad_norm": 0.6080892678836386, + "learning_rate": 3.7853231106243156e-05, + "loss": 0.3526, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1652260422706604, + "step": 865, + "valid_targets_mean": 4076.8, + "valid_targets_min": 2659 + }, + { + "epoch": 0.667433831990794, + "grad_norm": 0.6120351841668956, + "learning_rate": 3.807228915662651e-05, + "loss": 0.343, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16574092209339142, + "step": 870, + "valid_targets_mean": 4040.6, + "valid_targets_min": 1349 + }, + { + "epoch": 0.6712696586114307, + "grad_norm": 0.6918766996306709, + "learning_rate": 3.8291347207009864e-05, + "loss": 0.3471, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1311621069908142, + "step": 875, + "valid_targets_mean": 3000.6, + "valid_targets_min": 1567 + }, + { + "epoch": 0.6751054852320675, + "grad_norm": 0.729663637489808, + "learning_rate": 3.851040525739321e-05, + "loss": 0.3575, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14107397198677063, + "step": 880, + "valid_targets_mean": 2628.4, + "valid_targets_min": 1500 + }, + { + "epoch": 0.6789413118527042, + "grad_norm": 0.6234588542925856, + "learning_rate": 3.8729463307776565e-05, + "loss": 0.3371, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15540353953838348, + "step": 885, + "valid_targets_mean": 3413.6, + "valid_targets_min": 2033 + }, + { + "epoch": 0.682777138473341, + "grad_norm": 0.9542773758921617, + "learning_rate": 3.894852135815991e-05, + "loss": 0.3624, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2030370533466339, + "step": 890, + "valid_targets_mean": 2852.0, + "valid_targets_min": 1346 + }, + { + "epoch": 0.6866129650939777, + "grad_norm": 0.8105226064917814, + "learning_rate": 3.9167579408543266e-05, + "loss": 0.3535, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19626829028129578, + "step": 895, + "valid_targets_mean": 2981.6, + "valid_targets_min": 1821 + }, + { + "epoch": 0.6904487917146145, + "grad_norm": 0.6620126648426952, + "learning_rate": 3.938663745892662e-05, + "loss": 0.3473, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1847713589668274, + "step": 900, + "valid_targets_mean": 3951.2, + "valid_targets_min": 2535 + }, + { + "epoch": 0.6942846183352512, + "grad_norm": 0.8248491338379917, + "learning_rate": 3.960569550930997e-05, + "loss": 0.3423, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1623857319355011, + "step": 905, + "valid_targets_mean": 3048.8, + "valid_targets_min": 2080 + }, + { + "epoch": 0.698120444955888, + "grad_norm": 0.7339468582922652, + "learning_rate": 3.982475355969332e-05, + "loss": 0.3426, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15624918043613434, + "step": 910, + "valid_targets_mean": 2975.6, + "valid_targets_min": 1153 + }, + { + "epoch": 0.7019562715765247, + "grad_norm": 0.6634954511688771, + "learning_rate": 3.999999853753796e-05, + "loss": 0.3394, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1523067206144333, + "step": 915, + "valid_targets_mean": 3207.6, + "valid_targets_min": 1858 + }, + { + "epoch": 0.7057920981971615, + "grad_norm": 0.6485833193647668, + "learning_rate": 3.999994735138874e-05, + "loss": 0.3505, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1639857292175293, + "step": 920, + "valid_targets_mean": 3643.6, + "valid_targets_min": 2197 + }, + { + "epoch": 0.7096279248177982, + "grad_norm": 0.5998541129064924, + "learning_rate": 3.999982304235102e-05, + "loss": 0.3478, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13881850242614746, + "step": 925, + "valid_targets_mean": 3769.4, + "valid_targets_min": 1394 + }, + { + "epoch": 0.713463751438435, + "grad_norm": 0.57676614839692, + "learning_rate": 3.9999625610879285e-05, + "loss": 0.3436, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16524070501327515, + "step": 930, + "valid_targets_mean": 4348.6, + "valid_targets_min": 1844 + }, + { + "epoch": 0.7172995780590717, + "grad_norm": 0.6619277096149477, + "learning_rate": 3.999935505769538e-05, + "loss": 0.3405, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19352124631404877, + "step": 935, + "valid_targets_mean": 4341.1, + "valid_targets_min": 1734 + }, + { + "epoch": 0.7211354046797085, + "grad_norm": 0.6395236274821362, + "learning_rate": 3.999901138378847e-05, + "loss": 0.337, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1886102557182312, + "step": 940, + "valid_targets_mean": 4252.6, + "valid_targets_min": 2077 + }, + { + "epoch": 0.7249712313003452, + "grad_norm": 0.5804599482318901, + "learning_rate": 3.99985945904151e-05, + "loss": 0.3408, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1625012457370758, + "step": 945, + "valid_targets_mean": 4177.9, + "valid_targets_min": 2349 + }, + { + "epoch": 0.728807057920982, + "grad_norm": 0.5740833430428686, + "learning_rate": 3.999810467909913e-05, + "loss": 0.335, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18336337804794312, + "step": 950, + "valid_targets_mean": 4160.2, + "valid_targets_min": 1607 + }, + { + "epoch": 0.7326428845416187, + "grad_norm": 0.708554178426048, + "learning_rate": 3.999754165163174e-05, + "loss": 0.352, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18162031471729279, + "step": 955, + "valid_targets_mean": 3273.0, + "valid_targets_min": 1377 + }, + { + "epoch": 0.7364787111622555, + "grad_norm": 0.6179574464179778, + "learning_rate": 3.9996905510071454e-05, + "loss": 0.3523, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15163150429725647, + "step": 960, + "valid_targets_mean": 3304.5, + "valid_targets_min": 1902 + }, + { + "epoch": 0.7403145377828922, + "grad_norm": 0.5994828495336388, + "learning_rate": 3.99961962567441e-05, + "loss": 0.3393, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1502002775669098, + "step": 965, + "valid_targets_mean": 3324.1, + "valid_targets_min": 654 + }, + { + "epoch": 0.744150364403529, + "grad_norm": 0.5857197883367778, + "learning_rate": 3.999541389424281e-05, + "loss": 0.3412, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1639762967824936, + "step": 970, + "valid_targets_mean": 3907.8, + "valid_targets_min": 2607 + }, + { + "epoch": 0.7479861910241657, + "grad_norm": 0.5903239377528937, + "learning_rate": 3.999455842542805e-05, + "loss": 0.3438, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18204987049102783, + "step": 975, + "valid_targets_mean": 4467.2, + "valid_targets_min": 1739 + }, + { + "epoch": 0.7518220176448025, + "grad_norm": 0.6526446250860098, + "learning_rate": 3.9993629853427506e-05, + "loss": 0.3406, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14916029572486877, + "step": 980, + "valid_targets_mean": 3224.1, + "valid_targets_min": 1140 + }, + { + "epoch": 0.7556578442654392, + "grad_norm": 0.7620928004933628, + "learning_rate": 3.9992628181636204e-05, + "loss": 0.3199, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16705787181854248, + "step": 985, + "valid_targets_mean": 3543.4, + "valid_targets_min": 2501 + }, + { + "epoch": 0.759493670886076, + "grad_norm": 0.6249127777868125, + "learning_rate": 3.9991553413716414e-05, + "loss": 0.3335, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17633020877838135, + "step": 990, + "valid_targets_mean": 4364.1, + "valid_targets_min": 1298 + }, + { + "epoch": 0.7633294975067128, + "grad_norm": 0.5980931151405485, + "learning_rate": 3.9990405553597644e-05, + "loss": 0.3149, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16861015558242798, + "step": 995, + "valid_targets_mean": 4377.9, + "valid_targets_min": 1101 + }, + { + "epoch": 0.7671653241273494, + "grad_norm": 0.6417180167699816, + "learning_rate": 3.998918460547664e-05, + "loss": 0.3559, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1895521879196167, + "step": 1000, + "valid_targets_mean": 3894.2, + "valid_targets_min": 2079 + }, + { + "epoch": 0.7710011507479861, + "grad_norm": 1.059144681178496, + "learning_rate": 3.998789057381739e-05, + "loss": 0.3294, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13229483366012573, + "step": 1005, + "valid_targets_mean": 3054.4, + "valid_targets_min": 1743 + }, + { + "epoch": 0.7748369773686229, + "grad_norm": 0.5800052728185836, + "learning_rate": 3.9986523463351075e-05, + "loss": 0.3392, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18021135032176971, + "step": 1010, + "valid_targets_mean": 3688.0, + "valid_targets_min": 1234 + }, + { + "epoch": 0.7786728039892596, + "grad_norm": 0.612711097231469, + "learning_rate": 3.998508327907605e-05, + "loss": 0.3424, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17737287282943726, + "step": 1015, + "valid_targets_mean": 3790.0, + "valid_targets_min": 1822 + }, + { + "epoch": 0.7825086306098964, + "grad_norm": 0.5861354751606369, + "learning_rate": 3.9983570026257844e-05, + "loss": 0.3455, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17364472150802612, + "step": 1020, + "valid_targets_mean": 4106.9, + "valid_targets_min": 1651 + }, + { + "epoch": 0.7863444572305331, + "grad_norm": 0.6484452705470919, + "learning_rate": 3.998198371042917e-05, + "loss": 0.3372, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18413937091827393, + "step": 1025, + "valid_targets_mean": 3287.2, + "valid_targets_min": 1374 + }, + { + "epoch": 0.7901802838511699, + "grad_norm": 0.611932476515898, + "learning_rate": 3.9980324337389824e-05, + "loss": 0.3219, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17605528235435486, + "step": 1030, + "valid_targets_mean": 4399.2, + "valid_targets_min": 2725 + }, + { + "epoch": 0.7940161104718066, + "grad_norm": 0.6270742069544508, + "learning_rate": 3.9978591913206724e-05, + "loss": 0.3269, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17606815695762634, + "step": 1035, + "valid_targets_mean": 4269.0, + "valid_targets_min": 1966 + }, + { + "epoch": 0.7978519370924434, + "grad_norm": 0.6172862734338425, + "learning_rate": 3.9976786444213895e-05, + "loss": 0.3324, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18050189316272736, + "step": 1040, + "valid_targets_mean": 3716.2, + "valid_targets_min": 1561 + }, + { + "epoch": 0.8016877637130801, + "grad_norm": 0.6165906735088483, + "learning_rate": 3.99749079370124e-05, + "loss": 0.325, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16749659180641174, + "step": 1045, + "valid_targets_mean": 3750.1, + "valid_targets_min": 1444 + }, + { + "epoch": 0.8055235903337169, + "grad_norm": 0.5858940024657702, + "learning_rate": 3.997295639847036e-05, + "loss": 0.3119, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1426488757133484, + "step": 1050, + "valid_targets_mean": 3042.1, + "valid_targets_min": 1504 + }, + { + "epoch": 0.8093594169543536, + "grad_norm": 0.6392593873311246, + "learning_rate": 3.997093183572288e-05, + "loss": 0.3201, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15939974784851074, + "step": 1055, + "valid_targets_mean": 3153.0, + "valid_targets_min": 858 + }, + { + "epoch": 0.8131952435749904, + "grad_norm": 0.6256458730548907, + "learning_rate": 3.99688342561721e-05, + "loss": 0.3328, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18561862409114838, + "step": 1060, + "valid_targets_mean": 4063.5, + "valid_targets_min": 1288 + }, + { + "epoch": 0.8170310701956272, + "grad_norm": 0.5722986784047474, + "learning_rate": 3.996666366748708e-05, + "loss": 0.3542, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15849706530570984, + "step": 1065, + "valid_targets_mean": 3914.9, + "valid_targets_min": 1537 + }, + { + "epoch": 0.8208668968162639, + "grad_norm": 0.596570104135805, + "learning_rate": 3.996442007760382e-05, + "loss": 0.3265, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17926135659217834, + "step": 1070, + "valid_targets_mean": 4236.5, + "valid_targets_min": 2972 + }, + { + "epoch": 0.8247027234369007, + "grad_norm": 0.6315360399234435, + "learning_rate": 3.996210349472524e-05, + "loss": 0.3408, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18373149633407593, + "step": 1075, + "valid_targets_mean": 4411.0, + "valid_targets_min": 2013 + }, + { + "epoch": 0.8285385500575374, + "grad_norm": 0.6989642598213596, + "learning_rate": 3.9959713927321125e-05, + "loss": 0.3203, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16920846700668335, + "step": 1080, + "valid_targets_mean": 3548.4, + "valid_targets_min": 1744 + }, + { + "epoch": 0.8323743766781742, + "grad_norm": 0.5711513747114408, + "learning_rate": 3.99572513841281e-05, + "loss": 0.332, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14355367422103882, + "step": 1085, + "valid_targets_mean": 4129.6, + "valid_targets_min": 1914 + }, + { + "epoch": 0.8362102032988109, + "grad_norm": 0.6015431904585796, + "learning_rate": 3.9954715874149596e-05, + "loss": 0.3246, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14213445782661438, + "step": 1090, + "valid_targets_mean": 3826.4, + "valid_targets_min": 2414 + }, + { + "epoch": 0.8400460299194477, + "grad_norm": 0.6259375602694764, + "learning_rate": 3.9952107406655835e-05, + "loss": 0.3103, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17785485088825226, + "step": 1095, + "valid_targets_mean": 4195.0, + "valid_targets_min": 1937 + }, + { + "epoch": 0.8438818565400844, + "grad_norm": 0.6996300488192579, + "learning_rate": 3.9949425991183775e-05, + "loss": 0.3397, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18839314579963684, + "step": 1100, + "valid_targets_mean": 3154.9, + "valid_targets_min": 1649 + }, + { + "epoch": 0.8477176831607212, + "grad_norm": 0.6824021949212935, + "learning_rate": 3.9946671637537084e-05, + "loss": 0.322, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14219292998313904, + "step": 1105, + "valid_targets_mean": 3088.5, + "valid_targets_min": 1463 + }, + { + "epoch": 0.8515535097813579, + "grad_norm": 0.6269087597183562, + "learning_rate": 3.99438443557861e-05, + "loss": 0.3381, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16178417205810547, + "step": 1110, + "valid_targets_mean": 3241.4, + "valid_targets_min": 1426 + }, + { + "epoch": 0.8553893364019947, + "grad_norm": 0.7074219731072543, + "learning_rate": 3.9940944156267803e-05, + "loss": 0.3375, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14191371202468872, + "step": 1115, + "valid_targets_mean": 2768.9, + "valid_targets_min": 538 + }, + { + "epoch": 0.8592251630226314, + "grad_norm": 0.5632045714943364, + "learning_rate": 3.9937971049585766e-05, + "loss": 0.3316, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17113135755062103, + "step": 1120, + "valid_targets_mean": 4483.5, + "valid_targets_min": 1646 + }, + { + "epoch": 0.8630609896432682, + "grad_norm": 0.6557115877212156, + "learning_rate": 3.9934925046610134e-05, + "loss": 0.3625, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19142764806747437, + "step": 1125, + "valid_targets_mean": 3761.9, + "valid_targets_min": 2169 + }, + { + "epoch": 0.8668968162639049, + "grad_norm": 0.6762211354301243, + "learning_rate": 3.993180615847755e-05, + "loss": 0.3292, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1453298181295395, + "step": 1130, + "valid_targets_mean": 3168.4, + "valid_targets_min": 1625 + }, + { + "epoch": 0.8707326428845417, + "grad_norm": 0.6104030319444549, + "learning_rate": 3.992861439659116e-05, + "loss": 0.3383, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13507860898971558, + "step": 1135, + "valid_targets_mean": 3449.0, + "valid_targets_min": 1997 + }, + { + "epoch": 0.8745684695051784, + "grad_norm": 0.6887551314481509, + "learning_rate": 3.9925349772620536e-05, + "loss": 0.3422, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20485205948352814, + "step": 1140, + "valid_targets_mean": 4406.4, + "valid_targets_min": 2949 + }, + { + "epoch": 0.8784042961258152, + "grad_norm": 0.5327487589446982, + "learning_rate": 3.992201229850164e-05, + "loss": 0.3382, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18603554368019104, + "step": 1145, + "valid_targets_mean": 5358.5, + "valid_targets_min": 2652 + }, + { + "epoch": 0.8822401227464519, + "grad_norm": 0.5638475478133825, + "learning_rate": 3.99186019864368e-05, + "loss": 0.339, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19718316197395325, + "step": 1150, + "valid_targets_mean": 4314.5, + "valid_targets_min": 1381 + }, + { + "epoch": 0.8860759493670886, + "grad_norm": 0.6016582995249167, + "learning_rate": 3.9915118848894633e-05, + "loss": 0.3371, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13380110263824463, + "step": 1155, + "valid_targets_mean": 3395.9, + "valid_targets_min": 637 + }, + { + "epoch": 0.8899117759877253, + "grad_norm": 0.6902965232139197, + "learning_rate": 3.991156289861004e-05, + "loss": 0.3371, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14711695909500122, + "step": 1160, + "valid_targets_mean": 3256.4, + "valid_targets_min": 719 + }, + { + "epoch": 0.8937476026083621, + "grad_norm": 0.5720292432742352, + "learning_rate": 3.9907934148584105e-05, + "loss": 0.3342, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.145595520734787, + "step": 1165, + "valid_targets_mean": 4725.1, + "valid_targets_min": 3210 + }, + { + "epoch": 0.8975834292289988, + "grad_norm": 0.6685578550036175, + "learning_rate": 3.9904232612084104e-05, + "loss": 0.3266, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17669783532619476, + "step": 1170, + "valid_targets_mean": 3775.4, + "valid_targets_min": 1372 + }, + { + "epoch": 0.9014192558496356, + "grad_norm": 0.6795381419547154, + "learning_rate": 3.990045830264343e-05, + "loss": 0.3125, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1864616572856903, + "step": 1175, + "valid_targets_mean": 4266.1, + "valid_targets_min": 2768 + }, + { + "epoch": 0.9052550824702723, + "grad_norm": 0.6077511849230323, + "learning_rate": 3.989661123406154e-05, + "loss": 0.3285, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12511369585990906, + "step": 1180, + "valid_targets_mean": 2540.5, + "valid_targets_min": 530 + }, + { + "epoch": 0.9090909090909091, + "grad_norm": 0.619059959259929, + "learning_rate": 3.9892691420403906e-05, + "loss": 0.3212, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16179001331329346, + "step": 1185, + "valid_targets_mean": 3903.2, + "valid_targets_min": 1841 + }, + { + "epoch": 0.9129267357115458, + "grad_norm": 0.6066348002536585, + "learning_rate": 3.988869887600197e-05, + "loss": 0.3373, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14762656390666962, + "step": 1190, + "valid_targets_mean": 3476.8, + "valid_targets_min": 1728 + }, + { + "epoch": 0.9167625623321826, + "grad_norm": 0.7236022096603401, + "learning_rate": 3.988463361545309e-05, + "loss": 0.331, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18812285363674164, + "step": 1195, + "valid_targets_mean": 3712.2, + "valid_targets_min": 1362 + }, + { + "epoch": 0.9205983889528193, + "grad_norm": 0.609712858005834, + "learning_rate": 3.988049565362048e-05, + "loss": 0.326, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1683117300271988, + "step": 1200, + "valid_targets_mean": 3693.8, + "valid_targets_min": 852 + }, + { + "epoch": 0.9244342155734561, + "grad_norm": 0.7073932019932628, + "learning_rate": 3.987628500563318e-05, + "loss": 0.332, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.160975843667984, + "step": 1205, + "valid_targets_mean": 2822.6, + "valid_targets_min": 1195 + }, + { + "epoch": 0.9282700421940928, + "grad_norm": 0.6137874710905233, + "learning_rate": 3.987200168688595e-05, + "loss": 0.3262, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19224178791046143, + "step": 1210, + "valid_targets_mean": 4747.8, + "valid_targets_min": 1397 + }, + { + "epoch": 0.9321058688147296, + "grad_norm": 0.6126688872195202, + "learning_rate": 3.9867645713039275e-05, + "loss": 0.3163, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1229599341750145, + "step": 1215, + "valid_targets_mean": 3487.6, + "valid_targets_min": 1975 + }, + { + "epoch": 0.9359416954353663, + "grad_norm": 0.5908602924449757, + "learning_rate": 3.9863217100019264e-05, + "loss": 0.3091, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17327263951301575, + "step": 1220, + "valid_targets_mean": 4387.6, + "valid_targets_min": 2564 + }, + { + "epoch": 0.9397775220560031, + "grad_norm": 0.5469723199830279, + "learning_rate": 3.9858715864017604e-05, + "loss": 0.3142, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15003232657909393, + "step": 1225, + "valid_targets_mean": 6404.1, + "valid_targets_min": 2595 + }, + { + "epoch": 0.9436133486766398, + "grad_norm": 0.553212169828876, + "learning_rate": 3.985414202149151e-05, + "loss": 0.3234, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16989347338676453, + "step": 1230, + "valid_targets_mean": 4013.2, + "valid_targets_min": 2868 + }, + { + "epoch": 0.9474491752972766, + "grad_norm": 0.6812525656138777, + "learning_rate": 3.984949558916366e-05, + "loss": 0.3113, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1385374218225479, + "step": 1235, + "valid_targets_mean": 2886.0, + "valid_targets_min": 993 + }, + { + "epoch": 0.9512850019179133, + "grad_norm": 0.6163118593532194, + "learning_rate": 3.984477658402212e-05, + "loss": 0.3141, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1523301601409912, + "step": 1240, + "valid_targets_mean": 3805.1, + "valid_targets_min": 1575 + }, + { + "epoch": 0.9551208285385501, + "grad_norm": 0.6306704195441023, + "learning_rate": 3.9839985023320305e-05, + "loss": 0.3191, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18678873777389526, + "step": 1245, + "valid_targets_mean": 3722.0, + "valid_targets_min": 1360 + }, + { + "epoch": 0.9589566551591868, + "grad_norm": 0.5776498993324056, + "learning_rate": 3.9835120924576885e-05, + "loss": 0.3127, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14847475290298462, + "step": 1250, + "valid_targets_mean": 3669.4, + "valid_targets_min": 2414 + }, + { + "epoch": 0.9627924817798236, + "grad_norm": 0.5920013216656755, + "learning_rate": 3.9830184305575775e-05, + "loss": 0.3393, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1859714388847351, + "step": 1255, + "valid_targets_mean": 4327.4, + "valid_targets_min": 1454 + }, + { + "epoch": 0.9666283084004603, + "grad_norm": 0.6115969037988346, + "learning_rate": 3.9825175184366e-05, + "loss": 0.3286, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17300094664096832, + "step": 1260, + "valid_targets_mean": 3619.9, + "valid_targets_min": 1720 + }, + { + "epoch": 0.9704641350210971, + "grad_norm": 0.7104831874079424, + "learning_rate": 3.9820093579261685e-05, + "loss": 0.3279, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16959045827388763, + "step": 1265, + "valid_targets_mean": 3105.5, + "valid_targets_min": 1278 + }, + { + "epoch": 0.9742999616417338, + "grad_norm": 0.7363063487321052, + "learning_rate": 3.981493950884196e-05, + "loss": 0.3106, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17351767420768738, + "step": 1270, + "valid_targets_mean": 3147.8, + "valid_targets_min": 1858 + }, + { + "epoch": 0.9781357882623706, + "grad_norm": 0.6680280833885431, + "learning_rate": 3.9809712991950905e-05, + "loss": 0.3224, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12678033113479614, + "step": 1275, + "valid_targets_mean": 2984.5, + "valid_targets_min": 628 + }, + { + "epoch": 0.9819716148830073, + "grad_norm": 0.6530937935962138, + "learning_rate": 3.980441404769746e-05, + "loss": 0.3191, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1918300986289978, + "step": 1280, + "valid_targets_mean": 2602.6, + "valid_targets_min": 1039 + }, + { + "epoch": 0.9858074415036441, + "grad_norm": 0.5833944559435277, + "learning_rate": 3.979904269545538e-05, + "loss": 0.3287, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15227816998958588, + "step": 1285, + "valid_targets_mean": 4317.8, + "valid_targets_min": 2530 + }, + { + "epoch": 0.9896432681242808, + "grad_norm": 0.5927576767148522, + "learning_rate": 3.979359895486318e-05, + "loss": 0.3228, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18668025732040405, + "step": 1290, + "valid_targets_mean": 3711.0, + "valid_targets_min": 809 + }, + { + "epoch": 0.9934790947449176, + "grad_norm": 0.5928493603666899, + "learning_rate": 3.978808284582399e-05, + "loss": 0.3208, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1409824788570404, + "step": 1295, + "valid_targets_mean": 3450.5, + "valid_targets_min": 1780 + }, + { + "epoch": 0.9973149213655543, + "grad_norm": 0.6099778119305631, + "learning_rate": 3.9782494388505564e-05, + "loss": 0.3289, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1790844351053238, + "step": 1300, + "valid_targets_mean": 4356.6, + "valid_targets_min": 2185 + }, + { + "epoch": 1.0007671653241272, + "grad_norm": 0.48769210117329836, + "learning_rate": 3.977683360334015e-05, + "loss": 0.2988, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13263380527496338, + "step": 1305, + "valid_targets_mean": 8215.2, + "valid_targets_min": 6163 + }, + { + "epoch": 1.004602991944764, + "grad_norm": 0.550489506960781, + "learning_rate": 3.9771100511024473e-05, + "loss": 0.273, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13865600526332855, + "step": 1310, + "valid_targets_mean": 8594.4, + "valid_targets_min": 6006 + }, + { + "epoch": 1.0084388185654007, + "grad_norm": 0.5979518775129362, + "learning_rate": 3.976529513251959e-05, + "loss": 0.2708, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12106271833181381, + "step": 1315, + "valid_targets_mean": 3578.4, + "valid_targets_min": 289 + }, + { + "epoch": 1.0122746451860376, + "grad_norm": 0.44957595091499913, + "learning_rate": 3.9759417489050854e-05, + "loss": 0.2607, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12876835465431213, + "step": 1320, + "valid_targets_mean": 7019.9, + "valid_targets_min": 4985 + }, + { + "epoch": 1.0161104718066742, + "grad_norm": 0.5132179230666586, + "learning_rate": 3.975346760210784e-05, + "loss": 0.2398, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12546518445014954, + "step": 1325, + "valid_targets_mean": 6956.6, + "valid_targets_min": 1457 + }, + { + "epoch": 1.019946298427311, + "grad_norm": 0.40942445031677144, + "learning_rate": 3.974744549344426e-05, + "loss": 0.247, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11723187565803528, + "step": 1330, + "valid_targets_mean": 7361.6, + "valid_targets_min": 4961 + }, + { + "epoch": 1.0237821250479477, + "grad_norm": 0.5298231268647452, + "learning_rate": 3.9741351185077856e-05, + "loss": 0.2641, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15574219822883606, + "step": 1335, + "valid_targets_mean": 7058.5, + "valid_targets_min": 5985 + }, + { + "epoch": 1.0276179516685846, + "grad_norm": 0.43398091478249523, + "learning_rate": 3.9735184699290365e-05, + "loss": 0.2598, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14043277502059937, + "step": 1340, + "valid_targets_mean": 7664.2, + "valid_targets_min": 5169 + }, + { + "epoch": 1.0314537782892212, + "grad_norm": 0.4512818880897203, + "learning_rate": 3.972894605862742e-05, + "loss": 0.2477, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12542393803596497, + "step": 1345, + "valid_targets_mean": 7111.8, + "valid_targets_min": 5348 + }, + { + "epoch": 1.035289604909858, + "grad_norm": 0.4970063476548721, + "learning_rate": 3.972263528589843e-05, + "loss": 0.2461, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11628326028585434, + "step": 1350, + "valid_targets_mean": 4238.9, + "valid_targets_min": 152 + }, + { + "epoch": 1.0391254315304947, + "grad_norm": 0.44622072437313, + "learning_rate": 3.971625240417656e-05, + "loss": 0.2407, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12750333547592163, + "step": 1355, + "valid_targets_mean": 7199.4, + "valid_targets_min": 4258 + }, + { + "epoch": 1.0429612581511316, + "grad_norm": 0.4277554493223129, + "learning_rate": 3.9709797436798623e-05, + "loss": 0.2426, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11847192049026489, + "step": 1360, + "valid_targets_mean": 7243.4, + "valid_targets_min": 5831 + }, + { + "epoch": 1.0467970847717682, + "grad_norm": 0.4616577691079256, + "learning_rate": 3.970327040736495e-05, + "loss": 0.2387, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11849498748779297, + "step": 1365, + "valid_targets_mean": 6950.6, + "valid_targets_min": 3862 + }, + { + "epoch": 1.0506329113924051, + "grad_norm": 0.4800013355715912, + "learning_rate": 3.9696671339739395e-05, + "loss": 0.2365, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1271313726902008, + "step": 1370, + "valid_targets_mean": 6991.9, + "valid_targets_min": 5696 + }, + { + "epoch": 1.0544687380130418, + "grad_norm": 0.42541074890244024, + "learning_rate": 3.9690000258049144e-05, + "loss": 0.2295, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12747277319431305, + "step": 1375, + "valid_targets_mean": 6963.5, + "valid_targets_min": 4947 + }, + { + "epoch": 1.0583045646336786, + "grad_norm": 0.4432902211164894, + "learning_rate": 3.96832571866847e-05, + "loss": 0.2339, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11154384166002274, + "step": 1380, + "valid_targets_mean": 6436.9, + "valid_targets_min": 4988 + }, + { + "epoch": 1.0621403912543153, + "grad_norm": 0.4111541036245483, + "learning_rate": 3.967644215029979e-05, + "loss": 0.2372, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11755017936229706, + "step": 1385, + "valid_targets_mean": 7518.0, + "valid_targets_min": 5476 + }, + { + "epoch": 1.0659762178749521, + "grad_norm": 0.4109604711754066, + "learning_rate": 3.966955517381121e-05, + "loss": 0.2286, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10815656185150146, + "step": 1390, + "valid_targets_mean": 7422.6, + "valid_targets_min": 5226 + }, + { + "epoch": 1.0698120444955888, + "grad_norm": 0.45709620280736846, + "learning_rate": 3.9662596282398824e-05, + "loss": 0.2485, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1526232808828354, + "step": 1395, + "valid_targets_mean": 7080.0, + "valid_targets_min": 3753 + }, + { + "epoch": 1.0736478711162256, + "grad_norm": 0.552618347243215, + "learning_rate": 3.965556550150541e-05, + "loss": 0.2602, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13363657891750336, + "step": 1400, + "valid_targets_mean": 6064.6, + "valid_targets_min": 5161 + }, + { + "epoch": 1.0774836977368623, + "grad_norm": 0.41919690018749334, + "learning_rate": 3.964846285683658e-05, + "loss": 0.2187, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11345550417900085, + "step": 1405, + "valid_targets_mean": 7388.1, + "valid_targets_min": 5795 + }, + { + "epoch": 1.0813195243574991, + "grad_norm": 0.4140158690084107, + "learning_rate": 3.964128837436071e-05, + "loss": 0.2228, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10604599118232727, + "step": 1410, + "valid_targets_mean": 7696.0, + "valid_targets_min": 4658 + }, + { + "epoch": 1.0851553509781358, + "grad_norm": 0.41507085099932906, + "learning_rate": 3.963404208030879e-05, + "loss": 0.2294, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12562406063079834, + "step": 1415, + "valid_targets_mean": 7799.1, + "valid_targets_min": 5508 + }, + { + "epoch": 1.0889911775987726, + "grad_norm": 0.4318637034178312, + "learning_rate": 3.962672400117441e-05, + "loss": 0.2407, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11667871475219727, + "step": 1420, + "valid_targets_mean": 7341.6, + "valid_targets_min": 6454 + }, + { + "epoch": 1.0928270042194093, + "grad_norm": 0.4472161427596332, + "learning_rate": 3.961933416371358e-05, + "loss": 0.2316, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1036418229341507, + "step": 1425, + "valid_targets_mean": 6818.0, + "valid_targets_min": 4392 + }, + { + "epoch": 1.0966628308400461, + "grad_norm": 0.4401008250263132, + "learning_rate": 3.961187259494469e-05, + "loss": 0.2308, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10189579427242279, + "step": 1430, + "valid_targets_mean": 7857.6, + "valid_targets_min": 5394 + }, + { + "epoch": 1.1004986574606828, + "grad_norm": 0.42313702436305756, + "learning_rate": 3.960433932214839e-05, + "loss": 0.2019, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09276679903268814, + "step": 1435, + "valid_targets_mean": 8314.1, + "valid_targets_min": 4837 + }, + { + "epoch": 1.1043344840813196, + "grad_norm": 0.3511374600630408, + "learning_rate": 3.9596734372867474e-05, + "loss": 0.2047, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0873468816280365, + "step": 1440, + "valid_targets_mean": 7368.2, + "valid_targets_min": 6027 + }, + { + "epoch": 1.1081703107019563, + "grad_norm": 0.4954203062888038, + "learning_rate": 3.958905777490682e-05, + "loss": 0.214, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1156146377325058, + "step": 1445, + "valid_targets_mean": 6435.0, + "valid_targets_min": 4238 + }, + { + "epoch": 1.112006137322593, + "grad_norm": 1.959649768779798, + "learning_rate": 3.9581309556333243e-05, + "loss": 0.2369, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10030440986156464, + "step": 1450, + "valid_targets_mean": 247.8, + "valid_targets_min": 137 + }, + { + "epoch": 1.1158419639432298, + "grad_norm": 0.46229218579617753, + "learning_rate": 3.957348974547543e-05, + "loss": 0.2408, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11166730523109436, + "step": 1455, + "valid_targets_mean": 6807.0, + "valid_targets_min": 5165 + }, + { + "epoch": 1.1196777905638666, + "grad_norm": 0.6003531291310348, + "learning_rate": 3.956559837092381e-05, + "loss": 0.2451, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1283479630947113, + "step": 1460, + "valid_targets_mean": 6989.9, + "valid_targets_min": 5104 + }, + { + "epoch": 1.1235136171845033, + "grad_norm": 0.4232140506822345, + "learning_rate": 3.955763546153047e-05, + "loss": 0.2433, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13194653391838074, + "step": 1465, + "valid_targets_mean": 6692.2, + "valid_targets_min": 5068 + }, + { + "epoch": 1.12734944380514, + "grad_norm": 0.4451534045599726, + "learning_rate": 3.9549601046409024e-05, + "loss": 0.2365, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14130067825317383, + "step": 1470, + "valid_targets_mean": 7419.6, + "valid_targets_min": 5578 + }, + { + "epoch": 1.1311852704257768, + "grad_norm": 0.42929425287467604, + "learning_rate": 3.9541495154934544e-05, + "loss": 0.2257, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11511807143688202, + "step": 1475, + "valid_targets_mean": 7536.5, + "valid_targets_min": 5075 + }, + { + "epoch": 1.1350210970464134, + "grad_norm": 0.39668088712526145, + "learning_rate": 3.953331781674341e-05, + "loss": 0.2349, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10644698143005371, + "step": 1480, + "valid_targets_mean": 7221.5, + "valid_targets_min": 5841 + }, + { + "epoch": 1.1388569236670503, + "grad_norm": 0.4171910716149015, + "learning_rate": 3.952506906173323e-05, + "loss": 0.2282, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11281067132949829, + "step": 1485, + "valid_targets_mean": 7287.5, + "valid_targets_min": 5151 + }, + { + "epoch": 1.142692750287687, + "grad_norm": 0.41817085718756786, + "learning_rate": 3.951674892006272e-05, + "loss": 0.2392, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11073216795921326, + "step": 1490, + "valid_targets_mean": 6738.4, + "valid_targets_min": 4042 + }, + { + "epoch": 1.1465285769083238, + "grad_norm": 0.45839119739229733, + "learning_rate": 3.9508357422151604e-05, + "loss": 0.2291, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12807488441467285, + "step": 1495, + "valid_targets_mean": 6331.0, + "valid_targets_min": 5496 + }, + { + "epoch": 1.1503644035289604, + "grad_norm": 0.4632037404051386, + "learning_rate": 3.949989459868049e-05, + "loss": 0.2331, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11391602456569672, + "step": 1500, + "valid_targets_mean": 6271.1, + "valid_targets_min": 4695 + }, + { + "epoch": 1.1542002301495973, + "grad_norm": 0.45215555777668026, + "learning_rate": 3.949136048059077e-05, + "loss": 0.2342, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12079563736915588, + "step": 1505, + "valid_targets_mean": 5790.8, + "valid_targets_min": 5021 + }, + { + "epoch": 1.158036056770234, + "grad_norm": 0.556574604651013, + "learning_rate": 3.9482755099084496e-05, + "loss": 0.2258, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09364157915115356, + "step": 1510, + "valid_targets_mean": 3386.8, + "valid_targets_min": 2279 + }, + { + "epoch": 1.1618718833908708, + "grad_norm": 0.45627810686768605, + "learning_rate": 3.947407848562425e-05, + "loss": 0.227, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11292421817779541, + "step": 1515, + "valid_targets_mean": 6240.5, + "valid_targets_min": 5093 + }, + { + "epoch": 1.1657077100115074, + "grad_norm": 0.4642428930434613, + "learning_rate": 3.946533067193309e-05, + "loss": 0.2268, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11852425336837769, + "step": 1520, + "valid_targets_mean": 6329.5, + "valid_targets_min": 4589 + }, + { + "epoch": 1.1695435366321443, + "grad_norm": 0.43196553453524744, + "learning_rate": 3.945651168999435e-05, + "loss": 0.2258, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12109851837158203, + "step": 1525, + "valid_targets_mean": 7589.2, + "valid_targets_min": 4963 + }, + { + "epoch": 1.173379363252781, + "grad_norm": 0.42750412631740975, + "learning_rate": 3.94476215720516e-05, + "loss": 0.2261, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09737605601549149, + "step": 1530, + "valid_targets_mean": 5577.9, + "valid_targets_min": 4458 + }, + { + "epoch": 1.1772151898734178, + "grad_norm": 0.5180593938958705, + "learning_rate": 3.9438660350608475e-05, + "loss": 0.2249, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10786409676074982, + "step": 1535, + "valid_targets_mean": 5432.5, + "valid_targets_min": 4433 + }, + { + "epoch": 1.1810510164940544, + "grad_norm": 0.4479551728991794, + "learning_rate": 3.9429628058428586e-05, + "loss": 0.229, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11652961373329163, + "step": 1540, + "valid_targets_mean": 6361.8, + "valid_targets_min": 5201 + }, + { + "epoch": 1.1848868431146913, + "grad_norm": 0.45092129187772206, + "learning_rate": 3.942052472853538e-05, + "loss": 0.2301, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10579784214496613, + "step": 1545, + "valid_targets_mean": 5824.4, + "valid_targets_min": 5129 + }, + { + "epoch": 1.188722669735328, + "grad_norm": 0.9633846543438807, + "learning_rate": 3.941135039421204e-05, + "loss": 0.204, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1606774926185608, + "step": 1550, + "valid_targets_mean": 1637.6, + "valid_targets_min": 137 + }, + { + "epoch": 1.1925584963559648, + "grad_norm": 0.4515900091304907, + "learning_rate": 3.940210508900134e-05, + "loss": 0.225, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11362193524837494, + "step": 1555, + "valid_targets_mean": 6474.6, + "valid_targets_min": 4751 + }, + { + "epoch": 1.1963943229766014, + "grad_norm": 0.4276257323855524, + "learning_rate": 3.939278884670553e-05, + "loss": 0.222, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12210103124380112, + "step": 1560, + "valid_targets_mean": 7186.2, + "valid_targets_min": 5358 + }, + { + "epoch": 1.2002301495972383, + "grad_norm": 0.4449515231704409, + "learning_rate": 3.938340170138625e-05, + "loss": 0.228, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11385686695575714, + "step": 1565, + "valid_targets_mean": 7126.8, + "valid_targets_min": 4078 + }, + { + "epoch": 1.204065976217875, + "grad_norm": 0.4714017644795156, + "learning_rate": 3.937394368736433e-05, + "loss": 0.2196, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11061680316925049, + "step": 1570, + "valid_targets_mean": 7022.1, + "valid_targets_min": 4940 + }, + { + "epoch": 1.2079018028385118, + "grad_norm": 0.4489449717619331, + "learning_rate": 3.936441483921974e-05, + "loss": 0.2142, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1146257072687149, + "step": 1575, + "valid_targets_mean": 5900.6, + "valid_targets_min": 4779 + }, + { + "epoch": 1.2117376294591484, + "grad_norm": 0.44635720643305316, + "learning_rate": 3.935481519179141e-05, + "loss": 0.2247, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1149950698018074, + "step": 1580, + "valid_targets_mean": 6290.1, + "valid_targets_min": 4620 + }, + { + "epoch": 1.2155734560797853, + "grad_norm": 0.990989136251679, + "learning_rate": 3.934514478017713e-05, + "loss": 0.2482, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10060764104127884, + "step": 1585, + "valid_targets_mean": 1162.2, + "valid_targets_min": 788 + }, + { + "epoch": 1.219409282700422, + "grad_norm": 0.8655052465654122, + "learning_rate": 3.933540363973341e-05, + "loss": 0.2448, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11634095013141632, + "step": 1590, + "valid_targets_mean": 1500.6, + "valid_targets_min": 957 + }, + { + "epoch": 1.2232451093210588, + "grad_norm": 1.0246807412270007, + "learning_rate": 3.9325591806075364e-05, + "loss": 0.231, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10916225612163544, + "step": 1595, + "valid_targets_mean": 1371.2, + "valid_targets_min": 796 + }, + { + "epoch": 1.2270809359416954, + "grad_norm": 2.099162244909047, + "learning_rate": 3.9315709315076576e-05, + "loss": 0.2322, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1309526562690735, + "step": 1600, + "valid_targets_mean": 1433.6, + "valid_targets_min": 697 + }, + { + "epoch": 1.230916762562332, + "grad_norm": 0.8453701171361812, + "learning_rate": 3.930575620286894e-05, + "loss": 0.2292, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10963514447212219, + "step": 1605, + "valid_targets_mean": 1578.9, + "valid_targets_min": 638 + }, + { + "epoch": 1.234752589182969, + "grad_norm": 0.8219501297819789, + "learning_rate": 3.929573250584258e-05, + "loss": 0.2275, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11498934030532837, + "step": 1610, + "valid_targets_mean": 1746.4, + "valid_targets_min": 727 + }, + { + "epoch": 1.2385884158036058, + "grad_norm": 0.9090973334321243, + "learning_rate": 3.928563826064568e-05, + "loss": 0.2357, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0986706018447876, + "step": 1615, + "valid_targets_mean": 1309.9, + "valid_targets_min": 761 + }, + { + "epoch": 1.2424242424242424, + "grad_norm": 1.208443023228996, + "learning_rate": 3.927547350418433e-05, + "loss": 0.218, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13086728751659393, + "step": 1620, + "valid_targets_mean": 1368.6, + "valid_targets_min": 714 + }, + { + "epoch": 1.246260069044879, + "grad_norm": 0.8677514947109423, + "learning_rate": 3.926523827362247e-05, + "loss": 0.2374, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11222252249717712, + "step": 1625, + "valid_targets_mean": 1491.8, + "valid_targets_min": 867 + }, + { + "epoch": 1.250095895665516, + "grad_norm": 0.8993884767554339, + "learning_rate": 3.9254932606381675e-05, + "loss": 0.2258, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12630558013916016, + "step": 1630, + "valid_targets_mean": 1994.0, + "valid_targets_min": 790 + }, + { + "epoch": 1.2539317222861528, + "grad_norm": 0.8885589598260926, + "learning_rate": 3.9244556540141045e-05, + "loss": 0.219, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11382272094488144, + "step": 1635, + "valid_targets_mean": 1523.4, + "valid_targets_min": 896 + }, + { + "epoch": 1.2577675489067894, + "grad_norm": 0.8373640951116866, + "learning_rate": 3.9234110112837086e-05, + "loss": 0.2315, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11628646403551102, + "step": 1640, + "valid_targets_mean": 1790.1, + "valid_targets_min": 1009 + }, + { + "epoch": 1.261603375527426, + "grad_norm": 0.7811231155206202, + "learning_rate": 3.922359336266354e-05, + "loss": 0.2212, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09118906408548355, + "step": 1645, + "valid_targets_mean": 1394.0, + "valid_targets_min": 717 + }, + { + "epoch": 1.265439202148063, + "grad_norm": 1.2799661575562056, + "learning_rate": 3.9213006328071265e-05, + "loss": 0.211, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0861155316233635, + "step": 1650, + "valid_targets_mean": 1129.0, + "valid_targets_min": 659 + }, + { + "epoch": 1.2692750287686996, + "grad_norm": 0.8821642793252942, + "learning_rate": 3.9202349047768094e-05, + "loss": 0.2103, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10283654183149338, + "step": 1655, + "valid_targets_mean": 1466.9, + "valid_targets_min": 642 + }, + { + "epoch": 1.2731108553893364, + "grad_norm": 0.8433591153992017, + "learning_rate": 3.919162156071868e-05, + "loss": 0.2194, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07667336612939835, + "step": 1660, + "valid_targets_mean": 1129.4, + "valid_targets_min": 765 + }, + { + "epoch": 1.276946682009973, + "grad_norm": 1.1191910806221428, + "learning_rate": 3.918082390614437e-05, + "loss": 0.2297, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10648314654827118, + "step": 1665, + "valid_targets_mean": 1322.8, + "valid_targets_min": 631 + }, + { + "epoch": 1.28078250863061, + "grad_norm": 1.1213717077264254, + "learning_rate": 3.916995612352305e-05, + "loss": 0.2302, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1296577751636505, + "step": 1670, + "valid_targets_mean": 1612.0, + "valid_targets_min": 623 + }, + { + "epoch": 1.2846183352512466, + "grad_norm": 0.7600603056344629, + "learning_rate": 3.915901825258902e-05, + "loss": 0.2222, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08274514973163605, + "step": 1675, + "valid_targets_mean": 1402.1, + "valid_targets_min": 940 + }, + { + "epoch": 1.2884541618718834, + "grad_norm": 0.9447989451889589, + "learning_rate": 3.9148010333332814e-05, + "loss": 0.2224, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09882551431655884, + "step": 1680, + "valid_targets_mean": 1337.8, + "valid_targets_min": 776 + }, + { + "epoch": 1.29228998849252, + "grad_norm": 0.8639211080140133, + "learning_rate": 3.913693240600107e-05, + "loss": 0.2199, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12831756472587585, + "step": 1685, + "valid_targets_mean": 1525.5, + "valid_targets_min": 623 + }, + { + "epoch": 1.296125815113157, + "grad_norm": 0.8709025299192902, + "learning_rate": 3.912578451109641e-05, + "loss": 0.2251, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12210875004529953, + "step": 1690, + "valid_targets_mean": 1441.0, + "valid_targets_min": 693 + }, + { + "epoch": 1.2999616417337936, + "grad_norm": 0.9133409599346469, + "learning_rate": 3.9114566689377246e-05, + "loss": 0.2203, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.062201060354709625, + "step": 1695, + "valid_targets_mean": 847.9, + "valid_targets_min": 611 + }, + { + "epoch": 1.3037974683544304, + "grad_norm": 0.8758256116344582, + "learning_rate": 3.910327898185768e-05, + "loss": 0.2137, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11523973941802979, + "step": 1700, + "valid_targets_mean": 1524.9, + "valid_targets_min": 793 + }, + { + "epoch": 1.307633294975067, + "grad_norm": 0.8216109115468431, + "learning_rate": 3.909192142980729e-05, + "loss": 0.2127, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11683434247970581, + "step": 1705, + "valid_targets_mean": 1779.4, + "valid_targets_min": 563 + }, + { + "epoch": 1.311469121595704, + "grad_norm": 0.7812668913348632, + "learning_rate": 3.908049407475104e-05, + "loss": 0.2131, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12163171917200089, + "step": 1710, + "valid_targets_mean": 1741.2, + "valid_targets_min": 1341 + }, + { + "epoch": 1.3153049482163406, + "grad_norm": 1.51012771288664, + "learning_rate": 3.90689969584691e-05, + "loss": 0.2138, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13025221228599548, + "step": 1715, + "valid_targets_mean": 1875.6, + "valid_targets_min": 1107 + }, + { + "epoch": 1.3191407748369774, + "grad_norm": 0.9130807133900251, + "learning_rate": 3.905743012299671e-05, + "loss": 0.224, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10225005447864532, + "step": 1720, + "valid_targets_mean": 1419.5, + "valid_targets_min": 680 + }, + { + "epoch": 1.322976601457614, + "grad_norm": 0.77861083173254, + "learning_rate": 3.904579361062399e-05, + "loss": 0.2076, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.146154522895813, + "step": 1725, + "valid_targets_mean": 1968.2, + "valid_targets_min": 981 + }, + { + "epoch": 1.326812428078251, + "grad_norm": 0.8091599793846472, + "learning_rate": 3.9034087463895826e-05, + "loss": 0.2131, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10693266242742538, + "step": 1730, + "valid_targets_mean": 1568.9, + "valid_targets_min": 937 + }, + { + "epoch": 1.3306482546988876, + "grad_norm": 0.8916280996789813, + "learning_rate": 3.90223117256117e-05, + "loss": 0.2169, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11624979227781296, + "step": 1735, + "valid_targets_mean": 1645.2, + "valid_targets_min": 804 + }, + { + "epoch": 1.3344840813195242, + "grad_norm": 0.8175668941914893, + "learning_rate": 3.9010466438825505e-05, + "loss": 0.219, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10415971279144287, + "step": 1740, + "valid_targets_mean": 1574.9, + "valid_targets_min": 1019 + }, + { + "epoch": 1.338319907940161, + "grad_norm": 0.8970225449345844, + "learning_rate": 3.899855164684546e-05, + "loss": 0.2097, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09496060013771057, + "step": 1745, + "valid_targets_mean": 1121.6, + "valid_targets_min": 882 + }, + { + "epoch": 1.342155734560798, + "grad_norm": 0.8160041311482332, + "learning_rate": 3.898656739323386e-05, + "loss": 0.207, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11721847951412201, + "step": 1750, + "valid_targets_mean": 1818.1, + "valid_targets_min": 1168 + }, + { + "epoch": 1.3459915611814346, + "grad_norm": 0.91043146878187, + "learning_rate": 3.897451372180699e-05, + "loss": 0.2009, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09932956844568253, + "step": 1755, + "valid_targets_mean": 1417.8, + "valid_targets_min": 678 + }, + { + "epoch": 1.3498273878020712, + "grad_norm": 0.8868773414171788, + "learning_rate": 3.896239067663494e-05, + "loss": 0.2147, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08291667699813843, + "step": 1760, + "valid_targets_mean": 1051.0, + "valid_targets_min": 676 + }, + { + "epoch": 1.353663214422708, + "grad_norm": 0.7781496676157917, + "learning_rate": 3.895019830204141e-05, + "loss": 0.2099, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09344226866960526, + "step": 1765, + "valid_targets_mean": 1431.1, + "valid_targets_min": 794 + }, + { + "epoch": 1.357499041043345, + "grad_norm": 0.8833066745958096, + "learning_rate": 3.8937936642603616e-05, + "loss": 0.2118, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11165735125541687, + "step": 1770, + "valid_targets_mean": 1486.9, + "valid_targets_min": 742 + }, + { + "epoch": 1.3613348676639816, + "grad_norm": 0.805470116661206, + "learning_rate": 3.892560574315207e-05, + "loss": 0.2128, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12586958706378937, + "step": 1775, + "valid_targets_mean": 1734.5, + "valid_targets_min": 741 + }, + { + "epoch": 1.3651706942846182, + "grad_norm": 0.9665034054973424, + "learning_rate": 3.891320564877044e-05, + "loss": 0.2218, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12657983601093292, + "step": 1780, + "valid_targets_mean": 1712.5, + "valid_targets_min": 1157 + }, + { + "epoch": 1.369006520905255, + "grad_norm": 0.8490558498013122, + "learning_rate": 3.890073640479536e-05, + "loss": 0.2006, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14033156633377075, + "step": 1785, + "valid_targets_mean": 1723.8, + "valid_targets_min": 862 + }, + { + "epoch": 1.372842347525892, + "grad_norm": 0.8214134983679201, + "learning_rate": 3.888819805681635e-05, + "loss": 0.2233, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1332060694694519, + "step": 1790, + "valid_targets_mean": 1633.1, + "valid_targets_min": 1019 + }, + { + "epoch": 1.3766781741465286, + "grad_norm": 0.7423723339219671, + "learning_rate": 3.887559065067551e-05, + "loss": 0.2142, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10133840143680573, + "step": 1795, + "valid_targets_mean": 1376.1, + "valid_targets_min": 1058 + }, + { + "epoch": 1.3805140007671652, + "grad_norm": 0.9020722767900204, + "learning_rate": 3.886291423246746e-05, + "loss": 0.2287, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11525247246026993, + "step": 1800, + "valid_targets_mean": 1576.5, + "valid_targets_min": 618 + }, + { + "epoch": 1.384349827387802, + "grad_norm": 0.7389783306562234, + "learning_rate": 3.885016884853915e-05, + "loss": 0.2034, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12375020235776901, + "step": 1805, + "valid_targets_mean": 1615.8, + "valid_targets_min": 629 + }, + { + "epoch": 1.3881856540084387, + "grad_norm": 0.7648633211232435, + "learning_rate": 3.883735454548967e-05, + "loss": 0.2118, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10940219461917877, + "step": 1810, + "valid_targets_mean": 1881.2, + "valid_targets_min": 1167 + }, + { + "epoch": 1.3920214806290756, + "grad_norm": 0.8304070829944429, + "learning_rate": 3.882447137017007e-05, + "loss": 0.2134, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11188043653964996, + "step": 1815, + "valid_targets_mean": 1367.2, + "valid_targets_min": 699 + }, + { + "epoch": 1.3958573072497122, + "grad_norm": 0.804082501509465, + "learning_rate": 3.8811519369683225e-05, + "loss": 0.2004, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0827021598815918, + "step": 1820, + "valid_targets_mean": 1174.9, + "valid_targets_min": 679 + }, + { + "epoch": 1.399693133870349, + "grad_norm": 0.773606954349157, + "learning_rate": 3.8798498591383665e-05, + "loss": 0.2053, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09892909228801727, + "step": 1825, + "valid_targets_mean": 1556.2, + "valid_targets_min": 504 + }, + { + "epoch": 1.4035289604909857, + "grad_norm": 0.7504512007415264, + "learning_rate": 3.878540908287734e-05, + "loss": 0.2108, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11856793612241745, + "step": 1830, + "valid_targets_mean": 1885.0, + "valid_targets_min": 1331 + }, + { + "epoch": 1.4073647871116226, + "grad_norm": 0.7658737741682649, + "learning_rate": 3.877225089202151e-05, + "loss": 0.211, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07717673480510712, + "step": 1835, + "valid_targets_mean": 1306.1, + "valid_targets_min": 661 + }, + { + "epoch": 1.4112006137322592, + "grad_norm": 0.8129456959202752, + "learning_rate": 3.875902406692456e-05, + "loss": 0.2145, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0804230123758316, + "step": 1840, + "valid_targets_mean": 1330.1, + "valid_targets_min": 958 + }, + { + "epoch": 1.415036440352896, + "grad_norm": 0.7752896544533973, + "learning_rate": 3.87457286559458e-05, + "loss": 0.2023, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08119921386241913, + "step": 1845, + "valid_targets_mean": 1465.0, + "valid_targets_min": 963 + }, + { + "epoch": 1.4188722669735327, + "grad_norm": 0.8528728875822169, + "learning_rate": 3.873236470769529e-05, + "loss": 0.2092, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10215046256780624, + "step": 1850, + "valid_targets_mean": 1428.0, + "valid_targets_min": 793 + }, + { + "epoch": 1.4227080935941696, + "grad_norm": 0.8494624626073306, + "learning_rate": 3.8718932271033684e-05, + "loss": 0.2098, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13439273834228516, + "step": 1855, + "valid_targets_mean": 1633.1, + "valid_targets_min": 917 + }, + { + "epoch": 1.4265439202148062, + "grad_norm": 0.8075813686422219, + "learning_rate": 3.8705431395072045e-05, + "loss": 0.2032, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08263307809829712, + "step": 1860, + "valid_targets_mean": 1246.5, + "valid_targets_min": 872 + }, + { + "epoch": 1.4303797468354431, + "grad_norm": 0.6938279044215061, + "learning_rate": 3.869186212917166e-05, + "loss": 0.2107, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10818065702915192, + "step": 1865, + "valid_targets_mean": 1903.6, + "valid_targets_min": 1220 + }, + { + "epoch": 1.4342155734560798, + "grad_norm": 0.8402285351352858, + "learning_rate": 3.867822452294385e-05, + "loss": 0.1886, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09851537644863129, + "step": 1870, + "valid_targets_mean": 1391.6, + "valid_targets_min": 712 + }, + { + "epoch": 1.4380514000767166, + "grad_norm": 0.868472750325986, + "learning_rate": 3.86645186262498e-05, + "loss": 0.2028, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11110392212867737, + "step": 1875, + "valid_targets_mean": 1295.0, + "valid_targets_min": 605 + }, + { + "epoch": 1.4418872266973533, + "grad_norm": 0.7159859759201983, + "learning_rate": 3.865074448920039e-05, + "loss": 0.2094, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06926972419023514, + "step": 1880, + "valid_targets_mean": 1414.8, + "valid_targets_min": 746 + }, + { + "epoch": 1.4457230533179901, + "grad_norm": 0.8565351116195955, + "learning_rate": 3.8636902162155975e-05, + "loss": 0.1961, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08805161714553833, + "step": 1885, + "valid_targets_mean": 1059.4, + "valid_targets_min": 723 + }, + { + "epoch": 1.4495588799386268, + "grad_norm": 0.775956610015204, + "learning_rate": 3.862299169572624e-05, + "loss": 0.2099, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11050696671009064, + "step": 1890, + "valid_targets_mean": 1634.2, + "valid_targets_min": 1327 + }, + { + "epoch": 1.4533947065592634, + "grad_norm": 0.8115168244277858, + "learning_rate": 3.860901314077e-05, + "loss": 0.205, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09108547866344452, + "step": 1895, + "valid_targets_mean": 1412.5, + "valid_targets_min": 902 + }, + { + "epoch": 1.4572305331799003, + "grad_norm": 0.9796952515414742, + "learning_rate": 3.8594966548394994e-05, + "loss": 0.21, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09728436917066574, + "step": 1900, + "valid_targets_mean": 1558.8, + "valid_targets_min": 803 + }, + { + "epoch": 1.4610663598005371, + "grad_norm": 0.7947858135086905, + "learning_rate": 3.858085196995774e-05, + "loss": 0.2063, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10410543531179428, + "step": 1905, + "valid_targets_mean": 1441.2, + "valid_targets_min": 1079 + }, + { + "epoch": 1.4649021864211738, + "grad_norm": 0.7957769368930663, + "learning_rate": 3.8566669457063295e-05, + "loss": 0.2028, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0980774387717247, + "step": 1910, + "valid_targets_mean": 1449.8, + "valid_targets_min": 1263 + }, + { + "epoch": 1.4687380130418104, + "grad_norm": 0.7448914021886405, + "learning_rate": 3.855241906156513e-05, + "loss": 0.2144, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1052919328212738, + "step": 1915, + "valid_targets_mean": 1487.4, + "valid_targets_min": 844 + }, + { + "epoch": 1.4725738396624473, + "grad_norm": 0.9052530155348918, + "learning_rate": 3.853810083556487e-05, + "loss": 0.2095, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09099961817264557, + "step": 1920, + "valid_targets_mean": 1421.4, + "valid_targets_min": 667 + }, + { + "epoch": 1.4764096662830841, + "grad_norm": 0.7859099872048122, + "learning_rate": 3.8523714831412166e-05, + "loss": 0.2139, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12568005919456482, + "step": 1925, + "valid_targets_mean": 1822.6, + "valid_targets_min": 1041 + }, + { + "epoch": 1.4802454929037208, + "grad_norm": 0.9022825997056168, + "learning_rate": 3.850926110170446e-05, + "loss": 0.21, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09726682305335999, + "step": 1930, + "valid_targets_mean": 1327.4, + "valid_targets_min": 777 + }, + { + "epoch": 1.4840813195243574, + "grad_norm": 0.7836046857124543, + "learning_rate": 3.849473969928681e-05, + "loss": 0.2116, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10368139296770096, + "step": 1935, + "valid_targets_mean": 1823.8, + "valid_targets_min": 1283 + }, + { + "epoch": 1.4879171461449943, + "grad_norm": 0.8531567378531933, + "learning_rate": 3.8480150677251706e-05, + "loss": 0.1988, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06959441304206848, + "step": 1940, + "valid_targets_mean": 924.0, + "valid_targets_min": 620 + }, + { + "epoch": 1.4917529727656311, + "grad_norm": 0.7523130153573612, + "learning_rate": 3.846549408893887e-05, + "loss": 0.2072, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08927688747644424, + "step": 1945, + "valid_targets_mean": 1480.2, + "valid_targets_min": 656 + }, + { + "epoch": 1.4955887993862678, + "grad_norm": 0.7530871603065321, + "learning_rate": 3.8450769987935024e-05, + "loss": 0.2069, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08240532130002975, + "step": 1950, + "valid_targets_mean": 1415.5, + "valid_targets_min": 697 + }, + { + "epoch": 1.4994246260069044, + "grad_norm": 0.7354144775203162, + "learning_rate": 3.8435978428073764e-05, + "loss": 0.2011, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10189288854598999, + "step": 1955, + "valid_targets_mean": 1655.8, + "valid_targets_min": 816 + }, + { + "epoch": 1.5032604526275413, + "grad_norm": 0.7392478915524048, + "learning_rate": 3.842111946343531e-05, + "loss": 0.1997, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09638693928718567, + "step": 1960, + "valid_targets_mean": 1553.0, + "valid_targets_min": 850 + }, + { + "epoch": 1.5070962792481781, + "grad_norm": 0.828194553106562, + "learning_rate": 3.840619314834633e-05, + "loss": 0.1946, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09910672158002853, + "step": 1965, + "valid_targets_mean": 1353.9, + "valid_targets_min": 747 + }, + { + "epoch": 1.5109321058688148, + "grad_norm": 0.7878753522023645, + "learning_rate": 3.8391199537379714e-05, + "loss": 0.2129, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08896408975124359, + "step": 1970, + "valid_targets_mean": 1420.9, + "valid_targets_min": 933 + }, + { + "epoch": 1.5147679324894514, + "grad_norm": 0.8339593355444966, + "learning_rate": 3.837613868535444e-05, + "loss": 0.2021, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09718890488147736, + "step": 1975, + "valid_targets_mean": 1394.9, + "valid_targets_min": 827 + }, + { + "epoch": 1.5186037591100883, + "grad_norm": 0.7176336429625116, + "learning_rate": 3.836101064733528e-05, + "loss": 0.1936, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08817388117313385, + "step": 1980, + "valid_targets_mean": 1383.2, + "valid_targets_min": 909 + }, + { + "epoch": 1.5224395857307251, + "grad_norm": 0.7486209757738533, + "learning_rate": 3.834581547863268e-05, + "loss": 0.1946, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11141055822372437, + "step": 1985, + "valid_targets_mean": 1682.4, + "valid_targets_min": 1094 + }, + { + "epoch": 1.5262754123513618, + "grad_norm": 0.8702125552934588, + "learning_rate": 3.833055323480253e-05, + "loss": 0.1872, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09034430235624313, + "step": 1990, + "valid_targets_mean": 1118.4, + "valid_targets_min": 720 + }, + { + "epoch": 1.5301112389719984, + "grad_norm": 0.7586063445888539, + "learning_rate": 3.831522397164592e-05, + "loss": 0.2032, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10598678141832352, + "step": 1995, + "valid_targets_mean": 1538.2, + "valid_targets_min": 610 + }, + { + "epoch": 1.5339470655926353, + "grad_norm": 0.7138003134014441, + "learning_rate": 3.829982774520901e-05, + "loss": 0.2, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10905133932828903, + "step": 2000, + "valid_targets_mean": 1570.1, + "valid_targets_min": 713 + }, + { + "epoch": 1.537782892213272, + "grad_norm": 0.7198369762784996, + "learning_rate": 3.828436461178278e-05, + "loss": 0.2067, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09731920063495636, + "step": 2005, + "valid_targets_mean": 1537.8, + "valid_targets_min": 874 + }, + { + "epoch": 1.5416187188339086, + "grad_norm": 0.7345471636592713, + "learning_rate": 3.826883462790282e-05, + "loss": 0.1986, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09961222112178802, + "step": 2010, + "valid_targets_mean": 1471.2, + "valid_targets_min": 765 + }, + { + "epoch": 1.5454545454545454, + "grad_norm": 0.7441430737169329, + "learning_rate": 3.825323785034915e-05, + "loss": 0.2013, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08664283901453018, + "step": 2015, + "valid_targets_mean": 1254.4, + "valid_targets_min": 816 + }, + { + "epoch": 1.5492903720751823, + "grad_norm": 0.7873078181519199, + "learning_rate": 3.823757433614599e-05, + "loss": 0.2057, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09963281452655792, + "step": 2020, + "valid_targets_mean": 1482.9, + "valid_targets_min": 1182 + }, + { + "epoch": 1.553126198695819, + "grad_norm": 0.5673533798949199, + "learning_rate": 3.822184414256156e-05, + "loss": 0.2949, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1389979124069214, + "step": 2025, + "valid_targets_mean": 4566.8, + "valid_targets_min": 702 + }, + { + "epoch": 1.5569620253164556, + "grad_norm": 0.5647675240386916, + "learning_rate": 3.820604732710786e-05, + "loss": 0.3163, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18257680535316467, + "step": 2030, + "valid_targets_mean": 5524.9, + "valid_targets_min": 1005 + }, + { + "epoch": 1.5607978519370924, + "grad_norm": 0.5262952885005203, + "learning_rate": 3.819018394754051e-05, + "loss": 0.3042, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12346282601356506, + "step": 2035, + "valid_targets_mean": 4319.8, + "valid_targets_min": 2054 + }, + { + "epoch": 1.5646336785577293, + "grad_norm": 0.5350379928442917, + "learning_rate": 3.8174254061858436e-05, + "loss": 0.3033, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12029622495174408, + "step": 2040, + "valid_targets_mean": 3498.9, + "valid_targets_min": 1561 + }, + { + "epoch": 1.568469505178366, + "grad_norm": 0.5208159448124051, + "learning_rate": 3.815825772830377e-05, + "loss": 0.2997, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18742284178733826, + "step": 2045, + "valid_targets_mean": 5020.6, + "valid_targets_min": 1970 + }, + { + "epoch": 1.5723053317990026, + "grad_norm": 0.467948758229714, + "learning_rate": 3.814219500536157e-05, + "loss": 0.311, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14820121228694916, + "step": 2050, + "valid_targets_mean": 6153.9, + "valid_targets_min": 1651 + }, + { + "epoch": 1.5761411584196394, + "grad_norm": 0.4306590976594248, + "learning_rate": 3.812606595175963e-05, + "loss": 0.2933, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15877076983451843, + "step": 2055, + "valid_targets_mean": 7671.2, + "valid_targets_min": 2231 + }, + { + "epoch": 1.5799769850402763, + "grad_norm": 0.5006428678911244, + "learning_rate": 3.810987062646825e-05, + "loss": 0.2991, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10005389153957367, + "step": 2060, + "valid_targets_mean": 3124.8, + "valid_targets_min": 467 + }, + { + "epoch": 1.583812811660913, + "grad_norm": 0.5055932753138395, + "learning_rate": 3.809360908870003e-05, + "loss": 0.3053, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17013373970985413, + "step": 2065, + "valid_targets_mean": 6102.0, + "valid_targets_min": 1748 + }, + { + "epoch": 1.5876486382815496, + "grad_norm": 0.5140873519502539, + "learning_rate": 3.807728139790967e-05, + "loss": 0.3062, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1433097869157791, + "step": 2070, + "valid_targets_mean": 4508.4, + "valid_targets_min": 1697 + }, + { + "epoch": 1.5914844649021864, + "grad_norm": 0.6125983989292069, + "learning_rate": 3.8060887613793716e-05, + "loss": 0.3027, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1580856442451477, + "step": 2075, + "valid_targets_mean": 3941.2, + "valid_targets_min": 1223 + }, + { + "epoch": 1.5953202915228233, + "grad_norm": 0.4803744521286387, + "learning_rate": 3.804442779629035e-05, + "loss": 0.2997, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15367859601974487, + "step": 2080, + "valid_targets_mean": 5561.2, + "valid_targets_min": 1552 + }, + { + "epoch": 1.59915611814346, + "grad_norm": 0.5514827103335576, + "learning_rate": 3.802790200557923e-05, + "loss": 0.3042, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16528496146202087, + "step": 2085, + "valid_targets_mean": 4325.5, + "valid_targets_min": 1546 + }, + { + "epoch": 1.6029919447640966, + "grad_norm": 0.5745774591300502, + "learning_rate": 3.801131030208118e-05, + "loss": 0.2988, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17140871286392212, + "step": 2090, + "valid_targets_mean": 4611.9, + "valid_targets_min": 1957 + }, + { + "epoch": 1.6068277713847334, + "grad_norm": 0.5468302868563809, + "learning_rate": 3.799465274645802e-05, + "loss": 0.3071, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12387045472860336, + "step": 2095, + "valid_targets_mean": 3394.4, + "valid_targets_min": 1538 + }, + { + "epoch": 1.6106635980053703, + "grad_norm": 0.625739854965364, + "learning_rate": 3.797792939961235e-05, + "loss": 0.3075, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16092948615550995, + "step": 2100, + "valid_targets_mean": 3414.6, + "valid_targets_min": 1645 + }, + { + "epoch": 1.614499424626007, + "grad_norm": 0.62574315675758, + "learning_rate": 3.796114032268729e-05, + "loss": 0.3044, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14611348509788513, + "step": 2105, + "valid_targets_mean": 3189.5, + "valid_targets_min": 935 + }, + { + "epoch": 1.6183352512466436, + "grad_norm": 0.6033468340393319, + "learning_rate": 3.79442855770663e-05, + "loss": 0.3141, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15901070833206177, + "step": 2110, + "valid_targets_mean": 3584.8, + "valid_targets_min": 1269 + }, + { + "epoch": 1.6221710778672804, + "grad_norm": 0.6797875189387327, + "learning_rate": 3.7927365224372925e-05, + "loss": 0.3108, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1211296021938324, + "step": 2115, + "valid_targets_mean": 2098.1, + "valid_targets_min": 1173 + }, + { + "epoch": 1.6260069044879173, + "grad_norm": 0.5352302018072561, + "learning_rate": 3.791037932647057e-05, + "loss": 0.306, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1388864815235138, + "step": 2120, + "valid_targets_mean": 4135.4, + "valid_targets_min": 1670 + }, + { + "epoch": 1.629842731108554, + "grad_norm": 0.6089667720489725, + "learning_rate": 3.7893327945462315e-05, + "loss": 0.3072, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14827892184257507, + "step": 2125, + "valid_targets_mean": 3372.5, + "valid_targets_min": 1329 + }, + { + "epoch": 1.6336785577291906, + "grad_norm": 0.6058587675669085, + "learning_rate": 3.787621114369062e-05, + "loss": 0.2955, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18027259409427643, + "step": 2130, + "valid_targets_mean": 3950.1, + "valid_targets_min": 2106 + }, + { + "epoch": 1.6375143843498274, + "grad_norm": 0.49108926576115325, + "learning_rate": 3.7859028983737154e-05, + "loss": 0.2813, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12763871252536774, + "step": 2135, + "valid_targets_mean": 4317.1, + "valid_targets_min": 2222 + }, + { + "epoch": 1.6413502109704643, + "grad_norm": 0.601130036934732, + "learning_rate": 3.784178152842254e-05, + "loss": 0.3023, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14744964241981506, + "step": 2140, + "valid_targets_mean": 3463.2, + "valid_targets_min": 1565 + }, + { + "epoch": 1.645186037591101, + "grad_norm": 0.6054464019412783, + "learning_rate": 3.782446884080614e-05, + "loss": 0.3015, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15163356065750122, + "step": 2145, + "valid_targets_mean": 3481.2, + "valid_targets_min": 771 + }, + { + "epoch": 1.6490218642117376, + "grad_norm": 0.5501289999823216, + "learning_rate": 3.7807090984185795e-05, + "loss": 0.309, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14549870789051056, + "step": 2150, + "valid_targets_mean": 4016.5, + "valid_targets_min": 1774 + }, + { + "epoch": 1.6528576908323744, + "grad_norm": 0.6372010280781291, + "learning_rate": 3.7789648022097625e-05, + "loss": 0.287, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12309291958808899, + "step": 2155, + "valid_targets_mean": 2433.4, + "valid_targets_min": 997 + }, + { + "epoch": 1.656693517453011, + "grad_norm": 0.5794955668893225, + "learning_rate": 3.777214001831579e-05, + "loss": 0.2875, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14511778950691223, + "step": 2160, + "valid_targets_mean": 3968.9, + "valid_targets_min": 1741 + }, + { + "epoch": 1.6605293440736477, + "grad_norm": 0.6080987174050998, + "learning_rate": 3.775456703685226e-05, + "loss": 0.3035, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1448732167482376, + "step": 2165, + "valid_targets_mean": 3747.4, + "valid_targets_min": 863 + }, + { + "epoch": 1.6643651706942846, + "grad_norm": 0.5943401177527081, + "learning_rate": 3.773692914195655e-05, + "loss": 0.2939, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13823895156383514, + "step": 2170, + "valid_targets_mean": 3097.6, + "valid_targets_min": 1410 + }, + { + "epoch": 1.6682009973149214, + "grad_norm": 0.5025027870805127, + "learning_rate": 3.7719226398115515e-05, + "loss": 0.283, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12884891033172607, + "step": 2175, + "valid_targets_mean": 4840.6, + "valid_targets_min": 2110 + }, + { + "epoch": 1.672036823935558, + "grad_norm": 0.7555025323097405, + "learning_rate": 3.770145887005313e-05, + "loss": 0.3001, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14058586955070496, + "step": 2180, + "valid_targets_mean": 3410.2, + "valid_targets_min": 1318 + }, + { + "epoch": 1.6758726505561947, + "grad_norm": 0.5570721803502897, + "learning_rate": 3.768362662273021e-05, + "loss": 0.3003, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15026965737342834, + "step": 2185, + "valid_targets_mean": 4229.5, + "valid_targets_min": 1876 + }, + { + "epoch": 1.6797084771768316, + "grad_norm": 0.5996393309634074, + "learning_rate": 3.766572972134419e-05, + "loss": 0.2908, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17062075436115265, + "step": 2190, + "valid_targets_mean": 3783.2, + "valid_targets_min": 2605 + }, + { + "epoch": 1.6835443037974684, + "grad_norm": 0.5856615996742313, + "learning_rate": 3.764776823132891e-05, + "loss": 0.301, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13136203587055206, + "step": 2195, + "valid_targets_mean": 3068.0, + "valid_targets_min": 1446 + }, + { + "epoch": 1.687380130418105, + "grad_norm": 0.5755975846301241, + "learning_rate": 3.762974221835433e-05, + "loss": 0.2993, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14469008147716522, + "step": 2200, + "valid_targets_mean": 3860.8, + "valid_targets_min": 1270 + }, + { + "epoch": 1.6912159570387417, + "grad_norm": 0.5121593731887478, + "learning_rate": 3.761165174832634e-05, + "loss": 0.2959, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15652057528495789, + "step": 2205, + "valid_targets_mean": 5586.5, + "valid_targets_min": 2033 + }, + { + "epoch": 1.6950517836593786, + "grad_norm": 0.5908341036403723, + "learning_rate": 3.7593496887386485e-05, + "loss": 0.2871, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16908356547355652, + "step": 2210, + "valid_targets_mean": 4080.6, + "valid_targets_min": 1924 + }, + { + "epoch": 1.6988876102800154, + "grad_norm": 0.6143923773018835, + "learning_rate": 3.757527770191173e-05, + "loss": 0.2957, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15083667635917664, + "step": 2215, + "valid_targets_mean": 3680.1, + "valid_targets_min": 2098 + }, + { + "epoch": 1.702723436900652, + "grad_norm": 0.63452316254781, + "learning_rate": 3.7556994258514235e-05, + "loss": 0.2973, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19680596888065338, + "step": 2220, + "valid_targets_mean": 4332.0, + "valid_targets_min": 1363 + }, + { + "epoch": 1.7065592635212887, + "grad_norm": 0.6019234005866202, + "learning_rate": 3.753864662404107e-05, + "loss": 0.2887, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15563538670539856, + "step": 2225, + "valid_targets_mean": 3358.1, + "valid_targets_min": 1087 + }, + { + "epoch": 1.7103950901419256, + "grad_norm": 0.6164844267770644, + "learning_rate": 3.752023486557401e-05, + "loss": 0.2988, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1812639832496643, + "step": 2230, + "valid_targets_mean": 4630.6, + "valid_targets_min": 2483 + }, + { + "epoch": 1.7142309167625625, + "grad_norm": 0.5804964380658618, + "learning_rate": 3.75017590504293e-05, + "loss": 0.2929, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16715167462825775, + "step": 2235, + "valid_targets_mean": 4551.2, + "valid_targets_min": 2316 + }, + { + "epoch": 1.718066743383199, + "grad_norm": 0.5388129767463182, + "learning_rate": 3.748321924615736e-05, + "loss": 0.2848, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14066383242607117, + "step": 2240, + "valid_targets_mean": 3923.0, + "valid_targets_min": 2022 + }, + { + "epoch": 1.7219025700038357, + "grad_norm": 0.6395959110738638, + "learning_rate": 3.746461552054256e-05, + "loss": 0.2797, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10199883580207825, + "step": 2245, + "valid_targets_mean": 2890.6, + "valid_targets_min": 1294 + }, + { + "epoch": 1.7257383966244726, + "grad_norm": 0.6282928539272661, + "learning_rate": 3.7445947941603e-05, + "loss": 0.3008, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17430326342582703, + "step": 2250, + "valid_targets_mean": 3638.6, + "valid_targets_min": 1700 + }, + { + "epoch": 1.7295742232451095, + "grad_norm": 0.6284167515282237, + "learning_rate": 3.742721657759021e-05, + "loss": 0.2866, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12495638430118561, + "step": 2255, + "valid_targets_mean": 2722.4, + "valid_targets_min": 1586 + }, + { + "epoch": 1.733410049865746, + "grad_norm": 0.6801017574003632, + "learning_rate": 3.740842149698896e-05, + "loss": 0.2942, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16624921560287476, + "step": 2260, + "valid_targets_mean": 3360.6, + "valid_targets_min": 2407 + }, + { + "epoch": 1.7372458764863827, + "grad_norm": 0.6242292032404948, + "learning_rate": 3.7389562768516945e-05, + "loss": 0.2985, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15616175532341003, + "step": 2265, + "valid_targets_mean": 4063.0, + "valid_targets_min": 2290 + }, + { + "epoch": 1.7410817031070196, + "grad_norm": 0.5792611011474298, + "learning_rate": 3.737064046112458e-05, + "loss": 0.2962, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18799430131912231, + "step": 2270, + "valid_targets_mean": 5192.1, + "valid_targets_min": 2189 + }, + { + "epoch": 1.7449175297276565, + "grad_norm": 0.5953374820885332, + "learning_rate": 3.735165464399475e-05, + "loss": 0.2901, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1848529875278473, + "step": 2275, + "valid_targets_mean": 4296.2, + "valid_targets_min": 2609 + }, + { + "epoch": 1.748753356348293, + "grad_norm": 0.613751207348813, + "learning_rate": 3.733260538654252e-05, + "loss": 0.302, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.207084521651268, + "step": 2280, + "valid_targets_mean": 4534.8, + "valid_targets_min": 2011 + }, + { + "epoch": 1.7525891829689297, + "grad_norm": 0.5491158125445209, + "learning_rate": 3.731349275841491e-05, + "loss": 0.2794, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11974848061800003, + "step": 2285, + "valid_targets_mean": 3611.1, + "valid_targets_min": 1912 + }, + { + "epoch": 1.7564250095895666, + "grad_norm": 0.5788957556803295, + "learning_rate": 3.729431682949063e-05, + "loss": 0.2798, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16036933660507202, + "step": 2290, + "valid_targets_mean": 3801.9, + "valid_targets_min": 2266 + }, + { + "epoch": 1.7602608362102035, + "grad_norm": 0.5255631530995183, + "learning_rate": 3.727507766987983e-05, + "loss": 0.2789, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13088299334049225, + "step": 2295, + "valid_targets_mean": 4454.2, + "valid_targets_min": 904 + }, + { + "epoch": 1.76409666283084, + "grad_norm": 0.6207273354145584, + "learning_rate": 3.725577534992385e-05, + "loss": 0.2884, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16278964281082153, + "step": 2300, + "valid_targets_mean": 3787.4, + "valid_targets_min": 1219 + }, + { + "epoch": 1.7679324894514767, + "grad_norm": 0.6788321354744823, + "learning_rate": 3.7236409940194936e-05, + "loss": 0.2917, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1381135880947113, + "step": 2305, + "valid_targets_mean": 4048.5, + "valid_targets_min": 1658 + }, + { + "epoch": 1.7717683160721136, + "grad_norm": 0.6138190324860064, + "learning_rate": 3.721698151149601e-05, + "loss": 0.292, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14290864765644073, + "step": 2310, + "valid_targets_mean": 3499.8, + "valid_targets_min": 1423 + }, + { + "epoch": 1.7756041426927502, + "grad_norm": 0.53321521444176, + "learning_rate": 3.719749013486041e-05, + "loss": 0.2856, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11363449692726135, + "step": 2315, + "valid_targets_mean": 3917.6, + "valid_targets_min": 2320 + }, + { + "epoch": 1.7794399693133869, + "grad_norm": 0.5673810543381013, + "learning_rate": 3.71779358815516e-05, + "loss": 0.2977, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13837997615337372, + "step": 2320, + "valid_targets_mean": 3650.2, + "valid_targets_min": 1188 + }, + { + "epoch": 1.7832757959340237, + "grad_norm": 0.556285341712241, + "learning_rate": 3.715831882306294e-05, + "loss": 0.2959, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12447488307952881, + "step": 2325, + "valid_targets_mean": 3416.4, + "valid_targets_min": 618 + }, + { + "epoch": 1.7871116225546606, + "grad_norm": 0.7756940500796206, + "learning_rate": 3.7138639031117426e-05, + "loss": 0.291, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13555549085140228, + "step": 2330, + "valid_targets_mean": 2667.1, + "valid_targets_min": 1364 + }, + { + "epoch": 1.7909474491752972, + "grad_norm": 0.685521126401889, + "learning_rate": 3.7118896577667414e-05, + "loss": 0.2823, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1474594622850418, + "step": 2335, + "valid_targets_mean": 3505.4, + "valid_targets_min": 1397 + }, + { + "epoch": 1.7947832757959339, + "grad_norm": 0.5384825994083828, + "learning_rate": 3.709909153489435e-05, + "loss": 0.276, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14623813331127167, + "step": 2340, + "valid_targets_mean": 4666.4, + "valid_targets_min": 1415 + }, + { + "epoch": 1.7986191024165707, + "grad_norm": 0.5729521998414476, + "learning_rate": 3.707922397520852e-05, + "loss": 0.284, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15341173112392426, + "step": 2345, + "valid_targets_mean": 3801.8, + "valid_targets_min": 621 + }, + { + "epoch": 1.8024549290372076, + "grad_norm": 0.640658305392353, + "learning_rate": 3.705929397124878e-05, + "loss": 0.2748, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1091216653585434, + "step": 2350, + "valid_targets_mean": 3552.0, + "valid_targets_min": 1290 + }, + { + "epoch": 1.8062907556578442, + "grad_norm": 0.687371115682397, + "learning_rate": 3.703930159588229e-05, + "loss": 0.2749, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1161675676703453, + "step": 2355, + "valid_targets_mean": 2124.8, + "valid_targets_min": 642 + }, + { + "epoch": 1.810126582278481, + "grad_norm": 0.6756849168573179, + "learning_rate": 3.701924692220427e-05, + "loss": 0.2756, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1516008973121643, + "step": 2360, + "valid_targets_mean": 3453.9, + "valid_targets_min": 2109 + }, + { + "epoch": 1.8139624088991178, + "grad_norm": 0.607972387942167, + "learning_rate": 3.699913002353769e-05, + "loss": 0.2883, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20429207384586334, + "step": 2365, + "valid_targets_mean": 4593.1, + "valid_targets_min": 2786 + }, + { + "epoch": 1.8177982355197546, + "grad_norm": 0.6004412179117972, + "learning_rate": 3.6978950973433014e-05, + "loss": 0.3029, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12453005462884903, + "step": 2370, + "valid_targets_mean": 3278.0, + "valid_targets_min": 1510 + }, + { + "epoch": 1.8216340621403913, + "grad_norm": 0.7013739900392476, + "learning_rate": 3.695870984566799e-05, + "loss": 0.2852, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1814599484205246, + "step": 2375, + "valid_targets_mean": 3222.0, + "valid_targets_min": 987 + }, + { + "epoch": 1.825469888761028, + "grad_norm": 0.5646730648474158, + "learning_rate": 3.693840671424727e-05, + "loss": 0.2886, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12595924735069275, + "step": 2380, + "valid_targets_mean": 3659.5, + "valid_targets_min": 1609 + }, + { + "epoch": 1.8293057153816648, + "grad_norm": 0.5478228499036945, + "learning_rate": 3.6918041653402246e-05, + "loss": 0.282, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15491420030593872, + "step": 2385, + "valid_targets_mean": 4652.5, + "valid_targets_min": 1700 + }, + { + "epoch": 1.8331415420023016, + "grad_norm": 0.47005227719941683, + "learning_rate": 3.6897614737590716e-05, + "loss": 0.2808, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1260187327861786, + "step": 2390, + "valid_targets_mean": 5438.1, + "valid_targets_min": 1767 + }, + { + "epoch": 1.8369773686229383, + "grad_norm": 0.4975714036726387, + "learning_rate": 3.687712604149662e-05, + "loss": 0.2857, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1307644248008728, + "step": 2395, + "valid_targets_mean": 4504.2, + "valid_targets_min": 2381 + }, + { + "epoch": 1.840813195243575, + "grad_norm": 0.709426947098586, + "learning_rate": 3.685657564002981e-05, + "loss": 0.2768, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12030431628227234, + "step": 2400, + "valid_targets_mean": 2680.2, + "valid_targets_min": 1542 + }, + { + "epoch": 1.8446490218642118, + "grad_norm": 0.5483485943748535, + "learning_rate": 3.68359636083257e-05, + "loss": 0.2868, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10411541163921356, + "step": 2405, + "valid_targets_mean": 3123.9, + "valid_targets_min": 1476 + }, + { + "epoch": 1.8484848484848486, + "grad_norm": 0.6007809252358458, + "learning_rate": 3.681529002174505e-05, + "loss": 0.2829, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15618902444839478, + "step": 2410, + "valid_targets_mean": 3869.0, + "valid_targets_min": 1711 + }, + { + "epoch": 1.8523206751054853, + "grad_norm": 0.5896129966300073, + "learning_rate": 3.6794554955873695e-05, + "loss": 0.291, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13634750247001648, + "step": 2415, + "valid_targets_mean": 3236.9, + "valid_targets_min": 1669 + }, + { + "epoch": 1.856156501726122, + "grad_norm": 0.6580381597858896, + "learning_rate": 3.6773758486522215e-05, + "loss": 0.2976, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13955433666706085, + "step": 2420, + "valid_targets_mean": 3207.0, + "valid_targets_min": 1318 + }, + { + "epoch": 1.8599923283467588, + "grad_norm": 0.7135644964979131, + "learning_rate": 3.6752900689725714e-05, + "loss": 0.292, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17298159003257751, + "step": 2425, + "valid_targets_mean": 2656.8, + "valid_targets_min": 1002 + }, + { + "epoch": 1.8638281549673956, + "grad_norm": 0.6346834827387328, + "learning_rate": 3.673198164174351e-05, + "loss": 0.3091, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1140647828578949, + "step": 2430, + "valid_targets_mean": 2533.6, + "valid_targets_min": 845 + }, + { + "epoch": 1.8676639815880323, + "grad_norm": 0.5378223305556128, + "learning_rate": 3.6711001419058855e-05, + "loss": 0.2826, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1484578251838684, + "step": 2435, + "valid_targets_mean": 4455.6, + "valid_targets_min": 2611 + }, + { + "epoch": 1.871499808208669, + "grad_norm": 0.5929308517414741, + "learning_rate": 3.668996009837868e-05, + "loss": 0.3031, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13273712992668152, + "step": 2440, + "valid_targets_mean": 3460.1, + "valid_targets_min": 1241 + }, + { + "epoch": 1.8753356348293058, + "grad_norm": 0.6291321948562841, + "learning_rate": 3.666885775663331e-05, + "loss": 0.3067, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19320976734161377, + "step": 2445, + "valid_targets_mean": 3751.0, + "valid_targets_min": 992 + }, + { + "epoch": 1.8791714614499426, + "grad_norm": 0.6254081791655244, + "learning_rate": 3.664769447097614e-05, + "loss": 0.2869, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14107190072536469, + "step": 2450, + "valid_targets_mean": 3815.0, + "valid_targets_min": 1133 + }, + { + "epoch": 1.883007288070579, + "grad_norm": 0.5964459860191034, + "learning_rate": 3.662647031878341e-05, + "loss": 0.2953, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16643989086151123, + "step": 2455, + "valid_targets_mean": 4047.4, + "valid_targets_min": 2256 + }, + { + "epoch": 1.886843114691216, + "grad_norm": 0.5673046941108633, + "learning_rate": 3.660518537765388e-05, + "loss": 0.2929, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11981415003538132, + "step": 2460, + "valid_targets_mean": 3817.2, + "valid_targets_min": 1919 + }, + { + "epoch": 1.8906789413118528, + "grad_norm": 0.6197203825736891, + "learning_rate": 3.6583839725408584e-05, + "loss": 0.2993, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1451350748538971, + "step": 2465, + "valid_targets_mean": 3366.4, + "valid_targets_min": 2763 + }, + { + "epoch": 1.8945147679324894, + "grad_norm": 0.7004814867528292, + "learning_rate": 3.6562433440090514e-05, + "loss": 0.2976, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14817191660404205, + "step": 2470, + "valid_targets_mean": 2599.4, + "valid_targets_min": 617 + }, + { + "epoch": 1.898350594553126, + "grad_norm": 0.5211413845290879, + "learning_rate": 3.654096659996435e-05, + "loss": 0.273, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13184261322021484, + "step": 2475, + "valid_targets_mean": 3771.2, + "valid_targets_min": 3147 + }, + { + "epoch": 1.902186421173763, + "grad_norm": 0.5823338759133069, + "learning_rate": 3.651943928351616e-05, + "loss": 0.2801, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14209085702896118, + "step": 2480, + "valid_targets_mean": 3780.9, + "valid_targets_min": 1844 + }, + { + "epoch": 1.9060222477943998, + "grad_norm": 0.5314895001542878, + "learning_rate": 3.6497851569453116e-05, + "loss": 0.2865, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13384810090065002, + "step": 2485, + "valid_targets_mean": 3895.0, + "valid_targets_min": 2330 + }, + { + "epoch": 1.9098580744150364, + "grad_norm": 0.795577253207001, + "learning_rate": 3.647620353670325e-05, + "loss": 0.2916, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14383789896965027, + "step": 2490, + "valid_targets_mean": 2856.8, + "valid_targets_min": 1066 + }, + { + "epoch": 1.913693901035673, + "grad_norm": 0.5873857027062859, + "learning_rate": 3.64544952644151e-05, + "loss": 0.2801, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1259903609752655, + "step": 2495, + "valid_targets_mean": 3496.4, + "valid_targets_min": 1713 + }, + { + "epoch": 1.91752972765631, + "grad_norm": 0.605983728971514, + "learning_rate": 3.643272683195744e-05, + "loss": 0.2943, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14265413582324982, + "step": 2500, + "valid_targets_mean": 3918.8, + "valid_targets_min": 981 + }, + { + "epoch": 1.9213655542769468, + "grad_norm": 0.5417501197751011, + "learning_rate": 3.6410898318919026e-05, + "loss": 0.2847, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15600088238716125, + "step": 2505, + "valid_targets_mean": 4680.2, + "valid_targets_min": 3325 + }, + { + "epoch": 1.9252013808975834, + "grad_norm": 0.690889225143732, + "learning_rate": 3.6389009805108263e-05, + "loss": 0.2908, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1505012959241867, + "step": 2510, + "valid_targets_mean": 3138.9, + "valid_targets_min": 1353 + }, + { + "epoch": 1.92903720751822, + "grad_norm": 0.519677474296737, + "learning_rate": 3.636706137055292e-05, + "loss": 0.2835, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1220041811466217, + "step": 2515, + "valid_targets_mean": 3986.6, + "valid_targets_min": 2273 + }, + { + "epoch": 1.932873034138857, + "grad_norm": 0.5892576396920717, + "learning_rate": 3.6345053095499865e-05, + "loss": 0.2769, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11959492415189743, + "step": 2520, + "valid_targets_mean": 3204.4, + "valid_targets_min": 2106 + }, + { + "epoch": 1.9367088607594938, + "grad_norm": 0.5567093909016859, + "learning_rate": 3.632298506041474e-05, + "loss": 0.2694, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12005296349525452, + "step": 2525, + "valid_targets_mean": 3788.4, + "valid_targets_min": 1687 + }, + { + "epoch": 1.9405446873801304, + "grad_norm": 0.6185352269217175, + "learning_rate": 3.630085734598167e-05, + "loss": 0.2781, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15298223495483398, + "step": 2530, + "valid_targets_mean": 3502.8, + "valid_targets_min": 2246 + }, + { + "epoch": 1.944380514000767, + "grad_norm": 0.5959233529109299, + "learning_rate": 3.6278670033103e-05, + "loss": 0.2793, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1341932862997055, + "step": 2535, + "valid_targets_mean": 3663.8, + "valid_targets_min": 906 + }, + { + "epoch": 1.948216340621404, + "grad_norm": 0.6247630103833505, + "learning_rate": 3.6256423202898974e-05, + "loss": 0.2751, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1733320653438568, + "step": 2540, + "valid_targets_mean": 4811.0, + "valid_targets_min": 1874 + }, + { + "epoch": 1.9520521672420408, + "grad_norm": 0.5927228538302737, + "learning_rate": 3.62341169367074e-05, + "loss": 0.2711, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14768466353416443, + "step": 2545, + "valid_targets_mean": 3815.6, + "valid_targets_min": 873 + }, + { + "epoch": 1.9558879938626774, + "grad_norm": 0.6331658646777687, + "learning_rate": 3.6211751316083447e-05, + "loss": 0.2798, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12714754045009613, + "step": 2550, + "valid_targets_mean": 3252.1, + "valid_targets_min": 1697 + }, + { + "epoch": 1.959723820483314, + "grad_norm": 0.8477096739584258, + "learning_rate": 3.6189326422799254e-05, + "loss": 0.2773, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11219899356365204, + "step": 2555, + "valid_targets_mean": 2395.0, + "valid_targets_min": 897 + }, + { + "epoch": 1.963559647103951, + "grad_norm": 0.6223093809111021, + "learning_rate": 3.616684233884371e-05, + "loss": 0.2977, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1319318413734436, + "step": 2560, + "valid_targets_mean": 3104.0, + "valid_targets_min": 1695 + }, + { + "epoch": 1.9673954737245878, + "grad_norm": 0.6253752258508035, + "learning_rate": 3.614429914642206e-05, + "loss": 0.2841, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1328519582748413, + "step": 2565, + "valid_targets_mean": 3211.9, + "valid_targets_min": 1712 + }, + { + "epoch": 1.9712313003452244, + "grad_norm": 0.582570850897299, + "learning_rate": 3.612169692795571e-05, + "loss": 0.2801, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12549902498722076, + "step": 2570, + "valid_targets_mean": 3500.5, + "valid_targets_min": 993 + }, + { + "epoch": 1.975067126965861, + "grad_norm": 0.5709602212073817, + "learning_rate": 3.6099035766081836e-05, + "loss": 0.2725, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14696191251277924, + "step": 2575, + "valid_targets_mean": 4181.4, + "valid_targets_min": 1679 + }, + { + "epoch": 1.978902953586498, + "grad_norm": 0.5420138789229847, + "learning_rate": 3.607631574365316e-05, + "loss": 0.2806, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1264781653881073, + "step": 2580, + "valid_targets_mean": 3779.9, + "valid_targets_min": 2250 + }, + { + "epoch": 1.9827387802071348, + "grad_norm": 0.5629334321546393, + "learning_rate": 3.6053536943737555e-05, + "loss": 0.2871, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1410553753376007, + "step": 2585, + "valid_targets_mean": 4175.8, + "valid_targets_min": 2892 + }, + { + "epoch": 1.9865746068277714, + "grad_norm": 0.5402934528999608, + "learning_rate": 3.6030699449617846e-05, + "loss": 0.2904, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1486661285161972, + "step": 2590, + "valid_targets_mean": 4485.5, + "valid_targets_min": 2965 + }, + { + "epoch": 1.990410433448408, + "grad_norm": 0.5680356902597896, + "learning_rate": 3.6007803344791415e-05, + "loss": 0.2816, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11320558935403824, + "step": 2595, + "valid_targets_mean": 3737.5, + "valid_targets_min": 1817 + }, + { + "epoch": 1.994246260069045, + "grad_norm": 0.5433812736749285, + "learning_rate": 3.598484871296996e-05, + "loss": 0.2836, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16301412880420685, + "step": 2600, + "valid_targets_mean": 4726.0, + "valid_targets_min": 1727 + }, + { + "epoch": 1.9980820866896818, + "grad_norm": 0.5258878606871368, + "learning_rate": 3.5961835638079136e-05, + "loss": 0.2784, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12243422865867615, + "step": 2605, + "valid_targets_mean": 3801.2, + "valid_targets_min": 1959 + }, + { + "epoch": 2.0015343306482545, + "grad_norm": 0.4259148994288393, + "learning_rate": 3.5938764204258306e-05, + "loss": 0.2503, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1014866977930069, + "step": 2610, + "valid_targets_mean": 8514.0, + "valid_targets_min": 6788 + }, + { + "epoch": 2.0053701572688913, + "grad_norm": 0.4586019508133762, + "learning_rate": 3.5915634495860174e-05, + "loss": 0.2312, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11351286619901657, + "step": 2615, + "valid_targets_mean": 6668.0, + "valid_targets_min": 5933 + }, + { + "epoch": 2.009205983889528, + "grad_norm": 0.5355826662436876, + "learning_rate": 3.589244659745052e-05, + "loss": 0.2221, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10232991725206375, + "step": 2620, + "valid_targets_mean": 4107.0, + "valid_targets_min": 161 + }, + { + "epoch": 2.013041810510165, + "grad_norm": 0.38681533487536696, + "learning_rate": 3.5869200593807886e-05, + "loss": 0.2175, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09863924980163574, + "step": 2625, + "valid_targets_mean": 7946.0, + "valid_targets_min": 5596 + }, + { + "epoch": 2.0168776371308015, + "grad_norm": 0.4251220893116979, + "learning_rate": 3.584589656992322e-05, + "loss": 0.2089, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11178788542747498, + "step": 2630, + "valid_targets_mean": 7392.8, + "valid_targets_min": 5629 + }, + { + "epoch": 2.0207134637514383, + "grad_norm": 0.41140039652117055, + "learning_rate": 3.5822534610999645e-05, + "loss": 0.2128, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10888037085533142, + "step": 2635, + "valid_targets_mean": 6879.6, + "valid_targets_min": 5113 + }, + { + "epoch": 2.024549290372075, + "grad_norm": 0.46065799440154953, + "learning_rate": 3.5799114802452064e-05, + "loss": 0.2303, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11086660623550415, + "step": 2640, + "valid_targets_mean": 6919.5, + "valid_targets_min": 5080 + }, + { + "epoch": 2.028385116992712, + "grad_norm": 0.4128247898588806, + "learning_rate": 3.577563722990691e-05, + "loss": 0.223, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09855577349662781, + "step": 2645, + "valid_targets_mean": 6377.0, + "valid_targets_min": 3939 + }, + { + "epoch": 2.0322209436133485, + "grad_norm": 0.4035224703888014, + "learning_rate": 3.5752101979201816e-05, + "loss": 0.2111, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10646621137857437, + "step": 2650, + "valid_targets_mean": 6392.2, + "valid_targets_min": 4355 + }, + { + "epoch": 2.0360567702339853, + "grad_norm": 0.48881316624884236, + "learning_rate": 3.572850913638527e-05, + "loss": 0.2054, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1203937828540802, + "step": 2655, + "valid_targets_mean": 5196.5, + "valid_targets_min": 137 + }, + { + "epoch": 2.039892596854622, + "grad_norm": 0.41276856423959996, + "learning_rate": 3.5704858787716344e-05, + "loss": 0.2094, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09328220039606094, + "step": 2660, + "valid_targets_mean": 7344.4, + "valid_targets_min": 5244 + }, + { + "epoch": 2.043728423475259, + "grad_norm": 0.3854003482659584, + "learning_rate": 3.5681151019664364e-05, + "loss": 0.2029, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08878053724765778, + "step": 2665, + "valid_targets_mean": 7921.0, + "valid_targets_min": 5311 + }, + { + "epoch": 2.0475642500958955, + "grad_norm": 0.4268066886584198, + "learning_rate": 3.5657385918908566e-05, + "loss": 0.2126, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10712914913892746, + "step": 2670, + "valid_targets_mean": 7931.2, + "valid_targets_min": 6015 + }, + { + "epoch": 2.0514000767165324, + "grad_norm": 0.4524027150252451, + "learning_rate": 3.563356357233783e-05, + "loss": 0.2026, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09540455043315887, + "step": 2675, + "valid_targets_mean": 7640.8, + "valid_targets_min": 5248 + }, + { + "epoch": 2.055235903337169, + "grad_norm": 0.46919147009791706, + "learning_rate": 3.560968406705034e-05, + "loss": 0.2011, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10013915598392487, + "step": 2680, + "valid_targets_mean": 7249.6, + "valid_targets_min": 5217 + }, + { + "epoch": 2.059071729957806, + "grad_norm": 0.4128397043485362, + "learning_rate": 3.55857474903532e-05, + "loss": 0.2092, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11581048369407654, + "step": 2685, + "valid_targets_mean": 7619.9, + "valid_targets_min": 5094 + }, + { + "epoch": 2.0629075565784425, + "grad_norm": 0.38160480951027365, + "learning_rate": 3.5561753929762276e-05, + "loss": 0.2029, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09671203792095184, + "step": 2690, + "valid_targets_mean": 7732.8, + "valid_targets_min": 5842 + }, + { + "epoch": 2.0667433831990794, + "grad_norm": 0.40235340572735895, + "learning_rate": 3.553770347300169e-05, + "loss": 0.205, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10450824350118637, + "step": 2695, + "valid_targets_mean": 7952.9, + "valid_targets_min": 5494 + }, + { + "epoch": 2.070579209819716, + "grad_norm": 0.39306509800561873, + "learning_rate": 3.5513596208003625e-05, + "loss": 0.2218, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10905580222606659, + "step": 2700, + "valid_targets_mean": 8001.0, + "valid_targets_min": 5454 + }, + { + "epoch": 2.074415036440353, + "grad_norm": 0.5081587117422275, + "learning_rate": 3.548943222290794e-05, + "loss": 0.226, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07942081242799759, + "step": 2705, + "valid_targets_mean": 2985.1, + "valid_targets_min": 1896 + }, + { + "epoch": 2.0782508630609895, + "grad_norm": 0.3858625489546898, + "learning_rate": 3.546521160606192e-05, + "loss": 0.1872, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08901523053646088, + "step": 2710, + "valid_targets_mean": 6714.6, + "valid_targets_min": 4917 + }, + { + "epoch": 2.0820866896816264, + "grad_norm": 0.3981566886090765, + "learning_rate": 3.544093444601985e-05, + "loss": 0.2023, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11059673130512238, + "step": 2715, + "valid_targets_mean": 7595.2, + "valid_targets_min": 5488 + }, + { + "epoch": 2.085922516302263, + "grad_norm": 0.4562140502546654, + "learning_rate": 3.5416600831542763e-05, + "loss": 0.2028, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11004103720188141, + "step": 2720, + "valid_targets_mean": 7046.1, + "valid_targets_min": 5579 + }, + { + "epoch": 2.0897583429229, + "grad_norm": 0.43724151712971254, + "learning_rate": 3.539221085159811e-05, + "loss": 0.2104, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1022379919886589, + "step": 2725, + "valid_targets_mean": 7108.0, + "valid_targets_min": 4963 + }, + { + "epoch": 2.0935941695435365, + "grad_norm": 0.4229479295056299, + "learning_rate": 3.536776459535942e-05, + "loss": 0.206, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10431510210037231, + "step": 2730, + "valid_targets_mean": 6810.1, + "valid_targets_min": 5666 + }, + { + "epoch": 2.0974299961641734, + "grad_norm": 0.3868425529854389, + "learning_rate": 3.5343262152205966e-05, + "loss": 0.1981, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09625992178916931, + "step": 2735, + "valid_targets_mean": 8068.5, + "valid_targets_min": 5524 + }, + { + "epoch": 2.1012658227848102, + "grad_norm": 0.43808206452842624, + "learning_rate": 3.531870361172245e-05, + "loss": 0.1793, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08531318604946136, + "step": 2740, + "valid_targets_mean": 6553.8, + "valid_targets_min": 3621 + }, + { + "epoch": 2.105101649405447, + "grad_norm": 0.4119367718285352, + "learning_rate": 3.529408906369868e-05, + "loss": 0.1844, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09665936976671219, + "step": 2745, + "valid_targets_mean": 6762.5, + "valid_targets_min": 4968 + }, + { + "epoch": 2.1089374760260835, + "grad_norm": 0.5009646745027726, + "learning_rate": 3.5269418598129254e-05, + "loss": 0.1967, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09908580034971237, + "step": 2750, + "valid_targets_mean": 6143.9, + "valid_targets_min": 4917 + }, + { + "epoch": 2.1127733026467204, + "grad_norm": 0.5402941539103558, + "learning_rate": 3.524469230521317e-05, + "loss": 0.2037, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1397203654050827, + "step": 2755, + "valid_targets_mean": 4583.8, + "valid_targets_min": 215 + }, + { + "epoch": 2.1166091292673572, + "grad_norm": 0.37223399419054704, + "learning_rate": 3.521991027535357e-05, + "loss": 0.21, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10517659783363342, + "step": 2760, + "valid_targets_mean": 7317.1, + "valid_targets_min": 4948 + }, + { + "epoch": 2.1204449558879936, + "grad_norm": 0.42209973021450164, + "learning_rate": 3.519507259915738e-05, + "loss": 0.2235, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10653809458017349, + "step": 2765, + "valid_targets_mean": 6437.6, + "valid_targets_min": 4510 + }, + { + "epoch": 2.1242807825086305, + "grad_norm": 0.4538081500555624, + "learning_rate": 3.5170179367434964e-05, + "loss": 0.2153, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11629869788885117, + "step": 2770, + "valid_targets_mean": 7707.9, + "valid_targets_min": 5796 + }, + { + "epoch": 2.1281166091292674, + "grad_norm": 0.37931375675483614, + "learning_rate": 3.514523067119982e-05, + "loss": 0.2039, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10018473863601685, + "step": 2775, + "valid_targets_mean": 7900.9, + "valid_targets_min": 5643 + }, + { + "epoch": 2.1319524357499042, + "grad_norm": 0.4159165214917134, + "learning_rate": 3.512022660166822e-05, + "loss": 0.2061, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11328540742397308, + "step": 2780, + "valid_targets_mean": 7627.2, + "valid_targets_min": 5742 + }, + { + "epoch": 2.1357882623705406, + "grad_norm": 0.39340511243690524, + "learning_rate": 3.509516725025889e-05, + "loss": 0.2103, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1197398453950882, + "step": 2785, + "valid_targets_mean": 8725.6, + "valid_targets_min": 5141 + }, + { + "epoch": 2.1396240889911775, + "grad_norm": 0.42111664604371457, + "learning_rate": 3.507005270859269e-05, + "loss": 0.2062, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11561287939548492, + "step": 2790, + "valid_targets_mean": 6967.8, + "valid_targets_min": 6088 + }, + { + "epoch": 2.1434599156118144, + "grad_norm": 0.3866554139606131, + "learning_rate": 3.504488306849224e-05, + "loss": 0.2098, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09691400825977325, + "step": 2795, + "valid_targets_mean": 7140.1, + "valid_targets_min": 4834 + }, + { + "epoch": 2.1472957422324512, + "grad_norm": 0.4214821230672719, + "learning_rate": 3.501965842198163e-05, + "loss": 0.2081, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09585778415203094, + "step": 2800, + "valid_targets_mean": 6706.9, + "valid_targets_min": 5108 + }, + { + "epoch": 2.1511315688530876, + "grad_norm": 0.45789236883064205, + "learning_rate": 3.499437886128605e-05, + "loss": 0.2116, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10418568551540375, + "step": 2805, + "valid_targets_mean": 5931.4, + "valid_targets_min": 4708 + }, + { + "epoch": 2.1549673954737245, + "grad_norm": 0.39428340338579154, + "learning_rate": 3.496904447883147e-05, + "loss": 0.2027, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09633439779281616, + "step": 2810, + "valid_targets_mean": 7145.6, + "valid_targets_min": 5280 + }, + { + "epoch": 2.1588032220943614, + "grad_norm": 0.728529846155519, + "learning_rate": 3.49436553672443e-05, + "loss": 0.1986, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06942793726921082, + "step": 2815, + "valid_targets_mean": 1554.9, + "valid_targets_min": 592 + }, + { + "epoch": 2.1626390487149982, + "grad_norm": 0.47653585534724785, + "learning_rate": 3.491821161935103e-05, + "loss": 0.2009, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11830389499664307, + "step": 2820, + "valid_targets_mean": 7069.5, + "valid_targets_min": 4765 + }, + { + "epoch": 2.1664748753356347, + "grad_norm": 0.44234984276986156, + "learning_rate": 3.4892713328177936e-05, + "loss": 0.2036, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10221248865127563, + "step": 2825, + "valid_targets_mean": 6193.4, + "valid_targets_min": 5587 + }, + { + "epoch": 2.1703107019562715, + "grad_norm": 0.4277702499080298, + "learning_rate": 3.486716058695069e-05, + "loss": 0.2058, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10692169517278671, + "step": 2830, + "valid_targets_mean": 6612.2, + "valid_targets_min": 4894 + }, + { + "epoch": 2.1741465285769084, + "grad_norm": 0.4392408239723064, + "learning_rate": 3.484155348909405e-05, + "loss": 0.1989, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0955607146024704, + "step": 2835, + "valid_targets_mean": 6605.9, + "valid_targets_min": 3795 + }, + { + "epoch": 2.1779823551975452, + "grad_norm": 0.44123781645715887, + "learning_rate": 3.481589212823151e-05, + "loss": 0.2039, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10345818102359772, + "step": 2840, + "valid_targets_mean": 7186.8, + "valid_targets_min": 4890 + }, + { + "epoch": 2.1818181818181817, + "grad_norm": 0.4612942811808649, + "learning_rate": 3.4790176598184966e-05, + "loss": 0.2069, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1038781926035881, + "step": 2845, + "valid_targets_mean": 6544.4, + "valid_targets_min": 4615 + }, + { + "epoch": 2.1856540084388185, + "grad_norm": 0.43045032849009285, + "learning_rate": 3.476440699297436e-05, + "loss": 0.2097, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10975253582000732, + "step": 2850, + "valid_targets_mean": 6339.2, + "valid_targets_min": 4798 + }, + { + "epoch": 2.1894898350594554, + "grad_norm": 0.593218865806326, + "learning_rate": 3.473858340681732e-05, + "loss": 0.1761, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11805373430252075, + "step": 2855, + "valid_targets_mean": 4558.8, + "valid_targets_min": 148 + }, + { + "epoch": 2.1933256616800922, + "grad_norm": 0.4339425341705898, + "learning_rate": 3.471270593412887e-05, + "loss": 0.2052, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11250495165586472, + "step": 2860, + "valid_targets_mean": 6807.9, + "valid_targets_min": 5164 + }, + { + "epoch": 2.1971614883007287, + "grad_norm": 0.4146937846006198, + "learning_rate": 3.4686774669521035e-05, + "loss": 0.1985, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10236822813749313, + "step": 2865, + "valid_targets_mean": 7011.8, + "valid_targets_min": 4974 + }, + { + "epoch": 2.2009973149213655, + "grad_norm": 0.42513715243643857, + "learning_rate": 3.4660789707802505e-05, + "loss": 0.2062, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.093147411942482, + "step": 2870, + "valid_targets_mean": 5794.2, + "valid_targets_min": 4809 + }, + { + "epoch": 2.2048331415420024, + "grad_norm": 0.4309010076873453, + "learning_rate": 3.4634751143978317e-05, + "loss": 0.1941, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09122869372367859, + "step": 2875, + "valid_targets_mean": 6401.8, + "valid_targets_min": 4842 + }, + { + "epoch": 2.2086689681626392, + "grad_norm": 0.4406677340477164, + "learning_rate": 3.4608659073249465e-05, + "loss": 0.2, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10238547623157501, + "step": 2880, + "valid_targets_mean": 6182.1, + "valid_targets_min": 5116 + }, + { + "epoch": 2.2125047947832757, + "grad_norm": 0.39227887564395447, + "learning_rate": 3.458251359101257e-05, + "loss": 0.2002, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09204268455505371, + "step": 2885, + "valid_targets_mean": 6165.1, + "valid_targets_min": 5764 + }, + { + "epoch": 2.2163406214039125, + "grad_norm": 0.8214859058746761, + "learning_rate": 3.455631479285956e-05, + "loss": 0.2199, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08113489300012589, + "step": 2890, + "valid_targets_mean": 1428.2, + "valid_targets_min": 1062 + }, + { + "epoch": 2.2201764480245494, + "grad_norm": 0.8045043225949363, + "learning_rate": 3.453006277457726e-05, + "loss": 0.1987, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09542155265808105, + "step": 2895, + "valid_targets_mean": 1663.0, + "valid_targets_min": 960 + }, + { + "epoch": 2.224012274645186, + "grad_norm": 0.8838631540618118, + "learning_rate": 3.4503757632147094e-05, + "loss": 0.1955, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09389881044626236, + "step": 2900, + "valid_targets_mean": 1476.5, + "valid_targets_min": 933 + }, + { + "epoch": 2.2278481012658227, + "grad_norm": 0.8884148491344682, + "learning_rate": 3.447739946174474e-05, + "loss": 0.1961, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09700118750333786, + "step": 2905, + "valid_targets_mean": 1478.1, + "valid_targets_min": 881 + }, + { + "epoch": 2.2316839278864595, + "grad_norm": 0.7328042541051814, + "learning_rate": 3.445098835973969e-05, + "loss": 0.197, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1147749125957489, + "step": 2910, + "valid_targets_mean": 1640.6, + "valid_targets_min": 869 + }, + { + "epoch": 2.2355197545070964, + "grad_norm": 0.8618844507349648, + "learning_rate": 3.4424524422695036e-05, + "loss": 0.1984, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10487952828407288, + "step": 2915, + "valid_targets_mean": 1709.0, + "valid_targets_min": 1090 + }, + { + "epoch": 2.2393555811277333, + "grad_norm": 0.7896346788878136, + "learning_rate": 3.439800774736699e-05, + "loss": 0.1871, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07611280679702759, + "step": 2920, + "valid_targets_mean": 1157.9, + "valid_targets_min": 761 + }, + { + "epoch": 2.2431914077483697, + "grad_norm": 0.7284271341019, + "learning_rate": 3.437143843070462e-05, + "loss": 0.2054, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09404310584068298, + "step": 2925, + "valid_targets_mean": 1553.6, + "valid_targets_min": 1241 + }, + { + "epoch": 2.2470272343690065, + "grad_norm": 0.7848896128035879, + "learning_rate": 3.434481656984941e-05, + "loss": 0.1897, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08897452801465988, + "step": 2930, + "valid_targets_mean": 1538.1, + "valid_targets_min": 733 + }, + { + "epoch": 2.2508630609896434, + "grad_norm": 0.7481738405489077, + "learning_rate": 3.431814226213501e-05, + "loss": 0.1952, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05928608775138855, + "step": 2935, + "valid_targets_mean": 1250.6, + "valid_targets_min": 835 + }, + { + "epoch": 2.25469888761028, + "grad_norm": 0.7458961209433825, + "learning_rate": 3.429141560508679e-05, + "loss": 0.187, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10857290029525757, + "step": 2940, + "valid_targets_mean": 1596.2, + "valid_targets_min": 1046 + }, + { + "epoch": 2.2585347142309167, + "grad_norm": 0.762107575883253, + "learning_rate": 3.426463669642153e-05, + "loss": 0.1987, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09376601874828339, + "step": 2945, + "valid_targets_mean": 1646.5, + "valid_targets_min": 797 + }, + { + "epoch": 2.2623705408515535, + "grad_norm": 0.7684355309444909, + "learning_rate": 3.423780563404703e-05, + "loss": 0.1869, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09529495984315872, + "step": 2950, + "valid_targets_mean": 1448.9, + "valid_targets_min": 570 + }, + { + "epoch": 2.2662063674721904, + "grad_norm": 0.9604513970392569, + "learning_rate": 3.4210922516061804e-05, + "loss": 0.1816, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06986501812934875, + "step": 2955, + "valid_targets_mean": 1164.1, + "valid_targets_min": 548 + }, + { + "epoch": 2.270042194092827, + "grad_norm": 0.8394753058253241, + "learning_rate": 3.418398744075467e-05, + "loss": 0.1795, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09428368508815765, + "step": 2960, + "valid_targets_mean": 1338.0, + "valid_targets_min": 682 + }, + { + "epoch": 2.2738780207134637, + "grad_norm": 0.7581382485267572, + "learning_rate": 3.415700050660441e-05, + "loss": 0.1931, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09718036651611328, + "step": 2965, + "valid_targets_mean": 1632.8, + "valid_targets_min": 896 + }, + { + "epoch": 2.2777138473341005, + "grad_norm": 0.805266112570685, + "learning_rate": 3.412996181227941e-05, + "loss": 0.1954, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1228247582912445, + "step": 2970, + "valid_targets_mean": 1806.5, + "valid_targets_min": 597 + }, + { + "epoch": 2.2815496739547374, + "grad_norm": 0.7750390034620877, + "learning_rate": 3.410287145663731e-05, + "loss": 0.2004, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08355341106653214, + "step": 2975, + "valid_targets_mean": 1289.9, + "valid_targets_min": 706 + }, + { + "epoch": 2.285385500575374, + "grad_norm": 0.7409532267026048, + "learning_rate": 3.4075729538724605e-05, + "loss": 0.1899, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06073107570409775, + "step": 2980, + "valid_targets_mean": 1354.1, + "valid_targets_min": 754 + }, + { + "epoch": 2.2892213271960107, + "grad_norm": 0.7462747618584279, + "learning_rate": 3.404853615777635e-05, + "loss": 0.1872, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08665506541728973, + "step": 2985, + "valid_targets_mean": 1461.9, + "valid_targets_min": 819 + }, + { + "epoch": 2.2930571538166475, + "grad_norm": 0.7277898614344414, + "learning_rate": 3.402129141321572e-05, + "loss": 0.1929, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07234615087509155, + "step": 2990, + "valid_targets_mean": 1225.5, + "valid_targets_min": 684 + }, + { + "epoch": 2.2968929804372844, + "grad_norm": 0.816909671666538, + "learning_rate": 3.3993995404653714e-05, + "loss": 0.1992, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09638668596744537, + "step": 2995, + "valid_targets_mean": 1282.2, + "valid_targets_min": 711 + }, + { + "epoch": 2.300728807057921, + "grad_norm": 0.7311245214352197, + "learning_rate": 3.3966648231888726e-05, + "loss": 0.1856, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10013262182474136, + "step": 3000, + "valid_targets_mean": 1634.8, + "valid_targets_min": 1190 + }, + { + "epoch": 2.3045646336785577, + "grad_norm": 0.7701845771127994, + "learning_rate": 3.393924999490624e-05, + "loss": 0.1825, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09571722149848938, + "step": 3005, + "valid_targets_mean": 1615.9, + "valid_targets_min": 774 + }, + { + "epoch": 2.3084004602991945, + "grad_norm": 0.7263217652862675, + "learning_rate": 3.3911800793878434e-05, + "loss": 0.1856, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09647343307733536, + "step": 3010, + "valid_targets_mean": 1848.4, + "valid_targets_min": 734 + }, + { + "epoch": 2.3122362869198314, + "grad_norm": 0.8140912172547758, + "learning_rate": 3.388430072916381e-05, + "loss": 0.1847, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09785354137420654, + "step": 3015, + "valid_targets_mean": 1570.6, + "valid_targets_min": 1011 + }, + { + "epoch": 2.316072113540468, + "grad_norm": 0.7557225584264688, + "learning_rate": 3.385674990130685e-05, + "loss": 0.1897, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10898664593696594, + "step": 3020, + "valid_targets_mean": 1697.0, + "valid_targets_min": 803 + }, + { + "epoch": 2.3199079401611047, + "grad_norm": 0.7895366692363555, + "learning_rate": 3.3829148411037614e-05, + "loss": 0.1883, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08174154162406921, + "step": 3025, + "valid_targets_mean": 1307.1, + "valid_targets_min": 782 + }, + { + "epoch": 2.3237437667817415, + "grad_norm": 0.7481947661156256, + "learning_rate": 3.380149635927141e-05, + "loss": 0.1828, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10013271868228912, + "step": 3030, + "valid_targets_mean": 1596.6, + "valid_targets_min": 646 + }, + { + "epoch": 2.327579593402378, + "grad_norm": 0.7891306359380644, + "learning_rate": 3.37737938471084e-05, + "loss": 0.191, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13017162680625916, + "step": 3035, + "valid_targets_mean": 1817.5, + "valid_targets_min": 922 + }, + { + "epoch": 2.331415420023015, + "grad_norm": 0.9139176047405607, + "learning_rate": 3.3746040975833226e-05, + "loss": 0.1877, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10471281409263611, + "step": 3040, + "valid_targets_mean": 1822.2, + "valid_targets_min": 796 + }, + { + "epoch": 2.3352512466436517, + "grad_norm": 0.7269894571229862, + "learning_rate": 3.371823784691467e-05, + "loss": 0.1807, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09405454248189926, + "step": 3045, + "valid_targets_mean": 1542.4, + "valid_targets_min": 903 + }, + { + "epoch": 2.3390870732642886, + "grad_norm": 0.8590026785608862, + "learning_rate": 3.369038456200525e-05, + "loss": 0.1807, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1041131317615509, + "step": 3050, + "valid_targets_mean": 1459.5, + "valid_targets_min": 780 + }, + { + "epoch": 2.3429228998849254, + "grad_norm": 0.7822664347974504, + "learning_rate": 3.366248122294087e-05, + "loss": 0.1779, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09538240730762482, + "step": 3055, + "valid_targets_mean": 1581.1, + "valid_targets_min": 783 + }, + { + "epoch": 2.346758726505562, + "grad_norm": 0.7315821057915752, + "learning_rate": 3.3634527931740435e-05, + "loss": 0.182, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07160693407058716, + "step": 3060, + "valid_targets_mean": 1342.1, + "valid_targets_min": 662 + }, + { + "epoch": 2.3505945531261987, + "grad_norm": 0.7560772669041117, + "learning_rate": 3.360652479060548e-05, + "loss": 0.1837, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08140251040458679, + "step": 3065, + "valid_targets_mean": 1269.1, + "valid_targets_min": 724 + }, + { + "epoch": 2.3544303797468356, + "grad_norm": 0.9096166571148661, + "learning_rate": 3.357847190191981e-05, + "loss": 0.1782, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09872934222221375, + "step": 3070, + "valid_targets_mean": 1510.1, + "valid_targets_min": 966 + }, + { + "epoch": 2.358266206367472, + "grad_norm": 0.8520424726176813, + "learning_rate": 3.35503693682491e-05, + "loss": 0.185, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0789322555065155, + "step": 3075, + "valid_targets_mean": 1327.8, + "valid_targets_min": 773 + }, + { + "epoch": 2.362102032988109, + "grad_norm": 0.7968212109171694, + "learning_rate": 3.352221729234054e-05, + "loss": 0.1824, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07611842453479767, + "step": 3080, + "valid_targets_mean": 1353.2, + "valid_targets_min": 807 + }, + { + "epoch": 2.3659378596087457, + "grad_norm": 0.7988261847060599, + "learning_rate": 3.3494015777122476e-05, + "loss": 0.1869, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06927424669265747, + "step": 3085, + "valid_targets_mean": 1142.5, + "valid_targets_min": 782 + }, + { + "epoch": 2.3697736862293826, + "grad_norm": 1.1447412225165903, + "learning_rate": 3.346576492570397e-05, + "loss": 0.1796, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10291219502687454, + "step": 3090, + "valid_targets_mean": 1849.9, + "valid_targets_min": 634 + }, + { + "epoch": 2.3736095128500194, + "grad_norm": 0.7816522129917189, + "learning_rate": 3.34374648413745e-05, + "loss": 0.1933, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09530937671661377, + "step": 3095, + "valid_targets_mean": 1599.6, + "valid_targets_min": 573 + }, + { + "epoch": 2.377445339470656, + "grad_norm": 0.757342241491061, + "learning_rate": 3.340911562760352e-05, + "loss": 0.1846, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10052375495433807, + "step": 3100, + "valid_targets_mean": 1994.9, + "valid_targets_min": 1102 + }, + { + "epoch": 2.3812811660912927, + "grad_norm": 0.7921306030486037, + "learning_rate": 3.338071738804014e-05, + "loss": 0.1915, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0690990537405014, + "step": 3105, + "valid_targets_mean": 1239.2, + "valid_targets_min": 841 + }, + { + "epoch": 2.3851169927119296, + "grad_norm": 0.8308384833024958, + "learning_rate": 3.3352270226512684e-05, + "loss": 0.1783, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09149043262004852, + "step": 3110, + "valid_targets_mean": 1467.5, + "valid_targets_min": 956 + }, + { + "epoch": 2.388952819332566, + "grad_norm": 2.2926323538046454, + "learning_rate": 3.332377424702837e-05, + "loss": 0.1874, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09627960622310638, + "step": 3115, + "valid_targets_mean": 1300.9, + "valid_targets_min": 857 + }, + { + "epoch": 2.392788645953203, + "grad_norm": 0.9890908827754561, + "learning_rate": 3.3295229553772874e-05, + "loss": 0.1849, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09650065004825592, + "step": 3120, + "valid_targets_mean": 1389.8, + "valid_targets_min": 808 + }, + { + "epoch": 2.3966244725738397, + "grad_norm": 0.782692152273952, + "learning_rate": 3.326663625111001e-05, + "loss": 0.1723, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07787799835205078, + "step": 3125, + "valid_targets_mean": 1303.8, + "valid_targets_min": 677 + }, + { + "epoch": 2.4004602991944766, + "grad_norm": 0.7892779378814854, + "learning_rate": 3.3237994443581274e-05, + "loss": 0.1766, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07788711786270142, + "step": 3130, + "valid_targets_mean": 1220.1, + "valid_targets_min": 727 + }, + { + "epoch": 2.404296125815113, + "grad_norm": 0.8272604020889924, + "learning_rate": 3.3209304235905545e-05, + "loss": 0.1813, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07552957534790039, + "step": 3135, + "valid_targets_mean": 1262.0, + "valid_targets_min": 726 + }, + { + "epoch": 2.40813195243575, + "grad_norm": 0.8431716809062236, + "learning_rate": 3.318056573297864e-05, + "loss": 0.1872, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10624980926513672, + "step": 3140, + "valid_targets_mean": 1615.0, + "valid_targets_min": 669 + }, + { + "epoch": 2.4119677790563867, + "grad_norm": 0.6996336027546233, + "learning_rate": 3.315177903987295e-05, + "loss": 0.1816, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10259701311588287, + "step": 3145, + "valid_targets_mean": 1806.5, + "valid_targets_min": 792 + }, + { + "epoch": 2.4158036056770236, + "grad_norm": 0.8072687127194107, + "learning_rate": 3.312294426183706e-05, + "loss": 0.1785, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09070362150669098, + "step": 3150, + "valid_targets_mean": 1627.6, + "valid_targets_min": 736 + }, + { + "epoch": 2.41963943229766, + "grad_norm": 0.944328514179292, + "learning_rate": 3.3094061504295355e-05, + "loss": 0.1844, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.128530353307724, + "step": 3155, + "valid_targets_mean": 1822.9, + "valid_targets_min": 694 + }, + { + "epoch": 2.423475258918297, + "grad_norm": 1.0155642532938178, + "learning_rate": 3.306513087284765e-05, + "loss": 0.1774, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11388645321130753, + "step": 3160, + "valid_targets_mean": 1848.9, + "valid_targets_min": 1437 + }, + { + "epoch": 2.4273110855389337, + "grad_norm": 0.8313328738038496, + "learning_rate": 3.303615247326879e-05, + "loss": 0.1783, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11169891059398651, + "step": 3165, + "valid_targets_mean": 1852.4, + "valid_targets_min": 809 + }, + { + "epoch": 2.4311469121595706, + "grad_norm": 3.0784925619130674, + "learning_rate": 3.300712641150828e-05, + "loss": 0.1767, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08253724873065948, + "step": 3170, + "valid_targets_mean": 1375.5, + "valid_targets_min": 790 + }, + { + "epoch": 2.434982738780207, + "grad_norm": 1.3167695309668417, + "learning_rate": 3.2978052793689845e-05, + "loss": 0.167, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08265050500631332, + "step": 3175, + "valid_targets_mean": 1338.8, + "valid_targets_min": 750 + }, + { + "epoch": 2.438818565400844, + "grad_norm": 0.9759521081457373, + "learning_rate": 3.2948931726111144e-05, + "loss": 0.1805, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11058226227760315, + "step": 3180, + "valid_targets_mean": 1778.8, + "valid_targets_min": 826 + }, + { + "epoch": 2.4426543920214807, + "grad_norm": 1.211659005047726, + "learning_rate": 3.291976331524326e-05, + "loss": 0.1786, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08184228837490082, + "step": 3185, + "valid_targets_mean": 1600.8, + "valid_targets_min": 907 + }, + { + "epoch": 2.4464902186421176, + "grad_norm": 0.8778106236798546, + "learning_rate": 3.289054766773041e-05, + "loss": 0.1737, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0944768562912941, + "step": 3190, + "valid_targets_mean": 1474.5, + "valid_targets_min": 1093 + }, + { + "epoch": 2.450326045262754, + "grad_norm": 1.0257262445928954, + "learning_rate": 3.28612848903895e-05, + "loss": 0.1813, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09383895248174667, + "step": 3195, + "valid_targets_mean": 1577.2, + "valid_targets_min": 1119 + }, + { + "epoch": 2.454161871883391, + "grad_norm": 0.9038167519446773, + "learning_rate": 3.2831975090209746e-05, + "loss": 0.1815, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10605060309171677, + "step": 3200, + "valid_targets_mean": 1512.1, + "valid_targets_min": 700 + }, + { + "epoch": 2.4579976985040277, + "grad_norm": 0.8435619729351531, + "learning_rate": 3.2802618374352295e-05, + "loss": 0.1796, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07273175567388535, + "step": 3205, + "valid_targets_mean": 1288.2, + "valid_targets_min": 810 + }, + { + "epoch": 2.461833525124664, + "grad_norm": 0.9085528777949627, + "learning_rate": 3.277321485014983e-05, + "loss": 0.1807, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08656062185764313, + "step": 3210, + "valid_targets_mean": 1328.9, + "valid_targets_min": 635 + }, + { + "epoch": 2.465669351745301, + "grad_norm": 0.8417758224849488, + "learning_rate": 3.274376462510616e-05, + "loss": 0.1765, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10687453299760818, + "step": 3215, + "valid_targets_mean": 1542.5, + "valid_targets_min": 673 + }, + { + "epoch": 2.469505178365938, + "grad_norm": 0.9778887485976885, + "learning_rate": 3.271426780689584e-05, + "loss": 0.1883, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0995025634765625, + "step": 3220, + "valid_targets_mean": 1215.6, + "valid_targets_min": 641 + }, + { + "epoch": 2.4733410049865747, + "grad_norm": 0.9031386339168607, + "learning_rate": 3.2684724503363784e-05, + "loss": 0.1756, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0795653834939003, + "step": 3225, + "valid_targets_mean": 1222.6, + "valid_targets_min": 701 + }, + { + "epoch": 2.4771768316072116, + "grad_norm": 0.8444590743201631, + "learning_rate": 3.265513482252487e-05, + "loss": 0.193, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09787701070308685, + "step": 3230, + "valid_targets_mean": 1850.6, + "valid_targets_min": 1202 + }, + { + "epoch": 2.481012658227848, + "grad_norm": 0.79081308734619, + "learning_rate": 3.262549887256351e-05, + "loss": 0.1817, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0710320919752121, + "step": 3235, + "valid_targets_mean": 1265.5, + "valid_targets_min": 657 + }, + { + "epoch": 2.484848484848485, + "grad_norm": 0.8146997436007908, + "learning_rate": 3.2595816761833326e-05, + "loss": 0.1844, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08464309573173523, + "step": 3240, + "valid_targets_mean": 1388.0, + "valid_targets_min": 672 + }, + { + "epoch": 2.4886843114691217, + "grad_norm": 0.8975853317242833, + "learning_rate": 3.256608859885667e-05, + "loss": 0.1777, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10940493643283844, + "step": 3245, + "valid_targets_mean": 1813.6, + "valid_targets_min": 1019 + }, + { + "epoch": 2.492520138089758, + "grad_norm": 0.8759640752632625, + "learning_rate": 3.2536314492324294e-05, + "loss": 0.1754, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07067206501960754, + "step": 3250, + "valid_targets_mean": 1114.6, + "valid_targets_min": 732 + }, + { + "epoch": 2.496355964710395, + "grad_norm": 0.8409903849329381, + "learning_rate": 3.250649455109491e-05, + "loss": 0.1823, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06327490508556366, + "step": 3255, + "valid_targets_mean": 1194.5, + "valid_targets_min": 634 + }, + { + "epoch": 2.500191791331032, + "grad_norm": 0.7096785925194998, + "learning_rate": 3.2476628884194834e-05, + "loss": 0.174, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08943314105272293, + "step": 3260, + "valid_targets_mean": 1756.5, + "valid_targets_min": 905 + }, + { + "epoch": 2.5040276179516687, + "grad_norm": 1.1197035412140595, + "learning_rate": 3.244671760081752e-05, + "loss": 0.1748, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10225990414619446, + "step": 3265, + "valid_targets_mean": 1554.9, + "valid_targets_min": 596 + }, + { + "epoch": 2.5078634445723056, + "grad_norm": 1.005174889515448, + "learning_rate": 3.2416760810323246e-05, + "loss": 0.1703, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07093896716833115, + "step": 3270, + "valid_targets_mean": 1214.6, + "valid_targets_min": 687 + }, + { + "epoch": 2.511699271192942, + "grad_norm": 0.9325769335602396, + "learning_rate": 3.238675862223865e-05, + "loss": 0.1858, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1015724390745163, + "step": 3275, + "valid_targets_mean": 1955.1, + "valid_targets_min": 1199 + }, + { + "epoch": 2.515535097813579, + "grad_norm": 0.7981708031877003, + "learning_rate": 3.2356711146256344e-05, + "loss": 0.1755, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09081659466028214, + "step": 3280, + "valid_targets_mean": 1758.6, + "valid_targets_min": 1214 + }, + { + "epoch": 2.5193709244342157, + "grad_norm": 0.7202532185868572, + "learning_rate": 3.232661849223455e-05, + "loss": 0.1722, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09128421545028687, + "step": 3285, + "valid_targets_mean": 1711.5, + "valid_targets_min": 1122 + }, + { + "epoch": 2.523206751054852, + "grad_norm": 0.9654813304640025, + "learning_rate": 3.229648077019662e-05, + "loss": 0.1687, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08783195912837982, + "step": 3290, + "valid_targets_mean": 1489.6, + "valid_targets_min": 850 + }, + { + "epoch": 2.527042577675489, + "grad_norm": 0.7336153294450161, + "learning_rate": 3.2266298090330725e-05, + "loss": 0.1612, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07636342942714691, + "step": 3295, + "valid_targets_mean": 1572.4, + "valid_targets_min": 1318 + }, + { + "epoch": 2.530878404296126, + "grad_norm": 0.8842659686582853, + "learning_rate": 3.223607056298939e-05, + "loss": 0.1802, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10130960494279861, + "step": 3300, + "valid_targets_mean": 1500.0, + "valid_targets_min": 1043 + }, + { + "epoch": 2.5347142309167623, + "grad_norm": 0.7565342528964171, + "learning_rate": 3.220579829868912e-05, + "loss": 0.179, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11019685864448547, + "step": 3305, + "valid_targets_mean": 1774.9, + "valid_targets_min": 920 + }, + { + "epoch": 2.538550057537399, + "grad_norm": 0.7516051593647939, + "learning_rate": 3.217548140810995e-05, + "loss": 0.1738, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09045585989952087, + "step": 3310, + "valid_targets_mean": 1547.1, + "valid_targets_min": 847 + }, + { + "epoch": 2.542385884158036, + "grad_norm": 0.6844226729557716, + "learning_rate": 3.2145120002095125e-05, + "loss": 0.1776, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09415974467992783, + "step": 3315, + "valid_targets_mean": 1769.4, + "valid_targets_min": 738 + }, + { + "epoch": 2.546221710778673, + "grad_norm": 0.7927590509546564, + "learning_rate": 3.211471419165063e-05, + "loss": 0.1774, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09813860058784485, + "step": 3320, + "valid_targets_mean": 1540.1, + "valid_targets_min": 594 + }, + { + "epoch": 2.5500575373993097, + "grad_norm": 0.6738844136996163, + "learning_rate": 3.208426408794476e-05, + "loss": 0.1798, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09529316425323486, + "step": 3325, + "valid_targets_mean": 2228.6, + "valid_targets_min": 1393 + }, + { + "epoch": 2.553893364019946, + "grad_norm": 0.6031719858557032, + "learning_rate": 3.2053769802307815e-05, + "loss": 0.2856, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14896118640899658, + "step": 3330, + "valid_targets_mean": 3692.9, + "valid_targets_min": 770 + }, + { + "epoch": 2.557729190640583, + "grad_norm": 0.5314663255387657, + "learning_rate": 3.202323144623159e-05, + "loss": 0.2738, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14624986052513123, + "step": 3335, + "valid_targets_mean": 5237.4, + "valid_targets_min": 1799 + }, + { + "epoch": 2.56156501726122, + "grad_norm": 0.5170971208889502, + "learning_rate": 3.199264913136902e-05, + "loss": 0.2667, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14288058876991272, + "step": 3340, + "valid_targets_mean": 5888.0, + "valid_targets_min": 3015 + }, + { + "epoch": 2.5654008438818563, + "grad_norm": 0.5141777992860577, + "learning_rate": 3.196202296953375e-05, + "loss": 0.2628, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13418129086494446, + "step": 3345, + "valid_targets_mean": 6029.1, + "valid_targets_min": 1681 + }, + { + "epoch": 2.569236670502493, + "grad_norm": 0.558185659572512, + "learning_rate": 3.1931353072699756e-05, + "loss": 0.2715, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1168038547039032, + "step": 3350, + "valid_targets_mean": 3299.6, + "valid_targets_min": 1374 + }, + { + "epoch": 2.57307249712313, + "grad_norm": 0.5702934356956282, + "learning_rate": 3.1900639553000914e-05, + "loss": 0.2755, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13735756278038025, + "step": 3355, + "valid_targets_mean": 3858.8, + "valid_targets_min": 1951 + }, + { + "epoch": 2.576908323743767, + "grad_norm": 0.49504510002313196, + "learning_rate": 3.186988252273056e-05, + "loss": 0.2595, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12802627682685852, + "step": 3360, + "valid_targets_mean": 5250.8, + "valid_targets_min": 2042 + }, + { + "epoch": 2.5807441503644037, + "grad_norm": 0.4918013897797534, + "learning_rate": 3.1839082094341156e-05, + "loss": 0.2647, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11478286236524582, + "step": 3365, + "valid_targets_mean": 4937.4, + "valid_targets_min": 1831 + }, + { + "epoch": 2.58457997698504, + "grad_norm": 0.5606821261574483, + "learning_rate": 3.180823838044381e-05, + "loss": 0.2744, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14047250151634216, + "step": 3370, + "valid_targets_mean": 4049.4, + "valid_targets_min": 2077 + }, + { + "epoch": 2.588415803605677, + "grad_norm": 0.5233843033964137, + "learning_rate": 3.177735149380787e-05, + "loss": 0.2696, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11958551406860352, + "step": 3375, + "valid_targets_mean": 4086.2, + "valid_targets_min": 2022 + }, + { + "epoch": 2.592251630226314, + "grad_norm": 0.5656257681120709, + "learning_rate": 3.174642154736058e-05, + "loss": 0.2737, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14817357063293457, + "step": 3380, + "valid_targets_mean": 4897.6, + "valid_targets_min": 1499 + }, + { + "epoch": 2.5960874568469503, + "grad_norm": 0.48842552363875585, + "learning_rate": 3.171544865418658e-05, + "loss": 0.2604, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13842380046844482, + "step": 3385, + "valid_targets_mean": 4821.6, + "valid_targets_min": 1013 + }, + { + "epoch": 2.599923283467587, + "grad_norm": 0.5753130890441258, + "learning_rate": 3.168443292752753e-05, + "loss": 0.2682, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1317327618598938, + "step": 3390, + "valid_targets_mean": 4676.2, + "valid_targets_min": 1939 + }, + { + "epoch": 2.603759110088224, + "grad_norm": 0.5797872029854289, + "learning_rate": 3.165337448078171e-05, + "loss": 0.2641, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15582017600536346, + "step": 3395, + "valid_targets_mean": 4010.2, + "valid_targets_min": 1820 + }, + { + "epoch": 2.607594936708861, + "grad_norm": 0.6091811929052324, + "learning_rate": 3.162227342750359e-05, + "loss": 0.2731, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15132489800453186, + "step": 3400, + "valid_targets_mean": 3824.9, + "valid_targets_min": 1589 + }, + { + "epoch": 2.6114307633294978, + "grad_norm": 0.5729071924098447, + "learning_rate": 3.15911298814034e-05, + "loss": 0.2654, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12214220315217972, + "step": 3405, + "valid_targets_mean": 4046.5, + "valid_targets_min": 2668 + }, + { + "epoch": 2.615266589950134, + "grad_norm": 0.5742947616792958, + "learning_rate": 3.1559943956346764e-05, + "loss": 0.2665, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1295974850654602, + "step": 3410, + "valid_targets_mean": 3592.6, + "valid_targets_min": 1269 + }, + { + "epoch": 2.619102416570771, + "grad_norm": 0.6485523421437455, + "learning_rate": 3.15287157663542e-05, + "loss": 0.2798, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13355399668216705, + "step": 3415, + "valid_targets_mean": 3423.4, + "valid_targets_min": 1695 + }, + { + "epoch": 2.622938243191408, + "grad_norm": 0.5854279606808739, + "learning_rate": 3.1497445425600795e-05, + "loss": 0.273, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12540721893310547, + "step": 3420, + "valid_targets_mean": 3522.2, + "valid_targets_min": 1128 + }, + { + "epoch": 2.6267740698120443, + "grad_norm": 0.6059799971557155, + "learning_rate": 3.146613304841575e-05, + "loss": 0.2643, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13703130185604095, + "step": 3425, + "valid_targets_mean": 3533.5, + "valid_targets_min": 1438 + }, + { + "epoch": 2.630609896432681, + "grad_norm": 0.5846617983861214, + "learning_rate": 3.143477874928191e-05, + "loss": 0.2713, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12760914862155914, + "step": 3430, + "valid_targets_mean": 3842.9, + "valid_targets_min": 1267 + }, + { + "epoch": 2.634445723053318, + "grad_norm": 0.5742983887749702, + "learning_rate": 3.140338264283544e-05, + "loss": 0.2606, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12054377794265747, + "step": 3435, + "valid_targets_mean": 3671.9, + "valid_targets_min": 1026 + }, + { + "epoch": 2.638281549673955, + "grad_norm": 0.6868647058278101, + "learning_rate": 3.137194484386534e-05, + "loss": 0.2482, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11949074268341064, + "step": 3440, + "valid_targets_mean": 2866.2, + "valid_targets_min": 1403 + }, + { + "epoch": 2.6421173762945918, + "grad_norm": 0.6154442380653024, + "learning_rate": 3.134046546731304e-05, + "loss": 0.2632, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14729031920433044, + "step": 3445, + "valid_targets_mean": 3882.1, + "valid_targets_min": 1588 + }, + { + "epoch": 2.645953202915228, + "grad_norm": 0.6248462455823339, + "learning_rate": 3.1308944628271986e-05, + "loss": 0.2691, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1184033453464508, + "step": 3450, + "valid_targets_mean": 2827.6, + "valid_targets_min": 1276 + }, + { + "epoch": 2.649789029535865, + "grad_norm": 0.5051292753811327, + "learning_rate": 3.127738244198724e-05, + "loss": 0.2598, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11317639797925949, + "step": 3455, + "valid_targets_mean": 4232.4, + "valid_targets_min": 1631 + }, + { + "epoch": 2.653624856156502, + "grad_norm": 0.548753254964154, + "learning_rate": 3.1245779023854996e-05, + "loss": 0.2611, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1588626503944397, + "step": 3460, + "valid_targets_mean": 5069.9, + "valid_targets_min": 3090 + }, + { + "epoch": 2.6574606827771383, + "grad_norm": 0.607609561698509, + "learning_rate": 3.121413448942223e-05, + "loss": 0.2526, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15929926931858063, + "step": 3465, + "valid_targets_mean": 3732.4, + "valid_targets_min": 1802 + }, + { + "epoch": 2.661296509397775, + "grad_norm": 0.6996389026831162, + "learning_rate": 3.1182448954386234e-05, + "loss": 0.2694, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15847238898277283, + "step": 3470, + "valid_targets_mean": 3490.8, + "valid_targets_min": 1685 + }, + { + "epoch": 2.665132336018412, + "grad_norm": 0.8015644349774539, + "learning_rate": 3.115072253459421e-05, + "loss": 0.2516, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11799109727144241, + "step": 3475, + "valid_targets_mean": 3660.1, + "valid_targets_min": 1596 + }, + { + "epoch": 2.6689681626390485, + "grad_norm": 0.6266413112617583, + "learning_rate": 3.111895534604284e-05, + "loss": 0.2573, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1455337405204773, + "step": 3480, + "valid_targets_mean": 4306.4, + "valid_targets_min": 2251 + }, + { + "epoch": 2.6728039892596853, + "grad_norm": 0.7278995276830391, + "learning_rate": 3.1087147504877835e-05, + "loss": 0.2565, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13282249867916107, + "step": 3485, + "valid_targets_mean": 3107.1, + "valid_targets_min": 1458 + }, + { + "epoch": 2.676639815880322, + "grad_norm": 0.6477222620757672, + "learning_rate": 3.105529912739358e-05, + "loss": 0.2632, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12455359101295471, + "step": 3490, + "valid_targets_mean": 3664.6, + "valid_targets_min": 1162 + }, + { + "epoch": 2.680475642500959, + "grad_norm": 0.6529257529553747, + "learning_rate": 3.1023410330032645e-05, + "loss": 0.256, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12170784920454025, + "step": 3495, + "valid_targets_mean": 2834.2, + "valid_targets_min": 1480 + }, + { + "epoch": 2.684311469121596, + "grad_norm": 0.6518745039554584, + "learning_rate": 3.099148122938539e-05, + "loss": 0.266, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13689857721328735, + "step": 3500, + "valid_targets_mean": 3795.1, + "valid_targets_min": 1229 + }, + { + "epoch": 2.6881472957422323, + "grad_norm": 0.5847567917738533, + "learning_rate": 3.095951194218953e-05, + "loss": 0.2637, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11791917681694031, + "step": 3505, + "valid_targets_mean": 3559.6, + "valid_targets_min": 1276 + }, + { + "epoch": 2.691983122362869, + "grad_norm": 0.6281391534230302, + "learning_rate": 3.092750258532968e-05, + "loss": 0.2583, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.116336390376091, + "step": 3510, + "valid_targets_mean": 3479.6, + "valid_targets_min": 1357 + }, + { + "epoch": 2.695818948983506, + "grad_norm": 0.6275251643620454, + "learning_rate": 3.089545327583701e-05, + "loss": 0.2543, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13213187456130981, + "step": 3515, + "valid_targets_mean": 2925.9, + "valid_targets_min": 924 + }, + { + "epoch": 2.6996547756041425, + "grad_norm": 0.6065684237112581, + "learning_rate": 3.086336413088872e-05, + "loss": 0.2628, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1364710032939911, + "step": 3520, + "valid_targets_mean": 4030.5, + "valid_targets_min": 1674 + }, + { + "epoch": 2.7034906022247793, + "grad_norm": 0.6320308800334179, + "learning_rate": 3.083123526780767e-05, + "loss": 0.2582, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12773147225379944, + "step": 3525, + "valid_targets_mean": 3173.9, + "valid_targets_min": 1643 + }, + { + "epoch": 2.707326428845416, + "grad_norm": 0.5440338429336724, + "learning_rate": 3.079906680406192e-05, + "loss": 0.2513, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1373293548822403, + "step": 3530, + "valid_targets_mean": 4259.1, + "valid_targets_min": 1244 + }, + { + "epoch": 2.711162255466053, + "grad_norm": 0.5733123610526826, + "learning_rate": 3.0766858857264336e-05, + "loss": 0.2681, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12298949807882309, + "step": 3535, + "valid_targets_mean": 4065.8, + "valid_targets_min": 730 + }, + { + "epoch": 2.71499808208669, + "grad_norm": 0.6906773070374556, + "learning_rate": 3.073461154517214e-05, + "loss": 0.2556, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11435170471668243, + "step": 3540, + "valid_targets_mean": 2945.9, + "valid_targets_min": 977 + }, + { + "epoch": 2.7188339087073263, + "grad_norm": 0.5909088511228201, + "learning_rate": 3.070232498568647e-05, + "loss": 0.2532, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14356869459152222, + "step": 3545, + "valid_targets_mean": 5116.1, + "valid_targets_min": 2091 + }, + { + "epoch": 2.722669735327963, + "grad_norm": 0.5960714109733091, + "learning_rate": 3.066999929685196e-05, + "loss": 0.2453, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11380893737077713, + "step": 3550, + "valid_targets_mean": 3125.4, + "valid_targets_min": 1654 + }, + { + "epoch": 2.7265055619486, + "grad_norm": 0.5918060303967756, + "learning_rate": 3.063763459685631e-05, + "loss": 0.2655, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13227030634880066, + "step": 3555, + "valid_targets_mean": 4081.6, + "valid_targets_min": 1432 + }, + { + "epoch": 2.7303413885692365, + "grad_norm": 0.555973431647015, + "learning_rate": 3.060523100402984e-05, + "loss": 0.2475, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11935532093048096, + "step": 3560, + "valid_targets_mean": 3868.6, + "valid_targets_min": 1261 + }, + { + "epoch": 2.7341772151898733, + "grad_norm": 0.651977358846617, + "learning_rate": 3.057278863684509e-05, + "loss": 0.2681, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15310674905776978, + "step": 3565, + "valid_targets_mean": 4079.6, + "valid_targets_min": 2737 + }, + { + "epoch": 2.73801304181051, + "grad_norm": 0.5899735495826726, + "learning_rate": 3.054030761391634e-05, + "loss": 0.2554, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10847122967243195, + "step": 3570, + "valid_targets_mean": 3471.4, + "valid_targets_min": 1219 + }, + { + "epoch": 2.741848868431147, + "grad_norm": 0.6174467186748583, + "learning_rate": 3.0507788053999225e-05, + "loss": 0.269, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14045050740242004, + "step": 3575, + "valid_targets_mean": 4087.4, + "valid_targets_min": 2058 + }, + { + "epoch": 2.745684695051784, + "grad_norm": 0.9520090920433673, + "learning_rate": 3.0475230075990258e-05, + "loss": 0.2483, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11757412552833557, + "step": 3580, + "valid_targets_mean": 2736.6, + "valid_targets_min": 916 + }, + { + "epoch": 2.7495205216724203, + "grad_norm": 0.5853572273866698, + "learning_rate": 3.0442633798926418e-05, + "loss": 0.2677, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11146454513072968, + "step": 3585, + "valid_targets_mean": 3695.1, + "valid_targets_min": 1302 + }, + { + "epoch": 2.753356348293057, + "grad_norm": 0.6539311949497894, + "learning_rate": 3.0409999341984734e-05, + "loss": 0.2503, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16135677695274353, + "step": 3590, + "valid_targets_mean": 4246.1, + "valid_targets_min": 1500 + }, + { + "epoch": 2.757192174913694, + "grad_norm": 0.5484668005774171, + "learning_rate": 3.037732682448179e-05, + "loss": 0.2456, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14510150253772736, + "step": 3595, + "valid_targets_mean": 4837.6, + "valid_targets_min": 2803 + }, + { + "epoch": 2.7610280015343305, + "grad_norm": 0.5792990348690134, + "learning_rate": 3.0344616365873345e-05, + "loss": 0.2386, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12269844859838486, + "step": 3600, + "valid_targets_mean": 3349.0, + "valid_targets_min": 878 + }, + { + "epoch": 2.7648638281549673, + "grad_norm": 0.5759103209120651, + "learning_rate": 3.0311868085753883e-05, + "loss": 0.2645, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11581222712993622, + "step": 3605, + "valid_targets_mean": 3753.9, + "valid_targets_min": 2662 + }, + { + "epoch": 2.768699654775604, + "grad_norm": 0.5507487446535961, + "learning_rate": 3.027908210385616e-05, + "loss": 0.2543, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12100072205066681, + "step": 3610, + "valid_targets_mean": 3965.1, + "valid_targets_min": 1086 + }, + { + "epoch": 2.7725354813962406, + "grad_norm": 0.5709938318251102, + "learning_rate": 3.0246258540050762e-05, + "loss": 0.2611, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14102187752723694, + "step": 3615, + "valid_targets_mean": 5292.5, + "valid_targets_min": 1566 + }, + { + "epoch": 2.7763713080168775, + "grad_norm": 0.6050130305426726, + "learning_rate": 3.021339751434571e-05, + "loss": 0.2547, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14905810356140137, + "step": 3620, + "valid_targets_mean": 4169.2, + "valid_targets_min": 2504 + }, + { + "epoch": 2.7802071346375143, + "grad_norm": 0.6127751903951855, + "learning_rate": 3.0180499146885976e-05, + "loss": 0.265, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1587723195552826, + "step": 3625, + "valid_targets_mean": 4600.9, + "valid_targets_min": 1548 + }, + { + "epoch": 2.784042961258151, + "grad_norm": 0.612473792434459, + "learning_rate": 3.014756355795306e-05, + "loss": 0.2588, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12079291045665741, + "step": 3630, + "valid_targets_mean": 3289.5, + "valid_targets_min": 1352 + }, + { + "epoch": 2.787878787878788, + "grad_norm": 0.5376756320165984, + "learning_rate": 3.011459086796453e-05, + "loss": 0.2523, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13263943791389465, + "step": 3635, + "valid_targets_mean": 5279.6, + "valid_targets_min": 2339 + }, + { + "epoch": 2.7917146144994245, + "grad_norm": 0.6272572999211624, + "learning_rate": 3.008158119747364e-05, + "loss": 0.2489, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12514540553092957, + "step": 3640, + "valid_targets_mean": 4053.0, + "valid_targets_min": 2015 + }, + { + "epoch": 2.7955504411200613, + "grad_norm": 0.6143509227839162, + "learning_rate": 3.004853466716883e-05, + "loss": 0.2469, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14922448992729187, + "step": 3645, + "valid_targets_mean": 4836.5, + "valid_targets_min": 1731 + }, + { + "epoch": 2.799386267740698, + "grad_norm": 0.6512024102418059, + "learning_rate": 3.0015451397873307e-05, + "loss": 0.25, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12092310190200806, + "step": 3650, + "valid_targets_mean": 3052.5, + "valid_targets_min": 1460 + }, + { + "epoch": 2.8032220943613346, + "grad_norm": 0.5636577532719578, + "learning_rate": 2.998233151054458e-05, + "loss": 0.2393, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13410170376300812, + "step": 3655, + "valid_targets_mean": 4621.9, + "valid_targets_min": 1282 + }, + { + "epoch": 2.8070579209819715, + "grad_norm": 0.586760776635638, + "learning_rate": 2.9949175126274075e-05, + "loss": 0.2382, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10557278990745544, + "step": 3660, + "valid_targets_mean": 3373.4, + "valid_targets_min": 1421 + }, + { + "epoch": 2.8108937476026084, + "grad_norm": 0.6033954167659192, + "learning_rate": 2.9915982366286642e-05, + "loss": 0.2447, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1518811285495758, + "step": 3665, + "valid_targets_mean": 4483.5, + "valid_targets_min": 1764 + }, + { + "epoch": 2.814729574223245, + "grad_norm": 0.7263926922243463, + "learning_rate": 2.9882753351940115e-05, + "loss": 0.264, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1531134992837906, + "step": 3670, + "valid_targets_mean": 3288.8, + "valid_targets_min": 1175 + }, + { + "epoch": 2.818565400843882, + "grad_norm": 0.5219157055448373, + "learning_rate": 2.9849488204724902e-05, + "loss": 0.2566, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.114308200776577, + "step": 3675, + "valid_targets_mean": 4656.6, + "valid_targets_min": 2546 + }, + { + "epoch": 2.8224012274645185, + "grad_norm": 0.6969014291276663, + "learning_rate": 2.9816187046263495e-05, + "loss": 0.2559, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15941005945205688, + "step": 3680, + "valid_targets_mean": 3833.2, + "valid_targets_min": 1244 + }, + { + "epoch": 2.8262370540851554, + "grad_norm": 0.5731304700369895, + "learning_rate": 2.9782849998310067e-05, + "loss": 0.2523, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12771131098270416, + "step": 3685, + "valid_targets_mean": 4403.5, + "valid_targets_min": 1856 + }, + { + "epoch": 2.830072880705792, + "grad_norm": 0.5976653901032873, + "learning_rate": 2.974947718275e-05, + "loss": 0.2512, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1466556042432785, + "step": 3690, + "valid_targets_mean": 3836.6, + "valid_targets_min": 1948 + }, + { + "epoch": 2.8339087073264286, + "grad_norm": 0.5317279041355414, + "learning_rate": 2.971606872159944e-05, + "loss": 0.2443, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10621809959411621, + "step": 3695, + "valid_targets_mean": 3336.2, + "valid_targets_min": 1248 + }, + { + "epoch": 2.8377445339470655, + "grad_norm": 0.5168858637435392, + "learning_rate": 2.9682624737004877e-05, + "loss": 0.2478, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1172504648566246, + "step": 3700, + "valid_targets_mean": 4304.4, + "valid_targets_min": 1145 + }, + { + "epoch": 2.8415803605677024, + "grad_norm": 0.6034306088905654, + "learning_rate": 2.9649145351242663e-05, + "loss": 0.2504, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11782073229551315, + "step": 3705, + "valid_targets_mean": 3231.1, + "valid_targets_min": 1428 + }, + { + "epoch": 2.845416187188339, + "grad_norm": 0.5914462100289156, + "learning_rate": 2.96156306867186e-05, + "loss": 0.2491, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1437692642211914, + "step": 3710, + "valid_targets_mean": 4426.2, + "valid_targets_min": 1708 + }, + { + "epoch": 2.849252013808976, + "grad_norm": 0.5952132350077493, + "learning_rate": 2.958208086596746e-05, + "loss": 0.2453, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08684082329273224, + "step": 3715, + "valid_targets_mean": 2934.0, + "valid_targets_min": 1301 + }, + { + "epoch": 2.8530878404296125, + "grad_norm": 0.6593437812328166, + "learning_rate": 2.954849601165255e-05, + "loss": 0.2676, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11009576916694641, + "step": 3720, + "valid_targets_mean": 3015.5, + "valid_targets_min": 1971 + }, + { + "epoch": 2.8569236670502494, + "grad_norm": 0.5883558583665127, + "learning_rate": 2.951487624656528e-05, + "loss": 0.2621, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15695665776729584, + "step": 3725, + "valid_targets_mean": 4887.2, + "valid_targets_min": 1054 + }, + { + "epoch": 2.8607594936708862, + "grad_norm": 0.6032632916416757, + "learning_rate": 2.948122169362468e-05, + "loss": 0.2597, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14216041564941406, + "step": 3730, + "valid_targets_mean": 3527.2, + "valid_targets_min": 1612 + }, + { + "epoch": 2.8645953202915226, + "grad_norm": 0.5855590555433633, + "learning_rate": 2.944753247587699e-05, + "loss": 0.2666, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10704144835472107, + "step": 3735, + "valid_targets_mean": 3531.1, + "valid_targets_min": 721 + }, + { + "epoch": 2.8684311469121595, + "grad_norm": 0.5785531724483994, + "learning_rate": 2.9413808716495173e-05, + "loss": 0.2574, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1297263652086258, + "step": 3740, + "valid_targets_mean": 3899.6, + "valid_targets_min": 1625 + }, + { + "epoch": 2.8722669735327964, + "grad_norm": 0.5598280722928962, + "learning_rate": 2.9380050538778498e-05, + "loss": 0.2674, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12826499342918396, + "step": 3745, + "valid_targets_mean": 4175.0, + "valid_targets_min": 1381 + }, + { + "epoch": 2.8761028001534332, + "grad_norm": 0.6461654923691023, + "learning_rate": 2.9346258066152056e-05, + "loss": 0.2691, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12764230370521545, + "step": 3750, + "valid_targets_mean": 3235.0, + "valid_targets_min": 998 + }, + { + "epoch": 2.87993862677407, + "grad_norm": 0.6495825564042996, + "learning_rate": 2.931243142216633e-05, + "loss": 0.2588, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1035550981760025, + "step": 3755, + "valid_targets_mean": 2957.1, + "valid_targets_min": 1186 + }, + { + "epoch": 2.8837744533947065, + "grad_norm": 0.5982379164968065, + "learning_rate": 2.927857073049674e-05, + "loss": 0.2583, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15188246965408325, + "step": 3760, + "valid_targets_mean": 3583.0, + "valid_targets_min": 1195 + }, + { + "epoch": 2.8876102800153434, + "grad_norm": 0.6631341548507479, + "learning_rate": 2.9244676114943203e-05, + "loss": 0.2601, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1301260143518448, + "step": 3765, + "valid_targets_mean": 2696.1, + "valid_targets_min": 885 + }, + { + "epoch": 2.8914461066359802, + "grad_norm": 0.6244466112931015, + "learning_rate": 2.921074769942965e-05, + "loss": 0.2642, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12420439720153809, + "step": 3770, + "valid_targets_mean": 3362.2, + "valid_targets_min": 814 + }, + { + "epoch": 2.8952819332566166, + "grad_norm": 0.5826567118281526, + "learning_rate": 2.9176785608003584e-05, + "loss": 0.2603, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11268092691898346, + "step": 3775, + "valid_targets_mean": 3612.5, + "valid_targets_min": 2174 + }, + { + "epoch": 2.8991177598772535, + "grad_norm": 0.5372817574070816, + "learning_rate": 2.9142789964835655e-05, + "loss": 0.2392, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14259761571884155, + "step": 3780, + "valid_targets_mean": 5472.4, + "valid_targets_min": 3025 + }, + { + "epoch": 2.9029535864978904, + "grad_norm": 0.67158743265752, + "learning_rate": 2.9108760894219173e-05, + "loss": 0.2548, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12312570214271545, + "step": 3785, + "valid_targets_mean": 3297.9, + "valid_targets_min": 1434 + }, + { + "epoch": 2.906789413118527, + "grad_norm": 0.6424086985499734, + "learning_rate": 2.9074698520569654e-05, + "loss": 0.247, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12526191771030426, + "step": 3790, + "valid_targets_mean": 3262.1, + "valid_targets_min": 1074 + }, + { + "epoch": 2.9106252397391636, + "grad_norm": 0.5956912216460154, + "learning_rate": 2.9040602968424386e-05, + "loss": 0.2588, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10052645206451416, + "step": 3795, + "valid_targets_mean": 3145.2, + "valid_targets_min": 980 + }, + { + "epoch": 2.9144610663598005, + "grad_norm": 0.5750259563536442, + "learning_rate": 2.9006474362441962e-05, + "loss": 0.2543, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14225074648857117, + "step": 3800, + "valid_targets_mean": 4575.0, + "valid_targets_min": 2366 + }, + { + "epoch": 2.9182968929804374, + "grad_norm": 0.6049854287747998, + "learning_rate": 2.8972312827401824e-05, + "loss": 0.2595, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13191381096839905, + "step": 3805, + "valid_targets_mean": 4999.9, + "valid_targets_min": 3139 + }, + { + "epoch": 2.9221327196010742, + "grad_norm": 0.6553843610850724, + "learning_rate": 2.89381184882038e-05, + "loss": 0.2533, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15510523319244385, + "step": 3810, + "valid_targets_mean": 4034.4, + "valid_targets_min": 2138 + }, + { + "epoch": 2.9259685462217107, + "grad_norm": 0.602769383370285, + "learning_rate": 2.890389146986767e-05, + "loss": 0.2534, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12889744341373444, + "step": 3815, + "valid_targets_mean": 3610.9, + "valid_targets_min": 2014 + }, + { + "epoch": 2.9298043728423475, + "grad_norm": 0.6167123853458945, + "learning_rate": 2.8869631897532676e-05, + "loss": 0.2569, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14353178441524506, + "step": 3820, + "valid_targets_mean": 3795.6, + "valid_targets_min": 1084 + }, + { + "epoch": 2.9336401994629844, + "grad_norm": 0.5755839958662072, + "learning_rate": 2.8835339896457104e-05, + "loss": 0.2453, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1287807822227478, + "step": 3825, + "valid_targets_mean": 3684.4, + "valid_targets_min": 1618 + }, + { + "epoch": 2.937476026083621, + "grad_norm": 0.647392686751687, + "learning_rate": 2.8801015592017787e-05, + "loss": 0.2408, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11894913762807846, + "step": 3830, + "valid_targets_mean": 3378.8, + "valid_targets_min": 1193 + }, + { + "epoch": 2.9413118527042577, + "grad_norm": 0.5637992662734296, + "learning_rate": 2.8766659109709664e-05, + "loss": 0.2488, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13283199071884155, + "step": 3835, + "valid_targets_mean": 4732.5, + "valid_targets_min": 2092 + }, + { + "epoch": 2.9451476793248945, + "grad_norm": 0.5683235157674276, + "learning_rate": 2.8732270575145336e-05, + "loss": 0.2475, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14965961873531342, + "step": 3840, + "valid_targets_mean": 4235.9, + "valid_targets_min": 2786 + }, + { + "epoch": 2.9489835059455314, + "grad_norm": 0.5687114767579186, + "learning_rate": 2.8697850114054584e-05, + "loss": 0.2376, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10389884561300278, + "step": 3845, + "valid_targets_mean": 4036.9, + "valid_targets_min": 1246 + }, + { + "epoch": 2.9528193325661682, + "grad_norm": 0.5722749499578341, + "learning_rate": 2.8663397852283904e-05, + "loss": 0.2406, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10533909499645233, + "step": 3850, + "valid_targets_mean": 3434.8, + "valid_targets_min": 1484 + }, + { + "epoch": 2.9566551591868047, + "grad_norm": 0.5774355995681378, + "learning_rate": 2.8628913915796083e-05, + "loss": 0.2509, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1330336332321167, + "step": 3855, + "valid_targets_mean": 4156.9, + "valid_targets_min": 1502 + }, + { + "epoch": 2.9604909858074415, + "grad_norm": 0.5596539287256942, + "learning_rate": 2.8594398430669697e-05, + "loss": 0.2446, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09604993462562561, + "step": 3860, + "valid_targets_mean": 3785.8, + "valid_targets_min": 1530 + }, + { + "epoch": 2.9643268124280784, + "grad_norm": 0.8015694984530333, + "learning_rate": 2.8559851523098695e-05, + "loss": 0.2744, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13911165297031403, + "step": 3865, + "valid_targets_mean": 3245.6, + "valid_targets_min": 1312 + }, + { + "epoch": 2.968162639048715, + "grad_norm": 0.5503913995236867, + "learning_rate": 2.8525273319391868e-05, + "loss": 0.2439, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13613417744636536, + "step": 3870, + "valid_targets_mean": 5026.8, + "valid_targets_min": 2738 + }, + { + "epoch": 2.9719984656693517, + "grad_norm": 0.6174027267472711, + "learning_rate": 2.8490663945972467e-05, + "loss": 0.2447, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12955054640769958, + "step": 3875, + "valid_targets_mean": 4030.1, + "valid_targets_min": 1759 + }, + { + "epoch": 2.9758342922899885, + "grad_norm": 0.59670481268521, + "learning_rate": 2.84560235293777e-05, + "loss": 0.2435, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10838741064071655, + "step": 3880, + "valid_targets_mean": 3422.8, + "valid_targets_min": 2155 + }, + { + "epoch": 2.9796701189106254, + "grad_norm": 0.7022222319204282, + "learning_rate": 2.842135219625826e-05, + "loss": 0.2524, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12163539975881577, + "step": 3885, + "valid_targets_mean": 3223.8, + "valid_targets_min": 2048 + }, + { + "epoch": 2.9835059455312622, + "grad_norm": 0.6363234951097119, + "learning_rate": 2.838665007337788e-05, + "loss": 0.2517, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10442635416984558, + "step": 3890, + "valid_targets_mean": 2658.6, + "valid_targets_min": 1233 + }, + { + "epoch": 2.9873417721518987, + "grad_norm": 0.5566194128463843, + "learning_rate": 2.8351917287612858e-05, + "loss": 0.2561, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11471602320671082, + "step": 3895, + "valid_targets_mean": 3698.2, + "valid_targets_min": 1383 + }, + { + "epoch": 2.9911775987725355, + "grad_norm": 0.5747453446332396, + "learning_rate": 2.8317153965951628e-05, + "loss": 0.2525, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11890333145856857, + "step": 3900, + "valid_targets_mean": 3766.5, + "valid_targets_min": 1800 + }, + { + "epoch": 2.9950134253931724, + "grad_norm": 0.6117954908348223, + "learning_rate": 2.8282360235494244e-05, + "loss": 0.2528, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1160469502210617, + "step": 3905, + "valid_targets_mean": 3114.6, + "valid_targets_min": 2327 + }, + { + "epoch": 2.998849252013809, + "grad_norm": 0.6291441654444433, + "learning_rate": 2.8247536223451935e-05, + "loss": 0.2536, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1273503601551056, + "step": 3910, + "valid_targets_mean": 3558.5, + "valid_targets_min": 1214 + }, + { + "epoch": 3.002301495972382, + "grad_norm": 0.5033730409057148, + "learning_rate": 2.8212682057146666e-05, + "loss": 0.2096, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1007581353187561, + "step": 3915, + "valid_targets_mean": 5435.1, + "valid_targets_min": 3516 + }, + { + "epoch": 3.006137322593019, + "grad_norm": 0.39524631172938907, + "learning_rate": 2.817779786401063e-05, + "loss": 0.2067, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09965745359659195, + "step": 3920, + "valid_targets_mean": 7654.2, + "valid_targets_min": 6354 + }, + { + "epoch": 3.0099731492136557, + "grad_norm": 0.4185196082331958, + "learning_rate": 2.814288377158582e-05, + "loss": 0.1999, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11382512748241425, + "step": 3925, + "valid_targets_mean": 7341.6, + "valid_targets_min": 5754 + }, + { + "epoch": 3.013808975834292, + "grad_norm": 0.37667913265237973, + "learning_rate": 2.810793990752352e-05, + "loss": 0.1939, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08618589490652084, + "step": 3930, + "valid_targets_mean": 7666.2, + "valid_targets_min": 5666 + }, + { + "epoch": 3.017644802454929, + "grad_norm": 0.40342105077450796, + "learning_rate": 2.8072966399583897e-05, + "loss": 0.1951, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10084332525730133, + "step": 3935, + "valid_targets_mean": 7469.0, + "valid_targets_min": 5646 + }, + { + "epoch": 3.021480629075566, + "grad_norm": 0.4408728070261714, + "learning_rate": 2.8037963375635473e-05, + "loss": 0.1958, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1152171865105629, + "step": 3940, + "valid_targets_mean": 7046.9, + "valid_targets_min": 5571 + }, + { + "epoch": 3.0253164556962027, + "grad_norm": 0.4480434476388113, + "learning_rate": 2.8002930963654705e-05, + "loss": 0.2118, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11240038275718689, + "step": 3945, + "valid_targets_mean": 7827.6, + "valid_targets_min": 5405 + }, + { + "epoch": 3.029152282316839, + "grad_norm": 0.4399829340638329, + "learning_rate": 2.7967869291725474e-05, + "loss": 0.1976, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10300078243017197, + "step": 3950, + "valid_targets_mean": 7023.5, + "valid_targets_min": 5664 + }, + { + "epoch": 3.032988108937476, + "grad_norm": 0.41716836866774326, + "learning_rate": 2.793277848803867e-05, + "loss": 0.1916, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09341755509376526, + "step": 3955, + "valid_targets_mean": 7069.8, + "valid_targets_min": 5103 + }, + { + "epoch": 3.036823935558113, + "grad_norm": 0.41774296786985154, + "learning_rate": 2.789765868089167e-05, + "loss": 0.1804, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09248331189155579, + "step": 3960, + "valid_targets_mean": 6712.4, + "valid_targets_min": 5469 + }, + { + "epoch": 3.0406597621787497, + "grad_norm": 0.38486070021062263, + "learning_rate": 2.7862509998687895e-05, + "loss": 0.1939, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09309627115726471, + "step": 3965, + "valid_targets_mean": 7289.2, + "valid_targets_min": 5513 + }, + { + "epoch": 3.044495588799386, + "grad_norm": 0.39986359008750244, + "learning_rate": 2.7827332569936345e-05, + "loss": 0.1847, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09149394929409027, + "step": 3970, + "valid_targets_mean": 7022.1, + "valid_targets_min": 5691 + }, + { + "epoch": 3.048331415420023, + "grad_norm": 0.38791047786184435, + "learning_rate": 2.7792126523251122e-05, + "loss": 0.1951, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10011675208806992, + "step": 3975, + "valid_targets_mean": 6736.0, + "valid_targets_min": 5550 + }, + { + "epoch": 3.05216724204066, + "grad_norm": 0.40753487493107066, + "learning_rate": 2.7756891987350945e-05, + "loss": 0.1819, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08908520638942719, + "step": 3980, + "valid_targets_mean": 6346.9, + "valid_targets_min": 5067 + }, + { + "epoch": 3.0560030686612967, + "grad_norm": 0.39874207583863774, + "learning_rate": 2.7721629091058724e-05, + "loss": 0.1859, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09568531066179276, + "step": 3985, + "valid_targets_mean": 6784.0, + "valid_targets_min": 5618 + }, + { + "epoch": 3.059838895281933, + "grad_norm": 0.4071520023843487, + "learning_rate": 2.7686337963301027e-05, + "loss": 0.1934, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1086595207452774, + "step": 3990, + "valid_targets_mean": 7568.4, + "valid_targets_min": 5856 + }, + { + "epoch": 3.06367472190257, + "grad_norm": 0.41006881114112465, + "learning_rate": 2.765101873310765e-05, + "loss": 0.1847, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08992061018943787, + "step": 3995, + "valid_targets_mean": 7198.5, + "valid_targets_min": 4793 + }, + { + "epoch": 3.067510548523207, + "grad_norm": 0.405630634991384, + "learning_rate": 2.761567152961115e-05, + "loss": 0.1901, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09625566005706787, + "step": 4000, + "valid_targets_mean": 6680.5, + "valid_targets_min": 5350 + }, + { + "epoch": 3.0713463751438437, + "grad_norm": 0.4458039987144032, + "learning_rate": 2.758029648204635e-05, + "loss": 0.2044, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10076257586479187, + "step": 4005, + "valid_targets_mean": 6770.6, + "valid_targets_min": 5753 + }, + { + "epoch": 3.07518220176448, + "grad_norm": 0.8394528489023428, + "learning_rate": 2.754489371974986e-05, + "loss": 0.2014, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07304803282022476, + "step": 4010, + "valid_targets_mean": 1419.8, + "valid_targets_min": 289 + }, + { + "epoch": 3.079018028385117, + "grad_norm": 0.4498930826579513, + "learning_rate": 2.7509463372159636e-05, + "loss": 0.1687, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09865590929985046, + "step": 4015, + "valid_targets_mean": 6784.2, + "valid_targets_min": 5042 + }, + { + "epoch": 3.082853855005754, + "grad_norm": 0.39350147663000284, + "learning_rate": 2.7474005568814495e-05, + "loss": 0.184, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09137275815010071, + "step": 4020, + "valid_targets_mean": 6774.9, + "valid_targets_min": 3875 + }, + { + "epoch": 3.0866896816263907, + "grad_norm": 0.41553174242237856, + "learning_rate": 2.743852043935363e-05, + "loss": 0.1874, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09747269749641418, + "step": 4025, + "valid_targets_mean": 7047.6, + "valid_targets_min": 6134 + }, + { + "epoch": 3.090525508247027, + "grad_norm": 0.42438016141963625, + "learning_rate": 2.7403008113516127e-05, + "loss": 0.1928, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1028500646352768, + "step": 4030, + "valid_targets_mean": 6886.6, + "valid_targets_min": 5869 + }, + { + "epoch": 3.094361334867664, + "grad_norm": 0.3769271935238977, + "learning_rate": 2.736746872114053e-05, + "loss": 0.1866, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08708025515079498, + "step": 4035, + "valid_targets_mean": 7559.8, + "valid_targets_min": 6145 + }, + { + "epoch": 3.098197161488301, + "grad_norm": 0.3481264124696832, + "learning_rate": 2.7331902392164332e-05, + "loss": 0.177, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0859161764383316, + "step": 4040, + "valid_targets_mean": 8920.5, + "valid_targets_min": 4780 + }, + { + "epoch": 3.1020329881089377, + "grad_norm": 0.3625432962817374, + "learning_rate": 2.7296309256623508e-05, + "loss": 0.1647, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07687647640705109, + "step": 4045, + "valid_targets_mean": 7393.4, + "valid_targets_min": 5116 + }, + { + "epoch": 3.105868814729574, + "grad_norm": 0.3723357643911091, + "learning_rate": 2.7260689444652048e-05, + "loss": 0.1678, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08329356461763382, + "step": 4050, + "valid_targets_mean": 8477.5, + "valid_targets_min": 4033 + }, + { + "epoch": 3.109704641350211, + "grad_norm": 0.4188982766133878, + "learning_rate": 2.7225043086481463e-05, + "loss": 0.1879, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0944136381149292, + "step": 4055, + "valid_targets_mean": 6736.6, + "valid_targets_min": 4646 + }, + { + "epoch": 3.113540467970848, + "grad_norm": 0.4346556730770208, + "learning_rate": 2.7189370312440343e-05, + "loss": 0.1787, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0967080146074295, + "step": 4060, + "valid_targets_mean": 7551.5, + "valid_targets_min": 5461 + }, + { + "epoch": 3.1173762945914847, + "grad_norm": 0.42423784967837885, + "learning_rate": 2.715367125295383e-05, + "loss": 0.1976, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11129596829414368, + "step": 4065, + "valid_targets_mean": 7091.0, + "valid_targets_min": 4879 + }, + { + "epoch": 3.121212121212121, + "grad_norm": 0.4246171709395171, + "learning_rate": 2.7117946038543203e-05, + "loss": 0.2004, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08972804248332977, + "step": 4070, + "valid_targets_mean": 6269.8, + "valid_targets_min": 4606 + }, + { + "epoch": 3.125047947832758, + "grad_norm": 0.39279504340029175, + "learning_rate": 2.7082194799825338e-05, + "loss": 0.1989, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0884200781583786, + "step": 4075, + "valid_targets_mean": 6818.4, + "valid_targets_min": 5525 + }, + { + "epoch": 3.128883774453395, + "grad_norm": 0.40800157372815443, + "learning_rate": 2.704641766751227e-05, + "loss": 0.1847, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09282584488391876, + "step": 4080, + "valid_targets_mean": 7365.4, + "valid_targets_min": 5000 + }, + { + "epoch": 3.1327196010740312, + "grad_norm": 0.44150067178189184, + "learning_rate": 2.701061477241071e-05, + "loss": 0.1913, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08472222834825516, + "step": 4085, + "valid_targets_mean": 6212.1, + "valid_targets_min": 3988 + }, + { + "epoch": 3.136555427694668, + "grad_norm": 0.38435880659210764, + "learning_rate": 2.6974786245421555e-05, + "loss": 0.1923, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09298259019851685, + "step": 4090, + "valid_targets_mean": 8396.9, + "valid_targets_min": 6131 + }, + { + "epoch": 3.140391254315305, + "grad_norm": 0.42809696615122095, + "learning_rate": 2.693893221753942e-05, + "loss": 0.1904, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09648600220680237, + "step": 4095, + "valid_targets_mean": 7651.0, + "valid_targets_min": 4061 + }, + { + "epoch": 3.144227080935942, + "grad_norm": 0.4309422136388789, + "learning_rate": 2.6903052819852163e-05, + "loss": 0.1931, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09368512034416199, + "step": 4100, + "valid_targets_mean": 7090.1, + "valid_targets_min": 5203 + }, + { + "epoch": 3.1480629075565782, + "grad_norm": 0.4757590670484199, + "learning_rate": 2.6867148183540375e-05, + "loss": 0.1933, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09000277519226074, + "step": 4105, + "valid_targets_mean": 7001.0, + "valid_targets_min": 5362 + }, + { + "epoch": 3.151898734177215, + "grad_norm": 0.407452659716199, + "learning_rate": 2.683121843987695e-05, + "loss": 0.1943, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09470850229263306, + "step": 4110, + "valid_targets_mean": 6359.2, + "valid_targets_min": 5236 + }, + { + "epoch": 3.155734560797852, + "grad_norm": 0.43338149690507816, + "learning_rate": 2.6795263720226548e-05, + "loss": 0.1903, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10303812474012375, + "step": 4115, + "valid_targets_mean": 5939.8, + "valid_targets_min": 4693 + }, + { + "epoch": 3.159570387418489, + "grad_norm": 1.2302232334671155, + "learning_rate": 2.675928415604519e-05, + "loss": 0.163, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.036698415875434875, + "step": 4120, + "valid_targets_mean": 208.6, + "valid_targets_min": 134 + }, + { + "epoch": 3.1634062140391253, + "grad_norm": 0.4753286503008569, + "learning_rate": 2.6723279878879683e-05, + "loss": 0.193, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.099032923579216, + "step": 4125, + "valid_targets_mean": 6008.5, + "valid_targets_min": 4956 + }, + { + "epoch": 3.167242040659762, + "grad_norm": 0.41354783960231317, + "learning_rate": 2.668725102036721e-05, + "loss": 0.1853, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08385391533374786, + "step": 4130, + "valid_targets_mean": 6042.0, + "valid_targets_min": 4817 + }, + { + "epoch": 3.171077867280399, + "grad_norm": 0.448787961690868, + "learning_rate": 2.665119771223484e-05, + "loss": 0.1915, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09320216625928879, + "step": 4135, + "valid_targets_mean": 6608.6, + "valid_targets_min": 5102 + }, + { + "epoch": 3.174913693901036, + "grad_norm": 0.41656097065539743, + "learning_rate": 2.661512008629903e-05, + "loss": 0.1838, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09460050612688065, + "step": 4140, + "valid_targets_mean": 6670.1, + "valid_targets_min": 4869 + }, + { + "epoch": 3.1787495205216723, + "grad_norm": 0.4190530901724773, + "learning_rate": 2.6579018274465113e-05, + "loss": 0.1842, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08419083058834076, + "step": 4145, + "valid_targets_mean": 6088.4, + "valid_targets_min": 4946 + }, + { + "epoch": 3.182585347142309, + "grad_norm": 0.43226522688278585, + "learning_rate": 2.6542892408726886e-05, + "loss": 0.1936, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08998145163059235, + "step": 4150, + "valid_targets_mean": 6228.4, + "valid_targets_min": 4856 + }, + { + "epoch": 3.186421173762946, + "grad_norm": 0.4569821593303352, + "learning_rate": 2.6506742621166086e-05, + "loss": 0.193, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09325650334358215, + "step": 4155, + "valid_targets_mean": 5376.6, + "valid_targets_min": 4145 + }, + { + "epoch": 3.190257000383583, + "grad_norm": 0.5023273582988683, + "learning_rate": 2.64705690439519e-05, + "loss": 0.1577, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09473643451929092, + "step": 4160, + "valid_targets_mean": 6309.2, + "valid_targets_min": 5305 + }, + { + "epoch": 3.1940928270042193, + "grad_norm": 0.4261571837295929, + "learning_rate": 2.64343718093405e-05, + "loss": 0.1856, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0912657082080841, + "step": 4165, + "valid_targets_mean": 6221.5, + "valid_targets_min": 4636 + }, + { + "epoch": 3.197928653624856, + "grad_norm": 0.5073478513431653, + "learning_rate": 2.6398151049674542e-05, + "loss": 0.1864, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0989779606461525, + "step": 4170, + "valid_targets_mean": 6831.5, + "valid_targets_min": 5668 + }, + { + "epoch": 3.201764480245493, + "grad_norm": 0.4184509727955677, + "learning_rate": 2.6361906897382716e-05, + "loss": 0.1881, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09788381308317184, + "step": 4175, + "valid_targets_mean": 6437.1, + "valid_targets_min": 5004 + }, + { + "epoch": 3.20560030686613, + "grad_norm": 0.4094748986095778, + "learning_rate": 2.6325639484979225e-05, + "loss": 0.1791, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08726285398006439, + "step": 4180, + "valid_targets_mean": 6205.8, + "valid_targets_min": 5341 + }, + { + "epoch": 3.2094361334867663, + "grad_norm": 0.41943853892115307, + "learning_rate": 2.6289348945063306e-05, + "loss": 0.1858, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0990147590637207, + "step": 4185, + "valid_targets_mean": 6748.0, + "valid_targets_min": 5363 + }, + { + "epoch": 3.213271960107403, + "grad_norm": 0.43117934343560044, + "learning_rate": 2.6253035410318775e-05, + "loss": 0.1835, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08680430054664612, + "step": 4190, + "valid_targets_mean": 5210.5, + "valid_targets_min": 791 + }, + { + "epoch": 3.21710778672804, + "grad_norm": 0.831801813699186, + "learning_rate": 2.62166990135135e-05, + "loss": 0.1949, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07522955536842346, + "step": 4195, + "valid_targets_mean": 1279.5, + "valid_targets_min": 746 + }, + { + "epoch": 3.220943613348677, + "grad_norm": 0.8510402703750597, + "learning_rate": 2.618033988749895e-05, + "loss": 0.175, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09460344910621643, + "step": 4200, + "valid_targets_mean": 1635.8, + "valid_targets_min": 661 + }, + { + "epoch": 3.2247794399693133, + "grad_norm": 0.8450151932137892, + "learning_rate": 2.6143958165209695e-05, + "loss": 0.1714, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08368884027004242, + "step": 4205, + "valid_targets_mean": 1420.2, + "valid_targets_min": 826 + }, + { + "epoch": 3.22861526658995, + "grad_norm": 0.9252890726026557, + "learning_rate": 2.6107553979662906e-05, + "loss": 0.1685, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07627316564321518, + "step": 4210, + "valid_targets_mean": 1399.9, + "valid_targets_min": 742 + }, + { + "epoch": 3.232451093210587, + "grad_norm": 0.801051042957654, + "learning_rate": 2.60711274639579e-05, + "loss": 0.1774, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06959507614374161, + "step": 4215, + "valid_targets_mean": 1309.9, + "valid_targets_min": 718 + }, + { + "epoch": 3.2362869198312234, + "grad_norm": 0.7477586099604203, + "learning_rate": 2.603467875127563e-05, + "loss": 0.1742, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10063307732343674, + "step": 4220, + "valid_targets_mean": 1729.4, + "valid_targets_min": 764 + }, + { + "epoch": 3.2401227464518603, + "grad_norm": 0.7876038222075419, + "learning_rate": 2.5998207974878202e-05, + "loss": 0.1611, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08368270099163055, + "step": 4225, + "valid_targets_mean": 1651.4, + "valid_targets_min": 814 + }, + { + "epoch": 3.243958573072497, + "grad_norm": 0.8727111596402826, + "learning_rate": 2.5961715268108395e-05, + "loss": 0.1803, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0664374977350235, + "step": 4230, + "valid_targets_mean": 1192.0, + "valid_targets_min": 610 + }, + { + "epoch": 3.247794399693134, + "grad_norm": 0.809602428339692, + "learning_rate": 2.5925200764389163e-05, + "loss": 0.1663, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09473729133605957, + "step": 4235, + "valid_targets_mean": 1879.0, + "valid_targets_min": 817 + }, + { + "epoch": 3.251630226313771, + "grad_norm": 0.6967764787202592, + "learning_rate": 2.5888664597223168e-05, + "loss": 0.169, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07764310389757156, + "step": 4240, + "valid_targets_mean": 1500.4, + "valid_targets_min": 837 + }, + { + "epoch": 3.2554660529344073, + "grad_norm": 0.8302394618708009, + "learning_rate": 2.585210690019225e-05, + "loss": 0.1689, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11238352954387665, + "step": 4245, + "valid_targets_mean": 1661.8, + "valid_targets_min": 593 + }, + { + "epoch": 3.259301879555044, + "grad_norm": 0.7209471561348492, + "learning_rate": 2.5815527806956984e-05, + "loss": 0.1657, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08015020191669464, + "step": 4250, + "valid_targets_mean": 1537.8, + "valid_targets_min": 704 + }, + { + "epoch": 3.263137706175681, + "grad_norm": 0.733686436770308, + "learning_rate": 2.577892745125618e-05, + "loss": 0.1601, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06279109418392181, + "step": 4255, + "valid_targets_mean": 1227.0, + "valid_targets_min": 722 + }, + { + "epoch": 3.2669735327963174, + "grad_norm": 0.7301097585834266, + "learning_rate": 2.5742305966906374e-05, + "loss": 0.1595, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0718720406293869, + "step": 4260, + "valid_targets_mean": 1364.5, + "valid_targets_min": 642 + }, + { + "epoch": 3.2708093594169543, + "grad_norm": 0.8629239479927083, + "learning_rate": 2.5705663487801347e-05, + "loss": 0.1584, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07906883209943771, + "step": 4265, + "valid_targets_mean": 1358.5, + "valid_targets_min": 518 + }, + { + "epoch": 3.274645186037591, + "grad_norm": 0.8680318906412322, + "learning_rate": 2.5669000147911645e-05, + "loss": 0.1694, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09543141722679138, + "step": 4270, + "valid_targets_mean": 1590.6, + "valid_targets_min": 728 + }, + { + "epoch": 3.278481012658228, + "grad_norm": 0.7874132209262358, + "learning_rate": 2.56323160812841e-05, + "loss": 0.1707, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10341475903987885, + "step": 4275, + "valid_targets_mean": 1791.4, + "valid_targets_min": 820 + }, + { + "epoch": 3.2823168392788644, + "grad_norm": 0.8505064233987827, + "learning_rate": 2.5595611422041308e-05, + "loss": 0.1769, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09316158294677734, + "step": 4280, + "valid_targets_mean": 1470.8, + "valid_targets_min": 818 + }, + { + "epoch": 3.2861526658995013, + "grad_norm": 0.7703259392615531, + "learning_rate": 2.5558886304381142e-05, + "loss": 0.1615, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06890230625867844, + "step": 4285, + "valid_targets_mean": 1278.6, + "valid_targets_min": 625 + }, + { + "epoch": 3.289988492520138, + "grad_norm": 0.8494964493534117, + "learning_rate": 2.5522140862576307e-05, + "loss": 0.1667, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10680123418569565, + "step": 4290, + "valid_targets_mean": 2313.5, + "valid_targets_min": 790 + }, + { + "epoch": 3.293824319140775, + "grad_norm": 0.7870302634782272, + "learning_rate": 2.5485375230973796e-05, + "loss": 0.1722, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11990931630134583, + "step": 4295, + "valid_targets_mean": 2231.9, + "valid_targets_min": 1335 + }, + { + "epoch": 3.2976601457614114, + "grad_norm": 0.7506702374598613, + "learning_rate": 2.544858954399442e-05, + "loss": 0.1655, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09394356608390808, + "step": 4300, + "valid_targets_mean": 1729.5, + "valid_targets_min": 719 + }, + { + "epoch": 3.3014959723820483, + "grad_norm": 0.7562629219502657, + "learning_rate": 2.5411783936132305e-05, + "loss": 0.159, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06409837305545807, + "step": 4305, + "valid_targets_mean": 1222.9, + "valid_targets_min": 697 + }, + { + "epoch": 3.305331799002685, + "grad_norm": 0.823627572137894, + "learning_rate": 2.5374958541954443e-05, + "loss": 0.1619, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09509257972240448, + "step": 4310, + "valid_targets_mean": 1521.1, + "valid_targets_min": 998 + }, + { + "epoch": 3.309167625623322, + "grad_norm": 0.7923675004410574, + "learning_rate": 2.5338113496100136e-05, + "loss": 0.1627, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08676047623157501, + "step": 4315, + "valid_targets_mean": 1693.9, + "valid_targets_min": 629 + }, + { + "epoch": 3.3130034522439584, + "grad_norm": 0.7968626444399963, + "learning_rate": 2.530124893328055e-05, + "loss": 0.1633, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08823008835315704, + "step": 4320, + "valid_targets_mean": 1630.2, + "valid_targets_min": 1078 + }, + { + "epoch": 3.3168392788645953, + "grad_norm": 0.7320825906589101, + "learning_rate": 2.5264364988278215e-05, + "loss": 0.1646, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08281449973583221, + "step": 4325, + "valid_targets_mean": 1843.0, + "valid_targets_min": 1075 + }, + { + "epoch": 3.320675105485232, + "grad_norm": 0.8042385599960803, + "learning_rate": 2.52274617959465e-05, + "loss": 0.1672, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09079332649707794, + "step": 4330, + "valid_targets_mean": 1848.2, + "valid_targets_min": 1317 + }, + { + "epoch": 3.324510932105869, + "grad_norm": 0.8084730604164897, + "learning_rate": 2.519053949120917e-05, + "loss": 0.1605, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09392134100198746, + "step": 4335, + "valid_targets_mean": 1727.4, + "valid_targets_min": 957 + }, + { + "epoch": 3.3283467587265054, + "grad_norm": 0.8516891257997655, + "learning_rate": 2.5153598209059858e-05, + "loss": 0.1612, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06505317986011505, + "step": 4340, + "valid_targets_mean": 1115.4, + "valid_targets_min": 643 + }, + { + "epoch": 3.3321825853471423, + "grad_norm": 0.8290995610232987, + "learning_rate": 2.5116638084561583e-05, + "loss": 0.168, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08371862769126892, + "step": 4345, + "valid_targets_mean": 1547.2, + "valid_targets_min": 1111 + }, + { + "epoch": 3.336018411967779, + "grad_norm": 0.8877616263405244, + "learning_rate": 2.507965925284625e-05, + "loss": 0.1557, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07133730500936508, + "step": 4350, + "valid_targets_mean": 1450.0, + "valid_targets_min": 705 + }, + { + "epoch": 3.3398542385884156, + "grad_norm": 1.0307331243287057, + "learning_rate": 2.5042661849114175e-05, + "loss": 0.16, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08405070006847382, + "step": 4355, + "valid_targets_mean": 1539.0, + "valid_targets_min": 1288 + }, + { + "epoch": 3.3436900652090524, + "grad_norm": 0.7792726123318077, + "learning_rate": 2.5005646008633552e-05, + "loss": 0.1513, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07645267248153687, + "step": 4360, + "valid_targets_mean": 1321.2, + "valid_targets_min": 753 + }, + { + "epoch": 3.3475258918296893, + "grad_norm": 0.7729611057694937, + "learning_rate": 2.496861186674e-05, + "loss": 0.1605, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07624319195747375, + "step": 4365, + "valid_targets_mean": 1480.0, + "valid_targets_min": 817 + }, + { + "epoch": 3.351361718450326, + "grad_norm": 0.8130852658086701, + "learning_rate": 2.4931559558836046e-05, + "loss": 0.1579, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06702481210231781, + "step": 4370, + "valid_targets_mean": 1293.9, + "valid_targets_min": 974 + }, + { + "epoch": 3.355197545070963, + "grad_norm": 0.8629210900629497, + "learning_rate": 2.4894489220390648e-05, + "loss": 0.1615, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0881173238158226, + "step": 4375, + "valid_targets_mean": 1529.2, + "valid_targets_min": 657 + }, + { + "epoch": 3.3590333716915994, + "grad_norm": 0.9926260156957495, + "learning_rate": 2.485740098693866e-05, + "loss": 0.1627, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09785797446966171, + "step": 4380, + "valid_targets_mean": 1678.0, + "valid_targets_min": 880 + }, + { + "epoch": 3.3628691983122363, + "grad_norm": 0.9326324493002796, + "learning_rate": 2.482029499408038e-05, + "loss": 0.1624, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09621813893318176, + "step": 4385, + "valid_targets_mean": 1609.4, + "valid_targets_min": 1050 + }, + { + "epoch": 3.366705024932873, + "grad_norm": 0.830007317147164, + "learning_rate": 2.4783171377481033e-05, + "loss": 0.1557, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07370647042989731, + "step": 4390, + "valid_targets_mean": 1449.0, + "valid_targets_min": 682 + }, + { + "epoch": 3.3705408515535096, + "grad_norm": 0.732377717133225, + "learning_rate": 2.4746030272870284e-05, + "loss": 0.1611, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07061082124710083, + "step": 4395, + "valid_targets_mean": 1344.0, + "valid_targets_min": 703 + }, + { + "epoch": 3.3743766781741464, + "grad_norm": 0.9779602749968135, + "learning_rate": 2.470887181604174e-05, + "loss": 0.1687, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07591959089040756, + "step": 4400, + "valid_targets_mean": 1469.4, + "valid_targets_min": 934 + }, + { + "epoch": 3.3782125047947833, + "grad_norm": 0.764350323742031, + "learning_rate": 2.467169614285242e-05, + "loss": 0.1743, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12555918097496033, + "step": 4405, + "valid_targets_mean": 1829.1, + "valid_targets_min": 734 + }, + { + "epoch": 3.38204833141542, + "grad_norm": 0.7758400876337053, + "learning_rate": 2.463450338922234e-05, + "loss": 0.1573, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10159923881292343, + "step": 4410, + "valid_targets_mean": 2010.8, + "valid_targets_min": 1437 + }, + { + "epoch": 3.3858841580360566, + "grad_norm": 0.9010304807398007, + "learning_rate": 2.4597293691133918e-05, + "loss": 0.1555, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.061382949352264404, + "step": 4415, + "valid_targets_mean": 1405.2, + "valid_targets_min": 551 + }, + { + "epoch": 3.3897199846566934, + "grad_norm": 0.8396548605056326, + "learning_rate": 2.4560067184631555e-05, + "loss": 0.1683, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09914135932922363, + "step": 4420, + "valid_targets_mean": 1817.6, + "valid_targets_min": 931 + }, + { + "epoch": 3.3935558112773303, + "grad_norm": 0.7559502227155085, + "learning_rate": 2.4522824005821075e-05, + "loss": 0.1573, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07862640172243118, + "step": 4425, + "valid_targets_mean": 1599.1, + "valid_targets_min": 904 + }, + { + "epoch": 3.397391637897967, + "grad_norm": 0.855639200324802, + "learning_rate": 2.4485564290869298e-05, + "loss": 0.1518, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08711159229278564, + "step": 4430, + "valid_targets_mean": 1774.8, + "valid_targets_min": 1158 + }, + { + "epoch": 3.4012274645186036, + "grad_norm": 0.8036449251276796, + "learning_rate": 2.444828817600347e-05, + "loss": 0.1486, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07083986699581146, + "step": 4435, + "valid_targets_mean": 1229.2, + "valid_targets_min": 756 + }, + { + "epoch": 3.4050632911392404, + "grad_norm": 0.9577819826502114, + "learning_rate": 2.441099579751081e-05, + "loss": 0.1687, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09361335635185242, + "step": 4440, + "valid_targets_mean": 1536.9, + "valid_targets_min": 854 + }, + { + "epoch": 3.4088991177598773, + "grad_norm": 0.7273210044587625, + "learning_rate": 2.4373687291738003e-05, + "loss": 0.1631, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08687122166156769, + "step": 4445, + "valid_targets_mean": 1739.5, + "valid_targets_min": 1462 + }, + { + "epoch": 3.412734944380514, + "grad_norm": 0.8055899043617806, + "learning_rate": 2.4336362795090694e-05, + "loss": 0.1561, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06997404992580414, + "step": 4450, + "valid_targets_mean": 1475.9, + "valid_targets_min": 944 + }, + { + "epoch": 3.4165707710011506, + "grad_norm": 1.7633921341241074, + "learning_rate": 2.4299022444032986e-05, + "loss": 0.1558, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0717436894774437, + "step": 4455, + "valid_targets_mean": 1482.6, + "valid_targets_min": 1227 + }, + { + "epoch": 3.4204065976217874, + "grad_norm": 0.7687651333660249, + "learning_rate": 2.4261666375086957e-05, + "loss": 0.1618, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06661325693130493, + "step": 4460, + "valid_targets_mean": 1276.5, + "valid_targets_min": 740 + }, + { + "epoch": 3.4242424242424243, + "grad_norm": 0.7048964135944301, + "learning_rate": 2.4224294724832152e-05, + "loss": 0.1594, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09349218755960464, + "step": 4465, + "valid_targets_mean": 2024.8, + "valid_targets_min": 881 + }, + { + "epoch": 3.428078250863061, + "grad_norm": 0.8700047805533708, + "learning_rate": 2.418690762990508e-05, + "loss": 0.1571, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07991793751716614, + "step": 4470, + "valid_targets_mean": 1288.6, + "valid_targets_min": 636 + }, + { + "epoch": 3.4319140774836976, + "grad_norm": 0.8285212623873, + "learning_rate": 2.414950522699872e-05, + "loss": 0.158, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05591881275177002, + "step": 4475, + "valid_targets_mean": 1103.9, + "valid_targets_min": 814 + }, + { + "epoch": 3.4357499041043345, + "grad_norm": 0.8557612237853773, + "learning_rate": 2.4112087652862018e-05, + "loss": 0.1506, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07987368106842041, + "step": 4480, + "valid_targets_mean": 1392.9, + "valid_targets_min": 820 + }, + { + "epoch": 3.4395857307249713, + "grad_norm": 0.8221790298446408, + "learning_rate": 2.4074655044299395e-05, + "loss": 0.1582, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07046188414096832, + "step": 4485, + "valid_targets_mean": 1392.4, + "valid_targets_min": 661 + }, + { + "epoch": 3.4434215573456077, + "grad_norm": 0.8096403111964474, + "learning_rate": 2.4037207538170223e-05, + "loss": 0.1604, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07357046008110046, + "step": 4490, + "valid_targets_mean": 1441.8, + "valid_targets_min": 728 + }, + { + "epoch": 3.4472573839662446, + "grad_norm": 0.8362503815998606, + "learning_rate": 2.399974527138837e-05, + "loss": 0.1551, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07548461109399796, + "step": 4495, + "valid_targets_mean": 1413.5, + "valid_targets_min": 692 + }, + { + "epoch": 3.4510932105868815, + "grad_norm": 0.8257735857367823, + "learning_rate": 2.3962268380921635e-05, + "loss": 0.1608, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0798138901591301, + "step": 4500, + "valid_targets_mean": 1323.5, + "valid_targets_min": 840 + }, + { + "epoch": 3.4549290372075183, + "grad_norm": 0.9070952500020759, + "learning_rate": 2.3924777003791318e-05, + "loss": 0.1622, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07953017950057983, + "step": 4505, + "valid_targets_mean": 1488.4, + "valid_targets_min": 950 + }, + { + "epoch": 3.458764863828155, + "grad_norm": 0.778731343502711, + "learning_rate": 2.3887271277071655e-05, + "loss": 0.1561, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08750265091657639, + "step": 4510, + "valid_targets_mean": 1935.9, + "valid_targets_min": 878 + }, + { + "epoch": 3.4626006904487916, + "grad_norm": 0.8538679864006609, + "learning_rate": 2.3849751337889377e-05, + "loss": 0.1599, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.082953542470932, + "step": 4515, + "valid_targets_mean": 1502.1, + "valid_targets_min": 836 + }, + { + "epoch": 3.4664365170694285, + "grad_norm": 0.82835132319106, + "learning_rate": 2.381221732342315e-05, + "loss": 0.1549, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08478623628616333, + "step": 4520, + "valid_targets_mean": 1535.1, + "valid_targets_min": 1013 + }, + { + "epoch": 3.4702723436900653, + "grad_norm": 0.877020550544337, + "learning_rate": 2.3774669370903113e-05, + "loss": 0.1623, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07010465860366821, + "step": 4525, + "valid_targets_mean": 1346.4, + "valid_targets_min": 748 + }, + { + "epoch": 3.4741081703107017, + "grad_norm": 0.8685133323265621, + "learning_rate": 2.3737107617610376e-05, + "loss": 0.158, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08062994480133057, + "step": 4530, + "valid_targets_mean": 1543.9, + "valid_targets_min": 729 + }, + { + "epoch": 3.4779439969313386, + "grad_norm": 0.7341874392221571, + "learning_rate": 2.369953220087649e-05, + "loss": 0.1687, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07027609646320343, + "step": 4535, + "valid_targets_mean": 1541.9, + "valid_targets_min": 1219 + }, + { + "epoch": 3.4817798235519755, + "grad_norm": 0.8025473341522802, + "learning_rate": 2.3661943258082952e-05, + "loss": 0.1607, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08055227994918823, + "step": 4540, + "valid_targets_mean": 1632.8, + "valid_targets_min": 916 + }, + { + "epoch": 3.4856156501726123, + "grad_norm": 0.7790540281264153, + "learning_rate": 2.362434092666075e-05, + "loss": 0.1594, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07335825264453888, + "step": 4545, + "valid_targets_mean": 1349.9, + "valid_targets_min": 646 + }, + { + "epoch": 3.489451476793249, + "grad_norm": 0.8575752309013551, + "learning_rate": 2.3586725344089785e-05, + "loss": 0.1573, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08786715567111969, + "step": 4550, + "valid_targets_mean": 1383.0, + "valid_targets_min": 906 + }, + { + "epoch": 3.4932873034138856, + "grad_norm": 0.8123236639165956, + "learning_rate": 2.3549096647898436e-05, + "loss": 0.1511, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07350484281778336, + "step": 4555, + "valid_targets_mean": 1577.5, + "valid_targets_min": 848 + }, + { + "epoch": 3.4971231300345225, + "grad_norm": 0.8444583778464302, + "learning_rate": 2.3511454975662992e-05, + "loss": 0.1627, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07474963366985321, + "step": 4560, + "valid_targets_mean": 1183.5, + "valid_targets_min": 692 + }, + { + "epoch": 3.5009589566551593, + "grad_norm": 0.8525282692986003, + "learning_rate": 2.347380046500723e-05, + "loss": 0.1535, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07284164428710938, + "step": 4565, + "valid_targets_mean": 1373.9, + "valid_targets_min": 821 + }, + { + "epoch": 3.5047947832757957, + "grad_norm": 0.762074736753951, + "learning_rate": 2.343613325360182e-05, + "loss": 0.1508, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06769385188817978, + "step": 4570, + "valid_targets_mean": 1428.1, + "valid_targets_min": 1013 + }, + { + "epoch": 3.5086306098964326, + "grad_norm": 0.8094048459172393, + "learning_rate": 2.33984534791639e-05, + "loss": 0.1516, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07062892615795135, + "step": 4575, + "valid_targets_mean": 1420.9, + "valid_targets_min": 781 + }, + { + "epoch": 3.5124664365170695, + "grad_norm": 0.9293796518935323, + "learning_rate": 2.336076127945653e-05, + "loss": 0.1669, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08843650668859482, + "step": 4580, + "valid_targets_mean": 1255.1, + "valid_targets_min": 548 + }, + { + "epoch": 3.5163022631377063, + "grad_norm": 0.820923476477777, + "learning_rate": 2.3323056792288198e-05, + "loss": 0.1506, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07857775688171387, + "step": 4585, + "valid_targets_mean": 1411.0, + "valid_targets_min": 694 + }, + { + "epoch": 3.520138089758343, + "grad_norm": 0.7061432731269989, + "learning_rate": 2.3285340155512313e-05, + "loss": 0.1574, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09083489328622818, + "step": 4590, + "valid_targets_mean": 1973.8, + "valid_targets_min": 1030 + }, + { + "epoch": 3.5239739163789796, + "grad_norm": 0.8546312280437895, + "learning_rate": 2.324761150702672e-05, + "loss": 0.1448, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06227022036910057, + "step": 4595, + "valid_targets_mean": 1580.2, + "valid_targets_min": 682 + }, + { + "epoch": 3.5278097429996165, + "grad_norm": 1.1312812073218301, + "learning_rate": 2.3209870984773157e-05, + "loss": 0.1466, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07357725501060486, + "step": 4600, + "valid_targets_mean": 1425.4, + "valid_targets_min": 990 + }, + { + "epoch": 3.5316455696202533, + "grad_norm": 0.8861809668916341, + "learning_rate": 2.3172118726736797e-05, + "loss": 0.1549, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05760674178600311, + "step": 4605, + "valid_targets_mean": 1299.6, + "valid_targets_min": 639 + }, + { + "epoch": 3.5354813962408898, + "grad_norm": 0.8371167116403342, + "learning_rate": 2.3134354870945698e-05, + "loss": 0.1544, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06614987552165985, + "step": 4610, + "valid_targets_mean": 1458.8, + "valid_targets_min": 721 + }, + { + "epoch": 3.5393172228615266, + "grad_norm": 1.0421248401763537, + "learning_rate": 2.3096579555470354e-05, + "loss": 0.1528, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07283510267734528, + "step": 4615, + "valid_targets_mean": 1329.8, + "valid_targets_min": 874 + }, + { + "epoch": 3.5431530494821635, + "grad_norm": 1.1321346895094822, + "learning_rate": 2.3058792918423122e-05, + "loss": 0.1551, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07509522885084152, + "step": 4620, + "valid_targets_mean": 1559.9, + "valid_targets_min": 686 + }, + { + "epoch": 3.5469888761028, + "grad_norm": 1.1702887322341096, + "learning_rate": 2.302099509795778e-05, + "loss": 0.156, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08168554306030273, + "step": 4625, + "valid_targets_mean": 1392.5, + "valid_targets_min": 713 + }, + { + "epoch": 3.5508247027234368, + "grad_norm": 0.88187346028639, + "learning_rate": 2.2983186232268966e-05, + "loss": 0.1658, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1410033255815506, + "step": 4630, + "valid_targets_mean": 2341.5, + "valid_targets_min": 920 + }, + { + "epoch": 3.5546605293440736, + "grad_norm": 0.6420489080732316, + "learning_rate": 2.2945366459591735e-05, + "loss": 0.267, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12536510825157166, + "step": 4635, + "valid_targets_mean": 5153.4, + "valid_targets_min": 746 + }, + { + "epoch": 3.5584963559647105, + "grad_norm": 0.4879229308596516, + "learning_rate": 2.290753591820099e-05, + "loss": 0.2461, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1246916651725769, + "step": 4640, + "valid_targets_mean": 5458.5, + "valid_targets_min": 1846 + }, + { + "epoch": 3.5623321825853473, + "grad_norm": 0.5229422936790133, + "learning_rate": 2.2869694746411016e-05, + "loss": 0.2386, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09194843471050262, + "step": 4645, + "valid_targets_mean": 4888.1, + "valid_targets_min": 1321 + }, + { + "epoch": 3.5661680092059838, + "grad_norm": 0.4881049885658451, + "learning_rate": 2.2831843082574984e-05, + "loss": 0.2386, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12018633633852005, + "step": 4650, + "valid_targets_mean": 5733.0, + "valid_targets_min": 2373 + }, + { + "epoch": 3.5700038358266206, + "grad_norm": 0.6599243957681933, + "learning_rate": 2.2793981065084396e-05, + "loss": 0.2492, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11306305229663849, + "step": 4655, + "valid_targets_mean": 4224.2, + "valid_targets_min": 2075 + }, + { + "epoch": 3.5738396624472575, + "grad_norm": 0.5392883735619209, + "learning_rate": 2.2756108832368614e-05, + "loss": 0.2439, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12436346709728241, + "step": 4660, + "valid_targets_mean": 5458.2, + "valid_targets_min": 1991 + }, + { + "epoch": 3.577675489067894, + "grad_norm": 0.6022451691306016, + "learning_rate": 2.2718226522894367e-05, + "loss": 0.2376, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10744770616292953, + "step": 4665, + "valid_targets_mean": 3290.8, + "valid_targets_min": 1748 + }, + { + "epoch": 3.5815113156885308, + "grad_norm": 0.5216248543077792, + "learning_rate": 2.268033427516521e-05, + "loss": 0.2416, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10728374123573303, + "step": 4670, + "valid_targets_mean": 4719.1, + "valid_targets_min": 1170 + }, + { + "epoch": 3.5853471423091676, + "grad_norm": 0.5452754026383936, + "learning_rate": 2.264243222772104e-05, + "loss": 0.2493, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10874859988689423, + "step": 4675, + "valid_targets_mean": 4463.9, + "valid_targets_min": 1241 + }, + { + "epoch": 3.5891829689298045, + "grad_norm": 0.5548902251696574, + "learning_rate": 2.260452051913757e-05, + "loss": 0.2428, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1284613013267517, + "step": 4680, + "valid_targets_mean": 4915.0, + "valid_targets_min": 2747 + }, + { + "epoch": 3.5930187955504413, + "grad_norm": 0.5277478312589667, + "learning_rate": 2.256659928802586e-05, + "loss": 0.2399, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10982024669647217, + "step": 4685, + "valid_targets_mean": 5663.9, + "valid_targets_min": 2144 + }, + { + "epoch": 3.5968546221710778, + "grad_norm": 0.6683477399087345, + "learning_rate": 2.252866867303177e-05, + "loss": 0.2362, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11507170647382736, + "step": 4690, + "valid_targets_mean": 5095.8, + "valid_targets_min": 2587 + }, + { + "epoch": 3.6006904487917146, + "grad_norm": 0.5427378198632373, + "learning_rate": 2.249072881283547e-05, + "loss": 0.2448, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12106771767139435, + "step": 4695, + "valid_targets_mean": 5127.1, + "valid_targets_min": 1366 + }, + { + "epoch": 3.6045262754123515, + "grad_norm": 0.6661025277417807, + "learning_rate": 2.2452779846150934e-05, + "loss": 0.2377, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1099114716053009, + "step": 4700, + "valid_targets_mean": 3455.5, + "valid_targets_min": 1721 + }, + { + "epoch": 3.608362102032988, + "grad_norm": 0.7593295134629314, + "learning_rate": 2.2414821911725433e-05, + "loss": 0.2476, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1130317747592926, + "step": 4705, + "valid_targets_mean": 3213.5, + "valid_targets_min": 746 + }, + { + "epoch": 3.6121979286536248, + "grad_norm": 0.6334187205405916, + "learning_rate": 2.2376855148339013e-05, + "loss": 0.2288, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11826205253601074, + "step": 4710, + "valid_targets_mean": 3643.8, + "valid_targets_min": 1538 + }, + { + "epoch": 3.6160337552742616, + "grad_norm": 0.6162018166484945, + "learning_rate": 2.233887969480402e-05, + "loss": 0.2371, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09771811962127686, + "step": 4715, + "valid_targets_mean": 3245.9, + "valid_targets_min": 1451 + }, + { + "epoch": 3.6198695818948985, + "grad_norm": 0.632526853720378, + "learning_rate": 2.230089568996456e-05, + "loss": 0.2492, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12717151641845703, + "step": 4720, + "valid_targets_mean": 4339.1, + "valid_targets_min": 2550 + }, + { + "epoch": 3.6237054085155354, + "grad_norm": 0.6202202756511548, + "learning_rate": 2.2262903272695998e-05, + "loss": 0.2453, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10400520265102386, + "step": 4725, + "valid_targets_mean": 3526.1, + "valid_targets_min": 1323 + }, + { + "epoch": 3.6275412351361718, + "grad_norm": 0.7742392534830955, + "learning_rate": 2.2224902581904476e-05, + "loss": 0.241, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13973087072372437, + "step": 4730, + "valid_targets_mean": 4593.9, + "valid_targets_min": 1921 + }, + { + "epoch": 3.6313770617568086, + "grad_norm": 0.580642225782057, + "learning_rate": 2.2186893756526366e-05, + "loss": 0.2258, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10396385192871094, + "step": 4735, + "valid_targets_mean": 4119.6, + "valid_targets_min": 2004 + }, + { + "epoch": 3.6352128883774455, + "grad_norm": 0.622485028259933, + "learning_rate": 2.2148876935527794e-05, + "loss": 0.2353, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09525234252214432, + "step": 4740, + "valid_targets_mean": 3539.5, + "valid_targets_min": 2377 + }, + { + "epoch": 3.639048714998082, + "grad_norm": 0.775124072015797, + "learning_rate": 2.2110852257904108e-05, + "loss": 0.2173, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1249440610408783, + "step": 4745, + "valid_targets_mean": 3529.1, + "valid_targets_min": 2183 + }, + { + "epoch": 3.6428845416187188, + "grad_norm": 0.6557081578746007, + "learning_rate": 2.2072819862679396e-05, + "loss": 0.2393, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.141234889626503, + "step": 4750, + "valid_targets_mean": 4312.6, + "valid_targets_min": 3031 + }, + { + "epoch": 3.6467203682393556, + "grad_norm": 0.5970527917201272, + "learning_rate": 2.2034779888905943e-05, + "loss": 0.2319, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08838079869747162, + "step": 4755, + "valid_targets_mean": 3192.2, + "valid_targets_min": 1475 + }, + { + "epoch": 3.650556194859992, + "grad_norm": 0.548606403788292, + "learning_rate": 2.199673247566376e-05, + "loss": 0.233, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09856103360652924, + "step": 4760, + "valid_targets_mean": 3245.2, + "valid_targets_min": 648 + }, + { + "epoch": 3.654392021480629, + "grad_norm": 0.5891018158821746, + "learning_rate": 2.1958677762060043e-05, + "loss": 0.2358, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1267816573381424, + "step": 4765, + "valid_targets_mean": 3821.4, + "valid_targets_min": 1233 + }, + { + "epoch": 3.6582278481012658, + "grad_norm": 0.6810692518870152, + "learning_rate": 2.1920615887228705e-05, + "loss": 0.2231, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1083626002073288, + "step": 4770, + "valid_targets_mean": 3471.6, + "valid_targets_min": 2831 + }, + { + "epoch": 3.6620636747219026, + "grad_norm": 0.6464202235091177, + "learning_rate": 2.1882546990329806e-05, + "loss": 0.2388, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11532345414161682, + "step": 4775, + "valid_targets_mean": 3616.6, + "valid_targets_min": 1781 + }, + { + "epoch": 3.6658995013425395, + "grad_norm": 0.670555614019718, + "learning_rate": 2.1844471210549102e-05, + "loss": 0.2295, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11831757426261902, + "step": 4780, + "valid_targets_mean": 3264.4, + "valid_targets_min": 1108 + }, + { + "epoch": 3.669735327963176, + "grad_norm": 0.6474403116404012, + "learning_rate": 2.1806388687097517e-05, + "loss": 0.2277, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08612930774688721, + "step": 4785, + "valid_targets_mean": 2588.5, + "valid_targets_min": 1913 + }, + { + "epoch": 3.673571154583813, + "grad_norm": 0.6572492500821666, + "learning_rate": 2.1768299559210617e-05, + "loss": 0.2279, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12740039825439453, + "step": 4790, + "valid_targets_mean": 3787.1, + "valid_targets_min": 1828 + }, + { + "epoch": 3.6774069812044496, + "grad_norm": 0.7044916399564258, + "learning_rate": 2.1730203966148123e-05, + "loss": 0.2318, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11263582110404968, + "step": 4795, + "valid_targets_mean": 4211.4, + "valid_targets_min": 1589 + }, + { + "epoch": 3.681242807825086, + "grad_norm": 0.6663777393856549, + "learning_rate": 2.1692102047193383e-05, + "loss": 0.2231, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10548977553844452, + "step": 4800, + "valid_targets_mean": 3294.1, + "valid_targets_min": 1328 + }, + { + "epoch": 3.685078634445723, + "grad_norm": 0.6419524458885351, + "learning_rate": 2.1653993941652898e-05, + "loss": 0.2341, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08727402240037918, + "step": 4805, + "valid_targets_mean": 2840.2, + "valid_targets_min": 1248 + }, + { + "epoch": 3.68891446106636, + "grad_norm": 0.6285885898761756, + "learning_rate": 2.1615879788855763e-05, + "loss": 0.2391, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10752232372760773, + "step": 4810, + "valid_targets_mean": 3823.9, + "valid_targets_min": 1727 + }, + { + "epoch": 3.6927502876869966, + "grad_norm": 0.5757399786887223, + "learning_rate": 2.157775972815319e-05, + "loss": 0.2257, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1208435446023941, + "step": 4815, + "valid_targets_mean": 4394.8, + "valid_targets_min": 1965 + }, + { + "epoch": 3.6965861143076335, + "grad_norm": 0.5869861398720324, + "learning_rate": 2.1539633898918004e-05, + "loss": 0.2274, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1049576848745346, + "step": 4820, + "valid_targets_mean": 3348.1, + "valid_targets_min": 1184 + }, + { + "epoch": 3.70042194092827, + "grad_norm": 0.6210122923313802, + "learning_rate": 2.15015024405441e-05, + "loss": 0.2285, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10667955875396729, + "step": 4825, + "valid_targets_mean": 4168.8, + "valid_targets_min": 2490 + }, + { + "epoch": 3.704257767548907, + "grad_norm": 0.6045166514196827, + "learning_rate": 2.1463365492445964e-05, + "loss": 0.2365, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11822743713855743, + "step": 4830, + "valid_targets_mean": 3463.2, + "valid_targets_min": 2500 + }, + { + "epoch": 3.7080935941695436, + "grad_norm": 0.7079333623640442, + "learning_rate": 2.1425223194058158e-05, + "loss": 0.227, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14823979139328003, + "step": 4835, + "valid_targets_mean": 3757.6, + "valid_targets_min": 1648 + }, + { + "epoch": 3.71192942079018, + "grad_norm": 0.6250568095035602, + "learning_rate": 2.13870756848348e-05, + "loss": 0.2308, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1153615340590477, + "step": 4840, + "valid_targets_mean": 3793.1, + "valid_targets_min": 1580 + }, + { + "epoch": 3.715765247410817, + "grad_norm": 0.6102613320049125, + "learning_rate": 2.134892310424906e-05, + "loss": 0.2276, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12046004831790924, + "step": 4845, + "valid_targets_mean": 3714.8, + "valid_targets_min": 1577 + }, + { + "epoch": 3.719601074031454, + "grad_norm": 0.537427206437267, + "learning_rate": 2.1310765591792647e-05, + "loss": 0.2181, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1036638393998146, + "step": 4850, + "valid_targets_mean": 4516.8, + "valid_targets_min": 2095 + }, + { + "epoch": 3.7234369006520907, + "grad_norm": 0.5927192888946426, + "learning_rate": 2.127260328697531e-05, + "loss": 0.225, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10659755766391754, + "step": 4855, + "valid_targets_mean": 3676.5, + "valid_targets_min": 1673 + }, + { + "epoch": 3.7272727272727275, + "grad_norm": 0.5385251528226888, + "learning_rate": 2.1234436329324306e-05, + "loss": 0.2316, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12931561470031738, + "step": 4860, + "valid_targets_mean": 4783.1, + "valid_targets_min": 2494 + }, + { + "epoch": 3.731108553893364, + "grad_norm": 0.6169841870093304, + "learning_rate": 2.1196264858383918e-05, + "loss": 0.2245, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10140418261289597, + "step": 4865, + "valid_targets_mean": 3445.6, + "valid_targets_min": 742 + }, + { + "epoch": 3.734944380514001, + "grad_norm": 0.6127033160656953, + "learning_rate": 2.115808901371493e-05, + "loss": 0.2364, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12366826832294464, + "step": 4870, + "valid_targets_mean": 4115.5, + "valid_targets_min": 1766 + }, + { + "epoch": 3.7387802071346377, + "grad_norm": 0.6199145954697589, + "learning_rate": 2.11199089348941e-05, + "loss": 0.2263, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11888039857149124, + "step": 4875, + "valid_targets_mean": 4264.9, + "valid_targets_min": 1504 + }, + { + "epoch": 3.742616033755274, + "grad_norm": 0.6088501879618458, + "learning_rate": 2.108172476151368e-05, + "loss": 0.2378, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12193469703197479, + "step": 4880, + "valid_targets_mean": 3899.8, + "valid_targets_min": 1170 + }, + { + "epoch": 3.746451860375911, + "grad_norm": 0.6418054444561634, + "learning_rate": 2.1043536633180894e-05, + "loss": 0.2256, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11018107831478119, + "step": 4885, + "valid_targets_mean": 2840.8, + "valid_targets_min": 1955 + }, + { + "epoch": 3.750287686996548, + "grad_norm": 0.5865524164566824, + "learning_rate": 2.100534468951742e-05, + "loss": 0.2331, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10905294865369797, + "step": 4890, + "valid_targets_mean": 3381.4, + "valid_targets_min": 1221 + }, + { + "epoch": 3.7541235136171847, + "grad_norm": 0.5665667091830437, + "learning_rate": 2.096714907015889e-05, + "loss": 0.2192, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09552112221717834, + "step": 4895, + "valid_targets_mean": 4147.0, + "valid_targets_min": 2756 + }, + { + "epoch": 3.7579593402378215, + "grad_norm": 0.5872883694662584, + "learning_rate": 2.0928949914754363e-05, + "loss": 0.2188, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10933491587638855, + "step": 4900, + "valid_targets_mean": 4371.0, + "valid_targets_min": 2526 + }, + { + "epoch": 3.761795166858458, + "grad_norm": 0.6266200746059385, + "learning_rate": 2.0890747362965847e-05, + "loss": 0.2156, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13050781190395355, + "step": 4905, + "valid_targets_mean": 4382.9, + "valid_targets_min": 464 + }, + { + "epoch": 3.765630993479095, + "grad_norm": 0.6000533845325107, + "learning_rate": 2.0852541554467763e-05, + "loss": 0.2351, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11303850263357162, + "step": 4910, + "valid_targets_mean": 3763.0, + "valid_targets_min": 1003 + }, + { + "epoch": 3.7694668200997317, + "grad_norm": 0.6651696410866639, + "learning_rate": 2.0814332628946415e-05, + "loss": 0.2276, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12612995505332947, + "step": 4915, + "valid_targets_mean": 3726.9, + "valid_targets_min": 1632 + }, + { + "epoch": 3.773302646720368, + "grad_norm": 0.5806901439389114, + "learning_rate": 2.0776120726099516e-05, + "loss": 0.2344, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11103049665689468, + "step": 4920, + "valid_targets_mean": 3750.4, + "valid_targets_min": 1969 + }, + { + "epoch": 3.777138473341005, + "grad_norm": 0.5771959135227241, + "learning_rate": 2.0737905985635693e-05, + "loss": 0.2256, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10986604541540146, + "step": 4925, + "valid_targets_mean": 3843.2, + "valid_targets_min": 1460 + }, + { + "epoch": 3.780974299961642, + "grad_norm": 0.6560726137669748, + "learning_rate": 2.0699688547273915e-05, + "loss": 0.2372, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11003892868757248, + "step": 4930, + "valid_targets_mean": 4060.6, + "valid_targets_min": 1172 + }, + { + "epoch": 3.7848101265822782, + "grad_norm": 0.6241808657000567, + "learning_rate": 2.0661468550743012e-05, + "loss": 0.232, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12715108692646027, + "step": 4935, + "valid_targets_mean": 4138.0, + "valid_targets_min": 1818 + }, + { + "epoch": 3.788645953202915, + "grad_norm": 0.6401174506499515, + "learning_rate": 2.062324613578119e-05, + "loss": 0.2211, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10739579051733017, + "step": 4940, + "valid_targets_mean": 3926.0, + "valid_targets_min": 1535 + }, + { + "epoch": 3.792481779823552, + "grad_norm": 0.6010706891296483, + "learning_rate": 2.0585021442135485e-05, + "loss": 0.2228, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14255794882774353, + "step": 4945, + "valid_targets_mean": 5372.1, + "valid_targets_min": 1814 + }, + { + "epoch": 3.796317606444189, + "grad_norm": 0.6851000989368039, + "learning_rate": 2.054679460956127e-05, + "loss": 0.2188, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12345459312200546, + "step": 4950, + "valid_targets_mean": 2874.2, + "valid_targets_min": 1073 + }, + { + "epoch": 3.8001534330648257, + "grad_norm": 0.6065234459598571, + "learning_rate": 2.0508565777821726e-05, + "loss": 0.2192, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11332584917545319, + "step": 4955, + "valid_targets_mean": 3723.6, + "valid_targets_min": 2420 + }, + { + "epoch": 3.803989259685462, + "grad_norm": 0.6320532823265058, + "learning_rate": 2.0470335086687353e-05, + "loss": 0.2109, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12114453315734863, + "step": 4960, + "valid_targets_mean": 4068.2, + "valid_targets_min": 1079 + }, + { + "epoch": 3.807825086306099, + "grad_norm": 0.6438752744315169, + "learning_rate": 2.0432102675935445e-05, + "loss": 0.2081, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0953713208436966, + "step": 4965, + "valid_targets_mean": 3211.1, + "valid_targets_min": 1143 + }, + { + "epoch": 3.811660912926736, + "grad_norm": 0.6063348153468884, + "learning_rate": 2.039386868534959e-05, + "loss": 0.2184, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09257456660270691, + "step": 4970, + "valid_targets_mean": 3013.6, + "valid_targets_min": 1853 + }, + { + "epoch": 3.8154967395473722, + "grad_norm": 0.6840662563972872, + "learning_rate": 2.035563325471915e-05, + "loss": 0.246, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10907180607318878, + "step": 4975, + "valid_targets_mean": 2993.5, + "valid_targets_min": 1410 + }, + { + "epoch": 3.819332566168009, + "grad_norm": 0.644191986280657, + "learning_rate": 2.0317396523838743e-05, + "loss": 0.2202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0978415235877037, + "step": 4980, + "valid_targets_mean": 2561.1, + "valid_targets_min": 1566 + }, + { + "epoch": 3.823168392788646, + "grad_norm": 0.6455698739452557, + "learning_rate": 2.027915863250775e-05, + "loss": 0.2316, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10561906546354294, + "step": 4985, + "valid_targets_mean": 3800.0, + "valid_targets_min": 2318 + }, + { + "epoch": 3.827004219409283, + "grad_norm": 0.5907883445330249, + "learning_rate": 2.0240919720529796e-05, + "loss": 0.2187, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08875202387571335, + "step": 4990, + "valid_targets_mean": 2651.1, + "valid_targets_min": 1255 + }, + { + "epoch": 3.8308400460299197, + "grad_norm": 0.6102446449407436, + "learning_rate": 2.0202679927712224e-05, + "loss": 0.2251, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13049189746379852, + "step": 4995, + "valid_targets_mean": 4281.9, + "valid_targets_min": 2258 + }, + { + "epoch": 3.834675872650556, + "grad_norm": 0.6451902313846355, + "learning_rate": 2.0164439393865614e-05, + "loss": 0.2202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0990615263581276, + "step": 5000, + "valid_targets_mean": 3365.5, + "valid_targets_min": 1281 + }, + { + "epoch": 3.838511699271193, + "grad_norm": 0.592991931538076, + "learning_rate": 2.012619825880325e-05, + "loss": 0.2204, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10597780346870422, + "step": 5005, + "valid_targets_mean": 3460.4, + "valid_targets_min": 1445 + }, + { + "epoch": 3.84234752589183, + "grad_norm": 0.6411157950712243, + "learning_rate": 2.008795666234061e-05, + "loss": 0.2245, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14322443306446075, + "step": 5010, + "valid_targets_mean": 4763.4, + "valid_targets_min": 3705 + }, + { + "epoch": 3.8461833525124662, + "grad_norm": 0.603696290010457, + "learning_rate": 2.0049714744294865e-05, + "loss": 0.2201, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12432338297367096, + "step": 5015, + "valid_targets_mean": 4469.2, + "valid_targets_min": 2717 + }, + { + "epoch": 3.850019179133103, + "grad_norm": 0.6294669347538074, + "learning_rate": 2.001147264448435e-05, + "loss": 0.2163, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11869427561759949, + "step": 5020, + "valid_targets_mean": 3554.1, + "valid_targets_min": 1508 + }, + { + "epoch": 3.85385500575374, + "grad_norm": 0.6121085008845595, + "learning_rate": 1.9973230502728087e-05, + "loss": 0.2361, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11348491907119751, + "step": 5025, + "valid_targets_mean": 4410.9, + "valid_targets_min": 3504 + }, + { + "epoch": 3.857690832374377, + "grad_norm": 0.56443239974138, + "learning_rate": 1.9934988458845227e-05, + "loss": 0.2316, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11127692461013794, + "step": 5030, + "valid_targets_mean": 4097.1, + "valid_targets_min": 2368 + }, + { + "epoch": 3.8615266589950137, + "grad_norm": 0.5858765835742038, + "learning_rate": 1.9896746652654574e-05, + "loss": 0.2374, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1459936499595642, + "step": 5035, + "valid_targets_mean": 4886.9, + "valid_targets_min": 3446 + }, + { + "epoch": 3.86536248561565, + "grad_norm": 0.6553931287671158, + "learning_rate": 1.985850522397407e-05, + "loss": 0.2365, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11522997915744781, + "step": 5040, + "valid_targets_mean": 3707.2, + "valid_targets_min": 1135 + }, + { + "epoch": 3.869198312236287, + "grad_norm": 0.6544600916189733, + "learning_rate": 1.982026431262026e-05, + "loss": 0.2331, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11998210102319717, + "step": 5045, + "valid_targets_mean": 4419.4, + "valid_targets_min": 1449 + }, + { + "epoch": 3.873034138856924, + "grad_norm": 0.6573602050777325, + "learning_rate": 1.978202405840781e-05, + "loss": 0.2375, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1105203703045845, + "step": 5050, + "valid_targets_mean": 3437.9, + "valid_targets_min": 1400 + }, + { + "epoch": 3.8768699654775602, + "grad_norm": 0.6149527128631458, + "learning_rate": 1.9743784601148988e-05, + "loss": 0.2399, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09850925952196121, + "step": 5055, + "valid_targets_mean": 3713.0, + "valid_targets_min": 1433 + }, + { + "epoch": 3.880705792098197, + "grad_norm": 0.5993790472171153, + "learning_rate": 1.9705546080653128e-05, + "loss": 0.2278, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0955912321805954, + "step": 5060, + "valid_targets_mean": 3649.5, + "valid_targets_min": 614 + }, + { + "epoch": 3.884541618718834, + "grad_norm": 0.6247774269633192, + "learning_rate": 1.966730863672616e-05, + "loss": 0.2286, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10776402056217194, + "step": 5065, + "valid_targets_mean": 3474.1, + "valid_targets_min": 2041 + }, + { + "epoch": 3.8883774453394704, + "grad_norm": 0.6586680346733291, + "learning_rate": 1.9629072409170075e-05, + "loss": 0.2383, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11428436636924744, + "step": 5070, + "valid_targets_mean": 3439.5, + "valid_targets_min": 1333 + }, + { + "epoch": 3.8922132719601072, + "grad_norm": 0.5894934882858609, + "learning_rate": 1.9590837537782394e-05, + "loss": 0.2297, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11921359598636627, + "step": 5075, + "valid_targets_mean": 4520.8, + "valid_targets_min": 1325 + }, + { + "epoch": 3.896049098580744, + "grad_norm": 0.5472575610755923, + "learning_rate": 1.95526041623557e-05, + "loss": 0.2289, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09395705163478851, + "step": 5080, + "valid_targets_mean": 3994.5, + "valid_targets_min": 2754 + }, + { + "epoch": 3.899884925201381, + "grad_norm": 0.6392640770068696, + "learning_rate": 1.9514372422677126e-05, + "loss": 0.217, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1028461828827858, + "step": 5085, + "valid_targets_mean": 3253.2, + "valid_targets_min": 1991 + }, + { + "epoch": 3.903720751822018, + "grad_norm": 0.6442122973188243, + "learning_rate": 1.9476142458527777e-05, + "loss": 0.2333, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13305522501468658, + "step": 5090, + "valid_targets_mean": 3927.8, + "valid_targets_min": 1076 + }, + { + "epoch": 3.9075565784426542, + "grad_norm": 0.5907034904097639, + "learning_rate": 1.94379144096823e-05, + "loss": 0.2117, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10904417932033539, + "step": 5095, + "valid_targets_mean": 4188.8, + "valid_targets_min": 1686 + }, + { + "epoch": 3.911392405063291, + "grad_norm": 0.588404934162781, + "learning_rate": 1.9399688415908333e-05, + "loss": 0.236, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11058816313743591, + "step": 5100, + "valid_targets_mean": 4221.0, + "valid_targets_min": 2185 + }, + { + "epoch": 3.915228231683928, + "grad_norm": 0.5953296498211396, + "learning_rate": 1.9361464616965994e-05, + "loss": 0.2269, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09004606306552887, + "step": 5105, + "valid_targets_mean": 3487.9, + "valid_targets_min": 2024 + }, + { + "epoch": 3.9190640583045644, + "grad_norm": 0.6381018987977479, + "learning_rate": 1.9323243152607388e-05, + "loss": 0.2329, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11940024793148041, + "step": 5110, + "valid_targets_mean": 3816.6, + "valid_targets_min": 1777 + }, + { + "epoch": 3.9228998849252013, + "grad_norm": 0.6240934985428387, + "learning_rate": 1.928502416257606e-05, + "loss": 0.2205, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09721286594867706, + "step": 5115, + "valid_targets_mean": 3629.8, + "valid_targets_min": 1965 + }, + { + "epoch": 3.926735711545838, + "grad_norm": 0.6133390157386737, + "learning_rate": 1.9246807786606534e-05, + "loss": 0.2339, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14093080163002014, + "step": 5120, + "valid_targets_mean": 5056.9, + "valid_targets_min": 2587 + }, + { + "epoch": 3.930571538166475, + "grad_norm": 0.5718039961636495, + "learning_rate": 1.920859416442378e-05, + "loss": 0.227, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11371390521526337, + "step": 5125, + "valid_targets_mean": 4733.9, + "valid_targets_min": 2762 + }, + { + "epoch": 3.934407364787112, + "grad_norm": 0.6424384767362437, + "learning_rate": 1.917038343574267e-05, + "loss": 0.2204, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10284560918807983, + "step": 5130, + "valid_targets_mean": 3052.5, + "valid_targets_min": 1035 + }, + { + "epoch": 3.9382431914077483, + "grad_norm": 0.6137934145318298, + "learning_rate": 1.913217574026752e-05, + "loss": 0.218, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1347580850124359, + "step": 5135, + "valid_targets_mean": 4380.9, + "valid_targets_min": 1876 + }, + { + "epoch": 3.942079018028385, + "grad_norm": 0.5858114438746841, + "learning_rate": 1.909397121769156e-05, + "loss": 0.223, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1097952350974083, + "step": 5140, + "valid_targets_mean": 3833.4, + "valid_targets_min": 2487 + }, + { + "epoch": 3.945914844649022, + "grad_norm": 0.6081372342225959, + "learning_rate": 1.9055770007696398e-05, + "loss": 0.2164, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10154423862695694, + "step": 5145, + "valid_targets_mean": 3927.4, + "valid_targets_min": 1360 + }, + { + "epoch": 3.9497506712696584, + "grad_norm": 0.5835337318788433, + "learning_rate": 1.9017572249951538e-05, + "loss": 0.2141, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11833646893501282, + "step": 5150, + "valid_targets_mean": 4537.6, + "valid_targets_min": 1496 + }, + { + "epoch": 3.9535864978902953, + "grad_norm": 0.6824615483591819, + "learning_rate": 1.8979378084113888e-05, + "loss": 0.2173, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12688958644866943, + "step": 5155, + "valid_targets_mean": 3578.1, + "valid_targets_min": 2059 + }, + { + "epoch": 3.957422324510932, + "grad_norm": 0.5693909638992001, + "learning_rate": 1.894118764982718e-05, + "loss": 0.2168, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09687596559524536, + "step": 5160, + "valid_targets_mean": 3722.0, + "valid_targets_min": 1152 + }, + { + "epoch": 3.961258151131569, + "grad_norm": 0.5929398339415562, + "learning_rate": 1.8903001086721538e-05, + "loss": 0.2261, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09495796263217926, + "step": 5165, + "valid_targets_mean": 3562.4, + "valid_targets_min": 1598 + }, + { + "epoch": 3.965093977752206, + "grad_norm": 0.5911973540452671, + "learning_rate": 1.8864818534412923e-05, + "loss": 0.243, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10482169687747955, + "step": 5170, + "valid_targets_mean": 4045.2, + "valid_targets_min": 2267 + }, + { + "epoch": 3.9689298043728423, + "grad_norm": 0.6468231202545552, + "learning_rate": 1.8826640132502607e-05, + "loss": 0.2147, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10068495571613312, + "step": 5175, + "valid_targets_mean": 3061.8, + "valid_targets_min": 1674 + }, + { + "epoch": 3.972765630993479, + "grad_norm": 0.6266528320519579, + "learning_rate": 1.8788466020576733e-05, + "loss": 0.2184, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10647514462471008, + "step": 5180, + "valid_targets_mean": 3347.2, + "valid_targets_min": 1972 + }, + { + "epoch": 3.976601457614116, + "grad_norm": 0.6023666969325486, + "learning_rate": 1.875029633820572e-05, + "loss": 0.2193, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09071183949708939, + "step": 5185, + "valid_targets_mean": 3000.5, + "valid_targets_min": 1664 + }, + { + "epoch": 3.9804372842347524, + "grad_norm": 0.5503737103727425, + "learning_rate": 1.8712131224943807e-05, + "loss": 0.2202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09134230017662048, + "step": 5190, + "valid_targets_mean": 4733.4, + "valid_targets_min": 3218 + }, + { + "epoch": 3.9842731108553893, + "grad_norm": 0.5987446731298943, + "learning_rate": 1.8673970820328542e-05, + "loss": 0.2315, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12355496734380722, + "step": 5195, + "valid_targets_mean": 3917.2, + "valid_targets_min": 2418 + }, + { + "epoch": 3.988108937476026, + "grad_norm": 0.6105934385307906, + "learning_rate": 1.8635815263880223e-05, + "loss": 0.2278, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12159065157175064, + "step": 5200, + "valid_targets_mean": 4429.4, + "valid_targets_min": 2100 + }, + { + "epoch": 3.991944764096663, + "grad_norm": 0.6062822855323677, + "learning_rate": 1.8597664695101455e-05, + "loss": 0.2257, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11565284430980682, + "step": 5205, + "valid_targets_mean": 3558.9, + "valid_targets_min": 1365 + }, + { + "epoch": 3.9957805907173, + "grad_norm": 0.6533370325847723, + "learning_rate": 1.8559519253476598e-05, + "loss": 0.228, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10582080483436584, + "step": 5210, + "valid_targets_mean": 3915.9, + "valid_targets_min": 1572 + }, + { + "epoch": 3.9996164173379363, + "grad_norm": 0.6353674475744329, + "learning_rate": 1.8521379078471248e-05, + "loss": 0.2322, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1307985931634903, + "step": 5215, + "valid_targets_mean": 4033.1, + "valid_targets_min": 876 + }, + { + "epoch": 4.003068661296509, + "grad_norm": 0.6058337862401332, + "learning_rate": 1.848324430953177e-05, + "loss": 0.1776, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15115858614444733, + "step": 5220, + "valid_targets_mean": 8029.9, + "valid_targets_min": 6142 + }, + { + "epoch": 4.006904487917146, + "grad_norm": 0.4551843447316291, + "learning_rate": 1.8445115086084756e-05, + "loss": 0.1918, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09132374823093414, + "step": 5225, + "valid_targets_mean": 6664.0, + "valid_targets_min": 5470 + }, + { + "epoch": 4.010740314537783, + "grad_norm": 0.4509386389033989, + "learning_rate": 1.8406991547536514e-05, + "loss": 0.1841, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09766881167888641, + "step": 5230, + "valid_targets_mean": 7126.9, + "valid_targets_min": 5368 + }, + { + "epoch": 4.01457614115842, + "grad_norm": 0.4077852725589552, + "learning_rate": 1.8368873833272574e-05, + "loss": 0.177, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0867834985256195, + "step": 5235, + "valid_targets_mean": 7634.9, + "valid_targets_min": 5614 + }, + { + "epoch": 4.018411967779056, + "grad_norm": 0.3970676902252668, + "learning_rate": 1.8330762082657185e-05, + "loss": 0.184, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08743724972009659, + "step": 5240, + "valid_targets_mean": 8630.8, + "valid_targets_min": 6362 + }, + { + "epoch": 4.022247794399693, + "grad_norm": 0.4334472730414226, + "learning_rate": 1.8292656435032762e-05, + "loss": 0.1859, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09218834340572357, + "step": 5245, + "valid_targets_mean": 7383.6, + "valid_targets_min": 5844 + }, + { + "epoch": 4.02608362102033, + "grad_norm": 0.40601448769531323, + "learning_rate": 1.8254557029719424e-05, + "loss": 0.1947, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10436390340328217, + "step": 5250, + "valid_targets_mean": 8017.0, + "valid_targets_min": 5283 + }, + { + "epoch": 4.0299194476409665, + "grad_norm": 0.3990768687192371, + "learning_rate": 1.821646400601448e-05, + "loss": 0.1817, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07754295319318771, + "step": 5255, + "valid_targets_mean": 7051.1, + "valid_targets_min": 4135 + }, + { + "epoch": 4.033755274261603, + "grad_norm": 0.43963293652963353, + "learning_rate": 1.8178377503191875e-05, + "loss": 0.1771, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07584546506404877, + "step": 5260, + "valid_targets_mean": 5370.4, + "valid_targets_min": 3919 + }, + { + "epoch": 4.03759110088224, + "grad_norm": 0.45075694629459007, + "learning_rate": 1.8140297660501745e-05, + "loss": 0.1634, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07687461376190186, + "step": 5265, + "valid_targets_mean": 6063.6, + "valid_targets_min": 4863 + }, + { + "epoch": 4.041426927502877, + "grad_norm": 0.44356443647460936, + "learning_rate": 1.8102224617169855e-05, + "loss": 0.1824, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08697506785392761, + "step": 5270, + "valid_targets_mean": 7001.2, + "valid_targets_min": 6020 + }, + { + "epoch": 4.045262754123514, + "grad_norm": 0.417737233848372, + "learning_rate": 1.8064158512397112e-05, + "loss": 0.1711, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08576381951570511, + "step": 5275, + "valid_targets_mean": 6181.9, + "valid_targets_min": 4669 + }, + { + "epoch": 4.04909858074415, + "grad_norm": 0.39210113545131237, + "learning_rate": 1.802609948535907e-05, + "loss": 0.1777, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07955572009086609, + "step": 5280, + "valid_targets_mean": 6250.2, + "valid_targets_min": 4900 + }, + { + "epoch": 4.052934407364787, + "grad_norm": 0.4158288560039495, + "learning_rate": 1.798804767520539e-05, + "loss": 0.1712, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07730640470981598, + "step": 5285, + "valid_targets_mean": 7156.4, + "valid_targets_min": 5323 + }, + { + "epoch": 4.056770233985424, + "grad_norm": 0.42048984731710154, + "learning_rate": 1.795000322105934e-05, + "loss": 0.1736, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0895126685500145, + "step": 5290, + "valid_targets_mean": 7215.8, + "valid_targets_min": 4720 + }, + { + "epoch": 4.0606060606060606, + "grad_norm": 0.4655383038301388, + "learning_rate": 1.7911966262017327e-05, + "loss": 0.1798, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08802001178264618, + "step": 5295, + "valid_targets_mean": 6550.8, + "valid_targets_min": 5739 + }, + { + "epoch": 4.064441887226697, + "grad_norm": 0.4287356034658328, + "learning_rate": 1.7873936937148312e-05, + "loss": 0.172, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08902718126773834, + "step": 5300, + "valid_targets_mean": 6726.8, + "valid_targets_min": 4540 + }, + { + "epoch": 4.068277713847334, + "grad_norm": 0.5243827493237017, + "learning_rate": 1.783591538549338e-05, + "loss": 0.1794, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09830628335475922, + "step": 5305, + "valid_targets_mean": 4573.4, + "valid_targets_min": 1039 + }, + { + "epoch": 4.072113540467971, + "grad_norm": 0.4635996610367619, + "learning_rate": 1.779790174606518e-05, + "loss": 0.1915, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09317825734615326, + "step": 5310, + "valid_targets_mean": 6430.0, + "valid_targets_min": 5412 + }, + { + "epoch": 4.075949367088608, + "grad_norm": 0.706027938625422, + "learning_rate": 1.775989615784742e-05, + "loss": 0.1728, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.027076993137598038, + "step": 5315, + "valid_targets_mean": 583.1, + "valid_targets_min": 158 + }, + { + "epoch": 4.079785193709244, + "grad_norm": 0.3912016261675463, + "learning_rate": 1.77218987597944e-05, + "loss": 0.1618, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09083262085914612, + "step": 5320, + "valid_targets_mean": 8230.9, + "valid_targets_min": 5061 + }, + { + "epoch": 4.083621020329881, + "grad_norm": 0.4635966759984746, + "learning_rate": 1.768390969083046e-05, + "loss": 0.1743, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09498058259487152, + "step": 5325, + "valid_targets_mean": 7238.9, + "valid_targets_min": 5112 + }, + { + "epoch": 4.087456846950518, + "grad_norm": 0.43749273124406546, + "learning_rate": 1.7645929089849476e-05, + "loss": 0.1715, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07731843739748001, + "step": 5330, + "valid_targets_mean": 6322.8, + "valid_targets_min": 4867 + }, + { + "epoch": 4.091292673571155, + "grad_norm": 0.4366899759564789, + "learning_rate": 1.760795709571438e-05, + "loss": 0.1803, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0873725414276123, + "step": 5335, + "valid_targets_mean": 7887.0, + "valid_targets_min": 6004 + }, + { + "epoch": 4.095128500191791, + "grad_norm": 0.4028799611796367, + "learning_rate": 1.756999384725665e-05, + "loss": 0.1701, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0781828910112381, + "step": 5340, + "valid_targets_mean": 7090.2, + "valid_targets_min": 4917 + }, + { + "epoch": 4.098964326812428, + "grad_norm": 0.41311248645436904, + "learning_rate": 1.753203948327575e-05, + "loss": 0.1645, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07959892600774765, + "step": 5345, + "valid_targets_mean": 7441.9, + "valid_targets_min": 4202 + }, + { + "epoch": 4.102800153433065, + "grad_norm": 0.4116716580123711, + "learning_rate": 1.7494094142538697e-05, + "loss": 0.1528, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08022089302539825, + "step": 5350, + "valid_targets_mean": 7901.8, + "valid_targets_min": 5060 + }, + { + "epoch": 4.106635980053702, + "grad_norm": 0.4323415326039323, + "learning_rate": 1.7456157963779478e-05, + "loss": 0.1539, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07285164296627045, + "step": 5355, + "valid_targets_mean": 8557.6, + "valid_targets_min": 6331 + }, + { + "epoch": 4.110471806674338, + "grad_norm": 0.5004205921213226, + "learning_rate": 1.741823108569863e-05, + "loss": 0.1805, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08340825885534286, + "step": 5360, + "valid_targets_mean": 4976.4, + "valid_targets_min": 2847 + }, + { + "epoch": 4.114307633294975, + "grad_norm": 0.48475367964075106, + "learning_rate": 1.738031364696266e-05, + "loss": 0.1628, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09691165387630463, + "step": 5365, + "valid_targets_mean": 7129.1, + "valid_targets_min": 5331 + }, + { + "epoch": 4.118143459915612, + "grad_norm": 0.4602347916265738, + "learning_rate": 1.7342405786203546e-05, + "loss": 0.182, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09371896833181381, + "step": 5370, + "valid_targets_mean": 6864.0, + "valid_targets_min": 4697 + }, + { + "epoch": 4.121979286536249, + "grad_norm": 0.45369079402392004, + "learning_rate": 1.7304507642018285e-05, + "loss": 0.188, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08609636127948761, + "step": 5375, + "valid_targets_mean": 6631.9, + "valid_targets_min": 4623 + }, + { + "epoch": 4.125815113156885, + "grad_norm": 0.42975007869192766, + "learning_rate": 1.726661935296833e-05, + "loss": 0.1808, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08153565227985382, + "step": 5380, + "valid_targets_mean": 6330.4, + "valid_targets_min": 5424 + }, + { + "epoch": 4.129650939777522, + "grad_norm": 0.4316430009354068, + "learning_rate": 1.7228741057579096e-05, + "loss": 0.1745, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0833243727684021, + "step": 5385, + "valid_targets_mean": 6814.8, + "valid_targets_min": 4964 + }, + { + "epoch": 4.133486766398159, + "grad_norm": 0.484245997162868, + "learning_rate": 1.7190872894339465e-05, + "loss": 0.1828, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09648758172988892, + "step": 5390, + "valid_targets_mean": 6915.8, + "valid_targets_min": 5321 + }, + { + "epoch": 4.137322593018795, + "grad_norm": 0.42270898002597446, + "learning_rate": 1.7153015001701284e-05, + "loss": 0.1758, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08905856311321259, + "step": 5395, + "valid_targets_mean": 6464.5, + "valid_targets_min": 5185 + }, + { + "epoch": 4.141158419639432, + "grad_norm": 0.4131814652804404, + "learning_rate": 1.711516751807883e-05, + "loss": 0.1786, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09088905900716782, + "step": 5400, + "valid_targets_mean": 7048.6, + "valid_targets_min": 4913 + }, + { + "epoch": 4.144994246260069, + "grad_norm": 0.43285987807803006, + "learning_rate": 1.7077330581848336e-05, + "loss": 0.1811, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09897810965776443, + "step": 5405, + "valid_targets_mean": 7469.2, + "valid_targets_min": 5426 + }, + { + "epoch": 4.148830072880706, + "grad_norm": 0.4345370232370801, + "learning_rate": 1.7039504331347477e-05, + "loss": 0.1815, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09374260902404785, + "step": 5410, + "valid_targets_mean": 6819.4, + "valid_targets_min": 5286 + }, + { + "epoch": 4.152665899501343, + "grad_norm": 0.39965664719961697, + "learning_rate": 1.700168890487484e-05, + "loss": 0.1792, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09366252273321152, + "step": 5415, + "valid_targets_mean": 7107.5, + "valid_targets_min": 5201 + }, + { + "epoch": 4.156501726121979, + "grad_norm": 0.4463529078921578, + "learning_rate": 1.6963884440689462e-05, + "loss": 0.1763, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0858970507979393, + "step": 5420, + "valid_targets_mean": 5642.6, + "valid_targets_min": 4838 + }, + { + "epoch": 4.160337552742616, + "grad_norm": 0.8759751774268991, + "learning_rate": 1.6926091077010287e-05, + "loss": 0.1463, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14921234548091888, + "step": 5425, + "valid_targets_mean": 2054.1, + "valid_targets_min": 134 + }, + { + "epoch": 4.164173379363253, + "grad_norm": 0.46266680247957004, + "learning_rate": 1.6888308952015658e-05, + "loss": 0.1798, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08233311772346497, + "step": 5430, + "valid_targets_mean": 6686.4, + "valid_targets_min": 5256 + }, + { + "epoch": 4.168009205983889, + "grad_norm": 0.42912723839675276, + "learning_rate": 1.685053820384287e-05, + "loss": 0.1732, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09341243654489517, + "step": 5435, + "valid_targets_mean": 6216.6, + "valid_targets_min": 5542 + }, + { + "epoch": 4.171845032604526, + "grad_norm": 0.4105054806697354, + "learning_rate": 1.6812778970587572e-05, + "loss": 0.1762, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08716951310634613, + "step": 5440, + "valid_targets_mean": 6486.1, + "valid_targets_min": 5531 + }, + { + "epoch": 4.175680859225163, + "grad_norm": 0.46792934978881817, + "learning_rate": 1.6775031390303352e-05, + "loss": 0.173, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08692094683647156, + "step": 5445, + "valid_targets_mean": 5797.4, + "valid_targets_min": 3911 + }, + { + "epoch": 4.1795166858458, + "grad_norm": 0.41856613665133585, + "learning_rate": 1.6737295601001182e-05, + "loss": 0.1722, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0881090983748436, + "step": 5450, + "valid_targets_mean": 6357.5, + "valid_targets_min": 5195 + }, + { + "epoch": 4.183352512466437, + "grad_norm": 0.4480211686300025, + "learning_rate": 1.6699571740648905e-05, + "loss": 0.1804, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0891503393650055, + "step": 5455, + "valid_targets_mean": 5808.9, + "valid_targets_min": 4959 + }, + { + "epoch": 4.187188339087073, + "grad_norm": 0.5521202176163402, + "learning_rate": 1.6661859947170776e-05, + "loss": 0.1755, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05632351338863373, + "step": 5460, + "valid_targets_mean": 2591.5, + "valid_targets_min": 842 + }, + { + "epoch": 4.19102416570771, + "grad_norm": 0.4677132689601809, + "learning_rate": 1.6624160358446928e-05, + "loss": 0.146, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08426553010940552, + "step": 5465, + "valid_targets_mean": 6006.9, + "valid_targets_min": 4769 + }, + { + "epoch": 4.194859992328347, + "grad_norm": 0.4385859712571102, + "learning_rate": 1.658647311231285e-05, + "loss": 0.171, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08442874997854233, + "step": 5470, + "valid_targets_mean": 6764.2, + "valid_targets_min": 5046 + }, + { + "epoch": 4.198695818948983, + "grad_norm": 0.4894127756053233, + "learning_rate": 1.6548798346558924e-05, + "loss": 0.1777, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09291423112154007, + "step": 5475, + "valid_targets_mean": 6055.8, + "valid_targets_min": 5182 + }, + { + "epoch": 4.2025316455696204, + "grad_norm": 0.39815622890266067, + "learning_rate": 1.651113619892991e-05, + "loss": 0.1718, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0855388194322586, + "step": 5480, + "valid_targets_mean": 7101.6, + "valid_targets_min": 5366 + }, + { + "epoch": 4.206367472190257, + "grad_norm": 0.4111676623356184, + "learning_rate": 1.647348680712441e-05, + "loss": 0.1682, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08801939338445663, + "step": 5485, + "valid_targets_mean": 7111.9, + "valid_targets_min": 5192 + }, + { + "epoch": 4.210203298810894, + "grad_norm": 0.4502327012455291, + "learning_rate": 1.6435850308794395e-05, + "loss": 0.174, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08616534620523453, + "step": 5490, + "valid_targets_mean": 5509.9, + "valid_targets_min": 4790 + }, + { + "epoch": 4.214039125431531, + "grad_norm": 0.9160425192063995, + "learning_rate": 1.639822684154472e-05, + "loss": 0.1717, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09516707807779312, + "step": 5495, + "valid_targets_mean": 1581.9, + "valid_targets_min": 864 + }, + { + "epoch": 4.217874952052167, + "grad_norm": 0.7840081042666655, + "learning_rate": 1.636061654293256e-05, + "loss": 0.1709, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07739853858947754, + "step": 5500, + "valid_targets_mean": 1333.2, + "valid_targets_min": 632 + }, + { + "epoch": 4.221710778672804, + "grad_norm": 0.972187128223758, + "learning_rate": 1.6323019550466958e-05, + "loss": 0.1569, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08127404749393463, + "step": 5505, + "valid_targets_mean": 1602.9, + "valid_targets_min": 875 + }, + { + "epoch": 4.225546605293441, + "grad_norm": 0.9860079943539749, + "learning_rate": 1.628543600160833e-05, + "loss": 0.1565, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08117218315601349, + "step": 5510, + "valid_targets_mean": 1340.4, + "valid_targets_min": 701 + }, + { + "epoch": 4.229382431914077, + "grad_norm": 0.8295045121721946, + "learning_rate": 1.62478660337679e-05, + "loss": 0.1494, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09056619554758072, + "step": 5515, + "valid_targets_mean": 1792.9, + "valid_targets_min": 898 + }, + { + "epoch": 4.2332182585347145, + "grad_norm": 0.8188032487720853, + "learning_rate": 1.6210309784307275e-05, + "loss": 0.1516, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07808853685855865, + "step": 5520, + "valid_targets_mean": 1433.1, + "valid_targets_min": 1070 + }, + { + "epoch": 4.237054085155351, + "grad_norm": 0.8097455240210127, + "learning_rate": 1.6172767390537874e-05, + "loss": 0.1581, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07835352420806885, + "step": 5525, + "valid_targets_mean": 1589.5, + "valid_targets_min": 970 + }, + { + "epoch": 4.240889911775987, + "grad_norm": 0.8538126262264486, + "learning_rate": 1.613523898972048e-05, + "loss": 0.1401, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0681949332356453, + "step": 5530, + "valid_targets_mean": 1270.5, + "valid_targets_min": 648 + }, + { + "epoch": 4.244725738396625, + "grad_norm": 0.8043233543715164, + "learning_rate": 1.6097724719064706e-05, + "loss": 0.1619, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08543618023395538, + "step": 5535, + "valid_targets_mean": 1674.5, + "valid_targets_min": 727 + }, + { + "epoch": 4.248561565017261, + "grad_norm": 0.7989244580438366, + "learning_rate": 1.6060224715728507e-05, + "loss": 0.1435, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05895886570215225, + "step": 5540, + "valid_targets_mean": 1406.0, + "valid_targets_min": 770 + }, + { + "epoch": 4.252397391637898, + "grad_norm": 0.8481825882151971, + "learning_rate": 1.602273911681766e-05, + "loss": 0.1519, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06739223748445511, + "step": 5545, + "valid_targets_mean": 1381.2, + "valid_targets_min": 572 + }, + { + "epoch": 4.256233218258535, + "grad_norm": 0.8043165861790226, + "learning_rate": 1.5985268059385314e-05, + "loss": 0.1512, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08851317316293716, + "step": 5550, + "valid_targets_mean": 1580.4, + "valid_targets_min": 633 + }, + { + "epoch": 4.260069044879171, + "grad_norm": 0.8083389984804482, + "learning_rate": 1.59478116804314e-05, + "loss": 0.1463, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07341709733009338, + "step": 5555, + "valid_targets_mean": 1478.8, + "valid_targets_min": 923 + }, + { + "epoch": 4.2639048714998085, + "grad_norm": 1.3965762892954492, + "learning_rate": 1.5910370116902223e-05, + "loss": 0.1381, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06662610173225403, + "step": 5560, + "valid_targets_mean": 1401.0, + "valid_targets_min": 740 + }, + { + "epoch": 4.267740698120445, + "grad_norm": 0.8020508948272961, + "learning_rate": 1.5872943505689914e-05, + "loss": 0.1411, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06651642173528671, + "step": 5565, + "valid_targets_mean": 1588.9, + "valid_targets_min": 870 + }, + { + "epoch": 4.271576524741081, + "grad_norm": 0.7927719862549989, + "learning_rate": 1.583553198363191e-05, + "loss": 0.1402, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10171519964933395, + "step": 5570, + "valid_targets_mean": 2008.4, + "valid_targets_min": 1042 + }, + { + "epoch": 4.275412351361719, + "grad_norm": 0.8052696382112644, + "learning_rate": 1.5798135687510517e-05, + "loss": 0.1461, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.058815598487854004, + "step": 5575, + "valid_targets_mean": 1209.0, + "valid_targets_min": 870 + }, + { + "epoch": 4.279248177982355, + "grad_norm": 0.8668675221829967, + "learning_rate": 1.576075475405235e-05, + "loss": 0.1528, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08240042626857758, + "step": 5580, + "valid_targets_mean": 1535.5, + "valid_targets_min": 740 + }, + { + "epoch": 4.283084004602992, + "grad_norm": 0.7093566528346401, + "learning_rate": 1.572338931992785e-05, + "loss": 0.1519, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07407762110233307, + "step": 5585, + "valid_targets_mean": 1642.9, + "valid_targets_min": 906 + }, + { + "epoch": 4.286919831223629, + "grad_norm": 0.8329087926054658, + "learning_rate": 1.5686039521750817e-05, + "loss": 0.1445, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.077165387570858, + "step": 5590, + "valid_targets_mean": 1570.5, + "valid_targets_min": 814 + }, + { + "epoch": 4.290755657844265, + "grad_norm": 0.7460786118812017, + "learning_rate": 1.5648705496077848e-05, + "loss": 0.1477, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0657590925693512, + "step": 5595, + "valid_targets_mean": 1722.1, + "valid_targets_min": 832 + }, + { + "epoch": 4.2945914844649025, + "grad_norm": 0.7643321265051494, + "learning_rate": 1.5611387379407906e-05, + "loss": 0.1479, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06141894310712814, + "step": 5600, + "valid_targets_mean": 1453.5, + "valid_targets_min": 891 + }, + { + "epoch": 4.298427311085539, + "grad_norm": 0.8552171061288394, + "learning_rate": 1.5574085308181777e-05, + "loss": 0.1454, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08809889107942581, + "step": 5605, + "valid_targets_mean": 1584.4, + "valid_targets_min": 734 + }, + { + "epoch": 4.302263137706175, + "grad_norm": 0.7920494378598284, + "learning_rate": 1.5536799418781563e-05, + "loss": 0.136, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06714694201946259, + "step": 5610, + "valid_targets_mean": 1495.2, + "valid_targets_min": 737 + }, + { + "epoch": 4.306098964326813, + "grad_norm": 0.819406426611117, + "learning_rate": 1.5499529847530236e-05, + "loss": 0.1425, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07912206649780273, + "step": 5615, + "valid_targets_mean": 1737.0, + "valid_targets_min": 872 + }, + { + "epoch": 4.309934790947449, + "grad_norm": 0.8583216727231345, + "learning_rate": 1.5462276730691085e-05, + "loss": 0.139, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07209743559360504, + "step": 5620, + "valid_targets_mean": 1431.6, + "valid_targets_min": 805 + }, + { + "epoch": 4.313770617568086, + "grad_norm": 0.7904007596665792, + "learning_rate": 1.5425040204467236e-05, + "loss": 0.1469, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08445294201374054, + "step": 5625, + "valid_targets_mean": 2060.2, + "valid_targets_min": 1253 + }, + { + "epoch": 4.317606444188723, + "grad_norm": 0.8163079256722341, + "learning_rate": 1.5387820405001174e-05, + "loss": 0.1498, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08525355160236359, + "step": 5630, + "valid_targets_mean": 2063.9, + "valid_targets_min": 1264 + }, + { + "epoch": 4.321442270809359, + "grad_norm": 0.7598390758316347, + "learning_rate": 1.5350617468374222e-05, + "loss": 0.1405, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.053884390741586685, + "step": 5635, + "valid_targets_mean": 1036.5, + "valid_targets_min": 613 + }, + { + "epoch": 4.3252780974299965, + "grad_norm": 0.7533073132434397, + "learning_rate": 1.531343153060604e-05, + "loss": 0.1419, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0684368684887886, + "step": 5640, + "valid_targets_mean": 1674.4, + "valid_targets_min": 768 + }, + { + "epoch": 4.329113924050633, + "grad_norm": 0.8612838807659969, + "learning_rate": 1.5276262727654144e-05, + "loss": 0.1406, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06439299881458282, + "step": 5645, + "valid_targets_mean": 1249.8, + "valid_targets_min": 714 + }, + { + "epoch": 4.332949750671269, + "grad_norm": 0.8176929086254164, + "learning_rate": 1.5239111195413416e-05, + "loss": 0.1478, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05252016335725784, + "step": 5650, + "valid_targets_mean": 1114.8, + "valid_targets_min": 800 + }, + { + "epoch": 4.336785577291907, + "grad_norm": 0.7814309742921794, + "learning_rate": 1.5201977069715568e-05, + "loss": 0.14, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08001722395420074, + "step": 5655, + "valid_targets_mean": 1780.9, + "valid_targets_min": 933 + }, + { + "epoch": 4.340621403912543, + "grad_norm": 0.8855366491744728, + "learning_rate": 1.5164860486328683e-05, + "loss": 0.1382, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.061440110206604004, + "step": 5660, + "valid_targets_mean": 1113.6, + "valid_targets_min": 746 + }, + { + "epoch": 4.34445723053318, + "grad_norm": 0.8689469294253976, + "learning_rate": 1.5127761580956723e-05, + "loss": 0.1305, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07270821928977966, + "step": 5665, + "valid_targets_mean": 1265.0, + "valid_targets_min": 814 + }, + { + "epoch": 4.348293057153817, + "grad_norm": 0.7995843103768819, + "learning_rate": 1.5090680489238983e-05, + "loss": 0.1426, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08606080710887909, + "step": 5670, + "valid_targets_mean": 1652.1, + "valid_targets_min": 635 + }, + { + "epoch": 4.352128883774453, + "grad_norm": 0.8806634248757328, + "learning_rate": 1.505361734674966e-05, + "loss": 0.139, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.056605733931064606, + "step": 5675, + "valid_targets_mean": 1231.2, + "valid_targets_min": 739 + }, + { + "epoch": 4.3559647103950905, + "grad_norm": 0.8916670001292972, + "learning_rate": 1.5016572288997306e-05, + "loss": 0.1408, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08641964197158813, + "step": 5680, + "valid_targets_mean": 1682.4, + "valid_targets_min": 831 + }, + { + "epoch": 4.359800537015727, + "grad_norm": 0.8217303072113938, + "learning_rate": 1.4979545451424357e-05, + "loss": 0.1433, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0690346211194992, + "step": 5685, + "valid_targets_mean": 1571.6, + "valid_targets_min": 1015 + }, + { + "epoch": 4.363636363636363, + "grad_norm": 3.019320664265735, + "learning_rate": 1.4942536969406645e-05, + "loss": 0.1391, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06208442896604538, + "step": 5690, + "valid_targets_mean": 1276.8, + "valid_targets_min": 711 + }, + { + "epoch": 4.367472190257001, + "grad_norm": 0.7631117824513171, + "learning_rate": 1.4905546978252865e-05, + "loss": 0.1384, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04727751389145851, + "step": 5695, + "valid_targets_mean": 1147.5, + "valid_targets_min": 733 + }, + { + "epoch": 4.371308016877637, + "grad_norm": 0.8633614828164144, + "learning_rate": 1.4868575613204139e-05, + "loss": 0.1437, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08766709268093109, + "step": 5700, + "valid_targets_mean": 1604.0, + "valid_targets_min": 953 + }, + { + "epoch": 4.3751438434982735, + "grad_norm": 0.7434481494185267, + "learning_rate": 1.4831623009433474e-05, + "loss": 0.1497, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07481752336025238, + "step": 5705, + "valid_targets_mean": 1784.0, + "valid_targets_min": 1011 + }, + { + "epoch": 4.378979670118911, + "grad_norm": 0.8159454679359223, + "learning_rate": 1.4794689302045262e-05, + "loss": 0.1519, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08777831494808197, + "step": 5710, + "valid_targets_mean": 1729.0, + "valid_targets_min": 1143 + }, + { + "epoch": 4.382815496739547, + "grad_norm": 0.8371323240824395, + "learning_rate": 1.4757774626074843e-05, + "loss": 0.1358, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06284742802381516, + "step": 5715, + "valid_targets_mean": 1483.9, + "valid_targets_min": 699 + }, + { + "epoch": 4.3866513233601845, + "grad_norm": 0.8878169906033956, + "learning_rate": 1.4720879116487963e-05, + "loss": 0.134, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06304629147052765, + "step": 5720, + "valid_targets_mean": 1198.4, + "valid_targets_min": 680 + }, + { + "epoch": 4.390487149980821, + "grad_norm": 0.8549460000205954, + "learning_rate": 1.4684002908180272e-05, + "loss": 0.1501, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07019459456205368, + "step": 5725, + "valid_targets_mean": 1365.2, + "valid_targets_min": 687 + }, + { + "epoch": 4.394322976601457, + "grad_norm": 0.8607283116541214, + "learning_rate": 1.4647146135976877e-05, + "loss": 0.1405, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0521932877600193, + "step": 5730, + "valid_targets_mean": 1142.9, + "valid_targets_min": 835 + }, + { + "epoch": 4.398158803222095, + "grad_norm": 0.857967798929198, + "learning_rate": 1.4610308934631827e-05, + "loss": 0.1308, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06068773195147514, + "step": 5735, + "valid_targets_mean": 1316.9, + "valid_targets_min": 829 + }, + { + "epoch": 4.401994629842731, + "grad_norm": 0.8141306870922265, + "learning_rate": 1.457349143882759e-05, + "loss": 0.1311, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07377788424491882, + "step": 5740, + "valid_targets_mean": 1739.2, + "valid_targets_min": 1138 + }, + { + "epoch": 4.4058304564633675, + "grad_norm": 0.8362471779565951, + "learning_rate": 1.4536693783174598e-05, + "loss": 0.1461, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06843001395463943, + "step": 5745, + "valid_targets_mean": 1592.2, + "valid_targets_min": 1174 + }, + { + "epoch": 4.409666283084005, + "grad_norm": 0.8214115389490098, + "learning_rate": 1.4499916102210766e-05, + "loss": 0.1429, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07091072201728821, + "step": 5750, + "valid_targets_mean": 1543.1, + "valid_targets_min": 554 + }, + { + "epoch": 4.413502109704641, + "grad_norm": 0.7854661090038673, + "learning_rate": 1.446315853040095e-05, + "loss": 0.1358, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06648841500282288, + "step": 5755, + "valid_targets_mean": 1430.0, + "valid_targets_min": 843 + }, + { + "epoch": 4.4173379363252785, + "grad_norm": 0.7119657234769702, + "learning_rate": 1.44264212021365e-05, + "loss": 0.1387, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0674772709608078, + "step": 5760, + "valid_targets_mean": 1622.9, + "valid_targets_min": 849 + }, + { + "epoch": 4.421173762945915, + "grad_norm": 0.8172390055359992, + "learning_rate": 1.4389704251734736e-05, + "loss": 0.1413, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.061807781457901, + "step": 5765, + "valid_targets_mean": 1429.0, + "valid_targets_min": 612 + }, + { + "epoch": 4.425009589566551, + "grad_norm": 0.8536655861113474, + "learning_rate": 1.4353007813438493e-05, + "loss": 0.1408, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.059383559972047806, + "step": 5770, + "valid_targets_mean": 1283.0, + "valid_targets_min": 782 + }, + { + "epoch": 4.428845416187189, + "grad_norm": 0.9060942049577573, + "learning_rate": 1.4316332021415613e-05, + "loss": 0.137, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07868949323892593, + "step": 5775, + "valid_targets_mean": 1457.0, + "valid_targets_min": 926 + }, + { + "epoch": 4.432681242807825, + "grad_norm": 0.7836300349637155, + "learning_rate": 1.4279677009758426e-05, + "loss": 0.1368, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.058946643024683, + "step": 5780, + "valid_targets_mean": 1399.6, + "valid_targets_min": 822 + }, + { + "epoch": 4.4365170694284615, + "grad_norm": 0.8238735384249218, + "learning_rate": 1.4243042912483302e-05, + "loss": 0.1302, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06600479781627655, + "step": 5785, + "valid_targets_mean": 1464.6, + "valid_targets_min": 670 + }, + { + "epoch": 4.440352896049099, + "grad_norm": 0.9236998932363395, + "learning_rate": 1.420642986353016e-05, + "loss": 0.143, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06835651397705078, + "step": 5790, + "valid_targets_mean": 1504.8, + "valid_targets_min": 647 + }, + { + "epoch": 4.444188722669735, + "grad_norm": 0.7689711132831754, + "learning_rate": 1.416983799676193e-05, + "loss": 0.1411, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05333807319402695, + "step": 5795, + "valid_targets_mean": 1276.9, + "valid_targets_min": 791 + }, + { + "epoch": 4.448024549290372, + "grad_norm": 0.8897342668485723, + "learning_rate": 1.4133267445964118e-05, + "loss": 0.1382, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06487174332141876, + "step": 5800, + "valid_targets_mean": 1291.4, + "valid_targets_min": 746 + }, + { + "epoch": 4.451860375911009, + "grad_norm": 0.8189115449820924, + "learning_rate": 1.40967183448443e-05, + "loss": 0.1403, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0557759664952755, + "step": 5805, + "valid_targets_mean": 1409.1, + "valid_targets_min": 892 + }, + { + "epoch": 4.455696202531645, + "grad_norm": 0.8980745708913338, + "learning_rate": 1.4060190827031605e-05, + "loss": 0.1457, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08061063289642334, + "step": 5810, + "valid_targets_mean": 1581.9, + "valid_targets_min": 914 + }, + { + "epoch": 4.459532029152283, + "grad_norm": 1.0352402920820307, + "learning_rate": 1.4023685026076278e-05, + "loss": 0.1391, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06457255780696869, + "step": 5815, + "valid_targets_mean": 1647.6, + "valid_targets_min": 636 + }, + { + "epoch": 4.463367855772919, + "grad_norm": 0.8521532712254869, + "learning_rate": 1.3987201075449153e-05, + "loss": 0.1356, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06841334700584412, + "step": 5820, + "valid_targets_mean": 1523.5, + "valid_targets_min": 641 + }, + { + "epoch": 4.4672036823935555, + "grad_norm": 0.9381538067175157, + "learning_rate": 1.3950739108541158e-05, + "loss": 0.1387, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07897339016199112, + "step": 5825, + "valid_targets_mean": 1629.4, + "valid_targets_min": 1020 + }, + { + "epoch": 4.471039509014193, + "grad_norm": 0.9072674213175376, + "learning_rate": 1.3914299258662873e-05, + "loss": 0.1377, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.062225092202425, + "step": 5830, + "valid_targets_mean": 1163.4, + "valid_targets_min": 729 + }, + { + "epoch": 4.474875335634829, + "grad_norm": 0.8803524336857421, + "learning_rate": 1.3877881659044006e-05, + "loss": 0.1447, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08284099400043488, + "step": 5835, + "valid_targets_mean": 1613.2, + "valid_targets_min": 1001 + }, + { + "epoch": 4.4787111622554665, + "grad_norm": 0.84628530466705, + "learning_rate": 1.3841486442832897e-05, + "loss": 0.1462, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07279478758573532, + "step": 5840, + "valid_targets_mean": 1567.0, + "valid_targets_min": 1175 + }, + { + "epoch": 4.482546988876103, + "grad_norm": 0.9057488508986264, + "learning_rate": 1.3805113743096082e-05, + "loss": 0.1428, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08599331229925156, + "step": 5845, + "valid_targets_mean": 1609.0, + "valid_targets_min": 892 + }, + { + "epoch": 4.486382815496739, + "grad_norm": 0.8724765767469137, + "learning_rate": 1.3768763692817734e-05, + "loss": 0.1433, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06701600551605225, + "step": 5850, + "valid_targets_mean": 1425.2, + "valid_targets_min": 801 + }, + { + "epoch": 4.490218642117377, + "grad_norm": 0.7760747180119547, + "learning_rate": 1.373243642489925e-05, + "loss": 0.1317, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06627722084522247, + "step": 5855, + "valid_targets_mean": 1462.4, + "valid_targets_min": 564 + }, + { + "epoch": 4.494054468738013, + "grad_norm": 0.8976577094010686, + "learning_rate": 1.3696132072158722e-05, + "loss": 0.1331, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06739483773708344, + "step": 5860, + "valid_targets_mean": 1749.1, + "valid_targets_min": 651 + }, + { + "epoch": 4.4978902953586495, + "grad_norm": 0.8545567972439262, + "learning_rate": 1.3659850767330438e-05, + "loss": 0.1401, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.058055490255355835, + "step": 5865, + "valid_targets_mean": 1161.0, + "valid_targets_min": 716 + }, + { + "epoch": 4.501726121979287, + "grad_norm": 0.8095380345936124, + "learning_rate": 1.3623592643064453e-05, + "loss": 0.134, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06450197845697403, + "step": 5870, + "valid_targets_mean": 1425.6, + "valid_targets_min": 744 + }, + { + "epoch": 4.505561948599923, + "grad_norm": 0.8185662947125141, + "learning_rate": 1.3587357831926049e-05, + "loss": 0.1302, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0603010319173336, + "step": 5875, + "valid_targets_mean": 1391.1, + "valid_targets_min": 1218 + }, + { + "epoch": 4.50939777522056, + "grad_norm": 0.8979998333312221, + "learning_rate": 1.3551146466395283e-05, + "loss": 0.1388, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08387155830860138, + "step": 5880, + "valid_targets_mean": 1724.5, + "valid_targets_min": 793 + }, + { + "epoch": 4.513233601841197, + "grad_norm": 0.8055951960435668, + "learning_rate": 1.3514958678866472e-05, + "loss": 0.1468, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06271986663341522, + "step": 5885, + "valid_targets_mean": 1366.8, + "valid_targets_min": 846 + }, + { + "epoch": 4.517069428461833, + "grad_norm": 0.8041904078077861, + "learning_rate": 1.3478794601647764e-05, + "loss": 0.1313, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07775819301605225, + "step": 5890, + "valid_targets_mean": 1772.1, + "valid_targets_min": 979 + }, + { + "epoch": 4.520905255082471, + "grad_norm": 0.8084368235348921, + "learning_rate": 1.3442654366960577e-05, + "loss": 0.1373, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0741063728928566, + "step": 5895, + "valid_targets_mean": 1778.9, + "valid_targets_min": 1093 + }, + { + "epoch": 4.524741081703107, + "grad_norm": 0.7894751444701532, + "learning_rate": 1.340653810693918e-05, + "loss": 0.1225, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06515541672706604, + "step": 5900, + "valid_targets_mean": 1510.8, + "valid_targets_min": 595 + }, + { + "epoch": 4.5285769083237435, + "grad_norm": 0.9010256150475929, + "learning_rate": 1.3370445953630195e-05, + "loss": 0.1337, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06379692256450653, + "step": 5905, + "valid_targets_mean": 1330.4, + "valid_targets_min": 868 + }, + { + "epoch": 4.532412734944381, + "grad_norm": 0.8333545656796055, + "learning_rate": 1.3334378038992078e-05, + "loss": 0.1331, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07732553035020828, + "step": 5910, + "valid_targets_mean": 1504.2, + "valid_targets_min": 914 + }, + { + "epoch": 4.536248561565017, + "grad_norm": 0.7842207592803885, + "learning_rate": 1.329833449489469e-05, + "loss": 0.1362, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0766078531742096, + "step": 5915, + "valid_targets_mean": 1972.9, + "valid_targets_min": 1080 + }, + { + "epoch": 4.540084388185654, + "grad_norm": 0.84222805835276, + "learning_rate": 1.326231545311879e-05, + "loss": 0.1385, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07629090547561646, + "step": 5920, + "valid_targets_mean": 1519.6, + "valid_targets_min": 658 + }, + { + "epoch": 4.543920214806291, + "grad_norm": 0.9542050786373134, + "learning_rate": 1.3226321045355524e-05, + "loss": 0.1392, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06195371598005295, + "step": 5925, + "valid_targets_mean": 1362.0, + "valid_targets_min": 1018 + }, + { + "epoch": 4.547756041426927, + "grad_norm": 0.7826948638625261, + "learning_rate": 1.3190351403206005e-05, + "loss": 0.1386, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07496462762355804, + "step": 5930, + "valid_targets_mean": 1623.0, + "valid_targets_min": 1041 + }, + { + "epoch": 4.551591868047565, + "grad_norm": 1.1856200675146584, + "learning_rate": 1.3154406658180782e-05, + "loss": 0.1722, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17681968212127686, + "step": 5935, + "valid_targets_mean": 5392.9, + "valid_targets_min": 2506 + }, + { + "epoch": 4.555427694668201, + "grad_norm": 0.7523898519471789, + "learning_rate": 1.3118486941699375e-05, + "loss": 0.2396, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1278199702501297, + "step": 5940, + "valid_targets_mean": 5558.8, + "valid_targets_min": 533 + }, + { + "epoch": 4.5592635212888375, + "grad_norm": 0.5827946313934804, + "learning_rate": 1.3082592385089815e-05, + "loss": 0.2304, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11986644566059113, + "step": 5945, + "valid_targets_mean": 4660.4, + "valid_targets_min": 1752 + }, + { + "epoch": 4.563099347909475, + "grad_norm": 0.6600050897979789, + "learning_rate": 1.3046723119588112e-05, + "loss": 0.2256, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12278605997562408, + "step": 5950, + "valid_targets_mean": 4029.9, + "valid_targets_min": 1201 + }, + { + "epoch": 4.566935174530111, + "grad_norm": 0.5509125981025357, + "learning_rate": 1.301087927633784e-05, + "loss": 0.2128, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10463835299015045, + "step": 5955, + "valid_targets_mean": 4551.0, + "valid_targets_min": 2259 + }, + { + "epoch": 4.570771001150748, + "grad_norm": 0.5708117533798637, + "learning_rate": 1.297506098638961e-05, + "loss": 0.2351, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14241449534893036, + "step": 5960, + "valid_targets_mean": 5568.8, + "valid_targets_min": 2171 + }, + { + "epoch": 4.574606827771385, + "grad_norm": 0.5156064753565629, + "learning_rate": 1.29392683807006e-05, + "loss": 0.2184, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12001025676727295, + "step": 5965, + "valid_targets_mean": 7026.1, + "valid_targets_min": 4500 + }, + { + "epoch": 4.578442654392021, + "grad_norm": 0.5554083160804721, + "learning_rate": 1.2903501590134097e-05, + "loss": 0.2255, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1278180480003357, + "step": 5970, + "valid_targets_mean": 5688.4, + "valid_targets_min": 1855 + }, + { + "epoch": 4.582278481012658, + "grad_norm": 0.6047174760249403, + "learning_rate": 1.2867760745459004e-05, + "loss": 0.2172, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11754312366247177, + "step": 5975, + "valid_targets_mean": 4019.0, + "valid_targets_min": 1024 + }, + { + "epoch": 4.586114307633295, + "grad_norm": 0.5629775434618268, + "learning_rate": 1.2832045977349337e-05, + "loss": 0.2302, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10742682218551636, + "step": 5980, + "valid_targets_mean": 4988.4, + "valid_targets_min": 2569 + }, + { + "epoch": 4.5899501342539315, + "grad_norm": 0.6317632866361073, + "learning_rate": 1.2796357416383798e-05, + "loss": 0.228, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12545615434646606, + "step": 5985, + "valid_targets_mean": 3892.2, + "valid_targets_min": 802 + }, + { + "epoch": 4.593785960874569, + "grad_norm": 0.45507246491958897, + "learning_rate": 1.2760695193045276e-05, + "loss": 0.2137, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11185461282730103, + "step": 5990, + "valid_targets_mean": 5692.4, + "valid_targets_min": 1597 + }, + { + "epoch": 4.597621787495205, + "grad_norm": 0.5893646829539114, + "learning_rate": 1.2725059437720338e-05, + "loss": 0.2254, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11180223524570465, + "step": 5995, + "valid_targets_mean": 3506.8, + "valid_targets_min": 1650 + }, + { + "epoch": 4.601457614115842, + "grad_norm": 0.5654565065398138, + "learning_rate": 1.2689450280698803e-05, + "loss": 0.2207, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10936104506254196, + "step": 6000, + "valid_targets_mean": 4337.4, + "valid_targets_min": 1385 + }, + { + "epoch": 4.605293440736479, + "grad_norm": 0.5785755166355194, + "learning_rate": 1.2653867852173223e-05, + "loss": 0.2158, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11755833029747009, + "step": 6005, + "valid_targets_mean": 3932.2, + "valid_targets_min": 2549 + }, + { + "epoch": 4.609129267357115, + "grad_norm": 0.5389747149400895, + "learning_rate": 1.2618312282238451e-05, + "loss": 0.2192, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1003919392824173, + "step": 6010, + "valid_targets_mean": 4973.0, + "valid_targets_min": 2052 + }, + { + "epoch": 4.612965093977753, + "grad_norm": 0.6192767648440826, + "learning_rate": 1.258278370089113e-05, + "loss": 0.2113, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10240796208381653, + "step": 6015, + "valid_targets_mean": 4029.6, + "valid_targets_min": 1210 + }, + { + "epoch": 4.616800920598389, + "grad_norm": 0.6291883289420841, + "learning_rate": 1.2547282238029213e-05, + "loss": 0.2189, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12469323724508286, + "step": 6020, + "valid_targets_mean": 4094.8, + "valid_targets_min": 1064 + }, + { + "epoch": 4.6206367472190255, + "grad_norm": 0.7517627521306819, + "learning_rate": 1.2511808023451523e-05, + "loss": 0.2273, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13166679441928864, + "step": 6025, + "valid_targets_mean": 3760.1, + "valid_targets_min": 1678 + }, + { + "epoch": 4.624472573839663, + "grad_norm": 0.6806459742971653, + "learning_rate": 1.2476361186857262e-05, + "loss": 0.2214, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1198461651802063, + "step": 6030, + "valid_targets_mean": 3665.2, + "valid_targets_min": 671 + }, + { + "epoch": 4.628308400460299, + "grad_norm": 0.6563795031938031, + "learning_rate": 1.2440941857845516e-05, + "loss": 0.2172, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09944385290145874, + "step": 6035, + "valid_targets_mean": 3325.0, + "valid_targets_min": 1864 + }, + { + "epoch": 4.632144227080936, + "grad_norm": 0.6632069350410055, + "learning_rate": 1.2405550165914806e-05, + "loss": 0.2023, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11336876451969147, + "step": 6040, + "valid_targets_mean": 3751.9, + "valid_targets_min": 1532 + }, + { + "epoch": 4.635980053701573, + "grad_norm": 0.6032220959309063, + "learning_rate": 1.2370186240462622e-05, + "loss": 0.2103, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0902489647269249, + "step": 6045, + "valid_targets_mean": 3773.9, + "valid_targets_min": 1051 + }, + { + "epoch": 4.639815880322209, + "grad_norm": 0.6153790610516952, + "learning_rate": 1.233485021078491e-05, + "loss": 0.2001, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10852416604757309, + "step": 6050, + "valid_targets_mean": 3679.9, + "valid_targets_min": 1345 + }, + { + "epoch": 4.643651706942846, + "grad_norm": 0.6307682652952609, + "learning_rate": 1.2299542206075641e-05, + "loss": 0.2164, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10726836323738098, + "step": 6055, + "valid_targets_mean": 4461.5, + "valid_targets_min": 2231 + }, + { + "epoch": 4.647487533563483, + "grad_norm": 0.6859005725005707, + "learning_rate": 1.226426235542633e-05, + "loss": 0.2179, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10614298284053802, + "step": 6060, + "valid_targets_mean": 3884.6, + "valid_targets_min": 2127 + }, + { + "epoch": 4.6513233601841195, + "grad_norm": 0.6454777041225607, + "learning_rate": 1.2229010787825529e-05, + "loss": 0.2032, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09876096993684769, + "step": 6065, + "valid_targets_mean": 3261.0, + "valid_targets_min": 1599 + }, + { + "epoch": 4.655159186804756, + "grad_norm": 0.6272430923296421, + "learning_rate": 1.2193787632158411e-05, + "loss": 0.207, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08384869992733002, + "step": 6070, + "valid_targets_mean": 2629.1, + "valid_targets_min": 908 + }, + { + "epoch": 4.658995013425393, + "grad_norm": 0.6901702311108918, + "learning_rate": 1.2158593017206263e-05, + "loss": 0.2091, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13275034725666046, + "step": 6075, + "valid_targets_mean": 3459.9, + "valid_targets_min": 1264 + }, + { + "epoch": 4.66283084004603, + "grad_norm": 0.6272169178162346, + "learning_rate": 1.2123427071646003e-05, + "loss": 0.2079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09067107737064362, + "step": 6080, + "valid_targets_mean": 2603.2, + "valid_targets_min": 926 + }, + { + "epoch": 4.666666666666667, + "grad_norm": 0.6714718389330004, + "learning_rate": 1.2088289924049766e-05, + "loss": 0.2119, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10703670978546143, + "step": 6085, + "valid_targets_mean": 3409.6, + "valid_targets_min": 1200 + }, + { + "epoch": 4.670502493287303, + "grad_norm": 0.6904300386184395, + "learning_rate": 1.2053181702884353e-05, + "loss": 0.2027, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10846846550703049, + "step": 6090, + "valid_targets_mean": 3607.5, + "valid_targets_min": 2019 + }, + { + "epoch": 4.67433831990794, + "grad_norm": 0.632038546151617, + "learning_rate": 1.2018102536510846e-05, + "loss": 0.2101, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12740489840507507, + "step": 6095, + "valid_targets_mean": 4661.0, + "valid_targets_min": 2908 + }, + { + "epoch": 4.678174146528577, + "grad_norm": 0.6281750935460643, + "learning_rate": 1.1983052553184083e-05, + "loss": 0.2046, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08864295482635498, + "step": 6100, + "valid_targets_mean": 3215.2, + "valid_targets_min": 1110 + }, + { + "epoch": 4.6820099731492135, + "grad_norm": 0.6665776385092145, + "learning_rate": 1.1948031881052187e-05, + "loss": 0.2034, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10864302515983582, + "step": 6105, + "valid_targets_mean": 3720.9, + "valid_targets_min": 513 + }, + { + "epoch": 4.685845799769851, + "grad_norm": 0.6432333134394118, + "learning_rate": 1.1913040648156144e-05, + "loss": 0.2064, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1111389696598053, + "step": 6110, + "valid_targets_mean": 4162.0, + "valid_targets_min": 2558 + }, + { + "epoch": 4.689681626390487, + "grad_norm": 0.6446335513342106, + "learning_rate": 1.1878078982429289e-05, + "loss": 0.2162, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10066306591033936, + "step": 6115, + "valid_targets_mean": 3291.6, + "valid_targets_min": 1673 + }, + { + "epoch": 4.693517453011124, + "grad_norm": 0.7205451501135408, + "learning_rate": 1.184314701169685e-05, + "loss": 0.2073, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09512203186750412, + "step": 6120, + "valid_targets_mean": 3464.6, + "valid_targets_min": 1860 + }, + { + "epoch": 4.697353279631761, + "grad_norm": 0.5904300803549164, + "learning_rate": 1.1808244863675498e-05, + "loss": 0.2063, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10711782425642014, + "step": 6125, + "valid_targets_mean": 5273.0, + "valid_targets_min": 2495 + }, + { + "epoch": 4.701189106252397, + "grad_norm": 0.6422144630860708, + "learning_rate": 1.1773372665972872e-05, + "loss": 0.2046, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1218777596950531, + "step": 6130, + "valid_targets_mean": 3913.6, + "valid_targets_min": 1945 + }, + { + "epoch": 4.705024932873034, + "grad_norm": 0.6289956884390581, + "learning_rate": 1.1738530546087086e-05, + "loss": 0.2106, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1052652895450592, + "step": 6135, + "valid_targets_mean": 4705.5, + "valid_targets_min": 1908 + }, + { + "epoch": 4.708860759493671, + "grad_norm": 0.6366588581164201, + "learning_rate": 1.17037186314063e-05, + "loss": 0.2057, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10649845004081726, + "step": 6140, + "valid_targets_mean": 3445.9, + "valid_targets_min": 1757 + }, + { + "epoch": 4.7126965861143075, + "grad_norm": 0.6156219213553654, + "learning_rate": 1.1668937049208251e-05, + "loss": 0.2039, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09350037574768066, + "step": 6145, + "valid_targets_mean": 4265.4, + "valid_targets_min": 2439 + }, + { + "epoch": 4.716532412734944, + "grad_norm": 0.6328156840309415, + "learning_rate": 1.1634185926659733e-05, + "loss": 0.2068, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0992661565542221, + "step": 6150, + "valid_targets_mean": 3829.6, + "valid_targets_min": 2482 + }, + { + "epoch": 4.720368239355581, + "grad_norm": 0.6255898756531633, + "learning_rate": 1.1599465390816232e-05, + "loss": 0.2019, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10773700475692749, + "step": 6155, + "valid_targets_mean": 3815.4, + "valid_targets_min": 1833 + }, + { + "epoch": 4.724204065976218, + "grad_norm": 0.6765265544327506, + "learning_rate": 1.1564775568621367e-05, + "loss": 0.2021, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08736778050661087, + "step": 6160, + "valid_targets_mean": 2736.6, + "valid_targets_min": 1182 + }, + { + "epoch": 4.728039892596855, + "grad_norm": 0.5972457689294021, + "learning_rate": 1.1530116586906456e-05, + "loss": 0.2044, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10141852498054504, + "step": 6165, + "valid_targets_mean": 4636.2, + "valid_targets_min": 1869 + }, + { + "epoch": 4.731875719217491, + "grad_norm": 0.6282824130231602, + "learning_rate": 1.1495488572390106e-05, + "loss": 0.2044, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1146717518568039, + "step": 6170, + "valid_targets_mean": 4108.8, + "valid_targets_min": 2498 + }, + { + "epoch": 4.735711545838128, + "grad_norm": 0.6596302105042832, + "learning_rate": 1.1460891651677637e-05, + "loss": 0.2154, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09300608932971954, + "step": 6175, + "valid_targets_mean": 3190.2, + "valid_targets_min": 988 + }, + { + "epoch": 4.739547372458765, + "grad_norm": 0.6958318522345084, + "learning_rate": 1.1426325951260746e-05, + "loss": 0.206, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1062602698802948, + "step": 6180, + "valid_targets_mean": 3140.5, + "valid_targets_min": 1361 + }, + { + "epoch": 4.7433831990794015, + "grad_norm": 0.6022692495141581, + "learning_rate": 1.1391791597516964e-05, + "loss": 0.2079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09696709364652634, + "step": 6185, + "valid_targets_mean": 4282.2, + "valid_targets_min": 2252 + }, + { + "epoch": 4.747219025700039, + "grad_norm": 0.7155393819663886, + "learning_rate": 1.1357288716709174e-05, + "loss": 0.209, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0978192687034607, + "step": 6190, + "valid_targets_mean": 3130.8, + "valid_targets_min": 1436 + }, + { + "epoch": 4.751054852320675, + "grad_norm": 0.6304071585730062, + "learning_rate": 1.1322817434985257e-05, + "loss": 0.2047, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08857513964176178, + "step": 6195, + "valid_targets_mean": 3714.9, + "valid_targets_min": 2200 + }, + { + "epoch": 4.754890678941312, + "grad_norm": 0.6509222484297947, + "learning_rate": 1.1288377878377514e-05, + "loss": 0.2022, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10461990535259247, + "step": 6200, + "valid_targets_mean": 4340.8, + "valid_targets_min": 1272 + }, + { + "epoch": 4.758726505561949, + "grad_norm": 0.6099274985226446, + "learning_rate": 1.1253970172802276e-05, + "loss": 0.2003, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10428176820278168, + "step": 6205, + "valid_targets_mean": 3423.1, + "valid_targets_min": 1319 + }, + { + "epoch": 4.762562332182585, + "grad_norm": 0.5659882549403806, + "learning_rate": 1.121959444405941e-05, + "loss": 0.1865, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09321720898151398, + "step": 6210, + "valid_targets_mean": 3725.6, + "valid_targets_min": 1194 + }, + { + "epoch": 4.766398158803222, + "grad_norm": 0.6652442940917075, + "learning_rate": 1.1185250817831879e-05, + "loss": 0.2209, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09109699726104736, + "step": 6215, + "valid_targets_mean": 2583.6, + "valid_targets_min": 1453 + }, + { + "epoch": 4.770233985423859, + "grad_norm": 0.5381959195554927, + "learning_rate": 1.1150939419685268e-05, + "loss": 0.2005, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08415363729000092, + "step": 6220, + "valid_targets_mean": 5286.5, + "valid_targets_min": 3139 + }, + { + "epoch": 4.7740698120444955, + "grad_norm": 0.6352622419979853, + "learning_rate": 1.1116660375067336e-05, + "loss": 0.2163, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08917569369077682, + "step": 6225, + "valid_targets_mean": 3116.1, + "valid_targets_min": 1256 + }, + { + "epoch": 4.777905638665132, + "grad_norm": 0.657352368923028, + "learning_rate": 1.1082413809307551e-05, + "loss": 0.2083, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1310250163078308, + "step": 6230, + "valid_targets_mean": 4477.9, + "valid_targets_min": 1608 + }, + { + "epoch": 4.781741465285769, + "grad_norm": 0.6343246198987194, + "learning_rate": 1.1048199847616634e-05, + "loss": 0.2147, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10516010224819183, + "step": 6235, + "valid_targets_mean": 3772.4, + "valid_targets_min": 1611 + }, + { + "epoch": 4.785577291906406, + "grad_norm": 0.6493308635800344, + "learning_rate": 1.10140186150861e-05, + "loss": 0.2073, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11766621470451355, + "step": 6240, + "valid_targets_mean": 4032.5, + "valid_targets_min": 2202 + }, + { + "epoch": 4.789413118527042, + "grad_norm": 0.6611054807552936, + "learning_rate": 1.0979870236687798e-05, + "loss": 0.2007, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12111271917819977, + "step": 6245, + "valid_targets_mean": 4019.0, + "valid_targets_min": 2582 + }, + { + "epoch": 4.793248945147679, + "grad_norm": 0.6753002846369952, + "learning_rate": 1.0945754837273461e-05, + "loss": 0.1979, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10814665257930756, + "step": 6250, + "valid_targets_mean": 3423.1, + "valid_targets_min": 2314 + }, + { + "epoch": 4.797084771768316, + "grad_norm": 0.6500877018919604, + "learning_rate": 1.0911672541574245e-05, + "loss": 0.1968, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10183213651180267, + "step": 6255, + "valid_targets_mean": 3991.5, + "valid_targets_min": 2343 + }, + { + "epoch": 4.800920598388953, + "grad_norm": 0.6123610631570575, + "learning_rate": 1.0877623474200273e-05, + "loss": 0.1927, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08726285398006439, + "step": 6260, + "valid_targets_mean": 2864.9, + "valid_targets_min": 1026 + }, + { + "epoch": 4.8047564250095895, + "grad_norm": 0.6666751953755633, + "learning_rate": 1.0843607759640182e-05, + "loss": 0.1942, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10687262564897537, + "step": 6265, + "valid_targets_mean": 4022.8, + "valid_targets_min": 1042 + }, + { + "epoch": 4.808592251630226, + "grad_norm": 0.6276174680137787, + "learning_rate": 1.0809625522260663e-05, + "loss": 0.1821, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09248840808868408, + "step": 6270, + "valid_targets_mean": 3428.8, + "valid_targets_min": 1274 + }, + { + "epoch": 4.812428078250863, + "grad_norm": 0.675242857318853, + "learning_rate": 1.0775676886306013e-05, + "loss": 0.1983, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10807399451732635, + "step": 6275, + "valid_targets_mean": 4167.0, + "valid_targets_min": 914 + }, + { + "epoch": 4.8162639048715, + "grad_norm": 0.6180578174965986, + "learning_rate": 1.0741761975897671e-05, + "loss": 0.2244, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08793118596076965, + "step": 6280, + "valid_targets_mean": 3927.2, + "valid_targets_min": 2093 + }, + { + "epoch": 4.820099731492137, + "grad_norm": 0.6693033983015216, + "learning_rate": 1.0707880915033777e-05, + "loss": 0.1976, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11797918379306793, + "step": 6285, + "valid_targets_mean": 3882.8, + "valid_targets_min": 1683 + }, + { + "epoch": 4.823935558112773, + "grad_norm": 0.6096680352849259, + "learning_rate": 1.067403382758871e-05, + "loss": 0.2066, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10478854179382324, + "step": 6290, + "valid_targets_mean": 4466.9, + "valid_targets_min": 2253 + }, + { + "epoch": 4.82777138473341, + "grad_norm": 0.6011261767156612, + "learning_rate": 1.0640220837312634e-05, + "loss": 0.1995, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09717799723148346, + "step": 6295, + "valid_targets_mean": 3929.2, + "valid_targets_min": 1789 + }, + { + "epoch": 4.831607211354047, + "grad_norm": 0.6608892197579915, + "learning_rate": 1.060644206783105e-05, + "loss": 0.2079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11043626815080643, + "step": 6300, + "valid_targets_mean": 4174.0, + "valid_targets_min": 1586 + }, + { + "epoch": 4.8354430379746836, + "grad_norm": 0.6841322585474678, + "learning_rate": 1.057269764264434e-05, + "loss": 0.2012, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11589368432760239, + "step": 6305, + "valid_targets_mean": 3746.6, + "valid_targets_min": 2033 + }, + { + "epoch": 4.83927886459532, + "grad_norm": 0.5972121815462891, + "learning_rate": 1.053898768512732e-05, + "loss": 0.1915, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09624233841896057, + "step": 6310, + "valid_targets_mean": 3945.5, + "valid_targets_min": 1616 + }, + { + "epoch": 4.843114691215957, + "grad_norm": 0.6112849268320905, + "learning_rate": 1.0505312318528804e-05, + "loss": 0.207, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08522311598062515, + "step": 6315, + "valid_targets_mean": 3534.4, + "valid_targets_min": 2156 + }, + { + "epoch": 4.846950517836594, + "grad_norm": 0.6257778105239956, + "learning_rate": 1.0471671665971104e-05, + "loss": 0.1947, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10826162993907928, + "step": 6320, + "valid_targets_mean": 3635.5, + "valid_targets_min": 1750 + }, + { + "epoch": 4.85078634445723, + "grad_norm": 0.7904008524068522, + "learning_rate": 1.0438065850449624e-05, + "loss": 0.2014, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12460195273160934, + "step": 6325, + "valid_targets_mean": 3780.1, + "valid_targets_min": 2126 + }, + { + "epoch": 4.854622171077867, + "grad_norm": 0.6863380890137596, + "learning_rate": 1.0404494994832421e-05, + "loss": 0.2148, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11547179520130157, + "step": 6330, + "valid_targets_mean": 3756.1, + "valid_targets_min": 1834 + }, + { + "epoch": 4.858457997698504, + "grad_norm": 0.6748472047522664, + "learning_rate": 1.0370959221859702e-05, + "loss": 0.2072, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12121252715587616, + "step": 6335, + "valid_targets_mean": 4082.5, + "valid_targets_min": 1571 + }, + { + "epoch": 4.862293824319141, + "grad_norm": 0.7112232601242412, + "learning_rate": 1.0337458654143417e-05, + "loss": 0.2148, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11240433901548386, + "step": 6340, + "valid_targets_mean": 3371.9, + "valid_targets_min": 1487 + }, + { + "epoch": 4.866129650939778, + "grad_norm": 0.6152377448236396, + "learning_rate": 1.0303993414166808e-05, + "loss": 0.2078, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08433739095926285, + "step": 6345, + "valid_targets_mean": 3824.5, + "valid_targets_min": 1970 + }, + { + "epoch": 4.869965477560414, + "grad_norm": 0.6223142393180298, + "learning_rate": 1.027056362428394e-05, + "loss": 0.2198, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09960272908210754, + "step": 6350, + "valid_targets_mean": 3956.8, + "valid_targets_min": 2492 + }, + { + "epoch": 4.873801304181051, + "grad_norm": 0.666348562427564, + "learning_rate": 1.0237169406719296e-05, + "loss": 0.213, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09271334111690521, + "step": 6355, + "valid_targets_mean": 3557.9, + "valid_targets_min": 1675 + }, + { + "epoch": 4.877637130801688, + "grad_norm": 0.64867643114164, + "learning_rate": 1.0203810883567255e-05, + "loss": 0.2142, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09892235696315765, + "step": 6360, + "valid_targets_mean": 3468.0, + "valid_targets_min": 2437 + }, + { + "epoch": 4.881472957422325, + "grad_norm": 0.7086209360608666, + "learning_rate": 1.0170488176791721e-05, + "loss": 0.2156, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1067817285656929, + "step": 6365, + "valid_targets_mean": 3597.0, + "valid_targets_min": 1969 + }, + { + "epoch": 4.885308784042961, + "grad_norm": 0.6502174103798045, + "learning_rate": 1.0137201408225663e-05, + "loss": 0.2047, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09159287810325623, + "step": 6370, + "valid_targets_mean": 3417.6, + "valid_targets_min": 1005 + }, + { + "epoch": 4.889144610663598, + "grad_norm": 0.7518198177834312, + "learning_rate": 1.0103950699570615e-05, + "loss": 0.2108, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0861232653260231, + "step": 6375, + "valid_targets_mean": 3225.5, + "valid_targets_min": 1207 + }, + { + "epoch": 4.892980437284235, + "grad_norm": 0.6511777635624048, + "learning_rate": 1.0070736172396286e-05, + "loss": 0.2141, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11640223860740662, + "step": 6380, + "valid_targets_mean": 4136.9, + "valid_targets_min": 1257 + }, + { + "epoch": 4.896816263904872, + "grad_norm": 0.7099242704375199, + "learning_rate": 1.0037557948140124e-05, + "loss": 0.2029, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10071162134408951, + "step": 6385, + "valid_targets_mean": 3068.1, + "valid_targets_min": 1357 + }, + { + "epoch": 4.900652090525508, + "grad_norm": 0.6297442223843981, + "learning_rate": 1.0004416148106808e-05, + "loss": 0.1935, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08636964857578278, + "step": 6390, + "valid_targets_mean": 4579.2, + "valid_targets_min": 1410 + }, + { + "epoch": 4.904487917146145, + "grad_norm": 0.6362187572078435, + "learning_rate": 9.97131089346788e-06, + "loss": 0.2105, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0942939966917038, + "step": 6395, + "valid_targets_mean": 3438.5, + "valid_targets_min": 1390 + }, + { + "epoch": 4.908323743766782, + "grad_norm": 0.655936071499079, + "learning_rate": 9.938242305261253e-06, + "loss": 0.1977, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08523450791835785, + "step": 6400, + "valid_targets_mean": 3037.1, + "valid_targets_min": 790 + }, + { + "epoch": 4.912159570387418, + "grad_norm": 0.6810307235781473, + "learning_rate": 9.905210504390755e-06, + "loss": 0.2149, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10180039703845978, + "step": 6405, + "valid_targets_mean": 3695.4, + "valid_targets_min": 1901 + }, + { + "epoch": 4.915995397008055, + "grad_norm": 0.7041689764547455, + "learning_rate": 9.87221561162577e-06, + "loss": 0.2062, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09847723692655563, + "step": 6410, + "valid_targets_mean": 2810.9, + "valid_targets_min": 1460 + }, + { + "epoch": 4.919831223628692, + "grad_norm": 0.6537349557659898, + "learning_rate": 9.83925774760068e-06, + "loss": 0.2098, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12227816879749298, + "step": 6415, + "valid_targets_mean": 5414.1, + "valid_targets_min": 2900 + }, + { + "epoch": 4.923667050249328, + "grad_norm": 0.8743698702493407, + "learning_rate": 9.806337032814533e-06, + "loss": 0.2053, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12008453905582428, + "step": 6420, + "valid_targets_mean": 3211.0, + "valid_targets_min": 1768 + }, + { + "epoch": 4.927502876869966, + "grad_norm": 0.6043036815345398, + "learning_rate": 9.773453587630536e-06, + "loss": 0.2058, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09489124268293381, + "step": 6425, + "valid_targets_mean": 4090.2, + "valid_targets_min": 1931 + }, + { + "epoch": 4.931338703490602, + "grad_norm": 0.5511344658154879, + "learning_rate": 9.740607532275603e-06, + "loss": 0.2038, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08979807049036026, + "step": 6430, + "valid_targets_mean": 5752.6, + "valid_targets_min": 1233 + }, + { + "epoch": 4.935174530111239, + "grad_norm": 0.5560983940978323, + "learning_rate": 9.707798986840003e-06, + "loss": 0.1994, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09843230992555618, + "step": 6435, + "valid_targets_mean": 5342.2, + "valid_targets_min": 1741 + }, + { + "epoch": 4.939010356731876, + "grad_norm": 0.6048642132499444, + "learning_rate": 9.675028071276826e-06, + "loss": 0.2017, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10471391677856445, + "step": 6440, + "valid_targets_mean": 3760.9, + "valid_targets_min": 856 + }, + { + "epoch": 4.942846183352512, + "grad_norm": 0.6264998083553485, + "learning_rate": 9.642294905401565e-06, + "loss": 0.2046, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10614514350891113, + "step": 6445, + "valid_targets_mean": 3722.1, + "valid_targets_min": 1062 + }, + { + "epoch": 4.946682009973149, + "grad_norm": 0.6150386330099883, + "learning_rate": 9.609599608891744e-06, + "loss": 0.1964, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09380221366882324, + "step": 6450, + "valid_targets_mean": 3910.2, + "valid_targets_min": 974 + }, + { + "epoch": 4.950517836593786, + "grad_norm": 0.6812505731890866, + "learning_rate": 9.576942301286393e-06, + "loss": 0.1959, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.097880519926548, + "step": 6455, + "valid_targets_mean": 3008.9, + "valid_targets_min": 1447 + }, + { + "epoch": 4.954353663214423, + "grad_norm": 0.6466225824394727, + "learning_rate": 9.544323101985663e-06, + "loss": 0.197, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11226776242256165, + "step": 6460, + "valid_targets_mean": 4858.2, + "valid_targets_min": 3617 + }, + { + "epoch": 4.95818948983506, + "grad_norm": 0.6647742673433029, + "learning_rate": 9.511742130250366e-06, + "loss": 0.1955, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10723144561052322, + "step": 6465, + "valid_targets_mean": 4037.0, + "valid_targets_min": 1515 + }, + { + "epoch": 4.962025316455696, + "grad_norm": 0.7431436798941812, + "learning_rate": 9.479199505201557e-06, + "loss": 0.2099, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10619507730007172, + "step": 6470, + "valid_targets_mean": 2778.0, + "valid_targets_min": 1349 + }, + { + "epoch": 4.965861143076333, + "grad_norm": 0.666122732626293, + "learning_rate": 9.446695345820085e-06, + "loss": 0.2147, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10945937037467957, + "step": 6475, + "valid_targets_mean": 3199.2, + "valid_targets_min": 1906 + }, + { + "epoch": 4.96969696969697, + "grad_norm": 0.6868093404308906, + "learning_rate": 9.414229770946166e-06, + "loss": 0.1943, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10278673470020294, + "step": 6480, + "valid_targets_mean": 3461.5, + "valid_targets_min": 2076 + }, + { + "epoch": 4.973532796317606, + "grad_norm": 0.6113069188307488, + "learning_rate": 9.381802899278942e-06, + "loss": 0.1915, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08219559490680695, + "step": 6485, + "valid_targets_mean": 3071.8, + "valid_targets_min": 1021 + }, + { + "epoch": 4.9773686229382434, + "grad_norm": 0.7178408784227869, + "learning_rate": 9.349414849376051e-06, + "loss": 0.2047, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09587296843528748, + "step": 6490, + "valid_targets_mean": 3088.9, + "valid_targets_min": 654 + }, + { + "epoch": 4.98120444955888, + "grad_norm": 0.6619604442919295, + "learning_rate": 9.317065739653193e-06, + "loss": 0.1982, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11185316741466522, + "step": 6495, + "valid_targets_mean": 4179.1, + "valid_targets_min": 1605 + }, + { + "epoch": 4.985040276179516, + "grad_norm": 0.7643461066426106, + "learning_rate": 9.284755688383695e-06, + "loss": 0.2124, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11014701426029205, + "step": 6500, + "valid_targets_mean": 2737.0, + "valid_targets_min": 815 + }, + { + "epoch": 4.988876102800154, + "grad_norm": 0.6678827302204615, + "learning_rate": 9.252484813698085e-06, + "loss": 0.2062, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09306727349758148, + "step": 6505, + "valid_targets_mean": 3384.8, + "valid_targets_min": 1878 + }, + { + "epoch": 4.99271192942079, + "grad_norm": 1.1196851985008065, + "learning_rate": 9.220253233583649e-06, + "loss": 0.2054, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10716482996940613, + "step": 6510, + "valid_targets_mean": 3888.4, + "valid_targets_min": 1793 + }, + { + "epoch": 4.996547756041427, + "grad_norm": 0.6970826237673434, + "learning_rate": 9.188061065884013e-06, + "loss": 0.2109, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09667366743087769, + "step": 6515, + "valid_targets_mean": 3238.4, + "valid_targets_min": 1217 + }, + { + "epoch": 5.0, + "grad_norm": 0.592970859883454, + "learning_rate": 9.155908428298696e-06, + "loss": 0.1923, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12662284076213837, + "step": 6520, + "valid_targets_mean": 7108.5, + "valid_targets_min": 2092 + }, + { + "epoch": 5.003835826620636, + "grad_norm": 0.6707123752033317, + "learning_rate": 9.123795438382698e-06, + "loss": 0.1772, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08760679513216019, + "step": 6525, + "valid_targets_mean": 7389.0, + "valid_targets_min": 5627 + }, + { + "epoch": 5.007671653241274, + "grad_norm": 0.7025440133739169, + "learning_rate": 9.091722213546059e-06, + "loss": 0.18, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.048427604138851166, + "step": 6530, + "valid_targets_mean": 1294.8, + "valid_targets_min": 422 + }, + { + "epoch": 5.01150747986191, + "grad_norm": 0.43076732804259654, + "learning_rate": 9.059688871053426e-06, + "loss": 0.1736, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09039315581321716, + "step": 6535, + "valid_targets_mean": 8539.9, + "valid_targets_min": 6761 + }, + { + "epoch": 5.015343306482547, + "grad_norm": 0.4189217545552735, + "learning_rate": 9.027695528023632e-06, + "loss": 0.1684, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07945355772972107, + "step": 6540, + "valid_targets_mean": 7579.0, + "valid_targets_min": 6174 + }, + { + "epoch": 5.019179133103184, + "grad_norm": 0.4015456545514661, + "learning_rate": 8.995742301429274e-06, + "loss": 0.1726, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07680564373731613, + "step": 6545, + "valid_targets_mean": 7328.2, + "valid_targets_min": 4839 + }, + { + "epoch": 5.02301495972382, + "grad_norm": 0.4470975300228331, + "learning_rate": 8.963829308096266e-06, + "loss": 0.18, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09324349462985992, + "step": 6550, + "valid_targets_mean": 6838.2, + "valid_targets_min": 5034 + }, + { + "epoch": 5.026850786344458, + "grad_norm": 0.42267247838595307, + "learning_rate": 8.931956664703427e-06, + "loss": 0.184, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09371136128902435, + "step": 6555, + "valid_targets_mean": 7248.2, + "valid_targets_min": 5567 + }, + { + "epoch": 5.030686612965094, + "grad_norm": 0.45065059053020445, + "learning_rate": 8.900124487782056e-06, + "loss": 0.1693, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07774712145328522, + "step": 6560, + "valid_targets_mean": 6234.0, + "valid_targets_min": 5603 + }, + { + "epoch": 5.03452243958573, + "grad_norm": 0.6824414781375057, + "learning_rate": 8.868332893715486e-06, + "loss": 0.1648, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.024296559393405914, + "step": 6565, + "valid_targets_mean": 602.0, + "valid_targets_min": 152 + }, + { + "epoch": 5.038358266206368, + "grad_norm": 0.42216439325482263, + "learning_rate": 8.836581998738705e-06, + "loss": 0.1567, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08029231429100037, + "step": 6570, + "valid_targets_mean": 7299.9, + "valid_targets_min": 6158 + }, + { + "epoch": 5.042194092827004, + "grad_norm": 0.42231130556134433, + "learning_rate": 8.804871918937852e-06, + "loss": 0.1721, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08066429197788239, + "step": 6575, + "valid_targets_mean": 6140.0, + "valid_targets_min": 4648 + }, + { + "epoch": 5.046029919447641, + "grad_norm": 0.45062424801472334, + "learning_rate": 8.773202770249871e-06, + "loss": 0.1647, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08607107400894165, + "step": 6580, + "valid_targets_mean": 6603.0, + "valid_targets_min": 4882 + }, + { + "epoch": 5.049865746068278, + "grad_norm": 0.40112030746239336, + "learning_rate": 8.741574668462053e-06, + "loss": 0.1668, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08525241911411285, + "step": 6585, + "valid_targets_mean": 8015.0, + "valid_targets_min": 6344 + }, + { + "epoch": 5.053701572688914, + "grad_norm": 0.40405378824246124, + "learning_rate": 8.709987729211604e-06, + "loss": 0.1595, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07438480108976364, + "step": 6590, + "valid_targets_mean": 6305.2, + "valid_targets_min": 4622 + }, + { + "epoch": 5.057537399309552, + "grad_norm": 0.4149733454918672, + "learning_rate": 8.678442067985244e-06, + "loss": 0.1673, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08778570592403412, + "step": 6595, + "valid_targets_mean": 7890.8, + "valid_targets_min": 5485 + }, + { + "epoch": 5.061373225930188, + "grad_norm": 0.4200940163832903, + "learning_rate": 8.64693780011877e-06, + "loss": 0.1686, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07702288776636124, + "step": 6600, + "valid_targets_mean": 7016.6, + "valid_targets_min": 4960 + }, + { + "epoch": 5.065209052550824, + "grad_norm": 0.4328497348712492, + "learning_rate": 8.615475040796631e-06, + "loss": 0.1635, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07945902645587921, + "step": 6605, + "valid_targets_mean": 6527.1, + "valid_targets_min": 5399 + }, + { + "epoch": 5.069044879171462, + "grad_norm": 0.5868492651221886, + "learning_rate": 8.58405390505154e-06, + "loss": 0.1704, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08253732323646545, + "step": 6610, + "valid_targets_mean": 4995.0, + "valid_targets_min": 618 + }, + { + "epoch": 5.072880705792098, + "grad_norm": 0.43764029916834996, + "learning_rate": 8.552674507763994e-06, + "loss": 0.1834, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0911393016576767, + "step": 6615, + "valid_targets_mean": 7219.4, + "valid_targets_min": 5764 + }, + { + "epoch": 5.076716532412735, + "grad_norm": 0.43098383995550293, + "learning_rate": 8.521336963661899e-06, + "loss": 0.1511, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08474987745285034, + "step": 6620, + "valid_targets_mean": 7998.5, + "valid_targets_min": 5902 + }, + { + "epoch": 5.080552359033372, + "grad_norm": 0.4458061723438398, + "learning_rate": 8.490041387320168e-06, + "loss": 0.1559, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06739877909421921, + "step": 6625, + "valid_targets_mean": 7377.4, + "valid_targets_min": 5015 + }, + { + "epoch": 5.084388185654008, + "grad_norm": 0.40950591595327346, + "learning_rate": 8.458787893160216e-06, + "loss": 0.1631, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06818641722202301, + "step": 6630, + "valid_targets_mean": 6794.0, + "valid_targets_min": 5283 + }, + { + "epoch": 5.088224012274646, + "grad_norm": 0.45163147685065796, + "learning_rate": 8.42757659544966e-06, + "loss": 0.1687, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09405025839805603, + "step": 6635, + "valid_targets_mean": 6656.4, + "valid_targets_min": 4364 + }, + { + "epoch": 5.092059838895282, + "grad_norm": 0.4096519260860122, + "learning_rate": 8.39640760830181e-06, + "loss": 0.1683, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07692734897136688, + "step": 6640, + "valid_targets_mean": 7178.4, + "valid_targets_min": 4448 + }, + { + "epoch": 5.095895665515918, + "grad_norm": 0.4459941877298365, + "learning_rate": 8.365281045675261e-06, + "loss": 0.1664, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09902428090572357, + "step": 6645, + "valid_targets_mean": 7619.8, + "valid_targets_min": 5737 + }, + { + "epoch": 5.099731492136556, + "grad_norm": 0.39712384680475216, + "learning_rate": 8.334197021373547e-06, + "loss": 0.1472, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0702294111251831, + "step": 6650, + "valid_targets_mean": 8754.4, + "valid_targets_min": 5752 + }, + { + "epoch": 5.103567318757192, + "grad_norm": 0.463129675246172, + "learning_rate": 8.303155649044643e-06, + "loss": 0.1498, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0776011049747467, + "step": 6655, + "valid_targets_mean": 6341.0, + "valid_targets_min": 4543 + }, + { + "epoch": 5.107403145377829, + "grad_norm": 0.440724844347778, + "learning_rate": 8.272157042180573e-06, + "loss": 0.1441, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08313284814357758, + "step": 6660, + "valid_targets_mean": 7181.9, + "valid_targets_min": 4563 + }, + { + "epoch": 5.111238971998466, + "grad_norm": 0.8505942919334928, + "learning_rate": 8.241201314117045e-06, + "loss": 0.1715, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.054243505001068115, + "step": 6665, + "valid_targets_mean": 1104.4, + "valid_targets_min": 161 + }, + { + "epoch": 5.115074798619102, + "grad_norm": 0.4412663840579128, + "learning_rate": 8.21028857803294e-06, + "loss": 0.1539, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07514876127243042, + "step": 6670, + "valid_targets_mean": 7018.2, + "valid_targets_min": 5424 + }, + { + "epoch": 5.118910625239739, + "grad_norm": 0.4507707813463081, + "learning_rate": 8.179418946950003e-06, + "loss": 0.173, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08843988180160522, + "step": 6675, + "valid_targets_mean": 7738.4, + "valid_targets_min": 5329 + }, + { + "epoch": 5.122746451860376, + "grad_norm": 0.49120605293565694, + "learning_rate": 8.148592533732365e-06, + "loss": 0.177, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08389255404472351, + "step": 6680, + "valid_targets_mean": 6426.6, + "valid_targets_min": 5560 + }, + { + "epoch": 5.1265822784810124, + "grad_norm": 0.4976455400893471, + "learning_rate": 8.117809451086119e-06, + "loss": 0.1716, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08443442732095718, + "step": 6685, + "valid_targets_mean": 6169.9, + "valid_targets_min": 3446 + }, + { + "epoch": 5.13041810510165, + "grad_norm": 0.42615588500547874, + "learning_rate": 8.087069811558976e-06, + "loss": 0.1668, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08629438281059265, + "step": 6690, + "valid_targets_mean": 7204.6, + "valid_targets_min": 5381 + }, + { + "epoch": 5.134253931722286, + "grad_norm": 0.4391954069786128, + "learning_rate": 8.0563737275398e-06, + "loss": 0.1717, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0896696075797081, + "step": 6695, + "valid_targets_mean": 7600.4, + "valid_targets_min": 5241 + }, + { + "epoch": 5.138089758342923, + "grad_norm": 0.5138973926833736, + "learning_rate": 8.025721311258171e-06, + "loss": 0.1673, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09089227765798569, + "step": 6700, + "valid_targets_mean": 7098.5, + "valid_targets_min": 5607 + }, + { + "epoch": 5.14192558496356, + "grad_norm": 0.4428213104456719, + "learning_rate": 7.995112674784078e-06, + "loss": 0.1703, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09079594910144806, + "step": 6705, + "valid_targets_mean": 6955.1, + "valid_targets_min": 5530 + }, + { + "epoch": 5.145761411584196, + "grad_norm": 0.45346793001110053, + "learning_rate": 7.964547930027395e-06, + "loss": 0.1715, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08868218958377838, + "step": 6710, + "valid_targets_mean": 6748.1, + "valid_targets_min": 5364 + }, + { + "epoch": 5.149597238204833, + "grad_norm": 0.46581052628474645, + "learning_rate": 7.934027188737541e-06, + "loss": 0.1713, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0824691653251648, + "step": 6715, + "valid_targets_mean": 6273.9, + "valid_targets_min": 5438 + }, + { + "epoch": 5.15343306482547, + "grad_norm": 0.4132679958693429, + "learning_rate": 7.903550562503049e-06, + "loss": 0.167, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07921819388866425, + "step": 6720, + "valid_targets_mean": 6419.5, + "valid_targets_min": 4879 + }, + { + "epoch": 5.1572688914461065, + "grad_norm": 0.438428229034428, + "learning_rate": 7.873118162751159e-06, + "loss": 0.1652, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07931020855903625, + "step": 6725, + "valid_targets_mean": 6483.4, + "valid_targets_min": 4309 + }, + { + "epoch": 5.161104718066744, + "grad_norm": 0.4763687825648046, + "learning_rate": 7.842730100747409e-06, + "loss": 0.1383, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09785845130681992, + "step": 6730, + "valid_targets_mean": 6825.0, + "valid_targets_min": 5205 + }, + { + "epoch": 5.16494054468738, + "grad_norm": 0.4352308316568513, + "learning_rate": 7.812386487595238e-06, + "loss": 0.1658, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07717172056436539, + "step": 6735, + "valid_targets_mean": 6729.4, + "valid_targets_min": 5200 + }, + { + "epoch": 5.168776371308017, + "grad_norm": 0.45819267568804395, + "learning_rate": 7.782087434235573e-06, + "loss": 0.1638, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07976675033569336, + "step": 6740, + "valid_targets_mean": 6094.0, + "valid_targets_min": 5302 + }, + { + "epoch": 5.172612197928654, + "grad_norm": 0.4340592159267031, + "learning_rate": 7.751833051446414e-06, + "loss": 0.1683, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08697172999382019, + "step": 6745, + "valid_targets_mean": 7140.0, + "valid_targets_min": 5270 + }, + { + "epoch": 5.17644802454929, + "grad_norm": 0.38344329753013867, + "learning_rate": 7.721623449842451e-06, + "loss": 0.163, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07733768224716187, + "step": 6750, + "valid_targets_mean": 7425.0, + "valid_targets_min": 5281 + }, + { + "epoch": 5.180283851169927, + "grad_norm": 0.44392699891306975, + "learning_rate": 7.691458739874636e-06, + "loss": 0.1681, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10598724335432053, + "step": 6755, + "valid_targets_mean": 7063.4, + "valid_targets_min": 5386 + }, + { + "epoch": 5.184119677790564, + "grad_norm": 0.4568683591520606, + "learning_rate": 7.6613390318298e-06, + "loss": 0.1684, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0864778459072113, + "step": 6760, + "valid_targets_mean": 6827.0, + "valid_targets_min": 4547 + }, + { + "epoch": 5.1879555044112005, + "grad_norm": 0.6741149155674138, + "learning_rate": 7.631264435830235e-06, + "loss": 0.1459, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03922869265079498, + "step": 6765, + "valid_targets_mean": 2047.0, + "valid_targets_min": 165 + }, + { + "epoch": 5.191791331031838, + "grad_norm": 0.4179728594767338, + "learning_rate": 7.601235061833294e-06, + "loss": 0.15, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07926439493894577, + "step": 6770, + "valid_targets_mean": 6754.1, + "valid_targets_min": 5196 + }, + { + "epoch": 5.195627157652474, + "grad_norm": 0.49912217895097505, + "learning_rate": 7.5712510196309965e-06, + "loss": 0.1623, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08202798664569855, + "step": 6775, + "valid_targets_mean": 5985.6, + "valid_targets_min": 5028 + }, + { + "epoch": 5.199462984273111, + "grad_norm": 0.46562053032962536, + "learning_rate": 7.54131241884962e-06, + "loss": 0.1693, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07801705598831177, + "step": 6780, + "valid_targets_mean": 5963.9, + "valid_targets_min": 5237 + }, + { + "epoch": 5.203298810893748, + "grad_norm": 0.4753162892110123, + "learning_rate": 7.5114193689493e-06, + "loss": 0.1632, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08535606414079666, + "step": 6785, + "valid_targets_mean": 6319.0, + "valid_targets_min": 5412 + }, + { + "epoch": 5.207134637514384, + "grad_norm": 0.4310759624576573, + "learning_rate": 7.481571979223634e-06, + "loss": 0.1555, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06834693253040314, + "step": 6790, + "valid_targets_mean": 7125.8, + "valid_targets_min": 5080 + }, + { + "epoch": 5.210970464135021, + "grad_norm": 0.4519087669854236, + "learning_rate": 7.451770358799273e-06, + "loss": 0.1666, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08345538377761841, + "step": 6795, + "valid_targets_mean": 6123.5, + "valid_targets_min": 4521 + }, + { + "epoch": 5.214806290755658, + "grad_norm": 1.0776803634102499, + "learning_rate": 7.4220146166355355e-06, + "loss": 0.1617, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09534299373626709, + "step": 6800, + "valid_targets_mean": 1886.0, + "valid_targets_min": 1111 + }, + { + "epoch": 5.2186421173762945, + "grad_norm": 0.8240218069947862, + "learning_rate": 7.392304861524e-06, + "loss": 0.1547, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07939254492521286, + "step": 6805, + "valid_targets_mean": 1497.6, + "valid_targets_min": 893 + }, + { + "epoch": 5.222477943996932, + "grad_norm": 0.8040561636805126, + "learning_rate": 7.362641202088103e-06, + "loss": 0.1422, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05186311528086662, + "step": 6810, + "valid_targets_mean": 1356.0, + "valid_targets_min": 914 + }, + { + "epoch": 5.226313770617568, + "grad_norm": 0.818063868958057, + "learning_rate": 7.333023746782757e-06, + "loss": 0.1389, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05730219930410385, + "step": 6815, + "valid_targets_mean": 1324.5, + "valid_targets_min": 680 + }, + { + "epoch": 5.230149597238205, + "grad_norm": 0.7890188247913464, + "learning_rate": 7.303452603893932e-06, + "loss": 0.1419, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06970693171024323, + "step": 6820, + "valid_targets_mean": 1931.8, + "valid_targets_min": 876 + }, + { + "epoch": 5.233985423858842, + "grad_norm": 0.85605531467771, + "learning_rate": 7.273927881538305e-06, + "loss": 0.1351, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06028727442026138, + "step": 6825, + "valid_targets_mean": 1292.2, + "valid_targets_min": 734 + }, + { + "epoch": 5.237821250479478, + "grad_norm": 0.9057403025273895, + "learning_rate": 7.244449687662787e-06, + "loss": 0.1449, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.059388667345047, + "step": 6830, + "valid_targets_mean": 1469.6, + "valid_targets_min": 655 + }, + { + "epoch": 5.241657077100115, + "grad_norm": 0.84299116733337, + "learning_rate": 7.215018130044202e-06, + "loss": 0.1249, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07243464887142181, + "step": 6835, + "valid_targets_mean": 1548.4, + "valid_targets_min": 737 + }, + { + "epoch": 5.245492903720752, + "grad_norm": 0.840206080518879, + "learning_rate": 7.185633316288862e-06, + "loss": 0.1456, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06134296581149101, + "step": 6840, + "valid_targets_mean": 1316.5, + "valid_targets_min": 759 + }, + { + "epoch": 5.2493287303413885, + "grad_norm": 0.9408888527236264, + "learning_rate": 7.156295353832161e-06, + "loss": 0.1334, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.052431635558605194, + "step": 6845, + "valid_targets_mean": 1249.6, + "valid_targets_min": 753 + }, + { + "epoch": 5.253164556962025, + "grad_norm": 0.8913775639909224, + "learning_rate": 7.127004349938234e-06, + "loss": 0.1363, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08337018638849258, + "step": 6850, + "valid_targets_mean": 1833.0, + "valid_targets_min": 927 + }, + { + "epoch": 5.257000383582662, + "grad_norm": 0.8245444502859905, + "learning_rate": 7.0977604116994795e-06, + "loss": 0.1355, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06416749954223633, + "step": 6855, + "valid_targets_mean": 1338.4, + "valid_targets_min": 679 + }, + { + "epoch": 5.260836210203299, + "grad_norm": 0.8621983933442725, + "learning_rate": 7.068563646036244e-06, + "loss": 0.131, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06324025988578796, + "step": 6860, + "valid_targets_mean": 1244.6, + "valid_targets_min": 771 + }, + { + "epoch": 5.264672036823936, + "grad_norm": 0.7992145804682786, + "learning_rate": 7.039414159696418e-06, + "loss": 0.1214, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06358686089515686, + "step": 6865, + "valid_targets_mean": 1732.5, + "valid_targets_min": 986 + }, + { + "epoch": 5.268507863444572, + "grad_norm": 0.8119328801341354, + "learning_rate": 7.010312059254998e-06, + "loss": 0.1263, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05751873552799225, + "step": 6870, + "valid_targets_mean": 1389.9, + "valid_targets_min": 1103 + }, + { + "epoch": 5.272343690065209, + "grad_norm": 0.9233857435648596, + "learning_rate": 6.981257451113741e-06, + "loss": 0.1296, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06699767708778381, + "step": 6875, + "valid_targets_mean": 1593.2, + "valid_targets_min": 675 + }, + { + "epoch": 5.276179516685846, + "grad_norm": 0.8068615401849676, + "learning_rate": 6.952250441500794e-06, + "loss": 0.1308, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07847777754068375, + "step": 6880, + "valid_targets_mean": 1926.6, + "valid_targets_min": 974 + }, + { + "epoch": 5.2800153433064825, + "grad_norm": 0.8632538285595641, + "learning_rate": 6.923291136470225e-06, + "loss": 0.1365, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0764370858669281, + "step": 6885, + "valid_targets_mean": 1953.1, + "valid_targets_min": 869 + }, + { + "epoch": 5.283851169927119, + "grad_norm": 0.8348460021353432, + "learning_rate": 6.894379641901734e-06, + "loss": 0.133, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.056976791471242905, + "step": 6890, + "valid_targets_mean": 1327.2, + "valid_targets_min": 1023 + }, + { + "epoch": 5.287686996547756, + "grad_norm": 0.8862932468783283, + "learning_rate": 6.865516063500199e-06, + "loss": 0.1323, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06726658344268799, + "step": 6895, + "valid_targets_mean": 1717.9, + "valid_targets_min": 709 + }, + { + "epoch": 5.291522823168393, + "grad_norm": 0.8680358525048916, + "learning_rate": 6.836700506795289e-06, + "loss": 0.1301, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05292157083749771, + "step": 6900, + "valid_targets_mean": 1139.9, + "valid_targets_min": 658 + }, + { + "epoch": 5.29535864978903, + "grad_norm": 0.8554534076448744, + "learning_rate": 6.807933077141147e-06, + "loss": 0.1334, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06177346780896187, + "step": 6905, + "valid_targets_mean": 1444.8, + "valid_targets_min": 662 + }, + { + "epoch": 5.299194476409666, + "grad_norm": 0.89412826145495, + "learning_rate": 6.779213879715902e-06, + "loss": 0.1318, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0476764440536499, + "step": 6910, + "valid_targets_mean": 1242.2, + "valid_targets_min": 859 + }, + { + "epoch": 5.303030303030303, + "grad_norm": 0.8929659579194519, + "learning_rate": 6.750543019521372e-06, + "loss": 0.1189, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07105942070484161, + "step": 6915, + "valid_targets_mean": 1569.2, + "valid_targets_min": 950 + }, + { + "epoch": 5.30686612965094, + "grad_norm": 0.9074629894478942, + "learning_rate": 6.721920601382652e-06, + "loss": 0.1263, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0618189312517643, + "step": 6920, + "valid_targets_mean": 1539.8, + "valid_targets_min": 686 + }, + { + "epoch": 5.3107019562715765, + "grad_norm": 0.8586698524929545, + "learning_rate": 6.693346729947687e-06, + "loss": 0.1289, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07187806069850922, + "step": 6925, + "valid_targets_mean": 1610.6, + "valid_targets_min": 702 + }, + { + "epoch": 5.314537782892213, + "grad_norm": 0.8740361576371894, + "learning_rate": 6.664821509686976e-06, + "loss": 0.1277, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04641083627939224, + "step": 6930, + "valid_targets_mean": 1190.4, + "valid_targets_min": 779 + }, + { + "epoch": 5.31837360951285, + "grad_norm": 0.8661217596208461, + "learning_rate": 6.636345044893116e-06, + "loss": 0.138, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06932976841926575, + "step": 6935, + "valid_targets_mean": 1467.1, + "valid_targets_min": 822 + }, + { + "epoch": 5.322209436133487, + "grad_norm": 0.8084355220566614, + "learning_rate": 6.607917439680427e-06, + "loss": 0.124, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.058125924319028854, + "step": 6940, + "valid_targets_mean": 1350.5, + "valid_targets_min": 868 + }, + { + "epoch": 5.326045262754123, + "grad_norm": 0.8657536202229338, + "learning_rate": 6.579538797984635e-06, + "loss": 0.1282, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.052100349217653275, + "step": 6945, + "valid_targets_mean": 1283.5, + "valid_targets_min": 781 + }, + { + "epoch": 5.32988108937476, + "grad_norm": 0.8332002229883131, + "learning_rate": 6.551209223562427e-06, + "loss": 0.1262, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05408904701471329, + "step": 6950, + "valid_targets_mean": 1303.2, + "valid_targets_min": 793 + }, + { + "epoch": 5.333716915995397, + "grad_norm": 0.7789070264182946, + "learning_rate": 6.52292881999107e-06, + "loss": 0.1287, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.053247950971126556, + "step": 6955, + "valid_targets_mean": 1309.1, + "valid_targets_min": 806 + }, + { + "epoch": 5.337552742616034, + "grad_norm": 0.7916312618889738, + "learning_rate": 6.494697690668094e-06, + "loss": 0.1247, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.052947789430618286, + "step": 6960, + "valid_targets_mean": 1397.6, + "valid_targets_min": 803 + }, + { + "epoch": 5.3413885692366705, + "grad_norm": 0.7935156357331365, + "learning_rate": 6.466515938810856e-06, + "loss": 0.1246, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.046653684228658676, + "step": 6965, + "valid_targets_mean": 1187.8, + "valid_targets_min": 660 + }, + { + "epoch": 5.345224395857307, + "grad_norm": 0.8059584223177221, + "learning_rate": 6.4383836674561766e-06, + "loss": 0.1176, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.056122783571481705, + "step": 6970, + "valid_targets_mean": 1547.0, + "valid_targets_min": 853 + }, + { + "epoch": 5.349060222477944, + "grad_norm": 0.8735776213649887, + "learning_rate": 6.410300979459976e-06, + "loss": 0.127, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06820593774318695, + "step": 6975, + "valid_targets_mean": 1704.2, + "valid_targets_min": 719 + }, + { + "epoch": 5.352896049098581, + "grad_norm": 0.8587865300447628, + "learning_rate": 6.382267977496886e-06, + "loss": 0.127, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0489940345287323, + "step": 6980, + "valid_targets_mean": 1255.0, + "valid_targets_min": 765 + }, + { + "epoch": 5.356731875719218, + "grad_norm": 0.9000259438790835, + "learning_rate": 6.3542847640598815e-06, + "loss": 0.1251, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06611358374357224, + "step": 6985, + "valid_targets_mean": 1550.8, + "valid_targets_min": 875 + }, + { + "epoch": 5.360567702339854, + "grad_norm": 0.8506946394971554, + "learning_rate": 6.326351441459908e-06, + "loss": 0.1239, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06112196668982506, + "step": 6990, + "valid_targets_mean": 1362.0, + "valid_targets_min": 739 + }, + { + "epoch": 5.364403528960491, + "grad_norm": 0.8902851740430925, + "learning_rate": 6.298468111825478e-06, + "loss": 0.1307, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07146982848644257, + "step": 6995, + "valid_targets_mean": 1666.8, + "valid_targets_min": 760 + }, + { + "epoch": 5.368239355581128, + "grad_norm": 0.8958433844286183, + "learning_rate": 6.270634877102357e-06, + "loss": 0.1216, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05248474329710007, + "step": 7000, + "valid_targets_mean": 1319.6, + "valid_targets_min": 1046 + }, + { + "epoch": 5.3720751822017645, + "grad_norm": 0.8249186454014623, + "learning_rate": 6.242851839053132e-06, + "loss": 0.1314, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07172630727291107, + "step": 7005, + "valid_targets_mean": 1727.9, + "valid_targets_min": 851 + }, + { + "epoch": 5.375911008822401, + "grad_norm": 0.8526204968101652, + "learning_rate": 6.2151190992568745e-06, + "loss": 0.1332, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06741802394390106, + "step": 7010, + "valid_targets_mean": 1643.2, + "valid_targets_min": 907 + }, + { + "epoch": 5.379746835443038, + "grad_norm": 0.827141105683187, + "learning_rate": 6.187436759108749e-06, + "loss": 0.1341, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05578827112913132, + "step": 7015, + "valid_targets_mean": 1478.5, + "valid_targets_min": 764 + }, + { + "epoch": 5.383582662063675, + "grad_norm": 0.7884380259470222, + "learning_rate": 6.159804919819658e-06, + "loss": 0.1196, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05576085299253464, + "step": 7020, + "valid_targets_mean": 1522.6, + "valid_targets_min": 822 + }, + { + "epoch": 5.387418488684311, + "grad_norm": 0.7956827952035188, + "learning_rate": 6.132223682415861e-06, + "loss": 0.1196, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.061332087963819504, + "step": 7025, + "valid_targets_mean": 1589.0, + "valid_targets_min": 982 + }, + { + "epoch": 5.391254315304948, + "grad_norm": 0.7980681927750569, + "learning_rate": 6.104693147738612e-06, + "loss": 0.133, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05970034748315811, + "step": 7030, + "valid_targets_mean": 1574.8, + "valid_targets_min": 816 + }, + { + "epoch": 5.395090141925585, + "grad_norm": 1.1526582568239008, + "learning_rate": 6.0772134164437855e-06, + "loss": 0.1217, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05711887776851654, + "step": 7035, + "valid_targets_mean": 1422.0, + "valid_targets_min": 721 + }, + { + "epoch": 5.398925968546222, + "grad_norm": 0.8410523316200055, + "learning_rate": 6.049784589001515e-06, + "loss": 0.1165, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0470406599342823, + "step": 7040, + "valid_targets_mean": 1150.6, + "valid_targets_min": 741 + }, + { + "epoch": 5.4027617951668585, + "grad_norm": 0.9515428158677373, + "learning_rate": 6.022406765695816e-06, + "loss": 0.1237, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07509981095790863, + "step": 7045, + "valid_targets_mean": 1684.2, + "valid_targets_min": 734 + }, + { + "epoch": 5.406597621787495, + "grad_norm": 0.8698426247360692, + "learning_rate": 5.995080046624229e-06, + "loss": 0.1271, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07555882632732391, + "step": 7050, + "valid_targets_mean": 1937.1, + "valid_targets_min": 744 + }, + { + "epoch": 5.410433448408132, + "grad_norm": 0.8587189736188438, + "learning_rate": 5.96780453169745e-06, + "loss": 0.1262, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0700686052441597, + "step": 7055, + "valid_targets_mean": 1842.2, + "valid_targets_min": 620 + }, + { + "epoch": 5.414269275028769, + "grad_norm": 0.8979330738961443, + "learning_rate": 5.940580320638956e-06, + "loss": 0.1189, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07197169214487076, + "step": 7060, + "valid_targets_mean": 1637.6, + "valid_targets_min": 785 + }, + { + "epoch": 5.418105101649405, + "grad_norm": 0.8119825965391393, + "learning_rate": 5.913407512984674e-06, + "loss": 0.1292, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06593205779790878, + "step": 7065, + "valid_targets_mean": 1969.9, + "valid_targets_min": 740 + }, + { + "epoch": 5.421940928270042, + "grad_norm": 0.8242130882281791, + "learning_rate": 5.886286208082559e-06, + "loss": 0.1194, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05897783115506172, + "step": 7070, + "valid_targets_mean": 1432.1, + "valid_targets_min": 711 + }, + { + "epoch": 5.425776754890679, + "grad_norm": 0.7793562592882609, + "learning_rate": 5.859216505092285e-06, + "loss": 0.1261, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06500230729579926, + "step": 7075, + "valid_targets_mean": 1695.6, + "valid_targets_min": 1122 + }, + { + "epoch": 5.429612581511316, + "grad_norm": 0.8709863543310534, + "learning_rate": 5.832198502984852e-06, + "loss": 0.1221, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06729629635810852, + "step": 7080, + "valid_targets_mean": 1524.5, + "valid_targets_min": 722 + }, + { + "epoch": 5.4334484081319525, + "grad_norm": 1.16898845221163, + "learning_rate": 5.805232300542245e-06, + "loss": 0.1193, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.054517507553100586, + "step": 7085, + "valid_targets_mean": 1289.2, + "valid_targets_min": 661 + }, + { + "epoch": 5.437284234752589, + "grad_norm": 0.8440609807088185, + "learning_rate": 5.778317996357048e-06, + "loss": 0.1195, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0603988915681839, + "step": 7090, + "valid_targets_mean": 1519.5, + "valid_targets_min": 1011 + }, + { + "epoch": 5.441120061373226, + "grad_norm": 0.8464963459596411, + "learning_rate": 5.751455688832108e-06, + "loss": 0.1292, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06925733387470245, + "step": 7095, + "valid_targets_mean": 1731.4, + "valid_targets_min": 1277 + }, + { + "epoch": 5.444955887993863, + "grad_norm": 0.8166941509494748, + "learning_rate": 5.724645476180151e-06, + "loss": 0.1214, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06345273554325104, + "step": 7100, + "valid_targets_mean": 1486.1, + "valid_targets_min": 807 + }, + { + "epoch": 5.448791714614499, + "grad_norm": 0.8220897147259472, + "learning_rate": 5.697887456423461e-06, + "loss": 0.1251, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0654323548078537, + "step": 7105, + "valid_targets_mean": 1549.9, + "valid_targets_min": 679 + }, + { + "epoch": 5.452627541235136, + "grad_norm": 0.8517163743354509, + "learning_rate": 5.671181727393467e-06, + "loss": 0.1279, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06445847451686859, + "step": 7110, + "valid_targets_mean": 1656.0, + "valid_targets_min": 1294 + }, + { + "epoch": 5.456463367855773, + "grad_norm": 0.8117674805118457, + "learning_rate": 5.644528386730424e-06, + "loss": 0.1254, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.057402558624744415, + "step": 7115, + "valid_targets_mean": 1368.8, + "valid_targets_min": 790 + }, + { + "epoch": 5.460299194476409, + "grad_norm": 0.8237275102774387, + "learning_rate": 5.617927531883072e-06, + "loss": 0.1254, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06831786036491394, + "step": 7120, + "valid_targets_mean": 1765.8, + "valid_targets_min": 592 + }, + { + "epoch": 5.4641350210970465, + "grad_norm": 0.8561703207865642, + "learning_rate": 5.591379260108214e-06, + "loss": 0.1227, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06770305335521698, + "step": 7125, + "valid_targets_mean": 1586.8, + "valid_targets_min": 662 + }, + { + "epoch": 5.467970847717683, + "grad_norm": 0.9221260632343898, + "learning_rate": 5.564883668470422e-06, + "loss": 0.1225, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.056591033935546875, + "step": 7130, + "valid_targets_mean": 1336.9, + "valid_targets_min": 622 + }, + { + "epoch": 5.47180667433832, + "grad_norm": 0.8869722883069383, + "learning_rate": 5.538440853841682e-06, + "loss": 0.1201, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04110845923423767, + "step": 7135, + "valid_targets_mean": 1097.1, + "valid_targets_min": 625 + }, + { + "epoch": 5.475642500958957, + "grad_norm": 0.9356845323143301, + "learning_rate": 5.512050912900968e-06, + "loss": 0.1338, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05644337087869644, + "step": 7140, + "valid_targets_mean": 1484.9, + "valid_targets_min": 888 + }, + { + "epoch": 5.479478327579593, + "grad_norm": 0.8952620804351472, + "learning_rate": 5.485713942133992e-06, + "loss": 0.132, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0712800920009613, + "step": 7145, + "valid_targets_mean": 1757.0, + "valid_targets_min": 635 + }, + { + "epoch": 5.48331415420023, + "grad_norm": 0.9140231496991617, + "learning_rate": 5.459430037832776e-06, + "loss": 0.1263, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06920260190963745, + "step": 7150, + "valid_targets_mean": 1476.1, + "valid_targets_min": 522 + }, + { + "epoch": 5.487149980820867, + "grad_norm": 0.8277940146766768, + "learning_rate": 5.433199296095302e-06, + "loss": 0.1232, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05237370356917381, + "step": 7155, + "valid_targets_mean": 1230.9, + "valid_targets_min": 697 + }, + { + "epoch": 5.490985807441504, + "grad_norm": 0.8170107109424672, + "learning_rate": 5.40702181282523e-06, + "loss": 0.1193, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0649838000535965, + "step": 7160, + "valid_targets_mean": 1671.0, + "valid_targets_min": 1010 + }, + { + "epoch": 5.4948216340621405, + "grad_norm": 0.8298324214885124, + "learning_rate": 5.380897683731452e-06, + "loss": 0.1189, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05617888644337654, + "step": 7165, + "valid_targets_mean": 1411.0, + "valid_targets_min": 769 + }, + { + "epoch": 5.498657460682777, + "grad_norm": 0.9036101326690158, + "learning_rate": 5.354827004327807e-06, + "loss": 0.1232, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0529860258102417, + "step": 7170, + "valid_targets_mean": 1298.6, + "valid_targets_min": 745 + }, + { + "epoch": 5.502493287303414, + "grad_norm": 0.8816686432394387, + "learning_rate": 5.328809869932736e-06, + "loss": 0.1169, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04725376144051552, + "step": 7175, + "valid_targets_mean": 1118.9, + "valid_targets_min": 680 + }, + { + "epoch": 5.506329113924051, + "grad_norm": 0.8412576685589915, + "learning_rate": 5.302846375668864e-06, + "loss": 0.1195, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0751207023859024, + "step": 7180, + "valid_targets_mean": 1905.9, + "valid_targets_min": 1347 + }, + { + "epoch": 5.510164940544687, + "grad_norm": 0.9212454385967688, + "learning_rate": 5.276936616462744e-06, + "loss": 0.1293, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08577333390712738, + "step": 7185, + "valid_targets_mean": 1636.0, + "valid_targets_min": 851 + }, + { + "epoch": 5.514000767165324, + "grad_norm": 0.822802499506228, + "learning_rate": 5.2510806870444476e-06, + "loss": 0.1207, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05182330310344696, + "step": 7190, + "valid_targets_mean": 1457.2, + "valid_targets_min": 903 + }, + { + "epoch": 5.517836593785961, + "grad_norm": 0.8698859352597403, + "learning_rate": 5.225278681947219e-06, + "loss": 0.1181, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06387019157409668, + "step": 7195, + "valid_targets_mean": 1674.4, + "valid_targets_min": 1240 + }, + { + "epoch": 5.521672420406597, + "grad_norm": 0.7460851476359572, + "learning_rate": 5.19953069550718e-06, + "loss": 0.1213, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.057586152106523514, + "step": 7200, + "valid_targets_mean": 1697.1, + "valid_targets_min": 917 + }, + { + "epoch": 5.5255082470272345, + "grad_norm": 0.7578018870102013, + "learning_rate": 5.173836821862937e-06, + "loss": 0.1088, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05455133318901062, + "step": 7205, + "valid_targets_mean": 1580.9, + "valid_targets_min": 930 + }, + { + "epoch": 5.529344073647871, + "grad_norm": 0.8397426567151561, + "learning_rate": 5.14819715495523e-06, + "loss": 0.1216, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06262408196926117, + "step": 7210, + "valid_targets_mean": 1655.1, + "valid_targets_min": 1227 + }, + { + "epoch": 5.533179900268507, + "grad_norm": 0.853264695722204, + "learning_rate": 5.12261178852665e-06, + "loss": 0.1176, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06053794175386429, + "step": 7215, + "valid_targets_mean": 1565.8, + "valid_targets_min": 870 + }, + { + "epoch": 5.537015726889145, + "grad_norm": 1.0787499843510944, + "learning_rate": 5.097080816121234e-06, + "loss": 0.1236, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06381325423717499, + "step": 7220, + "valid_targets_mean": 1564.6, + "valid_targets_min": 822 + }, + { + "epoch": 5.540851553509781, + "grad_norm": 0.9882978283936427, + "learning_rate": 5.0716043310841455e-06, + "loss": 0.129, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06041494384407997, + "step": 7225, + "valid_targets_mean": 1540.5, + "valid_targets_min": 1214 + }, + { + "epoch": 5.544687380130418, + "grad_norm": 0.8375619235677123, + "learning_rate": 5.046182426561344e-06, + "loss": 0.1161, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05357946455478668, + "step": 7230, + "valid_targets_mean": 1268.6, + "valid_targets_min": 591 + }, + { + "epoch": 5.548523206751055, + "grad_norm": 0.8309058802860864, + "learning_rate": 5.020815195499225e-06, + "loss": 0.1201, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04797566682100296, + "step": 7235, + "valid_targets_mean": 1232.5, + "valid_targets_min": 646 + }, + { + "epoch": 5.552359033371691, + "grad_norm": 1.104041119900565, + "learning_rate": 4.995502730644295e-06, + "loss": 0.1823, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11595666408538818, + "step": 7240, + "valid_targets_mean": 5097.2, + "valid_targets_min": 1971 + }, + { + "epoch": 5.5561948599923285, + "grad_norm": 0.8329673337125996, + "learning_rate": 4.970245124542823e-06, + "loss": 0.224, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11793982982635498, + "step": 7245, + "valid_targets_mean": 5217.0, + "valid_targets_min": 1631 + }, + { + "epoch": 5.560030686612965, + "grad_norm": 0.7313416817351651, + "learning_rate": 4.9450424695405085e-06, + "loss": 0.2266, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14277209341526031, + "step": 7250, + "valid_targets_mean": 5966.0, + "valid_targets_min": 3294 + }, + { + "epoch": 5.563866513233602, + "grad_norm": 0.6403773202285077, + "learning_rate": 4.919894857782141e-06, + "loss": 0.2072, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08462192118167877, + "step": 7255, + "valid_targets_mean": 4246.1, + "valid_targets_min": 2736 + }, + { + "epoch": 5.567702339854239, + "grad_norm": 0.7075621180084197, + "learning_rate": 4.8948023812112655e-06, + "loss": 0.2088, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08801675587892532, + "step": 7260, + "valid_targets_mean": 3896.2, + "valid_targets_min": 1439 + }, + { + "epoch": 5.571538166474875, + "grad_norm": 0.5795154639489792, + "learning_rate": 4.8697651315698365e-06, + "loss": 0.2209, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08651074767112732, + "step": 7265, + "valid_targets_mean": 4481.8, + "valid_targets_min": 1512 + }, + { + "epoch": 5.575373993095512, + "grad_norm": 0.5449553754539438, + "learning_rate": 4.8447832003979e-06, + "loss": 0.2009, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1103343516588211, + "step": 7270, + "valid_targets_mean": 5127.9, + "valid_targets_min": 2405 + }, + { + "epoch": 5.579209819716149, + "grad_norm": 0.5521927223498228, + "learning_rate": 4.819856679033244e-06, + "loss": 0.2135, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0837116539478302, + "step": 7275, + "valid_targets_mean": 3253.8, + "valid_targets_min": 1654 + }, + { + "epoch": 5.583045646336785, + "grad_norm": 0.5009650703364825, + "learning_rate": 4.794985658611069e-06, + "loss": 0.2048, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10188181698322296, + "step": 7280, + "valid_targets_mean": 5750.2, + "valid_targets_min": 4644 + }, + { + "epoch": 5.5868814729574225, + "grad_norm": 0.6003012244599504, + "learning_rate": 4.77017023006366e-06, + "loss": 0.2206, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10166530311107635, + "step": 7285, + "valid_targets_mean": 4141.6, + "valid_targets_min": 1544 + }, + { + "epoch": 5.590717299578059, + "grad_norm": 0.5321817466395727, + "learning_rate": 4.745410484120041e-06, + "loss": 0.2106, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10528967529535294, + "step": 7290, + "valid_targets_mean": 5869.8, + "valid_targets_min": 953 + }, + { + "epoch": 5.594553126198695, + "grad_norm": 0.5673900207047899, + "learning_rate": 4.720706511305664e-06, + "loss": 0.2082, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11161036789417267, + "step": 7295, + "valid_targets_mean": 4973.6, + "valid_targets_min": 1265 + }, + { + "epoch": 5.598388952819333, + "grad_norm": 0.5406657983301922, + "learning_rate": 4.6960584019420565e-06, + "loss": 0.2079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09905244410037994, + "step": 7300, + "valid_targets_mean": 5359.9, + "valid_targets_min": 1982 + }, + { + "epoch": 5.602224779439969, + "grad_norm": 0.5670053237175738, + "learning_rate": 4.671466246146501e-06, + "loss": 0.2086, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0887860655784607, + "step": 7305, + "valid_targets_mean": 3892.5, + "valid_targets_min": 1759 + }, + { + "epoch": 5.606060606060606, + "grad_norm": 0.6505969015392254, + "learning_rate": 4.64693013383171e-06, + "loss": 0.2034, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09527847170829773, + "step": 7310, + "valid_targets_mean": 3185.8, + "valid_targets_min": 1451 + }, + { + "epoch": 5.609896432681243, + "grad_norm": 1.172949072621199, + "learning_rate": 4.622450154705491e-06, + "loss": 0.203, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0987744852900505, + "step": 7315, + "valid_targets_mean": 4283.2, + "valid_targets_min": 1861 + }, + { + "epoch": 5.613732259301879, + "grad_norm": 0.6224570761858262, + "learning_rate": 4.598026398270416e-06, + "loss": 0.1999, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10847309976816177, + "step": 7320, + "valid_targets_mean": 4340.8, + "valid_targets_min": 2345 + }, + { + "epoch": 5.6175680859225166, + "grad_norm": 0.6268318717912089, + "learning_rate": 4.573658953823503e-06, + "loss": 0.2039, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1068105399608612, + "step": 7325, + "valid_targets_mean": 3923.9, + "valid_targets_min": 2767 + }, + { + "epoch": 5.621403912543153, + "grad_norm": 0.6702032899149752, + "learning_rate": 4.5493479104558835e-06, + "loss": 0.2107, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10821166634559631, + "step": 7330, + "valid_targets_mean": 3651.8, + "valid_targets_min": 1205 + }, + { + "epoch": 5.62523973916379, + "grad_norm": 0.5978333239287432, + "learning_rate": 4.5250933570524725e-06, + "loss": 0.202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08542361855506897, + "step": 7335, + "valid_targets_mean": 3835.4, + "valid_targets_min": 2004 + }, + { + "epoch": 5.629075565784427, + "grad_norm": 0.6314598131578465, + "learning_rate": 4.500895382291659e-06, + "loss": 0.2077, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.098983995616436, + "step": 7340, + "valid_targets_mean": 3778.0, + "valid_targets_min": 1020 + }, + { + "epoch": 5.632911392405063, + "grad_norm": 0.6442051461800672, + "learning_rate": 4.476754074644965e-06, + "loss": 0.1903, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1139015406370163, + "step": 7345, + "valid_targets_mean": 4307.8, + "valid_targets_min": 2308 + }, + { + "epoch": 5.6367472190257, + "grad_norm": 0.5992599419334531, + "learning_rate": 4.452669522376729e-06, + "loss": 0.191, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10413465648889542, + "step": 7350, + "valid_targets_mean": 4553.9, + "valid_targets_min": 2046 + }, + { + "epoch": 5.640583045646337, + "grad_norm": 0.6617597826137858, + "learning_rate": 4.428641813543779e-06, + "loss": 0.191, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1292639672756195, + "step": 7355, + "valid_targets_mean": 4660.2, + "valid_targets_min": 3373 + }, + { + "epoch": 5.644418872266973, + "grad_norm": 0.7122682709033852, + "learning_rate": 4.404671035995136e-06, + "loss": 0.2013, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09767237305641174, + "step": 7360, + "valid_targets_mean": 3727.4, + "valid_targets_min": 1747 + }, + { + "epoch": 5.648254698887611, + "grad_norm": 0.6673468561809532, + "learning_rate": 4.380757277371641e-06, + "loss": 0.2024, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09222052246332169, + "step": 7365, + "valid_targets_mean": 3326.1, + "valid_targets_min": 1570 + }, + { + "epoch": 5.652090525508247, + "grad_norm": 0.6382775656715328, + "learning_rate": 4.35690062510568e-06, + "loss": 0.1898, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09534966945648193, + "step": 7370, + "valid_targets_mean": 4600.8, + "valid_targets_min": 1537 + }, + { + "epoch": 5.655926352128883, + "grad_norm": 0.6141496369675626, + "learning_rate": 4.333101166420861e-06, + "loss": 0.1923, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08670054376125336, + "step": 7375, + "valid_targets_mean": 4328.9, + "valid_targets_min": 1934 + }, + { + "epoch": 5.659762178749521, + "grad_norm": 0.6792927801991392, + "learning_rate": 4.309358988331658e-06, + "loss": 0.197, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0882268100976944, + "step": 7380, + "valid_targets_mean": 3494.0, + "valid_targets_min": 2562 + }, + { + "epoch": 5.663598005370157, + "grad_norm": 0.651753582257499, + "learning_rate": 4.285674177643131e-06, + "loss": 0.1958, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0983646884560585, + "step": 7385, + "valid_targets_mean": 4076.8, + "valid_targets_min": 2659 + }, + { + "epoch": 5.6674338319907935, + "grad_norm": 0.609802029232181, + "learning_rate": 4.2620468209506116e-06, + "loss": 0.1896, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08816227316856384, + "step": 7390, + "valid_targets_mean": 4040.6, + "valid_targets_min": 1349 + }, + { + "epoch": 5.671269658611431, + "grad_norm": 0.6576699535180179, + "learning_rate": 4.238477004639336e-06, + "loss": 0.193, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07614319026470184, + "step": 7395, + "valid_targets_mean": 3000.6, + "valid_targets_min": 1567 + }, + { + "epoch": 5.675105485232067, + "grad_norm": 0.6872903531734094, + "learning_rate": 4.214964814884204e-06, + "loss": 0.1981, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07976074516773224, + "step": 7400, + "valid_targets_mean": 2628.4, + "valid_targets_min": 1500 + }, + { + "epoch": 5.678941311852705, + "grad_norm": 0.6067959604467803, + "learning_rate": 4.191510337649389e-06, + "loss": 0.1856, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08238235116004944, + "step": 7405, + "valid_targets_mean": 3413.6, + "valid_targets_min": 2033 + }, + { + "epoch": 5.682777138473341, + "grad_norm": 0.7975484158418957, + "learning_rate": 4.168113658688069e-06, + "loss": 0.1962, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11310118436813354, + "step": 7410, + "valid_targets_mean": 2852.0, + "valid_targets_min": 1346 + }, + { + "epoch": 5.686612965093977, + "grad_norm": 0.7825583136487398, + "learning_rate": 4.144774863542127e-06, + "loss": 0.1958, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11199402064085007, + "step": 7415, + "valid_targets_mean": 2981.6, + "valid_targets_min": 1821 + }, + { + "epoch": 5.690448791714615, + "grad_norm": 0.6603202617122221, + "learning_rate": 4.1214940375417755e-06, + "loss": 0.1927, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10025741159915924, + "step": 7420, + "valid_targets_mean": 3951.2, + "valid_targets_min": 2535 + }, + { + "epoch": 5.694284618335251, + "grad_norm": 0.8531105550595132, + "learning_rate": 4.098271265805298e-06, + "loss": 0.1953, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0924755409359932, + "step": 7425, + "valid_targets_mean": 3048.8, + "valid_targets_min": 2080 + }, + { + "epoch": 5.698120444955888, + "grad_norm": 0.6854471350424085, + "learning_rate": 4.075106633238748e-06, + "loss": 0.1904, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08661817014217377, + "step": 7430, + "valid_targets_mean": 2975.6, + "valid_targets_min": 1153 + }, + { + "epoch": 5.701956271576525, + "grad_norm": 0.6427713810889253, + "learning_rate": 4.0520002245355614e-06, + "loss": 0.1929, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08910036832094193, + "step": 7435, + "valid_targets_mean": 3207.6, + "valid_targets_min": 1858 + }, + { + "epoch": 5.705792098197161, + "grad_norm": 0.6556379940343808, + "learning_rate": 4.028952124176351e-06, + "loss": 0.1929, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08798359334468842, + "step": 7440, + "valid_targets_mean": 3643.6, + "valid_targets_min": 2197 + }, + { + "epoch": 5.709627924817799, + "grad_norm": 0.6311112325256589, + "learning_rate": 4.005962416428519e-06, + "loss": 0.1905, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07398124039173126, + "step": 7445, + "valid_targets_mean": 3769.4, + "valid_targets_min": 1394 + }, + { + "epoch": 5.713463751438435, + "grad_norm": 0.6507680559823559, + "learning_rate": 3.983031185345963e-06, + "loss": 0.1956, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10038575530052185, + "step": 7450, + "valid_targets_mean": 4348.6, + "valid_targets_min": 1844 + }, + { + "epoch": 5.717299578059071, + "grad_norm": 0.6666031086824297, + "learning_rate": 3.96015851476881e-06, + "loss": 0.1895, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1093379408121109, + "step": 7455, + "valid_targets_mean": 4341.1, + "valid_targets_min": 1734 + }, + { + "epoch": 5.721135404679709, + "grad_norm": 0.6707770034839663, + "learning_rate": 3.937344488323067e-06, + "loss": 0.1906, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10724876821041107, + "step": 7460, + "valid_targets_mean": 4252.6, + "valid_targets_min": 2077 + }, + { + "epoch": 5.724971231300345, + "grad_norm": 0.6255157071962757, + "learning_rate": 3.914589189420323e-06, + "loss": 0.1879, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09555348753929138, + "step": 7465, + "valid_targets_mean": 4177.9, + "valid_targets_min": 2349 + }, + { + "epoch": 5.7288070579209815, + "grad_norm": 0.6516226326878923, + "learning_rate": 3.891892701257463e-06, + "loss": 0.1897, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10840095579624176, + "step": 7470, + "valid_targets_mean": 4160.2, + "valid_targets_min": 1607 + }, + { + "epoch": 5.732642884541619, + "grad_norm": 0.7168120872662032, + "learning_rate": 3.869255106816339e-06, + "loss": 0.1957, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10281391441822052, + "step": 7475, + "valid_targets_mean": 3273.0, + "valid_targets_min": 1377 + }, + { + "epoch": 5.736478711162255, + "grad_norm": 0.6278103404696428, + "learning_rate": 3.846676488863483e-06, + "loss": 0.1944, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08866104483604431, + "step": 7480, + "valid_targets_mean": 3304.5, + "valid_targets_min": 1902 + }, + { + "epoch": 5.740314537782893, + "grad_norm": 0.6384605567195086, + "learning_rate": 3.824156929949809e-06, + "loss": 0.1942, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.086397685110569, + "step": 7485, + "valid_targets_mean": 3324.1, + "valid_targets_min": 654 + }, + { + "epoch": 5.744150364403529, + "grad_norm": 0.6597945056765726, + "learning_rate": 3.8016965124102778e-06, + "loss": 0.1974, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09828825294971466, + "step": 7490, + "valid_targets_mean": 3907.8, + "valid_targets_min": 2607 + }, + { + "epoch": 5.747986191024165, + "grad_norm": 0.6951110580457823, + "learning_rate": 3.7792953183636517e-06, + "loss": 0.1943, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10247915983200073, + "step": 7495, + "valid_targets_mean": 4467.2, + "valid_targets_min": 1739 + }, + { + "epoch": 5.751822017644803, + "grad_norm": 0.6712681459456963, + "learning_rate": 3.756953429712138e-06, + "loss": 0.1912, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0847967267036438, + "step": 7500, + "valid_targets_mean": 3224.1, + "valid_targets_min": 1140 + }, + { + "epoch": 5.755657844265439, + "grad_norm": 0.6359884200259689, + "learning_rate": 3.7346709281411263e-06, + "loss": 0.1822, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0914197564125061, + "step": 7505, + "valid_targets_mean": 3543.4, + "valid_targets_min": 2501 + }, + { + "epoch": 5.759493670886076, + "grad_norm": 0.6412336305732639, + "learning_rate": 3.712447895118876e-06, + "loss": 0.1918, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09600517898797989, + "step": 7510, + "valid_targets_mean": 4364.1, + "valid_targets_min": 1298 + }, + { + "epoch": 5.763329497506713, + "grad_norm": 0.6322292877280051, + "learning_rate": 3.690284411896219e-06, + "loss": 0.1776, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10184262692928314, + "step": 7515, + "valid_targets_mean": 4377.9, + "valid_targets_min": 1101 + }, + { + "epoch": 5.767165324127349, + "grad_norm": 0.6563232405010566, + "learning_rate": 3.6681805595062603e-06, + "loss": 0.2051, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10793612897396088, + "step": 7520, + "valid_targets_mean": 3894.2, + "valid_targets_min": 2079 + }, + { + "epoch": 5.771001150747987, + "grad_norm": 0.6454345937014524, + "learning_rate": 3.6461364187640948e-06, + "loss": 0.1928, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08285500854253769, + "step": 7525, + "valid_targets_mean": 3054.4, + "valid_targets_min": 1743 + }, + { + "epoch": 5.774836977368623, + "grad_norm": 0.6236622466184732, + "learning_rate": 3.6241520702664913e-06, + "loss": 0.1968, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10124188661575317, + "step": 7530, + "valid_targets_mean": 3688.0, + "valid_targets_min": 1234 + }, + { + "epoch": 5.778672803989259, + "grad_norm": 0.6811684985024322, + "learning_rate": 3.6022275943916205e-06, + "loss": 0.1982, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10320290923118591, + "step": 7535, + "valid_targets_mean": 3790.0, + "valid_targets_min": 1822 + }, + { + "epoch": 5.782508630609897, + "grad_norm": 0.6341916447300334, + "learning_rate": 3.58036307129874e-06, + "loss": 0.1989, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09771846234798431, + "step": 7540, + "valid_targets_mean": 4106.9, + "valid_targets_min": 1651 + }, + { + "epoch": 5.786344457230533, + "grad_norm": 0.674549329408804, + "learning_rate": 3.5585585809279155e-06, + "loss": 0.1944, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10214784741401672, + "step": 7545, + "valid_targets_mean": 3287.2, + "valid_targets_min": 1374 + }, + { + "epoch": 5.79018028385117, + "grad_norm": 0.6484070262638019, + "learning_rate": 3.536814202999723e-06, + "loss": 0.1839, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09671390056610107, + "step": 7550, + "valid_targets_mean": 4399.2, + "valid_targets_min": 2725 + }, + { + "epoch": 5.794016110471807, + "grad_norm": 0.6154673020204012, + "learning_rate": 3.515130017014956e-06, + "loss": 0.1871, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10245229303836823, + "step": 7555, + "valid_targets_mean": 4269.0, + "valid_targets_min": 1966 + }, + { + "epoch": 5.797851937092443, + "grad_norm": 0.725461141128357, + "learning_rate": 3.4935061022543403e-06, + "loss": 0.1838, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1014595702290535, + "step": 7560, + "valid_targets_mean": 3716.2, + "valid_targets_min": 1561 + }, + { + "epoch": 5.80168776371308, + "grad_norm": 0.7358066848693222, + "learning_rate": 3.4719425377782414e-06, + "loss": 0.1861, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10293035209178925, + "step": 7565, + "valid_targets_mean": 3750.1, + "valid_targets_min": 1444 + }, + { + "epoch": 5.805523590333717, + "grad_norm": 0.6353428522366301, + "learning_rate": 3.450439402426371e-06, + "loss": 0.1732, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07647493481636047, + "step": 7570, + "valid_targets_mean": 3042.1, + "valid_targets_min": 1504 + }, + { + "epoch": 5.809359416954353, + "grad_norm": 0.6914661368238604, + "learning_rate": 3.4289967748175033e-06, + "loss": 0.1736, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08658996224403381, + "step": 7575, + "valid_targets_mean": 3153.0, + "valid_targets_min": 858 + }, + { + "epoch": 5.813195243574991, + "grad_norm": 0.7445235127634592, + "learning_rate": 3.4076147333491895e-06, + "loss": 0.189, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10444004088640213, + "step": 7580, + "valid_targets_mean": 4063.5, + "valid_targets_min": 1288 + }, + { + "epoch": 5.817031070195627, + "grad_norm": 0.6809538433896142, + "learning_rate": 3.3862933561974676e-06, + "loss": 0.2079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08961853384971619, + "step": 7585, + "valid_targets_mean": 3914.9, + "valid_targets_min": 1537 + }, + { + "epoch": 5.820866896816264, + "grad_norm": 0.6910807595884583, + "learning_rate": 3.365032721316577e-06, + "loss": 0.184, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10179746150970459, + "step": 7590, + "valid_targets_mean": 4236.5, + "valid_targets_min": 2972 + }, + { + "epoch": 5.824702723436901, + "grad_norm": 0.6887530670932189, + "learning_rate": 3.343832906438671e-06, + "loss": 0.1962, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10917111486196518, + "step": 7595, + "valid_targets_mean": 4411.0, + "valid_targets_min": 2013 + }, + { + "epoch": 5.828538550057537, + "grad_norm": 0.694880102234136, + "learning_rate": 3.322693989073542e-06, + "loss": 0.1865, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09710349142551422, + "step": 7600, + "valid_targets_mean": 3548.4, + "valid_targets_min": 1744 + }, + { + "epoch": 5.832374376678175, + "grad_norm": 0.6577193659616001, + "learning_rate": 3.3016160465083293e-06, + "loss": 0.1957, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08475316315889359, + "step": 7605, + "valid_targets_mean": 4129.6, + "valid_targets_min": 1914 + }, + { + "epoch": 5.836210203298811, + "grad_norm": 0.7066622371392409, + "learning_rate": 3.280599155807229e-06, + "loss": 0.1879, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08015590906143188, + "step": 7610, + "valid_targets_mean": 3826.4, + "valid_targets_min": 2414 + }, + { + "epoch": 5.840046029919447, + "grad_norm": 0.6486414139336713, + "learning_rate": 3.2596433938112495e-06, + "loss": 0.1798, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10421285778284073, + "step": 7615, + "valid_targets_mean": 4195.0, + "valid_targets_min": 1937 + }, + { + "epoch": 5.843881856540085, + "grad_norm": 0.7075901864287605, + "learning_rate": 3.238748837137866e-06, + "loss": 0.1961, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10379116237163544, + "step": 7620, + "valid_targets_mean": 3154.9, + "valid_targets_min": 1649 + }, + { + "epoch": 5.847717683160721, + "grad_norm": 0.652580192628628, + "learning_rate": 3.217915562180802e-06, + "loss": 0.1819, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0820373147726059, + "step": 7625, + "valid_targets_mean": 3088.5, + "valid_targets_min": 1463 + }, + { + "epoch": 5.851553509781358, + "grad_norm": 0.6900303344047586, + "learning_rate": 3.1971436451097302e-06, + "loss": 0.1928, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09511121362447739, + "step": 7630, + "valid_targets_mean": 3241.4, + "valid_targets_min": 1426 + }, + { + "epoch": 5.855389336401995, + "grad_norm": 0.73804211385738, + "learning_rate": 3.1764331618699673e-06, + "loss": 0.1977, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08007380366325378, + "step": 7635, + "valid_targets_mean": 2768.9, + "valid_targets_min": 538 + }, + { + "epoch": 5.859225163022631, + "grad_norm": 0.6084221093034275, + "learning_rate": 3.155784188182229e-06, + "loss": 0.1957, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10092423856258392, + "step": 7640, + "valid_targets_mean": 4483.5, + "valid_targets_min": 1646 + }, + { + "epoch": 5.863060989643268, + "grad_norm": 0.7498049669500961, + "learning_rate": 3.1351967995423594e-06, + "loss": 0.2079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1169540286064148, + "step": 7645, + "valid_targets_mean": 3761.9, + "valid_targets_min": 2169 + }, + { + "epoch": 5.866896816263905, + "grad_norm": 0.6917681712694926, + "learning_rate": 3.114671071221005e-06, + "loss": 0.1901, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08705845475196838, + "step": 7650, + "valid_targets_mean": 3168.4, + "valid_targets_min": 1625 + }, + { + "epoch": 5.870732642884541, + "grad_norm": 0.6670330223500122, + "learning_rate": 3.094207078263405e-06, + "loss": 0.2057, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08457347750663757, + "step": 7655, + "valid_targets_mean": 3449.0, + "valid_targets_min": 1997 + }, + { + "epoch": 5.874568469505179, + "grad_norm": 0.6728809726376211, + "learning_rate": 3.073804895489061e-06, + "loss": 0.2007, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11987364292144775, + "step": 7660, + "valid_targets_mean": 4406.4, + "valid_targets_min": 2949 + }, + { + "epoch": 5.878404296125815, + "grad_norm": 0.5924210765709458, + "learning_rate": 3.0534645974914933e-06, + "loss": 0.1972, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10976015031337738, + "step": 7665, + "valid_targets_mean": 5358.5, + "valid_targets_min": 2652 + }, + { + "epoch": 5.882240122746452, + "grad_norm": 0.6533311602560773, + "learning_rate": 3.0331862586379813e-06, + "loss": 0.2063, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11076603084802628, + "step": 7670, + "valid_targets_mean": 4314.5, + "valid_targets_min": 1381 + }, + { + "epoch": 5.886075949367089, + "grad_norm": 2.03958596415384, + "learning_rate": 3.0129699530692335e-06, + "loss": 0.1943, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08127956837415695, + "step": 7675, + "valid_targets_mean": 3395.9, + "valid_targets_min": 637 + }, + { + "epoch": 5.889911775987725, + "grad_norm": 0.6761663011117922, + "learning_rate": 2.992815754699194e-06, + "loss": 0.197, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08665677905082703, + "step": 7680, + "valid_targets_mean": 3256.4, + "valid_targets_min": 719 + }, + { + "epoch": 5.893747602608362, + "grad_norm": 0.6360005208006169, + "learning_rate": 2.9727237372147177e-06, + "loss": 0.2011, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08652372658252716, + "step": 7685, + "valid_targets_mean": 4725.1, + "valid_targets_min": 3210 + }, + { + "epoch": 5.897583429228999, + "grad_norm": 0.6873841560077597, + "learning_rate": 2.952693974075298e-06, + "loss": 0.1893, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10174581408500671, + "step": 7690, + "valid_targets_mean": 3775.4, + "valid_targets_min": 1372 + }, + { + "epoch": 5.9014192558496354, + "grad_norm": 0.6903117964836543, + "learning_rate": 2.932726538512851e-06, + "loss": 0.185, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11197394877672195, + "step": 7695, + "valid_targets_mean": 4266.1, + "valid_targets_min": 2768 + }, + { + "epoch": 5.905255082470273, + "grad_norm": 0.6596333497824657, + "learning_rate": 2.9128215035313976e-06, + "loss": 0.1942, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07489095628261566, + "step": 7700, + "valid_targets_mean": 2540.5, + "valid_targets_min": 530 + }, + { + "epoch": 5.909090909090909, + "grad_norm": 0.6820749068178948, + "learning_rate": 2.8929789419067964e-06, + "loss": 0.1895, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10009552538394928, + "step": 7705, + "valid_targets_mean": 3903.2, + "valid_targets_min": 1841 + }, + { + "epoch": 5.912926735711546, + "grad_norm": 0.709752507095235, + "learning_rate": 2.8731989261865247e-06, + "loss": 0.2028, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09172187000513077, + "step": 7710, + "valid_targets_mean": 3476.8, + "valid_targets_min": 1728 + }, + { + "epoch": 5.916762562332183, + "grad_norm": 0.7293326718685008, + "learning_rate": 2.8534815286893614e-06, + "loss": 0.194, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10897614806890488, + "step": 7715, + "valid_targets_mean": 3712.2, + "valid_targets_min": 1362 + }, + { + "epoch": 5.920598388952819, + "grad_norm": 0.6291454717542756, + "learning_rate": 2.83382682150515e-06, + "loss": 0.1978, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09896582365036011, + "step": 7720, + "valid_targets_mean": 3693.8, + "valid_targets_min": 852 + }, + { + "epoch": 5.924434215573456, + "grad_norm": 0.708303841282401, + "learning_rate": 2.8142348764945325e-06, + "loss": 0.1948, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08722059428691864, + "step": 7725, + "valid_targets_mean": 2822.6, + "valid_targets_min": 1195 + }, + { + "epoch": 5.928270042194093, + "grad_norm": 0.7313558929412427, + "learning_rate": 2.7947057652886635e-06, + "loss": 0.1977, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1206844374537468, + "step": 7730, + "valid_targets_mean": 4747.8, + "valid_targets_min": 1397 + }, + { + "epoch": 5.9321058688147295, + "grad_norm": 0.7234310521253834, + "learning_rate": 2.775239559288996e-06, + "loss": 0.1921, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07865928113460541, + "step": 7735, + "valid_targets_mean": 3487.6, + "valid_targets_min": 1975 + }, + { + "epoch": 5.935941695435366, + "grad_norm": 0.7200261893839338, + "learning_rate": 2.7558363296669786e-06, + "loss": 0.1835, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10389940440654755, + "step": 7740, + "valid_targets_mean": 4387.6, + "valid_targets_min": 2564 + }, + { + "epoch": 5.939777522056003, + "grad_norm": 0.6397721123750132, + "learning_rate": 2.736496147363792e-06, + "loss": 0.1933, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09813384711742401, + "step": 7745, + "valid_targets_mean": 6404.1, + "valid_targets_min": 2595 + }, + { + "epoch": 5.94361334867664, + "grad_norm": 0.6489382892757688, + "learning_rate": 2.717219083090139e-06, + "loss": 0.1918, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10042206943035126, + "step": 7750, + "valid_targets_mean": 4013.2, + "valid_targets_min": 2868 + }, + { + "epoch": 5.947449175297277, + "grad_norm": 0.8185108755868531, + "learning_rate": 2.6980052073259313e-06, + "loss": 0.1835, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08330656588077545, + "step": 7755, + "valid_targets_mean": 2886.0, + "valid_targets_min": 993 + }, + { + "epoch": 5.951285001917913, + "grad_norm": 0.6760974377201443, + "learning_rate": 2.6788545903200545e-06, + "loss": 0.1827, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0883827731013298, + "step": 7760, + "valid_targets_mean": 3805.1, + "valid_targets_min": 1575 + }, + { + "epoch": 5.95512082853855, + "grad_norm": 0.6795297063354218, + "learning_rate": 2.6597673020901193e-06, + "loss": 0.1867, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10669929534196854, + "step": 7765, + "valid_targets_mean": 3722.0, + "valid_targets_min": 1360 + }, + { + "epoch": 5.958956655159187, + "grad_norm": 0.6594449775853662, + "learning_rate": 2.6407434124221887e-06, + "loss": 0.1838, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08820085227489471, + "step": 7770, + "valid_targets_mean": 3669.4, + "valid_targets_min": 2414 + }, + { + "epoch": 5.9627924817798235, + "grad_norm": 0.6951110678890519, + "learning_rate": 2.621782990870536e-06, + "loss": 0.2048, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11889301240444183, + "step": 7775, + "valid_targets_mean": 4327.4, + "valid_targets_min": 1454 + }, + { + "epoch": 5.966628308400461, + "grad_norm": 0.6629867810045232, + "learning_rate": 2.6028861067573807e-06, + "loss": 0.1913, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09352666139602661, + "step": 7780, + "valid_targets_mean": 3619.9, + "valid_targets_min": 1720 + }, + { + "epoch": 5.970464135021097, + "grad_norm": 0.7377249362426833, + "learning_rate": 2.584052829172645e-06, + "loss": 0.1882, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10007035732269287, + "step": 7785, + "valid_targets_mean": 3105.5, + "valid_targets_min": 1278 + }, + { + "epoch": 5.974299961641734, + "grad_norm": 0.7398388388869782, + "learning_rate": 2.5652832269736916e-06, + "loss": 0.1808, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10567589104175568, + "step": 7790, + "valid_targets_mean": 3147.8, + "valid_targets_min": 1858 + }, + { + "epoch": 5.978135788262371, + "grad_norm": 0.7051109957881178, + "learning_rate": 2.5465773687850792e-06, + "loss": 0.1899, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07767666131258011, + "step": 7795, + "valid_targets_mean": 2984.5, + "valid_targets_min": 628 + }, + { + "epoch": 5.981971614883007, + "grad_norm": 0.7339184622642564, + "learning_rate": 2.527935322998305e-06, + "loss": 0.1896, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10654906183481216, + "step": 7800, + "valid_targets_mean": 2602.6, + "valid_targets_min": 1039 + }, + { + "epoch": 5.985807441503644, + "grad_norm": 0.6776791774111509, + "learning_rate": 2.5093571577715593e-06, + "loss": 0.1983, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09552404284477234, + "step": 7805, + "valid_targets_mean": 4317.8, + "valid_targets_min": 2530 + }, + { + "epoch": 5.989643268124281, + "grad_norm": 0.6852431569013, + "learning_rate": 2.4908429410294786e-06, + "loss": 0.1936, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10980413109064102, + "step": 7810, + "valid_targets_mean": 3711.0, + "valid_targets_min": 809 + }, + { + "epoch": 5.9934790947449175, + "grad_norm": 0.6486778437674778, + "learning_rate": 2.4723927404628922e-06, + "loss": 0.1925, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0823063850402832, + "step": 7815, + "valid_targets_mean": 3450.5, + "valid_targets_min": 1780 + }, + { + "epoch": 5.997314921365554, + "grad_norm": 0.6572630091351028, + "learning_rate": 2.4540066235285733e-06, + "loss": 0.201, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10578109323978424, + "step": 7820, + "valid_targets_mean": 4356.6, + "valid_targets_min": 2185 + }, + { + "epoch": 6.001534330648255, + "grad_norm": 0.6427767410004253, + "learning_rate": 2.4356846574489977e-06, + "loss": 0.2057, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08114808052778244, + "step": 7825, + "valid_targets_mean": 8514.0, + "valid_targets_min": 6788 + }, + { + "epoch": 6.005370157268891, + "grad_norm": 0.6282791867341196, + "learning_rate": 2.417426909212095e-06, + "loss": 0.1814, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08914591372013092, + "step": 7830, + "valid_targets_mean": 6668.0, + "valid_targets_min": 5933 + }, + { + "epoch": 6.009205983889528, + "grad_norm": 0.620912453726633, + "learning_rate": 2.39923344557101e-06, + "loss": 0.1646, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07544026523828506, + "step": 7835, + "valid_targets_mean": 4107.0, + "valid_targets_min": 161 + }, + { + "epoch": 6.013041810510165, + "grad_norm": 0.46723987583589055, + "learning_rate": 2.3811043330438444e-06, + "loss": 0.1722, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0770735889673233, + "step": 7840, + "valid_targets_mean": 7946.0, + "valid_targets_min": 5596 + }, + { + "epoch": 6.0168776371308015, + "grad_norm": 0.4963501773612917, + "learning_rate": 2.363039637913427e-06, + "loss": 0.1655, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08916376531124115, + "step": 7845, + "valid_targets_mean": 7392.8, + "valid_targets_min": 5629 + }, + { + "epoch": 6.020713463751439, + "grad_norm": 0.46726227494970357, + "learning_rate": 2.345039426227067e-06, + "loss": 0.168, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08604732900857925, + "step": 7850, + "valid_targets_mean": 6879.6, + "valid_targets_min": 5113 + }, + { + "epoch": 6.024549290372075, + "grad_norm": 0.47383855125389657, + "learning_rate": 2.32710376379631e-06, + "loss": 0.1821, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08538611978292465, + "step": 7855, + "valid_targets_mean": 6919.5, + "valid_targets_min": 5080 + }, + { + "epoch": 6.028385116992712, + "grad_norm": 0.8020113123220727, + "learning_rate": 2.309232716196701e-06, + "loss": 0.1737, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07622009515762329, + "step": 7860, + "valid_targets_mean": 6377.0, + "valid_targets_min": 3939 + }, + { + "epoch": 6.032220943613349, + "grad_norm": 0.44316681173800093, + "learning_rate": 2.2914263487675402e-06, + "loss": 0.1643, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08276666700839996, + "step": 7865, + "valid_targets_mean": 6392.2, + "valid_targets_min": 4355 + }, + { + "epoch": 6.036056770233985, + "grad_norm": 0.5684229024573472, + "learning_rate": 2.273684726611662e-06, + "loss": 0.1484, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08441920578479767, + "step": 7870, + "valid_targets_mean": 5196.5, + "valid_targets_min": 137 + }, + { + "epoch": 6.039892596854622, + "grad_norm": 0.4524552693830371, + "learning_rate": 2.25600791459516e-06, + "loss": 0.1632, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07400171458721161, + "step": 7875, + "valid_targets_mean": 7344.4, + "valid_targets_min": 5244 + }, + { + "epoch": 6.043728423475259, + "grad_norm": 0.4234623340662572, + "learning_rate": 2.2383959773471765e-06, + "loss": 0.1582, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06870976835489273, + "step": 7880, + "valid_targets_mean": 7921.0, + "valid_targets_min": 5311 + }, + { + "epoch": 6.0475642500958955, + "grad_norm": 0.4607842392031124, + "learning_rate": 2.220848979259682e-06, + "loss": 0.167, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08654425293207169, + "step": 7885, + "valid_targets_mean": 7931.2, + "valid_targets_min": 6015 + }, + { + "epoch": 6.051400076716533, + "grad_norm": 0.4062273211957822, + "learning_rate": 2.2033669844871897e-06, + "loss": 0.1583, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07637645304203033, + "step": 7890, + "valid_targets_mean": 7640.8, + "valid_targets_min": 5248 + }, + { + "epoch": 6.055235903337169, + "grad_norm": 0.4179073534592395, + "learning_rate": 2.1859500569465818e-06, + "loss": 0.1571, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0793628990650177, + "step": 7895, + "valid_targets_mean": 7249.6, + "valid_targets_min": 5217 + }, + { + "epoch": 6.059071729957806, + "grad_norm": 0.44554314465653866, + "learning_rate": 2.1685982603168144e-06, + "loss": 0.1636, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09087036550045013, + "step": 7900, + "valid_targets_mean": 7619.9, + "valid_targets_min": 5094 + }, + { + "epoch": 6.062907556578443, + "grad_norm": 0.4205854476446412, + "learning_rate": 2.1513116580387304e-06, + "loss": 0.1586, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07565924525260925, + "step": 7905, + "valid_targets_mean": 7732.8, + "valid_targets_min": 5842 + }, + { + "epoch": 6.066743383199079, + "grad_norm": 0.4124860945972981, + "learning_rate": 2.1340903133148205e-06, + "loss": 0.1613, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08232372999191284, + "step": 7910, + "valid_targets_mean": 7952.9, + "valid_targets_min": 5494 + }, + { + "epoch": 6.070579209819716, + "grad_norm": 0.4459180196091079, + "learning_rate": 2.116934289108967e-06, + "loss": 0.1726, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08711044490337372, + "step": 7915, + "valid_targets_mean": 8001.0, + "valid_targets_min": 5454 + }, + { + "epoch": 6.074415036440353, + "grad_norm": 0.6259661067270786, + "learning_rate": 2.0998436481462315e-06, + "loss": 0.1791, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05852557718753815, + "step": 7920, + "valid_targets_mean": 2985.1, + "valid_targets_min": 1896 + }, + { + "epoch": 6.0782508630609895, + "grad_norm": 0.4418175411787294, + "learning_rate": 2.0828184529126473e-06, + "loss": 0.1333, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07027403265237808, + "step": 7925, + "valid_targets_mean": 6714.6, + "valid_targets_min": 4917 + }, + { + "epoch": 6.082086689681627, + "grad_norm": 0.49242330070355295, + "learning_rate": 2.065858765654931e-06, + "loss": 0.1587, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08668597042560577, + "step": 7930, + "valid_targets_mean": 7595.2, + "valid_targets_min": 5488 + }, + { + "epoch": 6.085922516302263, + "grad_norm": 0.4475096057027065, + "learning_rate": 2.048964648380325e-06, + "loss": 0.1576, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08579195290803909, + "step": 7935, + "valid_targets_mean": 7046.1, + "valid_targets_min": 5579 + }, + { + "epoch": 6.0897583429229, + "grad_norm": 0.4627632503584522, + "learning_rate": 2.0321361628563263e-06, + "loss": 0.1636, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07864826917648315, + "step": 7940, + "valid_targets_mean": 7108.0, + "valid_targets_min": 4963 + }, + { + "epoch": 6.093594169543537, + "grad_norm": 0.445546558784828, + "learning_rate": 2.015373370610456e-06, + "loss": 0.1618, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08102001994848251, + "step": 7945, + "valid_targets_mean": 6810.1, + "valid_targets_min": 5666 + }, + { + "epoch": 6.097429996164173, + "grad_norm": 0.4226881880327569, + "learning_rate": 1.9986763329300783e-06, + "loss": 0.1545, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07496386766433716, + "step": 7950, + "valid_targets_mean": 8068.5, + "valid_targets_min": 5524 + }, + { + "epoch": 6.10126582278481, + "grad_norm": 0.42487725151414046, + "learning_rate": 1.982045110862134e-06, + "loss": 0.1408, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06677807867527008, + "step": 7955, + "valid_targets_mean": 6553.8, + "valid_targets_min": 3621 + }, + { + "epoch": 6.105101649405447, + "grad_norm": 0.43600721313881335, + "learning_rate": 1.965479765212921e-06, + "loss": 0.1445, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07460106164216995, + "step": 7960, + "valid_targets_mean": 6762.5, + "valid_targets_min": 4968 + }, + { + "epoch": 6.1089374760260835, + "grad_norm": 0.47412154418161534, + "learning_rate": 1.9489803565479094e-06, + "loss": 0.1548, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07756656408309937, + "step": 7965, + "valid_targets_mean": 6143.9, + "valid_targets_min": 4917 + }, + { + "epoch": 6.112773302646721, + "grad_norm": 0.5545668440973357, + "learning_rate": 1.9325469451914714e-06, + "loss": 0.1432, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10113325715065002, + "step": 7970, + "valid_targets_mean": 4583.8, + "valid_targets_min": 215 + }, + { + "epoch": 6.116609129267357, + "grad_norm": 0.432636616094707, + "learning_rate": 1.9161795912266945e-06, + "loss": 0.1647, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08090570569038391, + "step": 7975, + "valid_targets_mean": 7317.1, + "valid_targets_min": 4948 + }, + { + "epoch": 6.120444955887994, + "grad_norm": 0.5362669522181512, + "learning_rate": 1.899878354495146e-06, + "loss": 0.1761, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08241428434848785, + "step": 7980, + "valid_targets_mean": 6437.6, + "valid_targets_min": 4510 + }, + { + "epoch": 6.124280782508631, + "grad_norm": 0.478301523152336, + "learning_rate": 1.8836432945966487e-06, + "loss": 0.1693, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09371457993984222, + "step": 7985, + "valid_targets_mean": 7707.9, + "valid_targets_min": 5796 + }, + { + "epoch": 6.128116609129267, + "grad_norm": 0.4056952398949876, + "learning_rate": 1.8674744708890969e-06, + "loss": 0.16, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07876624166965485, + "step": 7990, + "valid_targets_mean": 7900.9, + "valid_targets_min": 5643 + }, + { + "epoch": 6.131952435749904, + "grad_norm": 0.4943552184089174, + "learning_rate": 1.851371942488196e-06, + "loss": 0.1627, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08949854969978333, + "step": 7995, + "valid_targets_mean": 7627.2, + "valid_targets_min": 5742 + }, + { + "epoch": 6.135788262370541, + "grad_norm": 0.42206685871605587, + "learning_rate": 1.8353357682672613e-06, + "loss": 0.1676, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09566031396389008, + "step": 8000, + "valid_targets_mean": 8725.6, + "valid_targets_min": 5141 + }, + { + "epoch": 6.1396240889911775, + "grad_norm": 0.43934848863177295, + "learning_rate": 1.8193660068570284e-06, + "loss": 0.1623, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09056299924850464, + "step": 8005, + "valid_targets_mean": 6967.8, + "valid_targets_min": 6088 + }, + { + "epoch": 6.143459915611814, + "grad_norm": 0.4370307663339287, + "learning_rate": 1.803462716645399e-06, + "loss": 0.1657, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07625304162502289, + "step": 8010, + "valid_targets_mean": 7140.1, + "valid_targets_min": 4834 + }, + { + "epoch": 6.147295742232451, + "grad_norm": 0.48282351408105234, + "learning_rate": 1.7876259557772547e-06, + "loss": 0.1659, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07682505249977112, + "step": 8015, + "valid_targets_mean": 6706.9, + "valid_targets_min": 5108 + }, + { + "epoch": 6.151131568853088, + "grad_norm": 0.47035929243167274, + "learning_rate": 1.7718557821542303e-06, + "loss": 0.167, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08013999462127686, + "step": 8020, + "valid_targets_mean": 5931.4, + "valid_targets_min": 4708 + }, + { + "epoch": 6.154967395473725, + "grad_norm": 0.43502550341372626, + "learning_rate": 1.7561522534345132e-06, + "loss": 0.1577, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07337029278278351, + "step": 8025, + "valid_targets_mean": 7145.6, + "valid_targets_min": 5280 + }, + { + "epoch": 6.158803222094361, + "grad_norm": 0.6881649119278562, + "learning_rate": 1.7405154270326208e-06, + "loss": 0.1493, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04542398452758789, + "step": 8030, + "valid_targets_mean": 1554.9, + "valid_targets_min": 592 + }, + { + "epoch": 6.162639048714998, + "grad_norm": 0.4587618011893617, + "learning_rate": 1.7249453601192013e-06, + "loss": 0.1455, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09382027387619019, + "step": 8035, + "valid_targets_mean": 7069.5, + "valid_targets_min": 4765 + }, + { + "epoch": 6.166474875335635, + "grad_norm": 0.48662183246538754, + "learning_rate": 1.7094421096208136e-06, + "loss": 0.1593, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08097653836011887, + "step": 8040, + "valid_targets_mean": 6193.4, + "valid_targets_min": 5587 + }, + { + "epoch": 6.1703107019562715, + "grad_norm": 0.5030705794360709, + "learning_rate": 1.6940057322197367e-06, + "loss": 0.1621, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08493475615978241, + "step": 8045, + "valid_targets_mean": 6612.2, + "valid_targets_min": 4894 + }, + { + "epoch": 6.174146528576908, + "grad_norm": 0.4257971174784796, + "learning_rate": 1.6786362843537386e-06, + "loss": 0.1581, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07594906538724899, + "step": 8050, + "valid_targets_mean": 6605.9, + "valid_targets_min": 3795 + }, + { + "epoch": 6.177982355197545, + "grad_norm": 0.44600249013483345, + "learning_rate": 1.6633338222158891e-06, + "loss": 0.1604, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08152023702859879, + "step": 8055, + "valid_targets_mean": 7186.8, + "valid_targets_min": 4890 + }, + { + "epoch": 6.181818181818182, + "grad_norm": 0.4453051144034571, + "learning_rate": 1.64809840175435e-06, + "loss": 0.1628, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08163885772228241, + "step": 8060, + "valid_targets_mean": 6544.4, + "valid_targets_min": 4615 + }, + { + "epoch": 6.185654008438819, + "grad_norm": 0.4594906746792594, + "learning_rate": 1.6329300786721635e-06, + "loss": 0.1659, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08696261048316956, + "step": 8065, + "valid_targets_mean": 6339.2, + "valid_targets_min": 4798 + }, + { + "epoch": 6.189489835059455, + "grad_norm": 0.5540814550988761, + "learning_rate": 1.6178289084270571e-06, + "loss": 0.1213, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08570878952741623, + "step": 8070, + "valid_targets_mean": 4558.8, + "valid_targets_min": 148 + }, + { + "epoch": 6.193325661680092, + "grad_norm": 0.4521007132786052, + "learning_rate": 1.602794946231232e-06, + "loss": 0.1606, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0881035253405571, + "step": 8075, + "valid_targets_mean": 6807.9, + "valid_targets_min": 5164 + }, + { + "epoch": 6.197161488300729, + "grad_norm": 0.44058789809314286, + "learning_rate": 1.587828247051173e-06, + "loss": 0.1554, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08169705420732498, + "step": 8080, + "valid_targets_mean": 7011.8, + "valid_targets_min": 4974 + }, + { + "epoch": 6.2009973149213655, + "grad_norm": 0.5285563220163505, + "learning_rate": 1.5729288656074393e-06, + "loss": 0.1628, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07309045642614365, + "step": 8085, + "valid_targets_mean": 5794.2, + "valid_targets_min": 4809 + }, + { + "epoch": 6.204833141542002, + "grad_norm": 0.42136960572004, + "learning_rate": 1.5580968563744603e-06, + "loss": 0.1511, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06946075707674026, + "step": 8090, + "valid_targets_mean": 6401.8, + "valid_targets_min": 4842 + }, + { + "epoch": 6.208668968162639, + "grad_norm": 0.46400689254901295, + "learning_rate": 1.543332273580349e-06, + "loss": 0.158, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08014999330043793, + "step": 8095, + "valid_targets_mean": 6182.1, + "valid_targets_min": 5116 + }, + { + "epoch": 6.212504794783276, + "grad_norm": 0.49548304020992884, + "learning_rate": 1.528635171206696e-06, + "loss": 0.1573, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07311199605464935, + "step": 8100, + "valid_targets_mean": 6165.1, + "valid_targets_min": 5764 + }, + { + "epoch": 6.216340621403913, + "grad_norm": 0.8622590303794951, + "learning_rate": 1.5140056029883642e-06, + "loss": 0.1553, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05816476047039032, + "step": 8105, + "valid_targets_mean": 1428.2, + "valid_targets_min": 1062 + }, + { + "epoch": 6.220176448024549, + "grad_norm": 0.9489302844116957, + "learning_rate": 1.4994436224133235e-06, + "loss": 0.1396, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06753484159708023, + "step": 8110, + "valid_targets_mean": 1663.0, + "valid_targets_min": 960 + }, + { + "epoch": 6.224012274645186, + "grad_norm": 0.8046219596332626, + "learning_rate": 1.4849492827224054e-06, + "loss": 0.1342, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06619445979595184, + "step": 8115, + "valid_targets_mean": 1476.5, + "valid_targets_min": 933 + }, + { + "epoch": 6.227848101265823, + "grad_norm": 0.7726254808807752, + "learning_rate": 1.4705226369091485e-06, + "loss": 0.1316, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0677412748336792, + "step": 8120, + "valid_targets_mean": 1478.1, + "valid_targets_min": 881 + }, + { + "epoch": 6.2316839278864595, + "grad_norm": 0.8390634961532917, + "learning_rate": 1.4561637377196047e-06, + "loss": 0.1361, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.078497976064682, + "step": 8125, + "valid_targets_mean": 1640.6, + "valid_targets_min": 869 + }, + { + "epoch": 6.235519754507096, + "grad_norm": 0.8303573117525184, + "learning_rate": 1.4418726376521087e-06, + "loss": 0.134, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07179920375347137, + "step": 8130, + "valid_targets_mean": 1709.0, + "valid_targets_min": 1090 + }, + { + "epoch": 6.239355581127733, + "grad_norm": 0.8717291922462884, + "learning_rate": 1.4276493889571285e-06, + "loss": 0.1242, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04865536093711853, + "step": 8135, + "valid_targets_mean": 1157.9, + "valid_targets_min": 761 + }, + { + "epoch": 6.24319140774837, + "grad_norm": 0.7710062757831959, + "learning_rate": 1.4134940436370514e-06, + "loss": 0.1375, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06333539634943008, + "step": 8140, + "valid_targets_mean": 1553.6, + "valid_targets_min": 1241 + }, + { + "epoch": 6.247027234369007, + "grad_norm": 0.8277656710851524, + "learning_rate": 1.3994066534460005e-06, + "loss": 0.125, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0616399422287941, + "step": 8145, + "valid_targets_mean": 1538.1, + "valid_targets_min": 733 + }, + { + "epoch": 6.250863060989643, + "grad_norm": 0.8058424565514163, + "learning_rate": 1.3853872698896486e-06, + "loss": 0.1325, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04081512987613678, + "step": 8150, + "valid_targets_mean": 1250.6, + "valid_targets_min": 835 + }, + { + "epoch": 6.25469888761028, + "grad_norm": 1.5449181681998265, + "learning_rate": 1.371435944225017e-06, + "loss": 0.1226, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06998598575592041, + "step": 8155, + "valid_targets_mean": 1596.2, + "valid_targets_min": 1046 + }, + { + "epoch": 6.258534714230917, + "grad_norm": 0.8796734582125978, + "learning_rate": 1.3575527274603006e-06, + "loss": 0.1282, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.059625446796417236, + "step": 8160, + "valid_targets_mean": 1646.5, + "valid_targets_min": 797 + }, + { + "epoch": 6.2623705408515535, + "grad_norm": 0.8302817379646728, + "learning_rate": 1.3437376703546855e-06, + "loss": 0.1197, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.055894337594509125, + "step": 8165, + "valid_targets_mean": 1448.9, + "valid_targets_min": 570 + }, + { + "epoch": 6.26620636747219, + "grad_norm": 0.8108181019584033, + "learning_rate": 1.3299908234181412e-06, + "loss": 0.118, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04677771031856537, + "step": 8170, + "valid_targets_mean": 1164.1, + "valid_targets_min": 548 + }, + { + "epoch": 6.270042194092827, + "grad_norm": 0.8646453164358331, + "learning_rate": 1.3163122369112591e-06, + "loss": 0.1136, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.055200688540935516, + "step": 8175, + "valid_targets_mean": 1338.0, + "valid_targets_min": 682 + }, + { + "epoch": 6.273878020713464, + "grad_norm": 0.9040386645430797, + "learning_rate": 1.3027019608450652e-06, + "loss": 0.1291, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06950479745864868, + "step": 8180, + "valid_targets_mean": 1632.8, + "valid_targets_min": 896 + }, + { + "epoch": 6.2777138473341, + "grad_norm": 0.8830792093022971, + "learning_rate": 1.289160044980815e-06, + "loss": 0.1252, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0719609409570694, + "step": 8185, + "valid_targets_mean": 1806.5, + "valid_targets_min": 597 + }, + { + "epoch": 6.281549673954737, + "grad_norm": 0.8423397976326829, + "learning_rate": 1.275686538829848e-06, + "loss": 0.1305, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.052942417562007904, + "step": 8190, + "valid_targets_mean": 1289.9, + "valid_targets_min": 706 + }, + { + "epoch": 6.285385500575374, + "grad_norm": 0.7973090640723524, + "learning_rate": 1.2622814916533765e-06, + "loss": 0.1227, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04128693789243698, + "step": 8195, + "valid_targets_mean": 1354.1, + "valid_targets_min": 754 + }, + { + "epoch": 6.289221327196011, + "grad_norm": 0.9033464134136978, + "learning_rate": 1.248944952462312e-06, + "loss": 0.1211, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05419332534074783, + "step": 8200, + "valid_targets_mean": 1461.9, + "valid_targets_min": 819 + }, + { + "epoch": 6.2930571538166475, + "grad_norm": 0.8399788236929602, + "learning_rate": 1.2356769700171035e-06, + "loss": 0.1247, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04704103618860245, + "step": 8205, + "valid_targets_mean": 1225.5, + "valid_targets_min": 684 + }, + { + "epoch": 6.296892980437284, + "grad_norm": 0.9906630363205486, + "learning_rate": 1.2224775928275378e-06, + "loss": 0.1318, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05968429893255234, + "step": 8210, + "valid_targets_mean": 1282.2, + "valid_targets_min": 711 + }, + { + "epoch": 6.300728807057921, + "grad_norm": 0.8518985581417532, + "learning_rate": 1.2093468691525634e-06, + "loss": 0.1188, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06457693874835968, + "step": 8215, + "valid_targets_mean": 1634.8, + "valid_targets_min": 1190 + }, + { + "epoch": 6.304564633678558, + "grad_norm": 0.8659584406648598, + "learning_rate": 1.196284847000142e-06, + "loss": 0.1153, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06325021386146545, + "step": 8220, + "valid_targets_mean": 1615.9, + "valid_targets_min": 774 + }, + { + "epoch": 6.308400460299194, + "grad_norm": 0.8476254371915735, + "learning_rate": 1.1832915741270235e-06, + "loss": 0.1205, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06617705523967743, + "step": 8225, + "valid_targets_mean": 1848.4, + "valid_targets_min": 734 + }, + { + "epoch": 6.312236286919831, + "grad_norm": 0.8961294035518611, + "learning_rate": 1.1703670980386272e-06, + "loss": 0.1202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06661862879991531, + "step": 8230, + "valid_targets_mean": 1570.6, + "valid_targets_min": 1011 + }, + { + "epoch": 6.316072113540468, + "grad_norm": 0.9807556233933888, + "learning_rate": 1.1575114659888298e-06, + "loss": 0.1271, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07181344926357269, + "step": 8235, + "valid_targets_mean": 1697.0, + "valid_targets_min": 803 + }, + { + "epoch": 6.319907940161105, + "grad_norm": 0.8394312129654994, + "learning_rate": 1.1447247249797888e-06, + "loss": 0.1236, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05208613723516464, + "step": 8240, + "valid_targets_mean": 1307.1, + "valid_targets_min": 782 + }, + { + "epoch": 6.3237437667817415, + "grad_norm": 0.8504319284005705, + "learning_rate": 1.1320069217618125e-06, + "loss": 0.1194, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06412334740161896, + "step": 8245, + "valid_targets_mean": 1596.6, + "valid_targets_min": 646 + }, + { + "epoch": 6.327579593402378, + "grad_norm": 0.8867734556035588, + "learning_rate": 1.1193581028331457e-06, + "loss": 0.1242, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07959242910146713, + "step": 8250, + "valid_targets_mean": 1817.5, + "valid_targets_min": 922 + }, + { + "epoch": 6.331415420023015, + "grad_norm": 1.0424931788963272, + "learning_rate": 1.1067783144398115e-06, + "loss": 0.1235, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07270241528749466, + "step": 8255, + "valid_targets_mean": 1822.2, + "valid_targets_min": 796 + }, + { + "epoch": 6.335251246643652, + "grad_norm": 1.0302371627340443, + "learning_rate": 1.0942676025754628e-06, + "loss": 0.113, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.058831486850976944, + "step": 8260, + "valid_targets_mean": 1542.4, + "valid_targets_min": 903 + }, + { + "epoch": 6.339087073264288, + "grad_norm": 1.1869752110805334, + "learning_rate": 1.0818260129811863e-06, + "loss": 0.1174, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06435605138540268, + "step": 8265, + "valid_targets_mean": 1459.5, + "valid_targets_min": 780 + }, + { + "epoch": 6.342922899884925, + "grad_norm": 0.8823143880995131, + "learning_rate": 1.0694535911453508e-06, + "loss": 0.1148, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05788193270564079, + "step": 8270, + "valid_targets_mean": 1581.1, + "valid_targets_min": 783 + }, + { + "epoch": 6.346758726505562, + "grad_norm": 0.8490504823289065, + "learning_rate": 1.0571503823034356e-06, + "loss": 0.1159, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.049834828823804855, + "step": 8275, + "valid_targets_mean": 1342.1, + "valid_targets_min": 662 + }, + { + "epoch": 6.350594553126198, + "grad_norm": 0.922312462716022, + "learning_rate": 1.0449164314378702e-06, + "loss": 0.1161, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0508771687746048, + "step": 8280, + "valid_targets_mean": 1269.1, + "valid_targets_min": 724 + }, + { + "epoch": 6.3544303797468356, + "grad_norm": 0.9268650969991855, + "learning_rate": 1.0327517832778588e-06, + "loss": 0.1172, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06325769424438477, + "step": 8285, + "valid_targets_mean": 1510.1, + "valid_targets_min": 966 + }, + { + "epoch": 6.358266206367472, + "grad_norm": 1.0322140514933342, + "learning_rate": 1.0206564822992315e-06, + "loss": 0.1216, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.050257109105587006, + "step": 8290, + "valid_targets_mean": 1327.8, + "valid_targets_min": 773 + }, + { + "epoch": 6.362102032988109, + "grad_norm": 0.8778742433535476, + "learning_rate": 1.008630572724274e-06, + "loss": 0.1158, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.050442829728126526, + "step": 8295, + "valid_targets_mean": 1353.2, + "valid_targets_min": 807 + }, + { + "epoch": 6.365937859608746, + "grad_norm": 0.8593690619991964, + "learning_rate": 9.966740985215618e-07, + "loss": 0.1238, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04606330767273903, + "step": 8300, + "valid_targets_mean": 1142.5, + "valid_targets_min": 782 + }, + { + "epoch": 6.369773686229382, + "grad_norm": 0.9052009472888289, + "learning_rate": 9.847871034058065e-07, + "loss": 0.1166, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07434223592281342, + "step": 8305, + "valid_targets_mean": 1849.9, + "valid_targets_min": 634 + }, + { + "epoch": 6.373609512850019, + "grad_norm": 0.944763475085017, + "learning_rate": 9.729696308376946e-07, + "loss": 0.1241, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06288284063339233, + "step": 8310, + "valid_targets_mean": 1599.6, + "valid_targets_min": 573 + }, + { + "epoch": 6.377445339470656, + "grad_norm": 0.844293069893908, + "learning_rate": 9.612217240237244e-07, + "loss": 0.1222, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06743821501731873, + "step": 8315, + "valid_targets_mean": 1994.9, + "valid_targets_min": 1102 + }, + { + "epoch": 6.381281166091293, + "grad_norm": 0.7756146235571519, + "learning_rate": 9.495434259160552e-07, + "loss": 0.1221, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04445914924144745, + "step": 8320, + "valid_targets_mean": 1239.2, + "valid_targets_min": 841 + }, + { + "epoch": 6.38511699271193, + "grad_norm": 0.8508418820856422, + "learning_rate": 9.379347792123439e-07, + "loss": 0.1129, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.054076045751571655, + "step": 8325, + "valid_targets_mean": 1467.5, + "valid_targets_min": 956 + }, + { + "epoch": 6.388952819332566, + "grad_norm": 1.026889788553785, + "learning_rate": 9.263958263555884e-07, + "loss": 0.1244, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06989319622516632, + "step": 8330, + "valid_targets_mean": 1300.9, + "valid_targets_min": 857 + }, + { + "epoch": 6.392788645953203, + "grad_norm": 0.8302153085348426, + "learning_rate": 9.149266095339815e-07, + "loss": 0.1171, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0588388592004776, + "step": 8335, + "valid_targets_mean": 1389.8, + "valid_targets_min": 808 + }, + { + "epoch": 6.39662447257384, + "grad_norm": 0.8434884404583354, + "learning_rate": 9.035271706807447e-07, + "loss": 0.1105, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0520755872130394, + "step": 8340, + "valid_targets_mean": 1303.8, + "valid_targets_min": 677 + }, + { + "epoch": 6.400460299194476, + "grad_norm": 0.8412024640832646, + "learning_rate": 8.921975514739834e-07, + "loss": 0.1117, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04705450311303139, + "step": 8345, + "valid_targets_mean": 1220.1, + "valid_targets_min": 727 + }, + { + "epoch": 6.404296125815113, + "grad_norm": 0.822345337836382, + "learning_rate": 8.809377933365292e-07, + "loss": 0.117, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04495104029774666, + "step": 8350, + "valid_targets_mean": 1262.0, + "valid_targets_min": 726 + }, + { + "epoch": 6.40813195243575, + "grad_norm": 0.9097658389762161, + "learning_rate": 8.69747937435792e-07, + "loss": 0.1222, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06914593279361725, + "step": 8355, + "valid_targets_mean": 1615.0, + "valid_targets_min": 669 + }, + { + "epoch": 6.411967779056386, + "grad_norm": 0.7802580719569506, + "learning_rate": 8.586280246836098e-07, + "loss": 0.1163, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06151066720485687, + "step": 8360, + "valid_targets_mean": 1806.5, + "valid_targets_min": 792 + }, + { + "epoch": 6.415803605677024, + "grad_norm": 0.890544589865958, + "learning_rate": 8.475780957361013e-07, + "loss": 0.1155, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.058973848819732666, + "step": 8365, + "valid_targets_mean": 1627.6, + "valid_targets_min": 736 + }, + { + "epoch": 6.41963943229766, + "grad_norm": 1.0705663043784048, + "learning_rate": 8.365981909935028e-07, + "loss": 0.1217, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07973496615886688, + "step": 8370, + "valid_targets_mean": 1822.9, + "valid_targets_min": 694 + }, + { + "epoch": 6.423475258918297, + "grad_norm": 0.8715437240872397, + "learning_rate": 8.256883506000357e-07, + "loss": 0.113, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07627920806407928, + "step": 8375, + "valid_targets_mean": 1848.9, + "valid_targets_min": 1437 + }, + { + "epoch": 6.427311085538934, + "grad_norm": 1.1815333751742108, + "learning_rate": 8.148486144437661e-07, + "loss": 0.1159, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07057377696037292, + "step": 8380, + "valid_targets_mean": 1852.4, + "valid_targets_min": 809 + }, + { + "epoch": 6.43114691215957, + "grad_norm": 0.8912072872608118, + "learning_rate": 8.04079022156432e-07, + "loss": 0.1159, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05849696695804596, + "step": 8385, + "valid_targets_mean": 1375.5, + "valid_targets_min": 790 + }, + { + "epoch": 6.434982738780207, + "grad_norm": 0.9779761908763904, + "learning_rate": 7.933796131133253e-07, + "loss": 0.1086, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.052983663976192474, + "step": 8390, + "valid_targets_mean": 1338.8, + "valid_targets_min": 750 + }, + { + "epoch": 6.438818565400844, + "grad_norm": 1.0367302495322794, + "learning_rate": 7.827504264331298e-07, + "loss": 0.1195, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07091420888900757, + "step": 8395, + "valid_targets_mean": 1778.8, + "valid_targets_min": 826 + }, + { + "epoch": 6.44265439202148, + "grad_norm": 0.9272969274651961, + "learning_rate": 7.721915009777902e-07, + "loss": 0.1209, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06042829900979996, + "step": 8400, + "valid_targets_mean": 1600.8, + "valid_targets_min": 907 + }, + { + "epoch": 6.446490218642118, + "grad_norm": 0.8852151756918829, + "learning_rate": 7.617028753523725e-07, + "loss": 0.1102, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0627460926771164, + "step": 8405, + "valid_targets_mean": 1474.5, + "valid_targets_min": 1093 + }, + { + "epoch": 6.450326045262754, + "grad_norm": 0.8617021185650273, + "learning_rate": 7.512845879048968e-07, + "loss": 0.1194, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06679156422615051, + "step": 8410, + "valid_targets_mean": 1577.2, + "valid_targets_min": 1119 + }, + { + "epoch": 6.454161871883391, + "grad_norm": 0.8789316437972291, + "learning_rate": 7.409366767262316e-07, + "loss": 0.1193, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06459164619445801, + "step": 8415, + "valid_targets_mean": 1512.1, + "valid_targets_min": 700 + }, + { + "epoch": 6.457997698504028, + "grad_norm": 0.8327652916165331, + "learning_rate": 7.306591796499374e-07, + "loss": 0.1155, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04843812435865402, + "step": 8420, + "valid_targets_mean": 1288.2, + "valid_targets_min": 810 + }, + { + "epoch": 6.461833525124664, + "grad_norm": 0.9515098663217382, + "learning_rate": 7.204521342521209e-07, + "loss": 0.1188, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05283963680267334, + "step": 8425, + "valid_targets_mean": 1328.9, + "valid_targets_min": 635 + }, + { + "epoch": 6.465669351745301, + "grad_norm": 1.0677268018507977, + "learning_rate": 7.103155778513149e-07, + "loss": 0.1111, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06438931822776794, + "step": 8430, + "valid_targets_mean": 1542.5, + "valid_targets_min": 673 + }, + { + "epoch": 6.469505178365938, + "grad_norm": 0.9333652265042814, + "learning_rate": 7.002495475083337e-07, + "loss": 0.1201, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06262427568435669, + "step": 8435, + "valid_targets_mean": 1215.6, + "valid_targets_min": 641 + }, + { + "epoch": 6.473341004986574, + "grad_norm": 0.8434484082576346, + "learning_rate": 6.902540800261292e-07, + "loss": 0.1112, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05036976560950279, + "step": 8440, + "valid_targets_mean": 1222.6, + "valid_targets_min": 701 + }, + { + "epoch": 6.477176831607212, + "grad_norm": 0.861695945878246, + "learning_rate": 6.803292119496774e-07, + "loss": 0.1288, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06303419172763824, + "step": 8445, + "valid_targets_mean": 1850.6, + "valid_targets_min": 1202 + }, + { + "epoch": 6.481012658227848, + "grad_norm": 0.8085292380501945, + "learning_rate": 6.704749795658227e-07, + "loss": 0.1182, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04747471958398819, + "step": 8450, + "valid_targets_mean": 1265.5, + "valid_targets_min": 657 + }, + { + "epoch": 6.484848484848484, + "grad_norm": 0.89867395433849, + "learning_rate": 6.606914189031499e-07, + "loss": 0.1196, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05455479398369789, + "step": 8455, + "valid_targets_mean": 1388.0, + "valid_targets_min": 672 + }, + { + "epoch": 6.488684311469122, + "grad_norm": 0.8806597486228541, + "learning_rate": 6.509785657318723e-07, + "loss": 0.1155, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07261599600315094, + "step": 8460, + "valid_targets_mean": 1813.6, + "valid_targets_min": 1019 + }, + { + "epoch": 6.492520138089758, + "grad_norm": 0.9255145459363365, + "learning_rate": 6.413364555636769e-07, + "loss": 0.1104, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.046454936265945435, + "step": 8465, + "valid_targets_mean": 1114.6, + "valid_targets_min": 732 + }, + { + "epoch": 6.4963559647103954, + "grad_norm": 0.9199685918260776, + "learning_rate": 6.317651236515954e-07, + "loss": 0.117, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03977984935045242, + "step": 8470, + "valid_targets_mean": 1194.5, + "valid_targets_min": 634 + }, + { + "epoch": 6.500191791331032, + "grad_norm": 0.799828145505751, + "learning_rate": 6.222646049899039e-07, + "loss": 0.1099, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.056514985859394073, + "step": 8475, + "valid_targets_mean": 1756.5, + "valid_targets_min": 905 + }, + { + "epoch": 6.504027617951668, + "grad_norm": 1.0409877018761717, + "learning_rate": 6.128349343139506e-07, + "loss": 0.1111, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0693710595369339, + "step": 8480, + "valid_targets_mean": 1554.9, + "valid_targets_min": 596 + }, + { + "epoch": 6.507863444572306, + "grad_norm": 1.0331804490054297, + "learning_rate": 6.034761461000727e-07, + "loss": 0.1105, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0470346063375473, + "step": 8485, + "valid_targets_mean": 1214.6, + "valid_targets_min": 687 + }, + { + "epoch": 6.511699271192942, + "grad_norm": 0.8402331506271794, + "learning_rate": 5.941882745654393e-07, + "loss": 0.1217, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06711738556623459, + "step": 8490, + "valid_targets_mean": 1955.1, + "valid_targets_min": 1199 + }, + { + "epoch": 6.515535097813579, + "grad_norm": 0.8267376380872375, + "learning_rate": 5.849713536679357e-07, + "loss": 0.1117, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05870674178004265, + "step": 8495, + "valid_targets_mean": 1758.6, + "valid_targets_min": 1214 + }, + { + "epoch": 6.519370924434216, + "grad_norm": 0.8615426638914652, + "learning_rate": 5.758254171060507e-07, + "loss": 0.1103, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.061762236058712006, + "step": 8500, + "valid_targets_mean": 1711.5, + "valid_targets_min": 1122 + }, + { + "epoch": 6.523206751054852, + "grad_norm": 0.92916368801175, + "learning_rate": 5.667504983187311e-07, + "loss": 0.11, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05335118994116783, + "step": 8505, + "valid_targets_mean": 1489.6, + "valid_targets_min": 850 + }, + { + "epoch": 6.5270425776754895, + "grad_norm": 0.8766261644877806, + "learning_rate": 5.57746630485283e-07, + "loss": 0.1024, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05151817202568054, + "step": 8510, + "valid_targets_mean": 1572.4, + "valid_targets_min": 1318 + }, + { + "epoch": 6.530878404296126, + "grad_norm": 0.8926561482725245, + "learning_rate": 5.488138465252335e-07, + "loss": 0.1144, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06340505182743073, + "step": 8515, + "valid_targets_mean": 1500.0, + "valid_targets_min": 1043 + }, + { + "epoch": 6.534714230916762, + "grad_norm": 0.908360578034097, + "learning_rate": 5.399521790982131e-07, + "loss": 0.114, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07024265825748444, + "step": 8520, + "valid_targets_mean": 1774.9, + "valid_targets_min": 920 + }, + { + "epoch": 6.5385500575374, + "grad_norm": 0.8956572581823119, + "learning_rate": 5.311616606038472e-07, + "loss": 0.1147, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.058703839778900146, + "step": 8525, + "valid_targets_mean": 1547.1, + "valid_targets_min": 847 + }, + { + "epoch": 6.542385884158036, + "grad_norm": 0.8031001171206305, + "learning_rate": 5.224423231816223e-07, + "loss": 0.1162, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05934291332960129, + "step": 8530, + "valid_targets_mean": 1769.4, + "valid_targets_min": 738 + }, + { + "epoch": 6.546221710778672, + "grad_norm": 0.9141892618844573, + "learning_rate": 5.13794198710782e-07, + "loss": 0.1119, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06029961630702019, + "step": 8535, + "valid_targets_mean": 1540.1, + "valid_targets_min": 594 + }, + { + "epoch": 6.55005753739931, + "grad_norm": 0.9307676608336025, + "learning_rate": 5.052173188102005e-07, + "loss": 0.1122, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06259243190288544, + "step": 8540, + "valid_targets_mean": 2228.6, + "valid_targets_min": 1393 + }, + { + "epoch": 6.553893364019946, + "grad_norm": 0.95999825289373, + "learning_rate": 4.967117148382716e-07, + "loss": 0.2227, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11638356745243073, + "step": 8545, + "valid_targets_mean": 3692.9, + "valid_targets_min": 770 + }, + { + "epoch": 6.557729190640583, + "grad_norm": 0.9418208547949569, + "learning_rate": 4.882774178927951e-07, + "loss": 0.2159, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1179671511054039, + "step": 8550, + "valid_targets_mean": 5237.4, + "valid_targets_min": 1799 + }, + { + "epoch": 6.56156501726122, + "grad_norm": 0.8754119205852636, + "learning_rate": 4.799144588108595e-07, + "loss": 0.2124, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.115792416036129, + "step": 8555, + "valid_targets_mean": 5888.0, + "valid_targets_min": 3015 + }, + { + "epoch": 6.565400843881856, + "grad_norm": 0.7419835595989265, + "learning_rate": 4.716228681687285e-07, + "loss": 0.2041, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10393969714641571, + "step": 8560, + "valid_targets_mean": 6029.1, + "valid_targets_min": 1681 + }, + { + "epoch": 6.569236670502494, + "grad_norm": 0.8525823473107436, + "learning_rate": 4.63402676281739e-07, + "loss": 0.2134, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09036293625831604, + "step": 8565, + "valid_targets_mean": 3299.6, + "valid_targets_min": 1374 + }, + { + "epoch": 6.57307249712313, + "grad_norm": 0.8368956611636551, + "learning_rate": 4.5525391320417667e-07, + "loss": 0.2118, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10214373469352722, + "step": 8570, + "valid_targets_mean": 3858.8, + "valid_targets_min": 1951 + }, + { + "epoch": 6.576908323743766, + "grad_norm": 0.7460558648726231, + "learning_rate": 4.4717660872917624e-07, + "loss": 0.2048, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10001727938652039, + "step": 8575, + "valid_targets_mean": 5250.8, + "valid_targets_min": 2042 + }, + { + "epoch": 6.580744150364404, + "grad_norm": 0.7197207543781292, + "learning_rate": 4.391707923886035e-07, + "loss": 0.2052, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08998521417379379, + "step": 8580, + "valid_targets_mean": 4937.4, + "valid_targets_min": 1831 + }, + { + "epoch": 6.58457997698504, + "grad_norm": 0.7410766045164557, + "learning_rate": 4.312364934529623e-07, + "loss": 0.2136, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10909228771924973, + "step": 8585, + "valid_targets_mean": 4049.4, + "valid_targets_min": 2077 + }, + { + "epoch": 6.5884158036056775, + "grad_norm": 0.730096495748466, + "learning_rate": 4.233737409312655e-07, + "loss": 0.2121, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09737345576286316, + "step": 8590, + "valid_targets_mean": 4086.2, + "valid_targets_min": 2022 + }, + { + "epoch": 6.592251630226314, + "grad_norm": 0.7323691797308391, + "learning_rate": 4.155825635709509e-07, + "loss": 0.2109, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11733715236186981, + "step": 8595, + "valid_targets_mean": 4897.6, + "valid_targets_min": 1499 + }, + { + "epoch": 6.59608745684695, + "grad_norm": 0.6634710340739296, + "learning_rate": 4.078629898577635e-07, + "loss": 0.2018, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10772855579853058, + "step": 8600, + "valid_targets_mean": 4821.6, + "valid_targets_min": 1013 + }, + { + "epoch": 6.599923283467588, + "grad_norm": 0.7010660425868566, + "learning_rate": 4.00215048015653e-07, + "loss": 0.2056, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10270945727825165, + "step": 8605, + "valid_targets_mean": 4676.2, + "valid_targets_min": 1939 + }, + { + "epoch": 6.603759110088224, + "grad_norm": 0.6887580792440051, + "learning_rate": 3.9263876600667216e-07, + "loss": 0.1988, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11501763761043549, + "step": 8610, + "valid_targets_mean": 4010.2, + "valid_targets_min": 1820 + }, + { + "epoch": 6.6075949367088604, + "grad_norm": 0.7371373346254915, + "learning_rate": 3.851341715308765e-07, + "loss": 0.1999, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10866899788379669, + "step": 8615, + "valid_targets_mean": 3824.9, + "valid_targets_min": 1589 + }, + { + "epoch": 6.611430763329498, + "grad_norm": 0.6855704986699592, + "learning_rate": 3.777012920262202e-07, + "loss": 0.1977, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0911431536078453, + "step": 8620, + "valid_targets_mean": 4046.5, + "valid_targets_min": 2668 + }, + { + "epoch": 6.615266589950134, + "grad_norm": 0.6863599352967574, + "learning_rate": 3.7034015466845373e-07, + "loss": 0.197, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09315817058086395, + "step": 8625, + "valid_targets_mean": 3592.6, + "valid_targets_min": 1269 + }, + { + "epoch": 6.619102416570771, + "grad_norm": 0.7027634198203273, + "learning_rate": 3.630507863710309e-07, + "loss": 0.2065, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09664779901504517, + "step": 8630, + "valid_targets_mean": 3423.4, + "valid_targets_min": 1695 + }, + { + "epoch": 6.622938243191408, + "grad_norm": 0.6916242731541822, + "learning_rate": 3.55833213785004e-07, + "loss": 0.2003, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09179112315177917, + "step": 8635, + "valid_targets_mean": 3522.2, + "valid_targets_min": 1128 + }, + { + "epoch": 6.626774069812044, + "grad_norm": 0.7196329481381968, + "learning_rate": 3.486874632989312e-07, + "loss": 0.1971, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09786905348300934, + "step": 8640, + "valid_targets_mean": 3533.5, + "valid_targets_min": 1438 + }, + { + "epoch": 6.630609896432682, + "grad_norm": 0.6974513757839929, + "learning_rate": 3.4161356103877606e-07, + "loss": 0.2011, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09359021484851837, + "step": 8645, + "valid_targets_mean": 3842.9, + "valid_targets_min": 1267 + }, + { + "epoch": 6.634445723053318, + "grad_norm": 0.6966300456824807, + "learning_rate": 3.346115328678146e-07, + "loss": 0.19, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08716927468776703, + "step": 8650, + "valid_targets_mean": 3671.9, + "valid_targets_min": 1026 + }, + { + "epoch": 6.6382815496739545, + "grad_norm": 0.730818988189699, + "learning_rate": 3.2768140438654173e-07, + "loss": 0.1819, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08672162890434265, + "step": 8655, + "valid_targets_mean": 2866.2, + "valid_targets_min": 1403 + }, + { + "epoch": 6.642117376294592, + "grad_norm": 0.6904975439139568, + "learning_rate": 3.2082320093257623e-07, + "loss": 0.1916, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10881981253623962, + "step": 8660, + "valid_targets_mean": 3882.1, + "valid_targets_min": 1588 + }, + { + "epoch": 6.645953202915228, + "grad_norm": 0.7212447422891808, + "learning_rate": 3.140369475805627e-07, + "loss": 0.1977, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0845642238855362, + "step": 8665, + "valid_targets_mean": 2827.6, + "valid_targets_min": 1276 + }, + { + "epoch": 6.649789029535865, + "grad_norm": 0.5976228635177685, + "learning_rate": 3.0732266914209165e-07, + "loss": 0.1939, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08349532634019852, + "step": 8670, + "valid_targets_mean": 4232.4, + "valid_targets_min": 1631 + }, + { + "epoch": 6.653624856156502, + "grad_norm": 0.6534276488876793, + "learning_rate": 3.006803901655997e-07, + "loss": 0.1917, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11982393264770508, + "step": 8675, + "valid_targets_mean": 5069.9, + "valid_targets_min": 3090 + }, + { + "epoch": 6.657460682777138, + "grad_norm": 0.7037206267965201, + "learning_rate": 2.941101349362785e-07, + "loss": 0.1868, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11615725606679916, + "step": 8680, + "valid_targets_mean": 3732.4, + "valid_targets_min": 1802 + }, + { + "epoch": 6.661296509397776, + "grad_norm": 0.7554764993438118, + "learning_rate": 2.8761192747599474e-07, + "loss": 0.199, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1152307465672493, + "step": 8685, + "valid_targets_mean": 3490.8, + "valid_targets_min": 1685 + }, + { + "epoch": 6.665132336018412, + "grad_norm": 0.6678891552474201, + "learning_rate": 2.8118579154319923e-07, + "loss": 0.1842, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0914856418967247, + "step": 8690, + "valid_targets_mean": 3660.1, + "valid_targets_min": 1596 + }, + { + "epoch": 6.6689681626390485, + "grad_norm": 0.7398999412995361, + "learning_rate": 2.748317506328313e-07, + "loss": 0.1898, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11274268478155136, + "step": 8695, + "valid_targets_mean": 4306.4, + "valid_targets_min": 2251 + }, + { + "epoch": 6.672803989259686, + "grad_norm": 0.7168233740233785, + "learning_rate": 2.685498279762433e-07, + "loss": 0.185, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09546729177236557, + "step": 8700, + "valid_targets_mean": 3107.1, + "valid_targets_min": 1458 + }, + { + "epoch": 6.676639815880322, + "grad_norm": 0.6735080503962053, + "learning_rate": 2.6234004654111854e-07, + "loss": 0.1915, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09050801396369934, + "step": 8705, + "valid_targets_mean": 3664.6, + "valid_targets_min": 1162 + }, + { + "epoch": 6.680475642500959, + "grad_norm": 0.9414608126142952, + "learning_rate": 2.562024290313625e-07, + "loss": 0.1866, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08630982786417007, + "step": 8710, + "valid_targets_mean": 2834.2, + "valid_targets_min": 1480 + }, + { + "epoch": 6.684311469121596, + "grad_norm": 0.7704709016203939, + "learning_rate": 2.501369978870627e-07, + "loss": 0.1936, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10016943514347076, + "step": 8715, + "valid_targets_mean": 3795.1, + "valid_targets_min": 1229 + }, + { + "epoch": 6.688147295742232, + "grad_norm": 0.6699331368788622, + "learning_rate": 2.441437752843623e-07, + "loss": 0.1911, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08827116340398788, + "step": 8720, + "valid_targets_mean": 3559.6, + "valid_targets_min": 1276 + }, + { + "epoch": 6.691983122362869, + "grad_norm": 0.6435538136071092, + "learning_rate": 2.382227831354067e-07, + "loss": 0.1896, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08564750850200653, + "step": 8725, + "valid_targets_mean": 3479.6, + "valid_targets_min": 1357 + }, + { + "epoch": 6.695818948983506, + "grad_norm": 0.73419532510076, + "learning_rate": 2.323740430882615e-07, + "loss": 0.1876, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09410861134529114, + "step": 8730, + "valid_targets_mean": 2925.9, + "valid_targets_min": 924 + }, + { + "epoch": 6.6996547756041425, + "grad_norm": 0.6863108642128618, + "learning_rate": 2.2659757652681246e-07, + "loss": 0.1961, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1032608151435852, + "step": 8735, + "valid_targets_mean": 4030.5, + "valid_targets_min": 1674 + }, + { + "epoch": 6.70349060222478, + "grad_norm": 0.7112999612580819, + "learning_rate": 2.2089340457071672e-07, + "loss": 0.1888, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09143315255641937, + "step": 8740, + "valid_targets_mean": 3173.9, + "valid_targets_min": 1643 + }, + { + "epoch": 6.707326428845416, + "grad_norm": 0.6102602407967938, + "learning_rate": 2.152615480753073e-07, + "loss": 0.1813, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10092855244874954, + "step": 8745, + "valid_targets_mean": 4259.1, + "valid_targets_min": 1244 + }, + { + "epoch": 6.711162255466053, + "grad_norm": 0.6722544562247412, + "learning_rate": 2.0970202763151092e-07, + "loss": 0.1959, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0902620255947113, + "step": 8750, + "valid_targets_mean": 4065.8, + "valid_targets_min": 730 + }, + { + "epoch": 6.71499808208669, + "grad_norm": 0.719613075322354, + "learning_rate": 2.0421486356579923e-07, + "loss": 0.1894, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08508659899234772, + "step": 8755, + "valid_targets_mean": 2945.9, + "valid_targets_min": 977 + }, + { + "epoch": 6.718833908707326, + "grad_norm": 0.674058405957139, + "learning_rate": 1.9880007594008655e-07, + "loss": 0.1853, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10428628325462341, + "step": 8760, + "valid_targets_mean": 5116.1, + "valid_targets_min": 2091 + }, + { + "epoch": 6.722669735327964, + "grad_norm": 0.7050041497938554, + "learning_rate": 1.9345768455167224e-07, + "loss": 0.1783, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08329372107982635, + "step": 8765, + "valid_targets_mean": 3125.4, + "valid_targets_min": 1654 + }, + { + "epoch": 6.7265055619486, + "grad_norm": 0.7420572205650171, + "learning_rate": 1.881877089331563e-07, + "loss": 0.1946, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09684513509273529, + "step": 8770, + "valid_targets_mean": 4081.6, + "valid_targets_min": 1432 + }, + { + "epoch": 6.7303413885692365, + "grad_norm": 0.6678840202696662, + "learning_rate": 1.8299016835238826e-07, + "loss": 0.1822, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08364065736532211, + "step": 8775, + "valid_targets_mean": 3868.6, + "valid_targets_min": 1261 + }, + { + "epoch": 6.734177215189874, + "grad_norm": 0.7262375879434214, + "learning_rate": 1.7786508181237393e-07, + "loss": 0.1951, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11349351704120636, + "step": 8780, + "valid_targets_mean": 4079.6, + "valid_targets_min": 2737 + }, + { + "epoch": 6.73801304181051, + "grad_norm": 1.4235693523209338, + "learning_rate": 1.7281246805121998e-07, + "loss": 0.1857, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07879241555929184, + "step": 8785, + "valid_targets_mean": 3471.4, + "valid_targets_min": 1219 + }, + { + "epoch": 6.741848868431147, + "grad_norm": 0.7038851099723767, + "learning_rate": 1.678323455420605e-07, + "loss": 0.2035, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1055973470211029, + "step": 8790, + "valid_targets_mean": 4087.4, + "valid_targets_min": 2058 + }, + { + "epoch": 6.745684695051784, + "grad_norm": 0.7154115805877195, + "learning_rate": 1.6292473249298833e-07, + "loss": 0.1825, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08858203142881393, + "step": 8795, + "valid_targets_mean": 2736.6, + "valid_targets_min": 916 + }, + { + "epoch": 6.74952052167242, + "grad_norm": 0.6651230550381625, + "learning_rate": 1.5808964684699723e-07, + "loss": 0.1974, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08005672693252563, + "step": 8800, + "valid_targets_mean": 3695.1, + "valid_targets_min": 1302 + }, + { + "epoch": 6.753356348293057, + "grad_norm": 0.6835197335184454, + "learning_rate": 1.5332710628190196e-07, + "loss": 0.1829, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11640842258930206, + "step": 8805, + "valid_targets_mean": 4246.1, + "valid_targets_min": 1500 + }, + { + "epoch": 6.757192174913694, + "grad_norm": 0.8112111868362218, + "learning_rate": 1.4863712821028498e-07, + "loss": 0.1827, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10816378146409988, + "step": 8810, + "valid_targets_mean": 4837.6, + "valid_targets_min": 2803 + }, + { + "epoch": 6.7610280015343305, + "grad_norm": 0.6246020453137344, + "learning_rate": 1.440197297794299e-07, + "loss": 0.1746, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08940006792545319, + "step": 8815, + "valid_targets_mean": 3349.0, + "valid_targets_min": 878 + }, + { + "epoch": 6.764863828154967, + "grad_norm": 0.6319136547658165, + "learning_rate": 1.3947492787125483e-07, + "loss": 0.1985, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08778447657823563, + "step": 8820, + "valid_targets_mean": 3753.9, + "valid_targets_min": 2662 + }, + { + "epoch": 6.768699654775604, + "grad_norm": 0.6363893012673046, + "learning_rate": 1.350027391022568e-07, + "loss": 0.1884, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08916360884904861, + "step": 8825, + "valid_targets_mean": 3965.1, + "valid_targets_min": 1086 + }, + { + "epoch": 6.772535481396241, + "grad_norm": 0.6574430093934882, + "learning_rate": 1.306031798234475e-07, + "loss": 0.1982, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10965670645236969, + "step": 8830, + "valid_targets_mean": 5292.5, + "valid_targets_min": 1566 + }, + { + "epoch": 6.776371308016878, + "grad_norm": 0.6804884901588436, + "learning_rate": 1.262762661202932e-07, + "loss": 0.189, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11538819968700409, + "step": 8835, + "valid_targets_mean": 4169.2, + "valid_targets_min": 2504 + }, + { + "epoch": 6.780207134637514, + "grad_norm": 0.6926726566836009, + "learning_rate": 1.2202201381265487e-07, + "loss": 0.1963, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11999423801898956, + "step": 8840, + "valid_targets_mean": 4600.9, + "valid_targets_min": 1548 + }, + { + "epoch": 6.784042961258151, + "grad_norm": 0.7199954182057174, + "learning_rate": 1.1784043845473714e-07, + "loss": 0.1925, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09204265475273132, + "step": 8845, + "valid_targets_mean": 3289.5, + "valid_targets_min": 1352 + }, + { + "epoch": 6.787878787878788, + "grad_norm": 0.5937423118161487, + "learning_rate": 1.1373155533502156e-07, + "loss": 0.1876, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10010066628456116, + "step": 8850, + "valid_targets_mean": 5279.6, + "valid_targets_min": 2339 + }, + { + "epoch": 6.7917146144994245, + "grad_norm": 0.6732426906934017, + "learning_rate": 1.0969537947621789e-07, + "loss": 0.185, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09225790202617645, + "step": 8855, + "valid_targets_mean": 4053.0, + "valid_targets_min": 2015 + }, + { + "epoch": 6.795550441120062, + "grad_norm": 0.6705453168920024, + "learning_rate": 1.0573192563520851e-07, + "loss": 0.1809, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10675591230392456, + "step": 8860, + "valid_targets_mean": 4836.5, + "valid_targets_min": 1731 + }, + { + "epoch": 6.799386267740698, + "grad_norm": 0.7720182241715025, + "learning_rate": 1.0184120830299072e-07, + "loss": 0.18, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08718439936637878, + "step": 8865, + "valid_targets_mean": 3052.5, + "valid_targets_min": 1460 + }, + { + "epoch": 6.803222094361335, + "grad_norm": 0.6720418990024768, + "learning_rate": 9.80232417046234e-08, + "loss": 0.1787, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09809219837188721, + "step": 8870, + "valid_targets_mean": 4621.9, + "valid_targets_min": 1282 + }, + { + "epoch": 6.807057920981972, + "grad_norm": 0.718018832896278, + "learning_rate": 9.427803979918715e-08, + "loss": 0.1705, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07664323598146439, + "step": 8875, + "valid_targets_mean": 3373.4, + "valid_targets_min": 1421 + }, + { + "epoch": 6.810893747602608, + "grad_norm": 0.6677757278944716, + "learning_rate": 9.060561627971087e-08, + "loss": 0.1779, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11129924654960632, + "step": 8880, + "valid_targets_mean": 4483.5, + "valid_targets_min": 1764 + }, + { + "epoch": 6.814729574223245, + "grad_norm": 0.8008166778900557, + "learning_rate": 8.700598457314302e-08, + "loss": 0.198, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11546166241168976, + "step": 8885, + "valid_targets_mean": 3288.8, + "valid_targets_min": 1175 + }, + { + "epoch": 6.818565400843882, + "grad_norm": 0.6130766665506642, + "learning_rate": 8.347915784029159e-08, + "loss": 0.1899, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08555158972740173, + "step": 8890, + "valid_targets_mean": 4656.6, + "valid_targets_min": 2546 + }, + { + "epoch": 6.8224012274645185, + "grad_norm": 0.7965558937891689, + "learning_rate": 8.002514897577973e-08, + "loss": 0.1863, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11666454374790192, + "step": 8895, + "valid_targets_mean": 3833.2, + "valid_targets_min": 1244 + }, + { + "epoch": 6.826237054085155, + "grad_norm": 0.6546102531050428, + "learning_rate": 7.664397060799467e-08, + "loss": 0.1895, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09570662677288055, + "step": 8900, + "valid_targets_mean": 4403.5, + "valid_targets_min": 1856 + }, + { + "epoch": 6.830072880705792, + "grad_norm": 0.702900012318378, + "learning_rate": 7.333563509904329e-08, + "loss": 0.1881, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10926999151706696, + "step": 8905, + "valid_targets_mean": 3836.6, + "valid_targets_min": 1948 + }, + { + "epoch": 6.833908707326429, + "grad_norm": 0.62758040481807, + "learning_rate": 7.01001545447122e-08, + "loss": 0.1815, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07936488837003708, + "step": 8910, + "valid_targets_mean": 3336.2, + "valid_targets_min": 1248 + }, + { + "epoch": 6.837744533947066, + "grad_norm": 0.6015193035771662, + "learning_rate": 6.693754077441882e-08, + "loss": 0.1852, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08881563693284988, + "step": 8915, + "valid_targets_mean": 4304.4, + "valid_targets_min": 1145 + }, + { + "epoch": 6.841580360567702, + "grad_norm": 0.7290732098525768, + "learning_rate": 6.384780535116264e-08, + "loss": 0.1854, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08850069344043732, + "step": 8920, + "valid_targets_mean": 3231.1, + "valid_targets_min": 1428 + }, + { + "epoch": 6.845416187188339, + "grad_norm": 0.6776285660850613, + "learning_rate": 6.083095957149176e-08, + "loss": 0.184, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10522789508104324, + "step": 8925, + "valid_targets_mean": 4426.2, + "valid_targets_min": 1708 + }, + { + "epoch": 6.849252013808976, + "grad_norm": 0.7045055730322493, + "learning_rate": 5.788701446546308e-08, + "loss": 0.1777, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06047890707850456, + "step": 8930, + "valid_targets_mean": 2934.0, + "valid_targets_min": 1301 + }, + { + "epoch": 6.8530878404296125, + "grad_norm": 0.679164969577275, + "learning_rate": 5.50159807965911e-08, + "loss": 0.1998, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08315706253051758, + "step": 8935, + "valid_targets_mean": 3015.5, + "valid_targets_min": 1971 + }, + { + "epoch": 6.85692366705025, + "grad_norm": 0.6963288882766603, + "learning_rate": 5.2217869061816914e-08, + "loss": 0.1959, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11766372621059418, + "step": 8940, + "valid_targets_mean": 4887.2, + "valid_targets_min": 1054 + }, + { + "epoch": 6.860759493670886, + "grad_norm": 0.6567019943674285, + "learning_rate": 4.949268949146824e-08, + "loss": 0.1927, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10739628970623016, + "step": 8945, + "valid_targets_mean": 3527.2, + "valid_targets_min": 1612 + }, + { + "epoch": 6.864595320291523, + "grad_norm": 0.6831554082687543, + "learning_rate": 4.684045204922161e-08, + "loss": 0.1998, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07828736305236816, + "step": 8950, + "valid_targets_mean": 3531.1, + "valid_targets_min": 721 + }, + { + "epoch": 6.86843114691216, + "grad_norm": 0.6750894802129682, + "learning_rate": 4.42611664320669e-08, + "loss": 0.1923, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09905225038528442, + "step": 8955, + "valid_targets_mean": 3899.6, + "valid_targets_min": 1625 + }, + { + "epoch": 6.872266973532796, + "grad_norm": 0.6758818572972229, + "learning_rate": 4.1754842070269584e-08, + "loss": 0.2047, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09560565650463104, + "step": 8960, + "valid_targets_mean": 4175.0, + "valid_targets_min": 1381 + }, + { + "epoch": 6.876102800153433, + "grad_norm": 0.7189392892172067, + "learning_rate": 3.932148812733516e-08, + "loss": 0.1998, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09056779742240906, + "step": 8965, + "valid_targets_mean": 3235.0, + "valid_targets_min": 998 + }, + { + "epoch": 6.87993862677407, + "grad_norm": 0.800095077151153, + "learning_rate": 3.696111349998255e-08, + "loss": 0.1976, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08479700982570648, + "step": 8970, + "valid_targets_mean": 2957.1, + "valid_targets_min": 1186 + }, + { + "epoch": 6.8837744533947065, + "grad_norm": 0.7005009253178834, + "learning_rate": 3.4673726818101884e-08, + "loss": 0.1912, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10994219779968262, + "step": 8975, + "valid_targets_mean": 3583.0, + "valid_targets_min": 1195 + }, + { + "epoch": 6.887610280015343, + "grad_norm": 0.8211119280704406, + "learning_rate": 3.2459336444736754e-08, + "loss": 0.1933, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09459938108921051, + "step": 8980, + "valid_targets_mean": 2696.1, + "valid_targets_min": 885 + }, + { + "epoch": 6.89144610663598, + "grad_norm": 0.7024616340586688, + "learning_rate": 3.031795047603536e-08, + "loss": 0.1982, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09310254454612732, + "step": 8985, + "valid_targets_mean": 3362.2, + "valid_targets_min": 814 + }, + { + "epoch": 6.895281933256617, + "grad_norm": 0.6696826607364263, + "learning_rate": 2.8249576741234964e-08, + "loss": 0.1961, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08477000892162323, + "step": 8990, + "valid_targets_mean": 3612.5, + "valid_targets_min": 2174 + }, + { + "epoch": 6.899117759877253, + "grad_norm": 0.6046061617197763, + "learning_rate": 2.625422280263079e-08, + "loss": 0.1779, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10522358119487762, + "step": 8995, + "valid_targets_mean": 5472.4, + "valid_targets_min": 3025 + }, + { + "epoch": 6.90295358649789, + "grad_norm": 0.7194518390237327, + "learning_rate": 2.433189595554275e-08, + "loss": 0.1936, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09157722443342209, + "step": 9000, + "valid_targets_mean": 3297.9, + "valid_targets_min": 1434 + }, + { + "epoch": 6.906789413118527, + "grad_norm": 0.7466225165049727, + "learning_rate": 2.248260322829543e-08, + "loss": 0.1847, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.092948779463768, + "step": 9005, + "valid_targets_mean": 3262.1, + "valid_targets_min": 1074 + }, + { + "epoch": 6.910625239739164, + "grad_norm": 0.6702489737625144, + "learning_rate": 2.0706351382184796e-08, + "loss": 0.1961, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0748385637998581, + "step": 9010, + "valid_targets_mean": 3145.2, + "valid_targets_min": 980 + }, + { + "epoch": 6.9144610663598005, + "grad_norm": 0.6458174343121412, + "learning_rate": 1.9003146911462656e-08, + "loss": 0.1939, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10791381448507309, + "step": 9015, + "valid_targets_mean": 4575.0, + "valid_targets_min": 2366 + }, + { + "epoch": 6.918296892980437, + "grad_norm": 0.6553472006796922, + "learning_rate": 1.7372996043307777e-08, + "loss": 0.1962, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10535834729671478, + "step": 9020, + "valid_targets_mean": 4999.9, + "valid_targets_min": 3139 + }, + { + "epoch": 6.922132719601074, + "grad_norm": 0.7234520978729007, + "learning_rate": 1.581590473780148e-08, + "loss": 0.1924, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11775435507297516, + "step": 9025, + "valid_targets_mean": 4034.4, + "valid_targets_min": 2138 + }, + { + "epoch": 6.925968546221711, + "grad_norm": 0.7621965740679113, + "learning_rate": 1.4331878687912082e-08, + "loss": 0.1897, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09851715713739395, + "step": 9030, + "valid_targets_mean": 3610.9, + "valid_targets_min": 2014 + }, + { + "epoch": 6.929804372842348, + "grad_norm": 0.7007693268614058, + "learning_rate": 1.2920923319463819e-08, + "loss": 0.1957, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10604077577590942, + "step": 9035, + "valid_targets_mean": 3795.6, + "valid_targets_min": 1084 + }, + { + "epoch": 6.933640199462984, + "grad_norm": 0.6899997476695795, + "learning_rate": 1.1583043791130178e-08, + "loss": 0.1875, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09581349790096283, + "step": 9040, + "valid_targets_mean": 3684.4, + "valid_targets_min": 1618 + }, + { + "epoch": 6.937476026083621, + "grad_norm": 0.6839404418635214, + "learning_rate": 1.0318244994402815e-08, + "loss": 0.182, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08924812078475952, + "step": 9045, + "valid_targets_mean": 3378.8, + "valid_targets_min": 1193 + }, + { + "epoch": 6.941311852704258, + "grad_norm": 0.6180011286247561, + "learning_rate": 9.126531553582673e-09, + "loss": 0.1927, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10162334144115448, + "step": 9050, + "valid_targets_mean": 4732.5, + "valid_targets_min": 2092 + }, + { + "epoch": 6.9451476793248945, + "grad_norm": 0.6715414062119801, + "learning_rate": 8.007907825755557e-09, + "loss": 0.1849, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11447025835514069, + "step": 9055, + "valid_targets_mean": 4235.9, + "valid_targets_min": 2786 + }, + { + "epoch": 6.948983505945531, + "grad_norm": 0.6235143387235511, + "learning_rate": 6.962377900781026e-09, + "loss": 0.1764, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07447785139083862, + "step": 9060, + "valid_targets_mean": 4036.9, + "valid_targets_min": 1246 + }, + { + "epoch": 6.952819332566168, + "grad_norm": 0.7204514661723178, + "learning_rate": 5.98994560128352e-09, + "loss": 0.1778, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07502232491970062, + "step": 9065, + "valid_targets_mean": 3434.8, + "valid_targets_min": 1484 + }, + { + "epoch": 6.956655159186805, + "grad_norm": 0.6292539282012609, + "learning_rate": 5.09061448261905e-09, + "loss": 0.1869, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09701728820800781, + "step": 9070, + "valid_targets_mean": 4156.9, + "valid_targets_min": 1502 + }, + { + "epoch": 6.960490985807441, + "grad_norm": 0.6028978146109196, + "learning_rate": 4.264387832884076e-09, + "loss": 0.1825, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07229148596525192, + "step": 9075, + "valid_targets_mean": 3785.8, + "valid_targets_min": 1530 + }, + { + "epoch": 6.964326812428078, + "grad_norm": 0.731772516630842, + "learning_rate": 3.511268672888868e-09, + "loss": 0.207, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10390883684158325, + "step": 9080, + "valid_targets_mean": 3245.6, + "valid_targets_min": 1312 + }, + { + "epoch": 6.968162639048715, + "grad_norm": 0.609369938491558, + "learning_rate": 2.83125975615306e-09, + "loss": 0.1804, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10198619961738586, + "step": 9085, + "valid_targets_mean": 5026.8, + "valid_targets_min": 2738 + }, + { + "epoch": 6.971998465669352, + "grad_norm": 0.6682346965706782, + "learning_rate": 2.2243635688945496e-09, + "loss": 0.1788, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0923006609082222, + "step": 9090, + "valid_targets_mean": 4030.1, + "valid_targets_min": 1759 + }, + { + "epoch": 6.9758342922899885, + "grad_norm": 0.6698585031506065, + "learning_rate": 1.6905823300206181e-09, + "loss": 0.1818, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08009584248065948, + "step": 9095, + "valid_targets_mean": 3422.8, + "valid_targets_min": 2155 + }, + { + "epoch": 6.979670118910625, + "grad_norm": 0.7861237008405464, + "learning_rate": 1.2299179911168247e-09, + "loss": 0.1897, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09182140231132507, + "step": 9100, + "valid_targets_mean": 3223.8, + "valid_targets_min": 2048 + }, + { + "epoch": 6.983505945531262, + "grad_norm": 0.7140366103328384, + "learning_rate": 8.423722364425679e-10, + "loss": 0.1894, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07877764850854874, + "step": 9105, + "valid_targets_mean": 2658.6, + "valid_targets_min": 1233 + }, + { + "epoch": 6.987341772151899, + "grad_norm": 0.6440613670941303, + "learning_rate": 5.279464829244241e-10, + "loss": 0.1927, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0844639241695404, + "step": 9110, + "valid_targets_mean": 3698.2, + "valid_targets_min": 1383 + }, + { + "epoch": 6.991177598772536, + "grad_norm": 0.6468226856752733, + "learning_rate": 2.866418801517057e-10, + "loss": 0.1897, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08857706189155579, + "step": 9115, + "valid_targets_mean": 3766.5, + "valid_targets_min": 1800 + }, + { + "epoch": 6.995013425393172, + "grad_norm": 0.6967472864562262, + "learning_rate": 1.1845931037202107e-10, + "loss": 0.1898, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0863482803106308, + "step": 9120, + "valid_targets_mean": 3114.6, + "valid_targets_min": 2327 + }, + { + "epoch": 6.998849252013809, + "grad_norm": 0.6947034687914561, + "learning_rate": 2.3399388486833318e-11, + "loss": 0.1923, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09771628677845001, + "step": 9125, + "valid_targets_mean": 3558.5, + "valid_targets_min": 1214 + }, + { + "epoch": 7.0, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11453244835138321, + "step": 9127, + "total_flos": 2.7891230302346936e+18, + "train_loss": 0.027739192068844568, + "train_runtime": 33702.2835, + "train_samples_per_second": 4.331, + "train_steps_per_second": 0.271, + "valid_targets_mean": 7108.5, + "valid_targets_min": 2092 + } + ], + "logging_steps": 5, + "max_steps": 9128, + "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": 2.7891230302346936e+18, + "train_batch_size": 1, + "trial_name": null, + "trial_params": null +}